Political Ground Truth: How Personal Issue

Experience Counters Partisan Biases
Aaron B. Strauss
A Dissertation
Presented to the Faculty
of Princeton University
in Candidacy for the Degree
of Doctor of Philosophy
Recommended for Acceptance
by the Department of Politics
Adviser: Martin Gilens
September 2009
c ( Copyright by Aaron B. Strauss, 2009.
All Rights Reserved
Abstract
Partisan attachments create pervasive bias in the way citizens process information. Political
scientists, psychologists, and recently neuroscientists find that people will believe nearly
anything if a favored politician espouses the view. Yet, even though partisan affiliation is
one of the most, if not the most, stable political attitudes, large segments of the public
switch their vote choice from one party to the other between elections or split their tickets
within a single election. This dissertation examines one explanation for shifting political
views: personal experience with specific issues.
Campaigns work to attract more support, but partisan biases hinder their efforts when
predispositions lead voters to doubt statements made by disfavored politicians. This dis-
sertation explores the theory that campaigns can successfully target voters who have ex-
perience on a particular political issue. The voter will use her independent knowledge on
the topic to judge, or “ground truth,” the politician’s views; if the voter and the politician
agree, the voter will hold the candidate in higher esteem. With the advent of massive
campaign databases of information on voters, campaigns are now able to identify these
critical voter-issue linkages.
The Personal Experience Model explores why personal experience plays such a crucial
role in political judgments. This formal model is an extension of Zaller’s Receive-Accept-
Sample model. The theory behind the Personal Experience Model is presented, related to
existing theories, and supported by empirical evidence. Observational data from the 2000
presidential campaign, two survey experiments, and two field experiments all support the
model’s hypotheses. Finally, the strategic implications for campaigns, and the normative
implications for democracy, are considered.
iii
Acknowledgements
It is difficult to adequately express my appreciation for the many people who have supported
and guided me during my time at Princeton. It would feel appropriate to add about a dozen
names next to mine on the title page, but I will defer to rules and customs and put their
names in this section instead.
My graduate school colleagues made Princeton a wonderful place to study political
science. From “West Wing” viewings to games of Settlers of Catan, from softball defeat
to ultimate triumph, my friends demonstrated that the phrase “graduate school life” isn’t
always an oxymoron. Thank you Dustin, Dan, Ben, Glick, Tom D., Keya, Richard, Mi-
randa, Kevin, Andi, Will, Lauren, Tom C., Melody, Shana, Andrew, Nick, Grace, Alicia,
Gayle, Lisa, and Alistair who have all been awesome friends.
The graduate work of Gabriel Lenz and Amy Gershkoff provided excellent models for
my research. Their dissertations were the first ones I turned to when I experienced writer’s
block. Amy’s ability to turn esoteric political topics into accessible prose and Gabe’s
straightforward mathematical proofs were the trailmarkers that kept me on the right path.
I have attempted to live up to the high standard they set.
In addition, fellow graduate student Ben Lauderdale was an excellent sounding board.
Several ideas in this dissertation are the result of our long conversations on the third floor of
Robertson Hall. His amazing ability to generate a new idea every week was both humbling
and motivating. I could only take solace in my ability to run down fly balls faster than he
could on the softball field.
This dissertation is, in large part, a reflection of the professors who have taught me
these past four years. Thank you to Chris Achen for early inspiration on this project.
Larry Bartels provided incisive comments that provided context for my work. Professors
iv
at other institutions, including Steven Ansolabehere (my undergraduate advisor at MIT)
and Andrew Gelman of Columbia also generously gave their time to comment on chapters.
I am deeply indebted to my dissertation committee. Markus Prior’s class on campaigns
provided the first forum for the hypotheses that became the centerpiece of the dissertation.
Despite his own deadlines occasionally competing with mine, Markus always found time
to share his insights.
I thank Tali Mendelberg for her positive comments and attention to detail. I could
always count on a lunch with Tali to bring clarity to this project and to lift my spirits.
She made this dissertation better by challenging me to dig into the psyches of voters.
When the methodological portion of this dissertation hit a rocky patch of shore, Kosuke
Imai was there to right the ship. Thanks to Kosuke’s teaching on causal inference, I realize
that without a randomized experiment (no thanks!) I can’t be sure that this dissertation
wouldn’t have been completed without his help, but I have a strong prior opinion. I’m
pleased that we were able to co-author multiple papers, but even happier that I can call
him my friend.
Completing a dissertation under my adviser, Marty Gilens, was a pleasure. His door
was always open, his smile contagious, and his methods of motivation creative—such as
the time he invoked a promise from me to eat Turkey on Thanksgiving (and end 12 years of
vegetarianism) if I failed to complete a chapter draft. I have sung the praises of Marty to
many prospective graduate students, undergrads, and members of other universities that I
may have forgotten to express my feelings to him directly. So, Marty, here it is on acid-free
paper: thank you for being the best adviser a graduate student could ask for.
I must give special thanks to Chu Hwang and Hillary Hampton for supporting me
throughout this process. I’m glad they asked the question “How’s the dissertation going?”
v
as often as possible. That they both traveled hours to be at my defense speaks volumes
about what wonderful friends they are.
And finally, my family deserves thanks for being supportive and loving. My father
introduced my brother and me to educational games at an early age. Though we live on
opposite coasts, my brother Marc and I continue this family tradition by facing off in online
Scrabble matches. My mother, who fortunately is a professional editor, was a tremendous
help on this project. The grade I received on my first high school literature essay—an
F—suggests that my ability to write a dissertation is not innate. Thank you, Mom, for
patiently teaching me that the world is not just a set of equations.
vi
To my family—Mom, Dad, and Marc—who are responsible for my love of learning.
vii
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
1 Introduction 1
2 The Personal Experience Model 8
2.1 Theory and Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.1 Psychological Model of Personal Experience and Issue Stability . . . 11
2.1.2 Extending Zaller by Introducing Personal Experience . . . . . . . . . 15
2.1.3 Learning Model: Issue Positions . . . . . . . . . . . . . . . . . . . . 19
2.1.4 Voters’ Beliefs About Candidate Positions . . . . . . . . . . . . . . . 24
2.1.5 Voters’ Candidate Evaluations . . . . . . . . . . . . . . . . . . . . . 27
2.1.6 Issue Experience as Signals . . . . . . . . . . . . . . . . . . . . . . . 33
2.1.7 Model Extensions and Details . . . . . . . . . . . . . . . . . . . . . . 35
2.2 Concurrence with Other Theories in the Literature . . . . . . . . . . . . . . 37
2.2.1 Definitions of Concepts in the Literature . . . . . . . . . . . . . . . . 37
2.2.2 Personal Experience Yields Nuanced and Stable Opinions . . . . . . 39
2.2.3 Evidence for Cue-Taking and the Role of Experience . . . . . . . . . 41
2.2.4 Easy Issues and Retrospection Affect Political Evaluations . . . . . . 45
viii
2.3 Alternative Voter-Issue Linkages in the Literature . . . . . . . . . . . . . . . 47
3 Experienced Evaluations and Self-Interested Opinions 51
3.1 The Personal Experience Model and the 2000 Presidential Campaign . . . . 51
3.1.1 The Campaign about Nothing . . . . . . . . . . . . . . . . . . . . . 51
3.1.2 Issue-Driven Vote Choice . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1.3 Data from the 2000 Annenberg Survey . . . . . . . . . . . . . . . . . 58
3.2 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2.1 Learning Over the Course of the Campaign . . . . . . . . . . . . . . 59
3.2.2 Cue-Taking and Self-Interest . . . . . . . . . . . . . . . . . . . . . . 62
3.2.3 Experienced Voters Judge Politicians on Issues . . . . . . . . . . . . 67
3.2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4 Evidence from Survey Experiments 74
4.1 Theory and Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2 Survey Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.2.1 Overall Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.2.2 Specifics of Princeton Survey Design . . . . . . . . . . . . . . . . . . 80
4.2.3 Specifics of Nationwide Survey . . . . . . . . . . . . . . . . . . . . . 82
4.3 Design Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.3.1 Characteristics of the Survey Respondent Population . . . . . . . . . 85
4.3.2 Features of the Survey . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3.3 Definitions of Issue Experience . . . . . . . . . . . . . . . . . . . . . 91
4.4 Support for Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.4.1 Candidate Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.4.2 Issue Opinions and Experience . . . . . . . . . . . . . . . . . . . . . 98
ix
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5 Field Experiments: New Methodology and Evidence 104
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.1.1 Background of the Methodological Problem . . . . . . . . . . . . . . 107
5.2 The Formal Framework of GOTV Campaign Planning . . . . . . . . . . . . 108
5.2.1 The Planner’s Decision Problem . . . . . . . . . . . . . . . . . . . . 109
5.2.2 Data from a Randomized Field Experiment . . . . . . . . . . . . . . 113
5.2.3 The Bayesian Planner . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.2.4 Bayesian Optimal Campaign Planning at A Glance . . . . . . . . . . 116
5.3 The Optimal Nonpartisan Campaign Strategy . . . . . . . . . . . . . . . . . 117
5.3.1 The Optimization Method . . . . . . . . . . . . . . . . . . . . . . . . 118
5.3.2 The Statistical Method . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.4 The Optimal Partisan Campaign Strategy . . . . . . . . . . . . . . . . . . . 124
5.4.1 The Decision Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.4.2 Data Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.4.3 Derivation of the Optimal Strategy . . . . . . . . . . . . . . . . . . . 126
5.5 Empirical Evaluation of the Proposed Method . . . . . . . . . . . . . . . . . 127
5.5.1 Evaluation Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.5.2 A Nonpartisan GOTV Campaign with a Single Mobilization Method 130
5.5.3 Partisan Example: Parents and Education Spending . . . . . . . . . 135
5.5.4 Partisan Example: Disadvantaged Voters and the Economy . . . . . 139
5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6 Campaign and Normative Implications of Microtargeting 145
6.1 Campaign Microtargeting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
x
6.2 Implications for Campaign Strategy . . . . . . . . . . . . . . . . . . . . . . 148
6.2.1 Microtargeting and Message Control . . . . . . . . . . . . . . . . . . 148
6.2.2 Translating the Personal Experience Model into a Formal Game . . 150
6.2.3 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
6.2.4 Best Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.2.5 Edge Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.2.6 No Internal Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6.2.7 Model Extensions: Multiple Issues and Platform Decisions . . . . . . 164
6.2.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
6.3 Normative Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
6.3.1 Heuristics and Judging Democracy . . . . . . . . . . . . . . . . . . . 165
6.3.2 Positives for Democracy . . . . . . . . . . . . . . . . . . . . . . . . . 167
6.3.3 Negatives for Democracy . . . . . . . . . . . . . . . . . . . . . . . . 169
6.3.4 Further Discussion: Party Structure and Alignment . . . . . . . . . 170
6.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
7 Conclusion 173
A Appendix for Chapter 2 179
A.1 Details for Candidate Evaluation Simulation . . . . . . . . . . . . . . . . . . 179
B Appendix for Chapter 3 182
B.1 Regressions for the Cue-Taking Hypothesis . . . . . . . . . . . . . . . . . . 183
B.2 Regressions for the Candidate Evaluation Hypothesis . . . . . . . . . . . . . 184
C Appendix for Chapter 4 186
C.1 Question Wording in Nationwide Survey . . . . . . . . . . . . . . . . . . . . 186
xi
C.1.1 Candidate Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . 186
C.1.2 Candidates’ Issue Signals . . . . . . . . . . . . . . . . . . . . . . . . 187
C.1.3 Experience Questions . . . . . . . . . . . . . . . . . . . . . . . . . . 188
C.2 Auxiliary Regressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
C.3 Issue Experience Measures and the Hypotheses . . . . . . . . . . . . . . . . 190
D Appendix for Chapter 5 203
D.1 Computational Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
D.1.1 Nonpartisan Case: the Knapsack Problem . . . . . . . . . . . . . . . 203
D.1.2 Partisan Case: the Stochastic Knapsack Problem . . . . . . . . . . . 204
E Appendix for Chapter 6 206
E.1 Analytical Solution for Variance of Vote Share . . . . . . . . . . . . . . . . 206
xii
Chapter 1
Introduction
Prominent research in political science (Campbell et al., 1960; Bartels, 2002) emphasizes the
long-lasting attachments voters have to political parties. Yet large portions of the electorate
switch parties from election to election (Key, 1968) or split their ticket within one election
(Fiorina, 1996). These defections are often attributed to the preferences of voters relative
to the candidates (Downs, 1957; Enelow and Hinich, 1984; Carmines and Stimson, 1990);
the debate over the relative effects of partisan identification versus issue opinion continues
in recent research (Ansolabehere et al., 2008). This dissertation synthesizes the arguments
from both schools of thought, demonstrating how party identification and issue opinions
interact to form individuals’ political judgments. To provide observable predictions of the
theory, the model and the evidence presented highlight the role of personal experience in
political evaluations.
Two recent anecdotes illustrate this connection. From the 2004 to the 2006 elections,
the percentage of Hispanics identifying as Democrats jumped by 15 percentage points (from
43% to 58%), while non-Hispanics showed no increase in their inclination to be identified
1
as Democrats.
1
Hispanics’ responses to another question in the exit poll shed light on the
underlying reason for the massive movement: 37% of Latinos indicated that the issue of
“illegal immigration” was “extremely important” in their voting decision (compared to only
29% of the non-Latino population). While immigration policies might not personally affect
these voters (as they were already citizens), their involvement in the Hispanic community
most likely gave them firsthand experience with the immigration issue. And despite illegal
immigration falling off as a key concern in 2008, this trend of Hispanics favoring the
Democrats continued in the 2008 election. As measured by the National Exit Poll, Obama
garnered 67% of the Hispanic vote, compared to 53% for Kerry.
As Hispanics were flocking to the Democratic party, active duty military personnel
were leaving the Republican party in droves. A 2004 mail survey of subscribers to Military
Times found that 60% of active duty personnel identified as Republicans. The same survey
in 2006 found that this percentage had dropped to 46%. Republican affiliation among the
electorate at large (as measured by exit polls) dropped only 1.5 percentage points over
these two years. Almost assuredly, the wars in Iraq and Afghanistan had a profound
effect on the political views of the military. In 2008, Obama outperformed Kerry among
the active military and veterans despite the facts that both Kerry and McCain—Obama’s
opponent—were veterans.
This dissertation argues that an individual’s political allegiance is likely to shift when
the political parties take identifiable positions on issues with which the individual has signif-
icant personal experience. When a large group of people have experience with a prominent
issue (e.g., Hispanics and immigration, the military and the Iraq War), substantial changes
can occur in the political landscape. Campaigns can use recent technological advances to
1
Source: National Exit Polls
2
identify (i.e., microtarget) the segments of the population that have experience with certain
issues and that thus may be ripe for a political change of heart.
Generally, an individual’s partisan bias weighs heavily on her political decision-making
process as she adopts the positions of her party’s elites and does not adjust her political
worldview.
2
A telling example of this lack of critical thinking is provided by an experiment
performed by neuroscientists on partisans during the 2004 election (Westen et al., 2006).
Subjects were presented with contradictory statements by George W. Bush and John Kerry.
For example, the subject saw Kerry making two statements: one in 1996 that Congress had
to consider raising the retirement age to save Social Security, and a second in 2004 that
he would never consider raising the retirement age. When asked whether the candidates’
statements were internally inconsistent, partisans were quick to pass judgment on the
other party’s nominee but not their own. In fact, when evaluating their own party’s
nominee, a sector of partisans’ brains that dealt with reward (ventral striatum) showed
significant activity. Not only did partisans seek to diminish displeasure by explaining away
uncomfortable facts, but their brains actively sought positive reinforcement.
The brain’s attempt to diminish displeasure and seek positive reinforcement is the neu-
rological basis for humans’ preference for obtaining conclusions that are consistent with
their prior beliefs—a process psychologists label “motivated reasoning” by psychologists
(Lord et al., 1979). When citizens fall prey to this psychological process in the political
arena (Lodge and Taber, 2000), large partisan biases are seen. For instance, among Re-
publicans in 1996, twice as many thought the deficit had increased during Clinton’s first
term as thought it had decreased (Achen and Bartels, 2006). In fact, the deficit had fallen
dramatically, by over 90%.
2
Throughout this dissertation the female pronoun is used to identify an individual voter. The male
pronoun is used to identify an individual candidate.
3
These partisan biases help keep party affiliation more stable over time than political is-
sues (Ansolabehere et al., 2008). The public takes cues from like-minded elites and adopts
those positions on issues (Gilens and Murakawa, 2002). For instance, during the 2000
campaign, Democrats adopted the anti-privatization position espoused by Gore while Re-
publicans adopted the privatization position espoused by Bush (Richard Johnston, 2004).
The exception, as detailed in Chapter 3, was older Americans, who were more likely to
stay consistent in their view (which was, more often than not, anti-privatization). This
dissertation hypothesizes, and demonstrates with data from observational data, survey ex-
periments, and field experiments, that seniors’ personal experience with the Social Security
system moderated their motivated reasoning and cue-taking. Further, personal experience
with an issue enables voters to judge politicians on this issue and can counter partisan
biases that would usually hold.
Personal experience is defined as interaction with a political issue. The key feature
of personal experiences that they enable voters to have stable opinions on political issues
independent of their political predispositions. Without these experiences, voters are reliant
on political elites to point them in the right direction. This distinction is crucial: if
politicians find it more expedient to switch sides on an issue (e.g., the legitimacy of judicial
filibuster) and the voters follow, then the voters’ opinions can hardly be considered “stable.”
The theory presented here is premised on the idea that experience with an issue induces
stable opinions. Issue interactions often produce points of reference, or to use the political
science term, “considerations,” on which an individual can base her opinions. Because these
considerations are generated through an individual’s personal interactions, they are likely
to be trusted even if they go against the person’s preconceptions (i.e., “seeing is believing”).
Thus, personal experiences enable individuals to generate stable issue opinions independent
of their overall political views.
4
This dissertation elaborates this personal experience-issue opinion connection, derives
hypotheses, provides evidence in support of the hypotheses, and discusses the hypotheses’
implications. Chapter 2 presents the Personal Experience Model, a learning model based on
Zaller’s (1992) Receive-Accept-Sample (RAS) model that generates two main hypotheses.
First, if political elites start debating an issue, citizens with experience judge politicians on
the issue, comparing their experience to the politician’s argument. In effect, an individual
“ground truths” an elite’s views by evaluating how consistent the elite’s arguments are
with the voter’s first-hand experience.
The second main hypothesis is that voters with experience shift their issue opinion less
in the direction of a politician’s appeal than do individuals without experience. Simulations
illustrate the model at work in the campaign setting. Existing research is shown to be
consistent with the Personal Experience Model, noting how the foundations of the model
differ from research on issue publics and self-interest.
Chapter 3 highlights two observational cases from the 2000 presidential campaign that
illustrate the model in action: Social Security privatization and the Patient’s Bill of Rights.
Regarding the first issue, seniors, who have experience with Social Security, judged Bush
and Gore more on privatization than did younger voters. Younger voters were more apt
to adopt the position of their preferred candidate than were seniors. An analogous phe-
nomenon occurs among voters insured by HMOs and the Patients’ Bill of Rights.
Chapter 4 looks for the effects predicted by the Personal Experience Model in a survey
experiment setting. Two Internet panel surveys, one of Princeton students (n=273) and
one of the public nationwide (n=391), asked respondents about hypothetical candidates
and real-world issues. Respondents were also queried about their experiences relating to
these issues. The results are generally supportive of the Personal Experience Model’s two
main hypotheses but are noisy because of low n-sizes.
5
Chapter 5 develops a methodology for analyzing field experiments that identifies vot-
ers who are responsive to candidate appeals. Off-the-shelf non-parametric data-mining
techniques are applied to both the treatment group and the control group. For each seg-
ment of the population, the difference between the estimated effect for the two treatment
assignments is the treatment effect. Resource allocation algorithms are applied to these
estimates to determine the optimal campaign strategy. The method is general enough to
be applicable in both non-partisan (i.e., get-out-the-vote) and partisan settings. Examples
with experienced-voter segments demonstrate the power of this method and validate the
candidate evaluation hypothesis of the Personal Experience Model.
With the evidence for the Personal Experience Model laid out, the final chapter explores
the implications of the model for campaigns and democracy. In recent decades, campaigns
and election-oriented organizations have expanded their capacity to capture political data
to the point where both major parties have access to information about every voter na-
tionwide. Commercial and census data supplement these database and provide even more
knowledge about individuals. Rather than segment the population into distinct clusters
(as campaigns used to do), campaign practitioners now assign a probability score for a
certain trait (e.g., probability of being undecided) to each individual. This process, called
“microtargeting”, is related to the Personal Experience Model because the technology be-
hind microtargeting can be used to link voters to issues in a manner recommended by the
model.
A game theoretical model explicates the situations in which a campaign should spend
its money microtargeting voters rather than broadcasting its candidate’s position on an
issue to all voters. In general, broadcasting is a gamble that may have large returns for
a campaign, but more often may be useless or even backfire. Thus, campaigns that are
6
behind in the polls find this risky strategy useful, while advantaged campaigns often prefer
to microtarget.
The normative implications for democracy are mixed, depending on one’s perspective.
The Personal Experience Model, combined with modern campaign tools, increases the effi-
ciency of microtargeting. As with many tools, these advancements may be used for positive
or negative purposes. Information may be disseminated to the public in a more relevant
fashion as voters are matched with issues they care about. The information gain is viewed
as a positive for the democratic process. However, the increased party attachment that
microtargeting produces in some segments of the public may lead individuals to accept the
statements of political elites more easily even if those arguments are completely fabricated.
In sum, the Personal Experience Model explains a voter-issue linkage that limits the
partisan bias that pervades the political arena. The micro-level model provides deeper
understanding of the research on issue publics and self-interest. Campaigns can leverage
the implications of the model to increase the impact and efficiency of their appeals.
7
Chapter 2
The Personal Experience Model
2.1 Theory and Hypotheses
As demonstrated by Hispanics and the military after the 2004 election, large segments of
the population can be persuaded to change parties from one election to the next. However,
most research on political persuasion does not account for these changes, especially when
the changes are pronounced in specific groups. The Personal Experience Model seeks to
address this discrepancy, at least among a subset of voter-issue interactions.
The so-called “Michigan model” (Campbell et al., 1960) highlights the role of early-life
events and relationships in the formation of long-standing partisan identification. Interac-
tions with the political system at an early age form individuals’ partisan loyalties, which
are dubbed the “unmoved mover” by Campbell et al.: “only an event of extraordinary
intensity can arouse any significant part of the electorate to the point that its established
political loyalties are shaken” (Campbell et al., 1960, p. 151).
The mechanism for stable partisanship has been elucidated over the years and scholars
have reached a consensus. Consistent with the psychological concept of cognitive dissonance
8
(Festinger, 1957), individuals are more likely to accept and process political information
and assertions that are consistent with their current views. Prominent research (Zaller,
1992; Bartels, 2002; Taber and Lodge, 2006) applies this concept to politics, formalizing
the phenomenon and demonstrating how the predictions are consistent with observations of
public political opinions. Dozens of micro studies, including some from neuroscience (Knut-
son et al., 2006), have verified that partisan identification skews individuals’ perceptions
of the world in a self-confirming manner, leading to hardened political opinions.
One emblematic explanation for the micro-foundations of this “partisan bias” is offered
by Lupia and McCubbins’s (1998) model of persuasion. In this model, voters are persuaded
by elites with both “perceived common interest” to the voter and “perceived knowledge”
of the subject at hand. Listening to trusted politicians leads to further agreement between
the voter and the elite (i.e., cue-taking, Gilens and Murakawa, 2002), and thus more per-
ceived common interest. This positive feedback loop, or “cycle of partisan bias,” begs two
questions. First, what is the origin of “perceived common interest?” Second, if politicians
cannot persuade voters to cast a ballot for previously untrusted candidates or parties (i.e.,
those without perceived common interest), then why do some voters switch parties between
elections or split their votes within an election?
The Michigan model clearly answers the first question with the notion that voters
adopt the party identification of their parents and the social atmosphere in which they are
raised. But the answer to the second question, change in party identification, is less clear:
what is an “event of extraordinary intensity” (Campbell et al., 1960) that can break the
usually dominant role of party identification? The lack of theory on the determinants of
party-switching has left the door open to other approaches.
Two large areas of work that explain a shifting electorate are retrospective voting and
issue voting. Fiorina (1981) argues that voters are affected by how parties and candidates
9
performed their governing duties while in power. Retrospective voting, especially on the
economy, certainly plays a role in candidate and party evaluations (Mueller, 1970). How-
ever, these types of judgments should be relatively uniform across the populace, and do
not account for heterogeneous movements, such as with Hispanics and the military after
2004.
A second answer to the vote-switching question is presented by scholars who analyze
voters’ issue opinions. The Hotelling-Downs (Hotelling, 1929; Downs, 1957) model presents
voters as rational actors who vote for the party (or candidate) that most closely matches
their issue positions.
1
Downs’ uni-dimensional, one-valued utility is expanded into several
issues of varying importance (or “salience”) by Enelow and Hinich (1984). Certain segments
of the population altering their issue opinions or issue saliences could account for the
observed shifts in the electorate’s political views. However, if voters cue-take from preferred
politicians via the cycle of partisan bias, then changes in issue opinion will only polarize
the electorate. No shifts from one party to the other, or one candidate to the other, will
be observed.
Two other subjects discussed in the literature, self-interest (e.g., Sears and Funk, 1990)
and issue publics (e.g., Krosnick, 1990), would predict heterogeneous shifts in public opin-
ion. While both the model presented here and these theories emphasize voter-issue link,
this dissertation distinguishes itself from existing research in the following two ways, and
offers an alternative causal mechanism. First, in contrast to the research on self-interest,
this dissertation incorporates cue-taking directly into a formal model. Second, it details
the origins of voter-issue linkages, suggesting they are formed before an issue is politicized.
1
That voters are rational actors in formalized models such as Hotelling-Downs and Fiorina does not
rule out the possibility that voters are rational when succumbing to partisan bias. It may be rational to
minimize the displeasure of believing two seemingly contradictory ideas.
10
2.1.1 Psychological Model of Personal Experience and Issue Stability
A mechanism for breaking the cycle of partisan bias—which emphatically does not rule out
other causes—is personal experience. Individuals create and strengthen political opinions
through personal experience on a subset of the many issues that political elites must ad-
dress. This subset of issues forms a benchmark against which to judge parties, candidates,
and political elites, enabling an individual to develop a political worldview.
First, this chapter presents the psychological flows of political information, and then
I develop a learning model that formalizes these flows. The flows of information are a
superset of Zaller’s (1992) Receive-Accept-Sample (RAS) model, and I call my model the
“Personal Experience Model.” Figure 2.1 presents a box-and-arrow depiction of the Per-
sonal Experience Model.
A review of Zaller’s model is crucial for understanding the how the Personal Experi-
ence Model fits within the current political science literature. Under RAS, an individual
incorporates information into a political opinion by first receiving the information. As
noted by Downs (1957), individuals can either seek out this information or happen upon
it accidentally. More politically attentive individuals are more likely to receive political
information.
Next, individuals (either consciously or subconsciously) accept or reject a piece of po-
litical information based on their worldview. In Zaller’s words, “People tend to resist argu-
ments that are inconsistent with their political predispositions” (p. 44). This accept/reject
process, based on what I label an individual’s “worldview,” is the key for producing the
stable partisan affiliation and partisan biases found by Campbell et al. (1960), Bartels
(2002), and others. An individual’s worldview is an online aggregation of that person’s
evaluations of important political figures, candidates, and parties. Even Zaller acknowl-
edges that voters’ evaluations of other people (rather than issues) act more like an online
11
F
i
g
u
r
e
2
.
1
:
T
h
e

o
w
o
f
p
o
l
i
t
i
c
a
l
i
n
f
o
r
m
a
t
i
o
n
t
h
r
o
u
g
h
t
h
e
m
i
n
d
o
f
a
n
i
n
d
i
v
i
d
u
a
l
,
b
a
s
e
d
o
n
Z
a
l
l
e
r

s
(
1
9
9
2
)
R
e
c
e
i
v
e
-
A
c
c
e
p
t
-
S
a
m
p
l
e
(
R
A
S
)
m
o
d
e
l
.
R
e
c
t
a
n
g
l
e
s
r
e
p
r
e
s
e
n
t
e
p
h
e
m
e
r
a
l
p
r
o
c
e
s
s
e
s
t
h
a
t
i
n
f
o
r
m
a
t
i
o
n

o
w
s
t
h
r
o
u
g
h
;
o
v
a
l
s
r
e
p
r
e
-
s
e
n
t
s
t
a
b
l
e
s
t
o
r
e
s
o
f
i
n
f
o
r
m
a
t
i
o
n
(
o
r
s
o
u
r
c
e
s
o
f
i
n
f
o
r
m
a
t
i
o
n
)
.
S
o
l
i
d
a
r
r
o
w
s
i
n
d
i
c
a
t
e
i
n
f
o
r
m
a
t
i
o
n

o
w
;
t
h
e
o
n
e
d
a
s
h
e
d
a
r
r
o
w
i
n
d
i
c
a
t
e
s
a
c
a
u
s
e
a
n
d
e

e
c
t
(
i
.
e
.
,
p
a
r
t
i
s
a
n
i
d
e
n
t
i

c
a
t
i
o
n
a
n
d
w
o
r
l
d
v
i
e
w
a

e
c
t
w
h
i
c
h
p
i
e
c
e
s
o
f
i
n
f
o
r
m
a
t
i
o
n
a
r
e
a
c
c
e
p
t
e
d
o
r
r
e
j
e
c
t
e
d
)
.
T
h
e
F
I
F
O
s
y
m
b
o
l
u
n
d
e
r

S
e
t
o
f
C
o
n
s
i
d
e
r
a
t
i
o
n
s

r
e
p
r
e
s
e
n
t
s
t
h
e
c
o
m
p
u
t
e
r
s
c
i
e
n
c
e
t
e
r
m

F
i
r
s
t
I
n
-
F
i
r
s
t
O
u
t

q
u
e
u
e
a
n
d
i
s
a
s
i
m
p
l
i
s
t
i
c
r
e
p
r
e
s
e
n
t
a
t
i
o
n
o
f
t
h
e
h
u
m
a
n
b
r
a
i
n

s
m
e
m
o
r
y
c
a
p
a
b
i
l
i
t
y
.
12
model (p. 281).
2
If the argument (i.e., “consideration”) is accepted, then it is stored among
all other considerations. Recently accepted (or “thought about”) considerations are most
accessible to the individual.
Figure 2.1 depicts this relative availability of political arguments with a First-In-First-
Out (FIFO) queue, a computer science concept. The first political consideration heard (if
never thought of again) works its way down to the end of the queue (i.e., human memory)
until it is completely forgotten (or at least, inaccessible). Considerations already in memory
that are primed or re-accepted move to the front of the queue and become more accessible.
While some political psychologists (Weston, 2007) would take issue with this simplistic
view of the brain, it provides a useful framework.
When an individual is asked, either by a survey interviewer or a friend, to report her
issue opinion, Zaller asserts that the accessible considerations relevant to this issue are sam-
pled, and the average of these considerations is reported. Opinion stability can be defined
as the inverse of the variance of repeated sample averages of considerations. Individuals
often appear to have unstable opinions about issues (Converse, 1964) because at any point
in time very few considerations about an issue may be accessible and the accessible consid-
erations may depend on seemingly random life occurrences (e.g., radio report, water cooler
conversation). The model predicts that this instability will be especially acute when an
individual rarely receives and accepts arguments about that issue and further does not per-
ceive a link between her political worldview and considerations about the issue. Stability
increases when issues are averaged together, since a large set of considerations is relevant.
This prediction is consistent with recent evidence from surveys (Ansolabehere et al., 2008).
2
An alternative explanation is based on cognitive dissonance rather than source evaluation (or credi-
bility). In this case, voters accept only considerations that are consistent with their existing issue beliefs.
However, numerous studies find that voters often switch their position to adopt the stance of their favored
candidate. See discussions in Sections 2.2.3 and 3.2.2.
13
Although Zaller contrasts his theory with an online model (Lodge et al., 1989), the
differences between the two perspectives are not significant for the purposes of this dis-
sertation.
3
While memory-based and online models differ in the causal mechanisms of
political opinions, the outcomes are generally not in conflict. Specifically, Lodge et al.
(1989) argue that memories cannot be sampled and evaluated in an unbiased manner, but
Zaller’s 1992 Resistance Axiom incorporates this bias. Zaller argues that for political is-
sue opinions (as opposed to evaluations of politicians), the online model does not account
for the wide variance in people’s opinions over time. However, the online model never
specifies how prone the current evaluation is to change. Also, despite Zaller’s “top of the
head” language (p. 49), he allows the full history of consideration to be sampled (p. 121).
This time-invariant weighting of considerations compares favorably to the online model’s
integration of all considerations over time into one evaluation. Importantly, both models
(Zaller, 1992; Lodge and Taber, 2000) hypothesize motivated reasoning, in which individu-
als’ conclusions are biased by their preconceptions. Thus, while these two models disagree
on the inner workings of voters’ minds, they share fairly consistent predictions.
The theory outlined here is an extension of Zaller’s model rather than the online model
for two reasons. First, Zaller’s RAS model, especially the “accept” step, is consistent with
the cycle of partisan bias. Second, in stark contrast to the online model, Zaller’s “sample”
step implies that reported political judgements are draws from random variables.
This sampling also differentiates Zaller’s model from a Bayesian updating process (e.g.,
Gerber and Green, 1998). If voters were pure Bayesians, they would report a summary
statistic of their beliefs when queried on a survey. But as early psychological experiments
(Grant et al., 1951) persuasively demonstrate, individuals draw from their belief distribu-
tion when reporting opinions. For instance, in Grant et al. (1951), laboratory participants
3
In the online framework, voters maintain and update an aggregate opinion of issues, people, and
information sources; voters do not remember the reasons underlying such opinions.
14
were asked whether a light bulb would turn on or not. They quickly learned (in one set of
trials) that the light bulb would turn on 75% of the time, but instead of maximizing their
probability of being correct and always predicting the light would turn on, 25% of the time
they predicted the light would not turn on.
4
Given this empirical evidence, the Personal
Experience Model assumes that individuals draw from top-of-the-head considerations and
are not perfect Bayesians.
However, Bayes rule does provide a convenient mathematical way to combine new data
into a distribution. Bayes rule is therefore referenced when calculating the mean of posterior
distributions of considerations. Certainly, these theoretically determined distributions are
subject to human fallacies, such as forgetting information (i.e., considerations) at variable
rates. These errors are unmodeled and do not alter the basic hypotheses.
2.1.2 Extending Zaller by Introducing Personal Experience
To answer the questions of where partisan biases originate and why certain segments of the
population shift partisan loyalties, Zaller’s model is extended by adding personal experience
as a source of political considerations. In the Personal Experience Model, a steady flow
of experiences related to a political issue leads to a relatively constant set of accessible
considerations, and hence a stable political opinion. That claim is formally stated as a
postulate below; it is one of the four predictions of the Personal Experience Model, along
with two main hypotheses and one corollary.
4
By always predicting that the light bulb will turn on, a perfect Bayesian would be correct 75% of
the time. Zalleresque individuals, sampling from a set of considerations of previous light bulb results (the
FIFO queue), would predict the bulb to turn on 75% of the time (and be correct 62.5% of the time). In the
experiments, across the entire participant group, 75% of predictions were for the bulb to be on, meaning
that there was not a mix of Bayesians and Zalleresque individuals, but only Zalleresque individuals in the
population. These Zaller-following individuals would have to sample only one consideration to make the
“on” prediction 75% of the time. Zaller admits that sampling one consideration is possible (p. 49).
15
“Personal experience” is defined as interacting with a political issue, whether this in-
teraction is sought after (e.g., being an environmental activist) or merely incidental (e.g.,
being retired and receiving Social Security checks). Frequent conversations about political
issues or major life events also count as personal experiences. The key factors for deter-
mining whether an interaction is considered personal experience in this context are (1)
the individual plays an active role, and (2) the interaction produces an increased flow of
received considerations on the issue. Examples include owning a gun, immigrating, having
an abortion, being in an occupation that deals with a political issue (e.g., doctor, teacher),
and discussing politics with friends who have a large stake in a political outcome (e.g.,
gays and equal rights). An “active role” need not be a large role: cashing a Social Security
check counts as active (though barely), but listening to political news does not.
The first criterion, that the individual plays an active role, is necessary to lessen the
degree to which partisan filters affect which information is received. Considerations as
a result of personal experience are more likely to be believed and accepted, even if they
contradict preconceptions. Certainly, there are exceptions to the rule, in which prior beliefs
or norms dominate people’s first-hand perceptions of the world (Sherif, 1936). However,
experiments demonstrate that even young children understand the concept of “seeing is
believing” (Mitchella et al., 1997).
Receiving information from elite sources, such as the media, fails to meet this criterion
because (1) a person’s political worldview may affect which news sources she trusts and
(2) news reports often present political parties commenting on current events. Zaller cites
a journalist stating, “we don’t deal in facts, but in attributed opinions” (page 315). News
information flows through partisan filters and cue-taking continues unabated.
The second criterion for political experience, an increased flow of considerations (rep-
resented by the thick arrow in Figure 2.1), leads to stable considerations in three ways.
16
First, the experience is often based on repeated interactions or one important event, so the
distribution of considerations is narrow. Second, experience leads individuals to process
these considerations in a consistent manner (Fazio and Zanna, 1981; Wood, 1982). Third,
the increased frequency of acceptance of considerations leads to a larger number of political
arguments being accessible(Sherman et al., 1982). Thus, when the set of considerations
of an individual with issue experience is sampled (e.g., by a survey interviewer), a consis-
tent set of considerations is accessible and the resulting opinion distribution has a small
variance.
That some political experiences consist of interactions of a single type causes the re-
sulting considerations to be narrow in scope. For instance, consider a teacher who works
at a crumbling school. Experiences generated from this interaction are likely to indicate
that public schools need more funding. The single-sided considerations generated by this
personal experience contrast with the two-sided considerations presented in news reports
about school budgets or vouchers.
Not only do some individuals receive similar experiences over time, but the similar
situations in which they have these interactions are most likely to be stable. Consider a
new nurse in the health care field. As the nurse learns the ins and outs of the hospital
at which she works, she develops a better understanding of the successes and failures of
the health care system. A nurse and a patient who both witness a failure of the health
care bureaucracy may reach different conclusions about the system. In fact, the patient
might not be able to accurately identify the underlying problem. Individuals with repeating
personal experiences are better able to generate considerations from interactions.
5
The result of repeated, narrow experiences being processed under similar circumstances
leads to a locus of considerations, often on one side of the issue. Since the individual
5
Similar conceptions of repeated interactions are present in advertising literature, which emphasizes that
repeating pictures or phrases affect future actions (Sawyer, 1973).
17
is experiencing these considerations herself (rather than receiving the information from
another party), the considerations are likely to be accepted. While these considerations
may not be explicitly political, current events, from time to time, may encompass the
issues that a voter deals with on a daily basis.
6
Consequently, considerations generated by
experience are likely to be sampled when the issue is explicitly raised by a politician or a
survey interviewer.
7
The output of Zaller’s model under repeated experiences is a stable
(though not necessarily extreme) political opinion.
Stability Postulate: Individuals who have experience with an otherwise political issue
in their non-political lives will have more stable opinions on the issue.
The three components of the second criterion are not individually necessary for the
Stability Postulate to hold. For instance, the relationship between experience and stability
holds regardless of whether the experiences received are ideologically one-sided or two-
sided. In the former case, where the voter’s considerations are all on one side of the
ideological spectrum (e.g., owning a gun and wanting to keep it for hunting), the process
of developing a stable opinion is straightforward. However, the personal experience of the
nurse dealing with the health care system might lead her to understand both sides of the
ideological debate. In this case, the desire to hold a consistent worldview (the mechanism
behind cognitive dissonance) might lead the nurse to develop a nuanced view of health care
policy.
As detailed in Section 2.2.2, studies demonstrate that personal experience often leads
voters to develop a more sophisticated approach to the issue in question. The key feature
6
Zaller’s (1992, Chapter 4) example of a teacher who realizes schools’ need for funding yet is confused
about the Federal budget is an example of considerations generated through personal experience that are
not directly related to the interviewer’s political question.
7
Experienced voters may be more likely to pay attention when a politician raises a particular issue
(Hutchings, 2003).
18
of personal experience is the heightened flow of received, accepted, and accessible political
considerations that are independent of political predispositions.
A second set of circumstances in which personal experience may lead to stable issue
opinion without the effects of the three sub-criterion occurs when an individual experiences
by a single, life-altering incident. After an experience, such as having an abortion or being
the victim of a violent crime, the voter will carry considerations generated by the experi-
ence, which may never be forgotten and always be readily accessible. Thus, the experience
has the same effect of the first and third sub-criterion—narrow, frequent interactions—
in that the same considerations are always accessible when touched upon by the politics
events.
Returning to Zaller (1992), this stability framework is referenced when he discusses
attitude change (Chapter 7). When an individual receives political communication on an
issue, she combines this new information with existing considerations. If the existing con-
siderations carry a large weight relative to the new information, then the new information
will have little impact on the reported opinion. The next section further develops this idea
using a learning model, which will demonstrate why issue stability is crucial in the arena
of voter persuasion.
2.1.3 Learning Model: Issue Positions
Following the lead of Achen and Bartels (2006), the Personal Experience Model represented
in Figure 2.1 can be formalized using a learning model. In its simplest form, the model
includes one voter, one issue, and one candidate.
8
The model has two periods: before and
after the candidate sends a signal (i.e., political communication) on the issue. Prior to
receiving a candidate’s signal, the voter has a distribution of Zalleresque considerations
8
For clarity, I drop the usual subscript i to indicate a single voter.
19
about the issue, centered at δ
1
. This opinion has a precision (Bartels, 1993), τ
2
1
, which,
if the Stability Postulate is correct, increases with personal experience.
9
For simplicity, I
assume that this prior is normally distributed; hence, in period 1, the voter’s issue position,
x
1
, is the distribution,
x
1
∼ ^(δ
1
,
1
τ
2
1
). (2.1)
At the end of period 1, a candidate announces a position, γ, on the issue. Not all
signals are created equal. The persuadability (ψ
2
) of the candidate’s message (i.e., signal)
depends on political factors such as how forcefully the candidate argues for the position
and whether voters consider the position “cheap talk.”
Voters update their beliefs about the best policy by adding the candidate’s consideration
to the original distribution.
10
To avoid discretization of the posterior belief, I model the
candidate’s announcement as a normal curve centered on γ and with precision ψ
2
. The
voter’s period 2 issue opinion is a mixture distribution with a mean and a variance of
E[x
2
] =
δ
1
τ
2
1
+γψ
2
τ
2
1

2
(2.2)
Var[x
2
] = 1 +
τ
2
1

1
−E[x
2
])
2

2
(γ −E[x
2
]
2
)
τ
2
1

2
(2.3)
The posterior distribution (x
2
) does not follow Bayesian updating because Zaller’s frame-
work implies a mixture distribution (see discussion 2.1.1). However, both framework’s
predict the same mean (Equation 2.2).
9
Instead of modeling stability in terms of variance, I use precision (the inverse of variance) because
precision more closely aligns with the concepts of the model.
10
In this instantiation of the model, voters are assumed to retain all considerations. Forgetfulness (i.e.,
the FIFO queue) is revisited in succeeding sections. However, forgetting period 1 considerations can be
approximated by increasing ψ
2
, thus increasing the impact of the candidate signal.
20
A concrete example of this model, albeit outside of the campaign framework, is Al
Gore’s movie An Inconvenient Truth. Before the movie was released, the environment was
not often a topic in America’s political discourse. Even liberals, while generally believing
in environmentalism, may have held unstable beliefs about how many resources the United
States should commit to stop global warming. This uncertainty is depicted by the wide
distribution in the left panel of Figure 2.2. Gore’s documentary sends a strong signal to
the left of the voter’s mean prior distribution (center panel). The voter accepts this signal
and becomes a fervent environmentalist (right panel).
Voter's Prior Beliefs
Position on Environment
−1.0 −0.5 0.0 0.5 1.0
Candidate's Signal
Position on Environment
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Voter's Posterior Beliefs
Position on Environment
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Figure 2.2: A hypothetical liberal’s response after viewing An Inconvenient Truth. The
liberal’s initial position (left panel) is moved close to the position espoused by Gore (center
panel), who is a favored politician. The right panel displays the resulting issue opinion
distribution in bold. Parameters: x
1
= −0.2; τ
2
1
= 1/(0.3
2
); γ = −0.5; ψ
2
= 1/(0.15
2
).
Brody and Page (1972) would label this interaction between the liberal citizen and Al
Gore’s documentary “persuasion.” However, since the liberal knew little about environ-
mentalism before viewing the documentary, she could not have thought critically about
the subject. Instead, the liberal accepted the cue from Al Gore, with no ability to judge
21
the quality of the arguments. Following other scholars (e.g. Gilens and Murakawa, 2002),
I label the shift of opinion by the liberal as “cue-taking.” I reserve the term “persua-
sion” for circumstances akin to “central route processing” (Petty and Cacioppo, 1981), in
which a voter’s issue opinion shifts based on critical analysis of the current set of issue
considerations.
Because most people go about their lives generally oblivious to many political issues,
cue-taking is prevalent in politics. Campbell et al. (1960) emphasize a similar point when
they note, “For many voters the details of the political landscape may be quite blurred
until they are brought more into focus during the campaign period” (pp. 135-136). If the
campaign environment induces cue-taking, then a straightforward application of the model
predicts that campaigns with dueling messages should polarize public opinion when a new
issue is introduced to the political landscape.
Consider a campaign environment in which both the Democratic and the Republican
candidates provide opposing signals to two voters (also one Democrat and one Republican)
about an issue. The voters accept only the signal from the candidate they trust more.
11
If the voters have unstable opinions about the issue (i.e., little personal experience), then
their opinions will polarize (Figure 2.3). This conclusion, supported by existing research
(Ansolabehere et al., 1994; Parker-Stephen, 2007), is crucial: when voters do not have
experience on an issue and this issue is emphasized during the campaign, the electorate
polarizes and individuals’ existing political worldviews are not altered.
Alternatively, if an individual has experience on an issue, signals from elites may fail to
confirm the individual’s current worldview. Consider the following example of a campaign
between two candidates of the major parties who are debating the issue of gun control. The
Republican candidate has a moderate position, in contrast to the Democratic nominee’s
11
See Section 2.1.5 for a formalization of this statement.
22
Voters' Prior Beliefs
Issue Opinion Scale
(Blue=Dem, Red=GOP)
−1.0 −0.5 0.0 0.5 1.0
Candidates' Signals of Issue Stances
Issue Opinion Scale
(Blue=Dem, Red=GOP)
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Voters' Posterior Belief
Issue Opinion Scale
(Blue=Dem, Red=GOP)
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Figure 2.3: Issue polarization. When two opposing candidates send signals (center panel)
to two moderate voters of different parties (left panel), the voters’ opinions diverge in
opposing directions (right panel). Democratic voter and candidate parameters: x
1
= −0.1;
τ
2
1
= 1/(0.3
2
); γ = −0.7; ψ
2
= 1/(0.15
2
). Republican voter and candidate parameters:
x
1
= 0.1; τ
2
1
= 1/(0.3
2
); γ = 0.8; ψ
2
= 1/(0.25
2
).
more orthodox, left-wing view. Of the two voters analyzed in this example (Figure 2.4,
left panel), the Democrat is a hunter who has experience with the issue (i.e., precise prior
opinion) while the Republican does not (wide prior opinion). In between periods 1 and
2, the Democratic candidate sends a leftist signal and the Republican candidate sends a
moderate, right-of-center signal (Figure 2.4, center panel). While the Republican voter cue-
takes and moves slightly to the right to follow the Republican candidate, the Democratic
hunter does not move left to any significant degree (Figure 2.4, right panel). In general,
this phenomenon, derived mathematically as
∂|γ−E[x
2
]|
∂τ
2
< 0, forms the first of the two main
hypotheses:
Cue-Taking Hypothesis: Ceteris paribus, when a favored political elite provides a
signal about his position on an issue, the segments of the population that have the least
23
personal experience with the issue are most likely to cue-take (i.e., alter their issue position
in the direction of the signal).
Voters' Prior Beliefs
Issue Opinion Scale
(Blue=Dem, Red=GOP)
−1.0 −0.5 0.0 0.5 1.0
Candidates' Signals of Issue Stances
Issue Opinion Scale
(Blue=Dem, Red=GOP)
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Voters' Posterior Belief
Issue Opinion Scale
(Blue=Dem, Red=GOP)
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Figure 2.4: Experienced voters do not cue-take. The Democratic hunter does not cue-
take from her preferred politician on this issue because of her highly-informed prior on the
issue. The non-experienced Republican voter shifts her issue opinion toward the stance
of her preferred candidate. Democratic voter and candidate parameters: x
1
= −0.1;
τ
2
1
= 1/(0.13
2
); γ = −0.8; ψ
2
= 1/(0.25
2
). Republican voter and candidate parameters:
x
1
= 0.1; τ
2
1
= 1/(0.3
2
); γ = 0.15; ψ
2
= 1/(0.25
2
).
Returning to the example in Figure 2.4, the Democratic voter’s opinion is closer (on
average) to the Republican candidate’s view on this issue, causing a contradiction in the
previously loyal Democrat’s worldview. To calculate how much more negatively the Demo-
cratic voter would rate the Democratic nominee, one must consider both where the voter
thought the candidates stood on the issues in period 1 and how the voter judges politicians.
2.1.4 Voters’ Beliefs About Candidate Positions
Voters have beliefs about where candidates stand on the issues; these beliefs become more
certain after receiving signals about the candidates’ platforms. Before a signal is sent,
24
a voter’s beliefs about the candidate’s positions often suffer from “projection” (Brody
and Page, 1972), whereby the voter assumes that her preferred candidate holds the same
position that she does.
12
I do not explicitly model the determinants of projection, instead
allowing the center of the distribution of prior beliefs about the candidate’s position, µ
1
,
to be an input parameter. Under complete projection, µ
1
= δ
1
. Voter have certainty about
these beliefs, υ
2
1
, which most likely increase with political attentiveness and sophistication.
Similar to equation 2.1, the prior belief y
1
is assumed to be the normal distribution,
y
1
∼ ^(µ
1
,
1
υ
2
1
). (2.4)
The candidate’s signal helps the voter identify where the candidate stands on the issue.
The signal has the same center point, γ, and precision, ψ
2
, as in Section 2.1.3. A candidate’s
signal could be unpersuasive (i.e., not move an individual’s issue position) yet credible
(i.e., be informative about where the candidate stands), so that learning and cue-taking
are based on different interpretations of a single signal; for simplicity, I model these signal
characteristics as the same parameter. Certainly, these qualities could be correlated in
certain cases. For example, a politician who appears to be pandering would produce an
un-credible and non-persuasive issue appeal.
As with the voter’s own issue position, the voter updates her belief about the candidate’s
positions by introducing considerations generated by the candidate signal. Beliefs at period
1 are treated as priors in period 2. The updated, period 2, beliefs about the candidate are
12
In addition to projection, a voter’s prior belief on a candidate’s position may be based on general factors
such as the candidate’s party.
25
a mixture model with mean and variance,
E[y
2
] =
µ
1
υ
2
1
+γψ
2
υ
2
1

2
(2.5)
Var[y
2
] = 1 +
υ
2
1

1
−E[y
2
])
2

2
(γ −E[y
2
]
2
)
υ
2
1

2
(2.6)
Returning to the case of the Democratic hunter and gun control, imagine that the
Republican candidate microtargets the Democratic hunter, highlighting both candidates’
positions. Before the signals are sent, the voter might have weak priors on the candidates’
positions according to their parties (left panel, Figure 2.5). The Republican sends a mailing
to the gun owner’s household indicating a moderate-right stance for the Republican and
an extreme liberal position for the Democrat (center panel). The Republican has more
credibility in revealing his own position than his opponent’s because the Republican might
be held to account for his stated issue opinion if he wins (Ferejohn, 1986). These signals
are combined with the voter’s prior and result in a posterior belief about the candidates’
positions (right panel).
2.1.5 Voters’ Candidate Evaluations
Traditionally, candidate evaluation is modeled as utility maximization (Downs, 1957) over
a range of issues (Enelow and Hinich, 1984). These types of models, where voters place
weights on a range of issues, are consistent with the issue public literature (Krosnick, 1990).
Members of an issue public place above-average weight on a specific issue. These models
of voter utility require a lot from citizens. First, voters must know their own position on
all the issues. Second, voters must know the politician’s view on each issue. Third, voters
must assign a weight to each issue. The weight-based model requires that voters keep track
of all this information.
26
Prior Beliefs of
Candidates' Positions
Gun Control Opinion
(Blue=Dem Cand., Red=GOP Cand.)
−1.0 −0.5 0.0 0.5 1.0
Republican Cand.'s Signals of
Candidates' Positions
Gun Control Opinion
(Blue=Dem Cand., Red=GOP Cand.)
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Posterior Beliefs of
Candidates' Positions
Gun Control Opinion
(Blue=Dem Cand., Red=GOP Cand.)
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Figure 2.5: A Voter integrates her projection-based beliefs with one candidate’s signals
about both candidates’ opinions. The voter starts out with prior beliefs about both can-
didates’ positions based on their party. The Republican candidate signals that he has a
moderate-right position and signals, with less credibility, that the Democrat has an extreme
left position. The voter updates her beliefs about the candidates’ positions accordingly. Pa-
rameters for belief and signal about the Democratic candidate: y
1
= −0.5; υ
2
1
= 1/(0.4
2
);
γ = −0.8; ψ
2
= 1/(0.8
2
). Parameters for belief and signal about the Republican candidate:
y
1
= 0.5; υ
2
1
= 1/(0.4
2
); γ = 0.15; ψ
2
= 1/(0.25
2
).
Instead, I propose a more Zalleresque model of candidate evaluation. Voters compare a
candidate’s issue signal against their considerations of the issue. For instance, when a can-
didate delivers a signal about an issue, the voter updates her belief about the candidate’s
position, forming the posterior belief y
2
(as in Section 2.1.4). Next, to evaluate the candi-
date on this issue, the voter compares her prior belief about the issue to the candidate’s
perceived position. This comparison is formalized with the Kullback-Leibler distance,
KL(y
2
[[x
1
) =


−∞
y
2
(z)log
y
2
(z)
x
1
(z)
dz , (2.7)
27
where z is the issue dimension and, as above, x
1
is the voter’s issue position at period 1.
13
The use of the Kullback-Leibler distance is appropriate for a Zalleresque model because the
entire distribution of considerations is compared. This contrasts to a Bayesian model, in
which means (or other summary statistics) are compared (e.g., Enelow and Hinich, 1984)
yet the exact distribution is unimportant.
For Gaussian distributions, the Kullback-Leibler distance has the analytical solution,
KL(y
2
[[x
1
) =
1
2

2 log

υ
2
τ
1

+

τ
2
1
υ
2
2

+ (µ
2
−δ
1
)
2
τ
2
2
−1

, (2.8)
where µ
2
and υ
2
2
are the mean and precision of y
2
(Equations 2.5 and 2.6). However, x
2
and y
2
approximate an unmixed Gaussian distribution only in cases in which one of their
constituent Gaussian distributions dominate the others. In addition, Equation 2.8 is an
idealized, continuous version of the discrete process of comparing a voter’s considerations
about an issue to considerations of a candidate’s platform. Thus, Equation 2.8 is sometimes
a poor approximation for Equation 2.7.
Candidate stances on issues for which the voter has a vague prior will have little im-
pact on the voter’s worldview. The uncertainty of the voter’s position means the Kullback-
Leibler distance will not be extreme in either distance or closeness. Once the voter cue-takes
on this issue (Section 2.1.3), the K-L distance will shrink, reinforcing existing predisposi-
tions.
Candidate evaluations may not be reinforced when the voter has a precise prior on the
issue in question. In this case, the voter compares her beliefs with the candidate’s and the
difference may be stark. An apt analogy is that the voter treats her personal experience
as the “ground truth” for judging the elite’s reliability. One incongruent signal from a
13
x1(z) and y2(z) are the densities of x1 and y2 at z, respectively.
28
preferred candidate might be enough to induce the voter’s defection. Understanding the
full voter-candidate interaction requires the consideration of multiple issues.
To calculate the voter’s overall utility of a candidate, all considerations on the top of
a voter’s head must be taken into account, and these considerations may be on multiple
issues. Let the scope of J issues be represented by j = 1...J and the positions of the voter
and the candidate at time t be x
jt
and y
jt
, respectively. The voter need not have all these
issues on the top of her head at one particular time. Which combination of issues is on her
mind determines the voter’s candidate evaluation, which is formally,
U(y

[x

) = −
J
¸
j=1
δ
j
∗ KL(x
tj
[[y
tj
), (2.9)
where δ
j
is the proportion of a voter’s considerations dealing with issue j. The voter
need not take an active role in consciously setting δ
j
for all issues; for issues the voter is
unfamiliar with, δ
j
= 0. In a two-sided issue debate, the model assumes voters receive only
signals from the candidate with the higher U(y

[x

).
As noted, a candidate signal on a new, unfamiliar issue would lead to a small K-L
distance after cue-taking, thus reinforcing predispositions. However, if considerations of
this new issue replace considerations of an older issue for which the voter also took an elite
cue, candidate utility would not change at all.
The chance of a switch in candidate support is particularly acute in three situations. In
the first, the voter is nearly undecided in the first period; e.g., KL(x
1
[[y
D
1
) ≈ KL(x
1
[[y
R
1
),
where D and R superscripts represent the two candidates. In this case, a small change in
the relative K-L distances results in a change in vote choice.
Second, priming may occur; “priming” is a traditional political science explanation of
why voters alter their opinions of candidates once a signal is received (Iyengar and Kinder,
29
1989). The signal about a particular issue (j) causes considerations about that issue to
become more accessible in period 2. When asked to evaluate the candidates, the voter
relies more heavily on these primed considerations; formally, δ
j
increases.
Third, if the voter has a precise prior about only the one issue at hand (i.e., the
voter has no experience on other issues), then the voter is susceptible to micro-targeting
by the non-preferred candidate. Consider the example of the Democratic hunter. The
Republican’s signals about gun control indicate that the Republican candidate is closer
to the voter on this issue. Remaining loyal to the Democratic candidate would cause
cognitive dissonance; the hunter does not cue-take from her initially preferred candidate
(the Democrat) since she is set in her ways. The equations above demonstrate that the
voter will alter her perceptions of the candidates to favor the Republican. Transforming
the voter’s worldview and cue-taking on all the other (non-experiential) issues from the
Republican candidate causes no dissonance. Hence, this Democratic hunter is susceptible
to defection by a well-targeted Republican appeal.
This dissertation focuses on the third mechanism: voters’ political worldviews can
be altered by candidates’ stances on the issues with which the voter has experience. This
phenomenon, derived mathematically as
∂KL(y
2
||x
1
)
∂|µ
2
−δ
1
|∂τ
2
1
> 0, results in the following prediction,
which is the main hypothesis of this dissertation:
Candidate Evaluation Hypothesis: Ceteris paribus, after receiving a candidate’s
signal on an issue, the segments of the population that have the most personal experience
with the issue will alter their perceptions of the candidate more than the segments of the
population that have no personal experience.
Political worldviews originate with and change based on issues for which voters have
experience. For these issues, voters know the “ground truth” and need not rely on political
elites for an indication of what to believe.
30
This hypothesis might be particularly relevant during primaries, where no partisan dif-
ferentiation exists. Voters’ preferences between candidates with similar ideologies could
easily hinge on the evaluations generated by this ground truth effect. Especially in pri-
maries, candidates may agree on the salient, oft-discussed issues. Since the Candidate
Evaluation Hypothesis does not rely on the pivotal issue being of extrinsic importance to
the voter, the hypothesis still predicts changes in political judgments based on a familiar
issue.
Consider the hypothetical example of a primary election between two candidates with
an electorate of three voters, each of whom has experience on a particular issue. The 20-day
campaign moves through three phases and focuses on two issues: health care and education.
The three voters are Alice the architect, Ted the teacher, and Doris the doctor; they must
decide between two the candidates, Mandy and Nathan. Figure 2.6 tracks the hypothetical
voters’ candidate evaluations over the course of the campaign.
14
They start with opposing
predispositions about the candidates, with Alice favoring Mandy, Ted favoring Nathan,
and Doris undecided. These predispositions may be attributable to valence evaluations
(e.g., competence, experience) or other traits (e.g., shares my values), which in turn may
originate from identity politics (e.g., voter-candidate shared gender or race).
15
Candidate
traits are modeled just as issues are; for traits, voters are very certain about where they
stand (e.g., competence is good) but less certain about politicians’ stances (e.g., claims of
competence are cheap talk).
The left-most portion of Figure 2.6 displays these starting preferences. No issue con-
siderations are on the top of voters’ heads on the initial day of the campaign (phase 1).
14
See Table A.1 of Appendix A for the exact parameter values that generated the figure.
15
In a general election, these predispositions are often aligned with partisan allegiances and are very
strong. A primary campaign magnifies the effect of issues.
31
5 10 15 20

4

2
0
2
4
Days of Campaign
R
e
l
a
t
i
v
e

E
v
a
l
u
a
t
i
o
n

o
f

C
a
n
d
i
d
a
t
e
s

(
M
a
n
d
y



N
a
t
h
a
n
)
Alice the
Architect
Doris the
Doctor
Ted the
Teacher
S
t
a
r
t

o
f

C
a
m
p
a
i
g
n
Health Care Phase Education Phase
More Pro−Mandy
More Pro−Nathan
Figure 2.6: A hypothetical primary campaign with three voters, two candidates, and two
issue phases (three phases total). In the first issue phase, when health care is discussed,
the initially undecided doctor (Doris) sides with the candidate (Nathan) who concurs with
her stable opinion on health policy. The other two voters polarize. In the second phase,
when education is the hot topic, the teacher (Ted) switches candidate allegiances because
his previously preferred candidate (Nathan) takes what Ted considers an incorrect stance.
In the next phase of the campaign (days 2-10), the issue of health care becomes promi-
nent and politicians stake out positions. Considerations related to health care are added
to the tops of voters’ heads. The only voter with experience on this issue is Doris (who is
a doctor). She sides with candidate Nathan on the issue, and increases her relative eval-
uation of him. The other two voters polarize in their opinion of the candidates, as actual
issue information (rather than trait information) solidifies their respective choices.
After the candidates have finished debating health care, education policy becomes rel-
evant in phase 3, days 11-20. Alice and Doris cue-take from their preferred candidate;
Ted the teacher sides with his previously unfavored candidate, Mandy. The education
debate eventually crowds out health care considerations. By this point, on the last day of
32
the hypothetical campaign, Ted switches candidate allegiances and supports Mandy over
Nathan.
16
This simulation illustrates the futility of campaigns’ attempts to alter the minds of
voters on a host of issues with which the voters have no experience.
17
In the normal course
of a campaign, two-sided elite messaging on issues unfamiliar to voters leads to polarization,
just as Alice and Ted diverge in the health care phase of the campaign. A real-world
example from the 2008 general election is the sudden emphasis on wealth redistribution
after the third debate between Obama and McCain (in which “Joe the Plumber” was
referenced 24 times). Despite this shift in campaign dialogue, the polls barely moved.
18
The simulation also demonstrates a distinction between the Personal Experience Hy-
pothesis and other alternative hypothesis: issue salience is constant across voters yet het-
erogeneous political judgments occur. The critical factor for the an issue to induce political
ground truthing is a stable opinion on the issue. Personal experiences, which occur regard-
less of which issue is on the political front-burner, are an excellent generator of these stable
opinions.
2.1.6 Issue Experience as Signals
In the Personal Experience Model candidate signals and issue experiences are variations on
the same theme. They are both considerations that are received by voters and subsequently
accepted or rejected (Figure 2.1). The most important distinction between these two types
of considerations is that acceptance of candidate signals depends on the voter’s evaluation
16
The decrease in polarization for Alice and Doris at the beginning of the education phase is caused by a
decrease in polarized health-care considerations, which are partially offset by education considerations that
have yet to become fully polarized. As voters learn the candidates’ positions on education and adopt the
position of their preferred candidate, education becomes as polarized as health care and voters’ preferences
diverge.
17
As discussed in Chapter 6, this futility is only a problem for campaigns trailing in the polls.
18
See Charles Franklin’s moving average (http://www.pollster.com/polls/us/08-us-pres-ge-mvo.
php) for the flat McCain and Obama lines before and after the October 16, 2008 debate.
33
of the candidate, while acceptance of personal experiences hinges on the proximity of the
voter-issue interaction.
When a voter interacts with a non-politicized issue, her issue opinion proceeds analo-
gously to the logic of Section 2.1.3. A real-world experience (e.g., a teacher sees the roof
of a school cave in) generates an additional consideration for her issue opinion distribution
(e.g., schools need more funding because they are crumbling). How much this consideration
affects the voter’s opinion depends on the weight (i.e., precision) of the new consideration

2
) relative to the weight of existing considerations (τ
2
).
With the model formally explicated, the three properties of personal experience dis-
cussed in Section 2.1.2 can also be formally linked to greater issue stability. First, repeated
considerations at the same point on the issue spectrum create a narrow distribution of
experiences and a higher precision, ψ
2
. Second, easier processing of experiences leads to
higher reception of considerations, and again a tighter distribution of experiences and a
larger ψ
2
. Third, since voters forget considerations over time, repeated considerations lead
to voters having experienced-based considerations on the top of their head more often. As
long as the precision of experiential considerations (ψ
2
) is greater than the background
noise considerations (τ
2
), increasing the proportion of experiential considerations will lead
to lower issue opinion variance.
19
Formally, ψ
2
> τ
2

∂Var[x
2
]
∂ψ
2
< 0. Infrequent, forgettable
voter-issue interactions may increase the variance of a voter’s opinion.
The process of issue opinion formation becomes more complicated when the individual
has experience with an already politicized issue. For instance, assume a conservative parent
has a narrow prior on the issue of gay marriage because the issue is often discussed in the
political arena and this individual accepts only conservative arguments. If the parent’s child
19
Another type of experience is a one-time, high-impact experience. If these considerations are forgotten
at a slower rate than other considerations (e.g., mass media-based), then the math operates in the same
way as for repeated, more ephemeral experiences.
34
“comes out of the closet,” then the parent’s personal experience will directly conflict with
her existing issue opinion. If the parent accepts these personal experience considerations,
then her issue opinion distribution becomes bimodal, and the parent may have a difficult
time deciding where to stand on the issue (Figure 2.7).
20
This bimodal distribution does
not generate as clear predictions for candidate evaluations as the situations discussed in
Section 2.1.5 where personal experience occurs before issue politicization. As Figure 2.7
displays, personal experience can balance political predispositions – a result found even in
the highly politicized environment of Congress (Washington, 2008).
Voter's Prior Beliefs
Position on Gay Rights
−1.0 −0.5 0.0 0.5 1.0
Personal Experience
Position on Gay Rights
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Voter's Posterior Beliefs
Position on Gay Rights
D
e
n
s
i
t
y

o
f

B
e
l
i
e
f
s
−1.0 −0.5 0.0 0.5 1.0
Figure 2.7: A hypothetical conservative parent’s issue consideration distribution after her
child comes out of the closet. The conservative’s initial position on gay rights (left panel)
is balanced by personal experience (center panel). The right panel displays the resulting
issue opinion distribution in bold. Normally, the conservative would reject the liberal con-
siderations of the center panel, but these considerations are generated by a close, personal
situation. Parameters: x
1
= 0.5; τ
2
1
= 1/(0.15
2
); γ = −0.5; ψ
2
= 1/(0.1
2
).
20
The initial “if” in this sentence is a “big if” in that the parent’s prior distribution would normally mean
that she would not accept liberal consideration. Only because the source of these considerations is so close
to her might these considerations be accepted, and even then perhaps at a slow rate.
35
2.1.7 Model Extensions and Details
The model, as presented in its most limited form, can be applied in alternative situations.
The model works just as well when considering political parties, news organizations, or
politically active citizens as when considering candidates and is flexible enough to be used
in partisan and non-partisan situations.
As modeled, signals from non-preferred candidates do not contain negative informa-
tion. As Zaller (1992) demonstrates with opinions about the Vietnam War, if only the
conservative elite is communicating to the public, then even liberal voters will shift their
opinions toward the conservative end of the spectrum. The shift of liberal opinion is less
than the shift in conservative opinion and may reflect the lower probability of a liberal
accepting a consideration from a conservative elite.
The probability of a voter receiving a signal does not appear in the Personal Experience
Model, but the effects of paying more or less attention to politics can be captured by
the precision of a candidate’s signal. In fact, small acceptance probabilities are captured
in the hypothetical election simulations by assuming weak candidate signals (small ψ
2
).
Alternatively, an additional parameter and process could be explicitly added to the model.
The model can be applied beyond political issues. For example, character traits can be
a powerful force in politics because voters judge individuals in their daily lives. Consider
the trait of honesty. Voters’ ideal points, x
tj
are all essentially grouped at positive infinity
(or the upper limit of the trait dimension)—voters want their politicians to be extremely
honest. Also, most voters know with a high degree of certainty that they prefer honest
people to dishonest people (since they have experience dealing with both in their daily
lives), so τ
2
j
is large. Candidate cues on this topic may lack credibility (i.e., ψ
2
is low)
because politicians often cannot credibly state “I am honest.” Thus, a credible cue from
36
a media source that a candidate is dishonest can have a devastating effect on voters’
evaluations of that candidate.
Character traits are an example of an “issue” with which most of the electorate has
extensive experience. Individuals must judge the honesty, trustworthiness, and empathy
of others on a daily basis. The model’s implications extend to these widely familiar issues.
Aggregating the conclusion from the Candidate Evaluation Hypothesis across the electorate
implies that mass communication about issues with which more voters are familiar (i.e.,
the “easy issues” of Carmines and Stimson, 1990) will influence candidate evaluation more
than communicating about complex, niche issues.
Broadcast Corollary: Signals on familiar issues or candidate character traits influ-
ence voters’ perceptions more than information about more complex issues.
The Candidate Evaluation Hypothesis and the Broadcast Corollary have direct impli-
cations for political campaigns. If a campaign is aware that a particular voter has personal
experience with an issue, then the campaign would benefit from communicating to the
voter on that issue (provided the voter and the candidate have congruent opinions). How-
ever, campaigns are often unable to procure this type of information; instead, they rely on
mass communication such as television advertising. In the case of mass communication,
campaigns would do well to follow the Broadcast Corollary and advertise about issues that
intersect with the lives of the greatest number of people.
2.2 Concurrence with Other Theories in the Literature
The literature substantiates much of the Personal Experience Model, especially the claims
about issue opinion stability (Stability Postulate), cue-taking (Cue-Taking Hypothesis),
and easy issues (Broadcast Corollary). Circumstantial evidence supports the experience-
issue interaction (Candidate Evaluation Hypothesis), yet no study directly tests the effect
37
of personal experience on cue-taking and candidate evaluation. The following sections
review the existing evidence and identify the gap that the Personal Experience Model fills
with the Candidate Evaluation Hypothesis.
2.2.1 Definitions of Concepts in the Literature
Over the decades, various terms (e.g., “projection,” “persuasion”) have represented distinct
concepts. Before exploring the existing research and discussing the empirical results in later
chapters, it is useful to review and define some of these terms.
• Issue Opinion Stability is the precision of the distribution of repeated samples of
an individual’s accessible considerations of an issue. A helpful way to think about
opinion stability is to imagine asking an individual to report an opinion on a repeated
basis. The variance of her reported opinions is the inverse of her issue stability.
• Self-Reported Issue Opinion Confidence is an individual’s belief about her sta-
bility on the issue. Confidence should not be conflated with issue stability as indi-
viduals are often poor judges of their own characteristics.
• Candidate projection is the phenomenon whereby a voter assumes her preferred
candidate agrees with her on an issue that she has relatively little information on.
This mechanism often before candidates send strong signals on the issue. See “pro-
jection” in Markus and Converse (1979).
• Learning is the process by which the voter ascertains the true issue position of the
candidates. It occurs after the voter receives strong signals on the issue. See Lenz
(2006) for an excellent discussion.
• Cue-taking is the process whereby the voter agrees with her preferred candidate’s
issue position without any critical thinking. See “projection” in Iyengar and Kinder
38
(1989), “persuasion” in Markus and Converse (1979), and “peripheral route process-
ing” in Petty and Cacioppo (1981).
• Persuasion on an issue opinion occurs when the voter thinks critically about
issue considerations and rejects them in favor of other considerations. Replacement
of accessible considerations over time is not considered persuasion. See “central route
processing” in Petty and Cacioppo (1981). Persuasion does not play a large role in
the Personal Experience Model.
• Priming is the mechanism by which voters increase the salience of a particular issue
in the vote decision. This salience may be raised directly by stressing the importance
of an issue or indirectly by mentioning the issue and thus increasing the accessibility
of considerations related to the issue. See Iyengar and Kinder (1989).
2.2.2 Personal Experience Yields Nuanced and Stable Opinions
The finding that personal (or “direct”) experience with an issue induces attitude stabil-
ity originates in the psychology literature. Regan and Fazio (1977) employ both a field
experiment and a laboratory experiment to test the effect of direct experience on attitude-
behavior consistency. Their experiments randomly assigns the direct experience treatment.
In the field experiment, college students are assigned to permanent or temporary housing.
In the laboratory experiment, some participants are assigned puzzles to work on, while
others receive only descriptions of the puzzles. In both experiments, the reported attitudes
of the participants assigned to the “direct experience” treatment are more consistent with
the participants’ actions than for the students assigned to the peripheral experience.
Wood (1982) examines how existing experiences affect the amount of variation in issue
opinions. Subjects in this experiment are asked to “to list specific instances of times when
they had engaged in actions relating” to the environment and to report their opinions about
39
environmentalism. A month later, Wood requests that the subjects prepare a persuasive,
pro-environmentalism talk in exchange for five dollars. On average, this offer of money
induces subjects to report views that are more anti-environmentalism than their original
position (presumably because the offer of money makes the arguments appear fraudulent).
However, individuals with prior experience with the environmental movement are less likely
to change their opinion.
21
Observational studies demonstrate that individuals with experience on an issue have
more stable (and sometimes more nuanced) opinions, supporting the Stability Postulate.
Sotirovic (2001) conducts a telephone survey of 395 Wisconsin adults to examine their
attitudes toward crime policy and the death penalty. The respondent’s answers to open-
ended questions are coded for number of distinct ideas and integration between ideas;
the answers are then aggregated into a “complexity” measure. Respondents who have
experience with non-trivial crimes (either personal or through friends) demonstrate higher
thought complexity.
McFadyen (1998) conducts in-depth interviews with 67 employed Britons to probe
their views on unemployment. The author asks participants about stereotypes of the un-
employed, the efficacy of government actions, and their thoughts about possible solutions.
To gauge the level of experience that participants had with unemployment, McFadyen
queries the subjects about their relationships with unemployed family or “close” friends,
as well as any direct experience they have had with being unemployed. A multivariate
analysis finds that amount of experience with the unemployed is the best predictor of how
few stereotypes the respondents express, with experience being more predictive than either
ideology or class and leading to fewer stereotypes.
21
This change is not caused by differences in initial opinions between the experienced and unexperienced
groups.
40
In both the crime and unemployment studies, experience does not lead to polarization
of issue opinion. In fact, just the opposite occurs. In the crime study, Sotirivic finds that
the respondents with the most extreme views had the least complex thought processes.
And personal experience with the unemployed does not push the Britons in McFadyen’s
study to the left on unemployment policy; political ideology is a much better predictor of
issue opinion than level of issue involvement. Consistent with these findings, the direct
effect of experience on left-right issue position is not specified by the Personal Experience
Model.
The impact of personal experience on issue opinion stability is directly measured by van
Knippenberg and Daamen (1996). The authors poll the attitudes of the Dutch on energy
generation twice, through surveys two months apart. In each survey, respondents are given
a range of six methods of generating energy and are asked to pick two. In the first survey,
brief descriptions of each option are provided. Van Knippenberg and Daamen measured
issue involvement on a “slightly modified version” of the scale developed by Verplanken
(1991), which queries respondents about their activities with respect to the issue (e.g.,
talking with friends about the issue). While they do not provide exact statistics, the
authors report that higher involvement leads to increased stability across the two-month
period.
Taken in sum, these studies demonstrate that direct experience, personal involvement,
or even the experience of close friends or family members induces a more complex thought
process about an issue. The resulting opinions of individuals with personal experience are
more stable than those of the public at large. Thus, the evidence for the Stability Postulate
is strong.
41
2.2.3 Evidence for Cue-Taking and the Role of Experience
The psychological literature provides the basis for cue-taking. Similar to Bartels’ (2002)
findings that the same information generates polarized responses between members of op-
posing parties, Lord et al. (1979) show that conflicting studies on capital punishment have
a polarizing rather than moderating effect. Undergraduates who hold prior beliefs in fa-
vor of capital punishment consider the pro-capital punishment study more convincing;
consequently, their issue opinions become more extreme. The opposite holds for the par-
ticipants who enter the experiment with anti-death penalty views. Similar to the Al Gore
and global warming hypothetical in Section 2.1.3, individuals generally accept arguments
that are congruent with their existing beliefs and adjust their views accordingly.
Political science research provides several examples of cue-taking. Zaller (1992) demon-
strates how elite consensus leads to mass consensus. In contrast to Zaller’s one-stream
example, campaigns provide an example of a two-stream environment. As the election was
heating up in the summer of 2004, Democrats and Republicans diverged over even the
relatively factual issue of whether Saddam Hussein was involved in the September 11th
attacks (Parker-Stephen, 2007). In a striking example of issue opinion following elites from
one end of the spectrum to the other, Lauderdale (2008) finds that a change in the party
that controls the White House alters whether conservatives or liberals favor an activist
foreign policy.
In one of the clearest examples, Lupia (1994) provides evidence of cue-taking in the
1988 California car insurance elections. Voters who have little knowledge of the substance
of the ballot initiatives, yet know where various interest groups stood on the measures,
mirrored the behavior of high-knowledge voters. Individuals who lack both substantive
and endorsement awareness, on the other hand, deviate from the voting patterns of the
42
more knowledgeable groups.
22
Small cues from advocacy organizations help citizens cut
through the clutter of political issues.
The Cue-Taking Hypothesis, however, does not merely assert that cue-taking occurs;
it predicts that personal experience moderates the effects of cue-taking. Experiments
from psychology provide evidence. In Wood’s (1982) experiment on the environment,
the participants are asked to list their personal connections with an environmental issue,
e.g., membership in the Sierra Club or Audubon Society. All subjects are then exposed
to a fabricated recording of an interview with a graduate student in which the graduate
student presents several arguments against preservation. Wood finds that subjects who
report more personal experience with the environment are less susceptible to the graduate
student’s persuasion even when controlling for the subjects’ initial views.
One of the experiments used to differentiate between “central route” and “peripheral”
processing speaks directly to the Cue-Taking Hypothesis. Undergraduate participants are
presented with a set of arguments for a change in university policy. While some students
are under the impression that the policy would affect their school, others are told that the
policy is under consideration at a nearby institution. The arguments (all against the policy)
vary by logical soundness and overall appeal. Students who are personally involved in the
issue (i.e., those under the “own school” condition) discriminate between the compelling
and non-compelling arguments, while those not involved are persuaded by the speaker
regardless of the argument’s quality. Perhaps both self-interest and issue experience play a
role in inducing students to think critically about the arguments, but cue-taking is clearly
at work in the “other school” condition.
22
In this case, a cue from an untrusted source (e.g., the insurance industry) appears to provide information
to individuals about what not to believe. However, since two streams of information exist (e.g., Ralph Nader
countering the insurance industry), it is unclear whether the voters are using positive or negative cues.
43
The Cue-Taking Hypothesis dovetails well with the Lupia and McCubbins (1998) model
of persuasion, which predicts that less persuasion (i.e., cue-taking) will occur when “per-
ceived speaker knowledge” is low. If personal experience increases a voter’s knowledge (or
self-perceived knowledge) of an issue—a reasonable assumption—then the relative knowl-
edge of the speaker decreases and the Cue-Taking Hypothesis is supported by the experi-
ments that Lupia and McCubbins (1998) conduct in the laboratory.
Shifting to the Candidate Evaluation Hypothesis, Mutz (1992) finds an interesting
interplay between personal experience with the economy and political evaluations. She
traces the effects of experience with unemployment and knowledge gained from the mass
media through the perception of the economy and incumbent politicians (governor and
president). Consistent with the Personal Experience Model, being unemployed has an
impact on state-level judgments. At the national level, the mass media has much more
influence in shaping people’s opinion about the national economy. Specifically, Mutz writes,
“These overall patterns suggest that personal experience may indeed have an indirect
influence on political evaluations at the state level by means of their influence on personal
and ultimately social concerns. At the national level, however, the path that translates
these concerns to political significance is incomplete.” This latter finding might appear to
refute the Candidate Evaluation Hypothesis, but being unemployed is a stronger signal of
the state economy rather than the national economy (since the reason for the unemployment
might be localized). The finding that individuals are properly incorporating information
from multiple sources by discounting their personal experiences when making national
judgments is encouraging for any hypotheses grounded in learning models.
Examining political judgments from a similar angle, Fiorina (1981), divides retrospec-
tive evaluations into two categories: “simple,” which are the result of direct experience and
“mediated,” which are filtered through secondary sources such as the media. Simple ret-
44
rospective evaluations affect voters’ political evaluations and thus shade how they receive
mediated evaluations. Fiorina reaches analogous conclusions to the Personal Experience
Model, but from the perspective of voters’ experience with policy outcomes rather than
issue positions.
While the direct evidence that personal experience moderates cue-taking is established
in the psychology literature, minimal research exists that linking personal experience to
political evaluations. Gilens and Murakawa (2002) characterize the state of research by
noting, “We are not aware of any research that explicitly compares citizens’ decision-
making with regard to high-involvement versus low-involvement issues.”
23
The Personal
Experience Model and the empirical examples presented in later chapters aim to fill that
void.
2.2.4 Easy Issues and Retrospection Affect Political Evaluations
Existing research also supports the Broadcast Corollary. Kelleher and Wolak (2006) com-
bine presidential approval ratings and content analysis of media stories to examine the
priming effects of easy and hard issues. Since the dependent variable (presidential ap-
proval) is dichotomous, the Broadcast Corollary would predict that changes attributed to
priming would be observable only for familiar (or “easy”) issues. On complex (or “hard”) is-
sues, individuals who approve of the President would take the President’s side and continue
to approve of him; individuals who do not approve of the President would act analogously
(assuming two issue streams). Thus, even if considerations of the complex issue are more
likely to be on the top of a person’s head (Zaller, 1992), there would be no change in the
correlation between issue opinion and presidential approval before and after the increase
in media stories. On the other hand, individuals’ opinions of familiar issues are not as
23
Recent work (Hillygus and Shields, 1991) that compliments this dissertation makes this statement less
true in 2009.
45
susceptible to cue-taking; thus, issue priming by the media might alter their perceptions
of the president.
Kelleher and Wolak’s findings are consistent with this line of reasoning. When the
media reports on easy issues (e.g., the economy, the President’s character), Kelleher and
Wolak observe the priming effects. On the other hand, their data do not exhibit the effects
of priming when the media emphasizes hard issues (e.g., foreign policy).
Lenz (2006) finds similar results when he examines over a dozen cases of suspected
priming and determines that priming occurred in only a handful of cases. Of the 13
examples of supposed priming, Lenz argues that just four are actual examples of priming:
two character issues, the economy in 1992, and defense spending after 9/11. Scholars
consistently label character issues and the economy as easy issues. The priming of defense
spending is trickier: perhaps the lack of a competing stream of information prevented
Democrats from polarizing away from the President.
When elite signals on a topic are one-sided, a similar phenomenon to easy issues occurs.
Often, one-sided news occurs because of the resolution of an issue, such as when there is
consensus that a policy succeeded or failed (Canes-Wrone et al., 2001). This consensus
fills the tops of all voters’ heads with one-sided considerations; if these considerations are
prevalent (such as constant reporting on the war or economy), the resolved issue mimics an
easy issue. The Broadcast Corollary predicts that these resolved issues have large effects
on vote choice.
Empirical research confirms this prediction. Mueller (1970) finds that retrospective
voting on these oft-reported issues (in the spirit of Fiorina, 1981) have a consistent effect
on presidential voting, which is a highly partisan contest. Recent research (Achen and
Bartels, 2005) demonstrates that this effect can be so strong that the outcomes need not
be logically connected to actual policy. As the Broadcast Corollary indicates, candidates
46
who can take advantage of issues that resolve in their favor are probably better off focusing
on those issues rather than microtargeting certain experienced voters.
The Personal Experience Model, by detailing a fine-grained relationship between voters
and issues, predicts the conclusions reached by less granular studies that examine decisive
issues. Easy issues are issues for which the voter does not need to rely on elites to form
an opinion (which correspondents to the first criterion of personal experience) and are
often major issues such as race relations (which keeps considerations at the top of the
head, meeting the second criterion). While not all easy issues come about by the result of
personal experience, a voter with experience on an issue would consider that issue to be
easy. Hence, the results of the issue-centric political science research are consistent with
the Personal Experience Model.
These discussions demonstrate that existing psychology and political science literature
provides solid support for the Stability Postulate, the Cue-Taking Hypothesis, and the
Broadcast Corollary. There is also circumstantial evidence for the Candidate Evaluation
Hypothesis, although the political science literature is lacking on this subject.
2.3 Alternative Voter-Issue Linkages in the Literature
While much of the political science literature focuses on the characteristics of a voter (e.g.,
attentiveness) or an issue (e.g., complexity) separately, strains of the literature emphasize
linkages between voters and issues. Specifically, research on self-interest and issue publics
theorize why voters might rely on certain issues more than others in their political decisions.
However, studies demonstrate that self-interest does not affect vote choice in a consistent
manner.
The Personal Experience Model generally captures the evidence of the self-interest lit-
erature. With regard to issue publics, the Personal Experience Model provides a logical
47
foundation for the conclusions reached by the issue publics literature; this logic also broad-
ens the scope of the theory to include both issues and character traits.
One perspective on issue-voter interaction is that voters base their political decisions on
self-interest. Chong et al. (2001) demonstrate that self-interest in a policy can be primed or
unprimed depending on the political communication delivered. Individuals with a vested
interest in a certain policy change (e.g., domestic partner health benefits) display a higher
propensity to favor the change after their self-interest in the change is highlighted. Al-
though the authors do not measure voters’ evaluations of candidates who take a position
on the policy changes tested in the experiment, one logical conclusion might be that politi-
cians should communicate to voters on issues in which the voters have a self-interest and
that these communications should emphasize that connection. On the other hand, Chong
et al. also find that self-interested voters can be swayed against their direct economic
interests with sociotropic arguments. In a climate with conflicting message streams (one
that primes self-interest and one that primes sociotropic considerations), it is unclear how
much self-interest would affect political judgments.
Consistent with Chong et al.’s finding that political communication can reduce the
salience of self-interested consideration, Sears and Funk (1990) find that self-interest is
rarely a factor in vote decision; instead, symbolic politics often drives individuals’ policy
stances. The Personal Experience Model predicts the influence of symbolic politics: voters
cue-take from their respective group’s leaders, polarize, and do not alter their political
evaluations. In some cases, however, Sears and Funk find that self-interest does play a
role in politics—specifically when the benefits are clear or the stakes are high (i.e., easy
issues). By focusing on which issues are easy for which voters, the Personal Experience
Model extends the findings of Sears and Funk and yields more nuanced implications.
48
In a second thread of political science research, Krosnick (1990) develops the idea of
issue publics first presented by Converse (1964). Although the vast majority of voters do
not closely follow the political debate on most issues, many voters find one or two issues to
be important. Krosnick emphasizes self-interest, group identification, and personal values
as the bases for how voters choose issue publics. Voters in an issue public have more stable
opinions about the issues and base their political decisions on these issues; Gershkoff (2006)
confirms these findings in an extensive analysis of issue publics. Malhotra and Kuo (2009)
demonstrate that, in the aftermath of Hurricane Katrina, increased personal importance
of the hurricane devastation moderates the effects of partisan polarization and cue-taking.
The Personal Experience Model shifts and extends the idea of issue publics in two ways.
First, the level of personal experience explains why a person might enter an issue public.
For instance, in the preface of Gershkoff’s (2006) dissertation, she relates a story about
a Russian hairdresser who was in the immigration issue public. Left unsaid is the high
likelihood that the hairdresser was in the immigration issue public because of her personal
experience emigrating. Second, the Personal Experience Model does not rely on significant
differences in issue salience to affect political preferences.
In the issue publics framework, certain issues are crucial for political judgments because
of the conscious weight that a voter places on them. These issues are most likely to be
reported as “most important” on a political survey. In contrast, in the Personal Experience
Model, the mechanism by which some issues are more predictive of political preference
than others—the desire to avoid cognitive dissonance—is the result of life-happenstance.
Voters often have these experiential issues foisted upon them (e.g., receiving Social Security
checks, emigrating from Russia) rather than choosing to be a member of an issue public.
The Personal Experience model focuses on issues—big or small, inherently political or
not—that form the bases for voters’ “ground truth” for judging politicians.
49
Finally, Fazio and Zanna (1978) propose that confidence is an intervening factor between
the experience-issue stability link. Individuals with direct experience with a topic are more
likely to have higher confidence in their attitudes as well as to match their behavior to their
attitudes. More pertinently, when Fazio and Zanna exogenously increase confidence in an
attitude (independent of how the attitude was formed), the correlation between attitude
and subsequent behavior increases as well. Thus, voters who have more confidence in
all their opinions (regardless of whether this confidence is justified) may exhibit less cue-
taking than would be otherwise expected. If this research is correct, while the Personal
Experience Model would still hold as presented (because direct experience would lead
to confidence, which would moderate cue-taking), part of the picture would be unmodeled
(i.e., more confident people are less prone to cue-taking). This causal mechanism is explored
in Chapter 4.
In sum, the Personal Experience Model hypothesizes that personal experience plays
a crucial role in political evaluation. The important features of personal experience is
that (1) the individual plays an active role thereby circumventing partisan filters and (2)
the considerations generated by the individual’s experience are accessible. This type of
personal experience allows a voter to ground truth candidates’ statements and explains
some observed heterogeneous movements in party and candidate preferences.
The theory generates two main hypotheses. First, it predicts that personal experience
moderates voters’ adoption of political elites’ issue opinions (i.e., cue-taking) based on
an individual’s political predispositions. Second, a candidate stakes out a position on an
issue, voters with personal experience involving that issue will alter their evaluations of the
candidates more than voters who lack such experience will. These candidate preference
changes may run counter to existing partisan biases, which makes them important to
campaigns which aim to persuade voters.
50
Chapter 3
Experienced Evaluations and
Self-Interested Opinions:
Observational Evidence from the
2000 Presidential Campaign
3.1 The Personal Experience Model and the 2000 Presiden-
tial Campaign
3.1.1 The Campaign about Nothing
From the perspective of political practitioners, the 2000 presidential campaign occurred
at just the wrong moment in history. While it would be another two to six years before
microtargeting became prevalent in national politics, the 2000 electorate was highly mi-
crotargetable because no single issue dominated the campaign discourse. Gore and Bush
51
each focused their attention on several issues, sometimes talking past each other. While
the campaigns delivered competing talking points about Social Security privatization, Gore
also made a push for the Patients’ Bill of Rights. These two campaign issues have a com-
mon feature: they exhibit evidence of issue-driven vote preference among the segment of
the population with experience on the issue. This main finding is based on the Annenberg
2000 dataset, which includes panel data recorded before and after the parties’ nominating
conventions.
Specific segments of the electorate—individuals with personal experience on a political
issue—were susceptible to targeting by the campaigns. When Bush or Gore emphasized an
issue in his convention speech, experienced voters learned about the candidate’s stance on
that issue faster than the rest of the public and judged the candidate on that issue. This
evidence provides support for the Candidate Evaluation Hypothesis presented in Chapter
2.
Analyses indicate another phenomenon at play: self-interest. Voters with personal
experience with an issue may also have a stake in the policy outcome. Candidates may
prime these concerns and shift individuals’ issue opinions—an effect that contradicts the
predicted reduced opinion shift of the Cue-Taking Hypothesis. The data find evidence of
both phenomena occurring in the 2000 election.
Voters in the 2000 electorate could not agree on a single most important issue. In one
poll, the modal response (representing 20% of responses) to the question, “What is the
most important problem facing the country?” was education—an area in which the federal
government has little say.
1
This result contrasts with the 2004 campaign, when the war on
terrorism and the war in Iraq combined to become the single most important issue for a
1
ABC News. November 1, 2000. http://abcnews.go.com/images/pdf/836a16Tracking16.pdf
52
third of the electorate, with the economy not far behind. In the 2008 campaign, after the
mid-September financial crisis, the economy mattered most to over a third of the public.
In this varied issue landscape, it is not surprising that Bush’s August 3, 2000, conven-
tion speech was largely devoid of issues while Gore focused on several minor issues in his
August 17 convention speech. Bush spent the bulk of his speech defining who he is. For in-
stance, he explained the term “compassionate conservative” with multiple vignettes. This
tactic appeared to work, as Bush opened up a double-digit point lead after the convention.
However, the Republican nominee’s speech did touch upon four concrete policy proposals:
education, Social Security privatization, military readiness, and taxes.
Bush made an especially forceful push for Social Security privatization, saying:
Social Security has been called the “third rail of American politics,” the one
you’re not supposed to touch because it shocks you. But if you don’t touch
it, you can’t fix it. And I intend to fix it. To seniors in this country: You
earned your benefits, you made your plans, and President George W. Bush will
keep the promise of Social Security. No changes, no reductions, no way. Our
opponents will say otherwise. This is their last, parting ploy, and don’t believe
a word of it. Now is the time for Republicans and Democrats to end the politics
of fear and save Social Security, together. For younger workers, we will give
you the option—your choice—to put a part of your payroll taxes into sound,
responsible investments.
Gore, in direct contrast to Bush, touched on dozens of issues in his nomination accep-
tance speech, some rather minor (e.g, tobacco advertising to children, Internet privacy).
The convention as a whole was a success, giving Gore a lead in the polls that he would
hold until the first debate in October. Perhaps the lasting legacy of the convention was
Gore’s coining of the oft-parodied term “lockbox”:
53
At a time when most Americans will live to know even their great-grandchildren,
we will save and strengthen Social Security and Medicare—not only for this
generation, but for generations to come... .
And to me, family values means honoring our fathers and mothers, teaching
our children well, caring for the sick, respecting one another—giving people the
power to achieve what they want for their families.
Putting both Social Security and Medicare in an iron-clad lockbox where the
politicians can’t touch them—to me, that kind of common sense is a family
value.
Even though the media focused on the lockbox and hence Social Security, Gore actually
devoted more of his speech to vilifying health maintenance organizations (HMOs) and to
rallying support for a Patients’ Bill of Rights:
And I will never forget a little boy named Ian Malone—who suffered from a
medical mistake during childbirth, and needs full-time nursing care for sev-
eral years. I met him and his parents in Seattle, near their home in Everett,
Washington. Their HMO had told the Malones it would no longer pay for the
nurse they needed, and then told them they should consider giving Ian up for
adoption... .
Dylan and Christine Malone are here with us tonight. Ian’s here, too. And I say
to them, and to all the families of America: I will fight for a real, enforceable
Patients’ Bill of Rights.
It’s just wrong to have life and death medical decisions made by bean-counters
at HMOs who don’t have a license to practice medicine, and don’t have a right
to play God. It’s time to take the medical decisions away from the HMOs and
54
insurance companies and give them back to the doctors, the nurses, and the
health care professionals.
In the week after the convention, Gore produced two television commercials about the
Patients’ Bill of Rights, one of which focused on the case of Ian Malone.
3.1.2 Issue-Driven Vote Choice
Political science has its own version of the chicken and egg question: Which comes first,
the vote preference or the issue opinion? Researchers often attempt to find evidence of
“priming,” issue-driven effects of a campaign (or other political elites) emphasizing an
issue and thus altering voters’ political judgments through the issue’s increased salience
(e.g., Iyengar and Kinder, 1989). As discussed in depth by Lenz (2006), reverse causation
is a troubling problem when attempting to discern the effects of issue opinion on vote
preference. As with cue-taking (discussed in Chapter 2), voters often adopt the issue
opinions of their favored candidate; therefore, analyzing the correlations of simultaneous
issue opinion and vote choice is problematic.
A more useful approach involves correlating previous issue opinion with change in vote
using panel data. If the correlation increases over time, then issue opinion must be driving
vote choice and not vice versa. A change in vote choice in July, for example, cannot alter a
voter’s issue opinion in June. One caveat is that if the issue in question already dominates
the June vote choice and continues to dominate in July, the analysis will not detect an
effect even though the issue opinion is affecting vote choice. For analytic purposes, the
best issues to examine are those that increase in salience in between waves of a panel data
set. This empirical necessity dovetails with the Personal Experience Model, as the model
produces cleaner predictions when the issue starts with few elite-driven considerations (see
discussion in Section 2.1.6).
55
The issues of Social Security and the Patients’ Bill of Rights were selected for analysis
for two reasons. First, they meet the requirement of being lesser-known issues that were
emphasized during the 2000 campaign. Other issues, such as the licensing of handguns, also
fit that description. However, data on personal experience for these other issues are lacking.
For instance, the Annenberg survey did not ask respondents about gun ownership. The
three issues highlighted here also meet the requirement that data on personal experience
be available.
Both issues fit the assumptions of the Personal Experience Model and can be used
to test the model’s predictions. The model assumes that there is little to no political
discussion of the issue in period 1 and that subsequently a candidate delivers a signal
about the issue. For each issue, the analyses verify that the public entered the convention
phase of the campaign with little knowledge of where candidates stood on the issue and
that the electorate learned about these stances over the post-convention weeks. The text
of the convention speeches provides evidence that Bush sent signals about Social Security
and that Gore emphasized his positions on Social Security and the Patients’ Bill of Rights.
The Personal Experience Model also requires an observable measure of voters’ personal
experience with an issue. For Social Security, it is assumed that voters over the age of
65 have experience with Social Security since they receive a check from the Social Secu-
rity Administration each month. They have experience with how reliable the government
agency is and how much the money helps them pay for expenses. This measure is of de-
batable quality, however, because these voters do not have experience with Social Security
privatization.
For the Patients’ Bill of Rights, it is assumed that individuals with health care through
an HMO have experience with this issue. These voters know how helpful additional rights
(such as the ability to sue) would be to them in their dealings with the HMO. While the
56
Patients’ Bill of Rights would apply to all citizens (not just the HMO-insured), Gore closely
tied the Bill of Rights to HMOs, believing it to be most relevant for the HMOs’ customers.
With these assumptions validated, the Personal Experience Model makes two testable
predictions. First, experienced voters do not cue-take from politicians as much as non-
experienced voters. This prediction is difficult to verify with the data at hand because
of the self-interest connection between issue experience and issue position. For example,
retirees might believe that their Social Security checks are at risk under a privatization
plan. HMO customers would benefit from additional rights.
Thus, when a candidate emphasizes an issue, an experienced voter with a self-interest
on one side of the issue may be pulled in two different directions. On one hand, the voter’s
issue considerations from experience may help outweigh the candidate’s argument. On the
other hand, the candidate’s point may be of particular importance to the voter (Chong
et al., 2001).
The best example of this second phenomenon is Social Security privatization. Seniors
have no more experience with how privatization would be implemented than other voters
have. Thus, when Gore claims that privatization would costs seniors their benefits, these
older voters have no experiential considerations to counter this argument. Naturally, some
of them shift their issue position to be against privatization. This pattern is evidenced in
the data.
The second prediction of the model is that experienced voters judge politicians on the
issue for which they have experience (when a politician sends a signal on that issue). The
empirical test of this hypothesis is whether the correlation between first period (in this
case, pre-convention) issue opinion and change in vote preference is greater for those with
personal experience than for those without. This effect is labeled “differential evaluation”—
candidate evaluations among a certain segment of the population. Differential evaluation is
57
a result of experienced voters learning the candidates’ positions on an issue and, possibly,
the issue becoming more salient, or primed. This latter effect is not required, and the
analysis finds only weak support for a priming effect separate from learning.
Crucially, the self-interest effect counters the effects of observable differential evalu-
ation. A senior who shifts her opinion from pro- to anti-privatization and from Bush to
Gore reduces the correlation between prior issue opinion (pro-privatization, the Republican
stance) and change in vote (toward Gore, the Democrat). The key, verifiable assumption
is that experienced voters who shift their issue opinions in the direction of the Democratic
position do not, at the same time, switch their vote choice toward Bush (or vice versa). If
that consistency holds, evidence of differential evaluation among experienced, potentially
self-interested voters is likely to be understated.
As with many observational studies, a word of caution is necessary. The data are used to
support hypotheses that make causal claims, yet without random assignment of treatment,
causal statements are difficult to justify. While findings show that voters experienced with
issues differed in their actions from non-experienced voters, these differences in behavior
cannot be directly tied to personal experience. However, alternative hypotheses are tested
in efforts to rule out spurious relationships and provide more confidence in the conclusions
drawn.
3.1.3 Data from the 2000 Annenberg Survey
To isolate these effects, the Republican and Democratic National Convention panel data
from the Annenberg 2000 presidential survey are analyzed. The Annenberg Center initially
interviewed 1,197 respondents during July as a part of its normal cross-sectional study. The
Republican convention was held from July 31 to August 3 and the respondents were re-
interviewed from August 4 through 13. For the Democratic National Convention, 1,230
58
respondents were interviewed first from August 4 to 13 and then after the Democratic
National Convention, which ran from August 14 to 17.
Ideally, those labeled as having experience with Social Security would be respondents
who receive a Social Security check, but that question is not directly asked on the survey.
Instead, respondents 65 years old or older are assumed to receive Social Security checks.
This group comprises 16.6% of the cross-sectional national study and about the same pro-
portion in the panel studies. In the pre-convention period (May through July), seniors are
less likely to support privatization (31% support) than younger voters are (62% support).
The experience measure for the Patients’ Bill of Rights issue is whether the respondent
self-reports having her “main insurance through an HMO.” This group encompassed over
43% of survey respondents in the national cross section. Prior to the nominating conven-
tions, HMO customers were slightly more likely to favor “giving patients the right to sue
their health maintenance organization or HMO” than non-HMO customers by 78% to 75%.
3.2 Results and Analysis
3.2.1 Learning Over the Course of the Campaign
Data on Social Security and the Patients’ Bill of Rights indicate that voters learned about
these issues over the course of the campaign. Learning was especially prevalent among
experienced voters, a crucial point since the Personal Experience Model and other research
(Lenz, 2006) present learning as a necessary condition of issue-driven vote preference shifts.
If learning is more prevalent among experienced voters, then the implications for campaigns
of the Candidate Evaluation Hypothesis are even more stark than in a situation in which
all voters learn the candidates’ positions.
59
Voters initially are largely ignorant of Gore’s opposition to Social Security privatization.
In the first month that the survey asked about Gore’s position (May 19 to June 15, n=605),
only 41% of respondents answer “no” to the question “Al Gore–do you think he favors
or opposes allowing workers to invest some of their Social Security contributions in the
stock market?” By contrast, a majority (51%) of respondents correctly report that Bush
favored personal investment for the retirement program. Seniors, during this early stage
of the campaign, are not statistically significantly more knowledgeable than the rest of the
population, though slightly more (56%) know that Bush opposed privatization (n=95).
Voters learn about the candidates’ positions during the campaign, through nominating
conventions, debates, and advertisements. In the last three weeks before the election, 61%
of respondents know that Gore opposed privatization and 67% report that Bush favored
the policy change (n=2,753). Seniors learn about the candidates’ platforms to a larger
extent, with 72% and 77% correctly reporting Gore’s and Bush’s positions (n=460). The
knowledge differences on this issue between seniors and younger voters are statistically
significant.
Figure 3.1 illustrates learning over the course of the campaign. Learning is slow and
steady through the summer, with an acceleration during the nominating conventions. Se-
niors learn faster during the convention period than the rest of the population. Then, after
the conventions, voter knowledge stagnates, only to rise again—very quickly—starting with
the first debate in the beginning of October.
As with Social Security, the campaign begins with voters generally ignorant of the
candidates’ stances on the Patients’ Bill of Rights. The Annenberg survey asks voters
whether they thought Bush (and, in a separate question, Gore) supported a patient’s
“right to sue” her HMO. Bush’s position on this topic is genuinely ambiguous: he vetoed a
Patients’ Bill of Rights in Texas in 1995 but let a subsequent bill become law without his
60
Date
P
e
r
c
e
n
t
a
g
e

o
f

C
o
r
r
e
c
t

R
e
s
p
o
n
d
e
n
t
s
June 1 August 1 October 1
3
0
%
4
0
%
5
0
%
6
0
%
7
0
%
8
0
% Seniors
Only
All
Respondents
R
e
p
.

C
o
n
v
.
D
e
m
.

C
o
n
v
.
Low
Daily
Sample
Sizes
Bush's Stance
Gore's Stance
Both Stances
Figure 3.1: Loess smoothed curves of voters’ knowledge about Bush’s and Gore’s positions
on Social Security. While both young and old voters learn about their issue positions during
the course of the campaign, seniors gain knowledge faster. Daily sample sizes average 23
prior to July 4 (inclusive) and 149 after July 4.
signature when a veto-proof majority of the Texas legislature passed it. However, in the
third debate, Bush claimed credit for this law and touted its provision that gives patients
the right to “sue an HMO for denying you proper coverage.”
2
Gore’s position on the Patients’ Bill of Rights, however, is unambiguous: he supported
the right of patients to sue their HMOs. Despite this clear position, in the first month of
the Annenberg survey, only 42% of respondents know Gore’s position (with 38% claiming
not to know). By the final two weeks of the campaign, this figure increases to 48% (with
2
As President, Bush’s Department of Justice successfully argued in Aetna Health, Inc. v. Davila, 542
U.S. 200 (2004), that federal law superseded the Texas Patients’ Bill of Rights, thus disallowing patient
lawsuits.
61
31% “don’t know”). Learning among voters who used an HMO for health insurance is
slightly faster, increasing from 42% to 51% over the same time period. The difference in
knowledge at the end of the campaign between voters who are experienced with HMOs
and those who are not is statistically significant.
3.2.2 Cue-Taking and Self-Interest
The issues discussed in this chapter—Social Security and the Patients’ Bill of Rights—
are issues in which experienced voters are have vested personal interests. The Personal
Experience Model and theories of self-interest yield competing hypotheses, with the 2000
Annenberg data helping to adjudicate this dispute. The results, however, are mixed: both
experience and self-interest affect changes in voters’ opinions. In the three cases tested, one
case (Patients’ Bill of Rights) supports the Cue-Taking Hypothesis, another case (Social
Security during the Democratic convention) supports the theory of self-interest, and a third
case (Social Security during the Republican Convention) yields mixed evidence. Most likely,
both phenomena occur in all cases and the resulting effect depend on the strength of the
self-interest and the personal experience.
Before analyzing whether personal experience or self-interest hold sway over voters,
another hypothesis—issue-opinion cognitive dissonance—must be considered. If, as the
issue-opinion cognitive dissonance hypothesis predicts, voters accept only congruent sig-
nals on issues, then no voter would ever alter her issue position. If, on the other hand, the
Personal Experience Model’s assumption is accurate and voters accept signals only from
favored politicians, then voters should adopt the positions of these politicians as issues be-
come politicized (see discussion in Section 2.1.1). Theories of self-interest suggest a similar
voter shift, though with different causal connections: in that model, voters are primed on
self interest and change their issue opinion and candidate evaluations accordingly. Hence,
62
both the Cue-Taking Hypothesis and self-interest theories predict increased correlation
between issue opinion and candidate preference as an issue becomes politicized.
The data from the 2000 campaign support these latter two hypotheses. For each of
the four issue-convention pairs, the correlation between pre-convention vote preference
(trichotomous) and issue position increases over the course of the convention. The panel
nature of the data ensures that reverse causation does not affect these results (i.e., post-
convention vote preference is ignored). Only the Social Security correlation boost during
the Republican convention meets standard levels of statistical significance (p = 0.10, two-
tailed). However, as the left panel in Figure 3.2 shows, all four shifts in correlation are in
the expected direction, and jointly the analysis is significant (p = 0.09).
3
The right panel
of Figure 3.2 displays the correlations between the more granular 7-point party affiliation
and issue opinion, revealing a similar pattern.
Evidence for the Cue-Taking Hypothesis exists but is weak. Of the three issue-convention
pairs, two have the expected sign of the Cue-Taking Hypothesis, although only one, Pa-
tients’ Bill of Rights, is statistically significant. The other, Social Security during the
Democratic convention, is nearly exactly zero. For the Republican convention, seniors are
less likely than other voters to have stable opinions. This phenomenon is a result of seniors
shift, in a self-interested manner, from a pro- to an anti-privatization stance.
Support for the Cue-Taking Hypothesis is measured with a logistic regression of post-
convention issue opinion on pre-convention issue opinion with an experience interaction
term. Controls for convention watching, income, political interest, and party (including
their interaction terms) are included. Figure 3.3 displays the quantity of interest: the
difference in probability of issue opinion change between experienced and non-experienced
voters. Negative values are indicative of the cue-taking moderation: experienced voters
3
The calculation of a joint p-value assumes that the tests are independent. When comparing analyses
of separate groups of experienced voters on separate issues, this assumptions is probably valid.
63
q
q
q
q
q
q
Issue−Convention Panel Pairing
P
r
e

C
o
n
v
.

V
o
t
e

I
s
s
u
e

O
p
i
n
i
o
n

C
o
r
r
e
l
a
t
i
o
n
Social Security
GOP Conv.
Social Security
Dem Conv.
Patients' B.O.R.
Dem Conv.
0
.
0
0
.
1
0
.
2
0
.
3
0
.
4
q
q
q
q
q
q
Issue−Convention Panel Pairing
P
r
e

C
o
n
v
.

P
a
r
t
y

I
D

I
s
s
u
e

O
p
i
n
i
o
n

C
o
r
r
e
l
a
t
i
o
n
Social Security
GOP Conv.
Social Security
Dem Conv.
Patients' B.O.R.
Dem Conv.
0
.
0
0
.
1
0
.
2
0
.
3
0
.
4
Figure 3.2: Change in predisposition-issue correlations before and after the conventions.
Predispositions are measured by trichotomous vote choice (left panel) or 7-point party
affiliation (right panel). Issue opinion measures have three and two possible values for
Social Security and Patients’ Bill of Rights, respectively. Each correlation includes about
1,200 cases in each correlation.
shifting their opinion less than non-experienced voters. Regression coefficients are presented
in Table B.1 of Appendix B.
The positive result for the Cue-Taking Hypothesis in the case of the Patients’ Bill of
Rights holds even when controlling for alternative explanations. Potentially, voters who
watched the convention might have been more likely to shift their opinions, but such an
interaction term was in the opposite direction and was near zero. Also, more informed
voters may have had more stable opinions; the evidence supports this idea but does not
diminish the importance or significance of the experience. The positive coefficient is not
solely an artifact of there being fewer non-HMO individuals who are against a Patients’ Bill
of Rights; a lower proportion of experienced individuals defected from that position than
64
did non-experienced individuals. The result also disappears when experience with Social
Security is entered into the regression in place of support for a Patients’ Bill of Rights,
thus eliminating the possibility that HMO-experienced individuals happen to have more
stable opinions across the entire political spectrum.
q
q
q
Issue−Convention Panel Pairing
P
e
r
c
e
n
t
a
g
e

P
o
i
n
t

D
i
f
f
e
r
e
n
c
e

i
n

S
h
i
f
t

P
r
o
b
a
b
i
l
i
t
y
:
E
x
p
e
r
i
e
n
c
e

V
o
t
e
r
s

m
i
n
u
s

I
n
e
x
p
e
r
i
e
n
c
e
d

V
o
t
e
r
s
Social Security
GOP Conv.
Social Security
Dem Conv.
Patients' B.O.R.
Dem Conv.

0
.
1
0
.
0
0
.
1
0
.
2
0
.
3
Figure 3.3: Increased or decreased probability of issue opinion change for voters experienced
with the issue. Issue opinion is dichotomized. Point estimates (circles), one bootstrapped
standard error (bold lines), and 95%-confidence interval (non-bold lines) are displayed.
Regression details are presented in Table B.1.
Social Security recipients, on the other hand, exhibit less cue-taking moderation or
none at all. This finding is especially apparent in the Republican convention panel data,
in which a significant proportion of seniors were alienated by Bush’s privatization message.
Most likely, these seniors believed that their Social Security checks were endangered by
Bush’s plan, even though Bush explicitly protected seniors in his policy proposal.
65
For the Democratic convention data, shades of self-interest are apparent. Over a quarter
of seniors (27%) who begin the convention period in favor of privatization switch their
opinion by the end of the convention. Less than a tenth of younger voters do the same.
The experience interaction coefficient (Table B.1) is positive for this issue-convention pair
because among those initially anti-privatization, more seniors stick with that position.
The anti-privatization movement among seniors is consistent with Chong et al.’s (2001)
experiments on priming self-interest. The Annenberg survey asks respondents whether they
agree with the statement, “Allowing younger people to invest some of their Social Security
contributions in the stock market will reduce Social Security benefits for older people.”
Even at the end of the campaign (which is the only period during which the question is put
to respondents), after so much information about Social Security had been disseminated,
40% of voters agreed that seniors’ benefits would be reduced under privatization.
4
As discussed in Section 3.1.2, the finding that experienced voters are also influenced
by self-interest biases the results in Section 3.2.3 downward. While self-interest is the
likely cause for the issue opinion defection pattern of experienced voters, this bias would
exist regardless of the mechanism. The sufficient condition is that issue opinion shifts are
accompanied by like-minded voter preference shifts. The data clearly show that they are.
For example, Gore’s margin increases by 29 percentage points among seniors who shift
their issue opinions against privatization during the Democratic Convention.
Of all the personal experiences tested in this dissertation, receiving a Social Security
check is the weakest. Receiving a Social Security check provides evidence apart from
partisan predispositions that the system “works” (i.e., that they receive checks) but lends
little insight into how the system works. It is not surprising then that the evidence for
the Cue-Taking Hypothesis is mixed, at best. In sum, the results lean toward the opposite
4
The Gore campaign ran an advertisement claiming, perhaps disingenuously, that Bush’s plan “could
cut benefits for seniors.”
66
hypothesis preferred by the self-interest literature: seniors are inclined to shift their opinion
toward anti-privatization.
In contrast, being a member of an HMO does give an individual at least some insight
into the health insurance bureaucracy. For this issue, the evidence for the Cue-Taking
Hypothesis is strong. Accounting for the amount of time individuals watched the con-
vention, political knowledge, and pre-convention position disparities does not weaken the
relationship. In this case, experienced voters are less likely to adopt the position of elites.
3.2.3 Experienced Voters Judge Politicians on Issues
Consistent with the Candidate Evaluation Hypothesis, the Annenberg data demonstrate
that voters judge candidates on issues with which they have experience. Two of the three
issue-convention pairs’ results near statistical significance. When the negative effects of self-
interest are corrected for, all effects are in the expected direction. Analyses also examine
the role of priming apart from voter learning.
Support for the Candidate Evaluation Hypothesis is provided by variations on one or-
dinal regression. Change in vote from the pre-convention period to the post-convention
period is regressed on issue experience, pre-convention issue opinion, and the interaction
between these two terms. A negative coefficient for experience and a positive interac-
tion term indicate that experienced voters relied on their pre-convention issue position to
evaluate politicians more than did voters without experience.
Two of the three cases exhibit higher degrees of issue-driven candidate evaluation by
experienced voters. The Republican convention data indicate that the average undecided
voter who enters the convention against privatization has a 1% chance of shifting her vote
to Gore. A similarly situated senior has a 5% chance of altering her vote preference for
Gore. For the Democratic convention, an undecided voter who favors a patient’s right to
67
sue has a 14% chance of altering her vote in favor of Gore; a similarly situated voter who
belongs to an HMO health care plan has a 19% chance. Figure 3.4 displays the differences
in these probabilities (4 and 5 percentage points, respectively) added to the analogous
probabilities for voters who enter the conventions with opinions on the Republican side of
the spectrum.
q
q
q
Issue−Convention Panel Pairing
D
i
f
f
e
r
e
n
c
e

i
n

P
r
o
b
a
b
i
l
i
t
y

o
f

I
s
s
u
e

D
r
i
v
e
n

V
o
t
e

P
r
e
f
e
r
e
n
c
e

S
h
i
f
t
:
E
x
p
e
r
i
e
n
c
e

V
o
t
e
r
s

m
i
n
u
s

I
n
e
x
p
e
r
i
e
n
c
e
d

V
o
t
e
r
s
Social Security
GOP Conv.
Social Security
Dem Conv.
Patients' B.O.R.
Dem Conv.

0
.
1
0
.
0
0
.
1
0
.
2
0
.
3
q All voters
No Issue Shifters
No Issues Shifters or Learners
Figure 3.4: Increase in probability of an experienced individual shifting her vote because
of pre-convention issue opinion over the analogous probability for non-experienced vot-
ers. Point estimates (circles, squares, triangles), one bootstrapped standard error (bold
lines), and 95%-confidence interval (non-bold lines) are displayed. Regression details and
coefficients for the “no issue shifters” condition are presented in Table B.2.
As noted, the shift of experienced voters’ issue positions based on their self-interest
biases these results downward. To tease out issue-driven effects that are not a result of
self-interest, the regressions are restricted to voters who stick with their issue position
throughout the convention (Figure 3.4, squares). In all cases, voters with experience and
stable opinions are more likely to evaluate politicians on these issues than are their stable,
68
non-experienced counterparts. Voters who have an HMO and do not switch their position
on the Patients’ Bill of Rights are very likely to judge the candidates on that issue.
These analyses demonstrate that voters’ evaluations hinge on issues with which they
have experience. They do not, however, demonstrate an increase in the salience of the
issue, or priming. Lenz (2006) raises the point that learning, not priming, may be driving
vote choice. The Personal Experience Hypothesis takes a similar view: learning drives
both priming and changes in vote choice. To check for the possibility of priming in the
absence of learning, the regression is further restricted to those voters who answered the
same number (or fewer) of “candidate position” questions correctly.
The results from the Democratic convention indicate that no priming occurred apart
from learning (Figure 3.4, triangles). In each case, the difference between experienced
and non-experienced voters decreases, indicating that learning plays an important role in
issue-driven candidate evaluation.
The results include controls (including interaction terms) for income, convention watch-
ing, political interest and party affiliation. Experience does not cross issues; for instance,
the interaction of HMO customers and Social Security privatization during the Democratic
convention has a coefficient of nearly exactly zero.
Regression discontinuity (Thistlewaite and Campbell, 1960) analysis provides more ev-
idence that receiving a Social Security check generates experience with that issue and
tightens the relationship between that issue and Social Security. The analysis divides the
population into two group: those 65 years of age or older (who are assumed to receive
Social Security checks) and younger voters. If experience with Social Security is driving
the results then no issue-driven vote changes will be observed in the younger group—even
among those near retirement age in this group. To eliminate the possibility of self-interest
69
contaminating the results, the regressions are restricted to voters who did not change their
issue positions during the conventions.
The results (Figure 3.5) indicate that as a voter turns 65, the impact of pre-convention
opinion on privatization and change in vote during the convention does increase. For a
65-year-old voter, a change in her pre-convention opinion increases her chance of shifting
her vote toward Gore by 4 percentage points.
5
For an analogous 64-year-old voter, the
change in Gore probability is nearly exactly zero. This 3 to 4 percentage point difference
is consistent with the ordinal probit regression’s findings of Figure 3.4 (left-most square)
for Social Security during the Republican Convention.
6
Once experienced voter alter their candidate preferences, they cue-take from their pre-
ferred candidate. For example, members of HMO who have a stable opinion in favor of a
Patients’ Bill of Rights and shift their vote toward Gore during the Democratic convention
also become more against Social Security privatization. The rest of the respondent popu-
lation exhibits equal pro- and anti-privatization shifts.
7
This behavioral pattern supports
the dynamic nature of the Personal Experience Model (Section 2.1.5).
3.2.4 Discussion
Observational evidence from the 2000 presidential campaign indicates that experienced
voters act differently from voters who lack experience with a given political issue. Expe-
rienced voters learn about an issue faster and as they learn, they judge politicians on the
5
The regression coefficient that drives this calculation is not statistically significant (Table B.3).
6
To check for a spurious pattern in the data, the same regression discontinuity is applied to pre-convention
vote. The difference in predicted vote between the two coefficients is highly statistically insignificant. Low
n-sizes affects both of these analyses.
7
Among the experience subgroup, 22% become anti-privatization while 8% become pro-privatization
(n=49). Among other voters, the proportions are 11.3% and 11.8%, respectively (n=1,163). This difference
is statistically significant at conventional levels (p = 0.09, two-tailed). The difference holds when restricting
the latter group to pre-convention supporters of a Patients’ Bill of Rights.
70
Age (cutoff=65)
I
n
c
r
e
a
s
e
d

P
r
o
b
a
b
i
l
i
t
y

o
f

I
s
s
u
e

D
r
i
v
e
n

V
o
t
e

C
h
a
n
g
e
:

E
x
p
e
r
i
e
n
c
e
d

v
o
t
e
r
s

m
i
n
u
s

I
n
e
x
p
e
r
i
e
n
c
e
d

V
o
t
e
r
s
0
.
0
0
0
.
0
2
0
.
0
4
0
.
0
6
Bush
Gore
50 55 60 65 70 75 80 85
a a
a
p
p
a a a
p
p a a
p
a a p a
p a a
p a p
a p
a a p a
p
p
a
a a
a
p p p pp
a
aa p
p a
a
a
p p p
p a
a
p
a p p p p p p
a
p
a a
p p a
p
a
a
p p p a a a
a a a p a
a a
a
a a
a p
p p a
a
a p
Predicted effect
of experience:
3.3 percentage points
Figure 3.5: Regression discontinuity of voters around retirement age show issue-driven
effects for seniors only. Lines represent predicted impact of pre-convention Social Security
privatization opinion on the probability of a pro-Gore shift in vote preference. Points
labeled ”p” represent voters who are pro-privatization; points labeled ”a” are those anti-
privatization. Points on the top line are voters who alter their candidate preferences
toward Gore (with age jiggered); on the bottom line, toward Bush. Voters who do not
shift their voter preferences are not displayed (for clarity), but are included in the analysis.
Coefficients and regression specification reported in Table B.3.
issue. Issue-driven candidate evaluation independent of learning, such as priming, may
occur but the evidence is weaker.
Self-interest also plays a role in the issue-vote choice interaction. Candidates can em-
phasize a voter’s self-motivated preference for an issue position, making that voter more
likely to adopt that position. Since self-interested voters are often experienced voters, this
effect is in tension with the Cue-Taking Hypothesis. This chapter addresses this tension
by attempting to tease out the different effects of self-interest and experience among the
71
same group of voters. The next chapter, in contrast, identifies voters who have experience
with an issue but no self-interest in the outcome.
The analyses of this chapter, as well as all analyses of this dissertation, cannot state
truly causal claims. Causal inference relies on manipulation and randomization of units
(Fisher, 1926; Rubin, 1978; Holland, 1986). Tests of the Personal Experience Model suf-
fer from the problem that it is difficult to randomize personal experience. Instead, this
dissertation emphasizes conclusions of form: individuals with personal experience act in a
certain fashion. Results are consistent with the Personal Experience Model’s causal mecha-
nism and sufficient for campaign practitioners who care about outcomes rather than causal
chains.
Because of the interplay between experience and self-interest, evidence for the Cue-
Taking Hypothesis from the 2000 campaign is mixed. In one case, the Patients’ Bill of
Rights, voters experienced with HMOs moderate their cue-taking as the Personal Experi-
ence model would predict. For Social Security, however, seniors shift their issue opinions
toward their perceived self-interested position—against privatization. The differences in
the two levels of experience, along with being potentially unaware of the inner workings of
the Social Security trust fund, may explain some of this difference.
The evidence for the Candidate Evaluation Hypothesis is moderately strong. The panel
nature of the Annenberg data enables detection of the issue-driven candidate evaluation
(Lenz, 2006). Even though the effects of self-interest counteract these issue-driven vote
changes, experienced voters rely more on their issue position in two of the three cases
tested. When the effects of the Personal Experience Model are separated from self-interest
by examining a subset of the data, the findings become clearer.
For campaign practitioners, these analyses indicate that personal experience with an
issue can drive vote choice. But the 2000 campaign also provides a warning. Experienced
72
voters often have a stake in the policy outcome, and this self-interest may cause issue
opinion shifts that would not otherwise occur among experienced voters.
From the point of view of the Personal Experience Model, campaigns should target
experienced voters who agree with the candidate’s stance. The evidence demonstrates that
such targeting is a net positive. However, a more holistic view that took self-interest into
account would be more efficient at garnering voters.
73
Chapter 4
Survey Experiments of Princeton
Students and Adults Nationwide
A central aspect of many political campaigns is persuasion: convincing voters who would
otherwise cast a ballot for the opponent to support the campaign’s candidate. Several
micro-theories of American politics (Campbell et al., 1960; Zaller, 1992; Lupia and Mc-
Cubbins, 1998) detail the difficulty of changing an individual’s partisan predispositions
and imply that campaigns’ persuasive efforts face an uphill battle. The Personal Expe-
rience Model predicts that campaigns’ emphasis on issues with which voters have per-
sonal experience can alter the voters’ evaluations of the candidates counter to the voters’
predispositions. Two randomized survey experiments with hypothetical candidates and
real-world issues support the model’s predictions.
This analysis moves away from the partisan biases literature and focuses on why citi-
zens might change their vote choice during a campaign. The literature on partisan biases
has expanded from the fields of political science (Markus and Converse, 1979), political
psychology (Taber and Lodge, 2006), and neuroscience (Knutson et al., 2006). Dozens of
74
studies find that partisans react more positively to arguments from elites of their own party
(see Bartels, 2002, for a review). If Democratic-leaning voters listen only to Democratic
elites and Republican-leaning voters listen only to Republican elites, then the electorate po-
larizes. Indeed, this phenomenon occurred in the 2004 campaign over the issue of whether
Saddam Hussein was involved in the September 11 attacks (Parker-Stephen, 2007).
Partisan biases abound, but they are not helpful for campaigns seeking to persuade
voters to cast a ballot for a specific candidate. The goal of shifting the electorate (rather
than polarizing it) is more relevant for a campaign that is underperforming in public
opinion polls, but even leading candidates may want to increase their win percentage to
scare off challengers (Levitt and Wolfram, 1997). For a campaign to increase its vote share,
it must convince voters who would not have voted for its candidate to cast a ballot for
that candidate. Convincing these voters requires that the campaign increase the standing
of its candidate in the voters’ eyes relative to the other candidate(s).
1
Survey experiments
demonstrate that certain subgroups of voters (e.g., veterans) respond strongly to candidate
signals on issues with which they have personal experience (e.g., the Iraq war), and that
these interactions influence voters’ candidate evaluations even if they run counter to existing
partisan biases.
4.1 Theory and Hypotheses
Since at least as far back as Downs (1957), political scientists have modeled political
preferences as a function of ideology. This ideology can be thought of as a combination of
issues, with voters preferring those candidates whose issue opinions are closest to their own
opinions. Thus, when an individual receives information (i.e., a “signal”) from a candidate
1
These claims assume an electorate of fixed size, but in models that correlate uncertainty of choice with
abstention (Sanders, 2000), increasing a candidate’s standing among the electorate will increase turnout as
well as vote share.
75
that indicates that he holds a similar position to the individual, the individual should
improve her evaluation of the candidate. Analogously, her evaluation should deteriorate
when the respondent receives an incongruent cue.
Lemma 1: A congruent signal from a candidate will improve the respondent’s evalua-
tion of the candidate. An incongruent signal will lessen the respondent’s evaluation. (A
congruent signal is defined as information that informs an individual that the candidate
supports the same side of the issue that the individual supports.)
This Lemma assumes that voters receive and incorporate signals about politicians’ ide-
ologies into their issue beliefs. However, Zaller (1992) persuasively demonstrates that voters
filter out a great proportion of political information based on their partisan predispositions.
Research on deference to elites (Lupia and McCubbins, 1998) and cue-taking (Gilens and
Murakawa, 2002) shows how voters fundamentally alter their issue positions based on the
stances of their preferred elites. For example, a large proportion of Democrats were pro-
Social Security privatization (and vice versa for Republicans) before George W. Bush and
Al Gore took public stances during the 2000 campaign (Richard Johnston, 2004); after the
candidates made public statements, the partisans sorted into their respective positions.
2
As elaborated in Chapter 2, personal experience on an issue provides a basis on which a
voter can judge politicians independent of partisan biases. This political “ground truthing”
of candidates’ statements leads voters to critically evaluate politicians on issues with which
they have experience. Examples include retirees who receive Social Security checks, hunters
and guns, Hispanics about immigration, and veterans about war policy. This “responsive-
ness” to an issue can either be positive (i.e., holding the candidate in higher regard) or
negative. Recasting the Candidate Evaluation Hypothesis from Chapter 2 into a prediction
specific to candidate statements yields the first of two main hypotheses of this chapter:
2
For this experiment, a candidate’s signals will always contain new information about the candidate’s
stance.
76
Hypothesis 1 (Candidate Evaluation): Individuals with personal experience with an
issue are more responsive to a candidate signal on that issue than are those with no personal
experience. This responsiveness will be present even when signals run counter to existing
partisan biases (or candidate preferences).
Personal experience also moderates cue-taking from politicians on issues because the
voter’s experiences crowd out elite-generated considerations. For example, as shown in
Chapter 3, during the 2000 election, Republican retirees did not become ardent supports
of Social Security privatization because in their experience the system worked well for
them. This idea is formalized in the learning model presented in Chapter 2. Modifying the
Cue-Taking Hypothesis to fit the experiments presented in this chapter yields:
Hypothesis 2 (Cue-Taking): Individuals who have personal experience with a political
issue, when presented with information about their preferred candidate’s stance on the
issue, alter their issue opinion to match their preferred candidate’s opinion (i.e., cue-take)
less than those individuals who lack personal experience.
These hypotheses are extremely important for campaigns that seek to alter the pref-
erences of the electorate. If a losing campaign shifts the political discussion to an issue
that people know very little about (even if initial public opinion polls show support for
one position), voters will sort by their existing predispositions. Given these hypotheses,
campaigns would be more successful targeting voters with messages on issues with which
they have personal experience.
This survey also measures characteristics of individuals that relate to alternative and
mediating hypotheses to those above. Perhaps individuals with generally high political
knowledge are likely to have stable opinions on all opinions and experience has little or
no effect. Another possibility is that self-interest rather than personal experience causes
issue-driven shifts in political judgments. Also, experience may increase opinion certainty
77
(i.e., confidence that an opinion is correct) and this certainty may drive later behavior
(Fazio and Zanna, 1978).
4.2 Survey Design
4.2.1 Overall Goals
To test these hypotheses, two two-wave survey experiments are conducted: one pilot exper-
iment of Princeton undergraduates (n=273) and one survey of adults nationwide (n=391).
Both experiments are very similar in nature, with respondents of the first wave being
re-interviewed about a month later. All interviews are conducted online.
The surveys are conducted in two waves to better capture the effects of candidates
signaling on respondents’ issue opinion and candidate evaluation. The first wave provides
a baseline for respondents’ opinions absent the key parts of the candidates’ platforms. Over
a month later, respondents are re-asked the same questions, this time after the candidates
reveal their issue positions. Cross-tab and regression analyses of the change in respondents’
political opinions across waves test the validity of the two hypotheses. The month lag
between waves limits the respondent’s ability to remember her wave one responses and
lowers the possibility of the respondent attempting to game the survey.
3
The two hypotheses tested have different dependent variables of interest. The Candi-
date Evaluation Hypothesis requires that respondents judge politicians and the Cue-Taking
Hypothesis requires that respondents report issue positions. The survey asks respondents
to evaluate fictional candidates and real issues. Fictional candidates are used for two pri-
mary reasons: (1) to limit the amount of prior information, biases, or predispositions the
respondent has about the candidate, and (2) so that the candidate can plausibly take either
3
For instance, if no lag between the waves were included, respondents could purposely report the same
issue positive in both waves to prove that they hold a consistent opinion regardless of the cues they receive.
78
side of issues. Since the survey tests the interaction of personal experience with political
judgments, actual issues are used. Both surveys test two hypothetical candidates and three
real issues.
The basic structure of the surveys is: (1) respondents rate hypothetical candidates
based on paragraph descriptions, (2) respondents report issue opinions, (3) respondents
are queried on their issue experience, (4) a month later the same candidate paragraphs are
presented with the candidates taking stances on different issues, and (5) respondents rate
the candidates and report issue positions again. This setup allows the effect of candidate
position-taking on both candidate evaluation and issue opinion, and the interaction of
personal experience with these effects, to be tested.
Because the candidates are hypothetical and the issues are real, change in candidate
evaluation is more likely than change in issue opinion. Thus, the surveys are created with
the goal of fostering cue-taking. First, the candidate paragraphs attempt to mimic the par-
tisan division of the real political arena. (See Appendix C.1.1 for candidate descriptions.)
If Democrats feel an affinity to the hypothetical candidate, for example, they are more
likely to cue-take. Thus, every respondent sees the same hypothetical match-up between
one Democrat and one Republican. Second, although the candidate signals include no new
information on the issue, they are attempts to reflect the actual debate and are designed
to be forceful enough to move respondents. Third, the evaluations and issue opinions are
reported on 101- and 21-point scales respectively to detect small movements. The web
format facilitates this granularity by allowing respondents to choose their response via a
slider (Figure 4.1).
In the second wave, the respondent views the entire candidate paragraph description
again, but at the end of the paragraph sees the statement, “[Candidate name] recently
made the following comments about the debate over [Issue]: [Issue statement].” Each issue
79
Figure 4.1: Screenshot of survey question. Respondents report their issue opinion by sliding
a shuttle left or right across the screen. The web browser informs the respondent of the
shuttle current position (which, in the example above, is B:2, or 2 out of a maximum 10
on the B-side of the issue). Candidates are evaluated on a similar scale, though with more
granularity. For both issue opinion and candidate evaluation, each respondent uses the
vertical slider to indicate her certainty of her reported belief.
has two issue statements related to it, one liberal and one conservative. Candidates are
assigned issues and issue statements randomly with one constraint: candidates cannot take
two stances on the same issue. Thus, for every respondent, the Democratic candidate takes
a stance on one issue, the Republican candidate sends a signal on a different issue, and the
third issue is left unmentioned. This constraint allows for the source of any change in a
voter’s opinion to be clearly identified. Half the time the two candidates send two liberal
or two conservative signals (although again on different issues).
4.2.2 Specifics of Princeton Survey Design
The first wave of the Princeton survey was distributed to 1,600 students on December 6,
2007. A $100 gift certificate to either amazon.com or iTunes was offered as an incentive
to participate, with the winner to be chosen at random from students who completed the
80
survey. A reminder about the survey was sent out a week after the initial email, with a
second incentive (a $50 gift certificate) for the 400th respondent. The survey was open for
four weeks, during which 418 undergraduates completed the survey (26% response rate).
Students who completed the survey are referred to are “respondents.”
The candidate evaluation pitted a former governor (Richard Miller, Democrat) against a
Retired Admiral (Mark Jones, Republican) in a race for a U.S. Senate seat of an unspecified
state. Basic biographical information and endorsements were mentioned in the candidate
paragraphs.
Respondents were asked for their opinions on three issues: gay marriage, environmental
regulation, and foreign aid. The gay marriage question informed respondents that their
state supreme court had ruled that the state must offer same-sex couples either civil unions
or marriage. The respondent was asked to report which one of these options she preferred
and how strongly. The question about environmental regulation asked respondents to
choose between two solutions to global warming: a cap-and-trade system or a less restricted,
incentive-based solution. With respect to foreign aid, respondents were asked to indicate
whether they favored doubling or halving the foreign-aid budget.
The order of candidates and issues was not random, but differed by class year and
remained constant across both waves. For issues, the relative position of the traditionally
liberal side and conservative side was constant across the three issues and waves. No
statistically significant order effects were found.
To measure issue experience, respondents were asked about their personal habits. To
evaluate experience with gay marriage, respondents reported how often they communicated
with any gay and lesbian friends and family. With regard to the environment, students were
queried about the number of outdoor excursions they went on in the past year. To gauge
experience in foreign policy (specifically foreign aid), respondents indicated how much
81
time they had spent abroad (excluding Europe) in the past five years. Finally, respondents
were asked standard demographic questions about their gender, age, party affiliation, and
ideology, as well as how closely they follow politics.
On February 11, 2008, two months after the initial wave, an email alerted first wave
respondents that the second-wave survey would be available within a day. This email also
provided (with permission) the names of the gift certificate winners of the first survey. The
second-wave survey was delivered to students’ inboxes the following evening and again was
open for a month. Gift certificate incentives were provided, as with the first survey. Of
the 418 first-wave respondents, 283 completed at least some questions in the second wave
(273 finished the entire survey).
4.2.3 Specifics of Nationwide Survey
The first wave of the nationwide survey was distributed to 18,812 people on Survey Sam-
pling, Inc.’s email list on July 28, 2008. A standard incentive (points in Survey Sampling
Inc.’s program) helped motivate individuals to take the survey. A reminder was sent out
two days later. Of the recipients, 540 completed the survey within a week of the survey’s
launch, for a 3% response rate.
4
At the end of the survey, the respondents were told that
a second survey would be coming “next month.”
The candidate evaluation pitted a former governor (Richard Miller, Democrat) against
the current attorney general (Mark Jones, Republican) in a race for a U.S. Senate seat
of an unspecified state. In addition to biographical information and endorsements, each
4
Completion of the survey required not only going through all the questions, but also touching or
shifting at least one evaluation slider and at least one confidence slider. This eliminated respondents who
had technical difficulties. Another 10 respondents were eliminated from the analysis because their responses
and the amount of time it took them to record their responses indicated that they had not read the questions.
No wave two respondents were eliminated, reducing the possibility that this judgement affected the results.
82
description ended with a reason why opposing party leaders disliked the candidate in
question.
After completing the candidate evaluations, respondents were asked for their opin-
ions on three issues: immigration, the Iraq war, and gay marriage. For immigration, the
choices were between deporting illegal immigrants and providing a path to citizenship. On
the Iraq war, respondents chose between keeping troops in Iraq until that country is a
“stable, functioning democracy” and withdrawing troops “as quickly as the Defense De-
partment believes is safe.” For the issue of gay marriage, the respondents were presented
with the situation in which a state supreme court recently mandated legalizing same-sex
marriage. The respondents then indicated whether they supported passing a constitutional
amendment to overturn the decision or to recognize gay marriage.
The nationwide survey included a few additional features. First, for each evaluation
and opinion, respondents were asked to report the certainty of their response on a 0 to 10
scale (Figure 4.1). Also, after the candidate descriptions, respondents were asked to guess
what each candidate’s opinion would be on one issue (the same issue the candidate would
take a stance on in wave two).
Several measures of issue experience were recorded. First, respondents were asked to
think about their social network, specifically, those people they talk to when “important
issues or problems come up in life.” After reporting how many people were in this group
of close friends and family, respondents gauged what percentage of their important conver-
sations were with: individuals currently serving in the military, former military personnel,
immigrants, Hispanics, and gays or lesbians.
Other measures of issue experience were tested as well. With respect to experience with
immigration, Survey Sampling, Inc., provided the respondents’ race and ethnicity (i.e.,
whether the respondent was of Hispanic descent) and the respondent’s zip code, which was
83
matched with the percentage of Hispanics in the respondent’s zip code according to the U.S.
census. On wave two, respondents were asked what proportion of their co-workers were
Hispanic. For gay marriage, respondents were asked whether they considered themselves
gay or lesbian (at the end of wave two). For additional measures on experience with the
Iraq war issue, respondents were asked if they are currently serving (or had served) in the
military or whether this applied to another person in their household.
Finally, respondents were asked standard demographic questions about their party,
ideology, and political attentiveness. To determine whether the respondents represent
potential voters, they were asked whether they voted in 2006 and their intentions to vote
in 2008. At the end of the survey, the respondent was thanked for her time and told that
a follow-up survey would be delivered to her inbox in about three weeks.
On August 18, 2008, exactly three weeks after the initial wave, an email alerted first-
wave respondents to the availability of the new survey. Of the 540 first-wave respondents,
397 completed at least some questions in the second wave (391 finished the entire survey).
These numbers translate into a response rate of 74% on the second survey.
Finally, respondents were asked to evaluate and comment on the survey. Specifically,
respondents were asked if they felt “forced” to choose an option they “didn’t particularly
like.” Of the 391 respondents who reached that question, four objected to the issue ques-
tions in general, two to immigration specifically, one to the Iraq war, and two to the gay
marriage option for the gay marriage question. Also, 14 respondents specifically praised
the range of options provided for the issues.
The real-world political scene during the period between the surveys was relatively
quiet. The biggest news during the first wave of the survey was the residual effect of Barack
Obama’s Europe trip and the launching of John McCain’s “celebrity” advertisement (July
30). The second wave closed just as Obama announced the selection of his running mate,
84
Joe Biden. No major events occurred that were related to any of the three issues tested in
the survey.
4.3 Design Checks
4.3.1 Characteristics of the Survey Respondent Population
The characteristics of the Princeton survey’s population are constrained by the fact that
the respondents are all undergraduates at Princeton University. The population is 54%
female and a plurality are first-year students. Only 7% indicate that they follow politics
“a great deal,” with a plurality placing themselves in the middle of the scale (“some”).
For the national survey, the population is not restricted to registered voters or likely
voters because individuals’ decisions to register to vote or to cast a ballot are ongoing de-
cisions. For instance, someone might want to register or decide to vote only after receiving
positive information about a candidate (Sanders, 2000). That said, one goal of conducting
a nationwide survey is to produce a respondent population that is similar to the national
electorate.
5
Certainly very good social science can be performed in a laboratory with a par-
ticipant group that looks very different from the population at large. However, treatment
effects observed in one segment of the population might be counterbalanced by opposite
effects in other segments of the electorate that are not included in the experiment.
As a reference point, the demographic composition of the survey respondents is com-
pared to the composition of the National Election Pool’s 2004 general election exit poll.
6
Overall, the demographics of the two groups are strikingly similar (Table 4.1), especially
considering the nature of an Internet survey. One might expect older, poorer, and mi-
5
The electorate and not the adult population is used as a reference point because the theory’s effects
on nonvoters are inconsequential. Applications of this theory would be focused on likely voters and they
decide which candidate wins or loses.
6
As of June 2009, individual data for the 2008 exit poll is not available.
85
nority individuals to be underrepresented, as those groups traditionally use the Internet
less. However, these demographics are well-represented in the survey, meeting or exceeding
their percentages in the electorate. The only troubling statistic is the overrepresentation
of non-churchgoers in the survey experiment: 32% of respondents report that they never
attend church, while 15% of the electorate report the same. The overall similarity between
the two populations is a benefit of the fact that Survey Sampling, Inc., is a professional
organization that produces nationally representative surveys.
Because Democrats have gained significantly in party identification since November
2004, Pew Research Center’s August 2008 poll is used to compare partisanship. The two
surveys report approximately the same proportion of Democrats and Republicans, although
the survey experiment includes more self-reported Independents than Pew reports. This
result may be due to differences in question wording (all options, including “lean Demo-
crat/Republican,” are immediately presented to the respondent in the survey experiment,
unlike in Pew).
4.3.2 Features of the Survey
Candidate Descriptions. To foster cue-taking in the second wave of the survey, the
paragraph descriptions of the candidates are designed to elicit both positive evaluations
and partisan differences. The nationwide survey fixes some deficiencies of the Princeton
survey to achieve these goals.
For the Princeton survey, respondents of all political affiliations are likely to rate both
candidates positively, although Democrats rate the Democratic candidate more positively
than the Republican candidate (and vice versa for the Republicans). Independents fall in
between the partisans, rating the Democratic nominee more highly, although this difference
does not reach statistical significance (p = 0.15). On a scale from -10 (worst) to 10
86
2008 2004 2008 2004
Internet Electorate: Internet Electorate:
Attribute Survey Exit Poll Diff. Attribute Survey Exit Poll Diff.
Gender Church Attendance
Male 43% 46% -3% > Weekly 13% 16% -3%
Female 57% 54% 3% Weekly 19% 26% -7%
≤ Monthly 9% 14% -5%
Race* A few times
White 79% 77% 2% a year 26% 28% -2%
Afr-Amer 14% 11% 3% Never 32% 15% 17%
Latino 13% 8% 5%
Asian 4% 2% 2% Whites: Are you born again/evangelical?
Other 10% 2% 8% Yes 29% 23% 6%
Refused 4% 0% 4%
Have you ever served in the military?
Age Yes 16% 18% -2%
18-29 16% 17% -1%
30-44 25% 29% -4% Are you currently married?
45-59 31% 30% 1% Yes 59% 63% -4%
60+ 28% 24% 4%
Are you gay, lesbian, or bisexual?
Party ID† Yes 6% 2% 4%
Democrat 31% 37% -6% No 92% 45% 47%
Republican 25% 28% -3% Refused 2% 54% -52%
Ind/Oth 44% 35% 9%
Income
Education under $30K 24% 20% 4%
< H.S. Grad 2% 4% -2% $30K-50K 22% 18% 4%
H.S. Grad 18% 22% -4% $50K-75K 20% 21% -1%
Some College 38% 32% 6% $75K-100K 10% 12% -2%
Coll Grad 27% 26% 1% $100K-150K 7% 10% -3%
Post Coll 15% 16% -1% $150K+ 3% 7% -4%
Refused 15% 11% 4%
Religion
Protestant‡ 33% 37% -4% Region
Catholic 22% 19% 3% East 22% 22% 0%
Mormon/LDS 1% 1% 0% Midwest 20% 26% -6%
Jewish 4% 2% 2% South 39% 32% 7%
Muslim 0% 0% 0% West 20% 20% 0%
Other/None 40% 11% 29%
Refused 0% 29% -29%
*Internet survey allowed more than one race to be specified
†From Pew (August 2008) instead of exit poll
‡Includes other Christians
Table 4.1: Demographics of survey experiment are similar to those of the November 2004
electorate. Because of the similarities, the national survey is not weighted. The exit poll
results are weighted according to the documentation.
87
(best), the average rating for the Democratic candidate is 5.2 and for the Republican
candidate is 3.3 (p < 0.01). The correlation between the two ratings is positive (ρ = 0.14),
meaning that the descriptions did not elicit strong partisan views. This correlation indicates
that partisan feelings are often trumped by the amount of “trust” (or some other valence
characteristic) that respondents afford to hypothetical candidates for which they have only
a brief description.
Across all respondents, the hypothetical candidates are rated slightly positively, with
means of 58 and 56 for the Democrat and Republican respectively (Figure 4.2). When
regressed on 7-level partisanship, the relative evaluation of the Democrat at the intercept
is only 0.2 points above the Republican (a statistically insignificant difference). Thus, con-
trolling for the partisanship of the respondent, the candidate profiles are equally positive.
Within respondent, the correlation between the Democratic and Republican evaluations
is negative and statistically significant (ρ = −0.16), meaning that in rating the candi-
dates, partisan feelings trump any propensity to rate hypothetical candidates positively or
negatively in general.
Issue Opinion. In both surveys, opinions are relatively split across all three issues. In
the Princeton survey, no side of any issue garners more than 60% support. Students support
gay marriage over civil unions 55% to 36% (with 10% neutral). A smaller plurality supports
a cap-and-trade system to regulate greenhouses gases (49%) over encouraging businesses
to be green while letting the market rule (42%; 8% neutral; 1% refused). On the issue of
foreign aid, 59% of respondents want to increase the budget for foreign aid, 21% want to
decrease it, and 20% are neutral.
Similar splits are seen in the nationwide survey. On a +10 (most liberal) to -10 (most
conservative) scale, the mean responses are +1.4, +1.7, and -1.0 for immigration, the Iraq
war, and gay marriage, respectively. Across all issues, the highest proportion of respondents
88
Democratic Candidate Republican Candidate
Mean Evaluation of Hypothetical Candidates
By Respondents' Party Affiliations
M
e
a
n

C
a
n
d
i
d
a
t
e

E
v
a
l
u
a
t
i
o
n

(
0

t
o

1
0
0

s
c
a
l
e
)
0
2
0
4
0
6
0
8
0
1
0
0
All Respondents
Democratic Respondents
Independent Respondents
Republican Respondents
Figure 4.2: Mean evaluations of hypothetical candidates by party for nationwide survey.
N-size is 397 for all respondents, 258 for Democrats, 66 for independents and 90 for Re-
publicans. Thick lines show plus or minus one standard error; thin lines are plus or minus
two standard errors.
on one side of the issue is 62% favoring a path to citizenship for immigrants. All three
issues correlate as expected with party; the Iraq war polarizes partisans the most (Figure
4.3).
Candidate Signals. Hypothetical candidates’ attempts to persuade voters through
a static text cannot be expected to move voters’ issue opinions very far. Two other fac-
tors lower expectations even further. First, the persuasion is not directly aimed at the
respondent; the candidate only states his own issue opinion. Second, half of the signals
emanate from disliked candidates. While the signals generally work as expected, these low
expectations are met in both surveys.
In the Princeton survey, candidate signals prove somewhat effective in two of the three
issues: gay marriage and the environment. Despite the entire survey population moving
89
Immigration Iraq War Gay Marriage
Mean Issue Opinion
By Respondents' Party Affiliations
M
e
a
n

I
s
s
u
e

O
p
i
n
i
o
n

L
i
b
e
r
a
l
n
e
s
s

(

1
0

t
o

1
0

s
c
a
l
e
)

1
0

5
0
5
1
0
All Respondents
Democratic Respondents
Independent Respondents
Republican Respondents
Figure 4.3: Mean issue opinions by party for the nationwide survey. N-size is 397 for all
respondents, 258 for Democrats, 66 for independents and 90 for Republicans. Thick lines
show plus or minus one standard error; thin lines are plus or minus two standard errors.
more conservative on gay marriage and more liberal on the environment, candidate signals
move respondents about 0.5 points (on average) on a 21-point scale in the direction of
the signal (beyond any overall survey population movement). A similar magnitude is
seen for the effect of the conservative foreign aid signal, but the liberal signal also pushes
respondents in the conservative direction.
7
For the nationwide survey, the most any signal moves voters is 1 point (on average)
on a 21-point scale (Figure 4.4). Immigration is an ideal case, with the electorate not
shifting opinion between waves and the signals moving opinion by about a point. The
survey population shifts to the left on the Iraq war between waves, but the conservative
signal dampens this migration. Across the three issues, the liberal signal on gay marriage
7
Foreign aid is consistently an ineffective issue in the Princeton survey, which lead to its omission in the
nationwide survey.
90
is the only signal that does not shift opinion in the expected direction when compared to
the group that does not receive a signal. However, the liberal gay marriage signal proves
effective when the signal is delivered by a preferred candidate.
Immigration Iraq War Gay Marriage
Issue Opinion Shift
By Signal Direction
S
h
i
f
t

i
n

I
s
s
u
e

P
o
s
i
t
i
o
n

(
p
o
s
i
t
i
v
e
=
l
i
b
e
r
a
l
)

3

2

1
0
1
2
3
All Respondents
Liberal Signal
Conservative Signal
No signal
Figure 4.4: Mean shift in the liberal direction of issue opinions from wave one to wave
two in the nationwide survey. Although the relative effect of the liberal and conservative
signals are in the expected direction in nearly all cases, none of the signals are particularly
powerful.
4.3.3 Definitions of Issue Experience
Although the definition of issue experience is somewhat arbitrary, the surveys attempt
to elicit the amount of interaction between the respondent and a political issue. These
survey measures can certainly be called into question, especially for the Princeton survey.
The nationwide survey attempts to gauge issue experience more accurately via either life
experience (e.g., veteran and Iraq war) or contact with people directly affected by the issue
(e.g., having close gay friends or family and same-sex marriage).
91
Immigration Iraq War
Variable Loading Variable Loading
Perc. of convs. with Hispanics 0.88 Perc. of convs. with military .50
Respondent is Hispanic .55 Perc. of convs. with veterans 0.69
Perc. of convs. with immigrants .52 Someone in household serves 0.5
Perc. Hispanic in zip code 0.5
Perc. Hispanic in workplace .38
Table 4.2: Factor loadings for immigration and Iraq war experience measures.
The Princeton survey has admittedly poor measures of issue experience. For example,
number of outdoor excursions in the past year is the chosen proxy for experience with the
environment. Potentially worse, amount of time spent abroad (outside North America and
Europe) in the past five years is the measure for experience with foreign aid. The gay
marriage measure is probably a more accurate depiction of experience; respondents are
asked how often they communicate with gay and lesbian friends and family.
The nationwide survey contains several measures of experience, with an emphasis on
conversations with individuals affected by the issue in question. (See Appendix C.1.3 for
question wording.) These respondent characteristics are combined into one measure of
experience for each issue. For the issues of immigration and the Iraq war, the disparate
potential sources of issue experience are amalgamated into one measure using factor anal-
ysis. (Factor loadings are described in Table 4.3.3.) For measures in both surveys, when
issue experience is dichotomized for easier display in figures or hypothesis testing, the top
30% of respondents (approximately) in each of the three experience measures are labeled
“experienced.”
92
4.4 Support for Hypotheses
4.4.1 Candidate Evaluation
The changes in candidate evaluations are the crucial piece of this analysis; these shifts
indicate how the voters might take a second look at a previously non-preferred candidate
in an election. The signals work well, producing large shifts predicted by Lemma 1: when
the candidate announces an issue stance close to a respondent’s position, that respondent
evaluates the candidate more highly.
8
The opposite happens when the candidate announces
an issue stance far from a respondent’s own view. These two situations are described as the
candidate delivering a “congruent” or an “incongruent” signal. The large (and statistically
significant) difference between the effects of these two types of signals for the nationwide
survey is displayed in Figure 4.5. Similar effects are present in the Princeton survey.
9
Voters with personal experience on related issues exhibit larger swings in their can-
didate re-evaluations as the Candidate Evaluation Hypothesis predicts. The data can be
analyzed two ways: by candidate (combining issue signals) and by issue (combining can-
didate evaluations). As an example of the former, in the Princeton survey, the average
difference in the change in Democratic candidate evaluation is 2.6 points (21-point candi-
date scale). For individuals experienced on the issue on which the Democratic candidate
delivered his signal, this change in evaluation is higher: 3.75 points (p = 0.18, all re-
ported p-vales are two-tailed), as the Candidate Evaluation Hypothesis predicts. For the
Republican candidate’s evaluation, a similar effect is apparent: a change of 4.1 points for
8
When signal congruity appears in a dichotomous context, congruous signals are only those on the same
side of the issue as the respondent’s wave-one opinion. When the concept is measured as a continuous
variable, it is operationalized as the distance between the respondent’s wave one opinion and either +10
(for liberal signals) or -10 (for conservative signals).
9
The average re-evaluation of the Democratic candidate is 0.1 points higher than the wave one evaluation
when he delivers a congruent signal, and lower by 2.5 points when he delivers an incongruent signal (20-
point scale). The analogous effects for the Republican candidate and 1.6 and -2.1. These differences are
highly statistically significant.
93
Democratic Candidate Republican Candidate
Candidate Evaluation Shift
By Signal Congruency and Candidate Party
C
h
a
n
g
e

i
n

C
a
n
d
i
d
a
t
e

E
v
a
l
.

(
p
o
s
s
.

r
a
n
g
e


1
0
0

t
o

+
1
0
0
)

2
0

1
0
0
1
0
2
0
All respondents
Received congruent signal
Received inconguent signal
Figure 4.5: Mean difference between candidate evaluations in wave one and wave two.
Respondents who are neutral on an issue are mixed in with respondents who received
a congruent cue. N-sizes are 397 overall, with 200 receiving congruent cues from the
Democrat (197 incongruent) and 214 receiving congruent cues from the Republican (183
incongruent). Thick lines show plus or minus one standard error; thin lines are plus or
minus two standard errors.
experienced individuals and 2.8 points for non-experienced individuals (p = 0.32). These
results are noisy, but supportive of the hypothesis.
When the evaluations of the Democratic and Republican candidates are combined into
one evaluation margin (e.g., Democratic evaluation minus Republican evaluation) for each
issue, issue-specific effects are quantified. For instance, the average difference in Democratic
margin from wave one to wave two when the Democrat delivers a congruent signal on gay
marriage or the Republican delivers an incongruous signal is 5.1 points (21-point candidate
scale).
10
For those with experience on the issue, however, this margin jumps to 7.9 points
10
This figure includes the change in Republican margin when the Republican delivers a congruent signal
on gay marriage or the Democrat delivers an incongruous signal.
94
(p = 0.06). Analogously, for the environmental issue, experienced respondents alter their
candidate margin by 2.9 points in the direction of the congruity of the signal, while non-
experienced respondents alter their margin by 1.8 points (p = 0.62). Foreign aid exhibits
the characteristic of a weak experience variable, with the margin difference increasing from
3.1 for respondents without experience to 3.9 for respondents with experience (p = 0.82).
The noise in these supportive results is from the issues of environment and foreign aid.
The nationwide experiment produces similarly supportive but noisy results. As Figure
4.6 shows, for each issue, experienced voters adjust their evaluations of candidates in the
direction of the signal (i.e., upward for congruent signals and downward for incongruent
signals) further than non-experienced respondents. Assuming the tests are independent
(implied by the randomized treatment assignment) the relationship is significant at the
10% level (two-tailed) across all (but not within each) issues. Regressions (one for each
issue) with one interaction term (congruency crossed with experience), the requisite main
effects (in their continuous form), and no controls yield similar, though slightly less noisy,
results (Table C.1).
These results validate the main hypothesis, but two additional analyses are crucial
to bolstering the claim. First, alternative hypotheses, such as political attentiveness and
self-interest, are considered. Second, respondents who are reacting to traditional partisan
biases are weeded out so that potential vote switchers are isolated.
The data indicate that political attentive respondents hinge their candidate evaluations
on the candidates’ signals. Regression analysis (Table C.1) demonstrates that political
attention strongly interacts with candidate signals. Respondents who pay attention to
politics rely on candidate issue positions when judging these candidates to a greater degree
than their less-attentive counterparts.
95
Immigration Iraq War Gay Marriage
Candidate Evaluation Shift
By Issue and Respondent Experience
C
h
a
n
g
e

i
n

C
a
n
d
i
d
a
t
e

E
v
a
l
.

i
n

t
h
e

D
i
r
e
c
t
i
o
n

o
f

S
i
g
n
a
l
0
5
1
0
1
5
2
0
2
5
3
0
All Respondents
No Issue Experience
Issue Experience
Figure 4.6: Mean difference between candidate evaluations in wave one and wave two,
coded so that the direction of the shift matches the direction of the signal. Experienced
respondents are more responsive to the candidate signals than non-experienced respon-
dents. Thick lines show plus or minus one standard error; thin lines are plus or minus two
standard errors. National survey.
A similar conclusion cannot be reached for self-interest as the results are mixed. Two
of the three issues have measures of self-interest: the Iraq war (active military members)
and gay marriage (gay respondents).
11
Self-interest for gay marriage works as expected;
gay and lesbian respondents are more sensitive to the candidate’s stances on extending
marriage rights. The reverse is true for the Iraq war; current members of the military are
less responsive to the candidate’s position on this issue.
The Candidate Evaluation Hypothesis holds even when controlling for both of these
effects. All the signs are in the correct direction, and jointly the results are statistically
11
Answers to the vote questions indicate that all respondents are U.S. citizens; thus, no self-interest
measure is available for immigration. For the other two issues: 20 respondents currently serve in the
miliary and 24 respondents identify as gay, lesbian, bi-sexual, or transgender.
96
significant at conventional levels (p = 0.06, one-tailed). Self-interest soaks up much of the
effect of personal experience for the issue gay marriage. These two measures are highly
correlated (ρ = 0.46) so teasing out the independent effects of self-interest and personal
experience is more difficult.
Issue certainty also affects political judgements in the expected direction, but does
not appear to be a mediator for political experience. Respondents with more certainty
about their opinion on the issue are more sensitive to the candidate’s position on this
issue. The magnitudes of the effects of personal experience do not decrease when this
potential mediator is added the regression indicating that these effects are independent of
each other.
12
The effects of personal experience are apparent, though weaker, when the candidate
signals counter partisan biases. A signal is said to go against partisan biases if: (1) it
is a congruent signal delivered by a candidate that the respondent did not favor in the
first wave, or (2) it is an incongruent signal delivered by a candidate that the respondent
did favor in the first wave. An example would be a Democratic respondent who follows
her party affiliation in wave one and rates the Democratic candidate more highly but
receives a signal in wave two that the candidate opposes ending the Iraq war. As the Iraq
war is the most polarizing issue, the Democratic respondent mostly likely assumed that
the Democratic candidate supported ending the war. She now lowers her opinion of the
Democratic candidate, and, depending on the signal sent by the Republican candidate,
might even rate the Republican higher.
When restricting the analyses to these types of signals, the effect of personal experience
generally remains. Regression analyses of the two candidate evaluations, limited to anti-
partisan bias signals, finds positive and significant effects in the Princeton survey (joint
12
Causal mediation analysis would be a more proper method of assessing the causal mechanism.
97
p = 0.04, one-tailed).
13
The national survey analyses control for political attentiveness
and self-interest and find effects consistent with the Candidate Evaluation Hypothesis in
two of three issues (joint p = 0.15, one-tailed). With regard to the Iraq war, a change
from below-average experience to above-average experience (i.e., a two standard deviation
change) increases the relative candidate evaluations (100-point scales) by 6 points when
that candidate takes a congruent rather than incongruent stance on the war. (Details
for each group of interest are listed in Table C.3.) Shifting an individual’s views against
their partisan predispositions is difficult, but sending congruent signals about an issue with
which the voter has experience helps grease the wheels.
4.4.2 Issue Opinions and Experience
The Cue-Taking Hypothesis—that issue experience moderates cue-taking—is supported
by the evidence. The data are broken down by issue and whether the signal is received
from the more favored candidate (as defined by the evaluations in wave one), yielding six
tests. The Cue-Taking Hypothesis predicts that in each case, but especially when a favored
candidate delivers a signal, individuals without personal experience on the issue will shift
their opinion in the direction of the candidate signal more than those with experience. In
both surveys, this relationship is observed in five of the six cases.
The signals in the Princeton survey all produced (on average) very small movements in
issue opinion. The largest amount of cue-taking is among non-experienced respondents on
the gay marriage question; these respondents shifted less than half a point (on a 21-point
issue scale) in the direction of a favored candidate’s signal. When signals are delivered by
favored candidates, the average shift is always toward the candidate’s position except for
experience respondents on the environment issue, who counterargued against the signal
13
Only the two main effects, congruent signal and experience, and the interaction term are included in
these regressions.
98
(Figure 4.7). Only signals from unfavored candidates on the issue of foreign aid generated
an unexpected relationship between experience and cue-taking; that reverse relationship is
far from statistically significant.
Gay Marriage
(Friends)
Environment
(Excursions)
Foreign Aid
(Trips Abroad)
Cue Taking By Issue−Experience Pair
C
h
a
n
g
e

i
n

I
s
s
u
e

O
p
i
n
i
o
n

i
n

D
i
r
e
c
t
i
o
n

o
f

C
u
e

0
.
5
0
.
0
0
.
5
1
.
0
All
Cue from favored; no exp.
Cue from favored; exp.
Cue from unfavored; no exp.
Cue from unfavord; exp.
Figure 4.7: Mean shift, in the direction of the received signal, of issue opinions from wave
one to wave two in the Princeton survey. In five of six cases, non-experienced respondents
shift their opinions in the direction of the signal further than experienced respondents.
In the nationwide survey, the immigration signal from a favored candidate yields a stark
contrast between individuals with experience (who counterargue the signal) and those with-
out experience (who shift toward the candidate’s position by over one point on the 21-point
issue scale). All six relationships are displayed in Figure 4.8; only a signal from a favored
candidate on the Iraq war produces an unexpected relationship.
14
More counterarguing
of signals from unfavored candidates is apparent in the nationwide survey than in the
Princeton survey, perhaps because more partisan feelings were aroused by the candidate
14
When experience is measured dichotomously and favored/unfavored signals are combined, t-tests are
significant at the 5% level for immigration and gay marriage.
99
descriptions (which include more partisan language). Table C.3 breaks down these effects
by specific experience measures.
Immigration Iraq War Gay Marriage
Issue Opinion Shift in the Direction Of Signal
By Signal Source and Respondent Experience
S
h
i
f
t

t
o
w
a
r
d

s
i
g
n
a
l

(
p
o
s
s
.

r
a
n
g
e


2
0

t
o

+
2
0
)

1
0

5
0
5
1
0
All who rec'd signal
From favored/no exp.
From favored/exp.
From unfavored/no exp.
From unfavored/exp.
Figure 4.8: Mean shift, in the direction of the received signal, of issue opinions from
wave one to wave two in the nationwide survey. In five of six cases, the non-experienced
respondents demonstrated cue-taking more than the experienced respondents.
In contrast to the clear link between cue-taking and experience, there is much less of
a relationship between cue-taking and attentiveness. The phenomenon of attentiveness
mediating cue-taking is not statistically significant, and for the issue of the Iraq war the
relationship runs counter to the hypothesized relationship (see dichotomized results in
Figure 4.9). No consistent effect is found for self-interest either (Table C.2). Certainty
about one’s opinion has no effect on change in opinion demonstrating that individuals are
poor judges of themselves. Overall, the Cue-Taking Hypothesis is robust to alternative
explanations.
100
Immigration Iraq War Gay Marriage
Issue Opinion Shift in the Direction Of Signal
By Signal Source and Respondent Political Attentiveness
S
h
i
f
t

t
o
w
a
r
d

s
i
g
n
a
l

(
p
o
s
s
.

r
a
n
g
e


2
0

t
o

+
2
0
)

4

2
0
2
4
6
8
All who Rec'd Signal
From favored/Inattentive
From favored/Attentive
From unfavored/Inattentive
From unfavored/Attentive
Figure 4.9: Mean shift, in the direction of the received signal, of issue opinions from
wave one to wave two. Attentive respondents are defined as those who read a presidential
campaign story at least daily (n=193).
4.5 Discussion
The results of these two survey experiments generally confirm the Personal Experience
Model: personal experience with issues plays a large role in the formation of political
judgments. Even taking political attentiveness and self-interest into account, experience
with political issues leads to less cue-taking from politicians and to more issue-driven
evaluation of those politicians.
Campaigns should especially take note of the magnitude of the Candidate Evaluation
Hypothesis’s effects. In both the Princeton and nationwide surveys, the average increase in
the magnitude of candidate evaluation shift due to experience is large enough to switch the
preferences of an additional 3% of respondents. In many close races, 3% is the difference
between winning and losing. Further, the effects in the national survey occur for minor
101
issues; the economy is cited as the most important issue, by far, in contemporaneous
surveys.
15
Many of the individual results are not statistically significant at conventional levels.
Much of this noise is probably due to the use of hypothetical candidates as proxies for
real-world politicians and low sample size.
16
Interestingly, however, the cue-taking results
are less noisy and less susceptible (at least in the nationwide experiment) to alternative hy-
potheses than the candidate evaluation results. This difference runs counter to an intuitive
prior belief that because it should be easier to move individuals’ opinions of hypothetical
candidates, the differences in candidate evaluation among subgroups should be easier to
identify. While the premise of that belief is correct (both surveys showed that as a percent-
age of the scale, candidate evaluations shifted more on average than issue opinions), the
implication does not follow. Furthermore, while many of the results do not independently
reach statistical significance, nearly all point in the direction predicted by the hypotheses.
A secondary finding is that non-experienced respondents did not counter-argue signals
from unfavored politicians. The theory on this topic is mixed, with Zaller (1992) claiming
that in a one-issue environment all voters take cues (even from out-party politicians) and
Lupia (1994) finding that voters can glean negative information from endorsements by ide-
ologues on the other end of the spectrum. The evidence here lends some support to Zaller,
but perhaps the one-issue environment tilts the playing field unfairly in that direction.
17
Another more tentative finding (Table C.1) is that self-reported certainty may play a
role similar to personal experience when judging candidates. The data do not indicate
15
Cooper, Michael and Dalia Sussman. “Voters in Poll Want Priority to Be Economy, Their Top Issue.”
The New York Times. Also see August 18-24 Diageo/The Hotline poll.
16
Power analysis of the Princeton results indicates that a sample size of 1,000 would generate two-tailed
p-values of at most 0.05 about two-thirds of the time.
17
Lupia’s finding comes from a ballot initiative where even if only one side of the argument is heard,
clearly there are two active sides to the referendum (otherwise one side would not be spending money to
put out its message). In the case of the Princeton and nationwide survey experiments, each issue receives
at most one signal, thereby eliminating the possibility of two-sided debate.
102
that certainty is a mediator of the Personal Experience Model’s causal mechanisms; in-
stead, voters may evaluate candidates on issues about which they hold confident opinions
independent of personal experience. However, certainty about an individual’s own opinion
does not temper that individual’s propensity to change her opinion later.
103
Chapter 5
Randomized Field Experiments:
Optimizing Campaign Strategy
and Evidence for the Personal
Experience Model
1
5.1 Introduction
For campaigns or political organizations to apply the Personal Experience Model, they must
be able to identify and target individuals with personal experience. The widening scope of
campaign databases, which now include not just basic information such as age and gender,
but also commercial (e.g., magazine subscriptions), governmental (e.g., fishing licenses),
and census (e.g., percent poverty in census block) data, is both a help and a hindrance to
campaigns. While campaigns have more granular information at their fingertips, they may
1
Large sections of this chapter are adapted from Imai and Strauss (2009).
104
not be equipped to sift through the mounds of data. This chapter develops a methodological
process for identifying voters who are responsive to campaign appeals by utilizing and
discriminating among information gathered from randomized experiments. Second, this
chapter examines two cases in which the results of field experiments are consistent with
the Personal Experience Model.
Over the last decade, political scientists have shown renewed interest in the use of
randomized field experiments to study voter turnout and persuasion (e.g., Gerber and
Green, 2000; Nickerson, 2007, among many others). Building on work from more than a
half century ago (e.g., Gosnell, 1927; Hartmann, 1936; Eldersveld, 1956), these researchers
have developed creative methods for conducting field experiments with a large number of
voters in real electoral environments as a way to test various theories of voter turnout
(e.g., Gerber et al., 2003; Nickerson, 2008). The empirical findings of these studies have
the potential to significantly affect the practice of get-out-the-vote (GOTV) efforts and
persuasion campaigns in the real world (Green and Gerber, 2008).
Political science research that reports the results of field experiments tends to focus on
the statistical significance of the estimated overall average treatment effect (ATE) of each
campaign appeal. In addition, many researchers implicitly assume the constant additive
treatment effect across individual voters when using regression models for statistical anal-
ysis. This leads to the standard practice of reporting a single estimate (or at most a small
number of estimates) summarizing the efficacy of each mobilization method.
The findings based on such an approach may not be of much use for campaign planners
for two reasons. First, a planner must consider the problem of treatment effect heterogeneity
where each of the available campaign tactics may mobilize different voters to a varying
degree. The Personal Experience Model predicts one type of heterogeneity: Voters with
experience on an issue will be more responsive to campaign appeals on that issue. The
105
second and related problem is that a planner faces budget constraints and must evaluate
the cost-effectiveness of available strategies, which may vary across different voters. Thus,
uniformly applying a single appeal to the entire target population is at best suboptimal
and sometimes is not even feasible. This means that the standard practice of academic
researchers reporting only the estimated overall ATE may not provide useful information
from a practitioner’s perspective. Similarly, the common assumption of constant additive
treatment effect is too restrictive and does not serve the purpose of campaign planners.
This gap between academic research and policy making can be addressed through an
essential and yet missing methodological tool. A formal decision-theoretic framework (Imai
and Strauss, 2009) allows campaign planners, both partisan and nonpartisan, to use the
results of field experiments effectively when planning their mobilization campaigns. In
the proposed Bayesian-decision theoretic framework, a partisan planner maximizes the
posterior probability that the party’s own candidate wins the election subject to a budget
constraint, whereas the objective function of a nonpartisan planner is the posterior expected
turnout among a target population of voters. A planner starts with a prior belief about the
effectiveness of each mobilization technique under consideration for voters with different
characteristics and updates this belief based on the available experimental data. With this
framework, campaigns can use the data from randomized field experiments to develop the
optimal campaign strategy for each target population.
This procedure is applied in three cases. First, in a nonpartisan setting, the nonpara-
metric method finds that recipients of mobilizing text messages in 2006 (Dale and Strauss,
2009) who were between the ages of 20 and 24 were most responsive to the GOTV appeal.
Second, in the first of two examples that relate to the Personal Experience Hypothesis, a
mailing campaign about education attempting to persuade individuals to vote for a Ken-
tucky gubernatorial candidate had the most effect on parents. And finally, an anti-McCain
106
2008 mail and robo-call program that emphasized the economy had the most effect on
individuals living in economically distressed areas.
5.1.1 Background of the Methodological Problem
The problem of optimizing campaign tactics across voter characteristics was first introduced
by Kramer (1966) more than forty years ago. In that research, a campaign manager is
assumed to face the decision of whether to conduct blind or selective canvassing across
precincts subject to a budget constraint. After illustrating his decision-theoretic approach
with artificial data about the partisan make-up of precincts, Kramer (1966) concludes that
“the use of quantitative methods for policy analysis has proved to be fruitful in many
different fields, and these methods deserve to be more widely known, and used, in political
science (p. 160).”
Unfortunately, few scholars who conduct field experiments have followed up on Kramer’s
proposal to inform policy makers with a formal decision-theoretic analysis. One exception
is a small number of researchers who have examined the related question of which sub-
groups exhibit larger treatment effects (e.g., Gerber, 2004; Nickerson and Arceneaux, 2006;
Arceneaux and Kolodny, 2009). Campaign managers can use these findings as a basis for
planning their GOTV or persuasion campaign.
However, an important and well-known methodological problem is that if subgroups
are formed after the experiment is conducted in an ad hoc manner, the analysis runs a risk
of finding statistically significant results when no true relationship exists (Pocock et al.,
2002). Moreover, these studies do not examine the issue of cost-effectiveness in the presence
of a budget constraint, which is an essential consideration for campaign planners. Thus,
the planner needs a principled and systematic approach for deriving the optimal campaign
107
strategy from experimental data while avoiding the post-hoc subgroup analysis and related
statistical problems.
The proposed methodology extends Kramer’s pioneering work by placing the campaign
planner’s problem in the formal framework of statistical decision theory and applying
modern statistical methods. The decision-theoretic framework is based on literature about
treatment choice (e.g., Manski, 2005). This framework differs from the literature by adopt-
ing a Bayesian approach rather than a frequentist approach based on the maximin or
minimax-regret criteria (see Dehejia, 2005, for a notable exception). The standard linear
programming algorithm can be used to derive the optimal campaign strategy within this
framework.
To address the subgroup analysis problem, which has been somewhat neglected in the
treatment choice literature a new variable selection algorithm is presented and then used
in combination with nonparametric methods and cross-validation procedures in order to
avoid the over-fitting of statistical models. These methods are then extended to the cases
of optimal partisan campaign planning. This set of methods can be used to derive the
optimal campaign strategy from randomized field experiments.
5.2 The Formal Framework of GOTV Campaign Planning
The problem of deriving the optimal nonpartisan GOTV campaign strategy from exper-
imental data can be formalized as a statistical decision problem where the choices are
treatment applications (e.g., Manski, 2005). This framework can then be extended to the
optimal partisan persuasion (and possible partisan GOTV) campaign. Rather than taking
a frequentist approach, which is dominant in the treatment choice literature, this method
is based on Bayesian statistical decision theory (Berger, 1985) and assumes that a GOTV
campaign planner learns about the effectiveness of various mobilization methods for dif-
108
ferent voters by analyzing randomized field experiments. The planner then maximizes
the posterior expected turnout among a target population of voters subject to the budget
constraint.
2
5.2.1 The Planner’s Decision Problem
The first step of the approach is to formally state the decision problem faced by a nonpar-
tisan GOTV campaign planner. Let { denote this planner’s target population of voters
where this population is assumed to be finite and of size N. Typically, the target popu-
lation is the registered voters in the electoral district, whose complete list is available to
the planner (at least in the U.S.). Then, the planner’s decision problem is to assign one
of K available mobilization techniques (i.e., treatments) to each member i of this popula-
tion. If a GOTV campaign can be planned at the level of households or precincts, then i
indexes the appropriate units rather than individual voters. Although the units of analysis
are represented as individual voters for notational simplicity, the proposed methodology is
directly applicable to aggregate units.
An unordered set T ≡ ¦0, 1, . . . , K − 1¦ where K ≥ 2 denotes the finite set of mobi-
lization techniques from which the planner makes a selection for each voter in {. Note
that T = 0 represents the strategy of not mobilizing (i.e., doing nothing). For example,
the planner may consider three strategies (i.e., K = 3) where T = 2 and T = 1 represent
a GOTV method based on a phone call and a postcard, respectively, and T = 0 denotes a
status quo strategy that involves no such phone call or mailing. Alternatively, mobilization
techniques may differ in their frequency, timing, content of messages, and other aspects.
2
The following sections describe the proposed methodology as if a campaign planner both analyzes the
data and makes the decision. The methods apply equally to the situation where a planner makes a decision
based on the recommendations of a data analyst.
109
Next, suppose that the planner observes the J-dimensional (pre-treatment) covariates
X for each member of this population {. Since { is a finite population, this means that the
planner knows the population distribution of X, i.e., P(X), whose support is denoted by
A. For example, if the target list of voters is obtained from the voter roll, such covariates
may include age, gender, voting history, party registration, and zip code. Since the values
of these covariates are observable for every voter on the list, the distribution of X is known
to the planner.
Finally, following the statistical literature on causal inference, let Y
i
(t) represent the
potential turnout of voter i that will be realized if the planner applies the mobilization
technique t to this voter where i ∈ { and t ∈ T . Then, the observed turnout is given
by Y
i
= Y
i
(T
i
). The outcome variable is binary and is equal to 1 if voter i casts a ballot
and is equal to 0 if the voter abstains. For the sake of simplicity, any interference between
voters is ignored (Cox, 1958; Rubin, 1990). But, this assumption can easily be relaxed. If
a voter’s turnout decision depends on the treatment status of other voters within the same
household (Nickerson, 2008), for example, then an analysis needs to be conducted at the
household level.
Given this setup, the planner’s mobilization strategy is characterized by the function,
δ(, ) : (T A) →[0, 1], (5.1)
where the mobilization strategy δ(t, x) denotes the probability of receiving treatment t ∈ T
for a voter with X = x. Alternatively, δ(t, x) may represent the fraction of voters with
the observed covariate X = x who are contacted by the planner using the mobilization
method t. These two definitions become essentially identical when the number of voters is
large, but have different implications for the way the optimization is conducted. In either
case, voters with the same value of X are assumed to be exchangeable since the planner
110
does not have additional information to distinguish between them. Thus, the planner is
interested only in determining the value of the function δ(t, x) for each value of t ∈ T and
x ∈ A. Given this definition, the set of feasible mobilization strategies, ∆, is the collection
of functions δ(, ) that satisfy the complement property constraint,
∆ ≡

δ(, ) :
K−1
¸
t=0
δ(t, x) = 1 for every x ∈ A
¸
. (5.2)
A nonpartisan GOTV campaign planner’s goal is to derive the optimal mobilization
strategy to maximize turnout of the target electorate. The planner can achieve this by
deriving the strategy that maximizes the expected turnout given the observed covariate
information about X. Then, the planner’s objective function can be written as a function
of the mobilization strategy as well as the probability of a voter’s turnout given the values
of covariates and the actual mobilization strategy applied to the voter,
g(δ, ρ) ≡ E

N
¸
i=1
δ(t, X
i
)Y
i
(t)

X

= N
¸
x∈X
P(X = x)
K−1
¸
t=0
δ(t, x) ρ(t, x), (5.3)
where the turnout profile is denoted by ρ(, ) : (T A) →[0, 1] with ρ(t, x) ≡ Pr(Y (t) = 1 [
X = x). The turnout profile represents the turnout probability given the characteristics
of a voter and the mobilization strategy applied to this voter. Note that the function
ρ(, ) is unknown to the planner. If the turnout profile is known (and there is no budget
constraint), then the planner applies the mobilization technique t that yields the highest
value of the function ρ(t, x) given each voter’s covariate characteristics x. However, since
ρ(, ) is unknown in practice, the planner must make the decision under uncertainty by
learning about ρ(, ) from experimental data.
A typical voter mobilization method usually has a small effect on an individual’s turnout
probability relative to the individual’s baseline predisposition to vote. For this reason,
111
planners may cast the objective function of equation 5.3 in terms of treatment effect rather
than turnout profile. In this framework, the planner maximizes,
g(δ, ρ) = E

N
¸
i=1
δ(t, X
i
)(τ
i
(t) +Y
i
(0))

X

∝ N
¸
x∈X
P(X = x)
K−1
¸
t=1
δ(t, x) τ(t, x), (5.4)
where τ
i
(t) ≡ Y
i
(t) −Y
i
(0) is the treatment effect of mobilization strategy t on voter i and
τ(, ) : (T , A) → [−1, 1] with τ(t, x) ≡ Pr(Y (1) = 1 [ X = x) − Pr(Y (0) = 1 [ X = x).
This setup is mathematically equivalent to equation 5.3. The decision whether to maximize
turnout profile (equation 5.3) or treatment effect (equation 5.4) hinges on whether the
planner can elicit a prior more easily for the turnout profile or treatment effects (see
Section 5.3.2).
In practice, the planner cannot maximize the expected turnout without considering the
differing costs of various GOTV techniques. This concept is formalized by assuming that
the planner faces the following budget constraint,
K−1
¸
t=1
1

¸
x∈X
δ(t, x) = 0
¸
κ(t) +N
¸
x∈X
P(X = x)
K−1
¸
t=1
δ(t, x) ξ(t, x) ≤ C, (5.5)
where 1¦¦ is the indicator function and C is the fixed positive constant representing the
maximum cost allowed for the GOTV campaign. In this formulation, the planner needs to
consider two kinds of costs. The first is the fixed overhead cost denoted by the function,
κ() : T →[0, ∞) for each mobilization technique t. This cost is incurred so long as at least
one voter is assigned to the mobilization method. The second component is the cost per
voter, i.e., the cost of contacting each voter with covariate value x ∈ A, which is represented
by a known cost function ξ(, ) : (T A) → [0, ∞). Thus, the planner must determine
which mobilization technique is most cost-effective for different voters, and whether the
differences are large enough to warrant using multiple mobilization techniques. Finally,
112
since t = 0 corresponds to the status quo strategy of not mobilizing, both the overhead
and per voter costs are zero for this strategy, i.e., ξ(0, x) = κ(0) = 0 for all x ∈ A. This
is why the summation in equation 5.5 is taken with all possible values of t except t = 0.
For the other mobilization strategies, the cost per voter is assumed to be positive, i.e.,
ξ(t, x) > 0 for all x ∈ A and t > 0, whereas the overhead cost is non-negative, i.e., κ(t) ≥ 0
for all t > 0.
5.2.2 Data from a Randomized Field Experiment
Using the decision-theoretic framework, the planner can analyze the data from a random-
ized field experiment to derive the optimal GOTV campaign. Certain assumptions are
required to link a field experiment with a planner’s decision problem. First, the planner
must assume that the experiment is conducted on a representative sample of size n taken
from the target population {. Without such an assumption, the planner would be forced
to model the non-random sample selection mechanism to infer characteristics about { from
the experiment. The observed data is denoted by D = ¦
¯
Y
i
,
¯
T
i
,
¯
X
i
¦
n
i=1
where
¯
Y
i
is the binary
turnout variable,
¯
T
i
∈ T is the treatment variable representing mobilization techniques,
and
¯
X
i
∈ A is the same set of covariates used earlier. The potential outcomes are denoted
by
¯
Y
i
(t) where
¯
Y
i
=
¯
Y
i
(
¯
T
i
) for t ∈ T .
Another key assumption required for the planner to apply the results of a randomized
field experiment to the decision problem is that the joint distribution of potential outcomes
and covariates does not change, i.e., P(
¯
Y (t),
¯
X) = P(Y (t), X). This assumption may
be invalid if, for example, the election in which the experiment was conducted differs
significantly from the election for which the planner is designing the GOTV campaign.
Although in real world settings this assumption may hold only approximately, it is essential
for learning about the planner’s decision problem from a field experiment.
113
If these assumptions hold, the randomization of treatments in field experiments imply
that the turnout profile, ρ(, ), is identified, i.e., ρ(t, x) = P(
¯
Y (t) = 1 [
¯
X = x), for
all t ∈ T and x ∈ A. Although such large sample identification results are important,
in practice the planner must estimate ρ(, ) from a finite sample and make the decision
under uncertainty. (Similarly, the planner can identify τ(t, x) but must estimate it from a
finite sample.) This problem can be addressed by deriving the optimal nonpartisan GOTV
campaign strategy in this setting, where the planner is assumed to be Bayesian.
5.2.3 The Bayesian Planner
The framework of this method assumes that the planner is Bayesian and has a prior belief
on the space of functions of ρ(, ). This prior distribution is denoted by π(ρ). The Bayesian
planner will update her belief via Bayes rule after observing the data from the randomized
field experiment. This posterior belief about the turnout profile is represented by π(ρ [ D).
In the Bayesian statistical decision framework (Berger, 1985), the optimal nonpartisan
GOTV strategy δ

maximizes the posterior mean of the expected turnout,
δ

= argmax
δ∈∆

g(δ, ρ) dπ(ρ [ D), (5.6)
where the optimization is subject to the budget constraint given in equation 5.5.
There are several reasons why the Bayesian optimality criteria are used rather than
a frequentist approach based on maximin or minimax-regret criteria, which is popular in
the treatment choice literature. First, the Bayesian decision has a frequentist justification.
To see this, consider an alternative class of GOTV mobilization strategies that depends
directly on experimental data as well as the values of observed covariates. Such strategies
are called “statistical treatment rules” and are characterized by the function δ(, , ) :
(T A T) →[0, 1] (Manski, 2005). Thus, the set of feasible strategies ∆ equals the set
114
of all functions δ(, , ) that satisfy
¸
K−1
t=0
δ(t, x, D) = 1 for all (x, D) ∈ (A T). In this
setting, the frequentist objective function (i.e., risk) is given by,
g(δ, F
D
, ρ) =

D
N
¸
x∈X
P(X = x)
K−1
¸
t=0
δ(t, x, D) ρ(t, x) dF
D
(5.7)
with the following budget constraint,
K−1
¸
t=1
1

¸
x∈X
δ(t, x, D) = 0
¸
κ(t) +N
¸
x∈X
P(X = x)
K−1
¸
t=1
δ(t, x, D)ξ(t, x) ≤ C, (5.8)
for all D ∈ T. Again, the costs for the status quo strategy t = 0 are zero and are excluded
from the budget constraint. It has been shown that in most practically relevant situations,
the Bayesian decision δ

defined in equation 5.6 agrees with the decision that maximizes the
expected value of g(δ, F
D
, ρ) averaging over the prior distribution of D on T (see Berger,
1985, p. 159; Manski, 2005, p. 59). Also, if the size of the experimental data is large
(as in many GOTV randomized field experiments) and little prior information is available,
the Bayesian decision is essentially equivalent to the strategy that maximizes the expected
turnout. Thus, the Bayesian decision can be justified from a frequentist perspective.
Second, an alternative optimality criterion is Wald’s (1950) minimax regret principle
(see Savage, 1951; Manski, 2005). One important advantage of the minimax regret criterion
is that it avoids the subjectivity of Bayesian optimality because it does not require the use
of prior information. On the other hand, unlike Bayesian decision theory, frequentist
theory based on the minimax regret criterion typically does not lead to the unique optimal
decision, which practitioners may find problematic. In addition, the strategies that meet
the minimax regret criterion can include no-data rules, which do not depend on the data
at all (Stoye, 2009). Such strategies do not allow the planner to learn anything from the
available experimental data. Furthermore, a minimax regret rule can be viewed as a Bayes
115
rule with a prior (i.e., a least favorable prior) distribution (Berger, 1985, Chapter 5). Thus,
depending on the planner’s subjective belief, the Bayesian GOTV strategy can meet the
minimax regret criterion.
5.2.4 Bayesian Optimal Campaign Planning at A Glance
Figure 5.1 depicts the process by which a planner arrives at the optimal strategy via the
proposed Bayesian decision theoretic framework. The planner must determine the costs
of each mobilization strategy (both overhead and per voter) as well as the prior belief
about their effects on voters with different characteristics. Since the cost function inputs
are often exogenously determined (e.g., the cost of postage and phone calls), the planner’s
only meaningful decision might be determining a prior belief. In many cases, the planner
might use a diffuse prior centered around a belief that there is no a priori difference in
effects of a mobilization method across different voters. This is especially appropriate if
the mobilization technique has not been empirically tested (e.g., airplanes with reminder-
to-vote advertisements). If the treatment is an oft-used mobilization technique that has
been extensively studied in the past (e.g., canvassing), then the planner might center a
prior around the estimated effects in previous experiments. Note that the influence of
prior belief diminishes as the size of experimental data increases.
Once a prior belief is elicited and a randomized field experiment is conducted, these two
sources of information are combined via Bayes rule obtaining the posterior belief about the
effects of mobilization strategies on different voters. The planner now has an updated belief
about the most cost-effective way to mobilize each voter. Taking into account overhead
costs, the optimal strategy may be to implement a subset of the available mobilization
techniques, even if every technique is marginally optimal for at least one voter.
116
Optimize
Overhead
κ
Costs
ξ
π(ρ)
Prior Belief
Cost Function Input Values
Per Voter
Bayes Rule
Posterior
π(ρ | D)
Experimental
Data
Planner’s Inputs:
Optimal
Campaign
δ

D
Belief
Budget
Constraint
Turnout
Expected
Figure 5.1: An Overview of the Bayesian Optimal Campaign Planning Process. Inputs over
which the planner has direct control are represented by hexagons and are: (1) the prior
belief about the effects of various mobilization strategies on different voters, π(ρ), (2) the
overhead costs of each mobilization method, κ, and (3) the cost per voter for each strategy,
ξ. Data from randomized field experiments, D, are represented by the oval. These data and
the planner’s prior distribution are combined via Bayes rule to produce a posterior belief
about the effects of mobilization strategies π(ρ [ D). Finally, the proposed optimization
method uses this posterior belief and the exogenous costs, ξ, to find the optimal campaign
strategy, δ

, for the planner.
5.3 The Optimal Nonpartisan Campaign Strategy
Using the decision-theoretic framework, the optimal GOTV campaign strategy, as defined
in equation 5.6, can be derived. One relative advantage of the proposed Bayesian framework
over a frequentist’s approach is that the planner can completely separate the derivation
of the optimal campaign strategy from the analysis of experimental data. That is, the
planner first uses statistical models to obtain the posterior belief of the turnout profile,
ρ(, ). Conditional on this posterior turnout profile, the planner determines the optimal
campaign strategy by solving an optimization problem subject to a budget constraint.
117
5.3.1 The Optimization Method
Before describing the method to obtain the posterior turnout profile, this section describes
how to obtain the optimal nonpartisan GOTV campaign strategy given the posterior
turnout profile. Let ˜ ρ(t, x) be the posterior turnout profile for each t ∈ T and x ∈ A.
Then, the optimal campaign strategy can be obtained by solving the constrained optimiza-
tion problem,
δ

= argmax
δ∈∆
N
¸
x∈X
P(X = x)
K−1
¸
t=0
δ(t, x) ˜ ρ(t, x), (5.9)
subject to the budget constraint given in equation 5.5.
To solve this optimization problem, first consider the case of no budget constraint. In
this case, the solution is trivial because the most effective strategy for each stratum defined
by X is used. That is, the optimal campaign strategy is given by,
δ

(t, x) =

1 if t = argmax
s∈T
˜ ρ(s, x),
0 otherwise,
(5.10)
for any given x ∈ A. However, in many cases, the budget constraint may prevent the
planner from employing campaign strategy δ

.
Next, consider the case where the budget constraint is binding so that the strategy
defined by equation 5.10 is not feasible and yet there is no overhead cost. Unfortunately,
in this case, the derivation of the optimal strategy is no longer trivial. Thus, as a general
118
strategy, the planner solves the constrained linear optimization problem,
δ

= argmax
δ∈∆
N
¸
x∈X
P(X = x)
K−1
¸
t=0
δ(t, x) ˜ ρ(t, x), (5.11)
subject to

δ(t, x) ≥ 0 for all x ∈ A and all t ∈ T ,
¸
K−1
t=0
δ(t, x) = 1 for all x ∈ A,
N
¸
x∈X
P(X = x)
¸
K−1
t=1
δ(t, x) ξ(t, x) ≤ C.
(5.12)
The standard linear programming algorithm can then be used to obtain the optimal strat-
egy, δ

. If δ(t, x) represents the proportion of voters rather than the probability of treat-
ment assignment, then this can be formulated as a mixed integer programming problem,
which is more difficult but can be solved by applying an appropriate algorithm. Moreover,
as the sample size increases, this difference vanishes.
Finally, when the mobilization methods involve overhead costs, consider a case where
only a subset of mobilization techniques is applied to at least one voter. In this case, the
planner can solve the optimization problem in the same manner as in the case without
overhead costs except that the total overhead costs are subtracted from the maximum
budget allowed for the campaign, i.e., C −
¸
K−1
t=1
1
¸¸
x∈X
δ(t, x) = 0
¸
κ(t). The optimal
strategy is obtained by considering all possible subsets and their corresponding optimal
strategy, and then choosing the strategy that yields the greatest posterior expected turnout.
Although the proposed general strategy based on linear programming is easy to im-
plement, in some cases an approximate method, which is more computationally efficient,
may be preferred. Appendix D.1.1 details a fast approximate solution to the nonpartisan
planner’s optimization problem, which is used in the nonpartisan application presented in
Section 5.5.
3
3
The strategy obtained with this approximation is still labeled “optimal” for ease of language.
119
5.3.2 The Statistical Method
Numerous statistical models can be used to obtain the posterior distributions of the turnout
profile. When choosing a model, certain statistical issues need to be addressed. First, a
statistical model needs to be flexible to avoid strong functional form assumptions. In
particular, a nonparametric model is useful because a planner does not know the exact
functional form of the turnout profile a priori. Second, since many of the covariates avail-
able in GOTV field experiments are discrete (e.g., party registration, previous turnout
history), a statistical model should be able to accommodate discrete covariates. This ex-
cludes the use of some binary nonparametric regression models that require covariates to
be continuous. Third, the model fitting procedure should require a minimal number of
arbitrary decisions from a planner (or the data analyst who is working for the planner).
Finally, a model should not be overfitted to the data at hand and thus an appropriate set
of pre-treatment covariates must be carefully selected. Overfitting tends to yield a model
with poor performance in the actual election to which the derived mobilization strategy
will be applied. This is the main problem of the post-hoc subgroup analysis described
in Section 5.1. In practice, this consideration is important because the sample size may
not be large enough to accommodate a high number of pre-treatment covariates, which
are potentially useful for deriving the optimal campaign strategy. Thus, an appropriate
variable selection algorithm is needed as a part of the model selection procedure.
Moreover, as Gunter, Zhu, and Murphy (2007) point out, most variable selection algo-
rithms are developed for improving prediction rather than decision making. These concepts
are closely related, but are not the same. For example, a powerful predictor of the outcome,
i.e., a predictive variable, is not necessarily useful for decision making if its effect on the
outcome is constant between the treatment and control groups. Similarly, a variable that
explains a significant portion of treatment effect heterogeneity, i.e., a prescriptive variable,
120
may not be selected by standard variable selection procedures if it does not predict the
observable outcome (rather than the potential outcome) as well as other variables.
The proposed algorithm meets these criteria relatively well. The method is similar to
that proposed by Gunter et al. (2007) but differs from their algorithm in that it uses a
tree-based method (Breiman et al., 1984) rather than Lasso (in part because the outcome
variable in political applications is categorical) and the measure of importance for pre-
scriptive variables is somewhat different. It is also possible to use Bayesian regression tree
models (Chipman et al., 2008; Hill and McCulloch, 2008).
A Bayesian approach is used to model the turnout among voters with the same char-
acteristics of covariates via the binomial distribution. For the moment, assume that the
sample size is sufficiently large and thus there is no need for variable selection. Using a
conjugate prior, this turnout model is,
W
tx
[ T = t, X = x ∼ Binom(n
tx
, ρ(t, x)), (5.13)
ρ(t, x) [ X = x ∼ Beta(a
tx
, b
tx
), (5.14)
for each t ∈ T and x ∈ A where W
tx
is the number of voters with T
i
= t and X
i
= x who
turned out, n
tx
is the total number of such voters, and (a
tx
, b
tx
) are the prior parameters.
This model yields the familiar posterior distribution,
ρ(t, x) [ Y
i
, X
i
= x ∼ Beta(W
tx
+a
tx
, n
tx
−W
tx
+b
tx
), (5.15)
where the posterior mean of ρ(t, x) is given by (W
tx
+a
tx
)/(n
tx
+a
tx
+b
tx
).
4
Although its simplicity is attractive, this model is unlikely to work well in practice if
the sample size is small relative to the number of unique values the observable covariates
4
Alternatively, the planner may formalize a prior belief in terms of treatment effects. If a normal prior
distribution is used, then the posterior of τ(t, x) can also be approximated by a normal distribution.
121
X take. In particular, if one conditions upon irrelevant covariates, then the sample size
within each subgroup will be too small to yield informative inferences about ρ(t, x). Such
overfitting will then necessarily lead to a mobilization strategy that will perform poorly
in the actual election. On the other hand, if important covariates are not used to define
subgroups, the planner will fail to differentiate across voters and will choose a suboptimal
campaign strategy. Thus, in most practical cases, we seek a principled way to select relevant
variables and form subgroups before applying the above standard Bayesian model.
The proposed solution to this problem involves three steps. First, a variable selection
algorithm is applied to decide which variable needs to be conditioned upon when deriving
the optimal campaign strategy. Next, given the selected variables, a tree-based classifica-
tion method (Breiman et al., 1984) is fitted to each treatment/control group to identify
relevant subgroups within the group.
5
The data is cross-validated to avoid overfitting.
Finally, once all subgroups are identified in this way, the Binomial-Beta model is applied
within each subgroup to obtain the posterior distribution of ρ(t, x) for all t and x. This ap-
proach is labeled semi-Bayesian because the data are used twice—once to form subgroups
and again to calculate the posterior distribution. The details of the proposed method are:
Step 1: (Selection of Predictive Variables) Fit a classification tree to the entire sam-
ple using all pre-treatment covariates and the treatment variable. Use K-fold cross
validation on the misclassification rate to determine the value of the parameter that
controls the complexity of the tree, e.g., the complexity parameter in rpart() imple-
mentation in R (Ripley, 1996, Chapter 7). Denote the predictive variables that are
used in the final model by V , i.e., V ⊂ X.
Step 2: (Importance of Prescriptive Variables) Order each pre-treatment covariate, X
j
for j = 1, 2, . . . , J, based on the statistic, S
j
≡ g

j1
− g

j0
, where g

jk
is the optimal
overall turnout using the turnout profile ˆ ρ
jk
(t, x) for k = 0, 1. Obtain ˆ ρ
j1
(t, x) by
fitting a classification tree for the treated subset of the data (i.e.,
¯
T
i
≥ 1) and using V
and X
j
as covariates. Similarly, ˆ ρ
j0
(t, x) is obtained by fitting a classification tree on
5
This tree-based classification method is one of many nonparametric models. One disadvantage of
this approach is that gradual changes in treatment effects across covariate groups are modeled as sharp
discontinuities rather than smooth functionals.
122
the untreated subset of the data (i.e.,
¯
T
i
= 0). The value of the complexity parameter
from Step 1 is also used to fit the trees in Step 2.
Step 3: (Model Fitting) For each j = 1, . . . , J with S
j
> 0:
(a) Randomly divide the sample into K subsets for K-fold cross validation.
(b) Using K − 1 training sets, fit classification trees (without pruning) separately
to the treatment and control groups using V as well as the j most important
prescriptive variables in both models. Select the values of the complexity pa-
rameters for the two models based on the mean of the optimal overall turnout
across K validation sets. Denote the optimal overall turnout and the optimal
campaign strategy corresponding to the selected values of complexity parame-
ters by g

j
and δ

j
, respectively.
Step 4: (Derivation of Optimal Strategy) The optimal overall turnout is given by g

=
max
1≤j≤J
g

j
, whereas the optimal campaign strategy is given by δ

= δ

argmax
1≤j≤J
g

j
.
The first step of the algorithm selects predictive variables using a standard fitting
procedure of tree-based methods. The second step orders each pre-treatment covariate
(including those identified as predictive variables in Step 1) according to its importance as
a prescriptive variable. The statistic, S
j
, is designed to measure how much a planner can
increase the optimal overall turnout by interacting the value of X
j
with the treatment. This
statistic provides a measure of the ability of X
j
to explain heterogeneous treatment effects.
The third step uses the K-fold cross-validation procedure, given all predictive variables
and different subsets of prescriptive variables, in order to select the values of complexity
parameters for classification trees fitted separately to the treatment and control groups.
This is done by comparing the optimal overall turnout corresponding to different values of
the complexity parameters. Finally, Step 4 selects the final model among the ones chosen
in Step 3 by again comparing the resulting optimal overall turnout and thus determines
the optimal campaign strategy.
A main advantage of this semi-Bayesian approach is that it inherits the simplicity of
tree-based methods. In particular, practitioners can interpret each of the subgroups that
123
result from the final model. They can thus use available prior information within subgroups
by specifying the parameters of the beta prior distribution. The proposed approach also
addresses three key issues highlighted earlier. First, the tree-based classification models
are nonparametric and can handle discrete covariates effectively. Second, the use of cross
validation procedure avoids overfitting. Third, transparent algorithms such as the one
proposed here prevent planners from making arbitrary decisions when deriving the optimal
campaign strategy.
5.4 The Optimal Partisan Campaign Strategy
The proposed decision theory framework and the statistical and optimization methods can
be applied to the case of partisan persuasion and GOTV campaign planning. A Bayesian
planner can derive the optimal campaign strategy using randomized field experiments to
maximize the (posterior) expected chance of winning the election. This case assumes that
two major candidates are competing for the office. (Minor party candidates may exist, but
their probability of winning the election is assumed to be negligible.)
5.4.1 The Decision Problem
Using the notation introduced in Section 5.2, the decision problem of the partisan campaign
planner is to assign one of K different mobilization methods (including the status quo
strategy of doing nothing, which is denoted by T
i
= 0) to each member of the target
population { of finite size N. Again, the planner is assumed to know the distribution
of a certain set of covariates P(X). Thus, the planner’s mobilization strategy can be
characterized by δ(, ) (see equation 5.1) and the set of feasible strategies is equal to ∆
(defined in equation 5.2).
124
Unlike a nonpartisan campaign planner, a partisan campaign planner seeks the mobi-
lization strategy that will lead to electoral victory. For this decision problem, the outcome
variable Y
i
needs to be redefined. Let Y
i
(t) represent the potential voting behavior of voter
i that will be realized if the planner assigns mobilization method t to this voter where i ∈ {
and t ∈ T . The variable Y
i
(t) can take three different values; it equals 1 if voter i casts
a ballot for the candidate of the planner’s party, −1 if she votes for the opponent, and
0 otherwise (e.g., she votes for a third party candidate or abstains). Then, the planner’s
ultimate goal is to win the election, which can be represented as the indicator function,
h(δ, V ) ≡ 1

n
¸
i=1
δ(t, X
i
)Y
i
(t) > 0
¸
= 1

¸
x∈X
P(X = x)
K−1
¸
t=0
δ(t, x)ν(t, x) > 0
¸
, (5.16)
where ν(t, x) ≡
¸
i∈{i

:X
i
=x}
Y
i
(t)/
¸
N
i=1
1¦X
i
= x¦ is a random variable representing the
vote share differential for the candidate that will result among voters with covariates X = x
if the planner assigns mobilization method t to them. Clearly, h(δ, V ) is equal to 1 if the
candidate of the planner’s party wins the election, and 0 if he loses. In the statistical
decision theory literature, such an objective function is called “0 −1 loss function.”
Finally, the partisan planner typically faces a budget constraint similar to that con-
fronted by the nonpartisan planner, and therefore equation 5.5 also applies to the partisan
planner’s situation.
5.4.2 Data Requirements
As is the case for a nonpartisan GOTV campaign planner, certain assumptions are required
for a partisan campaign planner to be able to use randomized field experiments to reach
the optimal decision. These assumptions are essentially identical to those described in
Section 5.2.2: (1) a field experiment is conducted on a representative sample from the
125
same target population of voters {, and (2) the joint distribution of potential outcomes and
covariates, P(Y (t), X), remains identical between the experiment and the actual election.
However, one important difference is that the derivation of the optimal partisan campaign
requires vote choice data as well as turnout data for the voters who are the subjects of field
experiments; recall that in the case of a partisan campaign, Y (t) represents a trichotomous
variable rather than a binary variable. For example, in the United States, vote choice
data, unlike turnout data, are not publicly available and cannot be verified for each voter.
This means that a sample survey needs to be conducted to derive the optimal partisan
mobilization strategy (unless the entire analysis and strategy planning are conducted at
an aggregate level where validated election results are available).
5.4.3 Derivation of the Optimal Strategy
The optimal partisan campaign is derived via Bayes theorem strategy, as in the case of a
nonpartisan GOTV campaign. In particular, the optimal strategy maximizes the posterior
probability of winning the election,
δ

= argmax
δ∈∆

h(δ, ν) dπ(ν [ D), (5.17)
subject to the budget constraint given in equation 5.5 where π(ν [ D) is the posterior dis-
tribution of the vote share differential ν(, ). Using the classification method and variable
selection algorithm similar to those described in Section 5.3.2, the planner can estimate
ν(, ), except that the outcome variable is now trichotomous rather than binary. Most
classifiers including tree-based methods can handle such categorical variables even when
the number of categories is greater than two.
The problem, however, is that the optimization in equation 5.17 is not trivial for two
reasons. First, the integration cannot be explicitly solved. Second, the objective function
126
is an indicator function that is not continuous. These difficulties are often amplified by
the reality that the optimization must be conducted over a high-dimensional space if the
number of treatments and/or the number of subclasses is large. These computational con-
siderations prevented Kramer (1966) from using the probability of winning as the objective
function of a partisan campaign planner. Instead, he used the expected plurality of votes as
the objective function while acknowledging that it may not be appropriate. Kramer (1966)
noted that “the probabilistic objective is the more realistic. However, this formulation is
computationally quite difficult to work with” (p. 141). Indeed, to maximize the expected
plurality, the partisan planner can solve the following constrained optimization problem by
applying the standard linear programming algorithm, as explained in Section 5.3.1,
δ

= argmax
δ∈∆
N
¸
x∈X
P(X = x)
K−1
¸
t=0
δ(t, x)E(ν(t, x) [ D), (5.18)
subject to

δ(t, x) ≥ 0 for all x ∈ A and all t ∈ T ,
¸
K−1
t=0
δ(t, x) = 1 for all x ∈ A,
N
¸
x∈X
P(X = x)
¸
K−1
t=1
δ(t, x) ξ(t, x) ≤ C.
(5.19)
To overcome this computational difficulty, a fast and approximate solution to the
partisan planner’s optimization problem can be applied. This method, detailed in Ap-
pendix D.1.2, is not guaranteed to yield an optimal campaign strategy, but is relatively
fast and closely approximates an optimal strategy.
5.5 Empirical Evaluation of the Proposed Method
To asses how effective the proposed method of deriving the optimal GOTV campaign is
in real world applications, the method is applied to three data sets of randomized field
experiments. The intent is to use a randomly selected subset of the data as a test data set
127
and to obtain an unbiased estimate of the actual turnout (or the probability of winning) by
applying the resulting optimal campaign strategy derived from the rest of the data to this
test data set. This procedure mimics the real world situation by using the test data set as
the actual election to which the optimal campaign strategy is applied. Since the treatment
is randomized and the test data set is not used to derive the optimal strategy, the procedure
results in an unbiased evaluation of the empirical performance of the proposed methodology.
The three applications demonstrate the power and effectiveness of this method. All
examples consist of a single treatment; for more varied applications see (Imai and Strauss,
2009). In each case, the decision for the planner is which voters to treat with the appeal.
The first application is a nonpartisan text messaging GOTV experiment from the 2006
Congressional election. The second example is a randomized mail experiment from the 2007
Kentucky gubernatorial campaign. One of three persuasive mailings was sent to potential
voters; two of the mailings used a “control group” to analyze the heterogeneous effects of
the third mailing, which was on education. The third application is a program of anti-
McCain robo-calls and mailings that focused on the economy, which were sent to potential
voters in 2008. A randomized control group did not receive any of the treatments. In
both partisan examples (the 2007 and 2008 experiments), the outcome measure (candidate
support) was gathered via a survey following treatment delivery.
The partisan applications demonstrate the powerful implications of the Personal Expe-
rience model in the real world. As detailed in Sections 5.5.3 and 5.5.4, parents of school-
aged children were more likely to re-evaluate the candidates once they learned their stance
on education and individuals most likely affected by the economy were more likely to re-
evaluate McCain. In addition to using the nonparametric method, these results are verified
using more standard generalized linear models. The findings show strong support for the
128
Candidate Evaluation Hypothesis presented in Chapter 2. Matching voters to issues for
which they have experience increased the candidate’s margin by 10-15 percentage points.
5.5.1 Evaluation Method
To assess the effectiveness of the proposed method, an additional level of cross-validation
is added to the procedure described in Section 5.3. The aim is to cross-validate the whole
procedure (consisting of the three steps described in Section 5.3.2) and obtain an unbiased
estimate of the resulting turnout under the optimal strategy from test data that are not
used in derivation of the strategy. After randomly dividing the sample into L subsamples,
one subsample is set aside as a test set and the proposed methodology is applied to the
rest of the data. The derived optimal strategy is then applied to the test set to obtain an
unbiased estimate of the resulting overall turnout. The random assignment of treatments
and the random subsampling of the test set make the unbiased estimation possible. The
entire procedure is repeated L times using each subset as a test set. Finally, the average
value of the L estimated optimal turnout rates is taken as an estimate of the turnout that
would result under the proposed methodology.
Each application uses a normal-normal conjugate prior for the treatment effect τ based
on the setup defined in equation 5.4. The prior for each subgroup treatment effect distri-
bution is Gaussian and is centered on the estimated population average treatment effect.
The value of the prior variance is chosen so that it increases in proportion to the per
capita budget constraint. In each case, a grid search is implemented to approximate the
optimal complexity parameter. A 10-fold cross-validation procedure is used for Step 3 of
Section 5.3.2 to determine the optimality of each complexity parameter. In addition, the
optimization problem is solved using algorithms that yield approximate (but fast) solutions
(Appendices D.1.1 and D.1.2).
129
5.5.2 A Nonpartisan GOTV Campaign with a Single Mobilization Method
During the 2006 election, two nonpartisan organizations contributed the cell phone numbers
of newly registered individuals to an experiment that tested the efficacy of text messages
to mobilize voters (see Dale and Strauss, 2009, for details). The election was of moderate
interest, with at least one gubernatorial or senatorial campaign on the ballot in most states.
Subjects were included in the experiment when they registered to vote with a campus
representative of the Student PIRGs or when they registered online with Working Assets.
About 8,000 subjects nationwide were randomly assigned, with 50 percent probability, to
either the treatment group or control group.
The treatment group received a short text message the day before Election Day. An ex-
ample text message read “A friendly reminder that TOMORROW is Election Day. Democ-
racy depends on citizens like you-so please vote! -PIRG.” (The text message appeals were
varied slightly, but these differences are ignored for this analysis.) Subjects were matched
to the voter file using information on their registration forms.
The outcome variable is dichotomous: one for having voted in 2006, and zero for not
having voted. The estimated average treatment effect, or more precisely the overall intent-
to-treat effect, is 3.0 percentage points with a standard error of 1.1. Available covariates
for the analysis include gender, age, race, past voting history, log of county population
density, and registering organization.
Following the procedure outlined in Section 5.3, the optimal campaign strategy is de-
rived using this experimental data set. The classification tree produced in Step 1 chooses
the following variables as having predictive power (i.e., ν): age, log of county density, reg-
istering organization, whether the subject had voted in a previous election, and gender.
The prescriptive variables chosen in Step 2 are age, density, and Hispanic (in decreasing
order of S
j
). Not all these variables are included in the final classification tree produced,
130
however, as searching the complexity parameter space often finds that models with fewer
variables result in higher overall turnout on validation sets.
The resulting final trees are presented in Figure 5.2. The tree for the control group
is on the left; the tree for the treatment group is on the right. Left branches represent
voters who meet the criteria of the nodes; right branches represent voters who would falsify
the nodes’ inequalities. The leaves show the predicted probability of voting conditional on
their covariates; higher probabilities are to the right at each node. The control group tree
demonstrates that, in this group, voters above the age of 24 are predicted to vote at a rate of
62%; whether the participant’s age is known and county population density are important
determinants of the voting rates for voters aged 24 and under. In the treatment group,
the age cut-point is in between 19 and 20, with population density and voting history also
providing information about turnout rates.
131
|
age< 24.5
age.missing< 0.5
lg.dens>=6.043
lg.dens< 5.858
0.42
0.42 0.73
0.72
0.62
|
age< 19.5
lg.dens>=5.985
lg.dens< 5.852
0.38
0.42 0.79
0.66
Figure 5.2: Final Classification Trees for the Control Group (left panel) and Treatment
Group (right panel). The complexity parameters are chosen from 10-fold cross-validation
using the algorithm described in Section 5.3.2 so that the resulting optimal turnout is
maximized on the validation set. In this example, the planner’s budget allows treatment
of at most 10% of the population. At each node, subjects who meet the node’s criterion
are filtered through the left branch of the tree. Covariate abbreviations: age is the age in
years of the subject, age.missing is whether the age of the participant is unknown, and
lg.dens is the log of the subject’s county population density.
132
Much of the heterogeneous treatment effect can be explained by the voters’ ages. Po-
tential voters between the ages of 20 and 24 are very responsive to the treatment. In the
control group (left panel of Figure 5.2), these individuals vote at a rate of 62%. In contrast,
the classification tree for the treatment group assigns them a probability of voting of 66%
(the right branch of the right panel of Figure 5.2) – a four percentage point increase above
the average treatment effect. Also, the treatment model predicts a negligible or negative
treatment effect for 18- and 19-year-olds, as their probability of voting is assigned at most
42% under the treatment unless they live in a county with a density within a narrow
range. (Such age ranges could not be identified by a classic model such as logit with linear
explanatory variables.)
Figure 5.3 displays the performance of the proposed methodology given the maximum
proportion of voters that could be contacted. The turnout that would result under the
optimal strategy is estimated using the difference in means estimator between the treated
and untreated voters (solid lines with solid circles). The average turnout is computed for
the treated voters who are assigned to the treatment group as well as for the untreated
voters who are assigned to the control group under the optimal strategy. The latter is
then subtracted from the former to yield the estimated overall turnout under the optimal
strategy. The turnout rate achieved with the proposed method compares favorably with the
turnout rate achieved using a naive strategy where randomly selected voters are contacted.
This less-informed strategy, which is based solely on the estimated overall average treatment
effect (“ATE strategy”), completely ignores covariate information and thus assumes zero
treatment effect heterogeneity.
Figure 5.3 shows that the proposed method results in a higher overall turnout than an
ATE strategy (at least on average), if an organization can afford to treat only a subset
of the population. For a campaign that can afford to treat 10% of the population, for
133
Maximum Proportion of Voters Contacted
O
v
e
r
a
l
l

T
u
r
n
o
u
t

I
n
c
r
e
a
s
e
0% 20% 40% 60% 80% 100%
0
.
0
0
0
.
0
1
0
.
0
2
0
.
0
3
q
q
q
q
q
q
q
ATE Strategy
Optimal Strategy
Figure 5.3: Empirical Evaluation of the Performance of the Proposed Method for the Text
Messaging Experiment. The figure displays the overall turnout increase that results from
two campaign strategies as a function of the maximum proportion of voters contacted.
The first strategy is the average treatment effect or ATE strategy (solid line), which con-
tacts randomly selected voters. The second strategy is an optimal approach based on the
proposed methodology, which uses covariate characteristics of voters to determine which
voters receive the treatment. Solid circles represent the estimated optimal turnout using
the difference-in-means estimator. The estimator is applied to the test data which are not
used to devise of the optimal strategy.
134
example, the optimal campaign strategy achieves an overall turnout increase on average
over three times greater than the turnout increase under the ATE strategy. Because the
nonparametric procedure doe not isolate individuals who reacted negatively toward the
treatment, no gain over the ATE strategy is seen when treating the entire population.
This lack of negatively responsive individuals is not surprising given that little backlash was
found to the text messages in a post-treatment survey (Dale and Strauss, 2008). Overall, for
the text messaging application, organizations operating under tight budgetary constraints
would be wise to use the proposed nonparametric procedure.
5.5.3 Partisan Example: Parents and Education Spending
In September 2007, a union delivered three mailings to its members in Kentucky in a
randomized field experiment. Each participant received exactly one of the three mailings,
which focused on schools, change, and corruption, respectively, with no control group. The
mailings were sent out on September 13, and 1,321 participants were surveyed by telephone
between September 17 and September 20. As predicted by the Personal Experience Model,
the school mailing worked particularly well among parents.
The respondents to the survey were asked a standard vote question in which the parties
of the candidates, Democrat Steve Beshear and Republican Ernie Fletcher, were identified.
Respondents reported which candidate they currently supported, indicating whether their
support was “definite, probable, or lean.” Given that the respondents were union members,
it is not surprising that they generally preferred the Democrat: 69% supported Beshear
(including leaners), 17% supported Fletcher, and 13% were undecided. The results were
similar for the subset who received the school mailing: 70% Beshear, 15% Fletcher, and
15% undecided (n=448).
135
Part of the schools mailing is pictured in Figure 5.4. The Personal Experience Hypoth-
esis suggests that parents would be especially responsive to this piece because they are
able to independently judge politicians on the state of Kentucky schools. Unfortunately,
the survey of respondents did not ask whether the respondent was a parent. The survey
did ask about marital status, however, and a linked commercial database includes data on
children in the household. The 13% of respondents who have children in the household (as
indicated by the commercial data) was combined with the 18% of respondents who are not
single and are under 40 to compose the target group: the 26% of the population likely to
have children. This group was labeled (with some hesitation) “parents.”
Figure 5.4: Part of the Kentucky Mailing.
Parents who received the school mailing favored Beshear 75%-12% (n=97) as displayed
in Figure 5.5; parents who received other mailings favored Beshear 68%-22% (n=223). This
result is at the edge of statistical significance. A linear regression using the 7-point vote
preference, and controlling for the prior probability of supporting Beshear (the support
score demographic model), the school mailing*parent interaction coefficient is significant
with p=0.03, two-tailed. An ordinal regression with the trichotomous Beshear/Fletcher/undecided
outcome variable (i.e., less granularity) and no controls produced a p-score for the interac-
tion of 0.15, two-tailed. These regressions indicate that an independent parent increased
his/her probability of supporting Beshear by 11 percentage points after receiving the school
mailing (from 66% to 77%).
136
Mailing Appeal Type
B
e
s
h
e
a
r

M
a
r
g
i
n
Corruption & Change Schools
3
0
%
4
0
%
5
0
%
6
0
%
7
0
%
8
0
%
P
a
re
n
ts
Non−parents
Figure 5.5: Parents React Strongly to School Mailing. Beshear’s margin by type of mail-
ing and whether the recipient is believed to be a parent are shown. Point estimates are
represented by filled-in squares and triangles. Dashed lines represent one standard error;
dotted lines represent two standard errors.
The population of the experiment, union members, is not representative of the Kentucky
electorate. To maximize the mailing’s effect on Beshear’s margin of victory, Beshear is
assumed to have a baseline margin of -3 percentage points. The mailing has an average
treatment effect of 3.9 percentage points, thus treating 77% of union members (under the
ATE approach) would push Beshear over the 50% change of winning mark.
Figure 5.6 displays the significant cost savings provided by the nonparametric algo-
rithm. Beshear’s probability of winning passes the coin-flip mark with just 10% of the
population treated. With such a small sample, this 10% figure has a lot of noise; the fact
that treating 20% of the population decreases the chance of winning is an indication of this
noise. Nevertheless, the algorithm, which selects both the commercial data on parents and
the imputed data in various iterations, provides clear benefits.
137
q
q
q
q
q
q
q
0.0 0.2 0.4 0.6 0.8 1.0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
Maximum Proportion of Voters Contacted
P
r
o
b
a
b
i
l
i
t
y

o
f

B
e
s
h
e
a
r

W
i
n
n
i
n
g
Optimal Strategy
ATE Strategy
0.0 0.2 0.4 0.6 0.8 1.0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
Maximum Proportion of Voters Contacted
A
c
t
u
a
l

P
r
o
p
o
r
t
i
o
n

o
f

V
o
t
e
r
s

C
o
n
t
a
c
t
e
d
Optimal
Strategy
ATE Strategy
Figure 5.6: Empirical evaluation of the performance of the nonparametic method on the
Kentucky data. The results are calculated based on 10-fold cross-validation. The left panel
displays the estimated probability of Beshear winning under two treatment strategies plot-
ted over the maximum proportion of the electorate treatable under the budget constraint.
The dashed lines represent the “ATE strategy” in which random voters are contacted. The
solid lines represent the optimal strategy based on the proposed methodology, which uses
covariate characteristics of voters to determine which voters receive the treatment. The
optimal strategy outperforms the ATE approach strategy under tight budget constraints.
The right panel plots the actual proportion of voters contacted by the mailing against the
maximum proportion of voters contacted, which is determined by budget constraint. Only
in the situations where nearly all voters can be contacted, does the algorithm choose not
to canvass some voters—and it chooses these untargeted voters poorly.
Certainly there are alternative explanations of why parents are more responsive. First,
parents may be more likely to read the mailing because its subject is schools. This mech-
anism, however, is consistent with the Personal Experience Model: individuals might pay
more attention to a mailing that they realize will inform their political judgments, and they
are more capable of processing the political arguments of the mailing. Personal experience
with political issues is expected to strengthen both these processes.
138
A second possible explanation is that parents have a self-interest in school funding.
As with the observational data of Chapter 3, this experiment does not allow these two
mechanisms to be disentangled.
A third explanation is that the mailing convinces a voter to support Beshear solely on
the basis of her support of education spending. Since parents might be more inclined favor
of education spending, they may increase their vote for Beshear more than the general pop-
ulation. Two factors mitigate this alternative hypothesis. First, this explanation assumes
that voters do not cue-take from politicians; the evidence for cue-taking in Chapters 3 and
4 is strong. Second, Beshear’s margin does not increase in the non-parents group, meaning
any votes gained among non-parents who support education funding must be offset by non-
parents moving against Bashear. Certainly movement against the candidate delivering a
mail piece is possible, but the algorithm does not find large shifts against Bashear (Figure
5.6).
5.5.4 Partisan Example: Disadvantaged Voters and the Economy
In March 2008, while the Democratic presidential primary race was still undecided, a union
tested three anti-McCain mail pieces (along with a control group) in Ohio. The focus of the
mailings, with varying degrees of emphasis, was how McCain’s policies would damage the
economy and weaken the working- and middle-class. Figure 5.7 displays part of the most
straightforward of the mailings, which were mailed to union members. These mailings
were followed up with robocalls that reinforced the topic of the mailing. Participants
were surveyed by telephone shortly after the mail pieces and robocalls were received. The
Democrat’s margin was 23 points (Dem 54% - McCain 31%, n=2,967) among those who
received any mailing and 21 points (Dem 53% - McCain 32%, n=987) in the control group—
not a significant difference.
139
Figure 5.7: Part of the ant-McCain mailing.
The Personal Experience Hypothesis would predict that members who had personally
experienced the economic downturn (or were close to those who had) would react more
strongly to the mail and robocall program. The subgroup of interest is the set of individuals
who live in economically distressed areas, defined by several census characteristics, such as
percent unemployed and median household income. A factor analysis identifies the top 30%
of respondents who live in disadvantaged areas (after individuals who could not be matched
to census data were eliminated).
6
The median household income of the disadvantaged group
is about $32,500, which was significantly lower than the average of nearly $50,000 in the
advantaged group.
This key group’s support for the Democrat increased significantly after receiving the
mailings and robocalls: from 16 points in the control group (Dem 48%, McCain 32%,
n=227) to 28 points among anyone who received a mailing (Dem 58%, McCain 30%,
n=710). Among those who lived in the 70% most well-off places, this margin decreased
sightly between treatment and control, from 18 points to 16 points (Figure 5.8).
These data are suspicious because of the low Democratic margin among the disadvan-
taged in the control group. With only 227 disadvantaged control-group respondents, poor
randomization may be affecting the results. To reduce this possibility, vote preference is
regressed on the treatment condition and its interaction with the subgroup of interest, con-
6
Special thanks to Catalist, LLC for its help in matching voters to census information.
140
trolling for a demographic-based partisanship score provided by the union.
7
With a linear
regression on the 7-point vote question, the treatment effect for the advantaged subgroup is
nearly exactly zero, but for the disadvantaged subgroup the coefficient is large and nearly
statistically significant (p=0.15, two-tailed). The interaction coefficient can be interpreted
as: the average union member living in a disadvantaged area moved one-third of a survey
response unit (e.g., from “Dem: lean” to “Dem: probably”) after receiving the mailing.
An ordinal regression on the trichotomous outcome variable (i.e., the only options are
voting for the Democrat, McCain, or undecided) indicates that a member with median
characteristics would increase his/her probability of voting for the Democrat by 7 percent-
age points as a result of living in a disadvantaged area and receiving the treatment (from
38% to 45%).
When the nonparametric algorithm is applied to this data, both small- and large-budget
operations benefit. As Figure 5.9 displays on the left panel, treating 10% of the population
increases the Democrat’s probability of winning to 38%, while the ATE strategy produces a
win only 24% of the time.
8
In nearly every iteration, the algorithm selected the continuous
variable of economic distress.
Big gains are also found under loose budget constraints. The ATE is 2.4 percentage
points, yet treating 80% of the population under the proposed strategy would produce
an effect of 4.5 percentage points. A backlash is present among some of the population.
The algorithm, when the budget constraint is removed, estimates that 10% of the popula-
tion is negatively treatment responsive (Figure 5.9, right panel), but that is probably an
underestimate since the overall persuasion boost declines to 3.5 percentage points in this
case.
7
This score was developed before the experiment.
8
Again, the margin between the candidates is artificially tightened to examine differences between the
algorithms.
141
Figure 2: Members in Disadvantaged Neighborhoods
React Strongly to Economic Mailings
Treatment Condition
G
e
n
e
r
i
c

D
e
m
'
s

M
a
r
g
i
n

o
v
e
r

M
c
C
a
i
n
Control Any Mailing
0
%
1
0
%
2
0
%
3
0
%
D
is
a
d
v
a
n
ta
g
e
d
Advantaged
Figure 5.8: The Economically Disadvantaged React Strongly to Economy-focused Mailings.
Generic Democrat’s margin over McCain by treatment condition and whether the union
member lives in a disadvantaged area are shown. Point estimates are represented by filled-
in squares and triangles. Dashed lines represent one standard error; dotted lines represent
two standard errors.
As with the Kentucky test, alternative hypotheses are possible. Especially concerning
with this test is that disadvantaged members would react strongly to a variety of issues, not
just the economy. Unfortunately, this proposition cannot be tested in a rigorous manner
since there was no group that received a generic message. The only evidence offered is that
in a mail test later in 2008 by the same union dealing with Social Security, the economically
disadvantaged were less likely to react positively to the treatment than the economically
advantaged.
9
Also, as with the Kentucky test, prior issue position (in this case, belief that
the economy is performing badly) might generate the difference between the two groups.
9
This result is not statistically significant at conventional levels. Further, the Social Security mailing
cannot be used as a test of the Personal Experience Hypothesis because the mailing was delivered to seniors
only. Thus the experiment lacks crucial variance in the interacting variable of interest (seniors v. younger
voters and Social Security appeals).
142
q
q
q
q
q
q
q
0.0 0.2 0.4 0.6 0.8 1.0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
Maximum Proportion of Voters Contacted
P
r
o
b
a
b
i
l
i
t
y

o
f

O
b
a
m
a

W
i
n
n
i
n
g
Optimal Strategy
ATE Strategy
0.0 0.2 0.4 0.6 0.8 1.0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
Maximum Proportion of Voters Contacted
A
c
t
u
a
l

P
r
o
p
o
r
t
i
o
n

o
f

V
o
t
e
r
s

C
o
n
t
a
c
t
e
d
Optimal
Strategy
ATE Strategy
Figure 5.9: Empirical evaluation of the performance of the nonparametic method on the
anti-McCain experiment data. The results are calculated based on 10-fold cross-validation.
The left panel displays the estimated probability of the generic Democratic winning under
two treatment strategies plotted over the maximum proportion of the electorate treat-
able under the budget constraint. The algorithm successfully finds voters who are very
treatment responsive and negatively responsive. See Figure 5.6 for details.
Both partisan experiments demonstrate large effects of targeting participants with per-
sonal experience. Targeting parents with a school mailing rather than a generic mailing
increases the probability that a voter in this group shifts her preferred candidate by 11
percentage points. Targeting disadvantaged voters with an economic mailing increases
the probability of vote change by 7 percentage points. These shifts might appear large,
but often experiments with small average treatment effects significant subgroup differences
(e.g., Arceneaux and Kolodny, 2009). Also, the union clearly identified themselves as the
source of the mailing, and among members the union is likely a trusted source. This
member-organization relationship probably increased the credibility of the message.
143
5.6 Conclusion
More than forty years ago, Kramer (1966) observed,
In the past two decades, the use of quantitative methods as aids for decision-
making has become common in many fields. [...] By and large, however, these
efforts have not been made by political scientists. [...] This is unfortunate,
for many of the traditional concerns of political scientists appear to be quite
susceptible to this sort of analysis. (p.137)
Yet, political scientists have since generally neglected to use quantitative methods to di-
rectly inform policy makers who must make decisions using available data. This chapter
takes up Kramer’s proposal and show how modern statistical methods can be used to help
practitioners devise strategies and implement optimal policy using the Personal Experience
Hypothesis.
The proposed methodology is agnostic to theories (e.g., the Personal Experience Model)
about why certain subgroups should be more responsive to specific treatment. Nevertheless,
these theories can help practitioners form priors, identify sources of potentially useful data,
and better explain the results of the nonparemetric method. Such approaches demonstrate
one application of Kramer’s larger goal: arming campaign strategists with the theory and
methods to more accurately target voters who are susceptible to persuasive appeals.
144
Chapter 6
Campaign and Normative
Implications of Microtargeting
6.1 Campaign Microtargeting
Campaign microtargeting is a method for identifying a subset of voters to which to direct a
tailored message. The advent of electronic voter files and commercial databases has made
this strategy more efficient and more widely applicable to groups of voters (Malchow, 2003).
The Personal Experience Model is one theoretical connection between specific voters and
issues. However, several alternative group-issue connections are described in both the
academic (Krosnick, 1990; Chong et al., 2001) and practitioner literature (Sosnick et al.,
2006). This chapter outlines the situations in which campaigns should leverage these
theories to microtarget voters and the normative implications of that microtargeting on
democracy.
Microtargeting is a powerful tool. If voter-issue linkages are identified correctly, cam-
paigns can have an important effect on how voters judge candidates in an election. Chapters
145
3, 4, and 5 provide the empirical evidence to support this claim and demonstrate how cam-
paigns can leverage personal experience to garner support from voters who would otherwise
vote for the opposing candidate. In this chapter, first the situations in which campaigns
would not choose to microtarget are outlined. In the broad sense, campaigns that need to
change the landscape of a race dramatically would do best not to microtarget. Second, the
normative implications for microtargeting are discussed.
Various methods for identifying voters who will be responsive to narrow issue appeals
are available to campaigns. The easiest method, perhaps, is to use well-defined categories
from readily available voter files. For example, a campaign might use birth year to target
an age group, such as seniors. A second possibility is to use non-voter-file information (e.g.,
census data, licensing lists), match this data to the voter file, and identify specific groups,
such as hunters or those with low incomes. The third, and most complex, method is to
survey a group of voters, asking a question (e.g., “Do you have children under the age of
18?” or “What is the most important issue to you?”) the answer to which is not available
in any database. After a sufficient number of voters (perhaps on the order of 2,000) have
responded to a question, a campaign analyzes the data using either a parametric (e.g.,
logit) or a non-parametric (e.g., CHAID) method to correlate measures available for the
entire electorate to the survey response of interest (Malchow, 2003). The campaign can
then assign a probability of being in the specified subgroup to all voters. Similar to the
optimizations described in Chapter 5, the campaign would contact the voters with the
highest probability of being in the target group.
Figure 6.1 displays the results of a hypothetical microtargeting of undecided voters.
Voters in the top decile (by their probability of being undecided) are twice as likely to be
undecided as the population as a whole (20% to 10%).
146
Decile of Undecided Score
P
e
r
c
e
n
t

U
n
d
e
c
i
d
e
d
0
%
5
%
1
0
%
1
5
%
2
0
%
1 2 3 4 5 6 7 8 9 10
Figure 6.1: Example of the results of a campaign microtargeting undecideds. The top
decile of voters (10) includes twice as many undecided voters as the population average.
Targeting these voters is thus twice as efficient as a random strategy.
None of these methods perfectly identifies a list of voters who will change their votes
if targeted with an appeal. In an attempt to maximize the probability of a targeted voter
changing her vote choice, campaigns often also use the survey/CHAID method to identify
voters who are on (or near) the fence for their vote choice decisions. Further error is
induced when the targeted population is identified by survey, as electorate-wide covariates
may not predict survey responses well. The use of a training and test set can ensure that
microtargeting models do not over-fit and that they do identify targeted voters better than
if the campaign were to deliver messages randomly.
In essence, microtargeting increases the efficiency of campaign issue signals and the
linkage between a candidate’s stances and a shift in voters’ candidate preferences. If a voter
is experienced with an issue, belongs to an issue public, or has self-interest in an issue (all
microtargetable subgroups), that voter is more likely to pay attention to campaign appeals
147
on that issue. In all these cases, the voter more easily processes political arguments on
the issue (as a result of their experience, appetites, or self-interest) and thus is less likely
to ignore messages on the issue because the candidate’s appeal is either too complex or
uninteresting. Microtargeting increases the pace of voter learning about pivotal issues.
Applying the Personal Experience Model to various forms of voter targeting demon-
strates how microtargeting enhances the connection between candidate stances and vote
choice. Again, assume that campaigns target voters who have experience, who are in an
issue public, or who have self-interest. Because each of these types of targeted voters
has information or opinions that originate independently of political elites, the voter can
“ground-truth” the stances taken by the candidates (see Chapter 2). This independent
information allows the voter to judge the politicians on the issue when the voter cue-takes
from the politicians’ positions. Since cue-taking polarizes the electorate, and does not af-
fect vote preferences, an increase in microtargeting (and hence a reduction in cue-taking)
strengthens the relationship between pre-campaign voter attitudes, candidate stances and
candidate evaluation.
6.2 Implications for Campaign Strategy
6.2.1 Microtargeting and Message Control
If a campaign can identify a subgroup of voters with experience on an issue, it will be able
to deliver issue-specific messages to these voters. If the voters’ opinions are congruent with
the candidate’s platform, then the campaign’s appeal will raise the voters’ evaluations of
the candidate (on average). In some cases, these voters can make the difference between
losing and winning.
148
However, in some situations, microtargeting is a net loss for the campaign. Microtar-
geting and message delivery cost the campaign both in terms of money and opportunity
cost; thus the benefits must be proportional to the resources devoted to microtargeting.
In three situations, microtargeting is not helpful. First, independent of other options
for resource allocation, microtargeting is not necessarily beneficial for a campaign. For
instance, if all microtargetable voters disagree with the candidate on the issue in question,
then clearly the campaign should not raise the issue. If microtargetable voters do side
with the candidate on the issue, the campaign can identify these like-minded voters and
deliver messages only to them. (For the rest of this chapter, the term “microtargetable” is
restricted to this group of voters.)
Even if microtargetable, like-minded voters exist, sending these voters tailored messages
is useful only if these individuals are unlikely to vote for the candidate in the absence of
the appeal. Some of these individuals will likely abstain or vote for the opponent based
on their predispositions if the campaign does not target them. However, the proportion of
these voters in this circumstance can be small in high-salience, polarized elections where the
dimension of opinions on the issue in question is similar to the overall ideological dimension
of the campaign.
1
While potentially small in number, some portion of the electorate would
shift their vote choice only if microtargeted.
In addition to microtargeting, campaigns can spend resources attempting to control
their overall message agenda. One example is the 2002 midterm elections, when Democrats
pushed the issues of health care and corporate responsibility and Republicans attempted to
put taxes and national security at the fore of voters’ minds.
2
Because of cue-taking, these
1
A “high salience election” means few registered voters abstain, so little get-out-the-vote effect is pos-
sible. A polarized electorate means that few voters are undecided or potential defectors. “Similar issue
dimensions” means that few voters currently supporting the opponent are likely to be congruent with the
candidate for the issue in question.
2
Nagourney, Adam. “Domestic Concerns Take Center Stage In Congress Races.” New York Times.
September 1, 2002.
149
non-microtargeted issues often have little effect on the vote preferences of the electorate.
For instance, the sudden shift in dialogue in the 2008 presidential election after the third
debate to “spreading the wealth” and “Joe the Plumber” polarized the public on economic
issues and did not move the overall vote.
3
The exception to this rule is when the media
or public agree that one party (or candidate) has outperformed the other on an issue.
This occurs after the issue has resolved (Canes-Wrone et al., 2001) and accounts for the
importance of economic voting (Mueller, 1970) and retrospective voting in general (Fiorina,
1981). If a campaign chooses to spend resources to “broadcast” a message on a single issue
and that issue resolves in its favor, then the electoral payoff can be large.
6.2.2 Translating the Personal Experience Model into a Formal Game
To delineate the situations in which microtargeting is the optimal strategy, I develop a
model that explores the campaign resource allocation decision. In the broad sense, I
presume that a candidate (1) has decided to run for office, (2) has chosen a platform on
which to run, and (3) must decide how to spend the campaign’s limited resources (i.e.,
money). The candidate must choose between two methods of campaigning. First, the
campaign can microtarget voter-issue pairs for which the voters have stable opinions that
are congruent with the candidate’s platform. Alternatively, the campaign can broadcast
to all voters the candidate’s stance on an issue for which the voters have unstable opinions
but that may resolve in the candidate’s favor.
I also assume that the campaign has a wealth of knowledge about the electorate to
guide its decisions. Through public opinion polls (or other means), the campaign knows
the general predispositions of the electorate. The campaign also has knowledge about what
percentage of the electorate is microtargetable (i.e., has stable opinions) on each potential
3
See Gallup Polls on wealth redistribution and aggregate vote choice surveys on pollster.com.
150
campaign issue. In addition, the campaign has accurate beliefs about the probability that
certain issues will resolve in its favor by Election Day (e.g., an improving economy helps
the incumbent party).
The Personal Experience Model demonstrates how specific segments of the electorate
are disproportionately responsive to certain issue appeals when forming candidate evalua-
tions. Campaigns generally have the ability to identify these voters to some extent; perfect
identification is not necessary. If a campaign delivers targeted messages to these voters,
the campaign can shift some of these microtargeted voters’ candidate preferences. I also
assume that the more money a campaign invests in microtargeting, the more voters it can
reach.
Any voter-issue pairs in which some voters have a stable opinion is labeled a “potentially
microtargetable” pairing. The campaign whose platform is congruent with the voter’s
opinion in this pairing would prefer that the voter learn about the candidate’s stances on
the issue and have the issue on the top of her head come Election Day. However, the
campaign does not want to emphasize the issue among voters who have stable opinions
that are incongruous with its platform. Hence, even if wide swaths of the electorate have
stable opinion on issues, campaigns can maximize their efficiency by microtargeting only
the voters with congruent opinions.
Segments of the population who lack issue experience (or an alternative issue connection
such as self-interest) are not likely to shift their vote on the issue in question. Instead,
these voters cue-take from politicians, their vote choices polarize, and they do not change
their vote choice. This polarization mechanism requires two streams of considerations (i.e.,
an appeal from each ideological side), a criterion that is not always met. The canonical
example of one-sided issue streams is the early discussion of the Vietnam war (Zaller, 1992).
151
In this case, both liberals and conservatives alike supported the war as that was the only
position represented by political elites.
Message streams can become one-sided when an issue resolves. For instance, when the
country’s economy tanked in the fall of 2008, only 5% of the public rated the economic
situation of the country as “excellent” or ”good.”
4
Voters can then use the single stream
of information from the media to “ground truth” politicians performance. The evidence
that resolved issues affect large swaths of the electorate is clear when economic and war
data are correlated with election results on the national (Hibbs Jr., 2000) and state levels
(Campell, 1992; Cohen and King, 2004).
In the model, an issue can (1) resolve in favor of one candidate, (2) resolve in favor of
the other candidate, or (3) not resolve during the course of the campaign. Campaigns can
choose to focus their appeals on certain issues in an attempt to prime voters’ minds (i.e.,
place considerations about this issue on the tops of voters’ heads). In essence, campaigns
can place bets that certain issues will resolve in their favor.
One possible exception to the cue-taking rule is issue ownership (Petrocik, 1996). Re-
lated to the Broadcast Corollary discussed in Chapter 2, the theory of issue ownership
stipulates that wide swaths of the electorate favor one side of an issue. If a party “owns”
an issue, it can be considered an “easy issue” (Carmines and Stimson, 1990) in that voters
will likely have a stable opinion on the issue independent of cue-taking.
However, elections in which a large majority of voters side with one candidate on an
“easy” or “owned” issue are usually uncompetitive elections. The campaign that has this
majority of voters on its side most likely has a winning (and perhaps dominant) strategy to
emphasize this issue. Since close campaigns are more interesting, the model instantiations
4
“How would you rate economic conditions in this country today – as excellent, good, only fair, or
poor?” USA Today/Gallup Poll. Oct. 10-12, 2008. N=1,269 adults nationwide.
152
in Section 6.2.5 usually consider cases in which a small percentage of the electorate is
microtargetable.
6.2.3 The Model
In an election between two candidates, D and R, the campaigns decide how to allocate
resources across two issues (A and B) and whether to microtarget or broadcast their
messages about these issues. In the hypothetical world in which neither campaign makes
any effort, the predispositions of the voters dominate. Both campaigns have common beliefs
about their chance of winning in this no-campaign universe (i.e., a benchmark poll is taken
before the campaign to test the mood of the electorate). This belief is represented by a
mean percent of the vote for candidate i, µ
i
, and a common uncertainty, σ
2
(i ∈ D, R).
This representation is similar to a valence advantage in other models (Groseclose, 2001).
The vote for candidate i, V
i
, is distributed normally,
V
i
∼ ^(µ
i
, σ
2
). (6.1)
Candidate i’s utility, U
i
, is the campaign’s probability of winning,
U
i
= Pr(i wins) = Pr(V
i
> 0.5) = Φ(
µ
i
−0.5
σ
). (6.2)
To simplify the analysis, I assume only two candidates in the race, µ
D
= 1 − µ
R
, and a
fixed electorate with 100% turnout. The tiny tails of the vote distribution that extend
beyond the logical range of V
i
, [0, 1], are ignored.
Implicitly in the model, ties are broken with a coin flip. If the campaigns garner exactly
the same number of votes on average (i.e., µ = 0.5), then the probability of winning ap-
153
proaches 0.5 as the uncertainty of the result goes to zero (i.e., σ
2
→0). This characteristic
of the model is a result of the symmetry of the normal curve about its mean.
Under all circumstances, campaigns want to increase their share of the vote, µ
i
, since
∂U
i
∂µ
i
> 0. They attempt to increase this share by taking stances on issues. For the moment,
let the issues equal advantages for the campaigns. Issue A is an issue for which a portion
of the electorate has experience, although some experienced voters side with candidate D
and some with candidate R. Let the proportion of all voters who fall into one of those two
categories (i.e., who are microtargeted by the candidates) be m
D
and m
R
, respectively.
For now, assume a level playing field: m
D
= m
R
.
The other issue, B, is not microtargetable but has a chance of resolving by the election.
If this issue resolves in candidate D’s favor, an occurrence with probability of p
D
, q percent
of voters switch their preference from R to D. A symmetric switch of q percent of the
electorate occurs with probability p
R
; i.e., in this case, the issue resolves in candidate R’s
favor. Let p
D
= p
R
.
The campaigns do not have the resources to fully utilize the advantages offered by both
approaches. Instead, they must choose a combination of microtargeting and broadcasting.
Assume that both campaigns have equal resources, a budget of 1, and campaign i spends
δ
i
∈ [0, 1] on microtargeting. A campaign must spend its entire budget to gain the maxi-
mum votes from the strategy-issues combinations above.
5
The resulting distribution of the
percentage of votes for candidate D, given both campaigns disbursements, is
V
D

D
, δ
R
) ∼ ^(µ
i
, σ
2
) +δ
D
m
D
−δ
R
m
R
+ (2 −δ
D
−δ
R
)BS(p
D
, p
R
) (6.3)
V
D

D
, δ
R
) ∼ ^(µ
i
, σ
2
) +m(δ
D
−δ
R
) + (2 −δ
D
−δ
R
)BS(p, p) (6.4)
5
If one campaign spends all its resources on broadcasting, then q percent of voters switch. If both
campaigns only broadcast, then 2q voters switch.
154
where BS(p
1
, p
−1
) is a Bernoulli scheme with probability p
1
of outcome 1 and probability
p
−1
of outcome -1 (see Appendix E). The analogous equation shows the vote for candidate
R. The mean and variance of V
i
are,
mean(V
i

i
, δ
∼i
)) =
¯
V
i

i
, δ
∼i
) = µ +m(δ
i
−δ
∼i
) (6.5)
Var(V
i

i
, δ
∼i
)) = σ
2
+ 2pq
2
(2 −δ
i
−δ
∼i
)
2
(6.6)
If neither campaign spends any resources broadcasting (i.e., δ
i
= 1), each candidate’s
contribution to the variance (Equation 6.6) goes to 0. As before, the utility for candidate
i is the probability of winning, U
i
= Pr(V
i
> 0.5). For candidate D,
U
D
= (1 −2p)Φ((µ +m(δ
D
−δ
R
) −0.5)/σ) (6.7)
+pΦ((µ +m(δ
D
−δ
R
) +q(2 −δ
D
−δ
R
) −0.5)/σ)
+pΦ((µ +m(δ
D
−δ
R
) −q(2 −δ
D
−δ
R
) −0.5)/σ)
This model includes two assumptions that are rough approximations of the real world
and are not likely to hold in actual campaigns. First, the proportion of voters who will
switch their candidate preference because of microtargeting (m) does not vary based on the
resources spent on broadcasting (1 −δ), or vice versa (with q and δ). In essence, this sim-
plification assumes a uniform distribution of the electorate across the ideological spectrum
regardless of whether a candidate becomes more advantaged (i.e., the decision cutpoint on
the ideological spectrum shifts); the number of nearly undecided (i.e., indifferent) voters
is constant. Second, if the campaign microtargets voters who have a high probability of
shifting their vote choice, the rate of successful microtargeting is not affected by spending
on broadcasting. In other words, wasteful spending by campaigns that “doubles up” on
voters—contacting them with both microtargeting and broadcasting when only one of the
155
methods is necessary to shift the voter’s preference toward the candidate—is assumed not
to occur.
6
6.2.4 Best Response
The optimal strategy of candidate i, given the strategy of the other candidate (labeled ∼ i)
is the level of microtargeting (δ

i
) that maximizes candidate i’s utility,
δ

i

∼i
) = argmax
δ
i
U
i

i
, δ
∼i
) (6.8)
To determine the best response to the opponent’s strategy, δ
∼i
, the partial derivative
of candidate i’s utility is taken with respect to the campaign’s strategy,
∂U
i
∂δ
i

i
, δ
∼i
) = (1 −2p)
m
σ
φ((µ +m(δ
i
−δ
∼i
) −0.5)/σ) (6.9)
+p
m−q
σ
φ((µ +m(δ
i
−δ
∼i
) +q(2 −δ
i
−δ
∼i
) −0.5)/σ)
+p
m+q
σ
φ((µ +m(δ
i
−δ

i
) −q(2 −δ
i
−δ
∼i
) −0.5)/σ)
Under some circumstances, the resource allocation decision is trivial. If the number of
votes that can be shifted with microtargeting (m) is greater than the proportion of votes
possibly shifted if the issue resolves (q), then the campaign should spend all its money on
microtargeting. Formally, m > q →
∂U
i
∂δ
i
> 0 (see Equation 6.4) and the candidate’s utility
is increasing in δ.
As an intuition for the whether a campaign has an incentive to microtarget or broadcast,
consider the effect of broadcasting on the variance of V
i
. Pouring more resources into
6
A myriad of real-world considerations are not included in the model such as diminishing returns, fixed
costs, and economies of scale.
156
broadcasting increases the percentage of the population that will switch under a resolved
issue, and since issue B resolves stochastically, increased broadcasting increases the variance
of V
i
. For a mathematical intuition of this result, see equation 6.6 and note that
∂Var(V
i
)
∂δ
i
=
−4pq
2
(1 − δ
i
− δ
∼i
) is negative, so the variance of the vote increases with the amount of
resources spent on broadcasting.
Similar to the stochastic knapsack problem discussed (see discussion D.1.2), campaigns
want to increase the variance of the election result if they are losing (i.e., the expected vote
percentage is below 50%) and decrease the variance if they are winning (Carraway et al.,
1993). Applying this principle to the microtargeting vs. broadcasting model, we find an
incentive for broadcasting only when a campaign cannot reach 50% (on average) with a
combination of predispositions and microtargeting.
To prove this result formally, note that adding or subtracting a value to the point
on a Gaussian probability distribution has the property x < 0 ↔ φ(x + c) > φ(x −
c). Next, Equation 6.9 is rewritten to combine the terms influenced by the potential for
microtargeting, m, and broadcasting, q,
∂U
i
∂δ
i

i
, δ
∼i
) =
m
σ
[(1 −2p)φ((µ +m(δ
i
−δ
∼i
) −0.5)/σ) (6.10)
+pφ((µ +m(δ
i
−δ
∼i
) +q(2 −δ
i
−δ
∼i
) −0.5)/σ)
+pφ((µ +m(δ
i
−δ
∼i
) −q(2 −δ
i
−δ
∼i
) −0.5)/σ)]

q
σ
[pφ((µ +m(δ
i
−δ
∼i
) +q(2 −δ
i
−δ
∼i
) −0.5)/σ)
−pφ((µ +m(δ
i
−δ
∼i
) −q(2 −δ
i
−δ
∼i
) −0.5)/σ)].
For
∂U
i
∂δ
i
to be negative, the campaign must start in a losing position and not be able to
make up this ground with microtargeting alone: µ + m(δ
i
−δ
∼i
) < 0.5. That condition is
157
necessary, but not sufficient, for the optimal campaign strategy to be 100% broadcasting.
The other necessary condition is that broadcasting must be sufficiently more potent than
microtargeting (q >> m) that Equation 6.11 is negative.
If the campaigns start on equal footing (i.e., µ
D
= µ
R
= 0 as well as m
D
= m
R
= m,
p
D
= p
R
= p, and q
D
= q
R
= q), then the weakly dominating strategy is to spend all
resources on microtargeting; if campaign ∼ i spends all its resources on microtargeting

∼i
= 1), then any money campaign i spends on broadcasting will be wasted in the case
of issue B not resolving. When the issue does not resolve (which occurs with probability
1−2p), campaign i loses with probability Φ(
m(1−δ
i
)
σ
), which is greater than 50%. If the issue
resolves, then the half the time the issue resolves in campaign i’s favor nearly balances out
with the other half the time it does not. But even in this case, the result is a net negative
for campaign i since the mean point is negative and x < 0 →Φ(x +c) < (1 −Φ(x −c)). If
candidate ∼ i foolishly does not spend all his resources on microtargeting, then candidate
i can take advantage of that error and win over half the time by following this logic and
spending all his money on microtargeting.
Thus, in the case of a campaign in a marginal district with symmetric properties,
microtargeting is a weakly dominating strategy. The strategy is not strictly dominant
because of the case where issue B always resolves (p = 0.5). Figure 6.2 illustrates the
relationship between the probability of issue B resolving and the usefulness of broadcasting
relative to microtargeting. The utilities of candidate D for all possible combinations of
microtargeting and broadcasting are shown under three scenarios, each with increasing
probability of issue resolution. As the effect of microtargeting shrinks (m) relative to the
effectiveness of broadcasting (q), the utility of a microtargeting-only strategy and the utility
of a broadcasting-only strategy approach the same value: 50%.
158
Low Probability (20%) of
Issue Resolution
Proportion of Candidate D's of
resources spent microtargeting
P
r
o
b
a
b
i
l
i
t
y

t
h
a
t

C
a
n
d
i
d
a
t
e

D

w
i
n
s
0.0 0.2 0.4 0.6 0.8 1.0
0
.
3
8
0
.
4
0
0
.
4
2
0
.
4
4
0
.
4
6
0
.
4
8
0
.
5
0
High Probability (80%) of
Issue Resolution
Proportion of Candidate D's of
resources spent microtargeting
P
r
o
b
a
b
i
l
i
t
y

t
h
a
t

C
a
n
d
i
d
a
t
e

D

w
i
n
s
0.0 0.2 0.4 0.6 0.8 1.0
0
.
3
8
0
.
4
0
0
.
4
2
0
.
4
4
0
.
4
6
0
.
4
8
0
.
5
0
Issue Resolution is
a Certainty
Proportion of Candidate D's of
resources spent microtargeting
P
r
o
b
a
b
i
l
i
t
y

t
h
a
t

C
a
n
d
i
d
a
t
e

D

w
i
n
s
0.0 0.2 0.4 0.6 0.8 1.0
0
.
3
8
0
.
4
0
0
.
4
2
0
.
4
4
0
.
4
6
0
.
4
8
0
.
5
0
Figure 6.2: Candidate D’s utility under increasing probabilities of issue B resolving. In the
examples depicted, 5% of the electorate is microtargetable by each campaign (m = 0.05)
and 25% of the electorate shifts if the issue resolves (q = 0.25). The campaigns start on
equal footing (µ = 0.5), although uncertainty about the electorate is high (σ = 0.08).
The panels depict candidate D’s utility under the strategy specified by the x-axis given
candidate R spending all resources on microtargeting δ
R
= 1 and increasing probabilities
of issue B resolution (p = 0.1, p = 0.4, and p = 0.5, respectively).
6.2.5 Edge Equilibria
Extending this logic, in the case where candidates are on equal footing, only corner equilib-
ria exist. In nearly all circumstances, these equilibria occur when both campaigns micro-
target. In this section, restrictions on symmetry are relaxed; three scenarios are considered
and depicted in Figure 6.3.
The first scenario is similar to the symmetric case considered in the previous section
except that candidate D has an advantage with respect to broadcasting. More often the
issue will resolve in candidate D’s favor (p
D
= 0.7, p
R
= 0.3) and more voters will switch
to candidate D in the case of resolution (q
D
= 0.2, q
R
= 0.1). Thus, at high levels of
microtargeting by candidate R (approximately, δ
R
> 0.8), candidate D optimally puts all
159
0.0 0.2 0.4 0.6 0.8 1.0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
(a) Candidate R's Advantage Holds
Cand D Strategy: Pct. Micortarget
C
a
n
d
.

R

S
t
r
a
t
e
g
y
:

P
c
t
.

M
i
c
o
r
t
a
r
g
e
t
X
Candidate D Best Resp.
Candidate R Best Resp.
0.0 0.2 0.4 0.6 0.8 1.0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
(b) No Starting Advantage;
Cand D Issue Res. Adv.
Cand D Strategy: Pct. Micortarget
C
a
n
d
.

R

S
t
r
a
t
e
g
y
:

P
c
t
.

M
i
c
o
r
t
a
r
g
e
t
q
X
Candidate D Best Resp.
Candidate R Best Resp.
0.0 0.2 0.4 0.6 0.8 1.0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
(c) Cand R has advantage;
Cand D Needs Luck to Win
Cand D Strategy: Pct. Micortarget
C
a
n
d
.

R

S
t
r
a
t
e
g
y
:

P
c
t
.

M
i
c
o
r
t
a
r
g
e
t
q
q
X
Candidate D Best Resp.
Candidate R Best Resp.
0.0 0.2 0.4 0.6 0.8 1.0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
(d) Cand R Starting and Iss.
Res. Adv.; No Equilibrium
Cand D Strategy: Pct. Micortarget
C
a
n
d
.

R

S
t
r
a
t
e
g
y
:

P
c
t
.

M
i
c
o
r
t
a
r
g
e
t
q
Candidate D Best Resp.
Candidate R Best Resp.
Figure 6.3: Best Response Plots of Four Scenarios.
160
resources into broadcasting. As shown as an “X” in panel (a) of Figure 6.3, the equilibrium
is candidate D fully broadcasting and candidate R fully microtargeting.
When candidate i has an advantage in the electorate (i.e., µ
i
> 0.5) that can be
maintained when both campaigns microtarget (i.e., µ
i
+m
i
−m
∼i
> 0), then campaign i’s
best strategy is often to microtarget only. The microtargeting strategy is optimal because
it lowers the variance of the election result compared to broadcasting. The disadvantaged
campaign’s best response is often a combination of microtargeting and broadcasting; the
exact mix depends on the relative effectiveness of those two strategies.
An edge equilibrium of this type is depicted in Figure 6.3, panel (b). In this case can-
didate R has a pre-campaign edge (µ
R
= 0.54, σ = 0.04) and candidate D cannot recover
this edge via microtargeting (m
D
= m
R
= 0.03). In equilibrium, R fully microtargets to
bring his chance of winning up to 72%. Candidate D wants both a high vote mean and
a high variance; microtargeting provides the former and broadcasting (q
D
= q
R
= 0.2,
p
D
= p
R
= 0.25) provides the latter. Hence, candidate D spends 28% of his money on
microtargeting and the rest on broadcasting.
An exception to the general rule that advantaged candidates microtarget is presented
in panel (c). In this case, candidate R is advantaged both in the electorate’s predisposition

D
= 0.445, σ = 0.067) and issue resolution (q
R
= 0.1, q
D
= 0.1; p
R
= 0.7, p
D
= 0.2).
By combining a microtargeting and a broadcasting strategy, candidate R can increase
his probability of winning to 80%. Candidate D’s best response in this case is to fully
microtarget; this pair of strategies is an equilibrium.
In this example, candidate D has an interesting best response curve, which has three
regimes. When candidate R does not microtarget much (δ
R
< 0.2) then candidate R can
increase the mean vote to fairly close to a 50-50 election (m
D
= 0.044, m
R
= 0.008).
Also, the emphasis candidate R places on broadcasting in this case increases the variance
161
of the election, which also raises candidate D’s probability of winning. When candidate
R microtargets to a substantial degree (approx. m
D
> 0.4) then the benefit of raising
the mean by candidate D via microtargeting is greater than the benefit of increasing the
variance by broadcasting. In the middle of these two regimes, the need for candidate D to
raise the mean is small enough to encourage broadcasting. However, candidate D cannot
free-ride off candidate R’s full broadcasting effort, and it is worth candidate D’s while to
broadcast himself.
No equilibrium exists in the final example, which is depicted in panel (d). Candidate
R is advantaged in electorate predispositions and in issue resolution. This circumstance
is interesting because both campaigns may have an incentive to broadcast. Candidate R
can broadcast to increase the mean of his vote distribution. Candidate D can broadcast
to increase the variance of his vote share. (The latter strategy is helpful to candidate
D’s mean of vote share, which is always below 50%.) However, it is never the case that
both candidates broadcast at the same time. As candidate D spends more resources on
broadcasting only, he contributes to candidate D’s vote-share variance, so candidate D has
an incentive to microtarget. As candidate D spends more on broadcasting, he contributes
to candidate R’s vote-share mean, so candidate R has an incentive to microtarget and
lower his vote-share variance. In this manner, the candidates never reach equilibrium.
6.2.6 No Internal Equilibria
In the microtargeting-broadcasting game, there are no internal equilibria. Nor are there
edge equilibria in which one candidate fully broadcasts. To prove this formally for the
general, non-symmetric case, first I derive the candidate i’s utility and first derivative in
this case.
162
U
i

i
, δ
∼i
) = (1 −p
i
−p
∼i
)Φ((µ +m
i
δ
i
−m
∼i
δ
∼i
−0.5)/σ) (6.11)
+p
i
Φ((µ
i
+m
i
δ
i
−m
∼i
δ
∼i
+q
i
(2 −δ
i
−δ
∼i
) −0.5)/σ)
+p
∼i
Φ((µ +m
i
δ
i
−m
∼i
δ
∼i
−q
∼i
(2 −δ
i
−δ
∼i
) −0.5)/σ)
∂U
i
∂δ
i

i
, δ
∼i
) =
m
i
σ
(1 −p
i
−p
∼i
)φ((µ +m
i
δ
i
−m
∼i
δ
∼i
−0.5)/σ) (6.12)
+
m
i
−q
i
σ
p
i
φ((µ
i
+m
i
δ
i
−m
∼i
δ
∼i
+q
i
(2 −δ
i
−δ
∼i
) −0.5)/σ)
+
m
i
+q
∼i
σ
p
∼i
Φ((µ +m
i
δ
i
−m
∼i
δ
∼i
−q
∼i
(2 −δ
i
−δ
∼i
) −0.5)/σ)
The proof that no equilibria exist where both candidates spend resources on broad-
casting follows by contradiction. Assume that equilibrium of (δ

i
, δ

∼i
) exists where δ

i
<
1 ∧ δ

∼i
< 1. The contribution of microtargeting alone to the candidate’s utility is always
positive: limit
q
i
→0
∂U
i
∂δ
i
> 0. For candidate i’s optimal strategy (δ
i
∗) to include some broad-
casting, the first derivative is nonpositive:
∂U
i
∂δ
i


i
, δ

∼i
)) ≤ 0. Thus, the broadcasting term
of the first derivative is negative:
0 > −
q
i
σ
p
i
φ((µ
i
+m
i
δ
i
−m
∼i
δ
∼i
+q
i
(2 −δ
i
−δ
∼i
) −0.5)/σ) (6.13)
+
q
∼i
σ
p
∼i
Φ((µ +m
i
δ
i
−m
∼i
δ
∼i
−q
∼i
(2 −δ
i
−δ
∼i
) −0.5)/σ)
But analogous logic for candidate ∼ i yields the result that negation of the right-hand
side of Equation 6.13 must be negative. The negation of a negative cannot be negative.
Contradiction.
163
6.2.7 Model Extensions: Multiple Issues and Platform Decisions
If more than two issues are at play in an election, the model will also inform a candidate’s
decision about which issues to focus his efforts on. A candidate who is advantaged against
his opponent’s optimal strategy seeks to increase the mean of his vote share and decrease the
variance. Candidates who, on average, cannot win a majority of the vote seek to increase
both their vote share and the variance of the outcome. The model can be extended to
address multiple issues in a straightforward manner, although the strategy space would
increase to k − 1 dimensions, where k is the number of issues considered. The logic of
Section 6.2.6 holds within an issue: no two candidates would broadcast on the same issue,
although candidates might optimally broadcast on two different issues.
Incorporating multiple issues enables the model to be extended to platform choice.
Similar to the model of Groseclose (2001), candidates would have exogenous policy prefer-
ences and weights that they would place on winning the election vs. policy outcome. The
choice of which issues to emphasize (either with microtargeting or broadcasting) would be
a function not only of whether the issues can help the candidate win, but whether the
candidate’s positions on the issues are consistent with his policy preferences.
6.2.8 Discussion
The model has three major implications. First, in general, campaigns that are behind
microtarget and campaigns that are ahead broadcast. This implication may actually be
understated by the model, as advantaged campaigns would not want to microtarget just
to increase their vote share, but also to prevent voter defection. A natural extension of the
theory presented in Chapter 2 suggests that delivering microtargeted appeals before issue
resolution will prevent some voters from defecting. Thus, early microtargeting might be
even more effective than demonstrated at lowering the variance of election results.
164
Second, a losing campaign may be in the situation where it must lower its expected
vote share to increase its chance of winning. This circumstance was epitomized by the
description provided to John McCain of Sarah Palin during his running-mate decision
process: “high risk, high reward.”
7
Since losing campaigns take risks that on average do
not work out for them, more strategic thinking by campaigns may lead to fewer moderately
close elections.
Third, campaigns never purposely talk to the same voters about the same issues. Op-
posing campaigns may microtarget the same issue, but they are targeting mutually exclu-
sive groups: only voters who agree with the campaign on the issue. The empirical evidence
demonstrates that opposing campaigns seek to emphasize distinct sets of issues (Sellers,
1998), although current events often foil these plans and force campaigns to talk about the
same issues (Sigelman and Buell, 2004).
6.3 Normative Implications
Microtargeting increases the efficiency of individuals learning about candidates’ positions.
Whether or not the increased ability of campaigns to microtarget is helpful or detrimental to
democracy depends on context and situation. Fundamentally, microtargeting is a tool, and
like most tools (e.g., screwdrivers, TNT), it can be used for positive or negative purposes.
Microtargeting has both positive and negative implications for democracy.
6.3.1 Heuristics and Judging Democracy
A fundamental debate rages in the study of American politics: How much information
does a voter need to make an informed decision in the voting booth? On one side of the
ledger are those who believe that Americans are ill-informed (Delli Carpini and Keeter,
7
As publicly stated by A.B. Culvahouse, recounting his conversation with John McCain.
165
1996), that they are generally incapable of forming coherent opinions (Converse, 1964),
and that this lack of information is crucial to their decisions (Lau and Redlawsk, 2001;
Bartels, 2005). In the other camp are scholars who believe that cognitive shortcuts are
available (Popkin, 1994), voters seek the information that interests them (Krosnick, 1990),
and any errors cancel each other in collective public opinion (Page and Shapiro, 1992).
Often, when judging democracy, scholars analyze citizens’ vote preferences (Lau and
Redlawsk, 2001; Bartels, 1996). The argument from the “more information needed” camp
is that there is a standard by which citizens should judge candidates. Lau and Redlawsk
(2001) attempt to infer the weights that voters should place on various issues to determine
a “correct” vote. Bartels (1996) assumes that if low-informed voters become more knowl-
edgable, they will process information similarly to their high-information counterparts—an
assumption that is called into question by research on information processing in political
situations (Gilens, 2001).
These assumptions about which candidate preference a voter should have are trou-
blesome because they impose scholars’ views on how citizens choose among political al-
ternatives. A safer standard for judging democracy is to evaluate voters’ knowledge of
indisputable facts. The debate over heuristics is lively under this normative standard as
well (Gilens, 2001; Bartels, 2002; Caplan, 2008). The downside is that the outcome of
governments (democratic or otherwise) is policy decisions, and citizens’ issue opinions may
be farther from those policy positions than their vote decisions.
To balance these views, I use two standards for judging democracy. The primary
standard is the amount of accurate information voters have. Often the accuracy of political
opinion is in dispute; even when political elites agree on the optimal outcome (e.g., lower
unemployment), they disagree about how to achieve that outcome. Thus, for this first
166
standard, the analysis is constrained to the set of information that is verifiable. Included
in this set are candidate positions (which the candidate defines) and observable facts.
Second, I examine the more subjective standard of policy options and vote choice,
albeit under certain crucial assumptions. Often voters must judge candidates’ platforms
before knowing the effect of the policy stances contained within those platforms. I assume
that certain segments of the population have more knowledge about the probability of
positive resolution of an issue under the various proposed policy options. If this segment of
the population has more sway in the election outcome, the resultant government’s policy
choices are more likely to be optimal. (Optimal is defined as issue resolution that the most
people approve of, e.g., good economy or winning a war).
6.3.2 Positives for Democracy
Microtargeting increases the efficiency of voters learning the candidates’ positions. If cam-
paigns microtarget, a voter with experience on a political issue is more likely to receive
messages about that issue. Microtargeting increases the amount of information at voters’
disposal and is thus a positive for democracy under the first standard.
An example of this process is seniors learning about Bush’s and Gore’s positions on
Social Security (see Chapter 3 for details). At the beginning of the campaign (through
May 2000), 42% of seniors (ages 65 and older) could correctly identify both candidates on
Social Security, compared to 36% of younger voters: a difference of 6 percentage points. In
August, this difference was 8 percentage points; in September, 10 points; and in October,
12 points. This interaction between dichotomous age and time is significant at conventional
levels (logit regression, p = 0.05).
8
8
For two other issues tested, vouchers and taxes, the interaction effect is in the same direction, but is
half the magnitude and not statistically significant.
167
It is unclear how much of this effect was due to microtargeting by the Bush and Gore
campaigns, but clearly seniors were more actively engaged in the Social Security debate. If
the campaigns had been able to use the techniques developed just two cycles later, perhaps
more seniors would have learned about the candidates’ positions. Even by the end of the
campaign, over a third of seniors could not accurately identify both candidates’ positions
on Social Security.
Transitioning to the second standard—correct policy opinion—if experienced voters
have more knowledge about which policy options will resolve favorably, then the increased
information efficiency provided by microtargeting benefits democracy. In game theory
parlance, experienced voters have more knowledge about the “state of the world” on that
particular issue than non-experienced voters.
9
In a situation similar to the “Swing Voter’s
Curse” of Feddersen and Pesendorfer (1996), the non-experienced voters do not know
which policy proposal is best on a particular issue. However, if the pool of voters who judge
politicians on the issue for which they have experience (and thus, more in-depth knowledge)
is larger than any underlying partisan or incumbency biases, then the candidate with the
best policies will be elected.
As an illustration of why microtargeting helps in this case, consider a challenger fac-
ing an incumbent who is so incompetent that he is wrong on every issue. Assume that
the candidate cannot credibly prove the incompetency of his opponent; instead, he can
credibly offer proof of his opponent’s issue positions. This situation would occur when the
incumbent’s preferred policies have not been implement or have not yet resolved poorly.
If the challenger does not microtarget and instead broadcasts on one issue to all voters,
then only the small subset of voters with experience on that issue will be affected (unless
the issue resolves). The incumbent’s resources (e.g., higher name recognition, more funds)
9
If, for some reason, experienced voters are less likely to hold the correct policy position than the general
public, then the analysis works in the opposite direction and microtargeting is detrimental to democracy.
168
might be sufficient to overcome a small block of voters casting ballots for the challenger.
However, if the challenger uses modern technology to match voters to the issues they have
experience with, several subsets of voters will judge the incumbent poorly, potentially
increasing the challenger’s probability of success to greater than 50%. As demonstrated in
Section 6.3, the challenger will want to microtarget in this case—this strategy will prevent
the incumbent from winning again and implementing wrong-headed policies.
6.3.3 Negatives for Democracy
There are potential downsides to microtargeting as well. Foremost, microtargeting may
enhance the ability of candidates and parties to have individuals believe incorrect state-
ments. As shown in the survey experiments in Chapter 4, delivering congruent cues to
partisans increases their attachment to their party’s candidates. This attachment leads to
cue-taking (see Chapters 2 and 4), which means that individuals uncritically accept the
position of their preferred elite. Thus, they may be more prone to believe false information
(Bartels, 2002).
As an example of individuals believing verifiably incorrect information, the 2008 Coop-
erative Congressional Election Survey (CCES) asked potential voters whether they knew
Barack Obama’s religion. Despite several statements by Obama that he was Christian and
a highly covered debate in the spring of 2008 about controversial statements by Obama’s
pastor, over a quarter of the electorate believed (with at least some degree of certainty) that
Obama was a Muslim. Conservative elites (e.g., Fox News) were peddling this falsehood
and the individuals who believed it were nearly all conservative.
10
Another reason why microtargeting might be detrimental to democracy is that peo-
ple may have experience with issues that affect only a small part of their lives. For in-
10
Of the respondents who were able to place themselves somewhere other than 50 on the 0-100 scale of
liberal to conservative, 90% were in the conservative half.
169
stance, an environmentalist might be microtargeted on global warming and vote for the
pro-environment candidate even though she does not support that candidate’s economic
policies. Since global warming works over a decade- or century-long time frame, the econ-
omy is probably more relevant to the voter. But the voter may nonetheless voter for the
suboptimal candidate based on the issue with which she has experience.
Undoubtedly voters sometimes have policy opinions that are incorrect, even by their
own standards (with the benefit of hindsight). For instance, a third of the public went
from believing the Iraq War was justified to believing it was a mistake from 2003 to 2006.
Would a more informed electorate have had a different view of the war in 2003?
11
Given
that elites on both sides were advocating their respective positions and Americans were
filtering their consideration intake (e.g., via Zaller’s RAS model), it is not clear that a
more attentive electorate would have had a different opinion. Perhaps a more attentive
electorate would have had more considerations at the tops of their heads, but in the same
distributions.
The danger is that voters with strong party or elite affiliations may have tighter filters.
Since microtargeting increases polarization and elite affection (see Chapter 2 for the theory
and Chapter 4 for the empirical evidence), advancements in campaigns’ targeting abilities
may increase cue-taking from elites; elected officials’ subsequent policy decisions may lead
to negative outcomes.
6.3.4 Further Discussion: Party Structure and Alignment
The cue-taking effect that causes large segments of the population to believe falsehoods also
leads to closer party affiliation. Political parties in the United States often include different
11
This question differs from that of whether better intelligence would have changed the minds of the
elites. If more moderate Democrats (e.g., John Kerry, Hillary Clinton) had voted against the 2002 war
resolution, a lower percentage of the public would have supported the war.
170
“types” of voters (e.g., social conservatives, economic conservatives). Microtargeting on an
issue a voter has a stable opinion about (e.g., abortion) raises her opinion of her party and
leads her to more easily accept the party’s platform on other issues (e.g., the economy).
This artifact of microtargeting has both positive and negative implications for demo-
cratic participation. On one hand, voters polarize when affiliating with parties, thus in-
creasing political participation (Dalton, 2008). The number of parties is kept to a minimum
(since parties can better keep factions under one roof), which also increases political par-
ticipation (Blais and Dobrzynska, 1998). On the other hand, parties are less representative
of voters’ stable beliefs, which can produce negative results if the parties abuse the voters’
affiliations by pursuing rent-seeking or ideologically extreme policies that are detrimental
to the state of the nation (Sartori, 1976).
Whether parties that comprise disparate groups are beneficial to voters depends on the
ideological alignment of voters and parties. Converse (1964) finds that most citizens hold
ideologically incoherent opinions and that these opinions are highly unstable. The Personal
Experience Model theorizes that many voters have stable opinions on a (potentially small)
set of issues. If voters had personal experience with every issue would they agree with their
party on that issue? The answer to that question, which is generally unknowable, would
determine whether having a small number of big-tent parties is beneficial to voters.
12
6.4 Conclusion
The Personal Experience Model provides a theory explains and can inform campaigns’ mi-
crotargeting tactics. The more campaigns combine political science research with individual-
level data, the more they can control their own destinies. Instead of simply hoping that an
12
Even with increased cue-taking with microtargeting, voters may have ideologically incoherent issue
opinions. The Cue-Taking Hypothesis requires that voters’ know the parties’ positions on issues. As the
2000 Social Security example shows, this condition holds only when an issue has become heavily politicized.
171
issue will resolve in the campaign’s favor a week before the election, a campaign can plan
well in advance which voters they will microtarget and when.
Microtargeting is a powerful tool that helps identify the voters who are most likely to
change their vote preference as a result of a persuasive appeal. These voters are also more
likely to receive information about candidates’ stances that is relevant to their lives and
that therefore makes their voting decisions easier. However, those easy decisions can lead to
voters being too trusting of elites. Microtargeting thus offers both positives and negatives
for democracy. Which side has more weight, as always, depends on one’s perspective.
172
Chapter 7
Conclusion
Political scientists spend more time explaining why campaigns do not matter (e.g., Gelman
and King, 1993; Campbell, 2005) than why they do (Gerber et al., 2007). This perspective
remains the norm for studies of micro-level behavior. Thousands of pages have been devoted
to partisan bias (e.g., Campbell et al., 1960; Markus and Converse, 1979; Bartels, 2002;
Goren, 2007), motivated reasoning (e.g., Lodge and Taber, 2000), cognitive filters (e.g.,
Zaller, 1992), and cue-taking (e.g., Gilens and Murakawa, 2002). The field of neuroscience
has continued this trend, exploring how the brains of Democrats and Republicans process
information differently (Westen et al., 2006; Knutson et al., 2006).
However, research based on these paradigms cannot account explain why voters shift
from one party to another—a phenomenon that has a tremendous impact on policy out-
comes and the political trajectory of the country. Several presidential elections have yielded
popular-vote results in which the margin of victory was closer to one percentage point. Two
times in the last decade, control of the Senate has been determined in a single election.
Even small exceptions to the generally-accepted understanding of the role of partisan bi-
173
ases are extremely consequential. This dissertation’s primary purpose is to explore one of
those exceptions: personal experience.
The dissertation’s second goal is to bridge the divide between researchers and political
practitioners. The goal of all campaign managers is to win elections, and achieving that aim
often necessitates persuading voters who would normally vote one way to vote another way.
Research that demonstrates the pervasiveness of partisan biases is of little help to these
campaign managers. They need a way to counter these biases and persuade individuals to
alter their predispositions.
Political science does identify several mechanisms for persuasion, notably issue publics
and self-interest. In these cases, a voter might evaluate politicians on what they know
about an issue, a process here labeled political “ground truthing,” instead of uncritically
adopting the politician’s opinion (i.e., cue-taking). This dissertation differs from the issue
publics and self-interest threads of the literature in four ways. First, personal experience
is introduced as a mechanism that voters use to develop a stable opinion on an issue
independent of their political predispositions. Second, a micro-theory formally describes
the reasons why voters cue-take from politicians in general but judge these elites on issues
when the voters have issue experience. Third, the Personal Experience Model explains why
an individual might join an issue public and why self-interest might not always dominate.
Fourth, the Model is evaluated in terms of its usefulness to political practitioners, the
people whose job it is to counter political biases among the masses.
Personal experience with an issue, whether it is a politicized issue or not, leads to stable
opinions regarding that issue. The proximity of the experience to the individual must be
sufficient to circumvent the partisan filters that usually color voters’ perceptions. This
proximity is cast as the voter taking an active role in the experience, in effect demonstrating
174
the phrase “seeing is believing.” The experiential characteristics of frequent, consistent in
nature, and one-sided also help individuals form stable opinions.
The Personal Experience Model demonstrates how personal experience forms the basis
of political judgments. This model formalizes Zaller’s (1992) RAS framework as a learning
model. When an issue becomes politicized, voters with unstable opinions on the issue—
often individuals who lack personal experience—adopt the position of their preferred elite.
This cue-taking prevents voters from re-evaluating politicians on that issue.
In contrast, voters with experience have knowledge about the now-politicized issue apart
from their political predispositions. These voters can ground truth politicians’ stances and
alter their perceptions of these political elites. This segment of the electorate is a key
target for campaigns; experienced voters who agree with the candidate’s issue positions
are low-hanging fruit for persuasion while experiences voters with incongruent opinions are
potential defectors.
The Personal Experience Model provides a richer understanding of voter-issue link-
ages such as issue publics and self-interest. Individuals who have experience with a non-
politicized issue are likely candidates for the issue public if the issue becomes politicized.
The Personal Experience Model does not require an issue to be the most salient for an
individual to drive vote changes, which might explain why large portions of the public can
belong to issue publics (Gershkoff, 2006) while citing the same topic as the country’s “most
important issue.”
The two hypotheses of the Personal Experience model help explain an incongruity
in the literature on self-interest. Voters can be primed on self-interest in an artificial
setting (Chong et al., 2001), yet this effect is often unimportant in electoral contexts
(Sears and Funk, 1990). The explanation offered here is: if a voter does not have experience
understanding the complexities of an issue, the voter’s initial opinion matters as she adopts
175
the position of her preferred candidate. The voter will have a stable opinion and be able
to judge the candidates only on issues that offer clear benefits to the voter.
Unlike recent research on voter persuasion (Hillygus and Shields, 1991) that focuses on
voter movement apart from campaign treatment, this dissertation assesses the impact of
campaign action on shifts in voter behavior. These shifts are particularly important when
they counter existing partisan biases and lead to changes in vote choice.
Chapter 3 quotes the 2000 nomination speeches on Social Security and a Patients’ Bill
of Rights, finding that, on average, experienced voters increase their probability of defection
by three to four percentage points. Chapter 4 explores the effect of hypothetical candidates
taking random stances on issues and finds an increased probability of three percentage
points for experienced respondents. Chapter 5 examines the effect of campaign literature
on voters in real campaigns and finds increased persuasion of about seven percentage points
among experienced voters.
Each of these methods has its strengths and weaknesses. The observational data of
Chapter 3 are useful because they demonstrate the effects of actual campaign activity
during a very salient election. The data are noisy, however, and it is difficult to separate
out the effects of self-interest from those of personal experience. The survey experiments
of Chapter 4 discern the relative effects of personal experience, self-interest, and political
interest. The price of these clean tests is external validity; hypothetical candidates are a
poor substitute for real-life politicians. The field experiments of Chapter 5 demonstrate
that applying the lessons of the Personal Experience Model increases campaign efficiency.
However, the data provide no information on voters’ issue opinions, making it difficult to
test alternative hypotheses.
Also, all of these analyses rely on survey responses to represent actual voting behav-
iors. And none of the analyses randomize the variable of interest, personal experience;
176
thus, causal inference is challenging. Even with these caveats, the combined weight of the
evidence supports both hypotheses of the Personal Experience Model.
The Personal Experience Model presents a clear method for targeting voters. However,
it is not always in the campaign’s best interest to segment the electorate and microtarget.
Notably, campaigns that are inherently disadvantaged and are trailing badly in the polls
probably need a bigger boost than the single-digit returns that can be expected from
microtargeting.
This limited use mitigates the broad impact of increased campaign microtargeting and
hides its macro effects. Extensive use of microtargeting would probably lead to more voters
identifying as strong partisans as well as to increased shifts in party affiliation in the middle
of the partisan spectrum. Campaigns would target voters with experiences congruent to
their current party affiliation; this issue emphasis would lead voters to hold their party in
even higher esteem.
1
American politics has experienced increased polarization since the
advent of microtargeting tools, but other factors are certainly at play as well (McCarty
et al., 2006).
In contrast, voters who have experience on an issue and hold a view at odds with their
party are likely to defect if microtargeted. As campaigns generate more precise databases
of voters, these “cross-pressured partisans” (Hillygus and Shields, 1991) and independents
are likely to be persuaded to one side or another depending on the issues that arise during a
given campaign season. Viewing this trend over time would most likely require consistent,
multi-election panel data, which is generally not available. Consistent with these two large-
scale implications is that the proportions of Americans identifying as strong partisans and
independents are both above their historical averages (ANES data, 2000-2008).
1
Parties might microtarget their own partisans to increase turnout in low-salience elections or to en-
courage increased participation in party politics.
177
If microtargeting dominated macro trends, then strong partisans would be more loyal
on Election Day as parties would be able to find a congruent issue position to prevent de-
fection. The data are also consistent with this prediction, as the effect of strong partisans
on presidential vote choice is at a historical high (ANES data, 2000-2008, compared with
Bartels, 2000). The implication for partisan leaners is less clear as their party affiliation
might move in tandem with vote preference changes. These synchronized shifts are most
likely during a presidential election, which is also when the most resources are available
to microtarget, further reducing the probability of observing a macro trend. Overall, the
macro data are consistent with the implications of increased microtargeting, but microtar-
geting’s contribution to these trends cannot be teased out and probably is small.
As for democracy as an institution, the effects of microtargeting are mixed. Voters re-
ceive more information that they can use to evaluate politicians, but these targeted appeals
might skew their understanding of other, potentially more important, issues. Microtarget-
ing is neither good nor bad; rather, it is a tool that can be wielded for many purposes.
The Personal Experience Model and subsequent studies quantify just how powerful a tool
microtargeting is and why it works.
178
Appendix A
Appendix for Chapter 2
A.1 Details for Candidate Evaluation Simulation
The following tables display the initial parameter values of the simulation in Section 2.1.5.
Voters’ issue opinions, beliefs of candidate stances, and candidate evaluation are updated
each day of the hypothetical campaign using the formulae of the Personal Experience
Model.
Voters’ Prior Issue Opinions: Mean (Precision)
Voter Traits Health Care Education
Alice the Architect 0.95 (10,000) 0 (6.25) 0 (625)
Ted the Teacher 0.95 (10,000) 0 (6.25) 0.5 (100)
Doris the Doctor 0.95 (10,000) 0.5 (400) 0 (625)
Table A.1: The mean (δ
1
) and precision (τ
2
1
), in parentheses, of voter’s prior beliefs on
issues and traits.
The voters’ initial issue opinions are displayed in Table A.1. The precision on traits
are extremely high because voters are sure that they want high valence (e.g., competent,
179
trustworthy, shared-values) politicians. When occupation and policy intersect (e.g., Ted
the teacher and education) the voter’s precision on this issue is higher than otherwise.
Voter’s Prior Beliefs about Candidate Mandy’s Positions
Voter Traits Health Care Education
Alice the Architect 0.4 (6.25) 0 (6.25) 0 (6.25)
Ted the Teacher 0 (4) 0 (6.25) 0 (6.25)
Doris the Doctor 0.3 (4) 0 (6.25) 0 (6.25)
bottomrule
Voter’s Prior Beliefs about Candidate Nathan’s Positions
Voter Traits Health Care Education
Alice the Architect 0 (4) 0 (6.25) 0 (6.25)
Ted the Teacher 0.4 (6.25) 0 (6.25) 0 (6.25)
Doris the Doctor 0.3 (4) 0.5 (6.25) 0 (6.25)
Table A.2: The mean (µ
1
) and precision (ν
2
1
), in parentheses, of voter’s prior beliefs on
issues and traits.
Candidate Signals
Candidate Traits Health Care Education
Mandy No signal -0.5 (1.5) 0.5 (1.5)
Nathan No signal 0.5 (1.5) -0.5 (1.5)
Table A.3: The mean (γ
1
) and precision (ψ
2
1
), in parentheses, of the candidate’s signals
about where they stand on the issues.
The precisions are low because I simulation that the topic of conversation about each
issue lasts 10 days and that one signal is received by the voters each day from their preferred
candidate. Alternatively, I could have simulated a higher precision and an additional
parameter representing the chance that the voter receives a signal on a day that issue is
discussed. (If a voter prefers the candidates equally, there is a 50% chance of receiving the
signal from one candidate or the other.
At the beginning of the campaign, voters know nothing about the candidates’ issue
position and only care about traits. During the second phase of the campaign, a discussion
180
Per-Day Change in Issue Weight
Campaign Phase Traits Health Care Education
Phase 1 (day 1) +1.0 0 0
Phase 2 (days 2-10) 0 +0.02 -0.03
Phase 3 (days 11-20) No signal -0.02 +0.03
Table A.4: The weight (δ
j
) placed on each issue.
about health care begins, and voters add considerations about health care to their candidate
evaluations. The third and final phase of the campaign sees the conversation shift to
education to the detriment of health care. Weights are forced to be nonnegative.
181
Appendix B
Appendix for Chapter 3
182
B.1 Regressions for the Cue-Taking Hypothesis
Soc. Sec. Soc. Sec. PBR
Variable GOp Conv. Dem Conv. Dem. Conv. Scale
Experience 1 (0.3) 0.6 (0.3) -0.1 (0.3) 2
Experience*Opinion -0.5 (0.4) 0.7 (0.4) 0.5 (0.4)
Income*Opinion 0.8 (0.7) 2 (0.7) 2 (0.7)
Conv. Watch*Opinion 0.1 (0.4) -0.5 (0.4) 0.2 (0.5)
Pol. Int.*Opinion 0.5 (0.7) -0.1 (0.7) 0.4 (0.8)
Party*Opinion -0.2 (0.5) 0.7 (0.5) -0.4 (0.5)
Pre-Conv. Opinion 2 (0.6) 2 (0.6) 1 (0.7) 2
Income -0.5 (0.5) -2 (0.5) 0.4 (0.6) 9
Convention Watched -0.4 (0.3) 0.4 (0.3) -0.3 (0.4) 5
Political Interest -0.8 (0.5) 0.3 (0.5) -0.3 (0.6) 3
Party Affiliation 1 (0.4) 0.7 (0.3) 0.9 (0.4) 7
Female 0.2 (0.2) 0.3 (0.2) 0.2 (0.2) 2
Constant -2 (0.5) -2 (0.4) -0.7 (0.6)
n 1177 1209 1213
Table B.1: Evaluating the Cue-Taking Hypothesis with logistic regressions of post-
convention opinion regressed on pre-convention opinion, experience and interaction terms.
Coefficients with standard errors in parentheses are reported. All variables are on a zero-
to-one scale. The scale column indicates how many points are on the scale. For issues:
Democratic positions is unity. Bolded coefficients are the coefficients of interest; the Cue-
Taking Hypothesis predicts a negative coefficient for experience and a positive coefficient
for the experience interaction. As mentioned in the text, the cue-taking moderation ability
of HMO experience does not extend to Social Security: the main experience coefficient is
zero and the interaction term coefficient is negative.
183
B.2 Regressions for the Candidate Evaluation Hypothesis
Soc. Sec. Soc. Sec. PBR
Variable GOp Conv. Dem Conv. Dem. Conv.
Experience 0.2 (0.4) -0.1 (0.1) -0.4 (0.3)
Experience*Opinion 0.8 (0.6) 0.04 (0.1) 0.5 (0.3)
Income*Opinion 0.7 (0.9) -1.2 (0.4) 0.1 (0.5)
Conv. Watch*Opinion 0.2 (0.5) -0.3 (0.3) -0.5 (0.3)
Pol. Int.*Opinion 0.5 (0.9) 1.1 (0.4) 0.4 (0.6)
Party*Opinion 1.6 (0.6) -0.1 (0.3) 0.2 (0.4)
Pre-Conv. Opinion -2.1 (0.8) 0 (0.2) -0.4 (0.5)
Income -0.7 (0.5) 0.2 (0.2) -0.3 (0.5)
Convention Watched -0.5 (0.3) 0.5 (0.1) 0.8 (0.3)
Political Interest 0.5 (0.5) -0.9 (0.2) -0.9 (0.5)
Party Affiliation -0.5 (0.4) 0.2 (0.2) -0.2 (0.4)
Female 0.3 (0.2) 0.1 (0.1) 0.1 (0.1)
n 888 917 1,003
Table B.2: Evaluating the Candidate Evaluation Hypothesis with ordinal probit regressions
of post-convention opinion regressed on pre-convention opinion, experience and interaction
terms. Coefficients with standard errors in parentheses are reported. All variables are on
a zero-to-one scale.For issues: Democratic positions is unity. Bolded coefficients are the
coefficients of interest; the Candidate Evaluation Hypothesis predicts a negative coefficient
for experience and a positive coefficient for the experience interaction. Regression restricted
to voters who do not alter their issue opinion during the convention.
184
Variable Age < 65 Age ≥ 65
Pre-convention Opinion 0.05 (0.46) 0.54 (0.45)
Age -0.02 (0.03) 0.05 (0.04)
Age*opinion -0.003 (0.05) -0.02 (0.05)
DV = -2 [ DV = -1 -1.8 (0.3) na
DV = -1 [ DV = 0 -1.1 (0.3) -1.1 (0.3)
DV = 0 [ DV = 1 1.9 (0.3) 2.2 (0.4)
DV = 1 [ DV = 2 2.8 (0.4) na
n 220 135
Table B.3: Regression discontinuity analysis: Social Security and the Republican con-
vention. Ordinal probit regressions of voters who do not change their position on Social
Security privatization during the convention. Coefficient of interest bolded. Dependent
variable is change in vote (Bush, Gore, or undecided) before and after the Republican
convention. The coefficient of interest is the effect of support for privatization on change
in vote, which is higher in the second regression of seniors. Age is actual age minus 65,
the cutoff for treatment, so that the coefficients of interest are comparable. The regression
is restricted to those between the ages of (50,85)—lack of vote switchers precludes further
restriction of the sample closer to the cut point. No seniors switched their vote more than
one scale-point; hence, the na’s for two of the ordinal cut-offs.
185
Appendix C
Appendix for Chapter 4
C.1 Question Wording in Nationwide Survey
C.1.1 Candidate Descriptions
First, I’d like to get your feelings toward people – some real and some hypothetical. For
each, please rate that candidate using something we call the feeling thermometer. Ratings
between 51 degrees and 100 degrees mean that you feel favorable and warm toward the
person. Ratings between 0 and 49 degrees mean that you don’t feel favorable toward the
person and that you don’t care too much for that person. You would rate the person at 50
if you don’t feel particularly warm or cold toward the person. Use the slider (with either
your mouse or keyboard) to indicate your rating.
Then, on a second slider, I’d like you to rate your certainty of that feeling from 0
to 10. If you are very uncertain about your feeling, move the slider downward toward 0. If
you are perfectly certain about your feeling, move the slider upward toward 10.
• Former Governor Richard Miller has been nominated as the Democratic United
States Senate candidate for your state.
186
Gov. Richard Miller grew up in a working-class family, won a scholarship to a top
college, is now 52, is married and has two children. He served eight years as the state’s
governor, during which time he won re-election by a large margin. Gov. Miller won
praise for crafting an innovative health care policy, promoting economic growth that
outpaced the national average, and working well with state leaders of the opposite
party. He is endorsed by the nurses’ unions and several environmental organizations.
Republican leaders, however, claim he will raise taxes too much.
• Attorney General Mark Jones has been nominated as the Republican United States
Senate candidate for your state.
Attorney General Mark Jones grew up in a small town, graduated top of his class
in law school, and began his career in the local U.S. District Attorney’s office. After
rising quickly in the ranks to head that office, he was elected Attorney General.
As Attorney General, he reduced violent crime in the state and cracked down on
corrupt politicians in the state capital; he easily won re-election. He is 62, married,
has three children and five grandchildren. Jones is endorsed by the local farmers’
organization and the Fraternal Order of Police. Democratic leaders, however, claim
that his policies are too biased toward big business.
C.1.2 Candidates’ Issue Signals
• America has always been a nation of immigrants and today’s hard-working immi-
grants deserve the chance to achieve the American dream.
• Immigrants who entered the country illegally clearly have no respect for the rule of
law and must be sent back to their country of origin.
187
• Not only is Iraq a distraction from dangerous terrorists in other countries, but the
U.S. leaving Iraq will finally force the Iraqi government to take responsibility for their
county.
• Not only does staying in Iraq provide America security against terrorists, but America
also has a duty to see the Iraq situation through to a peaceful conclusion.
• We should recognize same-sex marriage so that all Americans are treated equally.
America has long outgrown its history of discrimination and we need to extend mar-
riage rights to all citizens.
• Marriage has always been between one man and one woman, and passing a consti-
tutional amendment affirming that principle ensures that the courts cannot override
the will of the people.
C.1.3 Experience Questions
• Next I’m going to ask about the characteristics of these people with whom you
discuss important matters. Think about the proportion of important discussion you
have with any person that fits the description below.
– About what proportion (percent) of your important discussions is with a person
who currently serves in the U.S. military?
– About what proportion (percent) of your important discussions is with a person
who once served in the U.S. military?
– About what proportion (percent) of your important discussions is with a person
who is an immigrant to the United States?
– About what proportion (percent) of your important discussions is with a person
who is Hispanic?
188
– About what proportion (percent) of your important discussions is with a person
who is gay or lesbian?
• Including yourself, has anyone living in your house or apartment ever served in the
U.S. military? This includes the National Guard and the Reserves. Check all that
apply.
• How often do you attend religious services, apart from special events like weddings
and funerals?
• If you are Christian: Would you say you have been ”born again” or have had a
”born again” experience, that is, a turning point in your life when you committed
yourself to Christ?
• Regardless of whether you now attend any religious services, do you ever think of
yourself as part of a particular church or denomination? Which one?
189
C.2 Auxiliary Regressions
Issue
Inter. Var. Immigration Iraq War Gay Marriage
No Controls
Issue Experience 0.24 (0.20) 0.15 (.20) 0.28 (0.22)
With Controls
Issue Experience 0.22 (0.20) 0.15 (.20) 0.09 (0.22)
Political Attentiveness 0.92 (.76) 1.4 (0.74) 2.9 (0.8)
Self-Interest – -1.7 (1.0) 1.7 (1.0)
With Certainty Mediator
Issue Experience 0.22 (0.20) 0.13 (.20) 0.09 (0.27)
Political Attentiveness 0.95 (.76) 1.4 (0.74) 2.9 (0.8)
Self-Interest – -1.7 (1.0) 1.4 (1.2)
Issue Certainty 0.59 (0.52) 0.9 (0.58) -0.03 (0.7)
Anti-Partisan Bias Signals Only
Issue Experience 0.19 (0.28) 0.15 (.20) -0.07 (0.22)
Political Attentiveness 1.6 (1.1) 2.0 (1.0) 3.0 (0.8)
Self-Interest – -0.6 (1.4) 2.6 (1.8)
Table C.1: OLS regressions of shift in candidate evaluation. Dependent variable’s potential
range is -100 to +100. Signal incongruity is the distance (magnitude) of the candidate’s
signal (either -10 or +10) to the respondent’s wave one position on that issue. All variables
listed are interacted with signal congruity, meaning the expected sign is positive. Main
effects are not shown; they are insignificant for all regressions. Political attentiveness and
issue certainty (from wave one) are measured on a 0 to 1 scale. Self-interest is dichotomous.
Top and bottom sets of coefficients represent two different regressions. N is about 270 for
each regression.
C.3 Issue Experience Measures and the Hypotheses
190
Issue
Exp. Var. Immigration Iraq War Gay Marriage
Without Controls
Issue Experience -0.54 (0.34) -0.50 (0.35) -0.84 (0.35)
Constant 0.72 (0.37) 0.41 (0.36) 0.45 (0.34)
With Controls
Issue Experience -0.47 (0.35) -0.56 (0.36) -0.94 (0.42)
From favored cand. 0.63 (0.74) -0.39 (0.73) 0.34 (0.69)
Attentiveness -0.24 (0.26) 0.26 (0.26) -0.38 (0.23)
Self interest – -0.26 (1.7) 1.3 (1.9)
Constant 1.2 (0.96) -0.21 (0.93) 1.4 (0.83)
With Certainty Mediator
Issue Experience -0.48 (0.35) -0.39 (0.73) -0.93 (0.42)
From favored cand. 0.65 (0.74) -0.41 (0.74) 0.47 (0.69)
Attentiveness -0.26 (0.26) 0.25 (0.26) -0.37 (0.24)
Self interest – -0.35 (1.7) 1.4 (1.9)
Certainty 0.10 (0.15) 0.09 (0.15) -0.02 (0.13)
Constant 0.47 (1.5) -0.89 (1.5) 1.4 (1.2)
Table C.2: OLS regressions of issue opinion shift in the direction of the delivered signal.
Dependent variable’s potential range is -20 to +20. “From favored cand.” is a dichotomous
variable indicating whether the signal is delivered by the candidate that the respondent
favored in the first wave. Political attentiveness and issue certainty (from wave 1) are
measured on a 0 to 1 scale. Self-interest is dichotomous. Top and bottom sets of coefficients
represent two different regressions. N is about 270 for each regression.
191
Group Congruity:Exp. Against Bias
Hispanic 0.57 (.59) 0.79 (0.80)
... in workplace (26%) -.32 (.16) -0.2 (0.2)
... in zipcode (31%) 0.0 (0.4) 0.14 (0.63)
Talk to hisp (32%) 0.46 (0.44) 0.35 (0.61)
Serve(d) in Mil. -0.16 (.56) -0.17 (0.75)
HH Served(s) 0.39 (.074) 0.40 (0.4)
Talk to vets/mil (33%) 0.13 (.44) 0.42 (0.59)
Resp is GLBT 1.5 (1.0) 2.3 (1.3)
Talks to GLBTs (30%) 0.6 (0.5) 0.33 (0.69)
Table C.3: OLS regression coefficients for various experience measures. (See Table C.1 for
regression details; includes control for attentiveness.) The first column lists the experience
measure and the percentage of the population included in the measure if there is not
a self-evident boundary. The second column is the regression coefficient (and standard
error), with all respondents who received a signal on that issue included (n is about 270).
The third column is the same regression as column two but only among respondents who
received a signal that went against their partisan predisposition (n is about 135).
192
Group Shift in Direction of Signal
Hispanic -0.9 pts (n.s)
... in workplace (26%) -1.1 pts (n.q.s)
... in zipcode (31%) no diff.
Talk to hisp (32%) -1.4 (p < .05)
Serve(d) in Mil. -0.1 (n.s.)
HH Served(s) -0.6 pts (n.q.s)
Talk to vets/mil (33%) -0.2 pts (n.s.)
Resp is GLBT -1.1 pts (n.q.s)
Talks to GLBTs -1.1 pts (p < 0.1)
Table C.4: Number of points in the direction of the signal (or cue) respondents shifted when
compared with individuals not in the group listed. The Cue-Taking hypothesis predicts
these values to be negative. When the values are significant at conventional levels, p-
values are listed, n.s. means “not significant,” and n.q.s means “not quite significant at
conventional levels.”
193
194
Thank you very much for taking my survey, which should only take 5 minutes of your
time.

Your participation is completely voluntary and you may withdraw your consent and
discontinue participant of the project at any time. Your refusal to participate will not
result in any penalty. This study has been approved by the University's Institutional
Review Panel for Human Subjects. If you have questions about the project please email
Aaron Strauss (abstraus@princeton.edu). For answers to any questions you may have
about your rights as a research subject, contact Joseph Broderick, Secretary,
Institutional Review Panel for Human Subjects (609.258.3976).

If you complete the survey you will be entered in a random drawing for a $100 gift
certificate to amazon.com or iTunes (your choice!).

The survey is on multiple pages; answer the first question below and click the "Next"
button below to navigate through the pages.
1. What is your class year at Princeton?
2011 (Freshman).......................................................... 19
2010 (Sophomore) ........................................................ 20
2009 (Junior) ............................................................... 25
2008 (Senior) ............................................................... 36

[Candidates Rotated Below] [Candidates Rotated Below] [Candidates Rotated Below] [Candidates Rotated Below]
Below are descriptions of two hypothetical candidates running for United States Senate.
For each, please rate that candidate using something we call the feeling scale. Positive
ratings between 1 and 10 mean that you feel favorable and warm toward the person,
with 10 being the warmest rating. Negative ratings between -1 and -10 mean that you
don't feel favorable toward the person and that you don't care too much for that person,
with -10 the coolest rating. You would rate the person at 0 if you don't feel particularly
warm or cold toward the person.

195
2. Former Governor Richard Miller has been nominated as the Democratic Democratic Democratic Democratic United
States Senate candidate for your state.

Gov. Miller grew up in a working-class family, won a scholarship to a top
college, is now 52, married and has two children. He served eight years as the
state’s governor, during which time he won re-election by a large margin. Gov.
Miller won praise for crafting an innovative health care policy, promoting
economic growth that outpaced the national average, and working well with
state leaders of the opposite party. He is endorsed by the state’s farmers
organization as well as the National Organization for Women.
-10: most negative................................ 0 +1: ................................................................ 3
-9: ................................ 0 +2: ................................................................ 6
-8: ................................ 0 +3: ................................................................ 6
-7: ................................ 0 +4: ................................................................ 6
-6: ................................ 0 +5: ................................................................ 15
-5: ................................ 0 +6: ................................................................ 11
-4: ................................ 1 +7: ................................................................ 15
-3: ................................ 1 +8: ................................................................ 14
-2: ................................ 1 +9: ................................................................ 7
-1: ................................ 2 7 +10: most positive................................ 7 89
0: neutral ................................ 4 Refused................................ 1

3. Retired Admiral Mark Jones has been nominated as the Republican Republican Republican Republican United
States Senate candidate for your state.

Admiral Jones grew up in a small town, graduated from the Naval Academy,
and climbed the ranks of the Navy quickly. He is 62, married, has three children
and five grandchildren. After retiring from service, he served as Secretary of
State of a Republican presidential administration. During his tenure as
Secretary, Admiral Jones won accolades for achieving peace in several long-
standing conflicts. Admiral Jones is endorsed by several local newspapers and
the Fraternal Order of Police.
-10: most negative................................ 0 +1: ................................................................ 4
-9: ................................ 0 +2: ................................................................ 11
-8: ................................ 0 +3: ................................................................ 8
-7: ................................ 1 +4: ................................................................ 11
-6: ................................ 0 +5: ................................................................ 11
-5: ................................ 1 +6: ................................................................ 10
-4: ................................ 2 +7: ................................................................ 7
-3: ................................ 3 +8: ................................................................ 6
-2: ................................ 4 +9: ................................................................ 3
-1: ................................ 4 16 +10: most positive................................ 3 75
0: neutral ................................ 9 Refused................................ 0
196

Prefer Democratic Candidate....................................... 63
Prefer Republican Candidate ....................................... 20
Prefer Equally.............................................................. 15
Refused on either evaluation........................................ 1


Next, please consider the following political issues.

Read the following two statements and again rate them on a scale similar to that of the
last two questions. If you agree with Statement A you have ten choices, from "A:1"
which means you slightly agree with Statement A all the way to "A:10" which means
you completely agree with Statement A. Similarly, if you agree with Statement B pick a
value between "B:1" (slightly agree with B) and "B:10" (completely agree with B).
Choose "0, Neutral" if you are unsure or agree with the two statements equally.

[Issue order rotated. A/B messages rotated] [Issue order rotated. A/B messages rotated] [Issue order rotated. A/B messages rotated] [Issue order rotated. A/B messages rotated]
4. With regard to U.S. foreign aid, do you think the federal government should:

A) A) A) A) Cut foreign aid in half because it is ineffective and costs taxpayers over 25
billion dollars a year.

B) B) B) B) Double foreign aid since it accounts for less than one percent of the federal
budget and helps those in need.
A:10, completely
agree w/ A ................................ 2
B:1, slightly
agree w/ B ................................ 5
A:9................................ 0 B:2................................................................ 7
A:8................................ 1 B:3................................................................ 9
A:7................................ 0 B:4................................................................ 8
A:6................................ 2 B:5................................................................ 6
A:5................................ 2 B:6................................................................ 7
A:4................................ 3 B:7................................................................ 5
A:3................................ 3 B:8................................................................ 4
A:2................................ 3 B:9................................................................ 2
A:1, slightly
agree w/ A ................................ 5 21
B:10, completely
agree w/ B ................................ 6 59
0: neutral ................................ 20 Refused................................ 0


197
5. On a different issue, do you think the federal federal federal federal government should:

A) A) A) A) Encourage businesses to invest in new, more energy efficient technology,
while letting the market shape the actions of businesses.

B) B) B) B) Implement a cap-and-trade system that would restrict the amount of
greenhouse gases released by businesses to a government-mandated level.
A:10, completely
agree w/ A ................................ 5
B:1, slightly
agree w/ B ................................ 2
A:9................................ 3 B:2................................................................ 5
A:8................................ 4 B:3................................................................ 5
A:7................................ 5 B:4................................................................ 4
A:6................................ 5 B:5................................................................ 8
A:5................................ 7 B:6................................................................ 6
A:4................................ 5 B:7................................................................ 6
A:3................................ 4 B:8................................................................ 5
A:2................................ 2 B:9................................................................ 2
A:1, slightly
agree w/ A ................................ 4 42
B:10, completely
agree w/ B ................................ 7 49
0: neutral ................................ 8 Refused................................ 1

198
6. A state's Supreme Court state's Supreme Court state's Supreme Court state's Supreme Court recently mandated that the state government must
legalize same-sex unions, either through civil unions or gay marriage.
In your opinion, should the state:
A) A) A) A) Implement civil unions--a less divisive solution that provides equal rights to
same-sex couples.

B) B) B) B) Recognize gay marriage; anything less would be discriminatory against gays.
A:10, completely
agree w/ A ................................ 11
B:1, slightly
agree w/ B ................................ 4
A:9................................ 1 B:2................................................................ 3
A:8................................ 2 B:3................................................................ 2
A:7................................ 2 B:4................................................................ 3
A:6................................ 3 B:5................................................................ 5
A:5................................ 4 B:6................................................................ 2
A:4................................ 3 B:7................................................................ 4
A:3................................ 4 B:8................................................................ 4
A:2................................ 3 B:9................................................................ 3
A:1, slightly
agree w/ A ................................ 4 36
B:10, completely
agree w/ B ................................ 25 55
0: neutral ................................ 10 Refused................................ 0


7. On average, how many times a month have you attended a spiritual or religious
event or service (e.g., church, synagogue, mosque, meditation) during Fall term?
Never ........................................................................... 40
Less than once a month............................................... 23
Once a month.............................................................. 6
Twice a month............................................................. 6
Four times a month..................................................... 15
More often than weekly ............................................... 8
Refused........................................................................ 1

199
8. Over the past twelve months, how many outdoor trips or excursions have you
gone on (e.g., camping, hiking, ski cabin, quiet beach)?
0: ................................................................................. 10
1: ................................................................................. 10
2: ................................................................................. 17
3: ................................................................................. 15
4: ................................................................................. 10
5: ................................................................................. 9
6: ................................................................................. 6
7: ................................................................................. 3
8: ................................................................................. 2
9: ................................................................................. 1
10 or more ................................................................15
Refused........................................................................ 1

9. In the past five years, about how much cumulative time have you spent outside
of the U.S., Canada, Mexico, and western Europe (e.g., South America, Asia,
Africa, eastern Europe)?
None ............................................................................ 21
One week................................................................ 6
Two weeks ................................................................ 6
Three or four weeks ..................................................... 8
One month................................................................ 7
Two months................................................................ 15
Three to size months ................................................... 19
Seven months to a year................................................ 5
More than one year...................................................... 12
Refused........................................................................ 1

10. Do you have friends or family members who are gay, lesbian, bisexual or
transgender?
No................................................................................ 28
Yes .............................................................................. 72
Refused........................................................................ 0

200
11. If so, how often do you communicate with these individuals? (Feel free to check
multiple frequencies below if they are applicable to different friends and/or
family.)
Never ........................................................................... 1
On occasion ................................................................ 16
Monthly....................................................................... 12
Weekly......................................................................... 11
Two or three times a week........................................... 11
Daily............................................................................ 21

Answered “No” in previous question............................. 28
Refused in either question............................................ 1

12. Generally speaking, do you think of yourself as a Republican, a Democrat, an
Independent, or another party? And how strong is your affiliation? If you are an
Independent, do you tend to favor (i.e., lean to) Republican candidates or
Democratic candidates?
Strong Republican ....................................................... 5
Not so strong Republican............................................. 7
Independent, lean Republican...................................... 10
Independent................................................................ 16
Independent, lean Democrat ........................................ 24
Not so strong Democrat............................................... 24
Strong Democrat.......................................................... 14
Other party ................................................................ 0
Refused........................................................................ 0


13.
When it comes to politics, do you usually think of yourself as:
Extremely Conservative............................................... 2
Conservative................................................................ 10
Slightly conservative.................................................... 11
Moderate; middle of the road....................................... 17
Slightly liberal ............................................................. 21
Liberal ......................................................................... 32
Extremely liberal ......................................................... 6
Other party ................................................................ 0
Refused........................................................................ 0



201

14.
Please indicate your sex:
Female......................................................................... 54
Male ............................................................................ 45
Refused........................................................................ 1

15.
Please Enter your age in the box to the right
17-18:........................................................................... 28
19: ............................................................................... 27
20: ............................................................................... 19
21: ............................................................................... 18
22-24:........................................................................... 6
Refused........................................................................ 1

16. Compared to other Princeton students, how much to do you pay attention to
political news?
A great deal ................................................................ 7
Quite a bit................................................................20
Some............................................................................ 39
Very little ................................................................ 28
Not at all ................................................................ 6
Refused........................................................................ 0

202


Wave Two Candidate Signals Wave Two Candidate Signals Wave Two Candidate Signals Wave Two Candidate Signals


[Respondent receives two signals on distinct issues.] [Respondent receives two signals on distinct issues.] [Respondent receives two signals on distinct issues.] [Respondent receives two signals on distinct issues.]
Signals: Signals: Signals: Signals:

Gay Marriage Gay Marriage Gay Marriage Gay Marriage
• Marriage has always been between one man and one woman. Civil unions provide
the same rights to same-sex couples without alienating the more traditional
segments of the population.
• We should recognize same-sex marriage so that all Americans are treated equally.
America has long outgrown its era of discrimination and we need to extend
marriage rights to all citizens.

Environment Environment Environment Environment
• To combat global warming, we need a cap-and-trade system that will restrict the
amount of dangerous greenhouse gasses released into the atmosphere. Businesses
will always try to cut costs; without tight restrictions, our environment will
continue to suffer.
• A market-based solution to climate change is the best way to ensure the health of
the planet while maintaining a strong economy. Innovative, eco-friendly
businesses will prosper, leading to a healthy environment and continued job
growth.

Foreign Aid Foreign Aid Foreign Aid Foreign Aid
• America has the world’s largest economy by far, and as global leaders we have the
responsibility to assist those less fortunate than us.
• Foreign aid is wasted on corrupt foreign governments. We should spent taxpayers
money in America, where we can better evaluate its effectiveness.

Appendix D
Appendix for Chapter 5
D.1 Computational Appendices
This appendix presents fast and approximate solutions to the planner’s optimization prob-
lem in both nonpartisan and partisan cases.
D.1.1 Nonpartisan Case: the Knapsack Problem
To approximate the solution to the nonpartisan planner’s optimization problem defined in
Section 5.3.1, the key is to notice that the linear optimization problem is identical to the
canonical knapsack problem. In the knapsack problem, one maximizes the total value of
objects to be placed in a knapsack of fixed sized, with each object having its own value and
size. The analogous case for the nonpartisan planner is maximizing the number of voters
given a budget constraint where each individual-treatment pairing may be thought of as
an object.
Following Dantzig (1957), the exact solution of this linear programming problem is
approximated by ordering the individual pairs by their maximum vote per dollar ratio and
203
treating the individuals with the highest such ratio first until the budget is exhausted. If
the ratio is non-positive (i.e., the best non-control treatment for an individual does not
outperform the control), this individual is not treated. In most cases, this approximation
yields solutions very close to the optimal result because the ratio of the per-use cost of the
most expensive treatment (e.g., $15 for a canvassing shift) is tiny compared to the overall
budget (usually at least $10,000). Thus, when the addition of an expensive and efficient
treatment runs just over budget and a cheaper yet less efficient tactic should be used in its
place, inefficiencies at the edge of the problem are of little importance.
D.1.2 Partisan Case: the Stochastic Knapsack Problem
To derive a fast and approximate solution to the partisan’s optimization problem defined
in Section 5.4.3, the key is to notice that this optimization problem is identical to the
stochastic knapsack problem, in which the probability is maximized that the total value of
items in the knapsack equals or exceeds a target value where each object has a random
value and a known size. As in the nonpartisan case, each individual-treatment pair can be
considered an item.
As an approximate solution to this problem, an algorithm based on Geoffrion (1967)
is used; subgroups are ordered by the weighted combination of the mean and standard
error of their posterior vote choice profile, π(ρ). Optimization is performed over the weight
parameter, which can take values between 1 (i.e., only the means of the posteriors matter)
and 0 (i.e., only the standard errors matter). (For a discussion of when this approximation
fails to yield the optimal result, see Henig (1990)). The intuition behind this algorithm can
be developed by considering the following scenarios. Campaigns with a natural advantage
(i.e., they would garner a majority of the vote without treatment) could further increase
their probability of winning by contacting voters who are highly responsive on average
204
and have a low variance of their treatment response. On the other hand, campaigns that
are behind aim to treat segments of the population that are both highly responsive and
have high variance. Thus, unlike in the nonpartisan case, the optimal subgroups to treat
change depending on the outcome under the control. The algorithm finds an approximate
solution by limiting its search to the subspace defined by the weight parameter, which
makes optimization feasible when the dimension of δ is large.
205
Appendix E
Appendix for Chapter 6
E.1 Analytical Solution for Variance of Vote Share
The general formula for the variance of a random variable X that is composed of two
Gaussian distributions that occur with probability p and 1 − p respectively is derived.
For an example, Figure E.1 shows the distribution with p = 0.5 and sub-distributions of
^(1, 0.16) and ^(−1, 0.16).
To determine the variance of X, let Y be a Bernoulli random variable that determines
which Gaussian produces x. If the two normal distributions that compose X are ^(µ
1
, σ
2
1
)
and ^(µ
2
, σ
2
s
), then
Var(X) = Var(E[X[Y ]) +E[Var(X[Y )] (E.1)
Var(E[X[Y ]) = p(µ
1

¯
X)
2
+ (1 −p)(µ
2

¯
X)
2
(E.2)
E[Var(X[Y )] = pσ
2
1
+ (1 −p)σ
2
2
(E.3)
Var(X) = p[(µ
1

¯
X)
2

2
1
] + (1 −p)[(µ
2

¯
X)
2

2
2
], (E.4)
206
−3 −2 −1 0 1 2 3
0
.
0
0
.
1
0
.
2
0
.
3
0
.
4
0
.
5
Hypothetical Random Variable X
P
r
o
b
a
b
i
l
i
t
y

D
e
n
s
i
t
y
Figure E.1: Hypothetical distribution of a variable composed of two Gaussian distributions
with a Bernoulli process arbiter.
where
¯
X = pµ
1
+ (1 − p)µ
2
. Each sub-distribution contributes its own variance and the
squared distance to the mean to the overall variance, in proportion to the probability that
the distribution is activated by the Bernoulli probability.
This derivation is important because the distribution of the vote function is the su-
perposition of four normal curves, each of which occurs with a certain probability. There
are four distributions because there are four possible outcomes, as both the Democratic
and Republican broadcasted issues have independent probabilities of either resolving or
not resolving.
The mean of the distribution is,
¯
V
D

D
, δ
R
) = µ +m(δ
D
−δ
R
)
207
The variance of V
i
is a weighted average of the squared distance of the four normal
distributions that constitute V
i
to
¯
V
i
. Using the derivation above, the variance is,
Var(V
D

D
, δ
R
)) = p
D

2
+ (q(2 −δ
D
−δ
R
))
2
)
+p
R

2
+ (q(2 −δ
D
−δ
R
))
2
)
+(1 −p
D
−p
R
)(σ
2
)
combining terms and assuming p
D
= p
R
= p
= σ
2
+ 2pq
2
(2 −δ
D
−δ
R
)
2
208
Bibliography
Achen, C. and Bartels, L. (2006). It feels like we’re thinking: The rationalizing voter and
electoral democracy. In Annual Meeting of the American Political Science Association,
Philadelphia.
Achen, C. H. and Bartels, L. M. (2005). Partisan hearts and gall bladders: Retrospection
and realignment in the wake of the great depression. Presented at 2005 the Annual
Meeting of the Midwest Political Science Association .
Ansolabehere, S., Iyengar, S., Simon, A., and Valentino, N. (1994). Does attack advertising
demobilize the electorate? The American Political Science Review 88, 4, 829–838.
Ansolabehere, S., Rodden, J., and Snyder, J. M. (2008). The strength of issues: Using
multiple measures to gauge preference stability, ideological constraint, and issue voting.
The American Political Science Review 102, 2, 215–232.
Arceneaux, K. and Kolodny, R. (2009). Educating the least informed: Group endorsements
in a grassroots campaign. American Journal of Political Science 53, 4.
Bartels, L. M. (1993). Messages received: The political impact of media exposure. The
American Political Science Review 87, 2, 267–285.
209
Bartels, L. M. (1996). Uninformed votes: Information effects in presidential elections.
American Journal of Political Science 40, 1, 194–230.
Bartels, L. M. (2000). Partisanship and voting behavior, 1952–1996. American Journal of
Political Science 44, 1, 35–50.
Bartels, L. M. (2002). Beyond the running tally: Partisan bias in political perceptions.
Political Behavior 24, 2, 117–150.
Bartels, L. M. (2005). Homer gets a tax cut: Inequality and public policy in the american
mind. Perspectives on Politics 3, 1, 15–31.
Berger, J. O. (1985). Statistical Decision Theory and Bayesian Analysis. Springer, New
York, 2nd edn.
Blais, A. and Dobrzynska, A. (1998). Turnout in electoral democracies. European Journal
of Political Research 33, 2, 239–262.
Breiman, L., Friedman, J., Stone, C. J., and Olshen, R. (1984). Classification and Regres-
sion Trees. Chapman & Hall, London.
Brody, R. A. and Page, B. I. (1972). Comment: The assessment of policy voting. The
American Political Science Review 66, 2, 450–458.
Campbell, A., Converse, P., Miller, W., and Stokes, D. (1960). The American Voter. John
Wiley & Sons Inc., New York.
Campbell, J. E. (2005). The fundamentals in us presidential elections: Public opinion, the
economy and incumbency in the 2004 presidential election. Journal of Elections, Public
Opinion & Parties 15, 1, 73–83.
210
Campell, J. E. (1992). Forecasting the presidential vote in the states. American Journal
of Political Science 36, 2, 386–407.
Canes-Wrone, B., Herron, M. C., and Shotts, K. W. (2001). Leadership and pandering: A
theory of executive policymaking. American Journal of Political Science 45, 3, 532–550.
Caplan, B. (2008). The Myth of the Rational Voter: Why Democracies Choose Bad Policies.
Princeton Univertsity Press, Princeton, NJ.
Carmines, E. and Stimson, J. (1990). Issue Evolution. Princeton University Press, Prince-
ton, NJ.
Carraway, R. L., Schmidt, R. L., and Weatherford, L. R. (1993). An algorithm for maximiz-
ing target achievement in the stochastic knapsack problem with normal returns. Naval
Research Logistics 40, 2, 161–173.
Chipman, H. A., George, E. I., and McCulloch, R. E. (2008). BART: Bayesian addi-
tive regression trees. Tech. rep., Department of Statistics, University of Pennsylvania.
Available at http://arxiv.org/abs/0806.3286v1.
Chong, D., Citrin, J., and Conley, P. (2001). When self–interest matters. Political Psy-
chology 22, 3, 541–570.
Cohen, J. E. and King, J. D. (2004). Relative unemployment and gubernatorial popularity.
The Journal of Politics 66, 4, 1267–1282.
Converse, P. E. (1964). The nature of belief systems in mass publics. In D. E. Apter, ed.,
Ideology and Discontent, 75–169. Free Press, New York.
Cox, D. R. (1958). Planning of Experiments. John Wiley & Sons, New York.
211
Dale, A. and Strauss, A. (2008). Mobilizing the mobiles: How text messaging can boost
turnout. In C. Panagopoulos, ed., Politicking Online: The Transformation of Election
Campaign Communications. Rutgers University Press, New Brunswick, NJ.
Dale, A. and Strauss, A. (2009). Don’t forget to vote: Text message reminders as a
mobilization tool. American Journal of Political Science 53, 4.
Dalton, R. J. (2008). The quantity and the quality of party systems: Party system po-
larization, its measurement, and its consequences. Comparative Political Studies 41, 7,
899–920.
Dantzig, G. B. (1957). Discrete–variable extremum problems. Operations Research 5, 2,
266–277.
Dehejia, R. (2005). Program evaluation as a decision problem. Journal of Econometrics
125, 141–173.
Delli Carpini, M. and Keeter, S. (1996). What Americans Know about Politics and Why
It Matters. Yale University Press, New Haven, CT.
Downs, A. (1957). An Economic Theory of Democracy. Harper & Row, New York.
Eldersveld, S. J. (1956). Experimental propaganda techniques and voting behavior. Amer-
ican Political Science Review 50, 1, 154–165.
Enelow, J. M. and Hinich, M. J. (1984). The Spatial Theory of Voting : An Introduction.
Cambridge University Press, New York.
Fazio, R. H. and Zanna, M. P. (1978). On the predictive validity of attitudes: The roles of
direct experience and confidence. Journal of Personality 46, 2, 228–243.
212
Fazio, R. H. and Zanna, M. P. (1981). Direct experience and attitude-behavior consistency.
Advances in Experimental Social Psychology 14, 161–202.
Feddersen, T. J. and Pesendorfer, W. (1996). The swing voter’s curse. The American
Economic Review 86, 3, 408–424.
Ferejohn, J. (1986). Incumbent performance and electoral control. Public Choice 50, 1,
5–25.
Festinger, L. (1957). A theory of cognitive dissonance. Stanford University Press, Stanford,
CA.
Fiorina, M. P. (1981). Retrospective Voting in American National Elections. Yale University
Press, New Haven, CT.
Fiorina, M. P. (1996). Divided Government. 2nd ed. Allyn and Bacon, Needham Heights,
MA.
Fisher, R. (1926). The arrangement of field experiments. Journal of the Ministry of
Agriculture 33, 503–513.
Gelman, A. and King, G. (1993). Why are american presidential election campaign polls
so variable when votes are so predictable? British Journal of Political Science 23, 4,
409–451.
Geoffrion, A. M. (1967). Solving bicriterion mathematical programs. Operations Research
15, 1, 39–54.
Gerber, A., Gimpel, J. G., Green, D. P., and Shaw, D. R. (2007). The influence of television
and radio advertising on candidate evaluations: Results from a large scale randomized
experiment. Unpublished manuscript .
213
Gerber, A. and Green, D. P. (1998). Rational learning and partisan attitudes. American
Journal of Political Science 42, 3, 794–818.
Gerber, A. S. (2004). Does campaign spending work?: Field experiments provide evidence
and suggest new theory. American Behavioral Scientist 47, 5, 541–574.
Gerber, A. S. and Green, D. P. (2000). The effects of canvassing, telephone calls, and
direct mail on voter turnout: A field experiment. American Political Science Review 94,
653–663.
Gerber, A. S., Green, D. P., and Shachar, R. (2003). Voting may be habit–forming:
Evidence from a randomized field experiment. American Journal of Political Science
47, 3, 540–550.
Gershkoff, A. R. (2006). How issue interest can resue the american public. Doctoral
Disseration, Princeton University, Department of Politics .
Gilens, M. (2001). Political ignorance and collective policy preferences. American Political
Science Review 95, 2, 379–396.
Gilens, M. and Murakawa, N. (2002). Elite cues and political decision–making. In M. X.
Delli Carpini, L. Huddy, and R. Y. Shapiro, eds., Research in Micropolitics: Political
Decision–Making, Deliberation, and Participation (Vol. 6), 15–49. JAI Press, New York.
Goren, P. (2007). Character weakness, partisan bias, and presidential evaluation: Modiff-
cations and extensions. Political Behavior 29, 305–325.
Gosnell, H. F. (1927). Getting–Out–the–Vote: An experiment in the stimulation of voting.
University of Chicago Press, Chicago.
214
Grant, D. A., Hake, H. W., and Hornseth, J. P. (1951). Acquisition and extinction of
a verbal conditioned response with differing percentages of reinforcement. Journal of
Experimental Psychology 41, 1.
Green, D. P. and Gerber, A. S. (2008). Get Out the Vote: How to Increase Voter Turnout.
Brookings Institution, Washington D.C., 2nd edn.
Groseclose, T. (2001). A model of candidate location when one candidate has a valence
advantage. American Journal of Political Science 45, 4, 862–886.
Gunter, L., Zhu, J., and Murphy, S. (2007). Variable selection for optimal decision making.
Poceedings of the 11th Conference on Artificial Intelligence in Medicine AIME 2007,
LNCS/LNAI 4594, 149–154.
Hartmann, G. W. (1936). A field experiment on the comparative effectiveness of “emo-
tional” and “rational” political leaflets in determining election results. Journal of Ab-
normal Psychology 31, 1, 99–114.
Henig, M. I. (1990). Risk criteria in a stochastic knapsack problem. Operations Research
38, 5, 820–825.
Hibbs Jr., D. A. (2000). Bread and peace voting in u.s. presidential elections. Public Choice
104, 1, 149–180.
Hill, J. L. and McCulloch, R. E. (2008). Bayesian nonparametric modeling for causal
inference. Tech. rep., Graduate School of Business, University of Chicago.
Hillygus, D. S. and Shields, T. G. (1991). The Persuadable Voter: Wedge Issues in Presi-
dential Campaigns. Princeton University Press, Princeton, NJ.
215
Holland, P. W. (1986). Statistics and causal inference. Journal of the American Statistical
Association 81, 945–960.
Hotelling, H. (1929). Stability in competition. The Economic Journal 39, 153, 41–57.
Hutchings, V. L. (2003). Public Opinion and Democratic Accountability: How Citizens
Learn about Politics. Princeton University Press, Princeton, NJ.
Imai, K. and Strauss, A. (2009). Planning the optimal get–out–the–vote campaign using
randomized field experiments. Tech. rep., Department of Politics, Princeton University.
available at http://imai.princeton.edu/research/choice.html.
Iyengar, S. and Kinder, D. (1989). News that matters : Television and American opinion.
University Of Chicago Press, Chicago, IL.
Kelleher, C. A. and Wolak, J. (2006). Priming presidential approval: The conditionality
of issue effects. Political Behavior 28, 3, 193–210.
Key, V. (1968). The Responsible Electorate: Rationality in Presidential Voting, 1936–1960.
Harvard University Press.
Knutson, K. M., Wood, J. N., Spampinato, M. V., and Grafman, J. (2006). Politics on the
brain: An fmri investigation. Social neuroscience 1, 1, 25–40.
Kramer, G. (1966). A decision–theoretic analysis of a problem in political campaigning.
In J. L. Bernd, ed., Mathematical Applications in Political Science, Vol. II, 137–160.
Southern Methodist University, Dallas, TX.
Krosnick, J. A. (1990). Government policy and citizen passion: A study of issue publics in
contemporary america. Political Behavior 12, 1, 59–92.
216
Lau, R. R. and Redlawsk, D. P. (2001). Advantages and disadvantages of cognitive heuris-
tics in political decision making. American Journal of Political Science 45, 4, 951–971.
Lauderdale, B. (2008). Bayesian social learning: Theoretical implications for the measure-
ment of projection phenomena. Presented in the Princeton American Politics Graduate
Research Seminar .
Lenz, G. (2006). What Politics is About. Ph.D. thesis, Princeton University.
Levitt, S. D. and Wolfram, C. D. (1997). Decomposing the sources of incumbency advantage
in the u. s. house. Legislative Studies Quarterly 22, 1, 45–60.
Lodge, M., McGraw, K. M., and Stroh, P. (1989). An impression–driven model of candidate
evaluation. The American Political Science Review 83, 2, 399–419.
Lodge, M. and Taber, C. (2000). Three steps toward a theory of motivated political
reasoning. In A. Lupia, M. McCubbins, and S. Popkin, eds., Elements of Reason: Cog-
nition, Choice, and the Bounds of Rationality, 183–213. Cambridge University Press,
Cambridge, MA.
Lord, C. G., Ross, L., and Lepper, M. R. (1979). Biased assimilation and attitude polar-
ization: The effects of prior theories on subsequently considered evidence. Journal of
Personality and Social Psychology 37, 11, 2098–2109.
Lupia, A. (1994). Shortcuts versus encyclopedias: Information and voting behavior in
california insurance reform elections. The American Political Science Review 88, 1,
63–76.
Lupia, A. and McCubbins, M. D. (1998). The Democratic Dilemma: Can Citizens Learn
What They Need To Know? Cambridge University Press, New York.
217
Malchow, H. (2003). The New Political Targeting. Campaigns & Elections, Washington,
D.C.
Malhotra, N. and Kuo, A. G. (2009). Emotions as moderators of information cue use.
American Politics Research 37, 2, 301–326.
Manski, C. F. (2005). Social Choice with Partial Knowledge of Treatment Response. Prince-
ton University Press.
Markus, G. B. and Converse, P. E. (1979). A dynamic simultaneous equation model of
electoral choice. The American Political Science Review 73, 4, 1055–1070.
McCarty, N., Poole, K. T., and Rosenthal, H. (2006). Polarized America: The Dance of
Ideology and Unequal Riches. MIT Press, Cambridge, MA.
McFadyen, R. G. (1998). Attitudes toward the unemployed. Human Relations 51, 2,
179–199.
Mitchella, P., Robinsona, E. J., Nyea, R. M., and Isaacsa, J. E. (1997). When speech
conflicts with seeing: Young children’s understanding of informational priority. Journal
of Experimental Child Psychology 64, 276–294.
Mueller, J. E. (1970). Presidential popularity from truman to johnson. The American
Political Science Review 64, 1, 18–34.
Mutz, D. C. (1992). Mass media and the depolitization of personal experience. American
Journal of Political Science 36, 2, 483–508.
Nickerson, D. and Arceneaux, K. (2006). Who is mobilized to vote? a meta–analysis of 7
field experiments. In Annual Meeting of the Midwest Political Science Association.
218
Nickerson, D. W. (2007). Quality is job one: Professional and volunteer voter mobilization
calls. American Journal of Political Science 51, 2, 269–282.
Nickerson, D. W. (2008). Is voting contagious?: Evidence from two field experiments.
American Political Science Review 102, 1, 49–57.
Page, B. and Shapiro, R. (1992). The Rational Public: Fifty Years of Trends in American’s
Policy Preferences. University Of Chicago Press.
Parker-Stephen, E. (2007). When perceptions polarize: How motives and information
shape partisan inference. In Annual Meeting of the Midwest Political Science Association,
Chicago.
Petrocik, J. R. (1996). Issue ownership in presidential elections, with a 1980 case study.
American Journal of Political Science 40, 3, 825–850.
Petty, R. E. and Cacioppo, J. T. (1981). Attitude and Persuasion: Classic and Contem-
porary Approaches. Wm. C. Brown, Dubuque, IA.
Pocock, S. J., Assmann, S. E., Enos, L. E., and Kasten, L. E. (2002). Subgroup analysis,
covariate adjustment and baseline comparisons in clinical trial reporting: current practice
and problems. Statistics in Medicine 21, 2917–2930.
Popkin, S. (1994). The Reasoning Voter: Communication and Persuasion in Presidential
Campaigns. University of Chicago Press, Chicago, IL.
Regan, D. T. and Fazio, R. H. (1977). On the consistency between attitudes and behavior:
Look to the method of attitude formation. Journal of Experimental Social Psychology
13, 1, 28–45.
219
Richard Johnston, Michael G. Hagen, K. H. J. (2004). The 2000 Presidential Election and
the Foundations of Party Politics. Cambridge University Press.
Ripley, B. D. (1996). Pattern Recognition and Neural Networks. Cambridge University
Press, Cambridge.
Rubin, D. B. (1978). Bayesian inference for causal effects: The role of randomization.
Annals of Statistics 6, 1, 34–58.
Rubin, D. B. (1990). Comments on “On the application of probability theory to agricultural
experiments. Essay on principles. Section 9” by J. Splawa–Neyman translated from the
Polish and edited by D. M. Dabrowska and T. P. Speed. Statistical Science 5, 472–480.
Sanders, M. S. (2000). Uncertainty and turnout. Political Analysis 9, 1, 45–57.
Sartori, G. (1976). Parties and party systems. Cambridge University Press, New York.
Savage, L. J. (1951). The theory of statistical decision. Journal of the American Statistical
Association 46, 253, 55–67.
Sawyer, A. G. (1973). The effects of repetition of refutational and supportive advertising
appeals. Journal of Marketing Research 10, 1, 23–33.
Sears, D. O. and Funk, C. L. (1990). Self–interest in americans’ political opinions. In J. J.
Mansbridge, ed., Beyond self–interest, 147–170. University Of Chicago Press, Chicago.
Sellers, P. J. (1998). Strategy and background in congressional campaigns. The American
Political Science Review 92, 1, 159–171.
Sherif, M. (1936). The psychology of Social Norms. Harper and Brothers, New York.
220
Sherman, S. J., Presson, C. C., Chassin, L., Bensenberg, M., Corty, E., and Olshavsky,
R. W. (1982). Smoking Intentions in Adolescents: Direct Experience and Predictability.
Pers Soc Psychol Bull 8, 2, 376–383.
Sigelman, L. and Buell, Emmett H., J. (2004). Avoidance or engagement? issue convergence
in u.s. presidential campaigns, 1960–2000. American Journal of Political Science 48, 4,
650–661.
Sosnick, D. B., Dowd, M. J., and Fournier, R. (2006). Applebees America: How Successful
Political, Business, and Religious Leaders Connect with the New American Community.
Simon & Schuster, New York.
Sotirovic, M. (2001). Effects of media use on complexity and extremity of attitudes toward
the death penalty and prisoners’ rehabilitation. Media Psychology 3, 1, 1–24.
Stoye, J. (2009). Minimax regret treatment choice with finite samples. Journal of Econo-
metrics Forthcoming.
Taber, C. S. and Lodge, M. (2006). Motivated skepticism in the evaluation of political
beliefs. American Journal of Political Sciencey 50, 3, 755–799.
Thistlewaite, D. and Campbell, D. (1960). Regression discontinuity analysis: An alterna-
tive to the ex–post–factor experiment. Journal of Educational Psychology 51, 309–317.
van Knippenberg, D. and Daamen, D. (1996). Providing information in public opinion
surveys: Motivation and ability effects in the information–and–choice questionnaire.
International Journal of Public Opinion Research 8, 1, 70–82.
Verplanken, B. (1991). Persuasive communication of risk information: A test of cue versus
message processing effects in a field experiment. Personality and Social Psychology
Bulletin 17, 2, 188–93.
221
Wald, A. (1950). Statistical Decision Functions. Wiley, New York.
Washington, E. (2008). Female socialization: How daughters affect their legislator fathers’
voting on women’s issues. American Economic Review 98, 311–332.
Westen, D., Blagov, P. S., Harenski, K., Kilts, C., and Hamann, S. (2006). Neural bases
of motivated reasoning: An fmri study of emotional constraints on partisan political
judgment in the 2004 u.s. presidential election. Journal of Cognitive Neuroscience 18,
11, 1947–1958.
Weston, D. (2007). The Political Brain: The Role of Emotion in Deciding the Fate of the
Nation. Public Affairs, Cambridge, MA.
Wood, W. (1982). Retrieval of attitude–relevant information from memory: Effects on
susceptibility to persuasion and on intrinsic motivation. Personality & social psychology
bulletin 42, 5, 798–810.
Zaller, J. R. (1992). The Nature and Origins of Mass Opinion. Cambridge University
Press, New York.
222

c Copyright by Aaron B. Strauss, 2009. All Rights Reserved

Abstract
Partisan attachments create pervasive bias in the way citizens process information. Political scientists, psychologists, and recently neuroscientists find that people will believe nearly anything if a favored politician espouses the view. Yet, even though partisan affiliation is one of the most, if not the most, stable political attitudes, large segments of the public switch their vote choice from one party to the other between elections or split their tickets within a single election. This dissertation examines one explanation for shifting political views: personal experience with specific issues. Campaigns work to attract more support, but partisan biases hinder their efforts when predispositions lead voters to doubt statements made by disfavored politicians. This dissertation explores the theory that campaigns can successfully target voters who have experience on a particular political issue. The voter will use her independent knowledge on the topic to judge, or “ground truth,” the politician’s views; if the voter and the politician agree, the voter will hold the candidate in higher esteem. With the advent of massive campaign databases of information on voters, campaigns are now able to identify these critical voter-issue linkages. The Personal Experience Model explores why personal experience plays such a crucial role in political judgments. This formal model is an extension of Zaller’s Receive-AcceptSample model. The theory behind the Personal Experience Model is presented, related to existing theories, and supported by empirical evidence. Observational data from the 2000 presidential campaign, two survey experiments, and two field experiments all support the model’s hypotheses. Finally, the strategic implications for campaigns, and the normative implications for democracy, are considered.

iii

Acknowledgements
It is difficult to adequately express my appreciation for the many people who have supported and guided me during my time at Princeton. It would feel appropriate to add about a dozen names next to mine on the title page, but I will defer to rules and customs and put their names in this section instead. My graduate school colleagues made Princeton a wonderful place to study political science. From “West Wing” viewings to games of Settlers of Catan, from softball defeat to ultimate triumph, my friends demonstrated that the phrase “graduate school life” isn’t always an oxymoron. Thank you Dustin, Dan, Ben, Glick, Tom D., Keya, Richard, Miranda, Kevin, Andi, Will, Lauren, Tom C., Melody, Shana, Andrew, Nick, Grace, Alicia, Gayle, Lisa, and Alistair who have all been awesome friends. The graduate work of Gabriel Lenz and Amy Gershkoff provided excellent models for my research. Their dissertations were the first ones I turned to when I experienced writer’s block. Amy’s ability to turn esoteric political topics into accessible prose and Gabe’s straightforward mathematical proofs were the trailmarkers that kept me on the right path. I have attempted to live up to the high standard they set. In addition, fellow graduate student Ben Lauderdale was an excellent sounding board. Several ideas in this dissertation are the result of our long conversations on the third floor of Robertson Hall. His amazing ability to generate a new idea every week was both humbling and motivating. I could only take solace in my ability to run down fly balls faster than he could on the softball field. This dissertation is, in large part, a reflection of the professors who have taught me these past four years. Thank you to Chris Achen for early inspiration on this project. Larry Bartels provided incisive comments that provided context for my work. Professors

iv

Kosuke Imai was there to right the ship. I thank Tali Mendelberg for her positive comments and attention to detail. Markus Prior’s class on campaigns provided the first forum for the hypotheses that became the centerpiece of the dissertation. When the methodological portion of this dissertation hit a rocky patch of shore. Markus always found time to share his insights. Marty Gilens. his smile contagious. I am deeply indebted to my dissertation committee. I’m glad they asked the question “How’s the dissertation going?” v . So. She made this dissertation better by challenging me to dig into the psyches of voters.at other institutions. Completing a dissertation under my adviser. I could always count on a lunch with Tali to bring clarity to this project and to lift my spirits. including Steven Ansolabehere (my undergraduate advisor at MIT) and Andrew Gelman of Columbia also generously gave their time to comment on chapters. Marty. and members of other universities that I may have forgotten to express my feelings to him directly. and his methods of motivation creative—such as the time he invoked a promise from me to eat Turkey on Thanksgiving (and end 12 years of vegetarianism) if I failed to complete a chapter draft. was a pleasure. I must give special thanks to Chu Hwang and Hillary Hampton for supporting me throughout this process. here it is on acid-free paper: thank you for being the best adviser a graduate student could ask for. His door was always open. I have sung the praises of Marty to many prospective graduate students. undergrads. I’m pleased that we were able to co-author multiple papers. Despite his own deadlines occasionally competing with mine. I realize that without a randomized experiment (no thanks!) I can’t be sure that this dissertation wouldn’t have been completed without his help. but I have a strong prior opinion. but even happier that I can call him my friend. Thanks to Kosuke’s teaching on causal inference.

Thank you. my brother Marc and I continue this family tradition by facing off in online Scrabble matches. That they both traveled hours to be at my defense speaks volumes about what wonderful friends they are. Though we live on opposite coasts. The grade I received on my first high school literature essay—an F—suggests that my ability to write a dissertation is not innate. was a tremendous help on this project. My father introduced my brother and me to educational games at an early age. My mother. Mom. vi . for patiently teaching me that the world is not just a set of equations. And finally.as often as possible. who fortunately is a professional editor. my family deserves thanks for being supportive and loving.

Dad. vii .To my family—Mom. and Marc—who are responsible for my love of learning.

. . . . . iii iv 1 8 8 11 15 19 24 27 33 35 37 37 39 41 45 Concurrence with Other Theories in the Literature .1. . . . . . . . . . . .7 2.2. Issue Experience as Signals .6 2. . . . .3 2. .1.4 Definitions of Concepts in the Literature . . . . . . . Evidence for Cue-Taking and the Role of Experience . . . . .Contents Abstract . . . .2 2.1. . . . . . . . . . . . . . . . . .2. . Learning Model: Issue Positions . . . . . viii . . . Voters’ Beliefs About Candidate Positions . . . . . . . . . . . Model Extensions and Details . . . . . . . . .1. Voters’ Candidate Evaluations . . . . 2. . . . . . . . . . . . . . . . . . . . . Acknowledgements . . .4 2. .2 Psychological Model of Personal Experience and Issue Stability . . . . . . .2. . Personal Experience Yields Nuanced and Stable Opinions . . . . .5 2. . 1 Introduction 2 The Personal Experience Model 2. .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Extending Zaller by Introducing Personal Experience . . . . . . . . . . . Easy Issues and Retrospection Affect Political Evaluations .1. 2. . . . . . . . .3 2. . .1. . . . . . . .1 Theory and Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . .1 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 2. . . . .1. . . . . . . . . . .2 2. . . .

. . . . .1 4. . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . Results and Analysis . . . . . . . . . . Issue Opinions and Experience . . . . . . . . . . . . . . .1 4. . .1. . . . . . . . . . . . . Discussion . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Definitions of Issue Experience . . .1. . . . . . . . . . . . . . . Experienced Voters Judge Politicians on Issues . . . . . . . . Design Checks 4. Specifics of Princeton Survey Design . . . . . . . . . .2 Candidate Evaluation .3 Characteristics of the Survey Respondent Population . . . . . . . . . 4. . . . . . . Issue-Driven Vote Choice . . . . . . . . .1 3. . . . . . .2 4. . . . . . . . . . .2. . . . Cue-Taking and Self-Interest .1 4. . . . . . . . .3 Alternative Voter-Issue Linkages in the Literature . . . . . . . . . . . 3. . . . . . . . . . . . . .2 3. .2. . ix . . . . . . . . . . .2 Theory and Hypotheses . . . . . . . .1 The Personal Experience Model and the 2000 Presidential Campaign . . . . . . . . . . . . .1 3. . . . . Features of the Survey . Specifics of Nationwide Survey . . . . . . . . . . . . . . . . . . . . . . . .4 Support for Hypotheses . . . . . . . . . . . Data from the 2000 Annenberg Survey . . .1. . . . . . . 47 51 51 51 55 58 59 59 62 67 70 74 75 78 78 80 82 85 85 86 91 92 92 98 3 Experienced Evaluations and Self-Interested Opinions 3. . . . . . .2. . . . . . . .3.2 The Campaign about Nothing . . . . . . . . . .3. . . . . .1 4. . . . Survey Design . . . . . . . . . . . . .3 3. . . . . . . . . . . . .2. . . . . . . .2. .3. . . . . . . . 4 Evidence from Survey Experiments 4. .4. . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 3. . . . . . . . . . . . . . . . .3 Overall Goals . . . . . . . . . . . . . . . . . . .2. . . . .2 4. . . .3 4. . .2.3 3. . . . . . . . . . . . .2. . . 4. . . . . . . . . . . .4 Learning Over the Course of the Campaign . . . . . . . . . . .

. . . . .2 5. . . 116 5. 124 5. . . .1 5. . . . .4. . . . . . .5. . . . . . . . . . . . . . . 120 5. . . .1 5. . . . . 107 5. .2. . . . . . . . . . .2. . . . . . 144 145 6 Campaign and Normative Implications of Microtargeting 6. . . . . . . . . . . . . . . . .6 Conclusion . . . . . .1 Campaign Microtargeting . . . . . . . . . . . . . . . . . . . 114 Bayesian Optimal Campaign Planning at A Glance . . . .4 Evaluation Method . . . . . . . . . . . .1 5.5. . . .3 The Optimal Nonpartisan Campaign Strategy . . . . . . . . . . . . . . . . . . . . . 145 x .2 The Formal Framework of GOTV Campaign Planning . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . .5. . . . . . . . . . . . . . . .5. . 127 5. . . . . 113 The Bayesian Planner .4. .2 5.5 Empirical Evaluation of the Proposed Method .3 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 The Optimization Method . . . . . . . . . . . . . . . . . . . . . . . .2 5. 129 A Nonpartisan GOTV Campaign with a Single Mobilization Method 130 Partisan Example: Parents and Education Spending . . . . . . . . . . . .3 5. 108 5. . . . . . . . . .1 Introduction . . . . . . . . . . . . . . . . . .3 The Decision Problem . . . . . . . . . . . . . .1 5. . . . . . . . . . . . . . . 124 Data Requirements . 125 Derivation of the Optimal Strategy . . . . . . . .1. . . . . . . . . . 104 5. . . . . . . . . . . . . . . . 101 104 5 Field Experiments: New Methodology and Evidence 5. . . . . . . .4 The Planner’s Decision Problem . . . . . . . . . . . . 126 5. . . . . . .4. .5 Discussion .3.4 The Optimal Partisan Campaign Strategy . .3. . 109 Data from a Randomized Field Experiment . . . . . . 135 Partisan Example: Disadvantaged Voters and the Economy . .2. .1 Background of the Methodological Problem . . . . . . . 117 5. . . . 139 5. . . . . 118 The Statistical Method . . . . . .4. . . . . . . .

170 6. . . . . . . .2. . . .2 6. . . . . . . . . . . 164 6. 162 Model Extensions: Multiple Issues and Platform Decisions .1 Details for Candidate Evaluation Simulation . . 165 Positives for Democracy . . . . . . . 179 B Appendix for Chapter 3 182 B. . . . . . .7 6. . . . . . . . . .1 6. . . . . .2. . . . . . .1 Question Wording in Nationwide Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 Edge Equilibria . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . .2. . 183 B. . . . . . . . . . . . . . 184 C Appendix for Chapter 4 186 C. . . . . . .3. . .2 6. . . . . .1 6.4 Conclusion . . . . .4 Heuristics and Judging Democracy . . . .6. . . . . . . . . . . . . . . . . . . 148 6. . . . . . . 153 Best Response . . . . . . . . . .2 Implications for Campaign Strategy .8 Microtargeting and Message Control . . . . . . . . . .2 Regressions for the Candidate Evaluation Hypothesis . .2. . . . . . . . . . . . .3. . . .3. . . . .4 6. . . . . . . .1 Regressions for the Cue-Taking Hypothesis . . . . . . . . . . . . . . 148 Translating the Personal Experience Model into a Formal Game . . . .6 6. . . . . . . . . . . . . . . . . . 171 173 179 7 Conclusion A Appendix for Chapter 2 A. . . 165 6. . . . . . . . . . . . . . . 169 Further Discussion: Party Structure and Alignment . . . . . . . . . . . . .3. . .2. . .3 6. . . .2. . . . . . . . . . . . . . . 150 The Model . . . . . . . . . . . . . . . 186 xi . . . .3 Normative Implications . . . . .3 6. . . . . . . 167 Negatives for Democracy . . . . . . 164 Discussion . .5 6. . . . . . . . . . . . . . . . . . . . . . . . . . .2. 159 No Internal Equilibria . . .

. . . . . . . . . . . . .1. 187 C. . . . .1 Candidate Descriptions . . . .2 Partisan Case: the Stochastic Knapsack Problem . . . . . . . . .1. . . . . 204 E Appendix for Chapter 6 206 E. . . . . . . . . .1 Computational Appendices . . . . . . . . . . . . . . .1. .3 Issue Experience Measures and the Hypotheses . . .C. . . . . . .1. . . . . . . . . 203 D. . . . . . . . . . . . . . . . . . . . . . 206 xii . . . . . .1 Analytical Solution for Variance of Vote Share . . . . . . 190 D Appendix for Chapter 5 203 D. . . . . . . .2 Candidates’ Issue Signals . . . 203 D. . . . . 188 C. . . . . . . . 186 C. . . . . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . 190 C. . . . .3 Experience Questions . .2 Auxiliary Regressions . . . . .1 Nonpartisan Case: the Knapsack Problem . . . . . . . . . . . . . . . . . . .

1984. Bartels. 1960. while non-Hispanics showed no increase in their inclination to be identified 1 . 2008). 1957.. These defections are often attributed to the preferences of voters relative to the candidates (Downs. the model and the evidence presented highlight the role of personal experience in political evaluations. 1990). Enelow and Hinich.Chapter 1 Introduction Prominent research in political science (Campbell et al. the debate over the relative effects of partisan identification versus issue opinion continues in recent research (Ansolabehere et al.. From the 2004 to the 2006 elections. 1968) or split their ticket within one election (Fiorina. To provide observable predictions of the theory. This dissertation synthesizes the arguments from both schools of thought. Carmines and Stimson. 2002) emphasizes the long-lasting attachments voters have to political parties. demonstrating how party identification and issue opinions interact to form individuals’ political judgments. the percentage of Hispanics identifying as Democrats jumped by 15 percentage points (from 43% to 58%). 1996). Yet large portions of the electorate switch parties from election to election (Key. Two recent anecdotes illustrate this connection.

5 percentage points over these two years. The same survey in 2006 found that this percentage had dropped to 46%. When a large group of people have experience with a prominent issue (e. This dissertation argues that an individual’s political allegiance is likely to shift when the political parties take identifiable positions on issues with which the individual has significant personal experience. Campaigns can use recent technological advances to 1 Source: National Exit Polls 2 . And despite illegal immigration falling off as a key concern in 2008. In 2008. compared to 53% for Kerry. the wars in Iraq and Afghanistan had a profound effect on the political views of the military. Obama outperformed Kerry among the active military and veterans despite the facts that both Kerry and McCain—Obama’s opponent—were veterans. A 2004 mail survey of subscribers to Military Times found that 60% of active duty personnel identified as Republicans. As Hispanics were flocking to the Democratic party. Republican affiliation among the electorate at large (as measured by exit polls) dropped only 1. While immigration policies might not personally affect these voters (as they were already citizens). this trend of Hispanics favoring the Democrats continued in the 2008 election.g.as Democrats. As measured by the National Exit Poll.1 Hispanics’ responses to another question in the exit poll shed light on the underlying reason for the massive movement: 37% of Latinos indicated that the issue of “illegal immigration” was “extremely important” in their voting decision (compared to only 29% of the non-Latino population). Obama garnered 67% of the Hispanic vote.. their involvement in the Hispanic community most likely gave them firsthand experience with the immigration issue. Hispanics and immigration. substantial changes can occur in the political landscape. active duty military personnel were leaving the Republican party in droves. the military and the Iraq War). Almost assuredly.

twice as many thought the deficit had increased during Clinton’s first term as thought it had decreased (Achen and Bartels. among Republicans in 1996. large partisan biases are seen. Not only did partisans seek to diminish displeasure by explaining away uncomfortable facts. the subject saw Kerry making two statements: one in 1996 that Congress had to consider raising the retirement age to save Social Security. 2006). 2 Throughout this dissertation the female pronoun is used to identify an individual voter. 2006).2 A telling example of this lack of critical thinking is provided by an experiment performed by neuroscientists on partisans during the 2004 election (Westen et al. 1979). an individual’s partisan bias weighs heavily on her political decision-making process as she adopts the positions of her party’s elites and does not adjust her political worldview. but their brains actively sought positive reinforcement. partisans were quick to pass judgment on the other party’s nominee but not their own. Generally. microtarget) the segments of the population that have experience with certain issues and that thus may be ripe for a political change of heart.e. by over 90%. When citizens fall prey to this psychological process in the political arena (Lodge and Taber. and a second in 2004 that he would never consider raising the retirement age. When asked whether the candidates’ statements were internally inconsistent. In fact. The male pronoun is used to identify an individual candidate. Bush and John Kerry. For instance. a sector of partisans’ brains that dealt with reward (ventral striatum) showed significant activity. 2000).. the deficit had fallen dramatically.. For example.. 3 . The brain’s attempt to diminish displeasure and seek positive reinforcement is the neurological basis for humans’ preference for obtaining conclusions that are consistent with their prior beliefs—a process psychologists label “motivated reasoning” by psychologists (Lord et al. Subjects were presented with contradictory statements by George W. In fact.identify (i. when evaluating their own party’s nominee.

the legitimacy of judicial filibuster) and the voters follow.These partisan biases help keep party affiliation more stable over time than political issues (Ansolabehere et al. survey experiments. as detailed in Chapter 3. Further. “considerations. 2002).” on which an individual can base her opinions. Because these considerations are generated through an individual’s personal interactions. 2004). Issue interactions often produce points of reference. The key feature of personal experiences that they enable voters to have stable opinions on political issues independent of their political predispositions.e.g. then the voters’ opinions can hardly be considered “stable. Personal experience is defined as interaction with a political issue. This distinction is crucial: if politicians find it more expedient to switch sides on an issue (e. The exception. that seniors’ personal experience with the Social Security system moderated their motivated reasoning and cue-taking. was older Americans. and demonstrates with data from observational data. 4 . Democrats adopted the anti-privatization position espoused by Gore while Republicans adopted the privatization position espoused by Bush (Richard Johnston. For instance. This dissertation hypothesizes. and field experiments. who were more likely to stay consistent in their view (which was. they are likely to be trusted even if they go against the person’s preconceptions (i. Thus. 2008). anti-privatization). personal experience with an issue enables voters to judge politicians on this issue and can counter partisan biases that would usually hold. voters are reliant on political elites to point them in the right direction.” The theory presented here is premised on the idea that experience with an issue induces stable opinions. The public takes cues from like-minded elites and adopts those positions on issues (Gilens and Murakawa. personal experiences enable individuals to generate stable issue opinions independent of their overall political views.. during the 2000 campaign... “seeing is believing”). Without these experiences. or to use the political science term. more often than not.

who have experience with Social Security. Chapter 3 highlights two observational cases from the 2000 presidential campaign that illustrate the model in action: Social Security privatization and the Patient’s Bill of Rights. The second main hypothesis is that voters with experience shift their issue opinion less in the direction of a politician’s appeal than do individuals without experience. 5 . derives hypotheses. provides evidence in support of the hypotheses. and discusses the hypotheses’ implications. Regarding the first issue. Chapter 2 presents the Personal Experience Model. Existing research is shown to be consistent with the Personal Experience Model. seniors.This dissertation elaborates this personal experience-issue opinion connection. citizens with experience judge politicians on the issue. if political elites start debating an issue. asked respondents about hypothetical candidates and real-world issues. Simulations illustrate the model at work in the campaign setting. First. one of Princeton students (n=273) and one of the public nationwide (n=391). The results are generally supportive of the Personal Experience Model’s two main hypotheses but are noisy because of low n-sizes. Chapter 4 looks for the effects predicted by the Personal Experience Model in a survey experiment setting. Younger voters were more apt to adopt the position of their preferred candidate than were seniors. Respondents were also queried about their experiences relating to these issues. comparing their experience to the politician’s argument. An analogous phenomenon occurs among voters insured by HMOs and the Patients’ Bill of Rights. In effect. judged Bush and Gore more on privatization than did younger voters. Two Internet panel surveys. a learning model based on Zaller’s (1992) Receive-Accept-Sample (RAS) model that generates two main hypotheses. an individual “ground truths” an elite’s views by evaluating how consistent the elite’s arguments are with the voter’s first-hand experience. noting how the foundations of the model differ from research on issue publics and self-interest.

Examples with experienced-voter segments demonstrate the power of this method and validate the candidate evaluation hypothesis of the Personal Experience Model.Chapter 5 develops a methodology for analyzing field experiments that identifies voters who are responsive to candidate appeals. get-out-the-vote) and partisan settings. This process. For each segment of the population. called “microtargeting”. but more often may be useless or even backfire. The method is general enough to be applicable in both non-partisan (i. broadcasting is a gamble that may have large returns for a campaign. campaigns and election-oriented organizations have expanded their capacity to capture political data to the point where both major parties have access to information about every voter nationwide.. Resource allocation algorithms are applied to these estimates to determine the optimal campaign strategy. In recent decades. With the evidence for the Personal Experience Model laid out. the difference between the estimated effect for the two treatment assignments is the treatment effect. A game theoretical model explicates the situations in which a campaign should spend its money microtargeting voters rather than broadcasting its candidate’s position on an issue to all voters. Rather than segment the population into distinct clusters (as campaigns used to do). the final chapter explores the implications of the model for campaigns and democracy. campaign practitioners now assign a probability score for a certain trait (e. is related to the Personal Experience Model because the technology behind microtargeting can be used to link voters to issues in a manner recommended by the model. probability of being undecided) to each individual. Off-the-shelf non-parametric data-mining techniques are applied to both the treatment group and the control group.e. In general. Thus. Commercial and census data supplement these database and provide even more knowledge about individuals. campaigns that are 6 .g..

Campaigns can leverage the implications of the model to increase the impact and efficiency of their appeals. 7 . The Personal Experience Model. combined with modern campaign tools. The normative implications for democracy are mixed. Information may be disseminated to the public in a more relevant fashion as voters are matched with issues they care about. while advantaged campaigns often prefer to microtarget. the Personal Experience Model explains a voter-issue linkage that limits the partisan bias that pervades the political arena. In sum. The information gain is viewed as a positive for the democratic process. The micro-level model provides deeper understanding of the research on issue publics and self-interest. the increased party attachment that microtargeting produces in some segments of the public may lead individuals to accept the statements of political elites more easily even if those arguments are completely fabricated. increases the efficiency of microtargeting. As with many tools.behind in the polls find this risky strategy useful. depending on one’s perspective. However. these advancements may be used for positive or negative purposes.

.: “only an event of extraordinary intensity can arouse any significant part of the electorate to the point that its established political loyalties are shaken” (Campbell et al. Consistent with the psychological concept of cognitive dissonance 8 .Chapter 2 The Personal Experience Model 2. The so-called “Michigan model” (Campbell et al. However. 1960) highlights the role of early-life events and relationships in the formation of long-standing partisan identification. most research on political persuasion does not account for these changes. at least among a subset of voter-issue interactions. 151). which are dubbed the “unmoved mover” by Campbell et al. especially when the changes are pronounced in specific groups. p.. large segments of the population can be persuaded to change parties from one election to the next. The Personal Experience Model seeks to address this discrepancy. The mechanism for stable partisanship has been elucidated over the years and scholars have reached a consensus.1 Theory and Hypotheses As demonstrated by Hispanics and the military after the 2004 election. Interactions with the political system at an early age form individuals’ partisan loyalties. 1960.

Taber and Lodge. 1960) that can break the usually dominant role of party identification? The lack of theory on the determinants of party-switching has left the door open to other approaches. then why do some voters switch parties between elections or split their votes within an election? The Michigan model clearly answers the first question with the notion that voters adopt the party identification of their parents and the social atmosphere in which they are raised. and thus more perceived common interest. 1992.e. 1957).e. including some from neuroscience (Knutson et al. 2006) applies this concept to politics.” begs two questions. those without perceived common interest). 2006). Prominent research (Zaller. individuals are more likely to accept and process political information and assertions that are consistent with their current views. First. is less clear: what is an “event of extraordinary intensity” (Campbell et al. formalizing the phenomenon and demonstrating how the predictions are consistent with observations of public political opinions. Fiorina (1981) argues that voters are affected by how parties and candidates 9 .. Bartels. But the answer to the second question. Two large areas of work that explain a shifting electorate are retrospective voting and issue voting. This positive feedback loop. In this model. Dozens of micro studies. or “cycle of partisan bias. cue-taking.. leading to hardened political opinions. Listening to trusted politicians leads to further agreement between the voter and the elite (i. change in party identification... if politicians cannot persuade voters to cast a ballot for previously untrusted candidates or parties (i.(Festinger. voters are persuaded by elites with both “perceived common interest” to the voter and “perceived knowledge” of the subject at hand. One emblematic explanation for the micro-foundations of this “partisan bias” is offered by Lupia and McCubbins’s (1998) model of persuasion. Gilens and Murakawa. what is the origin of “perceived common interest?” Second. 2002. 2002). have verified that partisan identification skews individuals’ perceptions of the world in a self-confirming manner.

one-valued utility is expanded into several issues of varying importance (or “salience”) by Enelow and Hinich (1984). certainly plays a role in candidate and party evaluations (Mueller. then changes in issue opinion will only polarize the electorate. suggesting they are formed before an issue is politicized. Second.g. especially on the economy. it details the origins of voter-issue linkages.performed their governing duties while in power. these types of judgments should be relatively uniform across the populace. will be observed. 1 10 . Krosnick. A second answer to the vote-switching question is presented by scholars who analyze voters’ issue opinions. 1929. this dissertation distinguishes itself from existing research in the following two ways. However. and do not account for heterogeneous movements. That voters are rational actors in formalized models such as Hotelling-Downs and Fiorina does not rule out the possibility that voters are rational when succumbing to partisan bias. or one candidate to the other. would predict heterogeneous shifts in public opinion. It may be rational to minimize the displeasure of believing two seemingly contradictory ideas. 1970).. self-interest (e.1 Downs’ uni-dimensional. Certain segments of the population altering their issue opinions or issue saliences could account for the observed shifts in the electorate’s political views. Sears and Funk. While both the model presented here and these theories emphasize voter-issue link. The Hotelling-Downs (Hotelling. First. and offers an alternative causal mechanism. 1990).. Two other subjects discussed in the literature. if voters cue-take from preferred politicians via the cycle of partisan bias. 1957) model presents voters as rational actors who vote for the party (or candidate) that most closely matches their issue positions. 1990) and issue publics (e.g. However. Retrospective voting. in contrast to the research on self-interest. Downs. this dissertation incorporates cue-taking directly into a formal model. such as with Hispanics and the military after 2004. No shifts from one party to the other.

this chapter presents the psychological flows of political information. individuals can either seek out this information or happen upon it accidentally. and then I develop a learning model that formalizes these flows. and parties. and others.1 Psychological Model of Personal Experience and Issue Stability A mechanism for breaking the cycle of partisan bias—which emphatically does not rule out other causes—is personal experience. Under RAS. Next. An individual’s worldview is an online aggregation of that person’s evaluations of important political figures. More politically attentive individuals are more likely to receive political information. This accept/reject process. (1960).2. Even Zaller acknowledges that voters’ evaluations of other people (rather than issues) act more like an online 11 . Individuals create and strengthen political opinions through personal experience on a subset of the many issues that political elites must address.” is the key for producing the stable partisan affiliation and partisan biases found by Campbell et al.” Figure 2. In Zaller’s words. an individual incorporates information into a political opinion by first receiving the information. candidates. The flows of information are a superset of Zaller’s (1992) Receive-Accept-Sample (RAS) model. individuals (either consciously or subconsciously) accept or reject a piece of political information based on their worldview.1 presents a box-and-arrow depiction of the Personal Experience Model. A review of Zaller’s model is crucial for understanding the how the Personal Experience Model fits within the current political science literature. “People tend to resist arguments that are inconsistent with their political predispositions” (p. As noted by Downs (1957). enabling an individual to develop a political worldview. This subset of issues forms a benchmark against which to judge parties. and I call my model the “Personal Experience Model. First. candidates. Bartels (2002). 44). based on what I label an individual’s “worldview. and political elites.1.

e. based on Zaller’s (1992) ReceiveAccept-Sample (RAS) model. Rectangles represent ephemeral processes that information flows through. the one dashed arrow indicates a cause and effect (i. Solid arrows indicate information flow. . ovals represent stable stores of information (or sources of information). partisan identification and worldview affect which pieces of information are accepted or rejected)..12 Figure 2.1: The flow of political information through the mind of an individual. The FIFO symbol under “Set of Considerations” represents the computer science term “First In-First Out” queue and is a simplistic representation of the human brain’s memory capability.

When an individual is asked. 2008). it provides a useful framework. inaccessible).2.3 and 3. numerous studies find that voters often switch their position to adopt the stance of their favored candidate. Individuals often appear to have unstable opinions about issues (Converse. Recently accepted (or “thought about”) considerations are most accessible to the individual. The first political consideration heard (if never thought of again) works its way down to the end of the queue (i. While some political psychologists (Weston. since a large set of considerations is relevant. In this case. 281). voters accept only considerations that are consistent with their existing issue beliefs. Zaller asserts that the accessible considerations relevant to this issue are sampled. water cooler conversation). This prediction is consistent with recent evidence from surveys (Ansolabehere et al. Considerations already in memory that are primed or re-accepted move to the front of the queue and become more accessible.model (p.. and the average of these considerations is reported. a computer science concept. radio report.2.1 depicts this relative availability of political arguments with a First-In-FirstOut (FIFO) queue. to report her issue opinion. Figure 2. 1964) because at any point in time very few considerations about an issue may be accessible and the accessible considerations may depend on seemingly random life occurrences (e. Opinion stability can be defined as the inverse of the variance of repeated sample averages of considerations. However.. Stability increases when issues are averaged together. either by a survey interviewer or a friend. “consideration”) is accepted. An alternative explanation is based on cognitive dissonance rather than source evaluation (or credibility)..2. The model predicts that this instability will be especially acute when an individual rarely receives and accepts arguments about that issue and further does not perceive a link between her political worldview and considerations about the issue. See discussions in Sections 2. then it is stored among all other considerations. 2 13 . 2007) would take issue with this simplistic view of the brain.e.e..2 If the argument (i. human memory) until it is completely forgotten (or at least.g.

If voters were pure Bayesians. Also. the online model never specifies how prone the current evaluation is to change. but Zaller’s 1992 Resistance Axiom incorporates this bias. 121). he allows the full history of consideration to be sampled (p.. the outcomes are generally not in conflict. voters maintain and update an aggregate opinion of issues. is consistent with the cycle of partisan bias. The theory outlined here is an extension of Zaller’s model rather than the online model for two reasons. people. 49). For instance.g... while these two models disagree on the inner workings of voters’ minds. Thus. in which individuals’ conclusions are biased by their preconceptions. This sampling also differentiates Zaller’s model from a Bayesian updating process (e. 1951) persuasively demonstrate. individuals draw from their belief distribution when reporting opinions. 2000) hypothesize motivated reasoning. But as early psychological experiments (Grant et al.Although Zaller contrasts his theory with an online model (Lodge et al. they would report a summary statistic of their beliefs when queried on a survey. in stark contrast to the online model. despite Zaller’s “top of the head” language (p. 1989). in Grant et al. This time-invariant weighting of considerations compares favorably to the online model’s integration of all considerations over time into one evaluation. they share fairly consistent predictions. Gerber and Green. and information sources.3 While memory-based and online models differ in the causal mechanisms of political opinions. 1992. Lodge et al. Specifically. 3 14 . Zaller’s “sample” step implies that reported political judgements are draws from random variables. Importantly. (1989) argue that memories cannot be sampled and evaluated in an unbiased manner. Second. voters do not remember the reasons underlying such opinions. First. both models (Zaller. the online model does not account for the wide variance in people’s opinions over time. Zaller’s RAS model. Zaller argues that for political issue opinions (as opposed to evaluations of politicians). the differences between the two perspectives are not significant for the purposes of this dissertation. 1998). laboratory participants In the online framework. Lodge and Taber. (1951). However. especially the “accept” step.

Certainly. but instead of maximizing their probability of being correct and always predicting the light would turn on. these theoretically determined distributions are subject to human fallacies.1.4 Given this empirical evidence. meaning that there was not a mix of Bayesians and Zalleresque individuals. Zalleresque individuals.were asked whether a light bulb would turn on or not. across the entire participant group. 25% of the time they predicted the light would not turn on. In the experiments. a perfect Bayesian would be correct 75% of the time. They quickly learned (in one set of trials) that the light bulb would turn on 75% of the time. but only Zalleresque individuals in the population. These Zaller-following individuals would have to sample only one consideration to make the “on” prediction 75% of the time. Zaller admits that sampling one consideration is possible (p. 4 15 . That claim is formally stated as a postulate below.2 Extending Zaller by Introducing Personal Experience To answer the questions of where partisan biases originate and why certain segments of the population shift partisan loyalties. 2. In the Personal Experience Model. would predict the bulb to turn on 75% of the time (and be correct 62.5% of the time). a steady flow of experiences related to a political issue leads to a relatively constant set of accessible considerations.. considerations) at variable rates. such as forgetting information (i. along with two main hypotheses and one corollary.e. sampling from a set of considerations of previous light bulb results (the FIFO queue). Bayes rule is therefore referenced when calculating the mean of posterior distributions of considerations. 49). Bayes rule does provide a convenient mathematical way to combine new data into a distribution. it is one of the four predictions of the Personal Experience Model. However. These errors are unmodeled and do not alter the basic hypotheses. Zaller’s model is extended by adding personal experience as a source of political considerations. By always predicting that the light bulb will turn on. 75% of predictions were for the bulb to be on. and hence a stable political opinion. the Personal Experience Model assumes that individuals draw from top-of-the-head considerations and are not perfect Bayesians.

However. News information flows through partisan filters and cue-taking continues unabated. is necessary to lessen the degree to which partisan filters affect which information is received. gays and equal rights). such as the media. being retired and receiving Social Security checks).g. but in attributed opinions” (page 315). fails to meet this criterion because (1) a person’s political worldview may affect which news sources she trusts and (2) news reports often present political parties commenting on current events. The first criterion.g. being in an occupation that deals with a political issue (e. experiments demonstrate that even young children understand the concept of “seeing is believing” (Mitchella et al. immigrating. being an environmental activist) or merely incidental (e... “we don’t deal in facts. The second criterion for political experience. there are exceptions to the rule. Examples include owning a gun. that the individual plays an active role. an increased flow of considerations (represented by the thick arrow in Figure 2.. Receiving information from elite sources. and (2) the interaction produces an increased flow of received considerations on the issue.1). Zaller cites a journalist stating. 16 . 1936). having an abortion. Frequent conversations about political issues or major life events also count as personal experiences. The key factors for determining whether an interaction is considered personal experience in this context are (1) the individual plays an active role. leads to stable considerations in three ways. Considerations as a result of personal experience are more likely to be believed and accepted. and discussing politics with friends who have a large stake in a political outcome (e. whether this interaction is sought after (e. even if they contradict preconceptions. in which prior beliefs or norms dominate people’s first-hand perceptions of the world (Sherif.g.g. teacher).. but listening to political news does not. An “active role” need not be a large role: cashing a Social Security check counts as active (though barely).“Personal experience” is defined as interacting with a political issue. doctor. Certainly.. 1997).

the increased frequency of acceptance of considerations leads to a larger number of political arguments being accessible(Sherman et al. Thus. That some political experiences consist of interactions of a single type causes the resulting considerations to be narrow in scope. For instance. In fact. 1982). Second.. Individuals with repeating personal experiences are better able to generate considerations from interactions. she develops a better understanding of the successes and failures of the health care system. 5 17 . As the nurse learns the ins and outs of the hospital at which she works. which emphasizes that repeating pictures or phrases affect future actions (Sawyer. by a survey interviewer). experience leads individuals to process these considerations in a consistent manner (Fazio and Zanna. often on one side of the issue. so the distribution of considerations is narrow. a consistent set of considerations is accessible and the resulting opinion distribution has a small variance. 1982). 1973). Not only do some individuals receive similar experiences over time. A nurse and a patient who both witness a failure of the health care bureaucracy may reach different conclusions about the system.First. but the similar situations in which they have these interactions are most likely to be stable.5 The result of repeated. narrow experiences being processed under similar circumstances leads to a locus of considerations. Third. Experiences generated from this interaction are likely to indicate that public schools need more funding. Wood. 1981. consider a teacher who works at a crumbling school. the experience is often based on repeated interactions or one important event. Since the individual Similar conceptions of repeated interactions are present in advertising literature. The single-sided considerations generated by this personal experience contrast with the two-sided considerations presented in news reports about school budgets or vouchers. the patient might not be able to accurately identify the underlying problem. when the set of considerations of an individual with issue experience is sampled (e..g. Consider a new nurse in the health care field.

Stability Postulate: Individuals who have experience with an otherwise political issue in their non-political lives will have more stable opinions on the issue.. However. the desire to hold a consistent worldview (the mechanism behind cognitive dissonance) might lead the nurse to develop a nuanced view of health care policy. current events. As detailed in Section 2. 18 . While these considerations may not be explicitly political.is experiencing these considerations herself (rather than receiving the information from another party). In this case.2. The three components of the second criterion are not individually necessary for the Stability Postulate to hold.2. The key feature 6 Zaller’s (1992. For instance. owning a gun and wanting to keep it for hunting). Chapter 4) example of a teacher who realizes schools’ need for funding yet is confused about the Federal budget is an example of considerations generated through personal experience that are not directly related to the interviewer’s political question.6 Consequently. the personal experience of the nurse dealing with the health care system might lead her to understand both sides of the ideological debate. where the voter’s considerations are all on one side of the ideological spectrum (e. the considerations are likely to be accepted. In the former case.7 The output of Zaller’s model under repeated experiences is a stable (though not necessarily extreme) political opinion. the relationship between experience and stability holds regardless of whether the experiences received are ideologically one-sided or twosided. 7 Experienced voters may be more likely to pay attention when a politician raises a particular issue (Hutchings. studies demonstrate that personal experience often leads voters to develop a more sophisticated approach to the issue in question. 2003). from time to time. considerations generated by experience are likely to be sampled when the issue is explicitly raised by a politician or a survey interviewer. may encompass the issues that a voter deals with on a daily basis.g. the process of developing a stable opinion is straightforward.

this stability framework is referenced when he discusses attitude change (Chapter 7). then the new information will have little impact on the reported opinion. The next section further develops this idea using a learning model. A second set of circumstances in which personal experience may lead to stable issue opinion without the effects of the three sub-criterion occurs when an individual experiences by a single. frequent interactions— in that the same considerations are always accessible when touched upon by the politics events. one issue. the Personal Experience Model represented in Figure 2. the voter has a distribution of Zalleresque considerations 8 For clarity.. Thus. she combines this new information with existing considerations. accepted. political communication) on the issue. and one candidate. In its simplest form. Returning to Zaller (1992). such as having an abortion or being the victim of a violent crime. the model includes one voter. I drop the usual subscript i to indicate a single voter.3 Learning Model: Issue Positions Following the lead of Achen and Bartels (2006). Prior to receiving a candidate’s signal. 19 .e.of personal experience is the heightened flow of received. When an individual receives political communication on an issue. After an experience. life-altering incident. which may never be forgotten and always be readily accessible.8 The model has two periods: before and after the candidate sends a signal (i. the voter will carry considerations generated by the experience. If the existing considerations carry a large weight relative to the new information.1 can be formalized using a learning model. 2. and accessible political considerations that are independent of political predispositions.1. which will demonstrate why issue stability is crucial in the arena of voter persuasion. the experience has the same effect of the first and third sub-criterion—narrow.

x1 ∼ N (δ1 .2 about the issue.1. I assume that this prior is normally distributed. I use precision (the inverse of variance) because precision more closely aligns with the concepts of the model.. voters are assumed to retain all considerations. on the issue. the voter’s issue position.e.9 For simplicity.” Voters update their beliefs about the best policy by adding the candidate’s consideration to the original distribution.3) The posterior distribution (x2 ) does not follow Bayesian updating because Zaller’s framework implies a mixture distribution (see discussion 2.2) (2. This opinion has a precision (Bartels.. Not all signals are created equal.1). hence. τ1 (2. I model the candidate’s announcement as a normal curve centered on γ and with precision ψ 2 . x1 . 9 20 . thus increasing the impact of the candidate signal. τ1 . The voter’s period 2 issue opinion is a mixture distribution with a mean and a variance of 2 δ1 τ1 + γψ 2 2 τ1 + ψ 2 τ 2 (δ1 − E[x2 ])2 + ψ 2 (γ − E[x2 ]2 ) Var[x2 ] = 1 + 1 2 τ1 + ψ 2 E[x2 ] = (2. 1 2 ). is the distribution.1) At the end of period 1. signal) depends on political factors such as how forcefully the candidate argues for the position and whether voters consider the position “cheap talk. which. However. 1993).10 To avoid discretization of the posterior belief. γ. Instead of modeling stability in terms of variance.2). forgetting period 1 considerations can be approximated by increasing ψ 2 . if the Stability Postulate is correct. However.e. increases with personal experience. in period 1. centered at δ1 . a candidate announces a position. Forgetfulness (i. The persuadability (ψ 2 ) of the candidate’s message (i. the FIFO queue) is revisited in succeeding sections. 10 In this instantiation of the model. both framework’s predict the same mean (Equation 2.

0 Position on Environment Position on Environment Position on Environment Figure 2.5. may have held unstable beliefs about how many resources the United States should commit to stop global warming. Brody and Page (1972) would label this interaction between the liberal citizen and Al Gore’s documentary “persuasion.2.152 ). The right panel displays the resulting issue opinion 2 distribution in bold. the environment was not often a topic in America’s political discourse. The voter accepts this signal and becomes a fervent environmentalist (right panel).5 1.5 1. since the liberal knew little about environmentalism before viewing the documentary.0 −1. Instead.0 0. The liberal’s initial position (left panel) is moved close to the position espoused by Gore (center panel).5 0.0 −0.2. Gore’s documentary sends a strong signal to the left of the voter’s mean prior distribution (center panel).5 0. is Al Gore’s movie An Inconvenient Truth. ψ 2 = 1/(0.0 −0.” However.0 0.32 ). Voter's Prior Beliefs Candidate's Signal Voter's Posterior Beliefs Density of Beliefs −1. she could not have thought critically about the subject.2: A hypothetical liberal’s response after viewing An Inconvenient Truth.0 Density of Beliefs −1. This uncertainty is depicted by the wide distribution in the left panel of Figure 2. Before the movie was released.0 0.5 0. with no ability to judge 21 . who is a favored politician.0 −0. while generally believing in environmentalism. Parameters: x1 = −0. the liberal accepted the cue from Al Gore.A concrete example of this model. γ = −0. τ1 = 1/(0. Even liberals. albeit outside of the campaign framework.5 1.

1.5 for a formalization of this statement. 2007). signals from elites may fail to confirm the individual’s current worldview. (1960) emphasize a similar point when they note. 2002). then their opinions will polarize (Figure 2.e. is crucial: when voters do not have experience on an issue and this issue is emphasized during the campaign. in contrast to the Democratic nominee’s 11 See Section 2.. if an individual has experience on an issue. Because most people go about their lives generally oblivious to many political issues. 22 . little personal experience). Alternatively. “For many voters the details of the political landscape may be quite blurred until they are brought more into focus during the campaign period” (pp. Consider the following example of a campaign between two candidates of the major parties who are debating the issue of gun control. The voters accept only the signal from the candidate they trust more. 1981).g. I label the shift of opinion by the liberal as “cue-taking. 135-136). cue-taking is prevalent in politics. the electorate polarizes and individuals’ existing political worldviews are not altered. If the campaign environment induces cue-taking. The Republican candidate has a moderate position.” I reserve the term “persuasion” for circumstances akin to “central route processing” (Petty and Cacioppo.. Parker-Stephen. This conclusion. supported by existing research (Ansolabehere et al.11 If the voters have unstable opinions about the issue (i.3). in which a voter’s issue opinion shifts based on critical analysis of the current set of issue considerations. Consider a campaign environment in which both the Democratic and the Republican candidates provide opposing signals to two voters (also one Democrat and one Republican) about an issue. Gilens and Murakawa. Following other scholars (e.the quality of the arguments. then a straightforward application of the model predicts that campaigns with dueling messages should polarize public opinion when a new issue is introduced to the political landscape. Campbell et al. 1994.

2 τ1 = 1/(0. In between periods 1 and 2. left-wing view. In general.e.1.4.0 Figure 2.5 0. derived mathematically as hypotheses: Cue-Taking Hypothesis: Ceteris paribus.5 0. Red=GOP) 1.8. τ1 = 1/(0.4. forms the first of the two main 23 . precise prior opinion) while the Republican does not (wide prior opinion).Voters' Prior Beliefs Candidates' Signals of Issue Stances Voters' Posterior Belief Density of Beliefs −1. more orthodox.5 0.152 ). right-of-center signal (Figure 2. left panel).0 0. this phenomenon.5 Issue Opinion Scale (Blue=Dem.0 −0. the Democratic candidate sends a leftist signal and the Republican candidate sends a moderate.7. the segments of the population that have the least ∂|γ−E[x2 ]| ∂τ 2 < 0.0 0. right panel). γ = −0. ψ 2 = 1/(0. center panel). γ = 0. Red=GOP) 1.0 Density of Beliefs −1. the Democratic hunter does not move left to any significant degree (Figure 2.0 0.0 −1. ψ 2 = 1/(0.32 ). when a favored political elite provides a signal about his position on an issue.5 Issue Opinion Scale (Blue=Dem. Democratic voter and candidate parameters: x1 = −0.0 −0. Republican voter and candidate parameters: 2 x1 = 0.252 ). Red=GOP) 1. the Democrat is a hunter who has experience with the issue (i. While the Republican voter cuetakes and moves slightly to the right to follow the Republican candidate.1.32 ). the voters’ opinions diverge in opposing directions (right panel)..3: Issue polarization. When two opposing candidates send signals (center panel) to two moderate voters of different parties (left panel). Of the two voters analyzed in this example (Figure 2.0 −0.5 Issue Opinion Scale (Blue=Dem.4.

32 ). these beliefs become more certain after receiving signals about the candidates’ platforms. Red=GOP) 1.1. The Democratic hunter does not cuetake from her preferred politician on this issue because of her highly-informed prior on the issue. ψ 2 = 1/(0. γ = −0.5 Issue Opinion Scale (Blue=Dem.0 Figure 2.5 Issue Opinion Scale (Blue=Dem. causing a contradiction in the previously loyal Democrat’s worldview. Before a signal is sent. Red=GOP) 1.8. ψ 2 = 1/(0.0 0.5 0.5 Issue Opinion Scale (Blue=Dem.5 0.4 Voters’ Beliefs About Candidate Positions Voters have beliefs about where candidates stand on the issues. γ = 0..0 −0. 2 τ1 = 1/(0. To calculate how much more negatively the Democratic voter would rate the Democratic nominee. Republican voter and candidate parameters: 2 x1 = 0. Returning to the example in Figure 2.0 −1.0 −0. Red=GOP) 1. τ1 = 1/(0.0 −0. 24 . one must consider both where the voter thought the candidates stood on the issues in period 1 and how the voter judges politicians.5 0.4: Experienced voters do not cue-take.0 Density of Beliefs −1.252 ).132 ).personal experience with the issue are most likely to cue-take (i.15.252 ).0 0. alter their issue position in the direction of the signal). Voters' Prior Beliefs Candidates' Signals of Issue Stances Voters' Posterior Belief Density of Beliefs −1. 2.1.1. the Democratic voter’s opinion is closer (on average) to the Republican candidate’s view on this issue.4. Democratic voter and candidate parameters: x1 = −0.e. The non-experienced Republican voter shifts her issue opinion toward the stance of her preferred candidate.0 0.

y1 ∼ N (µ1 . Similar to equation 2. these qualities could be correlated in certain cases.e. a voter’s prior belief on a candidate’s position may be based on general factors such as the candidate’s party. Under complete projection. Beliefs at period 1 are treated as priors in period 2. As with the voter’s own issue position. which most likely increase with political attentiveness and sophistication. period 2. υ1 (2.3. 1972). the prior belief y1 is assumed to be the normal distribution. whereby the voter assumes that her preferred candidate holds the same position that she does. and precision. Certainly. A candidate’s signal could be unpersuasive (i.12 I do not explicitly model the determinants of projection. for simplicity. beliefs about the candidate are In addition to projection. the voter updates her belief about the candidate’s positions by introducing considerations generated by the candidate signal. For example. ψ 2 . so that learning and cue-taking are based on different interpretations of a single signal.1. be informative about where the candidate stands).. υ1 .4) The candidate’s signal helps the voter identify where the candidate stands on the issue. a politician who appears to be pandering would produce an un-credible and non-persuasive issue appeal. 1 2 ).1. I model these signal characteristics as the same parameter.. γ. 12 25 . Voter have certainty about 2 these beliefs. as in Section 2. The updated. The signal has the same center point. to be an input parameter.e. instead allowing the center of the distribution of prior beliefs about the candidate’s position. not move an individual’s issue position) yet credible (i.a voter’s beliefs about the candidate’s positions often suffer from “projection” (Brody and Page. µ1 = δ1 . µ1 .

2 µ1 υ1 + γψ 2 2 υ1 + ψ 2 υ 2 (µ1 − E[y2 ])2 + ψ 2 (γ − E[y2 ]2 ) Var[y2 ] = 1 + 1 2 υ1 + ψ 2 E[y2 ] = (2. are consistent with the issue public literature (Krosnick. First. Members of an issue public place above-average weight on a specific issue. 1986). Second.a mixture model with mean and variance. voters must assign a weight to each issue. 26 .6) Returning to the case of the Democratic hunter and gun control. These signals are combined with the voter’s prior and result in a posterior belief about the candidates’ positions (right panel). The Republican has more credibility in revealing his own position than his opponent’s because the Republican might be held to account for his stated issue opinion if he wins (Ferejohn. 1957) over a range of issues (Enelow and Hinich.5).1. Before the signals are sent. candidate evaluation is modeled as utility maximization (Downs. Figure 2. the voter might have weak priors on the candidates’ positions according to their parties (left panel. voters must know their own position on all the issues. voters must know the politician’s view on each issue. 1990). imagine that the Republican candidate microtargets the Democratic hunter.5 Voters’ Candidate Evaluations Traditionally. highlighting both candidates’ positions. where voters place weights on a range of issues. Third.5) (2. 1984). These types of models. The weight-based model requires that voters keep track of all this information. These models of voter utility require a lot from citizens. 2. The Republican sends a mailing to the gun owner’s household indicating a moderate-right stance for the Republican and an extreme liberal position for the Democrat (center panel).

the voter compares her prior belief about the issue to the candidate’s perceived position. I propose a more Zalleresque model of candidate evaluation.5 1. γ = 0.0 Gun Control Opinion (Blue=Dem Cand.5 1. the voter updates her belief about the candidate’s position. The Republican candidate signals that he has a moderate-right position and signals. when a candidate delivers a signal about an issue.5 1. Red=GOP Cand. Next.5 0. x1 (z) (2.) Figure 2. γ = −0. ∞ KL(y2 ||x1 ) = y2 (z)log −∞ y2 (z) dz .42 ).15.5 0...0 −0. υ1 = 1/(0..5 0.0 Gun Control Opinion (Blue=Dem Cand.1.0 −0. The voter updates her beliefs about the candidates’ positions accordingly. ψ 2 = 1/(0.Prior Beliefs of Candidates' Positions Republican Cand. that the Democrat has an extreme left position.'s Signals of Candidates' Positions Posterior Beliefs of Candidates' Positions Density of Beliefs −1.) Density of Beliefs −1.42 ).0 0. Instead. ψ 2 = 1/(0. with less credibility.82 ). forming the posterior belief y2 (as in Section 2.0 Gun Control Opinion (Blue=Dem Cand.252 ).8. The voter starts out with prior beliefs about both candidates’ positions based on their party. For instance.0 0.4).0 0.5: A Voter integrates her projection-based beliefs with one candidate’s signals about both candidates’ opinions. Red=GOP Cand.5.5. υ1 = 1/(0.0 −0. Voters compare a candidate’s issue signal against their considerations of the issue. This comparison is formalized with the Kullback-Leibler distance. Red=GOP Cand.) −1.7) 27 . to evaluate the candidate on this issue. Pa2 rameters for belief and signal about the Democratic candidate: y1 = −0. Parameters for belief and signal about the Republican candidate: 2 y1 = 0.

One incongruent signal from a 13 x1 (z) and y2 (z) are the densities of x1 and y2 at z. Candidate stances on issues for which the voter has a vague prior will have little impact on the voter’s worldview. x2 and y2 approximate an unmixed Gaussian distribution only in cases in which one of their constituent Gaussian distributions dominate the others. For Gaussian distributions. 1984) yet the exact distribution is unimportant.3).8 is sometimes a poor approximation for Equation 2. This contrasts to a Bayesian model. reinforcing existing predispositions. Enelow and Hinich. The uncertainty of the voter’s position means the KullbackLeibler distance will not be extreme in either distance or closeness. the voter compares her beliefs with the candidate’s and the difference may be stark. x1 is the voter’s issue position at period 1. the Kullback-Leibler distance has the analytical solution. the K-L distance will shrink. Thus.7. (2. respectively. Candidate evaluations may not be reinforced when the voter has a precise prior on the issue in question.13 The use of the Kullback-Leibler distance is appropriate for a Zalleresque model because the entire distribution of considerations is compared. as above. Equation 2..8) 2 where µ2 and υ2 are the mean and precision of y2 (Equations 2.6). 1 2 υ2 τ1 2 τ1 2 υ2 KL(y2 ||x1 ) = 2 log + 2 + (µ2 − δ1 )2 τ2 − 1 . in which means (or other summary statistics) are compared (e. An apt analogy is that the voter treats her personal experience as the “ground truth” for judging the elite’s reliability. In this case. However.8 is an idealized.5 and 2. Equation 2. In addition.1. Once the voter cue-takes on this issue (Section 2. continuous version of the discrete process of comparing a voter’s considerations about an issue to considerations of a candidate’s platform.where z is the issue dimension and.g. 28 .

. thus reinforcing predispositions. candidate utility would not change at all. Let the scope of J issues be represented by j = 1. a candidate signal on a new. J U (yt· |xt· ) = − j=1 δj ∗ KL(xtj ||ytj ). In a two-sided issue debate. The chance of a switch in candidate support is particularly acute in three situations. a small change in the relative K-L distances results in a change in vote choice. In D R the first. However. To calculate the voter’s overall utility of a candidate. Second. δj = 0. the voter is nearly undecided in the first period. the model assumes voters receive only signals from the candidate with the higher U (yt· |xt· ). if considerations of this new issue replace considerations of an older issue for which the voter also took an elite cue. In this case. 29 . priming may occur. The voter need not take an active role in consciously setting δj for all issues. “priming” is a traditional political science explanation of why voters alter their opinions of candidates once a signal is received (Iyengar and Kinder.. Understanding the full voter-candidate interaction requires the consideration of multiple issues. KL(x1 ||y1 ) ≈ KL(x1 ||y1 ). e.. where D and R superscripts represent the two candidates. As noted. which is formally. for issues the voter is unfamiliar with.preferred candidate might be enough to induce the voter’s defection.J and the positions of the voter and the candidate at time t be xjt and yjt . and these considerations may be on multiple issues. unfamiliar issue would lead to a small K-L distance after cue-taking. Which combination of issues is on her mind determines the voter’s candidate evaluation. respectively. (2.9) where δj is the proportion of a voter’s considerations dealing with issue j.g. all considerations on the top of a voter’s head must be taken into account. The voter need not have all these issues on the top of her head at one particular time.

Transforming the voter’s worldview and cue-taking on all the other (non-experiential) issues from the Republican candidate causes no dissonance. formally. then the voter is susceptible to micro-targeting by the non-preferred candidate.e. voters know the “ground truth” and need not rely on political elites for an indication of what to believe. For these issues. the voter relies more heavily on these primed considerations. which is the main hypothesis of this dissertation: Candidate Evaluation Hypothesis: Ceteris paribus. This dissertation focuses on the third mechanism: voters’ political worldviews can be altered by candidates’ stances on the issues with which the voter has experience. When asked to evaluate the candidates. the segments of the population that have the most personal experience with the issue will alter their perceptions of the candidate more than the segments of the population that have no personal experience. this Democratic hunter is susceptible to defection by a well-targeted Republican appeal. after receiving a candidate’s signal on an issue. Remaining loyal to the Democratic candidate would cause cognitive dissonance. Third.1989). This phenomenon. Hence. 30 . The Republican’s signals about gun control indicate that the Republican candidate is closer to the voter on this issue. if the voter has a precise prior about only the one issue at hand (i. The signal about a particular issue (j) causes considerations about that issue to become more accessible in period 2. results in the following prediction. the voter has no experience on other issues). δj increases. Consider the example of the Democratic hunter. Political worldviews originate with and change based on issues for which voters have experience. the hunter does not cue-take from her initially preferred candidate (the Democrat) since she is set in her ways.. The equations above demonstrate that the voter will alter her perceptions of the candidates to favor the Republican. derived mathematically as ∂KL(y2 ||x1 ) 2 ∂|µ2 −δ1 |∂τ1 > 0.

6 displays these starting preferences. competence is good) but less certain about politicians’ stances (e. for traits.This hypothesis might be particularly relevant during primaries. and Doris the doctor. where no partisan differentiation exists. Ted favoring Nathan. Especially in primaries.. 15 14 31 .g.g.g. A primary campaign magnifies the effect of issues. Figure 2.g.. Mandy and Nathan. No issue considerations are on the top of voters’ heads on the initial day of the campaign (phase 1).14 They start with opposing predispositions about the candidates. claims of competence are cheap talk). these predispositions are often aligned with partisan allegiances and are very strong. See Table A.1 of Appendix A for the exact parameter values that generated the figure. Since the Candidate Evaluation Hypothesis does not rely on the pivotal issue being of extrinsic importance to the voter.. competence. Ted the teacher. each of whom has experience on a particular issue. with Alice favoring Mandy. The left-most portion of Figure 2. The 20-day campaign moves through three phases and focuses on two issues: health care and education. These predispositions may be attributable to valence evaluations (e. voter-candidate shared gender or race). and Doris undecided. voters are very certain about where they stand (e. In a general election. experience) or other traits (e.15 Candidate traits are modeled just as issues are. the hypothesis still predicts changes in political judgments based on a familiar issue.g. Consider the hypothetical example of a primary election between two candidates with an electorate of three voters. which in turn may originate from identity politics (e. candidates may agree on the salient. The three voters are Alice the architect.. oft-discussed issues. shares my values). they must decide between two the candidates.6 tracks the hypothetical voters’ candidate evaluations over the course of the campaign. Voters’ preferences between candidates with similar ideologies could easily hinge on the evaluations generated by this ground truth effect..

In the second phase. the teacher (Ted) switches candidate allegiances because his previously preferred candidate (Nathan) takes what Ted considers an incorrect stance. on the last day of 32 −4 . In the next phase of the campaign (days 2-10). days 11-20. when education is the hot topic. the initially undecided doctor (Doris) sides with the candidate (Nathan) who concurs with her stable opinion on health policy. when health care is discussed.Relative Evaluation of Candidates (Mandy − Nathan) 4 More Pro−Mandy 2 Alice the Architect 0 Doris the Doctor Start of Campaign Ted the Teacher −2 More Pro−Nathan Health Care Phase 5 10 Days of Campaign Education Phase 15 20 Figure 2. and increases her relative evaluation of him. education policy becomes relevant in phase 3. the issue of health care becomes prominent and politicians stake out positions. Considerations related to health care are added to the tops of voters’ heads. The education debate eventually crowds out health care considerations. Alice and Doris cue-take from their preferred candidate. two candidates. In the first issue phase. After the candidates have finished debating health care. By this point. Ted the teacher sides with his previously unfavored candidate. The only voter with experience on this issue is Doris (who is a doctor). The other two voters polarize in their opinion of the candidates. and two issue phases (three phases total). Mandy.6: A hypothetical primary campaign with three voters. The other two voters polarize. as actual issue information (rather than trait information) solidifies their respective choices. She sides with candidate Nathan on the issue.

education becomes as polarized as health care and voters’ preferences diverge. are an excellent generator of these stable opinions. A real-world example from the 2008 general election is the sudden emphasis on wealth redistribution after the third debate between Obama and McCain (in which “Joe the Plumber” was referenced 24 times). The most important distinction between these two types of considerations is that acceptance of candidate signals depends on the voter’s evaluation The decrease in polarization for Alice and Doris at the beginning of the education phase is caused by a decrease in polarized health-care considerations.1. which are partially offset by education considerations that have yet to become fully polarized.16 This simulation illustrates the futility of campaigns’ attempts to alter the minds of voters on a host of issues with which the voters have no experience. 16 33 .the hypothetical campaign. As voters learn the candidates’ positions on education and adopt the position of their preferred candidate.pollster. Despite this shift in campaign dialogue. which occur regardless of which issue is on the political front-burner. php) for the flat McCain and Obama lines before and after the October 16. 2. They are both considerations that are received by voters and subsequently accepted or rejected (Figure 2.1). this futility is only a problem for campaigns trailing in the polls. The critical factor for the an issue to induce political ground truthing is a stable opinion on the issue.6 Issue Experience as Signals In the Personal Experience Model candidate signals and issue experiences are variations on the same theme.18 The simulation also demonstrates a distinction between the Personal Experience Hypothesis and other alternative hypothesis: issue salience is constant across voters yet heterogeneous political judgments occur. 18 See Charles Franklin’s moving average (http://www. Personal experiences. 17 As discussed in Chapter 6. Ted switches candidate allegiances and supports Mandy over Nathan. the polls barely moved. just as Alice and Ted diverge in the health care phase of the campaign.com/polls/us/08-us-pres-ge-mvo. 2008 debate.17 In the normal course of a campaign. two-sided elite messaging on issues unfamiliar to voters leads to polarization.

g.1.. The process of issue opinion formation becomes more complicated when the individual has experience with an already politicized issue.e. As long as the precision of experiential considerations (ψ 2 ) is greater than the background noise considerations (τ 2 ). How much this consideration affects the voter’s opinion depends on the weight (i. With the model formally explicated. precision) of the new consideration (ψ 2 ) relative to the weight of existing considerations (τ 2 ).of the candidate. assume a conservative parent has a narrow prior on the issue of gay marriage because the issue is often discussed in the political arena and this individual accepts only conservative arguments. more ephemeral experiences. and again a tighter distribution of experiences and a larger ψ 2 . since voters forget considerations over time. When a voter interacts with a non-politicized issue..g.1. ψ 2 > τ 2 → ∂Var[x2 ] ∂ψ 2 < 0. her issue opinion proceeds analogously to the logic of Section 2. Third. then the math operates in the same way as for repeated. mass media-based). Infrequent.3. A real-world experience (e. forgettable voter-issue interactions may increase the variance of a voter’s opinion.. while acceptance of personal experiences hinges on the proximity of the voter-issue interaction. schools need more funding because they are crumbling).g.19 Formally.2 can also be formally linked to greater issue stability. If these considerations are forgotten at a slower rate than other considerations (e. For instance. 34 . high-impact experience. the three properties of personal experience discussed in Section 2. Second. If the parent’s child 19 Another type of experience is a one-time. repeated considerations lead to voters having experienced-based considerations on the top of their head more often. increasing the proportion of experiential considerations will lead to lower issue opinion variance. repeated considerations at the same point on the issue spectrum create a narrow distribution of experiences and a higher precision. ψ 2 .. First. easier processing of experiences leads to higher reception of considerations. a teacher sees the roof of a school cave in) generates an additional consideration for her issue opinion distribution (e.

0 −1.5 1.7: A hypothetical conservative parent’s issue consideration distribution after her child comes out of the closet. the conservative would reject the liberal considerations of the center panel.5 0. personal 2 situation.“comes out of the closet.0 −0.7). 2008).0 0.1. personal experience can balance political predispositions – a result found even in the highly politicized environment of Congress (Washington. and even then perhaps at a slow rate.0 −0.” then the parent’s personal experience will directly conflict with her existing issue opinion. but these considerations are generated by a close.5 1.0 Position on Gay Rights Position on Gay Rights Position on Gay Rights Figure 2. Parameters: x1 = 0.0 0. Normally.5 1.7 displays.152 ).0 −0. The right panel displays the resulting issue opinion distribution in bold. Voter's Prior Beliefs Personal Experience Voter's Posterior Beliefs Density of Beliefs −1.5 0.0 Density of Beliefs −1. As Figure 2. ψ 2 = 1/(0.12 ). then her issue opinion distribution becomes bimodal.20 This bimodal distribution does not generate as clear predictions for candidate evaluations as the situations discussed in Section 2. τ1 = 1/(0. 35 .5 0. and the parent may have a difficult time deciding where to stand on the issue (Figure 2.5. The conservative’s initial position on gay rights (left panel) is balanced by personal experience (center panel). γ = −0. Only because the source of these considerations is so close to her might these considerations be accepted.5 where personal experience occurs before issue politicization. 20 The initial “if” in this sentence is a “big if” in that the parent’s prior distribution would normally mean that she would not accept liberal consideration. If the parent accepts these personal experience considerations.5.0 0.

As modeled. an additional parameter and process could be explicitly added to the model. so τj is large. Voters’ ideal points. then even liberal voters will shift their opinions toward the conservative end of the spectrum.7 Model Extensions and Details The model. character traits can be a powerful force in politics because voters judge individuals in their daily lives. Also. most voters know with a high degree of certainty that they prefer honest people to dishonest people (since they have experience dealing with both in their daily 2 lives). a credible cue from 36 . news organizations. small acceptance probabilities are captured in the hypothetical election simulations by assuming weak candidate signals (small ψ 2 ). Candidate cues on this topic may lack credibility (i. as presented in its most limited form. As Zaller (1992) demonstrates with opinions about the Vietnam War. The model can be applied beyond political issues. In fact.e.” Thus.. but the effects of paying more or less attention to politics can be captured by the precision of a candidate’s signal. The probability of a voter receiving a signal does not appear in the Personal Experience Model. Consider the trait of honesty.2. For example. signals from non-preferred candidates do not contain negative information. The shift of liberal opinion is less than the shift in conservative opinion and may reflect the lower probability of a liberal accepting a consideration from a conservative elite. ψ 2 is low) because politicians often cannot credibly state “I am honest. Alternatively. xtj are all essentially grouped at positive infinity (or the upper limit of the trait dimension)—voters want their politicians to be extremely honest. can be applied in alternative situations. The model works just as well when considering political parties. if only the conservative elite is communicating to the public.1. or politically active citizens as when considering candidates and is flexible enough to be used in partisan and non-partisan situations.

However. they rely on mass communication such as television advertising. 1990) will influence candidate evaluation more than communicating about complex. campaigns would do well to follow the Broadcast Corollary and advertise about issues that intersect with the lives of the greatest number of people. Broadcast Corollary: Signals on familiar issues or candidate character traits influence voters’ perceptions more than information about more complex issues.2 Concurrence with Other Theories in the Literature The literature substantiates much of the Personal Experience Model. If a campaign is aware that a particular voter has personal experience with an issue. Circumstantial evidence supports the experienceissue interaction (Candidate Evaluation Hypothesis). and easy issues (Broadcast Corollary). Character traits are an example of an “issue” with which most of the electorate has extensive experience. trustworthiness. niche issues. then the campaign would benefit from communicating to the voter on that issue (provided the voter and the candidate have congruent opinions). yet no study directly tests the effect 37 . campaigns are often unable to procure this type of information. cue-taking (Cue-Taking Hypothesis). The model’s implications extend to these widely familiar issues. and empathy of others on a daily basis. Individuals must judge the honesty.. instead. 2. The Candidate Evaluation Hypothesis and the Broadcast Corollary have direct implications for political campaigns. In the case of mass communication. the “easy issues” of Carmines and Stimson.e.a media source that a candidate is dishonest can have a devastating effect on voters’ evaluations of that candidate. especially the claims about issue opinion stability (Stability Postulate). Aggregating the conclusion from the Candidate Evaluation Hypothesis across the electorate implies that mass communication about issues with which more voters are familiar (i.

1 Definitions of Concepts in the Literature Over the decades. various terms (e. A helpful way to think about opinion stability is to imagine asking an individual to report an opinion on a repeated basis.of personal experience on cue-taking and candidate evaluation. See “projection” in Markus and Converse (1979). • Cue-taking is the process whereby the voter agrees with her preferred candidate’s issue position without any critical thinking. Before exploring the existing research and discussing the empirical results in later chapters. “projection. See Lenz (2006) for an excellent discussion. it is useful to review and define some of these terms. The following sections review the existing evidence and identify the gap that the Personal Experience Model fills with the Candidate Evaluation Hypothesis. The variance of her reported opinions is the inverse of her issue stability. • Self-Reported Issue Opinion Confidence is an individual’s belief about her stability on the issue. • Candidate projection is the phenomenon whereby a voter assumes her preferred candidate agrees with her on an issue that she has relatively little information on.. See “projection” in Iyengar and Kinder 38 .2. • Learning is the process by which the voter ascertains the true issue position of the candidates. This mechanism often before candidates send strong signals on the issue. Confidence should not be conflated with issue stability as individuals are often poor judges of their own characteristics.g.” “persuasion”) have represented distinct concepts. It occurs after the voter receives strong signals on the issue. • Issue Opinion Stability is the precision of the distribution of repeated samples of an individual’s accessible considerations of an issue. 2.

In both experiments.2. • Persuasion on an issue opinion occurs when the voter thinks critically about issue considerations and rejects them in favor of other considerations. and “peripheral route processing” in Petty and Cacioppo (1981). Wood (1982) examines how existing experiences affect the amount of variation in issue opinions. college students are assigned to permanent or temporary housing. Regan and Fazio (1977) employ both a field experiment and a laboratory experiment to test the effect of direct experience on attitudebehavior consistency. some participants are assigned puzzles to work on. See “central route processing” in Petty and Cacioppo (1981).(1989). “persuasion” in Markus and Converse (1979).2 Personal Experience Yields Nuanced and Stable Opinions The finding that personal (or “direct”) experience with an issue induces attitude stability originates in the psychology literature. See Iyengar and Kinder (1989). Their experiments randomly assigns the direct experience treatment. In the field experiment. Persuasion does not play a large role in the Personal Experience Model. the reported attitudes of the participants assigned to the “direct experience” treatment are more consistent with the participants’ actions than for the students assigned to the peripheral experience. Subjects in this experiment are asked to “to list specific instances of times when they had engaged in actions relating” to the environment and to report their opinions about 39 . Replacement of accessible considerations over time is not considered persuasion. This salience may be raised directly by stressing the importance of an issue or indirectly by mentioning the issue and thus increasing the accessibility of considerations related to the issue. In the laboratory experiment. 2. while others receive only descriptions of the puzzles. • Priming is the mechanism by which voters increase the salience of a particular issue in the vote decision.

individuals with prior experience with the environmental movement are less likely to change their opinion. Wood requests that the subjects prepare a persuasive. McFadyen queries the subjects about their relationships with unemployed family or “close” friends. A multivariate analysis finds that amount of experience with the unemployed is the best predictor of how few stereotypes the respondents express. McFadyen (1998) conducts in-depth interviews with 67 employed Britons to probe their views on unemployment. supporting the Stability Postulate. The respondent’s answers to openended questions are coded for number of distinct ideas and integration between ideas. On average.environmentalism. with experience being more predictive than either ideology or class and leading to fewer stereotypes. the answers are then aggregated into a “complexity” measure.21 Observational studies demonstrate that individuals with experience on an issue have more stable (and sometimes more nuanced) opinions. Sotirovic (2001) conducts a telephone survey of 395 Wisconsin adults to examine their attitudes toward crime policy and the death penalty. The author asks participants about stereotypes of the unemployed. Respondents who have experience with non-trivial crimes (either personal or through friends) demonstrate higher thought complexity. and their thoughts about possible solutions. as well as any direct experience they have had with being unemployed. To gauge the level of experience that participants had with unemployment. the efficacy of government actions. this offer of money induces subjects to report views that are more anti-environmentalism than their original position (presumably because the offer of money makes the arguments appear fraudulent). 21 This change is not caused by differences in initial opinions between the experienced and unexperienced groups. A month later. 40 . pro-environmentalism talk in exchange for five dollars. However.

respondents are given a range of six methods of generating energy and are asked to pick two. The resulting opinions of individuals with personal experience are more stable than those of the public at large. The impact of personal experience on issue opinion stability is directly measured by van Knippenberg and Daamen (1996). While they do not provide exact statistics.In both the crime and unemployment studies. Van Knippenberg and Daamen measured issue involvement on a “slightly modified version” of the scale developed by Verplanken (1991). through surveys two months apart. In each survey. or even the experience of close friends or family members induces a more complex thought process about an issue. personal involvement. In the crime study. talking with friends about the issue). just the opposite occurs. which queries respondents about their activities with respect to the issue (e. these studies demonstrate that direct experience. Consistent with these findings. The authors poll the attitudes of the Dutch on energy generation twice. And personal experience with the unemployed does not push the Britons in McFadyen’s study to the left on unemployment policy. the authors report that higher involvement leads to increased stability across the two-month period. Sotirivic finds that the respondents with the most extreme views had the least complex thought processes. In fact.. political ideology is a much better predictor of issue opinion than level of issue involvement. the direct effect of experience on left-right issue position is not specified by the Personal Experience Model. Thus. 41 .g. Taken in sum. the evidence for the Stability Postulate is strong. experience does not lead to polarization of issue opinion. In the first survey. brief descriptions of each option are provided.

1. Undergraduates who hold prior beliefs in favor of capital punishment consider the pro-capital punishment study more convincing. In contrast to Zaller’s one-stream example. Lupia (1994) provides evidence of cue-taking in the 1988 California car insurance elections. consequently. In one of the clearest examples. deviate from the voting patterns of the 42 . individuals generally accept arguments that are congruent with their existing beliefs and adjust their views accordingly. Voters who have little knowledge of the substance of the ballot initiatives. their issue opinions become more extreme. on the other hand.3. Individuals who lack both substantive and endorsement awareness. 2007). Zaller (1992) demonstrates how elite consensus leads to mass consensus.3 Evidence for Cue-Taking and the Role of Experience The psychological literature provides the basis for cue-taking. Lauderdale (2008) finds that a change in the party that controls the White House alters whether conservatives or liberals favor an activist foreign policy. Democrats and Republicans diverged over even the relatively factual issue of whether Saddam Hussein was involved in the September 11th attacks (Parker-Stephen.2. Lord et al. Similar to the Al Gore and global warming hypothetical in Section 2. mirrored the behavior of high-knowledge voters. Political science research provides several examples of cue-taking. The opposite holds for the participants who enter the experiment with anti-death penalty views.2. campaigns provide an example of a two-stream environment. (1979) show that conflicting studies on capital punishment have a polarizing rather than moderating effect. In a striking example of issue opinion following elites from one end of the spectrum to the other. Similar to Bartels’ (2002) findings that the same information generates polarized responses between members of opposing parties. yet know where various interest groups stood on the measures. As the election was heating up in the summer of 2004.

it is unclear whether the voters are using positive or negative cues. a cue from an untrusted source (e. however. since two streams of information exist (e. Perhaps both self-interest and issue experience play a role in inducing students to think critically about the arguments.. e. While some students are under the impression that the policy would affect their school. However. 43 . Undergraduate participants are presented with a set of arguments for a change in university policy. 22 In this case. One of the experiments used to differentiate between “central route” and “peripheral” processing speaks directly to the Cue-Taking Hypothesis. All subjects are then exposed to a fabricated recording of an interview with a graduate student in which the graduate student presents several arguments against preservation. the insurance industry) appears to provide information to individuals about what not to believe.e.g. Students who are personally involved in the issue (i. The Cue-Taking Hypothesis. while those not involved are persuaded by the speaker regardless of the argument’s quality. others are told that the policy is under consideration at a nearby institution.. In Wood’s (1982) experiment on the environment.. The arguments (all against the policy) vary by logical soundness and overall appeal. Experiments from psychology provide evidence. those under the “own school” condition) discriminate between the compelling and non-compelling arguments.. it predicts that personal experience moderates the effects of cue-taking.more knowledgeable groups.g. Ralph Nader countering the insurance industry).22 Small cues from advocacy organizations help citizens cut through the clutter of political issues.g. membership in the Sierra Club or Audubon Society. Wood finds that subjects who report more personal experience with the environment are less susceptible to the graduate student’s persuasion even when controlling for the subjects’ initial views. the participants are asked to list their personal connections with an environmental issue. but cue-taking is clearly at work in the “other school” condition. does not merely assert that cue-taking occurs.

” which are filtered through secondary sources such as the media. Mutz (1992) finds an interesting interplay between personal experience with the economy and political evaluations. She traces the effects of experience with unemployment and knowledge gained from the mass media through the perception of the economy and incumbent politicians (governor and president). the path that translates these concerns to political significance is incomplete. divides retrospective evaluations into two categories: “simple. “These overall patterns suggest that personal experience may indeed have an indirect influence on political evaluations at the state level by means of their influence on personal and ultimately social concerns. however. Simple ret- 44 . which predicts that less persuasion (i. Examining political judgments from a similar angle. Shifting to the Candidate Evaluation Hypothesis. Fiorina (1981). the mass media has much more influence in shaping people’s opinion about the national economy.The Cue-Taking Hypothesis dovetails well with the Lupia and McCubbins (1998) model of persuasion.” which are the result of direct experience and “mediated. but being unemployed is a stronger signal of the state economy rather than the national economy (since the reason for the unemployment might be localized).e. cue-taking) will occur when “perceived speaker knowledge” is low.” This latter finding might appear to refute the Candidate Evaluation Hypothesis. If personal experience increases a voter’s knowledge (or self-perceived knowledge) of an issue—a reasonable assumption—then the relative knowledge of the speaker decreases and the Cue-Taking Hypothesis is supported by the experiments that Lupia and McCubbins (1998) conduct in the laboratory. being unemployed has an impact on state-level judgments. At the national level. Consistent with the Personal Experience Model.. Specifically. The finding that individuals are properly incorporating information from multiple sources by discounting their personal experiences when making national judgments is encouraging for any hypotheses grounded in learning models. Mutz writes. At the national level.

individuals who do not approve of the President would act analogously (assuming two issue streams). but from the perspective of voters’ experience with policy outcomes rather than issue positions. there would be no change in the correlation between issue opinion and presidential approval before and after the increase in media stories.2. individuals who approve of the President would take the President’s side and continue to approve of him. Fiorina reaches analogous conclusions to the Personal Experience Model. Gilens and Murakawa (2002) characterize the state of research by noting. 2. On the other hand.”23 The Personal Experience Model and the empirical examples presented in later chapters aim to fill that void.4 Easy Issues and Retrospection Affect Political Evaluations Existing research also supports the Broadcast Corollary. even if considerations of the complex issue are more likely to be on the top of a person’s head (Zaller. On complex (or “hard”) issues. “We are not aware of any research that explicitly compares citizens’ decisionmaking with regard to high-involvement versus low-involvement issues. individuals’ opinions of familiar issues are not as Recent work (Hillygus and Shields. Thus.rospective evaluations affect voters’ political evaluations and thus shade how they receive mediated evaluations. minimal research exists that linking personal experience to political evaluations. Since the dependent variable (presidential approval) is dichotomous. 23 45 . While the direct evidence that personal experience moderates cue-taking is established in the psychology literature. 1991) that compliments this dissertation makes this statement less true in 2009. the Broadcast Corollary would predict that changes attributed to priming would be observable only for familiar (or “easy”) issues. Kelleher and Wolak (2006) combine presidential approval ratings and content analysis of media stories to examine the priming effects of easy and hard issues. 1992).

if these considerations are prevalent (such as constant reporting on the war or economy). a similar phenomenon to easy issues occurs. candidates 46 .. their data do not exhibit the effects of priming when the media emphasizes hard issues (e. one-sided news occurs because of the resolution of an issue. Recent research (Achen and Bartels. 2005) demonstrates that this effect can be so strong that the outcomes need not be logically connected to actual policy.. 1981) have a consistent effect on presidential voting. This consensus fills the tops of all voters’ heads with one-sided considerations. When the media reports on easy issues (e.. Mueller (1970) finds that retrospective voting on these oft-reported issues (in the spirit of Fiorina. On the other hand. Often. the economy in 1992. and defense spending after 9/11. the economy. As the Broadcast Corollary indicates. Of the 13 examples of supposed priming. foreign policy).g. Lenz (2006) finds similar results when he examines over a dozen cases of suspected priming and determines that priming occurred in only a handful of cases. the President’s character). Scholars consistently label character issues and the economy as easy issues. The priming of defense spending is trickier: perhaps the lack of a competing stream of information prevented Democrats from polarizing away from the President. Lenz argues that just four are actual examples of priming: two character issues. the resolved issue mimics an easy issue. which is a highly partisan contest. Empirical research confirms this prediction. The Broadcast Corollary predicts that these resolved issues have large effects on vote choice. Kelleher and Wolak’s findings are consistent with this line of reasoning. issue priming by the media might alter their perceptions of the president. Kelleher and Wolak observe the priming effects. When elite signals on a topic are one-sided.g. 2001). such as when there is consensus that a policy succeeded or failed (Canes-Wrone et al. thus.susceptible to cue-taking.

While not all easy issues come about by the result of personal experience. The Personal Experience Model. meeting the second criterion). the results of the issue-centric political science research are consistent with the Personal Experience Model. With regard to issue publics. by detailing a fine-grained relationship between voters and issues. although the political science literature is lacking on this subject. However. Hence. a voter with experience on an issue would consider that issue to be easy.who can take advantage of issues that resolve in their favor are probably better off focusing on those issues rather than microtargeting certain experienced voters.3 Alternative Voter-Issue Linkages in the Literature While much of the political science literature focuses on the characteristics of a voter (e. The Personal Experience Model generally captures the evidence of the self-interest literature.g. attentiveness) or an issue (e. the Personal Experience Model provides a logical 47 . There is also circumstantial evidence for the Candidate Evaluation Hypothesis. These discussions demonstrate that existing psychology and political science literature provides solid support for the Stability Postulate. predicts the conclusions reached by less granular studies that examine decisive issues. Specifically.g. strains of the literature emphasize linkages between voters and issues. 2. studies demonstrate that self-interest does not affect vote choice in a consistent manner. the Cue-Taking Hypothesis. and the Broadcast Corollary. complexity) separately... research on self-interest and issue publics theorize why voters might rely on certain issues more than others in their political decisions. Easy issues are issues for which the voter does not need to rely on elites to form an opinion (which correspondents to the first criterion of personal experience) and are often major issues such as race relations (which keeps considerations at the top of the head.

The Personal Experience Model predicts the influence of symbolic politics: voters cue-take from their respective group’s leaders. easy issues). this logic also broadens the scope of the theory to include both issues and character traits. By focusing on which issues are easy for which voters. 48 . Individuals with a vested interest in a certain policy change (e. (2001) demonstrate that self-interest in a policy can be primed or unprimed depending on the political communication delivered. also find that self-interested voters can be swayed against their direct economic interests with sociotropic arguments. Sears and Funk (1990) find that self-interest is rarely a factor in vote decision. symbolic politics often drives individuals’ policy stances. Sears and Funk find that self-interest does play a role in politics—specifically when the benefits are clear or the stakes are high (i. instead.. it is unclear how much self-interest would affect political judgments. Chong et al. domestic partner health benefits) display a higher propensity to favor the change after their self-interest in the change is highlighted. however. and do not alter their political evaluations.. Although the authors do not measure voters’ evaluations of candidates who take a position on the policy changes tested in the experiment. Chong et al. the Personal Experience Model extends the findings of Sears and Funk and yields more nuanced implications. In a climate with conflicting message streams (one that primes self-interest and one that primes sociotropic considerations).foundation for the conclusions reached by the issue publics literature. One perspective on issue-voter interaction is that voters base their political decisions on self-interest. Consistent with Chong et al.’s finding that political communication can reduce the salience of self-interested consideration. one logical conclusion might be that politicians should communicate to voters on issues in which the voters have a self-interest and that these communications should emphasize that connection. On the other hand.g. polarize. In some cases.e.

For instance. Krosnick (1990) develops the idea of issue publics first presented by Converse (1964).. certain issues are crucial for political judgments because of the conscious weight that a voter places on them. in the aftermath of Hurricane Katrina. and personal values as the bases for how voters choose issue publics. First. inherently political or not—that form the bases for voters’ “ground truth” for judging politicians. Second. Left unsaid is the high likelihood that the hairdresser was in the immigration issue public because of her personal experience emigrating. Voters in an issue public have more stable opinions about the issues and base their political decisions on these issues. These issues are most likely to be reported as “most important” on a political survey. The Personal Experience Model shifts and extends the idea of issue publics in two ways. emigrating from Russia) rather than choosing to be a member of an issue public. increased personal importance of the hurricane devastation moderates the effects of partisan polarization and cue-taking. Voters often have these experiential issues foisted upon them (e. receiving Social Security checks. 49 .g. the mechanism by which some issues are more predictive of political preference than others—the desire to avoid cognitive dissonance—is the result of life-happenstance. Krosnick emphasizes self-interest. In the issue publics framework. she relates a story about a Russian hairdresser who was in the immigration issue public. Although the vast majority of voters do not closely follow the political debate on most issues. many voters find one or two issues to be important. in the Personal Experience Model. the level of personal experience explains why a person might enter an issue public. In contrast. in the preface of Gershkoff’s (2006) dissertation. the Personal Experience Model does not rely on significant differences in issue salience to affect political preferences. Malhotra and Kuo (2009) demonstrate that.In a second thread of political science research. The Personal Experience model focuses on issues—big or small. group identification. Gershkoff (2006) confirms these findings in an extensive analysis of issue publics.

e.Finally. the Personal Experience Model hypothesizes that personal experience plays a crucial role in political evaluation. which would moderate cue-taking). voters with personal experience involving that issue will alter their evaluations of the candidates more than voters who lack such experience will. More pertinently.. Second. Individuals with direct experience with a topic are more likely to have higher confidence in their attitudes as well as to match their behavior to their attitudes. 50 . which makes them important to campaigns which aim to persuade voters. First. cue-taking) based on an individual’s political predispositions. This type of personal experience allows a voter to ground truth candidates’ statements and explains some observed heterogeneous movements in party and candidate preferences.e. part of the picture would be unmodeled (i. This causal mechanism is explored in Chapter 4. voters who have more confidence in all their opinions (regardless of whether this confidence is justified) may exhibit less cuetaking than would be otherwise expected.. Fazio and Zanna (1978) propose that confidence is an intervening factor between the experience-issue stability link. These candidate preference changes may run counter to existing partisan biases. If this research is correct. Thus. when Fazio and Zanna exogenously increase confidence in an attitude (independent of how the attitude was formed). while the Personal Experience Model would still hold as presented (because direct experience would lead to confidence. it predicts that personal experience moderates voters’ adoption of political elites’ issue opinions (i. a candidate stakes out a position on an issue. The important features of personal experience is that (1) the individual plays an active role thereby circumventing partisan filters and (2) the considerations generated by the individual’s experience are accessible. The theory generates two main hypotheses. In sum. the correlation between attitude and subsequent behavior increases as well. more confident people are less prone to cue-taking).

Chapter 3 Experienced Evaluations and Self-Interested Opinions: Observational Evidence from the 2000 Presidential Campaign 3.1 The Personal Experience Model and the 2000 Presidential Campaign 3.1 The Campaign about Nothing From the perspective of political practitioners. the 2000 presidential campaign occurred at just the wrong moment in history. the 2000 electorate was highly microtargetable because no single issue dominated the campaign discourse. While it would be another two to six years before microtargeting became prevalent in national politics.1. Gore and Bush 51 .

com/images/pdf/836a16Tracking16. when the war on terrorism and the war in Iraq combined to become the single most important issue for a 1 ABC News.go. This main finding is based on the Annenberg 2000 dataset. experienced voters learned about the candidate’s stance on that issue faster than the rest of the public and judged the candidate on that issue. Specific segments of the electorate—individuals with personal experience on a political issue—were susceptible to targeting by the campaigns. which includes panel data recorded before and after the parties’ nominating conventions. This evidence provides support for the Candidate Evaluation Hypothesis presented in Chapter 2. Candidates may prime these concerns and shift individuals’ issue opinions—an effect that contradicts the predicted reduced opinion shift of the Cue-Taking Hypothesis. 2000. http://abcnews. Gore also made a push for the Patients’ Bill of Rights. In one poll. “What is the most important problem facing the country?” was education—an area in which the federal government has little say. Voters with personal experience with an issue may also have a stake in the policy outcome. Analyses indicate another phenomenon at play: self-interest. When Bush or Gore emphasized an issue in his convention speech. November 1. While the campaigns delivered competing talking points about Social Security privatization. These two campaign issues have a common feature: they exhibit evidence of issue-driven vote preference among the segment of the population with experience on the issue.1 This result contrasts with the 2004 campaign. sometimes talking past each other. The data find evidence of both phenomena occurring in the 2000 election.each focused their attention on several issues. Voters in the 2000 electorate could not agree on a single most important issue. the modal response (representing 20% of responses) to the question.pdf 52 .

responsible investments. it is not surprising that Bush’s August 3. convention speech was largely devoid of issues while Gore focused on several minor issues in his August 17 convention speech. But if you don’t touch it. The convention as a whole was a success. 2000. Internet privacy). Gore. Bush made an especially forceful push for Social Security privatization. Perhaps the lasting legacy of the convention was Gore’s coining of the oft-parodied term “lockbox”: 53 . with the economy not far behind. and don’t believe a word of it. as Bush opened up a double-digit point lead after the convention. Our opponents will say otherwise. giving Gore a lead in the polls that he would hold until the first debate in October.third of the electorate. the Republican nominee’s speech did touch upon four concrete policy proposals: education. In the 2008 campaign. the economy mattered most to over a third of the public. parting ploy. no way. No changes. Bush will keep the promise of Social Security. you made your plans. we will give you the option—your choice—to put a part of your payroll taxes into sound. after the mid-September financial crisis. This is their last.g. Bush spent the bulk of his speech defining who he is. and taxes.” the one you’re not supposed to touch because it shocks you. For younger workers. he explained the term “compassionate conservative” with multiple vignettes. This tactic appeared to work. touched on dozens of issues in his nomination acceptance speech. And I intend to fix it. military readiness. you can’t fix it. Now is the time for Republicans and Democrats to end the politics of fear and save Social Security. To seniors in this country: You earned your benefits. In this varied issue landscape. some rather minor (e. Social Security privatization. However. tobacco advertising to children. saying: Social Security has been called the “third rail of American politics. For instance. and President George W. together. in direct contrast to Bush. no reductions.

. caring for the sick. And I say to them.. It’s time to take the medical decisions away from the HMOs and 54 . and don’t have a right to play God. near their home in Everett. I met him and his parents in Seattle. Even though the media focused on the lockbox and hence Social Security. and needs full-time nursing care for several years. It’s just wrong to have life and death medical decisions made by bean-counters at HMOs who don’t have a license to practice medicine. Dylan and Christine Malone are here with us tonight. Gore actually devoted more of his speech to vilifying health maintenance organizations (HMOs) and to rallying support for a Patients’ Bill of Rights: And I will never forget a little boy named Ian Malone—who suffered from a medical mistake during childbirth. but for generations to come. enforceable Patients’ Bill of Rights.At a time when most Americans will live to know even their great-grandchildren. we will save and strengthen Social Security and Medicare—not only for this generation.. Their HMO had told the Malones it would no longer pay for the nurse they needed. Washington. that kind of common sense is a family value. and to all the families of America: I will fight for a real. too. teaching our children well.. Putting both Social Security and Medicare in an iron-clad lockbox where the politicians can’t touch them—to me. . respecting one another—giving people the power to achieve what they want for their families. And to me. family values means honoring our fathers and mothers. and then told them they should consider giving Ian up for adoption. Ian’s here..

Gore produced two television commercials about the Patients’ Bill of Rights. For analytic purposes.insurance companies and give them back to the doctors. the nurses.6).” issue-driven effects of a campaign (or other political elites) emphasizing an issue and thus altering voters’ political judgments through the issue’s increased salience (e. 55 . A change in vote choice in July. In the week after the convention. and the health care professionals. analyzing the correlations of simultaneous issue opinion and vote choice is problematic. As discussed in depth by Lenz (2006). as the model produces cleaner predictions when the issue starts with few elite-driven considerations (see discussion in Section 2.2 Issue-Driven Vote Choice Political science has its own version of the chicken and egg question: Which comes first. cannot alter a voter’s issue opinion in June.1. As with cue-taking (discussed in Chapter 2). voters often adopt the issue opinions of their favored candidate. Iyengar and Kinder. then issue opinion must be driving vote choice and not vice versa. reverse causation is a troubling problem when attempting to discern the effects of issue opinion on vote preference.g. the analysis will not detect an effect even though the issue opinion is affecting vote choice. for example. the vote preference or the issue opinion? Researchers often attempt to find evidence of “priming. One caveat is that if the issue in question already dominates the June vote choice and continues to dominate in July.. the best issues to examine are those that increase in salience in between waves of a panel data set. A more useful approach involves correlating previous issue opinion with change in vote using panel data. This empirical necessity dovetails with the Personal Experience Model.1. If the correlation increases over time. one of which focused on the case of Ian Malone. 3. 1989). therefore.

While the 56 . they meet the requirement of being lesser-known issues that were emphasized during the 2000 campaign. data on personal experience for these other issues are lacking. These voters know how helpful additional rights (such as the ability to sue) would be to them in their dealings with the HMO. such as the licensing of handguns. because these voters do not have experience with Social Security privatization. however. also fit that description. Other issues. For instance. This measure is of debatable quality. Both issues fit the assumptions of the Personal Experience Model and can be used to test the model’s predictions. it is assumed that voters over the age of 65 have experience with Social Security since they receive a check from the Social Security Administration each month. The three issues highlighted here also meet the requirement that data on personal experience be available.The issues of Social Security and the Patients’ Bill of Rights were selected for analysis for two reasons. The model assumes that there is little to no political discussion of the issue in period 1 and that subsequently a candidate delivers a signal about the issue. However. For the Patients’ Bill of Rights. The text of the convention speeches provides evidence that Bush sent signals about Social Security and that Gore emphasized his positions on Social Security and the Patients’ Bill of Rights. the Annenberg survey did not ask respondents about gun ownership. The Personal Experience Model also requires an observable measure of voters’ personal experience with an issue. it is assumed that individuals with health care through an HMO have experience with this issue. For each issue. First. They have experience with how reliable the government agency is and how much the money helps them pay for expenses. For Social Security. the analyses verify that the public entered the convention phase of the campaign with little knowledge of where candidates stood on the issue and that the electorate learned about these stances over the post-convention weeks.

This prediction is difficult to verify with the data at hand because of the self-interest connection between issue experience and issue position. experienced voters do not cue-take from politicians as much as nonexperienced voters. Gore closely tied the Bill of Rights to HMOs. Thus. retirees might believe that their Social Security checks are at risk under a privatization plan. some of them shift their issue position to be against privatization. the Personal Experience Model makes two testable predictions. HMO customers would benefit from additional rights. the candidate’s point may be of particular importance to the voter (Chong et al. pre-convention) issue opinion and change in vote preference is greater for those with personal experience than for those without. This effect is labeled “differential evaluation”— candidate evaluations among a certain segment of the population. The second prediction of the model is that experienced voters judge politicians on the issue for which they have experience (when a politician sends a signal on that issue). Differential evaluation is 57 . Thus. an experienced voter with a self-interest on one side of the issue may be pulled in two different directions. when a candidate emphasizes an issue. 2001).Patients’ Bill of Rights would apply to all citizens (not just the HMO-insured). when Gore claims that privatization would costs seniors their benefits. these older voters have no experiential considerations to counter this argument. The best example of this second phenomenon is Social Security privatization. With these assumptions validated. For example. Seniors have no more experience with how privatization would be implemented than other voters have. On the other hand. On one hand. believing it to be most relevant for the HMOs’ customers.. the voter’s issue considerations from experience may help outweigh the candidate’s argument. This pattern is evidenced in the data. Naturally. First. The empirical test of this hypothesis is whether the correlation between first period (in this case.

the issue becoming more salient. yet without random assignment of treatment. at the same time. The data are used to support hypotheses that make causal claims. evidence of differential evaluation among experienced. If that consistency holds. the Democrat). For the Democratic National Convention.1.to anti-privatization and from Bush to Gore reduces the correlation between prior issue opinion (pro-privatization. the Republican stance) and change in vote (toward Gore. the Republican and Democratic National Convention panel data from the Annenberg 2000 presidential survey are analyzed. potentially self-interested voters is likely to be understated. alternative hypotheses are tested in efforts to rule out spurious relationships and provide more confidence in the conclusions drawn. The key. Crucially. and the analysis finds only weak support for a priming effect separate from learning. The Republican convention was held from July 31 to August 3 and the respondents were reinterviewed from August 4 through 13. switch their vote choice toward Bush (or vice versa). a word of caution is necessary. 1. This latter effect is not required. verifiable assumption is that experienced voters who shift their issue opinions in the direction of the Democratic position do not. causal statements are difficult to justify. However. possibly. A senior who shifts her opinion from pro.197 respondents during July as a part of its normal cross-sectional study.a result of experienced voters learning the candidates’ positions on an issue and. While findings show that voters experienced with issues differed in their actions from non-experienced voters. 3.230 58 . the self-interest effect counters the effects of observable differential evaluation. The Annenberg Center initially interviewed 1. these differences in behavior cannot be directly tied to personal experience. or primed.3 Data from the 2000 Annenberg Survey To isolate these effects. As with many observational studies.

1 Results and Analysis Learning Over the Course of the Campaign Data on Social Security and the Patients’ Bill of Rights indicate that voters learned about these issues over the course of the campaign. 2006) present learning as a necessary condition of issue-driven vote preference shifts. HMO customers were slightly more likely to favor “giving patients the right to sue their health maintenance organization or HMO” than non-HMO customers by 78% to 75%. Prior to the nominating conventions.2 3.2. This group comprises 16. which ran from August 14 to 17. Learning was especially prevalent among experienced voters.” This group encompassed over 43% of survey respondents in the national cross section. 3.6% of the cross-sectional national study and about the same proportion in the panel studies. In the pre-convention period (May through July). respondents 65 years old or older are assumed to receive Social Security checks. If learning is more prevalent among experienced voters. 59 . Ideally. The experience measure for the Patients’ Bill of Rights issue is whether the respondent self-reports having her “main insurance through an HMO.respondents were interviewed first from August 4 to 13 and then after the Democratic National Convention. a crucial point since the Personal Experience Model and other research (Lenz. seniors are less likely to support privatization (31% support) than younger voters are (62% support). but that question is not directly asked on the survey. those labeled as having experience with Social Security would be respondents who receive a Social Security check. then the implications for campaigns of the Candidate Evaluation Hypothesis are even more stark than in a situation in which all voters learn the candidates’ positions. Instead.

Learning is slow and steady through the summer. n=605). As with Social Security. during this early stage of the campaign. In the first month that the survey asked about Gore’s position (May 19 to June 15. through nominating conventions. though slightly more (56%) know that Bush opposed privatization (n=95). with an acceleration during the nominating conventions. Seniors. and advertisements. with 72% and 77% correctly reporting Gore’s and Bush’s positions (n=460). the campaign begins with voters generally ignorant of the candidates’ stances on the Patients’ Bill of Rights. debates. after the conventions. In the last three weeks before the election.Voters initially are largely ignorant of Gore’s opposition to Social Security privatization. Then. in a separate question. only to rise again—very quickly—starting with the first debate in the beginning of October. Gore) supported a patient’s “right to sue” her HMO. voter knowledge stagnates. Bush’s position on this topic is genuinely ambiguous: he vetoed a Patients’ Bill of Rights in Texas in 1995 but let a subsequent bill become law without his 60 .1 illustrates learning over the course of the campaign. only 41% of respondents answer “no” to the question “Al Gore–do you think he favors or opposes allowing workers to invest some of their Social Security contributions in the stock market?” By contrast. a majority (51%) of respondents correctly report that Bush favored personal investment for the retirement program. are not statistically significantly more knowledgeable than the rest of the population. Seniors learn about the candidates’ platforms to a larger extent.753). 61% of respondents know that Gore opposed privatization and 67% report that Bush favored the policy change (n=2. Voters learn about the candidates’ positions during the campaign. The knowledge differences on this issue between seniors and younger voters are statistically significant. Seniors learn faster during the convention period than the rest of the population. Figure 3. The Annenberg survey asks voters whether they thought Bush (and.

Percentage of Correct Respondents

40%

50%

60%

Bush's Stance Gore's Stance Both Stances Low Daily Sample Sizes

Dem. Conv.

Rep. Conv.

80%

Seniors Only

70%

30%

All Respondents

June 1

August 1 Date

October 1

Figure 3.1: Loess smoothed curves of voters’ knowledge about Bush’s and Gore’s positions on Social Security. While both young and old voters learn about their issue positions during the course of the campaign, seniors gain knowledge faster. Daily sample sizes average 23 prior to July 4 (inclusive) and 149 after July 4. signature when a veto-proof majority of the Texas legislature passed it. However, in the third debate, Bush claimed credit for this law and touted its provision that gives patients the right to “sue an HMO for denying you proper coverage.”2 Gore’s position on the Patients’ Bill of Rights, however, is unambiguous: he supported the right of patients to sue their HMOs. Despite this clear position, in the first month of the Annenberg survey, only 42% of respondents know Gore’s position (with 38% claiming not to know). By the final two weeks of the campaign, this figure increases to 48% (with
2 As President, Bush’s Department of Justice successfully argued in Aetna Health, Inc. v. Davila, 542 U.S. 200 (2004), that federal law superseded the Texas Patients’ Bill of Rights, thus disallowing patient lawsuits.

61

31% “don’t know”). Learning among voters who used an HMO for health insurance is slightly faster, increasing from 42% to 51% over the same time period. The difference in knowledge at the end of the campaign between voters who are experienced with HMOs and those who are not is statistically significant.

3.2.2

Cue-Taking and Self-Interest

The issues discussed in this chapter—Social Security and the Patients’ Bill of Rights— are issues in which experienced voters are have vested personal interests. The Personal Experience Model and theories of self-interest yield competing hypotheses, with the 2000 Annenberg data helping to adjudicate this dispute. The results, however, are mixed: both experience and self-interest affect changes in voters’ opinions. In the three cases tested, one case (Patients’ Bill of Rights) supports the Cue-Taking Hypothesis, another case (Social Security during the Democratic convention) supports the theory of self-interest, and a third case (Social Security during the Republican Convention) yields mixed evidence. Most likely, both phenomena occur in all cases and the resulting effect depend on the strength of the self-interest and the personal experience. Before analyzing whether personal experience or self-interest hold sway over voters, another hypothesis—issue-opinion cognitive dissonance—must be considered. If, as the issue-opinion cognitive dissonance hypothesis predicts, voters accept only congruent signals on issues, then no voter would ever alter her issue position. If, on the other hand, the Personal Experience Model’s assumption is accurate and voters accept signals only from favored politicians, then voters should adopt the positions of these politicians as issues become politicized (see discussion in Section 2.1.1). Theories of self-interest suggest a similar voter shift, though with different causal connections: in that model, voters are primed on self interest and change their issue opinion and candidate evaluations accordingly. Hence,

62

both the Cue-Taking Hypothesis and self-interest theories predict increased correlation between issue opinion and candidate preference as an issue becomes politicized. The data from the 2000 campaign support these latter two hypotheses. For each of the four issue-convention pairs, the correlation between pre-convention vote preference (trichotomous) and issue position increases over the course of the convention. The panel nature of the data ensures that reverse causation does not affect these results (i.e., postconvention vote preference is ignored). Only the Social Security correlation boost during the Republican convention meets standard levels of statistical significance (p = 0.10, twotailed). However, as the left panel in Figure 3.2 shows, all four shifts in correlation are in the expected direction, and jointly the analysis is significant (p = 0.09).3 The right panel of Figure 3.2 displays the correlations between the more granular 7-point party affiliation and issue opinion, revealing a similar pattern. Evidence for the Cue-Taking Hypothesis exists but is weak. Of the three issue-convention pairs, two have the expected sign of the Cue-Taking Hypothesis, although only one, Patients’ Bill of Rights, is statistically significant. The other, Social Security during the Democratic convention, is nearly exactly zero. For the Republican convention, seniors are less likely than other voters to have stable opinions. This phenomenon is a result of seniors shift, in a self-interested manner, from a pro- to an anti-privatization stance. Support for the Cue-Taking Hypothesis is measured with a logistic regression of postconvention issue opinion on pre-convention issue opinion with an experience interaction term. Controls for convention watching, income, political interest, and party (including their interaction terms) are included. Figure 3.3 displays the quantity of interest: the difference in probability of issue opinion change between experienced and non-experienced voters. Negative values are indicative of the cue-taking moderation: experienced voters
The calculation of a joint p-value assumes that the tests are independent. When comparing analyses of separate groups of experienced voters on separate issues, this assumptions is probably valid.
3

63

0.4

Pre Conv. Party ID−Issue Opinion Correlation

Pre Conv. Vote−Issue Opinion Correlation

0.3

q q

0.3

0.4

q q

q

0.2

0.2 0.1

q

q

q

0.1

q q

q q

0.0

Social Security GOP Conv.

Social Security Dem Conv.

Patients' B.O.R. Dem Conv.

0.0
Social Security GOP Conv.

Social Security Dem Conv.

Patients' B.O.R. Dem Conv.

Issue−Convention Panel Pairing

Issue−Convention Panel Pairing

Figure 3.2: Change in predisposition-issue correlations before and after the conventions. Predispositions are measured by trichotomous vote choice (left panel) or 7-point party affiliation (right panel). Issue opinion measures have three and two possible values for Social Security and Patients’ Bill of Rights, respectively. Each correlation includes about 1,200 cases in each correlation. shifting their opinion less than non-experienced voters. Regression coefficients are presented in Table B.1 of Appendix B. The positive result for the Cue-Taking Hypothesis in the case of the Patients’ Bill of Rights holds even when controlling for alternative explanations. Potentially, voters who watched the convention might have been more likely to shift their opinions, but such an interaction term was in the opposite direction and was near zero. Also, more informed voters may have had more stable opinions; the evidence supports this idea but does not diminish the importance or significance of the experience. The positive coefficient is not solely an artifact of there being fewer non-HMO individuals who are against a Patients’ Bill of Rights; a lower proportion of experienced individuals defected from that position than

64

Point estimates (circles). Issue opinion is dichotomized. thus eliminating the possibility that HMO-experienced individuals happen to have more stable opinions across the entire political spectrum. The result also disappears when experience with Social Security is entered into the regression in place of support for a Patients’ Bill of Rights.1 Social Security GOP Conv.1. Regression details are presented in Table B. Most likely. even though Bush explicitly protected seniors in his policy proposal.2 q q −0. Percentage Point Difference in Shift Probability: Experience Voters minus Inexperienced Voters 0. and 95%-confidence interval (non-bold lines) are displayed. Dem Conv.did non-experienced individuals. 65 . these seniors believed that their Social Security checks were endangered by Bush’s plan. Issue−Convention Panel Pairing Figure 3.O. exhibit less cue-taking moderation or none at all.R. in which a significant proportion of seniors were alienated by Bush’s privatization message. This finding is especially apparent in the Republican convention panel data.0 0. on the other hand. Social Security Dem Conv. one bootstrapped standard error (bold lines).3: Increased or decreased probability of issue opinion change for voters experienced with the issue.3 q 0.1 0. Social Security recipients. Patients' B.

2. Less than a tenth of younger voters do the same. more seniors stick with that position.” 4 66 . “Allowing younger people to invest some of their Social Security contributions in the stock market will reduce Social Security benefits for older people.1. receiving a Social Security check is the weakest.e. that they receive checks) but lends little insight into how the system works. Of all the personal experiences tested in this dissertation.” Even at the end of the campaign (which is the only period during which the question is put to respondents). While self-interest is the likely cause for the issue opinion defection pattern of experienced voters. shades of self-interest are apparent. The Annenberg survey asks respondents whether they agree with the statement. this bias would exist regardless of the mechanism. In sum. Over a quarter of seniors (27%) who begin the convention period in favor of privatization switch their opinion by the end of the convention. It is not surprising then that the evidence for the Cue-Taking Hypothesis is mixed. The data clearly show that they are.’s (2001) experiments on priming self-interest. that Bush’s plan “could cut benefits for seniors.3 downward. at best. the results lean toward the opposite The Gore campaign ran an advertisement claiming. The sufficient condition is that issue opinion shifts are accompanied by like-minded voter preference shifts. the finding that experienced voters are also influenced by self-interest biases the results in Section 3. The anti-privatization movement among seniors is consistent with Chong et al. 40% of voters agreed that seniors’ benefits would be reduced under privatization. The experience interaction coefficient (Table B. For example.1) is positive for this issue-convention pair because among those initially anti-privatization.2. Gore’s margin increases by 29 percentage points among seniors who shift their issue opinions against privatization during the Democratic Convention. Receiving a Social Security check provides evidence apart from partisan predispositions that the system “works” (i. perhaps disingenuously.For the Democratic convention data. after so much information about Social Security had been disseminated..4 As discussed in Section 3.

an undecided voter who favors a patient’s right to 67 . 3. In this case. Two of the three issue-convention pairs’ results near statistical significance. the evidence for the Cue-Taking Hypothesis is strong. The Republican convention data indicate that the average undecided voter who enters the convention against privatization has a 1% chance of shifting her vote to Gore.2. all effects are in the expected direction. A negative coefficient for experience and a positive interaction term indicate that experienced voters relied on their pre-convention issue position to evaluate politicians more than did voters without experience. Support for the Candidate Evaluation Hypothesis is provided by variations on one ordinal regression. Two of the three cases exhibit higher degrees of issue-driven candidate evaluation by experienced voters. political knowledge. Accounting for the amount of time individuals watched the convention. being a member of an HMO does give an individual at least some insight into the health insurance bureaucracy. When the negative effects of selfinterest are corrected for. In contrast. For the Democratic convention. the Annenberg data demonstrate that voters judge candidates on issues with which they have experience. and the interaction between these two terms. pre-convention issue opinion. experienced voters are less likely to adopt the position of elites.hypothesis preferred by the self-interest literature: seniors are inclined to shift their opinion toward anti-privatization. Change in vote from the pre-convention period to the post-convention period is regressed on issue experience. Analyses also examine the role of priming apart from voter learning.3 Experienced Voters Judge Politicians on Issues Consistent with the Candidate Evaluation Hypothesis. For this issue. and pre-convention position disparities does not weaken the relationship. A similarly situated senior has a 5% chance of altering her vote preference for Gore.

the regressions are restricted to voters who stick with their issue position throughout the convention (Figure 3. As noted. Figure 3.4. To tease out issue-driven effects that are not a result of self-interest. Social Security GOP Conv.4 displays the differences in these probabilities (4 and 5 percentage points. triangles).2 0. squares). Difference in Probability of Issue−Driven Vote Preference Shift: Experience Voters minus Inexperienced Voters 0.4: Increase in probability of an experienced individual shifting her vote because of pre-convention issue opinion over the analogous probability for non-experienced voters. one bootstrapped standard error (bold lines). Point estimates (circles. squares.sue has a 14% chance of altering her vote in favor of Gore. In all cases.2. 68 . a similarly situated voter who belongs to an HMO health care plan has a 19% chance.1 All voters No Issue Shifters No Issues Shifters or Learners Social Security Dem Conv. Issue−Convention Panel Pairing Figure 3. and 95%-confidence interval (non-bold lines) are displayed.R. the shift of experienced voters’ issue positions based on their self-interest biases these results downward. Dem Conv. voters with experience and stable opinions are more likely to evaluate politicians on these issues than are their stable.0 q q −0.3 q q 0.1 0.O. Patients' B. respectively) added to the analogous probabilities for voters who enter the conventions with opinions on the Republican side of the spectrum. Regression details and coefficients for the “no issue shifters” condition are presented in Table B.

may be driving vote choice. 1960) analysis provides more evidence that receiving a Social Security check generates experience with that issue and tightens the relationship between that issue and Social Security. The results include controls (including interaction terms) for income. convention watching. the difference between experienced and non-experienced voters decreases.non-experienced counterparts. however. To check for the possibility of priming in the absence of learning. not priming.4. triangles). The Personal Experience Hypothesis takes a similar view: learning drives both priming and changes in vote choice. Voters who have an HMO and do not switch their position on the Patients’ Bill of Rights are very likely to judge the candidates on that issue. The results from the Democratic convention indicate that no priming occurred apart from learning (Figure 3. Regression discontinuity (Thistlewaite and Campbell. indicating that learning plays an important role in issue-driven candidate evaluation. for instance. demonstrate an increase in the salience of the issue. If experience with Social Security is driving the results then no issue-driven vote changes will be observed in the younger group—even among those near retirement age in this group. Lenz (2006) raises the point that learning. They do not. political interest and party affiliation. or priming. the regression is further restricted to those voters who answered the same number (or fewer) of “candidate position” questions correctly. The analysis divides the population into two group: those 65 years of age or older (who are assumed to receive Social Security checks) and younger voters. Experience does not cross issues. In each case. To eliminate the possibility of self-interest 69 . These analyses demonstrate that voters’ evaluations hinge on issues with which they have experience. the interaction of HMO customers and Social Security privatization during the Democratic convention has a coefficient of nearly exactly zero.

5 For an analogous 64-year-old voter. they judge politicians on the The regression coefficient that drives this calculation is not statistically significant (Table B. 22% become anti-privatization while 8% become pro-privatization (n=49).5).7 This behavioral pattern supports the dynamic nature of the Personal Experience Model (Section 2. Low n-sizes affects both of these analyses.5) indicate that as a voter turns 65. 6 5 70 .4 Discussion Observational evidence from the 2000 presidential campaign indicates that experienced voters act differently from voters who lack experience with a given political issue.3% and 11. members of HMO who have a stable opinion in favor of a Patients’ Bill of Rights and shift their vote toward Gore during the Democratic convention also become more against Social Security privatization. This 3 to 4 percentage point difference is consistent with the ordinal probit regression’s findings of Figure 3. the regressions are restricted to voters who did not change their issue positions during the conventions. respectively (n=1. The difference in predicted vote between the two coefficients is highly statistically insignificant.6 Once experienced voter alter their candidate preferences.and anti-privatization shifts. a change in her pre-convention opinion increases her chance of shifting her vote toward Gore by 4 percentage points.163). To check for a spurious pattern in the data. The results (Figure 3. This difference is statistically significant at conventional levels (p = 0. 7 Among the experience subgroup. Among other voters. the change in Gore probability is nearly exactly zero.1. The rest of the respondent population exhibits equal pro. The difference holds when restricting the latter group to pre-convention supporters of a Patients’ Bill of Rights. the same regression discontinuity is applied to pre-convention vote. the impact of pre-convention opinion on privatization and change in vote during the convention does increase. For example.4 (left-most square) for Social Security during the Republican Convention. For a 65-year-old voter.3). they cue-take from their preferred candidate.2. two-tailed).contaminating the results.09. Experienced voters learn about an issue faster and as they learn. the proportions are 11. 3.8%.

5: Regression discontinuity of voters around retirement age show issue-driven effects for seniors only. Voters who do not shift their voter preferences are not displayed (for clarity).3. points labeled ”a” are those antiprivatization. Since self-interested voters are often experienced voters. Coefficients and regression specification reported in Table B. but are included in the analysis.02 pp ap p a a Bush 50 55 60 65 70 75 80 85 Age (cutoff=65) Figure 3. Points on the top line are voters who alter their candidate preferences toward Gore (with age jiggered). issue.06 a a ap a a a p a aa a aa a pa a a ap a p Gore Predicted effect of experience: 3. such as priming. This chapter addresses this tension by attempting to tease out the different effects of self-interest and experience among the 71 . this effect is in tension with the Cue-Taking Hypothesis. Self-interest also plays a role in the issue-vote choice interaction. toward Bush. Candidates can emphasize a voter’s self-motivated preference for an issue position. Lines represent predicted impact of pre-convention Social Security privatization opinion on the probability of a pro-Gore shift in vote preference.00 0. Points labeled ”p” represent voters who are pro-privatization.Increased Probability of Issue−Driven Vote Change: Experienced voters minus Inexperienced Voters p p p pp pa p a a 0.04 0. Issue-driven candidate evaluation independent of learning. making that voter more likely to adopt that position.3 percentage points paa p pp p a ap a a p p p p p p p p p a a p ppp p p aa a p a a ap p aa aa a p a a aa p a a 0. on the bottom line. may occur but the evidence is weaker.

Rubin. these analyses indicate that personal experience with an issue can drive vote choice. the findings become clearer. identifies voters who have experience with an issue but no self-interest in the outcome. The evidence for the Candidate Evaluation Hypothesis is moderately strong. In one case. Even though the effects of self-interest counteract these issue-driven vote changes. Results are consistent with the Personal Experience Model’s causal mechanism and sufficient for campaign practitioners who care about outcomes rather than causal chains. the Patients’ Bill of Rights. Causal inference relies on manipulation and randomization of units (Fisher. may explain some of this difference. cannot state truly causal claims. When the effects of the Personal Experience Model are separated from self-interest by examining a subset of the data. Tests of the Personal Experience Model suffer from the problem that it is difficult to randomize personal experience. 1978. seniors shift their issue opinions toward their perceived self-interested position—against privatization. evidence for the CueTaking Hypothesis from the 2000 campaign is mixed. Because of the interplay between experience and self-interest. The analyses of this chapter. The differences in the two levels of experience. Instead. as well as all analyses of this dissertation. For Social Security. this dissertation emphasizes conclusions of form: individuals with personal experience act in a certain fashion. The next chapter. 2006). however. For campaign practitioners.same group of voters. Holland. in contrast. 1986). experienced voters rely more on their issue position in two of the three cases tested. But the 2000 campaign also provides a warning. The panel nature of the Annenberg data enables detection of the issue-driven candidate evaluation (Lenz. 1926. Experienced 72 . voters experienced with HMOs moderate their cue-taking as the Personal Experience model would predict. along with being potentially unaware of the inner workings of the Social Security trust fund.

and this self-interest may cause issue opinion shifts that would not otherwise occur among experienced voters. The evidence demonstrates that such targeting is a net positive. However. From the point of view of the Personal Experience Model. 73 . campaigns should target experienced voters who agree with the candidate’s stance. a more holistic view that took self-interest into account would be more efficient at garnering voters.voters often have a stake in the policy outcome.

1998) detail the difficulty of changing an individual’s partisan predispositions and imply that campaigns’ persuasive efforts face an uphill battle. The literature on partisan biases has expanded from the fields of political science (Markus and Converse. This analysis moves away from the partisan biases literature and focuses on why citizens might change their vote choice during a campaign. Dozens of 74 . 1979). The Personal Experience Model predicts that campaigns’ emphasis on issues with which voters have personal experience can alter the voters’ evaluations of the candidates counter to the voters’ predispositions. Several micro-theories of American politics (Campbell et al. Two randomized survey experiments with hypothetical candidates and real-world issues support the model’s predictions.. Zaller. political psychology (Taber and Lodge. 1960.. 1992.Chapter 4 Survey Experiments of Princeton Students and Adults Nationwide A central aspect of many political campaigns is persuasion: convincing voters who would otherwise cast a ballot for the opponent to support the campaign’s candidate. and neuroscience (Knutson et al. 2006). 2006). Lupia and McCubbins.

g. increasing a candidate’s standing among the electorate will increase turnout as well as vote share. Indeed. this phenomenon occurred in the 2004 campaign over the issue of whether Saddam Hussein was involved in the September 11 attacks (Parker-Stephen. when an individual receives information (i. Convincing these voters requires that the campaign increase the standing of its candidate in the voters’ eyes relative to the other candidate(s). This ideology can be thought of as a combination of issues. Partisan biases abound. 2000). Thus. veterans) respond strongly to candidate signals on issues with which they have personal experience (e. The goal of shifting the electorate (rather than polarizing it) is more relevant for a campaign that is underperforming in public opinion polls. but they are not helpful for campaigns seeking to persuade voters to cast a ballot for a specific candidate. If Democratic-leaning voters listen only to Democratic elites and Republican-leaning voters listen only to Republican elites. 1997). 2007).1 Survey experiments demonstrate that certain subgroups of voters (e..g. it must convince voters who would not have voted for its candidate to cast a ballot for that candidate.studies find that partisans react more positively to arguments from elites of their own party (see Bartels. and that these interactions influence voters’ candidate evaluations even if they run counter to existing partisan biases. the Iraq war). For a campaign to increase its vote share. political scientists have modeled political preferences as a function of ideology. 2002. 4.1 Theory and Hypotheses Since at least as far back as Downs (1957). but in models that correlate uncertainty of choice with abstention (Sanders. 1 75 . then the electorate polarizes.e. for a review). a “signal”) from a candidate These claims assume an electorate of fixed size. but even leading candidates may want to increase their win percentage to scare off challengers (Levitt and Wolfram... with voters preferring those candidates whose issue opinions are closest to their own opinions.

1998) and cue-taking (Gilens and Murakawa. a candidate’s signals will always contain new information about the candidate’s stance.) This Lemma assumes that voters receive and incorporate signals about politicians’ ideologies into their issue beliefs. Hispanics about immigration. Recasting the Candidate Evaluation Hypothesis from Chapter 2 into a prediction specific to candidate statements yields the first of two main hypotheses of this chapter: For this experiment. An incongruent signal will lessen the respondent’s evaluation. 2002) shows how voters fundamentally alter their issue positions based on the stances of their preferred elites. hunters and guns. after the candidates made public statements. her evaluation should deteriorate when the respondent receives an incongruent cue. Examples include retirees who receive Social Security checks. Bush and Al Gore took public stances during the 2000 campaign (Richard Johnston. 2 76 . and veterans about war policy. the partisans sorted into their respective positions. This “responsiveness” to an issue can either be positive (i.e.. (A congruent signal is defined as information that informs an individual that the candidate supports the same side of the issue that the individual supports. 2004). This political “ground truthing” of candidates’ statements leads voters to critically evaluate politicians on issues with which they have experience. holding the candidate in higher regard) or negative. personal experience on an issue provides a basis on which a voter can judge politicians independent of partisan biases. Lemma 1 : A congruent signal from a candidate will improve the respondent’s evaluation of the candidate. Research on deference to elites (Lupia and McCubbins. Analogously. For example. the individual should improve her evaluation of the candidate. a large proportion of Democrats were proSocial Security privatization (and vice versa for Republicans) before George W. However.2 As elaborated in Chapter 2.that indicates that he holds a similar position to the individual. Zaller (1992) persuasively demonstrates that voters filter out a great proportion of political information based on their partisan predispositions.

For example. If a losing campaign shifts the political discussion to an issue that people know very little about (even if initial public opinion polls show support for one position).Hypothesis 1 (Candidate Evaluation): Individuals with personal experience with an issue are more responsive to a candidate signal on that issue than are those with no personal experience. Given these hypotheses. alter their issue opinion to match their preferred candidate’s opinion (i. Republican retirees did not become ardent supports of Social Security privatization because in their experience the system worked well for them. These hypotheses are extremely important for campaigns that seek to alter the preferences of the electorate. Modifying the Cue-Taking Hypothesis to fit the experiments presented in this chapter yields: Hypothesis 2 (Cue-Taking): Individuals who have personal experience with a political issue. experience may increase opinion certainty 77 .. voters will sort by their existing predispositions.e. Another possibility is that self-interest rather than personal experience causes issue-driven shifts in political judgments. Personal experience also moderates cue-taking from politicians on issues because the voter’s experiences crowd out elite-generated considerations. This survey also measures characteristics of individuals that relate to alternative and mediating hypotheses to those above. Perhaps individuals with generally high political knowledge are likely to have stable opinions on all opinions and experience has little or no effect. when presented with information about their preferred candidate’s stance on the issue. This idea is formalized in the learning model presented in Chapter 2. This responsiveness will be present even when signals run counter to existing partisan biases (or candidate preferences). campaigns would be more successful targeting voters with messages on issues with which they have personal experience. Also. during the 2000 election. cue-take) less than those individuals who lack personal experience. as shown in Chapter 3.

The survey asks respondents to evaluate fictional candidates and real issues. two two-wave survey experiments are conducted: one pilot experiment of Princeton undergraduates (n=273) and one survey of adults nationwide (n=391). confidence that an opinion is correct) and this certainty may drive later behavior (Fazio and Zanna. 4. 78 . biases.e.(i. if no lag between the waves were included. Cross-tab and regression analyses of the change in respondents’ political opinions across waves test the validity of the two hypotheses. 1978). Both experiments are very similar in nature. and (2) so that the candidate can plausibly take either 3 For instance. All interviews are conducted online.1 Survey Design Overall Goals To test these hypotheses. Fictional candidates are used for two primary reasons: (1) to limit the amount of prior information. respondents could purposely report the same issue positive in both waves to prove that they hold a consistent opinion regardless of the cues they receive. The Candidate Evaluation Hypothesis requires that respondents judge politicians and the Cue-Taking Hypothesis requires that respondents report issue positions. The first wave provides a baseline for respondents’ opinions absent the key parts of the candidates’ platforms.2. The month lag between waves limits the respondent’s ability to remember her wave one responses and lowers the possibility of the respondent attempting to game the survey.3 The two hypotheses tested have different dependent variables of interest. with respondents of the first wave being re-interviewed about a month later. The surveys are conducted in two waves to better capture the effects of candidates signaling on respondents’ issue opinion and candidate evaluation. this time after the candidates reveal their issue positions. Over a month later. or predispositions the respondent has about the candidate.2 4. respondents are re-asked the same questions..

Third. Because the candidates are hypothetical and the issues are real. the candidate paragraphs attempt to mimic the partisan division of the real political arena. First. Both surveys test two hypothetical candidates and three real issues. and the interaction of personal experience with these effects. In the second wave.1 for candidate descriptions.and 21-point scales respectively to detect small movements.side of issues. This setup allows the effect of candidate position-taking on both candidate evaluation and issue opinion. they are attempts to reflect the actual debate and are designed to be forceful enough to move respondents. change in candidate evaluation is more likely than change in issue opinion. (2) respondents report issue opinions. actual issues are used. (See Appendix C. The basic structure of the surveys is: (1) respondents rate hypothetical candidates based on paragraph descriptions. the surveys are created with the goal of fostering cue-taking. Thus. to be tested. “[Candidate name] recently made the following comments about the debate over [Issue]: [Issue statement]. every respondent sees the same hypothetical match-up between one Democrat and one Republican.1). but at the end of the paragraph sees the statement. they are more likely to cue-take. although the candidate signals include no new information on the issue. Second. Thus. for example. and (5) respondents rate the candidates and report issue positions again. Since the survey tests the interaction of personal experience with political judgments. the evaluations and issue opinions are reported on 101. the respondent views the entire candidate paragraph description again.) If Democrats feel an affinity to the hypothetical candidate. (3) respondents are queried on their issue experience.” Each issue 79 .1. The web format facilitates this granularity by allowing respondents to choose their response via a slider (Figure 4. (4) a month later the same candidate paragraphs are presented with the candidates taking stances on different issues.

2. Half the time the two candidates send two liberal or two conservative signals (although again on different issues). and the third issue is left unmentioned. The web browser informs the respondent of the shuttle current position (which. each respondent uses the vertical slider to indicate her certainty of her reported belief. for every respondent. though with more granularity. Candidates are assigned issues and issue statements randomly with one constraint: candidates cannot take two stances on the same issue.2 Specifics of Princeton Survey Design The first wave of the Princeton survey was distributed to 1. A $100 gift certificate to either amazon. This constraint allows for the source of any change in a voter’s opinion to be clearly identified. is B:2. in the example above. Respondents report their issue opinion by sliding a shuttle left or right across the screen.600 students on December 6.1: Screenshot of survey question. 4. or 2 out of a maximum 10 on the B-side of the issue). with the winner to be chosen at random from students who completed the 80 .Figure 4. has two issue statements related to it. the Republican candidate sends a signal on a different issue. one liberal and one conservative. For both issue opinion and candidate evaluation. Candidates are evaluated on a similar scale. 2007. the Democratic candidate takes a stance on one issue.com or iTunes was offered as an incentive to participate. Thus.

To evaluate experience with gay marriage. and foreign aid. Respondents were asked for their opinions on three issues: gay marriage. The survey was open for four weeks. respondents were asked to indicate whether they favored doubling or halving the foreign-aid budget. Republican) in a race for a U. Basic biographical information and endorsements were mentioned in the candidate paragraphs.” The candidate evaluation pitted a former governor (Richard Miller. With regard to the environment. during which 418 undergraduates completed the survey (26% response rate). The respondent was asked to report which one of these options she preferred and how strongly. Students who completed the survey are referred to are “respondents. but differed by class year and remained constant across both waves.S. students were queried about the number of outdoor excursions they went on in the past year. with a second incentive (a $50 gift certificate) for the 400th respondent. Senate seat of an unspecified state. The gay marriage question informed respondents that their state supreme court had ruled that the state must offer same-sex couples either civil unions or marriage. The order of candidates and issues was not random. No statistically significant order effects were found. For issues. incentive-based solution. To measure issue experience. With respect to foreign aid. Democrat) against a Retired Admiral (Mark Jones. respondents were asked about their personal habits. environmental regulation. respondents indicated how much 81 . To gauge experience in foreign policy (specifically foreign aid). respondents reported how often they communicated with any gay and lesbian friends and family. The question about environmental regulation asked respondents to choose between two solutions to global warming: a cap-and-trade system or a less restricted. the relative position of the traditionally liberal side and conservative side was constant across the three issues and waves. A reminder about the survey was sent out a week after the initial email.survey.

age. The second-wave survey was delivered to students’ inboxes the following evening and again was open for a month. A reminder was sent out two days later.S. Gift certificate incentives were provided. 4 82 . 540 completed the survey within a week of the survey’s launch. reducing the possibility that this judgement affected the results. On February 11. This eliminated respondents who had technical difficulties.2. but also touching or shifting at least one evaluation slider and at least one confidence slider. 4 At the end of the survey. 283 completed at least some questions in the second wave (273 finished the entire survey). Finally. the respondents were told that a second survey would be coming “next month. respondents were asked standard demographic questions about their gender. 2008.” The candidate evaluation pitted a former governor (Richard Miller. Of the recipients. an email alerted first wave respondents that the second-wave survey would be available within a day. This email also provided (with permission) the names of the gift certificate winners of the first survey. 2008. and ideology. Another 10 respondents were eliminated from the analysis because their responses and the amount of time it took them to record their responses indicated that they had not read the questions. for a 3% response rate. In addition to biographical information and endorsements. as well as how closely they follow politics.time they had spent abroad (excluding Europe) in the past five years. party affiliation. A standard incentive (points in Survey Sampling Inc. each Completion of the survey required not only going through all the questions. Inc. as with the first survey.’s email list on July 28. two months after the initial wave.’s program) helped motivate individuals to take the survey. Democrat) against the current attorney general (Mark Jones. Republican) in a race for a U.812 people on Survey Sampling. Of the 418 first-wave respondents. 4.3 Specifics of Nationwide Survey The first wave of the nationwide survey was distributed to 18. Senate seat of an unspecified state. No wave two respondents were eliminated.

and gay marriage. respondents were asked to guess what each candidate’s opinion would be on one issue (the same issue the candidate would take a stance on in wave two). Also. respondents gauged what percentage of their important conversations were with: individuals currently serving in the military. Inc. On the Iraq war. provided the respondents’ race and ethnicity (i. the respondents were presented with the situation in which a state supreme court recently mandated legalizing same-sex marriage. First. For immigration. the Iraq war. respondents chose between keeping troops in Iraq until that country is a “stable. respondents were asked for their opinions on three issues: immigration.. With respect to experience with immigration. Survey Sampling. which was 83 . After completing the candidate evaluations.description ended with a reason why opposing party leaders disliked the candidate in question. former military personnel. whether the respondent was of Hispanic descent) and the respondent’s zip code. and gays or lesbians. Several measures of issue experience were recorded.1). Hispanics.” After reporting how many people were in this group of close friends and family. immigrants. respondents were asked to report the certainty of their response on a 0 to 10 scale (Figure 4. First. the choices were between deporting illegal immigrants and providing a path to citizenship.. those people they talk to when “important issues or problems come up in life. Other measures of issue experience were tested as well. respondents were asked to think about their social network. specifically. for each evaluation and opinion. after the candidate descriptions. The respondents then indicated whether they supported passing a constitutional amendment to overturn the decision or to recognize gay marriage. The nationwide survey included a few additional features.e. functioning democracy” and withdrawing troops “as quickly as the Defense Department believes is safe.” For the issue of gay marriage.

14 respondents specifically praised the range of options provided for the issues. Finally. the respondent was thanked for her time and told that a follow-up survey would be delivered to her inbox in about three weeks. respondents were asked to evaluate and comment on the survey. To determine whether the respondents represent potential voters. For additional measures on experience with the Iraq war issue. Finally. respondents were asked if they are currently serving (or had served) in the military or whether this applied to another person in their household. exactly three weeks after the initial wave. 2008. These numbers translate into a response rate of 74% on the second survey. Also. an email alerted firstwave respondents to the availability of the new survey. The second wave closed just as Obama announced the selection of his running mate. Of the 540 first-wave respondents. one to the Iraq war. and political attentiveness. and two to the gay marriage option for the gay marriage question. respondents were asked standard demographic questions about their party. Specifically. 84 . respondents were asked whether they considered themselves gay or lesbian (at the end of wave two).matched with the percentage of Hispanics in the respondent’s zip code according to the U. On wave two. respondents were asked what proportion of their co-workers were Hispanic. two to immigration specifically. they were asked whether they voted in 2006 and their intentions to vote in 2008. At the end of the survey. The biggest news during the first wave of the survey was the residual effect of Barack Obama’s Europe trip and the launching of John McCain’s “celebrity” advertisement (July 30).” Of the 391 respondents who reached that question. respondents were asked if they felt “forced” to choose an option they “didn’t particularly like. For gay marriage. census. On August 18. The real-world political scene during the period between the surveys was relatively quiet.S. 397 completed at least some questions in the second wave (391 finished the entire survey). four objected to the issue questions in general. ideology.

As a reference point.5 Certainly very good social science can be performed in a laboratory with a participant group that looks very different from the population at large. The population is 54% female and a plurality are first-year students. treatment effects observed in one segment of the population might be counterbalanced by opposite effects in other segments of the electorate that are not included in the experiment. individual data for the 2008 exit poll is not available. and miThe electorate and not the adult population is used as a reference point because the theory’s effects on nonvoters are inconsequential. For instance. someone might want to register or decide to vote only after receiving positive information about a candidate (Sanders. especially considering the nature of an Internet survey. For the national survey. the population is not restricted to registered voters or likely voters because individuals’ decisions to register to vote or to cast a ballot are ongoing decisions. 4.3. the demographic composition of the survey respondents is compared to the composition of the National Election Pool’s 2004 general election exit poll. poorer. One might expect older. Only 7% indicate that they follow politics “a great deal.Joe Biden. 6 As of June 2009. one goal of conducting a nationwide survey is to produce a respondent population that is similar to the national electorate. 2000). However.1). the demographics of the two groups are strikingly similar (Table 4. 5 85 .3 4. Applications of this theory would be focused on likely voters and they decide which candidate wins or loses.1 Design Checks Characteristics of the Survey Respondent Population The characteristics of the Princeton survey’s population are constrained by the fact that the respondents are all undergraduates at Princeton University.” with a plurality placing themselves in the middle of the scale (“some”). No major events occurred that were related to any of the three issues tested in the survey. That said.6 Overall.

although Democrats rate the Democratic candidate more positively than the Republican candidate (and vice versa for the Republicans). On a scale from -10 (worst) to 10 86 . The nationwide survey fixes some deficiencies of the Princeton survey to achieve these goals. is a professional organization that produces nationally representative surveys. although the survey experiment includes more self-reported Independents than Pew reports.2 Features of the Survey Candidate Descriptions. although this difference does not reach statistical significance (p = 0. while 15% of the electorate report the same. This result may be due to differences in question wording (all options. meeting or exceeding their percentages in the electorate.3. Independents fall in between the partisans.” are immediately presented to the respondent in the survey experiment. as those groups traditionally use the Internet less. rating the Democratic nominee more highly. these demographics are well-represented in the survey. However.15). the paragraph descriptions of the candidates are designed to elicit both positive evaluations and partisan differences. The overall similarity between the two populations is a benefit of the fact that Survey Sampling. unlike in Pew). For the Princeton survey. Because Democrats have gained significantly in party identification since November 2004.. including “lean Democrat/Republican. To foster cue-taking in the second wave of the survey. 4. Inc.nority individuals to be underrepresented. The two surveys report approximately the same proportion of Democrats and Republicans. The only troubling statistic is the overrepresentation of non-churchgoers in the survey experiment: 32% of respondents report that they never attend church. respondents of all political affiliations are likely to rate both candidates positively. Pew Research Center’s August 2008 poll is used to compare partisanship.

Grad H. or bisexual? Yes 6% 2% No 92% 45% Refused 2% 54% Income under $30K $30K-50K $50K-75K $75K-100K $100K-150K $150K+ Refused Region East Midwest South West -4% 31% 25% 44% 37% 28% 35% -6% -3% 9% 4% 47% -52% 2% 18% 38% 27% 15% 4% 22% 32% 26% 16% -2% -4% 6% 1% -1% 24% 22% 20% 10% 7% 3% 15% 20% 18% 21% 12% 10% 7% 11% 4% 4% -1% -2% -3% -4% 4% 33% 22% 1% 4% 0% 40% 0% 37% 19% 1% 2% 0% 11% 29% -4% 3% 0% 2% 0% 29% -29% 22% 20% 39% 20% 22% 26% 32% 20% 0% -6% 7% 0% *Internet survey allowed more than one race to be specified †From Pew (August 2008) instead of exit poll ‡Includes other Christians Table 4.Attribute Gender Male Female Race* White Afr-Amer Latino Asian Other Refused Age 18-29 30-44 45-59 60+ Party ID† Democrat Republican Ind/Oth Education < H. -3% 3% Attribute 2008 Internet Survey 2004 Electorate: Exit Poll 16% 26% 14% 28% 15% Diff. lesbian. the national survey is not weighted. -3% -7% -5% -2% 17% 79% 14% 13% 4% 10% 4% 77% 11% 8% 2% 2% 0% 2% 3% 5% 2% 8% 4% Church Attendance > Weekly 13% Weekly 19% ≤ Monthly 9% A few times a year 26% Never 32% Whites: Are you born again/evangelical? Yes 29% 23% 6% Have you ever served in the military? Yes 16% 18% -2% 16% 25% 31% 28% 17% 29% 30% 24% -1% -4% 1% 4% Are you currently married? Yes 59% 63% Are you gay.S. The exit poll results are weighted according to the documentation. 87 . Grad Some College Coll Grad Post Coll Religion Protestant‡ Catholic Mormon/LDS Jewish Muslim Other/None Refused 2008 Internet Survey 43% 57% 2004 Electorate: Exit Poll 46% 54% Diff.1: Demographics of survey experiment are similar to those of the November 2004 electorate. Because of the similarities.S.

partisan feelings trump any propensity to rate hypothetical candidates positively or negatively in general.(best). In both surveys.01). and gay marriage. 21% want to decrease it. Within respondent. meaning that the descriptions did not elicit strong partisan views. the correlation between the Democratic and Republican evaluations is negative and statistically significant (ρ = −0. controlling for the partisanship of the respondent. respectively. the Iraq war.2 and for the Republican candidate is 3.7. A smaller plurality supports a cap-and-trade system to regulate greenhouses gases (49%) over encouraging businesses to be green while letting the market rule (42%. On a +10 (most liberal) to -10 (most conservative) scale. This correlation indicates that partisan feelings are often trumped by the amount of “trust” (or some other valence characteristic) that respondents afford to hypothetical candidates for which they have only a brief description. the mean responses are +1.2). Students support gay marriage over civil unions 55% to 36% (with 10% neutral). and 20% are neutral. 59% of respondents want to increase the budget for foreign aid.2 points above the Republican (a statistically insignificant difference). Similar splits are seen in the nationwide survey. Across all respondents. the candidate profiles are equally positive. 8% neutral. 1% refused). The correlation between the two ratings is positive (ρ = 0. In the Princeton survey.3 (p < 0. the average rating for the Democratic candidate is 5. Thus. Issue Opinion. +1. the highest proportion of respondents 88 . When regressed on 7-level partisanship. Across all issues.4. meaning that in rating the candidates. no side of any issue garners more than 60% support.0 for immigration.14). the relative evaluation of the Democrat at the intercept is only 0.16). opinions are relatively split across all three issues. the hypothetical candidates are rated slightly positively. and -1. On the issue of foreign aid. with means of 58 and 56 for the Democrat and Republican respectively (Figure 4.

While the signals generally work as expected. Thick lines show plus or minus one standard error. these low expectations are met in both surveys. the persuasion is not directly aimed at the respondent. half of the signals emanate from disliked candidates. Hypothetical candidates’ attempts to persuade voters through a static text cannot be expected to move voters’ issue opinions very far. on one side of the issue is 62% favoring a path to citizenship for immigrants.2: Mean evaluations of hypothetical candidates by party for nationwide survey. Despite the entire survey population moving 89 . the candidate only states his own issue opinion. the Iraq war polarizes partisans the most (Figure 4. thin lines are plus or minus two standard errors. All three issues correlate as expected with party.3). In the Princeton survey. Second. N-size is 397 for all respondents. 66 for independents and 90 for Republicans. candidate signals prove somewhat effective in two of the three issues: gay marriage and the environment.Mean Evaluation of Hypothetical Candidates By Respondents' Party Affiliations Mean Candidate Evaluation (0 to 100 scale) 20 40 60 80 100 All Respondents Democratic Respondents Independent Respondents Republican Respondents 0 Democratic Candidate Republican Candidate Figure 4. Candidate Signals. 258 for Democrats. First. Two other factors lower expectations even further.

more conservative on gay marriage and more liberal on the environment. 258 for Democrats.Mean Issue Opinion By Respondents' Party Affiliations Mean Issue Opinion Liberalness (−10 to 10 scale) −10 −5 0 5 10 All Respondents Democratic Respondents Independent Respondents Republican Respondents Immigration Iraq War Gay Marriage Figure 4. the liberal signal on gay marriage Foreign aid is consistently an ineffective issue in the Princeton survey. with the electorate not shifting opinion between waves and the signals moving opinion by about a point. the most any signal moves voters is 1 point (on average) on a 21-point scale (Figure 4.4). 66 for independents and 90 for Republicans. which lead to its omission in the nationwide survey. but the liberal signal also pushes respondents in the conservative direction. The survey population shifts to the left on the Iraq war between waves. 7 90 .5 points (on average) on a 21-point scale in the direction of the signal (beyond any overall survey population movement). Immigration is an ideal case.3: Mean issue opinions by party for the nationwide survey. N-size is 397 for all respondents. A similar magnitude is seen for the effect of the conservative foreign aid signal. Thick lines show plus or minus one standard error. but the conservative signal dampens this migration.7 For the nationwide survey. thin lines are plus or minus two standard errors. candidate signals move respondents about 0. Across the three issues.

4. Although the relative effect of the liberal and conservative signals are in the expected direction in nearly all cases.g.g. 91 . the liberal gay marriage signal proves effective when the signal is delivered by a preferred candidate. The nationwide survey attempts to gauge issue experience more accurately via either life experience (e.3. none of the signals are particularly powerful.is the only signal that does not shift opinion in the expected direction when compared to the group that does not receive a signal.4: Mean shift in the liberal direction of issue opinions from wave one to wave two in the nationwide survey. having close gay friends or family and same-sex marriage). the surveys attempt to elicit the amount of interaction between the respondent and a political issue.3 Definitions of Issue Experience Although the definition of issue experience is somewhat arbitrary.. These survey measures can certainly be called into question. However. Issue Opinion Shift By Signal Direction 3 All Respondents Liberal Signal Conservative Signal No signal −3 Shift in Issue Position (positive=liberal) −2 −1 0 1 2 Immigration Iraq War Gay Marriage Figure 4. veteran and Iraq war) or contact with people directly affected by the issue (e.. especially for the Princeton survey.

amount of time spent abroad (outside North America and Europe) in the past five years is the measure for experience with foreign aid. with an emphasis on conversations with individuals affected by the issue in question.2: Factor loadings for immigration and Iraq war experience measures. the top 30% of respondents (approximately) in each of the three experience measures are labeled “experienced.3 for question wording.3. of convs.52 0. respondents are asked how often they communicate with gay and lesbian friends and family.69 0. of convs. number of outdoor excursions in the past year is the chosen proxy for experience with the environment. with military Perc. (Factor loadings are described in Table 4. Potentially worse. when issue experience is dichotomized for easier display in figures or hypothesis testing.50 0. The Princeton survey has admittedly poor measures of issue experience.5 Perc. of convs. (See Appendix C. The gay marriage measure is probably a more accurate depiction of experience. of convs. with Hispanics Respondent is Hispanic Perc.55 . For example. The nationwide survey contains several measures of experience. the disparate potential sources of issue experience are amalgamated into one measure using factor analysis.5 . with immigrants Perc. Hispanic in zip code Perc.3.38 Variable Iraq War Loading .1. with veterans Someone in household serves Table 4.) For measures in both surveys. For the issues of immigration and the Iraq war.” 92 . Hispanic in workplace Loading 0.Immigration Variable Perc.88 .) These respondent characteristics are combined into one measure of experience for each issue.

1.5 points when he delivers an incongruent signal (20point scale). as the Candidate Evaluation Hypothesis predicts.8 The opposite happens when the candidate announces an issue stance far from a respondent’s own view.18. 93 .6 and -2.5. For the Republican candidate’s evaluation.4. it is operationalized as the distance between the respondent’s wave one opinion and either +10 (for liberal signals) or -10 (for conservative signals).4 4. The signals work well. and lower by 2.6 points (21-point candidate scale). congruous signals are only those on the same side of the issue as the respondent’s wave-one opinion. all reported p-vales are two-tailed). The large (and statistically significant) difference between the effects of these two types of signals for the nationwide survey is displayed in Figure 4. For individuals experienced on the issue on which the Democratic candidate delivered his signal. this change in evaluation is higher: 3. When the concept is measured as a continuous variable. these shifts indicate how the voters might take a second look at a previously non-preferred candidate in an election. the average difference in the change in Democratic candidate evaluation is 2.1 Support for Hypotheses Candidate Evaluation The changes in candidate evaluations are the crucial piece of this analysis. a similar effect is apparent: a change of 4. Similar effects are present in the Princeton survey. These two situations are described as the candidate delivering a “congruent” or an “incongruent” signal. in the Princeton survey. The analogous effects for the Republican candidate and 1.1 points higher than the wave one evaluation when he delivers a congruent signal.75 points (p = 0. 9 The average re-evaluation of the Democratic candidate is 0. The data can be analyzed two ways: by candidate (combining issue signals) and by issue (combining candidate evaluations).4.1 points for 8 When signal congruity appears in a dichotomous context. that respondent evaluates the candidate more highly. producing large shifts predicted by Lemma 1: when the candidate announces an issue stance close to a respondent’s position. These differences are highly statistically significant. As an example of the former.9 Voters with personal experience on related issues exhibit larger swings in their candidate re-evaluations as the Candidate Evaluation Hypothesis predicts.

For instance. thin lines are plus or minus two standard errors.1 points (21-point candidate scale). this margin jumps to 7. however.Change in Candidate Eval. These results are noisy. range −100 to +100) −20 −10 0 10 20 Candidate Evaluation Shift By Signal Congruency and Candidate Party All respondents Received congruent signal Received inconguent signal Democratic Candidate Republican Candidate Figure 4.9 points This figure includes the change in Republican margin when the Republican delivers a congruent signal on gay marriage or the Democrat delivers an incongruous signal. issue-specific effects are quantified. but supportive of the hypothesis. with 200 receiving congruent cues from the Democrat (197 incongruent) and 214 receiving congruent cues from the Republican (183 incongruent). (poss. Thick lines show plus or minus one standard error.10 For those with experience on the issue.8 points for non-experienced individuals (p = 0.g. the average difference in Democratic margin from wave one to wave two when the Democrat delivers a congruent signal on gay marriage or the Republican delivers an incongruous signal is 5. 10 94 .32). N-sizes are 397 overall.5: Mean difference between candidate evaluations in wave one and wave two.. experienced individuals and 2. When the evaluations of the Democratic and Republican candidates are combined into one evaluation margin (e. Respondents who are neutral on an issue are mixed in with respondents who received a congruent cue. Democratic evaluation minus Republican evaluation) for each issue.

with the margin difference increasing from 3. alternative hypotheses. for each issue.6 shows. The noise in these supportive results is from the issues of environment and foreign aid.8 points (p = 0. These results validate the main hypothesis.1). while nonexperienced respondents alter their margin by 1. experienced voters adjust their evaluations of candidates in the direction of the signal (i. though slightly less noisy. upward for congruent signals and downward for incongruent signals) further than non-experienced respondents. and no controls yield similar.1 for respondents without experience to 3.06). As Figure 4. Assuming the tests are independent (implied by the randomized treatment assignment) the relationship is significant at the 10% level (two-tailed) across all (but not within each) issues. for the environmental issue.9 points in the direction of the congruity of the signal.9 for respondents with experience (p = 0. The data indicate that political attentive respondents hinge their candidate evaluations on the candidates’ signals. Second. Regressions (one for each issue) with one interaction term (congruency crossed with experience). results (Table C. respondents who are reacting to traditional partisan biases are weeded out so that potential vote switchers are isolated.(p = 0. Foreign aid exhibits the characteristic of a weak experience variable.. First. are considered. such as political attentiveness and self-interest. Respondents who pay attention to politics rely on candidate issue positions when judging these candidates to a greater degree than their less-attentive counterparts. Analogously.82).1) demonstrates that political attention strongly interacts with candidate signals. but two additional analyses are crucial to bolstering the claim.e. 95 . The nationwide experiment produces similarly supportive but noisy results. the requisite main effects (in their continuous form).62). experienced respondents alter their candidate margin by 2. Regression analysis (Table C.

gay and lesbian respondents are more sensitive to the candidate’s stances on extending marriage rights. bi-sexual.Candidate Evaluation Shift By Issue and Respondent Experience Change in Candidate Eval. All the signs are in the correct direction. The reverse is true for the Iraq war. 11 96 .S. The Candidate Evaluation Hypothesis holds even when controlling for both of these effects. thin lines are plus or minus two standard errors. thus. A similar conclusion cannot be reached for self-interest as the results are mixed. no self-interest measure is available for immigration. citizens. coded so that the direction of the shift matches the direction of the signal. or transgender. Experienced respondents are more responsive to the candidate signals than non-experienced respondents. For the other two issues: 20 respondents currently serve in the miliary and 24 respondents identify as gay.11 Self-interest for gay marriage works as expected. current members of the military are less responsive to the candidate’s position on this issue. lesbian. and jointly the results are statistically Answers to the vote questions indicate that all respondents are U. Thick lines show plus or minus one standard error. in the Direction of Signal 0 5 10 15 20 25 30 All Respondents No Issue Experience Issue Experience Immigration Iraq War Gay Marriage Figure 4. Two of the three issues have measures of self-interest: the Iraq war (active military members) and gay marriage (gay respondents). National survey.6: Mean difference between candidate evaluations in wave one and wave two.

As the Iraq war is the most polarizing issue. These two measures are highly correlated (ρ = 0. 97 . Regression analyses of the two candidate evaluations. when the candidate signals counter partisan biases. one-tailed). finds positive and significant effects in the Princeton survey (joint 12 Causal mediation analysis would be a more proper method of assessing the causal mechanism.06. might even rate the Republican higher. Self-interest soaks up much of the effect of personal experience for the issue gay marriage.12 The effects of personal experience are apparent. The magnitudes of the effects of personal experience do not decrease when this potential mediator is added the regression indicating that these effects are independent of each other. Respondents with more certainty about their opinion on the issue are more sensitive to the candidate’s position on this issue. limited to antipartisan bias signals. though weaker. She now lowers her opinion of the Democratic candidate.significant at conventional levels (p = 0. or (2) it is an incongruent signal delivered by a candidate that the respondent did favor in the first wave. A signal is said to go against partisan biases if: (1) it is a congruent signal delivered by a candidate that the respondent did not favor in the first wave. the Democratic respondent mostly likely assumed that the Democratic candidate supported ending the war.46) so teasing out the independent effects of self-interest and personal experience is more difficult. Issue certainty also affects political judgements in the expected direction. the effect of personal experience generally remains. An example would be a Democratic respondent who follows her party affiliation in wave one and rates the Democratic candidate more highly but receives a signal in wave two that the candidate opposes ending the Iraq war. depending on the signal sent by the Republican candidate. and. but does not appear to be a mediator for political experience. When restricting the analyses to these types of signals.

The Cue-Taking Hypothesis predicts that in each case.13 The national survey analyses control for political attentiveness and self-interest and find effects consistent with the Candidate Evaluation Hypothesis in two of three issues (joint p = 0. In both surveys. 13 98 . but especially when a favored candidate delivers a signal. The largest amount of cue-taking is among non-experienced respondents on the gay marriage question.2 Issue Opinions and Experience The Cue-Taking Hypothesis—that issue experience moderates cue-taking—is supported by the evidence. 4. but sending congruent signals about an issue with which the voter has experience helps grease the wheels. With regard to the Iraq war.4.3. who counterargued against the signal Only the two main effects.e.. the average shift is always toward the candidate’s position except for experience respondents on the environment issue. a two standard deviation change) increases the relative candidate evaluations (100-point scales) by 6 points when that candidate takes a congruent rather than incongruent stance on the war.p = 0.) Shifting an individual’s views against their partisan predispositions is difficult. congruent signal and experience. The signals in the Princeton survey all produced (on average) very small movements in issue opinion. (Details for each group of interest are listed in Table C. these respondents shifted less than half a point (on a 21-point issue scale) in the direction of a favored candidate’s signal. The data are broken down by issue and whether the signal is received from the more favored candidate (as defined by the evaluations in wave one). When signals are delivered by favored candidates. one-tailed).15. individuals without personal experience on the issue will shift their opinion in the direction of the candidate signal more than those with experience.04. one-tailed). this relationship is observed in five of the six cases. yielding six tests. and the interaction term are included in these regressions. a change from below-average experience to above-average experience (i.

In five of six cases. perhaps because more partisan feelings were aroused by the candidate When experience is measured dichotomously and favored/unfavored signals are combined. Cue from unfavord.5 0. no exp. 14 99 . Cue from favored. in the direction of the received signal.8.(Figure 4. Cue Taking By Issue−Experience Pair 1. Only signals from unfavored candidates on the issue of foreign aid generated an unexpected relationship between experience and cue-taking. exp. exp. of issue opinions from wave one to wave two in the Princeton survey. the immigration signal from a favored candidate yields a stark contrast between individuals with experience (who counterargue the signal) and those without experience (who shift toward the candidate’s position by over one point on the 21-point issue scale). only a signal from a favored candidate on the Iraq war produces an unexpected relationship.14 More counterarguing of signals from unfavored candidates is apparent in the nationwide survey than in the Princeton survey. In the nationwide survey.0 0. Cue from unfavored.7). Change in Issue Opinion in Direction of Cue −0.5 Gay Marriage (Friends) Environment (Excursions) Foreign Aid (Trips Abroad) Figure 4. no exp. that reverse relationship is far from statistically significant. non-experienced respondents shift their opinions in the direction of the signal further than experienced respondents.0 All Cue from favored. t-tests are significant at the 5% level for immigration and gay marriage.7: Mean shift. All six relationships are displayed in Figure 4.

3 breaks down these effects by specific experience measures. 100 . From favored/exp. there is much less of a relationship between cue-taking and attentiveness. From unfavored/no exp.descriptions (which include more partisan language). From unfavored/exp. of issue opinions from wave one to wave two in the nationwide survey. The phenomenon of attentiveness mediating cue-taking is not statistically significant.2). In contrast to the clear link between cue-taking and experience.9). the non-experienced respondents demonstrated cue-taking more than the experienced respondents.8: Mean shift. No consistent effect is found for self-interest either (Table C. in the direction of the received signal. Issue Opinion Shift in the Direction Of Signal By Signal Source and Respondent Experience Shift toward signal (poss. Overall. −10 Immigration Iraq War Gay Marriage Figure 4. Table C. Certainty about one’s opinion has no effect on change in opinion demonstrating that individuals are poor judges of themselves. and for the issue of the Iraq war the relationship runs counter to the hypothesized relationship (see dichotomized results in Figure 4. range −20 to +20) −5 0 5 10 All who rec'd signal From favored/no exp. the Cue-Taking Hypothesis is robust to alternative explanations. In five of six cases.

Even taking political attentiveness and self-interest into account. of issue opinions from wave one to wave two. experience with political issues leads to less cue-taking from politicians and to more issue-driven evaluation of those politicians. the effects in the national survey occur for minor 101 . 4. in the direction of the received signal. the average increase in the magnitude of candidate evaluation shift due to experience is large enough to switch the preferences of an additional 3% of respondents. In many close races. In both the Princeton and nationwide surveys. Further.9: Mean shift. Campaigns should especially take note of the magnitude of the Candidate Evaluation Hypothesis’s effects.Issue Opinion Shift in the Direction Of Signal By Signal Source and Respondent Political Attentiveness 8 Shift toward signal (poss. range −20 to +20) −2 0 2 4 6 All who Rec'd Signal From favored/Inattentive From favored/Attentive From unfavored/Inattentive From unfavored/Attentive −4 Immigration Iraq War Gay Marriage Figure 4. Attentive respondents are defined as those who read a presidential campaign story at least daily (n=193). 3% is the difference between winning and losing.5 Discussion The results of these two survey experiments generally confirm the Personal Experience Model: personal experience with issues plays a large role in the formation of political judgments.

issues. the economy is cited as the most important issue. with Zaller (1992) claiming that in a one-issue environment all voters take cues (even from out-party politicians) and Lupia (1994) finding that voters can glean negative information from endorsements by ideologues on the other end of the spectrum. Much of this noise is probably due to the use of hypothetical candidates as proxies for real-world politicians and low sample size. The evidence here lends some support to Zaller. This difference runs counter to an intuitive prior belief that because it should be easier to move individuals’ opinions of hypothetical candidates. while many of the results do not independently reach statistical significance.05 about two-thirds of the time. The theory on this topic is mixed. 102 . clearly there are two active sides to the referendum (otherwise one side would not be spending money to put out its message). by far. candidate evaluations shifted more on average than issue opinions). thereby eliminating the possibility of two-sided debate. While the premise of that belief is correct (both surveys showed that as a percentage of the scale. each issue receives at most one signal. the implication does not follow. Michael and Dalia Sussman. in contemporaneous surveys.1) is that self-reported certainty may play a role similar to personal experience when judging candidates. Furthermore. the cue-taking results are less noisy and less susceptible (at least in the nationwide experiment) to alternative hypotheses than the candidate evaluation results.16 Interestingly. A secondary finding is that non-experienced respondents did not counter-argue signals from unfavored politicians. In the case of the Princeton and nationwide survey experiments. 16 Power analysis of the Princeton results indicates that a sample size of 1. the differences in candidate evaluation among subgroups should be easier to identify. “Voters in Poll Want Priority to Be Economy. nearly all point in the direction predicted by the hypotheses. 17 Another more tentative finding (Table C. 17 Lupia’s finding comes from a ballot initiative where even if only one side of the argument is heard. The data do not indicate 15 Cooper.” The New York Times. but perhaps the one-issue environment tilts the playing field unfairly in that direction. Their Top Issue.000 would generate two-tailed p-values of at most 0.15 Many of the individual results are not statistically significant at conventional levels. Also see August 18-24 Diageo/The Hotline poll. however.

voters may evaluate candidates on issues about which they hold confident opinions independent of personal experience. instead. However.that certainty is a mediator of the Personal Experience Model’s causal mechanisms. 103 . certainty about an individual’s own opinion does not temper that individual’s propensity to change her opinion later.

percent poverty in census block) data.. magazine subscriptions). they must be able to identify and target individuals with personal experience.Chapter 5 Randomized Field Experiments: Optimizing Campaign Strategy and Evidence for the Personal Experience Model1 5. While campaigns have more granular information at their fingertips.g. The widening scope of campaign databases..g. which now include not just basic information such as age and gender. 104 . they may 1 Large sections of this chapter are adapted from Imai and Strauss (2009). governmental (e. but also commercial (e.1 Introduction For campaigns or political organizations to apply the Personal Experience Model. is both a help and a hindrance to campaigns. fishing licenses).. and census (e.g.

First.. The Personal Experience Model predicts one type of heterogeneity: Voters with experience on an issue will be more responsive to campaign appeals on that issue. 2003. The findings based on such an approach may not be of much use for campaign planners for two reasons. Building on work from more than a half century ago (e. Over the last decade. The empirical findings of these studies have the potential to significantly affect the practice of get-out-the-vote (GOTV) efforts and persuasion campaigns in the real world (Green and Gerber. Gerber and Green.. 1927. In addition. this chapter examines two cases in which the results of field experiments are consistent with the Personal Experience Model. Nickerson.g. Gosnell. Nickerson. Eldersveld. among many others)..g.not be equipped to sift through the mounds of data. Second. 2008).. a planner must consider the problem of treatment effect heterogeneity where each of the available campaign tactics may mobilize different voters to a varying degree. political scientists have shown renewed interest in the use of randomized field experiments to study voter turnout and persuasion (e.g. 2008). 2000. The 105 . many researchers implicitly assume the constant additive treatment effect across individual voters when using regression models for statistical analysis. 2007. 1936. Hartmann. This leads to the standard practice of reporting a single estimate (or at most a small number of estimates) summarizing the efficacy of each mobilization method. This chapter develops a methodological process for identifying voters who are responsive to campaign appeals by utilizing and discriminating among information gathered from randomized experiments. these researchers have developed creative methods for conducting field experiments with a large number of voters in real electoral environments as a way to test various theories of voter turnout (e. Political science research that reports the results of field experiments tends to focus on the statistical significance of the estimated overall average treatment effect (ATE) of each campaign appeal. Gerber et al. 1956).

First. an anti-McCain 106 . This means that the standard practice of academic researchers reporting only the estimated overall ATE may not provide useful information from a practitioner’s perspective. Second. the common assumption of constant additive treatment effect is too restrictive and does not serve the purpose of campaign planners. to use the results of field experiments effectively when planning their mobilization campaigns. a mailing campaign about education attempting to persuade individuals to vote for a Kentucky gubernatorial candidate had the most effect on parents. This procedure is applied in three cases. And finally. A planner starts with a prior belief about the effectiveness of each mobilization technique under consideration for voters with different characteristics and updates this belief based on the available experimental data. Similarly. 2009) who were between the ages of 20 and 24 were most responsive to the GOTV appeal. in the first of two examples that relate to the Personal Experience Hypothesis. This gap between academic research and policy making can be addressed through an essential and yet missing methodological tool. With this framework. which may vary across different voters. both partisan and nonpartisan. campaigns can use the data from randomized field experiments to develop the optimal campaign strategy for each target population. a partisan planner maximizes the posterior probability that the party’s own candidate wins the election subject to a budget constraint. whereas the objective function of a nonpartisan planner is the posterior expected turnout among a target population of voters. in a nonpartisan setting. the nonparametric method finds that recipients of mobilizing text messages in 2006 (Dale and Strauss. In the proposed Bayesian-decision theoretic framework. 2009) allows campaign planners. Thus. A formal decision-theoretic framework (Imai and Strauss. uniformly applying a single appeal to the entire target population is at best suboptimal and sometimes is not even feasible.second and related problem is that a planner faces budget constraints and must evaluate the cost-effectiveness of available strategies.

Kramer (1966) concludes that “the use of quantitative methods for policy analysis has proved to be fruitful in many different fields. few scholars who conduct field experiments have followed up on Kramer’s proposal to inform policy makers with a formal decision-theoretic analysis.1 Background of the Methodological Problem The problem of optimizing campaign tactics across voter characteristics was first introduced by Kramer (1966) more than forty years ago. an important and well-known methodological problem is that if subgroups are formed after the experiment is conducted in an ad hoc manner. and used. and these methods deserve to be more widely known. these studies do not examine the issue of cost-effectiveness in the presence of a budget constraint. 2004. One exception is a small number of researchers who have examined the related question of which subgroups exhibit larger treatment effects (e. which is an essential consideration for campaign planners.2008 mail and robo-call program that emphasized the economy had the most effect on individuals living in economically distressed areas.. 5. Moreover. After illustrating his decision-theoretic approach with artificial data about the partisan make-up of precincts. Nickerson and Arceneaux.g. Campaign managers can use these findings as a basis for planning their GOTV or persuasion campaign. Thus. Arceneaux and Kolodny. In that research. 2006. the analysis runs a risk of finding statistically significant results when no true relationship exists (Pocock et al. 2009). 2002).1.” Unfortunately.. a campaign manager is assumed to face the decision of whether to conduct blind or selective canvassing across precincts subject to a budget constraint. the planner needs a principled and systematic approach for deriving the optimal campaign 107 . in political science (p. However. Gerber. 160).

This framework differs from the literature by adopting a Bayesian approach rather than a frequentist approach based on the maximin or minimax-regret criteria (see Dehejia.. which is dominant in the treatment choice literature. This set of methods can be used to derive the optimal campaign strategy from randomized field experiments. 2005).strategy from experimental data while avoiding the post-hoc subgroup analysis and related statistical problems. Rather than taking a frequentist approach.2 The Formal Framework of GOTV Campaign Planning The problem of deriving the optimal nonpartisan GOTV campaign strategy from experimental data can be formalized as a statistical decision problem where the choices are treatment applications (e. 2005. This framework can then be extended to the optimal partisan persuasion (and possible partisan GOTV) campaign.. this method is based on Bayesian statistical decision theory (Berger. for a notable exception). The standard linear programming algorithm can be used to derive the optimal campaign strategy within this framework. 1985) and assumes that a GOTV campaign planner learns about the effectiveness of various mobilization methods for dif108 . Manski. The proposed methodology extends Kramer’s pioneering work by placing the campaign planner’s problem in the formal framework of statistical decision theory and applying modern statistical methods. Manski. To address the subgroup analysis problem. These methods are then extended to the cases of optimal partisan campaign planning.g. which has been somewhat neglected in the treatment choice literature a new variable selection algorithm is presented and then used in combination with nonparametric methods and cross-validation procedures in order to avoid the over-fitting of statistical models.g. 2005). 5. The decision-theoretic framework is based on literature about treatment choice (e.

. . treatments) to each member i of this population.1 The Planner’s Decision Problem The first step of the approach is to formally state the decision problem faced by a nonpartisan GOTV campaign planner. Typically. 2 109 . mobilization techniques may differ in their frequency. . content of messages.e.e. the proposed methodology is directly applicable to aggregate units. Although the units of analysis are represented as individual voters for notational simplicity.. Note that T = 0 represents the strategy of not mobilizing (i.e. K − 1} where K ≥ 2 denotes the finite set of mobilization techniques from which the planner makes a selection for each voter in P. For example. The methods apply equally to the situation where a planner makes a decision based on the recommendations of a data analyst. 1. the planner may consider three strategies (i. the planner’s decision problem is to assign one of K available mobilization techniques (i. and other aspects.2 5.2.S.. The planner then maximizes the posterior expected turnout among a target population of voters subject to the budget constraint.). If a GOTV campaign can be planned at the level of households or precincts. Then. An unordered set T ≡ {0. K = 3) where T = 2 and T = 1 represent a GOTV method based on a phone call and a postcard. whose complete list is available to the planner (at least in the U. timing. doing nothing). Let P denote this planner’s target population of voters where this population is assumed to be finite and of size N .. then i indexes the appropriate units rather than individual voters. Alternatively. respectively. The following sections describe the proposed methodology as if a campaign planner both analyzes the data and makes the decision. and T = 0 denotes a status quo strategy that involves no such phone call or mailing. the target population is the registered voters in the electoral district. .ferent voters by analyzing randomized field experiments.

Since P is a finite population. Given this setup. Since the values of these covariates are observable for every voter on the list.. and zip code. Rubin. following the statistical literature on causal inference. Then. but have different implications for the way the optimization is conducted. this assumption can easily be relaxed.Next. i. For the sake of simplicity. voters with the same value of X are assumed to be exchangeable since the planner 110 . For example. if the target list of voters is obtained from the voter roll. In either case. for example. the planner’s mobilization strategy is characterized by the function. 1]. x) may represent the fraction of voters with the observed covariate X = x who are contacted by the planner using the mobilization method t. then an analysis needs to be conducted at the household level. whose support is denoted by X . Alternatively.1) where the mobilization strategy δ(t. The outcome variable is binary and is equal to 1 if voter i casts a ballot and is equal to 0 if the voter abstains. (5. These two definitions become essentially identical when the number of voters is large. party registration. 1990). 2008). ·) : (T × X ) → [0. x) denotes the probability of receiving treatment t ∈ T for a voter with X = x. the observed turnout is given by Yi = Yi (Ti ).e. 1958. gender. δ(·. the distribution of X is known to the planner. this means that the planner knows the population distribution of X. δ(t. let Yi (t) represent the potential turnout of voter i that will be realized if the planner applies the mobilization technique t to this voter where i ∈ P and t ∈ T . such covariates may include age. Finally. suppose that the planner observes the J-dimensional (pre-treatment) covariates X for each member of this population P. If a voter’s turnout decision depends on the treatment status of other voters within the same household (Nickerson. P (X). voting history. any interference between voters is ignored (Cox. But.

the planner’s objective function can be written as a function of the mobilization strategy as well as the probability of a voter’s turnout given the values of covariates and the actual mobilization strategy applied to the voter. For this reason. (5. ·) from experimental data. then the planner applies the mobilization technique t that yields the highest value of the function ρ(t.2) A nonpartisan GOTV campaign planner’s goal is to derive the optimal mobilization strategy to maximize turnout of the target electorate. K−1 ∆ ≡ δ(·. 111 . Then. x) for each value of t ∈ T and x ∈ X . A typical voter mobilization method usually has a small effect on an individual’s turnout probability relative to the individual’s baseline predisposition to vote. x) = 1 for every x ∈ X . However. ·) : t=0 δ(t. ·) is unknown in practice. x) ρ(t. Xi )Yi (t) i=1 X = N x∈X P (X = x) t=0 δ(t.does not have additional information to distinguish between them. ∆. 1] with ρ(t. since ρ(·. the planner must make the decision under uncertainty by learning about ρ(·. If the turnout profile is known (and there is no budget constraint). Given this definition. Note that the function ρ(·. N K−1 g(δ. x) given each voter’s covariate characteristics x. ρ) ≡ E δ(t. ·) that satisfy the complement property constraint.3) where the turnout profile is denoted by ρ(·. The planner can achieve this by deriving the strategy that maximizes the expected turnout given the observed covariate information about X. is the collection of functions δ(·. the planner is interested only in determining the value of the function δ(t. the set of feasible mobilization strategies. (5. x) ≡ Pr(Y (t) = 1 | X = x). ·) : (T ×X ) → [0. ·) is unknown to the planner. x). The turnout profile represents the turnout probability given the characteristics of a voter and the mobilization strategy applied to this voter. Thus.

This concept is formalized by assuming that the planner faces the following budget constraint. (5.2). The first is the fixed overhead cost denoted by the function.3) or treatment effect (equation 5. x) ξ(t.3 in terms of treatment effect rather than turnout profile.4) hinges on whether the planner can elicit a prior more easily for the turnout profile or treatment effects (see Section 5.3. i.e. ∞) for each mobilization technique t.5) where 1{·} is the indicator function and C is the fixed positive constant representing the maximum cost allowed for the GOTV campaign. N K−1 g(δ. which is represented by a known cost function ξ(·. the planner maximizes.4) where τi (t) ≡ Yi (t) − Yi (0) is the treatment effect of mobilization strategy t on voter i and τ (·. Finally. Xi )(τi (t) + Yi (0)) X ∝ N P (X = x) x∈X t=1 δ(t. (5. the planner must determine which mobilization technique is most cost-effective for different voters. ·) : (T × X ) → [0. K−1 K−1 1 t=1 x∈X δ(t. x) = 0 κ(t) + N x∈X P (X = x) t=1 δ(t. x) τ (t.3. 1] with τ (t. x). In practice. x) ≤ C. κ(·) : T → [0. the cost of contacting each voter with covariate value x ∈ X . the planner cannot maximize the expected turnout without considering the differing costs of various GOTV techniques. ∞). ρ) = E i=1 δ(t..planners may cast the objective function of equation 5. 112 . The decision whether to maximize turnout profile (equation 5. Thus. In this framework. This cost is incurred so long as at least one voter is assigned to the mobilization method. The second component is the cost per voter. X ) → [−1. and whether the differences are large enough to warrant using multiple mobilization techniques. ·) : (T . This setup is mathematically equivalent to equation 5. x) ≡ Pr(Y (1) = 1 | X = x) − Pr(Y (0) = 1 | X = x). In this formulation. the planner needs to consider two kinds of costs.

. ξ(0. Although in real world settings this assumption may hold only approximately.. The observed data is denoted by D = {Yi . First. Xi }n where Yi is the binary i=1 turnout variable. This assumption may be invalid if.5 is taken with all possible values of t except t = 0. P (Y (t).e. ξ(t. the planner must assume that the experiment is conducted on a representative sample of size n taken from the target population P. X). The potential outcomes are denoted by Yi (t) where Yi = Yi (Ti ) for t ∈ T . Certain assumptions are required to link a field experiment with a planner’s decision problem. the planner would be forced to model the non-random sample selection mechanism to infer characteristics about P from the experiment. Without such an assumption. Ti ∈ T is the treatment variable representing mobilization techniques. X) = P (Y (t). the cost per voter is assumed to be positive. the election in which the experiment was conducted differs significantly from the election for which the planner is designing the GOTV campaign.2 Data from a Randomized Field Experiment Using the decision-theoretic framework. x) = κ(0) = 0 for all x ∈ X .e.. whereas the overhead cost is non-negative. Ti .since t = 0 corresponds to the status quo strategy of not mobilizing. i. κ(t) ≥ 0 for all t > 0. both the overhead and per voter costs are zero for this strategy. i. and Xi ∈ X is the same set of covariates used earlier.. Another key assumption required for the planner to apply the results of a randomized field experiment to the decision problem is that the joint distribution of potential outcomes and covariates does not change.e. it is essential for learning about the planner’s decision problem from a field experiment. x) > 0 for all x ∈ X and t > 0. 5. For the other mobilization strategies. i.e. This is why the summation in equation 5.2. i. 113 . for example. the planner can analyze the data from a randomized field experiment to derive the optimal GOTV campaign.

ρ(t. is identified. This prior distribution is denoted by π(ρ). In the Bayesian statistical decision framework (Berger. 5. ·).If these assumptions hold. the set of feasible strategies ∆ equals the set 114 . which is popular in the treatment choice literature. the planner can identify τ (t. x) but must estimate it from a finite sample. Although such large sample identification results are important.e. the Bayesian decision has a frequentist justification. (Similarly. Thus. ·). the optimal nonpartisan GOTV strategy δ ∗ maximizes the posterior mean of the expected turnout.. ·.3 The Bayesian Planner The framework of this method assumes that the planner is Bayesian and has a prior belief on the space of functions of ρ(·.6) where the optimization is subject to the budget constraint given in equation 5. To see this.5. First. in practice the planner must estimate ρ(·.2. This posterior belief about the turnout profile is represented by π(ρ | D). The Bayesian planner will update her belief via Bayes rule after observing the data from the randomized field experiment. consider an alternative class of GOTV mobilization strategies that depends directly on experimental data as well as the values of observed covariates. There are several reasons why the Bayesian optimality criteria are used rather than a frequentist approach based on maximin or minimax-regret criteria. where the planner is assumed to be Bayesian. ρ) dπ(ρ | D). (5. ·) from a finite sample and make the decision under uncertainty. the randomization of treatments in field experiments imply that the turnout profile. ρ(·. for all t ∈ T and x ∈ X . Such strategies are called “statistical treatment rules” and are characterized by the function δ(·. 1] (Manski. 2005). ·) : (T × X × D) → [0.) This problem can be addressed by deriving the optimal nonpartisan GOTV campaign strategy in this setting. δ ∗ = argmax δ∈∆ g(δ. i. 1985). x) = P (Y (t) = 1 | X = x).

FD . p. 59). K−1 g(δ. the costs for the status quo strategy t = 0 are zero and are excluded from the budget constraint. x. Furthermore. ρ) averaging over the prior distribution of D on D (see Berger. x) ≤ C. K−1 K−1 1 t=1 x∈X δ(t. which practitioners may find problematic. Manski. the frequentist objective function (i. In addition. One important advantage of the minimax regret criterion is that it avoids the subjectivity of Bayesian optimality because it does not require the use of prior information. On the other hand. 159. unlike Bayesian decision theory. Such strategies do not allow the planner to learn anything from the available experimental data. Second. ·. 1951. an alternative optimality criterion is Wald’s (1950) minimax regret principle (see Savage. 2005.e.6 agrees with the decision that maximizes the expected value of g(δ. if the size of the experimental data is large (as in many GOTV randomized field experiments) and little prior information is available. D) ∈ (X × D). x) dFD (5.8) for all D ∈ D. p. ·) that satisfy K−1 t=0 δ(t. a minimax regret rule can be viewed as a Bayes 115 . D) = 1 for all (x. FD . the Bayesian decision δ ∗ defined in equation 5. Again. Thus. x. frequentist theory based on the minimax regret criterion typically does not lead to the unique optimal decision. Also. 1985. risk) is given by. the Bayesian decision is essentially equivalent to the strategy that maximizes the expected turnout. which do not depend on the data at all (Stoye. In this setting. D) = 0 κ(t) + N x∈X P (X = x) t=1 δ(t. D)ξ(t. Manski. the strategies that meet the minimax regret criterion can include no-data rules. It has been shown that in most practically relevant situations. x. 2005).of all functions δ(·. 2009). D) ρ(t. (5. x. the Bayesian decision can be justified from a frequentist perspective.7) with the following budget constraint. ρ) = D N x∈X P (X = x) t=0 δ(t..

. 116 .rule with a prior (i. the Bayesian GOTV strategy can meet the minimax regret criterion.2. a least favorable prior) distribution (Berger. Since the cost function inputs are often exogenously determined (e.g. depending on the planner’s subjective belief. these two sources of information are combined via Bayes rule obtaining the posterior belief about the effects of mobilization strategies on different voters. even if every technique is marginally optimal for at least one voter. The planner now has an updated belief about the most cost-effective way to mobilize each voter.g. the planner might use a diffuse prior centered around a belief that there is no a priori difference in effects of a mobilization method across different voters. 5. the planner’s only meaningful decision might be determining a prior belief. Thus.. 1985. This is especially appropriate if the mobilization technique has not been empirically tested (e. the cost of postage and phone calls). If the treatment is an oft-used mobilization technique that has been extensively studied in the past (e. Taking into account overhead costs. In many cases.1 depicts the process by which a planner arrives at the optimal strategy via the proposed Bayesian decision theoretic framework. then the planner might center a prior around the estimated effects in previous experiments.. Note that the influence of prior belief diminishes as the size of experimental data increases. Chapter 5). Once a prior belief is elicited and a randomized field experiment is conducted.4 Bayesian Optimal Campaign Planning at A Glance Figure 5.e. The planner must determine the costs of each mobilization strategy (both overhead and per voter) as well as the prior belief about their effects on voters with different characteristics. the optimal strategy may be to implement a subset of the available mobilization techniques.g. canvassing). airplanes with reminderto-vote advertisements)..

Finally. the proposed optimization method uses this posterior belief and the exogenous costs. are represented by the oval. the planner first uses statistical models to obtain the posterior belief of the turnout profile.3 The Optimal Nonpartisan Campaign Strategy Using the decision-theoretic framework. These data and the planner’s prior distribution are combined via Bayes rule to produce a posterior belief about the effects of mobilization strategies π(ρ | D). π(ρ). (2) the overhead costs of each mobilization method. That is. for the planner.6. ξ. as defined in equation 5. to find the optimal campaign strategy. and (3) the cost per voter for each strategy. 117 . Conditional on this posterior turnout profile.1: An Overview of the Bayesian Optimal Campaign Planning Process. can be derived. D. 5. κ.Cost Function Input Values Planner’s Inputs: Prior Belief π(ρ) Overhead κ Per Voter Costs ξ Budget Constraint Posterior Belief Bayes Rule π(ρ | D) Optimize Expected Turnout Optimal Campaign δ∗ Experimental Data D Figure 5. One relative advantage of the proposed Bayesian framework over a frequentist’s approach is that the planner can completely separate the derivation of the optimal campaign strategy from the analysis of experimental data. δ ∗ . the planner determines the optimal campaign strategy by solving an optimization problem subject to a budget constraint. Data from randomized field experiments. ξ. ·). Inputs over which the planner has direct control are represented by hexagons and are: (1) the prior belief about the effects of various mobilization strategies on different voters. ρ(·. the optimal GOTV campaign strategy.

3. this section describes how to obtain the optimal nonpartisan GOTV campaign strategy given the posterior turnout profile. ˜ (5. Unfortunately. the optimal campaign strategy is given by.    1 if δ ∗ (t. x) ρ(t. In this case. as a general 118 . x).10 is not feasible and yet there is no overhead cost. the solution is trivial because the most effective strategy for each stratum defined by X is used. x) be the posterior turnout profile for each t ∈ T and x ∈ X .5. consider the case where the budget constraint is binding so that the strategy defined by equation 5. x) = t = argmaxs∈T ρ(s.5. the derivation of the optimal strategy is no longer trivial. To solve this optimization problem. for any given x ∈ X . the budget constraint may prevent the planner from employing campaign strategy δ ∗ .9) subject to the budget constraint given in equation 5. ˜ (5. Let ρ(t.10)   0 otherwise. However. Thus. x).1 The Optimization Method Before describing the method to obtain the posterior turnout profile. Next. in many cases. That is. first consider the case of no budget constraint. in this case. K−1 δ ∗ = argmax N δ∈∆ x∈X P (X = x) t=0 δ(t. ˜ Then. the optimal campaign strategy can be obtained by solving the constrained optimization problem.

K−1 δ∗ = argmax N δ∈∆ x∈X P (X = x) t=0 δ(t. x) ≤ C.e. 3 3 The strategy obtained with this approximation is still labeled “optimal” for ease of language. as the sample size increases. and then choosing the strategy that yields the greatest posterior expected turnout. the planner can solve the optimization problem in the same manner as in the case without overhead costs except that the total overhead costs are subtracted from the maximum budget allowed for the campaign.5.11)   δ(t. which is more computationally efficient. which is more difficult but can be solved by applying an appropriate algorithm. x) represents the proportion of voters rather than the probability of treatment assignment. may be preferred. Finally. ˜ (5..     K−1 subject to t=0 δ(t. δ ∗ . x). Appendix D. C − K−1 t=1 1 x∈X δ(t. x) ρ(t. If δ(t. i. 119 . The optimal strategy is obtained by considering all possible subsets and their corresponding optimal strategy. in some cases an approximate method. this difference vanishes. x) ξ(t. when the mobilization methods involve overhead costs. then this can be formulated as a mixed integer programming problem.1 details a fast approximate solution to the nonpartisan planner’s optimization problem.strategy. x) = 1 for all x ∈ X . Moreover. which is used in the nonpartisan application presented in Section 5. consider a case where only a subset of mobilization techniques is applied to at least one voter. x∈X P (X = x) (5. x) = 0 κ(t).1. the planner solves the constrained linear optimization problem.12) The standard linear programming algorithm can then be used to obtain the optimal strategy. Although the proposed general strategy based on linear programming is easy to implement. x) ≥ 0 for all x ∈ X and all t ∈ T . In this case.     K−1  N t=1 δ(t.

since many of the covariates available in GOTV field experiments are discrete (e. a model should not be overfitted to the data at hand and thus an appropriate set of pre-treatment covariates must be carefully selected. an appropriate variable selection algorithm is needed as a part of the model selection procedure. previous turnout history). a variable that explains a significant portion of treatment effect heterogeneity.e. this consideration is important because the sample size may not be large enough to accommodate a high number of pre-treatment covariates. which are potentially useful for deriving the optimal campaign strategy. a predictive variable. is not necessarily useful for decision making if its effect on the outcome is constant between the treatment and control groups. This is the main problem of the post-hoc subgroup analysis described in Section 5. a powerful predictor of the outcome. a statistical model should be able to accommodate discrete covariates.1. 120 . a prescriptive variable. These concepts are closely related. i. and Murphy (2007) point out. In particular. but are not the same. a statistical model needs to be flexible to avoid strong functional form assumptions. This excludes the use of some binary nonparametric regression models that require covariates to be continuous. Finally. In practice. the model fitting procedure should require a minimal number of arbitrary decisions from a planner (or the data analyst who is working for the planner). most variable selection algorithms are developed for improving prediction rather than decision making. When choosing a model. as Gunter.2 The Statistical Method Numerous statistical models can be used to obtain the posterior distributions of the turnout profile.3. Moreover. party registration. Zhu. i. Thus. For example. Third. a nonparametric model is useful because a planner does not know the exact functional form of the turnout profile a priori. First. Similarly.e. certain statistical issues need to be addressed. Overfitting tends to yield a model with poor performance in the actual election to which the derived mobilization strategy will be applied..5. Second.g...

assume that the sample size is sufficiently large and thus there is no need for variable selection. A Bayesian approach is used to model the turnout among voters with the same characteristics of covariates via the binomial distribution.15) Although its simplicity is attractive. where the posterior mean of ρ(t. btx ). Xi = x ∼ Beta(Wtx + atx .may not be selected by standard variable selection procedures if it does not predict the observable outcome (rather than the potential outcome) as well as other variables. (2007) but differs from their algorithm in that it uses a tree-based method (Breiman et al. 4 121 ..4 (5. X = x ∼ Binom(ntx . ρ(t. ρ(t. For the moment. x) is given by (Wtx + atx )/(ntx + atx + btx ). Wtx | T = t. 2008). Hill and McCulloch.14) for each t ∈ T and x ∈ X where Wtx is the number of voters with Ti = t and Xi = x who turned out. ρ(t. btx ) are the prior parameters. and (atx . It is also possible to use Bayesian regression tree models (Chipman et al. x) can also be approximated by a normal distribution. the planner may formalize a prior belief in terms of treatment effects. this turnout model is.. ntx is the total number of such voters. x) | Yi . The method is similar to that proposed by Gunter et al. This model yields the familiar posterior distribution. then the posterior of τ (t.13) (5. x)). (5. this model is unlikely to work well in practice if the sample size is small relative to the number of unique values the observable covariates Alternatively. Using a conjugate prior. 2008. 1984) rather than Lasso (in part because the outcome variable in political applications is categorical) and the measure of importance for prescriptive variables is somewhat different. x) | X = x ∼ Beta(atx . ntx − Wtx + btx ). If a normal prior distribution is used. The proposed algorithm meets these criteria relatively well.

given the selected variables. Chapter 7). The proposed solution to this problem involves three steps.. the planner will fail to differentiate across voters and will choose a suboptimal campaign strategy. 1984) is fitted to each treatment/control group to identify relevant subgroups within the group. 5 122 . . if one conditions upon irrelevant covariates. J. once all subgroups are identified in this way. Similarly. The details of the proposed method are: Step 1: (Selection of Predictive Variables) Fit a classification tree to the entire sample using all pre-treatment covariates and the treatment variable. Step 2: (Importance of Prescriptive Variables) Order each pre-treatment covariate. . This approach is labeled semi-Bayesian because the data are used twice—once to form subgroups and again to calculate the posterior distribution. ρj0 (t. Thus.5 The data is cross-validated to avoid overfitting. . a tree-based classification method (Breiman et al. if important covariates are not used to define subgroups..e. 1996. One disadvantage of this approach is that gradual changes in treatment effects across covariate groups are modeled as sharp discontinuities rather than smooth functionals.. i. Denote the predictive variables that are used in the final model by V . Use K-fold cross validation on the misclassification rate to determine the value of the parameter that controls the complexity of the tree. x). the complexity parameter in rpart() implementation in R (Ripley. x) for k = 0. x) by ˆ ˆ fitting a classification tree for the treated subset of the data (i.e. . we seek a principled way to select relevant variables and form subgroups before applying the above standard Bayesian model. a variable selection algorithm is applied to decide which variable needs to be conditioned upon when deriving the optimal campaign strategy. Finally.g. In particular. Ti ≥ 1) and using V and Xj as covariates. based on the statistic. Such overfitting will then necessarily lead to a mobilization strategy that will perform poorly in the actual election. e.X take. where gjk is the optimal overall turnout using the turnout profile ρjk (t. 1. Xj ∗ ∗ ∗ for j = 1. x) for all t and x. then the sample size within each subgroup will be too small to yield informative inferences about ρ(t. Sj ≡ gj1 − gj0 . 2. the Binomial-Beta model is applied within each subgroup to obtain the posterior distribution of ρ(t. Obtain ρj1 (t. x) is obtained by fitting a classification tree on ˆ This tree-based classification method is one of many nonparametric models.. Next. V ⊂ X. First. On the other hand. in most practical cases.

j The first step of the algorithm selects predictive variables using a standard fitting procedure of tree-based methods. . The statistic.e. This is done by comparing the optimal overall turnout corresponding to different values of the complexity parameters. The second step orders each pre-treatment covariate (including those identified as predictive variables in Step 1) according to its importance as a prescriptive variable. is designed to measure how much a planner can increase the optimal overall turnout by interacting the value of Xj with the treatment. given all predictive variables and different subsets of prescriptive variables. fit classification trees (without pruning) separately to the treatment and control groups using V as well as the j most important prescriptive variables in both models. In particular. Finally. practitioners can interpret each of the subgroups that 123 .. The third step uses the K-fold cross-validation procedure. Ti = 0). respectively. A main advantage of this semi-Bayesian approach is that it inherits the simplicity of tree-based methods. . The value of the complexity parameter from Step 1 is also used to fit the trees in Step 2. whereas the optimal campaign strategy is given by δ ∗ = δargmax1≤j≤J g∗ . Select the values of the complexity parameters for the two models based on the mean of the optimal overall turnout across K validation sets. Sj . Step 4 selects the final model among the ones chosen in Step 3 by again comparing the resulting optimal overall turnout and thus determines the optimal campaign strategy. Step 3: (Model Fitting) For each j = 1. Step 4: (Derivation of Optimal Strategy) The optimal overall turnout is given by g ∗ = ∗ ∗ max1≤j≤J gj . . Denote the optimal overall turnout and the optimal campaign strategy corresponding to the selected values of complexity parame∗ ∗ ters by gj and δj . J with Sj > 0: (a) Randomly divide the sample into K subsets for K-fold cross validation. This statistic provides a measure of the ability of Xj to explain heterogeneous treatment effects. in order to select the values of complexity parameters for classification trees fitted separately to the treatment and control groups. (b) Using K − 1 training sets.the untreated subset of the data (i. .

4 The Optimal Partisan Campaign Strategy The proposed decision theory framework and the statistical and optimization methods can be applied to the case of partisan persuasion and GOTV campaign planning.) 5. ·) (see equation 5. the use of cross validation procedure avoids overfitting. Thus. (Minor party candidates may exist. the planner is assumed to know the distribution of a certain set of covariates P (X). the planner’s mobilization strategy can be characterized by δ(·. transparent algorithms such as the one proposed here prevent planners from making arbitrary decisions when deriving the optimal campaign strategy. 5.2. They can thus use available prior information within subgroups by specifying the parameters of the beta prior distribution.2).result from the final model. The proposed approach also addresses three key issues highlighted earlier. Second. the decision problem of the partisan campaign planner is to assign one of K different mobilization methods (including the status quo strategy of doing nothing. First. 124 .1) and the set of feasible strategies is equal to ∆ (defined in equation 5. Again.4. which is denoted by Ti = 0) to each member of the target population P of finite size N . Third. but their probability of winning the election is assumed to be negligible. This case assumes that two major candidates are competing for the office.1 The Decision Problem Using the notation introduced in Section 5. A Bayesian planner can derive the optimal campaign strategy using randomized field experiments to maximize the (posterior) expected chance of winning the election. the tree-based classification models are nonparametric and can handle discrete covariates effectively.

such an objective function is called “0 − 1 loss function. and 0 if he loses. For this decision problem.” Finally. she votes for a third party candidate or abstains).2: (1) a field experiment is conducted on a representative sample from the 125 .4. Clearly. Let Yi (t) represent the potential voting behavior of voter i that will be realized if the planner assigns mobilization method t to this voter where i ∈ P and t ∈ T .. In the statistical decision theory literature. x) > 0(5. The variable Yi (t) can take three different values. Then. −1 if she votes for the opponent. the planner’s ultimate goal is to win the election. and 0 otherwise (e. Xi )Yi (t) > 0 i=1 = 1 x∈X P (X = x) t=0 . the outcome variable Yi needs to be redefined. h(δ.Unlike a nonpartisan campaign planner. 5.2. it equals 1 if voter i casts a ballot for the candidate of the planner’s party.16) i∈{i :Xi =x} Yi (t)/ N i=1 1{Xi = x} is a random variable representing the vote share differential for the candidate that will result among voters with covariates X = x if the planner assigns mobilization method t to them.5 also applies to the partisan planner’s situation. which can be represented as the indicator function. These assumptions are essentially identical to those described in Section 5. x)ν(t. certain assumptions are required for a partisan campaign planner to be able to use randomized field experiments to reach the optimal decision. a partisan campaign planner seeks the mobilization strategy that will lead to electoral victory. n K−1 h(δ.g. and therefore equation 5. the partisan planner typically faces a budget constraint similar to that confronted by the nonpartisan planner. δ(t. V ) is equal to 1 if the candidate of the planner’s party wins the election.2 Data Requirements As is the case for a nonpartisan GOTV campaign planner. V ) ≡ 1 where ν(t. x) ≡ δ(t.

are not publicly available and cannot be verified for each voter. Y (t) represents a trichotomous variable rather than a binary variable. Most classifiers including tree-based methods can handle such categorical variables even when the number of categories is greater than two. the integration cannot be explicitly solved. (5.4. ·). This means that a sample survey needs to be conducted to derive the optimal partisan mobilization strategy (unless the entire analysis and strategy planning are conducted at an aggregate level where validated election results are available). In particular. δ ∗ = argmax δ∈∆ h(δ. 5. First. is that the optimization in equation 5. Second. the optimal strategy maximizes the posterior probability of winning the election. P (Y (t).3. The problem. the planner can estimate ν(·. as in the case of a nonpartisan GOTV campaign. however. one important difference is that the derivation of the optimal partisan campaign requires vote choice data as well as turnout data for the voters who are the subjects of field experiments.5 where π(ν | D) is the posterior distribution of the vote share differential ν(·. unlike turnout data. ν) dπ(ν | D). remains identical between the experiment and the actual election. and (2) the joint distribution of potential outcomes and covariates. However.2. ·). recall that in the case of a partisan campaign.same target population of voters P. the objective function 126 . For example. vote choice data. except that the outcome variable is now trichotomous rather than binary.17 is not trivial for two reasons.17) subject to the budget constraint given in equation 5.3 Derivation of the Optimal Strategy The optimal partisan campaign is derived via Bayes theorem strategy. Using the classification method and variable selection algorithm similar to those described in Section 5. in the United States. X).

3. These difficulties are often amplified by the reality that the optimization must be conducted over a high-dimensional space if the number of treatments and/or the number of subclasses is large. the partisan planner can solve the following constrained optimization problem by applying the standard linear programming algorithm. but is relatively fast and closely approximates an optimal strategy. detailed in Appendix D. K−1 δ ∗ = argmax N δ∈∆ x∈X P (X = x) t=0 δ(t. x∈X P (X = x) (5. Indeed. he used the expected plurality of votes as the objective function while acknowledging that it may not be appropriate.     K−1  N t=1 δ(t. x) ξ(t.1. 5. x) | D). the method is applied to three data sets of randomized field experiments. However. this formulation is computationally quite difficult to work with” (p. 141). Instead. This method. x)E(ν(t. x) ≤ C. a fast and approximate solution to the partisan planner’s optimization problem can be applied. is not guaranteed to yield an optimal campaign strategy.19) To overcome this computational difficulty. Kramer (1966) noted that “the probabilistic objective is the more realistic. The intent is to use a randomly selected subset of the data as a test data set 127 .     K−1 subject to t=0 δ(t.1. These computational considerations prevented Kramer (1966) from using the probability of winning as the objective function of a partisan campaign planner. (5. x) = 1 for all x ∈ X .5 Empirical Evaluation of the Proposed Method To asses how effective the proposed method of deriving the optimal GOTV campaign is in real world applications. as explained in Section 5.18)   δ(t.is an indicator function that is not continuous. x) ≥ 0 for all x ∈ X and all t ∈ T . to maximize the expected plurality.2.

The three applications demonstrate the power and effectiveness of this method. 2009). parents of schoolaged children were more likely to re-evaluate the candidates once they learned their stance on education and individuals most likely affected by the economy were more likely to reevaluate McCain.and to obtain an unbiased estimate of the actual turnout (or the probability of winning) by applying the resulting optimal campaign strategy derived from the rest of the data to this test data set. the procedure results in an unbiased evaluation of the empirical performance of the proposed methodology. the outcome measure (candidate support) was gathered via a survey following treatment delivery. As detailed in Sections 5. the decision for the planner is which voters to treat with the appeal. In addition to using the nonparametric method. In each case. A randomized control group did not receive any of the treatments. The findings show strong support for the 128 . which were sent to potential voters in 2008. for more varied applications see (Imai and Strauss. All examples consist of a single treatment. Since the treatment is randomized and the test data set is not used to derive the optimal strategy. these results are verified using more standard generalized linear models. This procedure mimics the real world situation by using the test data set as the actual election to which the optimal campaign strategy is applied.3 and 5. The second example is a randomized mail experiment from the 2007 Kentucky gubernatorial campaign. One of three persuasive mailings was sent to potential voters. which was on education. two of the mailings used a “control group” to analyze the heterogeneous effects of the third mailing. The partisan applications demonstrate the powerful implications of the Personal Experience model in the real world.5.4. The third application is a program of antiMcCain robo-calls and mailings that focused on the economy.5. In both partisan examples (the 2007 and 2008 experiments). The first application is a nonpartisan text messaging GOTV experiment from the 2006 Congressional election.

an additional level of cross-validation is added to the procedure described in Section 5. 5.2) and obtain an unbiased estimate of the resulting turnout under the optimal strategy from test data that are not used in derivation of the strategy. a grid search is implemented to approximate the optimal complexity parameter.5. The entire procedure is repeated L times using each subset as a test set. Finally. The prior for each subgroup treatment effect distribution is Gaussian and is centered on the estimated population average treatment effect.4.1 Evaluation Method To assess the effectiveness of the proposed method. The derived optimal strategy is then applied to the test set to obtain an unbiased estimate of the resulting overall turnout. In addition. the average value of the L estimated optimal turnout rates is taken as an estimate of the turnout that would result under the proposed methodology.1.3. Matching voters to issues for which they have experience increased the candidate’s margin by 10-15 percentage points.1 and D. 129 .1.2). the optimization problem is solved using algorithms that yield approximate (but fast) solutions (Appendices D.2 to determine the optimality of each complexity parameter. The value of the prior variance is chosen so that it increases in proportion to the per capita budget constraint. The aim is to cross-validate the whole procedure (consisting of the three steps described in Section 5. In each case. After randomly dividing the sample into L subsamples. Each application uses a normal-normal conjugate prior for the treatment effect τ based on the setup defined in equation 5. A 10-fold cross-validation procedure is used for Step 3 of Section 5. The random assignment of treatments and the random subsampling of the test set make the unbiased estimation possible.3. one subsample is set aside as a test set and the proposed methodology is applied to the rest of the data.Candidate Evaluation Hypothesis presented in Chapter 2.3.

1. to either the treatment group or control group.3. registering organization. and Hispanic (in decreasing order of Sj ). 2009. log of county density.000 subjects nationwide were randomly assigned. past voting history. The treatment group received a short text message the day before Election Day. with at least one gubernatorial or senatorial campaign on the ballot in most states.e.” (The text message appeals were varied slightly. with 50 percent probability.5. The election was of moderate interest. The estimated average treatment effect. density. two nonpartisan organizations contributed the cell phone numbers of newly registered individuals to an experiment that tested the efficacy of text messages to mobilize voters (see Dale and Strauss. Democracy depends on citizens like you-so please vote! -PIRG. Available covariates for the analysis include gender. Subjects were included in the experiment when they registered to vote with a campus representative of the Student PIRGs or when they registered online with Working Assets. but these differences are ignored for this analysis. for details). The prescriptive variables chosen in Step 2 are age. race. log of county population density. and registering organization. is 3. whether the subject had voted in a previous election. Not all these variables are included in the final classification tree produced. The classification tree produced in Step 1 chooses the following variables as having predictive power (i.2 A Nonpartisan GOTV Campaign with a Single Mobilization Method During the 2006 election. the optimal campaign strategy is derived using this experimental data set.. ν): age. About 8. or more precisely the overall intentto-treat effect. Following the procedure outlined in Section 5. An example text message read “A friendly reminder that TOMORROW is Election Day. and gender.5. The outcome variable is dichotomous: one for having voted in 2006. and zero for not having voted. 130 . age.) Subjects were matched to the voter file using information on their registration forms.0 percentage points with a standard error of 1.

in this group. with population density and voting history also providing information about turnout rates. The resulting final trees are presented in Figure 5. The control group tree demonstrates that. The leaves show the predicted probability of voting conditional on their covariates. the age cut-point is in between 19 and 20. higher probabilities are to the right at each node. right branches represent voters who would falsify the nodes’ inequalities. In the treatment group. Left branches represent voters who meet the criteria of the nodes. The tree for the control group is on the left. as searching the complexity parameter space often finds that models with fewer variables result in higher overall turnout on validation sets. voters above the age of 24 are predicted to vote at a rate of 62%. whether the participant’s age is known and county population density are important determinants of the voting rates for voters aged 24 and under.however. the tree for the treatment group is on the right. 131 .2.

38 lg. and lg.dens is the log of the subject’s county population density.42 0.5 | lg.66 lg.73 0. In this example.dens< 5.dens< 5.2: Final Classification Trees for the Control Group (left panel) and Treatment Group (right panel).age< 24. Covariate abbreviations: age is the age in years of the subject.2 so that the resulting optimal turnout is maximized on the validation set.72 lg. The complexity parameters are chosen from 10-fold cross-validation using the algorithm described in Section 5.858 0.985 0.3.missing< 0.dens>=6.42 0. the planner’s budget allows treatment of at most 10% of the population.42 0.5 | age.missing is whether the age of the participant is unknown. 132 .dens>=5.5 0. age.043 0.852 0.62 age< 19. subjects who meet the node’s criterion are filtered through the left branch of the tree.79 Figure 5. At each node.

For a campaign that can afford to treat 10% of the population. the classification tree for the treatment group assigns them a probability of voting of 66% (the right branch of the right panel of Figure 5. these individuals vote at a rate of 62%. completely ignores covariate information and thus assumes zero treatment effect heterogeneity. if an organization can afford to treat only a subset of the population. which is based solely on the estimated overall average treatment effect (“ATE strategy”). Also.3 displays the performance of the proposed methodology given the maximum proportion of voters that could be contacted.and 19-year-olds. This less-informed strategy.3 shows that the proposed method results in a higher overall turnout than an ATE strategy (at least on average). The turnout rate achieved with the proposed method compares favorably with the turnout rate achieved using a naive strategy where randomly selected voters are contacted. The average turnout is computed for the treated voters who are assigned to the treatment group as well as for the untreated voters who are assigned to the control group under the optimal strategy. In contrast. In the control group (left panel of Figure 5.Much of the heterogeneous treatment effect can be explained by the voters’ ages. for 133 .) Figure 5.2). Potential voters between the ages of 20 and 24 are very responsive to the treatment. the treatment model predicts a negligible or negative treatment effect for 18. as their probability of voting is assigned at most 42% under the treatment unless they live in a county with a density within a narrow range. The latter is then subtracted from the former to yield the estimated overall turnout under the optimal strategy. (Such age ranges could not be identified by a classic model such as logit with linear explanatory variables. Figure 5.2) – a four percentage point increase above the average treatment effect. The turnout that would result under the optimal strategy is estimated using the difference in means estimator between the treated and untreated voters (solid lines with solid circles).

3: Empirical Evaluation of the Performance of the Proposed Method for the Text Messaging Experiment. The first strategy is the average treatment effect or ATE strategy (solid line). The estimator is applied to the test data which are not used to devise of the optimal strategy.0. which uses covariate characteristics of voters to determine which voters receive the treatment.02 q ATE Strategy q 0.03 q Optimal Strategy q q Overall Turnout Increase 0. Solid circles represent the estimated optimal turnout using the difference-in-means estimator.01 q q 0% 20% 40% 60% 80% 100% Maximum Proportion of Voters Contacted Figure 5. The second strategy is an optimal approach based on the proposed methodology. which contacts randomly selected voters. 134 . The figure displays the overall turnout increase that results from two campaign strategies as a function of the maximum proportion of voters contacted.00 0.

135 . The mailings were sent out on September 13. or lean. As predicted by the Personal Experience Model.” Given that the respondents were union members. The results were similar for the subset who received the school mailing: 70% Beshear. and 13% were undecided. 2008). Because the nonparametric procedure doe not isolate individuals who reacted negatively toward the treatment.example. and 15% undecided (n=448).321 participants were surveyed by telephone between September 17 and September 20. and corruption. Democrat Steve Beshear and Republican Ernie Fletcher. This lack of negatively responsive individuals is not surprising given that little backlash was found to the text messages in a post-treatment survey (Dale and Strauss. with no control group. Each participant received exactly one of the three mailings. it is not surprising that they generally preferred the Democrat: 69% supported Beshear (including leaners). no gain over the ATE strategy is seen when treating the entire population.3 Partisan Example: Parents and Education Spending In September 2007. 17% supported Fletcher. change. 5. respectively. were identified. for the text messaging application. probable. and 1. Respondents reported which candidate they currently supported. organizations operating under tight budgetary constraints would be wise to use the proposed nonparametric procedure. which focused on schools. 15% Fletcher.5. The respondents to the survey were asked a standard vote question in which the parties of the candidates. indicating whether their support was “definite. Overall. the school mailing worked particularly well among parents. a union delivered three mailings to its members in Kentucky in a randomized field experiment. the optimal campaign strategy achieves an overall turnout increase on average over three times greater than the turnout increase under the ATE strategy.

two-tailed.e. These regressions indicate that an independent parent increased his/her probability of supporting Beshear by 11 percentage points after receiving the school mailing (from 66% to 77%). The 13% of respondents who have children in the household (as indicated by the commercial data) was combined with the 18% of respondents who are not single and are under 40 to compose the target group: the 26% of the population likely to have children. Unfortunately.03.Part of the schools mailing is pictured in Figure 5. This result is at the edge of statistical significance.15.” Figure 5. 136 . An ordinal regression with the trichotomous Beshear/Fletcher/undecided outcome variable (i.4.4: Part of the Kentucky Mailing. This group was labeled (with some hesitation) “parents. and controlling for the prior probability of supporting Beshear (the support score demographic model). The Personal Experience Hypothesis suggests that parents would be especially responsive to this piece because they are able to independently judge politicians on the state of Kentucky schools. Parents who received the school mailing favored Beshear 75%-12% (n=97) as displayed in Figure 5. the school mailing*parent interaction coefficient is significant with p=0. The survey did ask about marital status. the survey of respondents did not ask whether the respondent was a parent. A linear regression using the 7-point vote preference. parents who received other mailings favored Beshear 68%-22% (n=223). however. two-tailed. and a linked commercial database includes data on children in the household.5. less granularity) and no controls produced a p-score for the interaction of 0..

which selects both the commercial data on parents and the imputed data in various iterations. The mailing has an average treatment effect of 3. Beshear’s probability of winning passes the coin-flip mark with just 10% of the population treated. this 10% figure has a lot of noise. the algorithm. Nevertheless. union members.Beshear Margin 60% 70% 80% Pare nts Corruption & Change Mailing Appeal Type 30% 40% 50% Non−parents Schools Figure 5.9 percentage points. provides clear benefits. thus treating 77% of union members (under the ATE approach) would push Beshear over the 50% change of winning mark. Dashed lines represent one standard error. To maximize the mailing’s effect on Beshear’s margin of victory. 137 . Point estimates are represented by filled-in squares and triangles. Beshear is assumed to have a baseline margin of -3 percentage points. The population of the experiment. Beshear’s margin by type of mailing and whether the recipient is believed to be a parent are shown. is not representative of the Kentucky electorate. dotted lines represent two standard errors. With such a small sample. the fact that treating 20% of the population decreases the chance of winning is an indication of this noise. Figure 5.5: Parents React Strongly to School Mailing.6 displays the significant cost savings provided by the nonparametric algorithm.

4 q 0. which is determined by budget constraint.0 0.6: Empirical evaluation of the performance of the nonparametic method on the Kentucky data. 138 . and they are more capable of processing the political arguments of the mailing.6 q q q q q q 0.4 0. The left panel displays the estimated probability of Beshear winning under two treatment strategies plotted over the maximum proportion of the electorate treatable under the budget constraint.4 0.6 Optimal Strategy q 0. Only in the situations where nearly all voters can be contacted.0 0. The dashed lines represent the “ATE strategy” in which random voters are contacted.8 0.0 0. This mechanism.2 0.6 0.8 Optimal Strategy 0.0 q 0.0 0.0 Maximum Proportion of Voters Contacted Maximum Proportion of Voters Contacted Figure 5.8 1. which uses covariate characteristics of voters to determine which voters receive the treatment. The right panel plots the actual proportion of voters contacted by the mailing against the maximum proportion of voters contacted. Certainly there are alternative explanations of why parents are more responsive. parents may be more likely to read the mailing because its subject is schools.1. Personal experience with political issues is expected to strengthen both these processes.2 ATE Strategy q q q 0.0 1. The solid lines represent the optimal strategy based on the proposed methodology. The results are calculated based on 10-fold cross-validation. however. First. is consistent with the Personal Experience Model: individuals might pay more attention to a mailing that they realize will inform their political judgments.0 Actual Proportion of Voters Contacted ATE Strategy q q Probability of Beshear Winning 0.2 0.6 0.2 0. The optimal strategy outperforms the ATE approach strategy under tight budget constraints.4 0.8 1. does the algorithm choose not to canvass some voters—and it chooses these untargeted voters poorly.

967) among those who received any mailing and 21 points (Dem 53% . Beshear’s margin does not increase in the non-parents group. 5. The Democrat’s margin was 23 points (Dem 54% .A second possible explanation is that parents have a self-interest in school funding. they may increase their vote for Beshear more than the general population. As with the observational data of Chapter 3. this experiment does not allow these two mechanisms to be disentangled. Certainly movement against the candidate delivering a mail piece is possible. n=2.McCain 32%. meaning any votes gained among non-parents who support education funding must be offset by nonparents moving against Bashear. The focus of the mailings. with varying degrees of emphasis. n=987) in the control group— not a significant difference. Two factors mitigate this alternative hypothesis. A third explanation is that the mailing convinces a voter to support Beshear solely on the basis of her support of education spending. First.4 Partisan Example: Disadvantaged Voters and the Economy In March 2008.5. but the algorithm does not find large shifts against Bashear (Figure 5.6).7 displays part of the most straightforward of the mailings. the evidence for cue-taking in Chapters 3 and 4 is strong. a union tested three anti-McCain mail pieces (along with a control group) in Ohio. 139 . Second. Since parents might be more inclined favor of education spending. These mailings were followed up with robocalls that reinforced the topic of the mailing.McCain 31%. Figure 5. this explanation assumes that voters do not cue-take from politicians.and middle-class. while the Democratic presidential primary race was still undecided. was how McCain’s policies would damage the economy and weaken the working. which were mailed to union members. Participants were surveyed by telephone shortly after the mail pieces and robocalls were received.

LLC for its help in matching voters to census information. Among those who lived in the 70% most well-off places. The Personal Experience Hypothesis would predict that members who had personally experienced the economic downturn (or were close to those who had) would react more strongly to the mail and robocall program. con6 Special thanks to Catalist.000 in the advantaged group. from 18 points to 16 points (Figure 5. this margin decreased sightly between treatment and control. which was significantly lower than the average of nearly $50. such as percent unemployed and median household income. This key group’s support for the Democrat increased significantly after receiving the mailings and robocalls: from 16 points in the control group (Dem 48%.8).6 The median household income of the disadvantaged group is about $32. With only 227 disadvantaged control-group respondents.500. To reduce this possibility. The subgroup of interest is the set of individuals who live in economically distressed areas. 140 . n=227) to 28 points among anyone who received a mailing (Dem 58%.Figure 5. poor randomization may be affecting the results. A factor analysis identifies the top 30% of respondents who live in disadvantaged areas (after individuals who could not be matched to census data were eliminated). defined by several census characteristics. These data are suspicious because of the low Democratic margin among the disadvantaged in the control group. vote preference is regressed on the treatment condition and its interaction with the subgroup of interest. McCain 32%. McCain 30%. n=710).7: Part of the ant-McCain mailing.

or undecided) indicates that a member with median characteristics would increase his/her probability of voting for the Democrat by 7 percentage points as a result of living in a disadvantaged area and receiving the treatment (from 38% to 45%).7 With a linear regression on the 7-point vote question. from “Dem: lean” to “Dem: probably”) after receiving the mailing.trolling for a demographic-based partisanship score provided by the union.8 In nearly every iteration.. An ordinal regression on the trichotomous outcome variable (i. the margin between the candidates is artificially tightened to examine differences between the algorithms.e.5 percentage points in this case. McCain.g. the algorithm selected the continuous variable of economic distress.9 displays on the left panel.15.5 percentage points. the treatment effect for the advantaged subgroup is nearly exactly zero. Big gains are also found under loose budget constraints. The algorithm. As Figure 5. A backlash is present among some of the population. but for the disadvantaged subgroup the coefficient is large and nearly statistically significant (p=0. but that is probably an underestimate since the overall persuasion boost declines to 3.and large-budget operations benefit.. The interaction coefficient can be interpreted as: the average union member living in a disadvantaged area moved one-third of a survey response unit (e.4 percentage points. treating 10% of the population increases the Democrat’s probability of winning to 38%. yet treating 80% of the population under the proposed strategy would produce an effect of 4. when the budget constraint is removed. When the nonparametric algorithm is applied to this data. 8 7 141 . This score was developed before the experiment. the only options are voting for the Democrat. The ATE is 2.9. right panel). estimates that 10% of the population is negatively treatment responsive (Figure 5. both small. two-tailed). Again. while the ATE strategy produces a win only 24% of the time.

9 This result is not statistically significant at conventional levels. alternative hypotheses are possible.9 Also. The only evidence offered is that in a mail test later in 2008 by the same union dealing with Social Security. the Social Security mailing cannot be used as a test of the Personal Experience Hypothesis because the mailing was delivered to seniors only. Generic Democrat’s margin over McCain by treatment condition and whether the union member lives in a disadvantaged area are shown. Especially concerning with this test is that disadvantaged members would react strongly to a variety of issues. Point estimates are represented by filledin squares and triangles. as with the Kentucky test. As with the Kentucky test. Dashed lines represent one standard error.8: The Economically Disadvantaged React Strongly to Economy-focused Mailings. belief that the economy is performing badly) might generate the difference between the two groups. Unfortunately. dotted lines represent two standard errors. 142 . not just the economy. prior issue position (in this case. this proposition cannot be tested in a rigorous manner since there was no group that received a generic message.Figure 2: Members in Disadvantaged Neighborhoods React Strongly to Economic Mailings Generic Dem's Margin over McCain 30% d tage dvan Disa Advantaged 0% Control Treatment Condition 10% 20% Any Mailing Figure 5. Thus the experiment lacks crucial variance in the interacting variable of interest (seniors v. younger voters and Social Security appeals). Further. the economically disadvantaged were less likely to react positively to the treatment than the economically advantaged.

4 0.0 Maximum Proportion of Voters Contacted Maximum Proportion of Voters Contacted Figure 5.6 q 0. These shifts might appear large.8 0.1.0 0.0 0. The left panel displays the estimated probability of the generic Democratic winning under two treatment strategies plotted over the maximum proportion of the electorate treatable under the budget constraint.6 0. 2009).4 q q q 0.6 q q Optimal Strategy 0. Targeting parents with a school mailing rather than a generic mailing increases the probability that a voter in this group shifts her preferred candidate by 11 percentage points.2 ATE Strategy q q q 0.8 q q Optimal Strategy 0. See Figure 5.0 0.g.6 for details.0 1. The results are calculated based on 10-fold cross-validation.0 q 0. but often experiments with small average treatment effects significant subgroup differences (e.2 0.2 0. The algorithm successfully finds voters who are very treatment responsive and negatively responsive.9: Empirical evaluation of the performance of the nonparametic method on the anti-McCain experiment data.8 1.4 q 0. and among members the union is likely a trusted source.. Also.6 0.0 0.8 1.0 ATE Strategy q Actual Proportion of Voters Contacted Probability of Obama Winning 0.2 0. Arceneaux and Kolodny. This member-organization relationship probably increased the credibility of the message. the union clearly identified themselves as the source of the mailing.4 0. Targeting disadvantaged voters with an economic mailing increases the probability of vote change by 7 percentage points. Both partisan experiments demonstrate large effects of targeting participants with personal experience. 143 .

Nevertheless.] This is unfortunate.137) Yet.. This chapter takes up Kramer’s proposal and show how modern statistical methods can be used to help practitioners devise strategies and implement optimal policy using the Personal Experience Hypothesis. The proposed methodology is agnostic to theories (e. Such approaches demonstrate one application of Kramer’s larger goal: arming campaign strategists with the theory and methods to more accurately target voters who are susceptible to persuasive appeals. for many of the traditional concerns of political scientists appear to be quite susceptible to this sort of analysis. these theories can help practitioners form priors. the Personal Experience Model) about why certain subgroups should be more responsive to specific treatment. [.. these efforts have not been made by political scientists.] By and large. In the past two decades..6 Conclusion More than forty years ago. Kramer (1966) observed.. however. and better explain the results of the nonparemetric method. the use of quantitative methods as aids for decisionmaking has become common in many fields.5. (p.g. [. identify sources of potentially useful data. 144 . political scientists have since generally neglected to use quantitative methods to directly inform policy makers who must make decisions using available data..

. 1990. campaigns can have an important effect on how voters judge candidates in an election. The Personal Experience Model is one theoretical connection between specific voters and issues. However. If voter-issue linkages are identified correctly. several alternative group-issue connections are described in both the academic (Krosnick. This chapter outlines the situations in which campaigns should leverage these theories to microtarget voters and the normative implications of that microtargeting on democracy. The advent of electronic voter files and commercial databases has made this strategy more efficient and more widely applicable to groups of voters (Malchow. 2003). 2001) and practitioner literature (Sosnick et al.Chapter 6 Campaign and Normative Implications of Microtargeting 6.. 2006). Chong et al. Microtargeting is a powerful tool.1 Campaign Microtargeting Campaign microtargeting is a method for identifying a subset of voters to which to direct a tailored message. Chapters 145 .

first the situations in which campaigns would not choose to microtarget are outlined. match this data to the voter file. and 5 provide the empirical evidence to support this claim and demonstrate how campaigns can leverage personal experience to garner support from voters who would otherwise vote for the opposing candidate. Figure 6. 4. campaigns that need to change the landscape of a race dramatically would do best not to microtarget. Voters in the top decile (by their probability of being undecided) are twice as likely to be undecided as the population as a whole (20% to 10%). 146 . CHAID) method to correlate measures available for the entire electorate to the survey response of interest (Malchow. such as hunters or those with low incomes. A second possibility is to use non-voter-file information (e. After a sufficient number of voters (perhaps on the order of 2.. and most complex.1 displays the results of a hypothetical microtargeting of undecided voters. Various methods for identifying voters who will be responsive to narrow issue appeals are available to campaigns. logit) or a non-parametric (e. licensing lists).g. The easiest method. The campaign can then assign a probability of being in the specified subgroup to all voters.g. census data. such as seniors. For example.000) have responded to a question. The third. Similar to the optimizations described in Chapter 5. perhaps. asking a question (e.. In this chapter. method is to survey a group of voters. the normative implications for microtargeting are discussed.. is to use well-defined categories from readily available voter files. 2003).3. and identify specific groups. “Do you have children under the age of 18?” or “What is the most important issue to you?”) the answer to which is not available in any database. the campaign would contact the voters with the highest probability of being in the target group.g. a campaign analyzes the data using either a parametric (e.g. In the broad sense.. Second. a campaign might use birth year to target an age group.

The use of a training and test set can ensure that microtargeting models do not over-fit and that they do identify targeted voters better than if the campaign were to deliver messages randomly.1: Example of the results of a campaign microtargeting undecideds.Percent Undecided 0% 5% 10% 15% 20% 1 2 3 4 5 6 7 8 9 10 Decile of Undecided Score Figure 6. If a voter is experienced with an issue. In an attempt to maximize the probability of a targeted voter changing her vote choice. The top decile of voters (10) includes twice as many undecided voters as the population average. In essence. as electorate-wide covariates may not predict survey responses well. Further error is induced when the targeted population is identified by survey. or has self-interest in an issue (all microtargetable subgroups). that voter is more likely to pay attention to campaign appeals 147 . Targeting these voters is thus twice as efficient as a random strategy. campaigns often also use the survey/CHAID method to identify voters who are on (or near) the fence for their vote choice decisions. microtargeting increases the efficiency of campaign issue signals and the linkage between a candidate’s stances and a shift in voters’ candidate preferences. belongs to an issue public. None of these methods perfectly identifies a list of voters who will change their votes if targeted with an appeal.

the voter more easily processes political arguments on the issue (as a result of their experience. appetites. or self-interest) and thus is less likely to ignore messages on the issue because the candidate’s appeal is either too complex or uninteresting.2. who are in an issue public. 6. Applying the Personal Experience Model to various forms of voter targeting demonstrates how microtargeting enhances the connection between candidate stances and vote choice. then the campaign’s appeal will raise the voters’ evaluations of the candidate (on average). Again. these voters can make the difference between losing and winning.1 Implications for Campaign Strategy Microtargeting and Message Control If a campaign can identify a subgroup of voters with experience on an issue. This independent information allows the voter to judge the politicians on the issue when the voter cue-takes from the politicians’ positions. or who have self-interest. candidate stances and candidate evaluation. the voter can “ground-truth” the stances taken by the candidates (see Chapter 2). assume that campaigns target voters who have experience. Microtargeting increases the pace of voter learning about pivotal issues. it will be able to deliver issue-specific messages to these voters. an increase in microtargeting (and hence a reduction in cue-taking) strengthens the relationship between pre-campaign voter attitudes. In some cases.2 6. If the voters’ opinions are congruent with the candidate’s platform. In all these cases. and does not affect vote preferences. Because each of these types of targeted voters has information or opinions that originate independently of political elites.on that issue. Since cue-taking polarizes the electorate. 148 .

Some of these individuals will likely abstain or vote for the opponent based on their predispositions if the campaign does not target them. 2002. First. in some situations. A polarized electorate means that few voters are undecided or potential defectors. September 1.1 While potentially small in number.) Even if microtargetable.2 Because of cue-taking. when Democrats pushed the issues of health care and corporate responsibility and Republicans attempted to put taxes and national security at the fore of voters’ minds. campaigns can spend resources attempting to control their overall message agenda. 2 Nagourney. if all microtargetable voters disagree with the candidate on the issue in question. the term “microtargetable” is restricted to this group of voters. For instance. microtargeting is a net loss for the campaign.” New York Times. independent of other options for resource allocation. Microtargeting and message delivery cost the campaign both in terms of money and opportunity cost. If microtargetable voters do side with the candidate on the issue. sending these voters tailored messages is useful only if these individuals are unlikely to vote for the candidate in the absence of the appeal. In addition to microtargeting. 1 149 . polarized elections where the dimension of opinions on the issue in question is similar to the overall ideological dimension of the campaign. then clearly the campaign should not raise the issue. some portion of the electorate would shift their vote choice only if microtargeted. so little get-out-the-vote effect is possible. the campaign can identify these like-minded voters and deliver messages only to them. microtargeting is not necessarily beneficial for a campaign. the proportion of these voters in this circumstance can be small in high-salience. like-minded voters exist. Adam. microtargeting is not helpful. “Domestic Concerns Take Center Stage In Congress Races. In three situations. One example is the 2002 midterm elections. (For the rest of this chapter. However.However. “Similar issue dimensions” means that few voters currently supporting the opponent are likely to be congruent with the candidate for the issue in question. these A “high salience election” means few registered voters abstain. thus the benefits must be proportional to the resources devoted to microtargeting.

First.. The candidate must choose between two methods of campaigning. I also assume that the campaign has a wealth of knowledge about the electorate to guide its decisions.2 Translating the Personal Experience Model into a Formal Game To delineate the situations in which microtargeting is the optimal strategy. Alternatively. the sudden shift in dialogue in the 2008 presidential election after the third debate to “spreading the wealth” and “Joe the Plumber” polarized the public on economic issues and did not move the overall vote. 1981). The campaign also has knowledge about what percentage of the electorate is microtargetable (i. has stable opinions) on each potential 3 See Gallup Polls on wealth redistribution and aggregate vote choice surveys on pollster. 150 . (2) has chosen a platform on which to run.e. This occurs after the issue has resolved (Canes-Wrone et al. 1970) and retrospective voting in general (Fiorina.non-microtargeted issues often have little effect on the vote preferences of the electorate.. I develop a model that explores the campaign resource allocation decision. Through public opinion polls (or other means). and (3) must decide how to spend the campaign’s limited resources (i. 6. For instance. then the electoral payoff can be large. I presume that a candidate (1) has decided to run for office.3 The exception to this rule is when the media or public agree that one party (or candidate) has outperformed the other on an issue.com. In the broad sense. money).2.. the campaign knows the general predispositions of the electorate.e. If a campaign chooses to spend resources to “broadcast” a message on a single issue and that issue resolves in its favor. the campaign can broadcast to all voters the candidate’s stance on an issue for which the voters have unstable opinions but that may resolve in the candidate’s favor. 2001) and accounts for the importance of economic voting (Mueller. the campaign can microtarget voter-issue pairs for which the voters have stable opinions that are congruent with the candidate’s platform.

an appeal from each ideological side). The canonical example of one-sided issue streams is the early discussion of the Vietnam war (Zaller. Instead. campaigns can maximize their efficiency by microtargeting only the voters with congruent opinions. 151 . a criterion that is not always met. If a campaign delivers targeted messages to these voters.. even if wide swaths of the electorate have stable opinion on issues.g. Campaigns generally have the ability to identify these voters to some extent. the more voters it can reach. Segments of the population who lack issue experience (or an alternative issue connection such as self-interest) are not likely to shift their vote on the issue in question. The campaign whose platform is congruent with the voter’s opinion in this pairing would prefer that the voter learn about the candidate’s stances on the issue and have the issue on the top of her head come Election Day. an improving economy helps the incumbent party). and they do not change their vote choice. these voters cue-take from politicians. the campaign does not want to emphasize the issue among voters who have stable opinions that are incongruous with its platform. the campaign can shift some of these microtargeted voters’ candidate preferences. I also assume that the more money a campaign invests in microtargeting. The Personal Experience Model demonstrates how specific segments of the electorate are disproportionately responsive to certain issue appeals when forming candidate evaluations. This polarization mechanism requires two streams of considerations (i. the campaign has accurate beliefs about the probability that certain issues will resolve in its favor by Election Day (e. Any voter-issue pairs in which some voters have a stable opinion is labeled a “potentially microtargetable” pairing. However. In addition.campaign issue.. perfect identification is not necessary. their vote choices polarize.e. Hence. 1992).

In this case, both liberals and conservatives alike supported the war as that was the only position represented by political elites. Message streams can become one-sided when an issue resolves. For instance, when the country’s economy tanked in the fall of 2008, only 5% of the public rated the economic situation of the country as “excellent” or ”good.”4 Voters can then use the single stream of information from the media to “ground truth” politicians performance. The evidence that resolved issues affect large swaths of the electorate is clear when economic and war data are correlated with election results on the national (Hibbs Jr., 2000) and state levels (Campell, 1992; Cohen and King, 2004). In the model, an issue can (1) resolve in favor of one candidate, (2) resolve in favor of the other candidate, or (3) not resolve during the course of the campaign. Campaigns can choose to focus their appeals on certain issues in an attempt to prime voters’ minds (i.e., place considerations about this issue on the tops of voters’ heads). In essence, campaigns can place bets that certain issues will resolve in their favor. One possible exception to the cue-taking rule is issue ownership (Petrocik, 1996). Related to the Broadcast Corollary discussed in Chapter 2, the theory of issue ownership stipulates that wide swaths of the electorate favor one side of an issue. If a party “owns” an issue, it can be considered an “easy issue” (Carmines and Stimson, 1990) in that voters will likely have a stable opinion on the issue independent of cue-taking. However, elections in which a large majority of voters side with one candidate on an “easy” or “owned” issue are usually uncompetitive elections. The campaign that has this majority of voters on its side most likely has a winning (and perhaps dominant) strategy to emphasize this issue. Since close campaigns are more interesting, the model instantiations
4 “How would you rate economic conditions in this country today – as excellent, good, only fair, or poor?” USA Today/Gallup Poll. Oct. 10-12, 2008. N=1,269 adults nationwide.

152

in Section 6.2.5 usually consider cases in which a small percentage of the electorate is microtargetable.

6.2.3

The Model

In an election between two candidates, D and R, the campaigns decide how to allocate resources across two issues (A and B) and whether to microtarget or broadcast their messages about these issues. In the hypothetical world in which neither campaign makes any effort, the predispositions of the voters dominate. Both campaigns have common beliefs about their chance of winning in this no-campaign universe (i.e., a benchmark poll is taken before the campaign to test the mood of the electorate). This belief is represented by a mean percent of the vote for candidate i, µi , and a common uncertainty, σ 2 (i ∈ D, R). This representation is similar to a valence advantage in other models (Groseclose, 2001). The vote for candidate i, Vi , is distributed normally, Vi ∼ N (µi , σ 2 ). Candidate i’s utility, Ui , is the campaign’s probability of winning, µi − 0.5 ). σ

(6.1)

Ui = Pr(i wins) = Pr(Vi > 0.5) = Φ(

(6.2)

To simplify the analysis, I assume only two candidates in the race, µD = 1 − µR , and a fixed electorate with 100% turnout. The tiny tails of the vote distribution that extend beyond the logical range of Vi , [0, 1], are ignored. Implicitly in the model, ties are broken with a coin flip. If the campaigns garner exactly the same number of votes on average (i.e., µ = 0.5), then the probability of winning ap-

153

proaches 0.5 as the uncertainty of the result goes to zero (i.e., σ 2 → 0). This characteristic of the model is a result of the symmetry of the normal curve about its mean. Under all circumstances, campaigns want to increase their share of the vote, µi , since
∂Ui ∂µi

> 0. They attempt to increase this share by taking stances on issues. For the moment,

let the issues equal advantages for the campaigns. Issue A is an issue for which a portion of the electorate has experience, although some experienced voters side with candidate D and some with candidate R. Let the proportion of all voters who fall into one of those two categories (i.e., who are microtargeted by the candidates) be mD and mR , respectively. For now, assume a level playing field: mD = mR . The other issue, B, is not microtargetable but has a chance of resolving by the election. If this issue resolves in candidate D’s favor, an occurrence with probability of pD , q percent of voters switch their preference from R to D. A symmetric switch of q percent of the electorate occurs with probability pR ; i.e., in this case, the issue resolves in candidate R’s favor. Let pD = pR . The campaigns do not have the resources to fully utilize the advantages offered by both approaches. Instead, they must choose a combination of microtargeting and broadcasting. Assume that both campaigns have equal resources, a budget of 1, and campaign i spends δi ∈ [0, 1] on microtargeting. A campaign must spend its entire budget to gain the maximum votes from the strategy-issues combinations above.5 The resulting distribution of the percentage of votes for candidate D, given both campaigns disbursements, is

VD (δD , δR ) ∼ N (µi , σ 2 ) + δD mD − δR mR + (2 − δD − δR )BS(pD , pR ) VD (δD , δR ) ∼ N (µi , σ 2 ) + m(δD − δR ) + (2 − δD − δR )BS(p, p)
5

(6.3) (6.4)

If one campaign spends all its resources on broadcasting, then q percent of voters switch. If both campaigns only broadcast, then 2q voters switch.

154

where BS(p1 , p−1 ) is a Bernoulli scheme with probability p1 of outcome 1 and probability p−1 of outcome -1 (see Appendix E). The analogous equation shows the vote for candidate R. The mean and variance of Vi are, ¯ mean(Vi (δi , δ∼i )) = Vi (δi , δ∼i ) = µ + m(δi − δ∼i ) Var(Vi (δi , δ∼i )) = σ 2 + 2pq 2 (2 − δi − δ∼i )2

(6.5) (6.6)

If neither campaign spends any resources broadcasting (i.e., δi = 1), each candidate’s contribution to the variance (Equation 6.6) goes to 0. As before, the utility for candidate i is the probability of winning, Ui = Pr(Vi > 0.5). For candidate D,

UD = (1 − 2p)Φ((µ + m(δD − δR ) − 0.5)/σ) +pΦ((µ + m(δD − δR ) + q(2 − δD − δR ) − 0.5)/σ) +pΦ((µ + m(δD − δR ) − q(2 − δD − δR ) − 0.5)/σ)

(6.7)

This model includes two assumptions that are rough approximations of the real world and are not likely to hold in actual campaigns. First, the proportion of voters who will switch their candidate preference because of microtargeting (m) does not vary based on the resources spent on broadcasting (1 − δ), or vice versa (with q and δ). In essence, this simplification assumes a uniform distribution of the electorate across the ideological spectrum regardless of whether a candidate becomes more advantaged (i.e., the decision cutpoint on the ideological spectrum shifts); the number of nearly undecided (i.e., indifferent) voters is constant. Second, if the campaign microtargets voters who have a high probability of shifting their vote choice, the rate of successful microtargeting is not affected by spending on broadcasting. In other words, wasteful spending by campaigns that “doubles up” on voters—contacting them with both microtargeting and broadcasting when only one of the 155

5)/σ) +p σ (6. Formally. m > q → is increasing in δ. the partial derivative of candidate i’s utility is taken with respect to the campaign’s strategy. then the campaign should spend all its money on microtargeting. δ∼i ) = (1 − 2p) φ((µ + m(δi − δ∼i ) − 0.5)/σ) σ m+q φ((µ + m(δi − δ∼i ) − q(2 − δi − δ∼i ) − 0. ∗ δi (δ∼i ) = argmax Ui (δi . δ∼i ) δi (6.4) and the candidate’s utility 156 . fixed costs.4 Best Response The optimal strategy of candidate i. the resource allocation decision is trivial. Pouring more resources into A myriad of real-world considerations are not included in the model such as diminishing returns.6 6. As an intuition for the whether a campaign has an incentive to microtarget or broadcast.methods is necessary to shift the voter’s preference toward the candidate—is assumed not to occur.2. and economies of scale. m ∂Ui (δi . 6 ∂Ui ∂δi > 0 (see Equation 6. If the number of votes that can be shifted with microtargeting (m) is greater than the proportion of votes possibly shifted if the issue resolves (q). given the strategy of the other candidate (labeled ∼ i) ∗ is the level of microtargeting (δi ) that maximizes candidate i’s utility.8) To determine the best response to the opponent’s strategy.5)/σ) ∂δi σ m−q +p φ((µ + m(δi − δ∼i ) + q(2 − δi − δ∼i ) − 0. δ∼i .9) Under some circumstances. consider the effect of broadcasting on the variance of Vi .

campaigns want to increase the variance of the election result if they are losing (i. note that adding or subtracting a value to the point on a Gaussian probability distribution has the property x < 0 ↔ φ(x + c) > φ(x − c). That condition is 157 .9 is rewritten to combine the terms influenced by the potential for microtargeting. the campaign must start in a losing position and not be able to make up this ground with microtargeting alone: µ + m(δi − δ∼i ) < 0. m. ∂Ui (δi .5)/σ)] q [pφ((µ + m(δi − δ∼i ) + q(2 − δi − δ∼i ) − 0.5)/σ)]. For ∂Ui ∂δi to be negative. Next.. see equation 6. increased broadcasting increases the variance of Vi . For a mathematical intuition of this result.2). To prove this result formally. we find an incentive for broadcasting only when a campaign cannot reach 50% (on average) with a combination of predispositions and microtargeting.5)/σ) +pφ((µ + m(δi − δ∼i ) − q(2 − δi − δ∼i ) − 0. Equation 6.5)/σ) σ +pφ((µ + m(δi − δ∼i ) + q(2 − δi − δ∼i ) − 0.10) − −pφ((µ + m(δi − δ∼i ) − q(2 − δi − δ∼i ) − 0. the expected vote percentage is below 50%) and decrease the variance if they are winning (Carraway et al.6 and note that ∂Var(Vi ) ∂δi = −4pq 2 (1 − δi − δ∼i ) is negative.5)/σ) σ (6. δ∼i ) = ∂δi m [(1 − 2p)φ((µ + m(δi − δ∼i ) − 0. Applying this principle to the microtargeting vs. and since issue B resolves stochastically.5.1. so the variance of the vote increases with the amount of resources spent on broadcasting. q. Similar to the stochastic knapsack problem discussed (see discussion D.broadcasting increases the percentage of the population that will switch under a resolved issue.e. 1993).. broadcasting model. and broadcasting.

and qD = qR = q). Figure 6. But even in this case. When the issue does not resolve (which occurs with probability 1−2p). but not sufficient.11 is negative. if campaign ∼ i spends all its resources on microtargeting (δ∼i = 1). If the campaigns start on equal footing (i.e. microtargeting is a weakly dominating strategy. then candidate i can take advantage of that error and win over half the time by following this logic and spending all his money on microtargeting. which is greater than 50%. campaign i loses with probability Φ( m(1−δi ) ). for the optimal campaign strategy to be 100% broadcasting. If the issue σ resolves. The strategy is not strictly dominant because of the case where issue B always resolves (p = 0.2 illustrates the relationship between the probability of issue B resolving and the usefulness of broadcasting relative to microtargeting. the utility of a microtargeting-only strategy and the utility of a broadcasting-only strategy approach the same value: 50%.5). then the half the time the issue resolves in campaign i’s favor nearly balances out with the other half the time it does not.necessary. 158 . Thus. pD = pR = p. the result is a net negative for campaign i since the mean point is negative and x < 0 → Φ(x + c) < (1 − Φ(x − c)). The other necessary condition is that broadcasting must be sufficiently more potent than microtargeting (q >> m) that Equation 6.. then any money campaign i spends on broadcasting will be wasted in the case of issue B not resolving. then the weakly dominating strategy is to spend all resources on microtargeting. The utilities of candidate D for all possible combinations of microtargeting and broadcasting are shown under three scenarios. each with increasing probability of issue resolution. µD = µR = 0 as well as mD = mR = m. If candidate ∼ i foolishly does not spend all his resources on microtargeting. in the case of a campaign in a marginal district with symmetric properties. As the effect of microtargeting shrinks (m) relative to the effectiveness of broadcasting (q).

0 0.1). qR = 0.4 0. δR > 0.42 0.46 0.05) and 25% of the electorate shifts if the issue resolves (q = 0.38 0. In this section. candidate D optimally puts all 159 . three scenarios are considered and depicted in Figure 6. only corner equilibria exist.50 Issue Resolution is a Certainty 0.7. The first scenario is similar to the symmetric case considered in the previous section except that candidate D has an advantage with respect to broadcasting. in the case where candidates are on equal footing.38 0. although uncertainty about the electorate is high (σ = 0.2: Candidate D’s utility under increasing probabilities of issue B resolving.8 1.44 0.8 1.2. More often the issue will resolve in candidate D’s favor (pD = 0.2 0.Low Probability (20%) of Issue Resolution 0.0 Proportion of Candidate D's of resources spent microtargeting 0.2 0. The campaigns start on equal footing (µ = 0.50 0.46 0.46 0.0 Proportion of Candidate D's of resources spent microtargeting 0.2. restrictions on symmetry are relaxed. at high levels of microtargeting by candidate R (approximately.48 Probability that Candidate D wins Probability that Candidate D wins Probability that Candidate D wins 0.5 Edge Equilibria Extending this logic. and p = 0. The panels depict candidate D’s utility under the strategy specified by the x-axis given candidate R spending all resources on microtargeting δR = 1 and increasing probabilities of issue B resolution (p = 0.48 0.6 0.3) and more voters will switch to candidate D in the case of resolution (qD = 0.42 0. these equilibria occur when both campaigns microtarget.38 0.0 0.42 0.4 0.5.6 0. In nearly all circumstances.8 1. respectively).2 0. 5% of the electorate is microtargetable by each campaign (m = 0. In the examples depicted.40 0.08).4. Thus.5).4 0.44 0. p = 0. 6.25).0 0.6 0.40 0.50 High Probability (80%) of Issue Resolution 0.3. pR = 0.8).40 0.44 0.0 Proportion of Candidate D's of resources spent microtargeting Figure 6.1.48 0.

Micortarget 1. Micortarget 0. 0.2 0.0 Cand D Strategy: Pct.8 q 0. R Strategy: Pct. Candidate R Best Resp.8 1.0 (d) Cand R Starting and Iss. Micortarget 1.4 X q q 0.0 Candidate D Best Resp.8 0. R Strategy: Pct. Candidate R Best Resp. 160 .0 0.4 Candidate D Best Resp.0 0.8 q q 0.6 0.2 0.6 0.4 0. Micortarget Figure 6. Cand D Needs Luck to Win 1.4 0.0 0.6 0.6 Candidate D Best Resp. R Strategy: Pct.2 0.8 1. Res. R Strategy: Pct. Micortarget Cand D Strategy: Pct. 0.0 0. 0.6 0.2 0.6 0. No Equilibrium 0.6 0.0 Cand. 0.0 Candidate D Best Resp.8 1. Adv.2 0. Cand. Cand D Issue Res. Candidate R Best Resp. Candidate R Best Resp.3: Best Response Plots of Four Scenarios.2 0. Micortarget Cand.4 0.8 1.0 X Cand.2 0.4 0..2 q q 0.0 Cand D Strategy: Pct. Adv.4 0. Micortarget Cand D Strategy: Pct.0 0.0 0.8 0.0 0.4 0.0 X q (b) No Starting Advantage.0 0.(a) Candidate R's Advantage Holds 1.6 0. Micortarget (c) Cand R has advantage.

the emphasis candidate R places on broadcasting in this case increases the variance 161 . microtargeting provides the former and broadcasting (qD = qR = 0.1. Also. candidate R can increase his probability of winning to 80%. Candidate D’s best response in this case is to fully microtarget. mR = 0.1. When candidate R does not microtarget much (δR < 0. An edge equilibrium of this type is depicted in Figure 6.2) then candidate R can increase the mean vote to fairly close to a 50-50 election (mD = 0. panel (b).445. Hence. candidate R is advantaged both in the electorate’s predisposition (µD = 0. which has three regimes.008).3. An exception to the general rule that advantaged candidates microtarget is presented in panel (c). pD = 0.03).. then campaign i’s best strategy is often to microtarget only. σ = 0.e. qD = 0.067) and issue resolution (qR = 0.resources into broadcasting.54.e. µi + mi − m∼i > 0). In this example.2. the equilibrium is candidate D fully broadcasting and candidate R fully microtargeting.5) that can be maintained when both campaigns microtarget (i.04) and candidate D cannot recover this edge via microtargeting (mD = mR = 0.. R fully microtargets to bring his chance of winning up to 72%.2).044. In this case candidate R has a pre-campaign edge (µR = 0. When candidate i has an advantage in the electorate (i. By combining a microtargeting and a broadcasting strategy. As shown as an “X” in panel (a) of Figure 6. candidate D spends 28% of his money on microtargeting and the rest on broadcasting. pR = 0. σ = 0. In this case. Candidate D wants both a high vote mean and a high variance. the exact mix depends on the relative effectiveness of those two strategies. pD = pR = 0.25) provides the latter. In equilibrium. candidate D has an interesting best response curve.3. The microtargeting strategy is optimal because it lowers the variance of the election result compared to broadcasting.7. The disadvantaged campaign’s best response is often a combination of microtargeting and broadcasting. this pair of strategies is an equilibrium. µi > 0.

162 .of the election. so candidate R has an incentive to microtarget and lower his vote-share variance. non-symmetric case.) However. the candidates never reach equilibrium. In this manner. (The latter strategy is helpful to candidate D’s mean of vote share. Candidate R can broadcast to increase the mean of his vote distribution. there are no internal equilibria.4) then the benefit of raising the mean by candidate D via microtargeting is greater than the benefit of increasing the variance by broadcasting. To prove this formally for the general. he contributes to candidate R’s vote-share mean. candidate D cannot free-ride off candidate R’s full broadcasting effort.2. which also raises candidate D’s probability of winning. mD > 0. the need for candidate D to raise the mean is small enough to encourage broadcasting. Candidate D can broadcast to increase the variance of his vote share. However. Nor are there edge equilibria in which one candidate fully broadcasts. When candidate R microtargets to a substantial degree (approx.6 No Internal Equilibria In the microtargeting-broadcasting game. which is always below 50%. Candidate R is advantaged in electorate predispositions and in issue resolution. This circumstance is interesting because both campaigns may have an incentive to broadcast. In the middle of these two regimes. he contributes to candidate D’s vote-share variance. first I derive the candidate i’s utility and first derivative in this case. No equilibrium exists in the final example. it is never the case that both candidates broadcast at the same time. As candidate D spends more on broadcasting. which is depicted in panel (d). so candidate D has an incentive to microtarget. As candidate D spends more resources on broadcasting only. and it is worth candidate D’s while to broadcast himself. 6.

5)/σ) σ q∼i + p∼i Φ((µ + mi δi − m∼i δ∼i − q∼i (2 − δi − δ∼i ) − 0.13 must be negative. δ∼i ) exists where δi < ∗ 1 ∧ δ∼i < 1. Thus.5)/σ) σ (6. δ∼i ) = ∂δi (6.13) But analogous logic for candidate ∼ i yields the result that negation of the right-hand side of Equation 6. The contribution of microtargeting alone to the candidate’s utility is always positive: limitqi →0 ∂Ui ∂δi > 0.Ui (δi .5)/σ) ∂Ui (δi . Assume that equilibrium of (δi .5)/σ) (6.5)/σ) The proof that no equilibria exist where both candidates spend resources on broad∗ ∗ ∗ casting follows by contradiction. 163 . The negation of a negative cannot be negative. δ∼i )) casting.5)/σ) σ +p∼i Φ((µ + mi δi − m∼i δ∼i − q∼i (2 − δi − δ∼i ) − 0. the first derivative is nonpositive: of the first derivative is negative: ≤ 0. Contradiction. the broadcasting term qi 0 > − pi φ((µi + mi δi − m∼i δ∼i + qi (2 − δi − δ∼i ) − 0.5)/σ) +pi Φ((µi + mi δi − m∼i δ∼i + qi (2 − δi − δ∼i ) − 0. For candidate i’s optimal strategy (δi ∗) to include some broad∂Ui ∗ ∗ ∂δi (δi .5)/σ) σ mi + q∼i + p∼i Φ((µ + mi δi − m∼i δ∼i − q∼i (2 − δi − δ∼i ) − 0.11) mi (1 − pi − p∼i )φ((µ + mi δi − m∼i δ∼i − 0.12) σ m i − qi + pi φ((µi + mi δi − m∼i δ∼i + qi (2 − δi − δ∼i ) − 0. δ∼i ) = (1 − pi − p∼i )Φ((µ + mi δi − m∼i δ∼i − 0.

candidates would have exogenous policy preferences and weights that they would place on winning the election vs.6 holds within an issue: no two candidates would broadcast on the same issue. the model will also inform a candidate’s decision about which issues to focus his efforts on. on average. Incorporating multiple issues enables the model to be extended to platform choice. cannot win a majority of the vote seek to increase both their vote share and the variance of the outcome. although candidates might optimally broadcast on two different issues. Similar to the model of Groseclose (2001). but whether the candidate’s positions on the issues are consistent with his policy preferences. The model can be extended to address multiple issues in a straightforward manner.7 Model Extensions: Multiple Issues and Platform Decisions If more than two issues are at play in an election. early microtargeting might be even more effective than demonstrated at lowering the variance of election results. 6.2. in general. but also to prevent voter defection. The logic of Section 6.2. where k is the number of issues considered. A candidate who is advantaged against his opponent’s optimal strategy seeks to increase the mean of his vote share and decrease the variance.8 Discussion The model has three major implications. Candidates who. as advantaged campaigns would not want to microtarget just to increase their vote share. although the strategy space would increase to k − 1 dimensions. campaigns that are behind microtarget and campaigns that are ahead broadcast. Thus.6. First. This implication may actually be understated by the model. A natural extension of the theory presented in Chapter 2 suggests that delivering microtargeted appeals before issue resolution will prevent some voters from defecting. 164 . The choice of which issues to emphasize (either with microtargeting or broadcasting) would be a function not only of whether the issues can help the candidate win. policy outcome.2.

screwdrivers. 2004). This circumstance was epitomized by the description provided to John McCain of Sarah Palin during his running-mate decision process: “high risk. 6. a losing campaign may be in the situation where it must lower its expected vote share to increase its chance of winning.B. but they are targeting mutually exclusive groups: only voters who agree with the campaign on the issue. Third. Fundamentally. and like most tools (e.g.”7 Since losing campaigns take risks that on average do not work out for them. TNT).3 Normative Implications Microtargeting increases the efficiency of individuals learning about candidates’ positions.1 Heuristics and Judging Democracy A fundamental debate rages in the study of American politics: How much information does a voter need to make an informed decision in the voting booth? On one side of the ledger are those who believe that Americans are ill-informed (Delli Carpini and Keeter. 165 . although current events often foil these plans and force campaigns to talk about the same issues (Sigelman and Buell. Microtargeting has both positive and negative implications for democracy. it can be used for positive or negative purposes. microtargeting is a tool. 7 As publicly stated by A. recounting his conversation with John McCain. The empirical evidence demonstrates that opposing campaigns seek to emphasize distinct sets of issues (Sellers. 6. Whether or not the increased ability of campaigns to microtarget is helpful or detrimental to democracy depends on context and situation.Second. more strategic thinking by campaigns may lead to fewer moderately close elections. campaigns never purposely talk to the same voters about the same issues. high reward.. Culvahouse.3. 1998). Opposing campaigns may microtarget the same issue.

1996). Often the accuracy of political opinion is in dispute. Lau and Redlawsk (2001) attempt to infer the weights that voters should place on various issues to determine a “correct” vote. that they are generally incapable of forming coherent opinions (Converse. Often. The argument from the “more information needed” camp is that there is a standard by which citizens should judge candidates. for this first 166 . 2001. Bartels. The downside is that the outcome of governments (democratic or otherwise) is policy decisions. In the other camp are scholars who believe that cognitive shortcuts are available (Popkin. 2008). The primary standard is the amount of accurate information voters have. lower unemployment). they disagree about how to achieve that outcome. 2001. 2001.g. they will process information similarly to their high-information counterparts—an assumption that is called into question by research on information processing in political situations (Gilens. Bartels. when judging democracy. and citizens’ issue opinions may be farther from those policy positions than their vote decisions. Thus.. and any errors cancel each other in collective public opinion (Page and Shapiro. I use two standards for judging democracy. Caplan. even when political elites agree on the optimal outcome (e. 2005). A safer standard for judging democracy is to evaluate voters’ knowledge of indisputable facts. scholars analyze citizens’ vote preferences (Lau and Redlawsk. Bartels. Bartels (1996) assumes that if low-informed voters become more knowledgable. and that this lack of information is crucial to their decisions (Lau and Redlawsk. 1990). 1996). These assumptions about which candidate preference a voter should have are troublesome because they impose scholars’ views on how citizens choose among political alternatives. voters seek the information that interests them (Krosnick. 2002. 1994). To balance these views. 1964). The debate over heuristics is lively under this normative standard as well (Gilens. 1992). 2001).

vouchers and taxes. the analysis is constrained to the set of information that is verifiable. 12 points. a voter with experience on a political issue is more likely to receive messages about that issue. 42% of seniors (ages 65 and older) could correctly identify both candidates on Social Security. 6. I assume that certain segments of the population have more knowledge about the probability of positive resolution of an issue under the various proposed policy options. good economy or winning a war). Included in this set are candidate positions (which the candidate defines) and observable facts. I examine the more subjective standard of policy options and vote choice. This interaction between dichotomous age and time is significant at conventional levels (logit regression. 10 points. Second. and in October. 8 167 .2 Positives for Democracy Microtargeting increases the efficiency of voters learning the candidates’ positions. Often voters must judge candidates’ platforms before knowing the effect of the policy stances contained within those platforms. but is half the magnitude and not statistically significant. (Optimal is defined as issue resolution that the most people approve of. Microtargeting increases the amount of information at voters’ disposal and is thus a positive for democracy under the first standard.standard. compared to 36% of younger voters: a difference of 6 percentage points. this difference was 8 percentage points. the resultant government’s policy choices are more likely to be optimal. in September. At the beginning of the campaign (through May 2000). In August.g. albeit under certain crucial assumptions. If campaigns microtarget. the interaction effect is in the same direction. p = 0.8 For two other issues tested. e. If this segment of the population has more sway in the election outcome..3.05). An example of this process is seniors learning about Bush’s and Gore’s positions on Social Security (see Chapter 3 for details).

instead. then the candidate with the best policies will be elected. If the campaigns had been able to use the techniques developed just two cycles later.It is unclear how much of this effect was due to microtargeting by the Bush and Gore campaigns. Transitioning to the second standard—correct policy opinion—if experienced voters have more knowledge about which policy options will resolve favorably. but clearly seniors were more actively engaged in the Social Security debate. higher name recognition. for some reason. experienced voters have more knowledge about the “state of the world” on that particular issue than non-experienced voters.g. consider a challenger facing an incumbent who is so incompetent that he is wrong on every issue.9 In a situation similar to the “Swing Voter’s Curse” of Feddersen and Pesendorfer (1996). over a third of seniors could not accurately identify both candidates’ positions on Social Security. This situation would occur when the incumbent’s preferred policies have not been implement or have not yet resolved poorly. The incumbent’s resources (e. Assume that the candidate cannot credibly prove the incompetency of his opponent. the non-experienced voters do not know which policy proposal is best on a particular issue. more funds) If. Even by the end of the campaign. 9 168 .. if the pool of voters who judge politicians on the issue for which they have experience (and thus. However. more in-depth knowledge) is larger than any underlying partisan or incumbency biases. If the challenger does not microtarget and instead broadcasts on one issue to all voters. he can credibly offer proof of his opponent’s issue positions. In game theory parlance. then the analysis works in the opposite direction and microtargeting is detrimental to democracy. As an illustration of why microtargeting helps in this case. experienced voters are less likely to hold the correct policy position than the general public. perhaps more seniors would have learned about the candidates’ positions. then only the small subset of voters with experience on that issue will be affected (unless the issue resolves). then the increased information efficiency provided by microtargeting benefits democracy.

over a quarter of the electorate believed (with at least some degree of certainty) that Obama was a Muslim.3 Negatives for Democracy There are potential downsides to microtargeting as well. Conservative elites (e.10 Another reason why microtargeting might be detrimental to democracy is that people may have experience with issues that affect only a small part of their lives. the 2008 Cooperative Congressional Election Survey (CCES) asked potential voters whether they knew Barack Obama’s religion. This attachment leads to cue-taking (see Chapters 2 and 4). However.g. 10 169 . the challenger will want to microtarget in this case—this strategy will prevent the incumbent from winning again and implementing wrong-headed policies.3. microtargeting may enhance the ability of candidates and parties to have individuals believe incorrect statements. if the challenger uses modern technology to match voters to the issues they have experience with.3. Thus. 2002).might be sufficient to overcome a small block of voters casting ballots for the challenger. Foremost. which means that individuals uncritically accept the position of their preferred elite. delivering congruent cues to partisans increases their attachment to their party’s candidates. As demonstrated in Section 6. Fox News) were peddling this falsehood and the individuals who believed it were nearly all conservative. Despite several statements by Obama that he was Christian and a highly covered debate in the spring of 2008 about controversial statements by Obama’s pastor. 6. For inOf the respondents who were able to place themselves somewhere other than 50 on the 0-100 scale of liberal to conservative. potentially increasing the challenger’s probability of success to greater than 50%. several subsets of voters will judge the incumbent poorly.. As an example of individuals believing verifiably incorrect information. 90% were in the conservative half. they may be more prone to believe false information (Bartels. As shown in the survey experiments in Chapter 4.

.3. Undoubtedly voters sometimes have policy opinions that are incorrect. Hillary Clinton) had voted against the 2002 war resolution. an environmentalist might be microtargeted on global warming and vote for the pro-environment candidate even though she does not support that candidate’s economic policies. John Kerry. a third of the public went from believing the Iraq War was justified to believing it was a mistake from 2003 to 2006. But the voter may nonetheless voter for the suboptimal candidate based on the issue with which she has experience. but in the same distributions.or century-long time frame. The danger is that voters with strong party or elite affiliations may have tighter filters. elected officials’ subsequent policy decisions may lead to negative outcomes. Since global warming works over a decade. a lower percentage of the public would have supported the war.. If more moderate Democrats (e. advancements in campaigns’ targeting abilities may increase cue-taking from elites. Political parties in the United States often include different This question differs from that of whether better intelligence would have changed the minds of the elites.stance. via Zaller’s RAS model). Would a more informed electorate have had a different view of the war in 2003?11 Given that elites on both sides were advocating their respective positions and Americans were filtering their consideration intake (e.g. 6.g. For instance. even by their own standards (with the benefit of hindsight). Perhaps a more attentive electorate would have had more considerations at the tops of their heads. it is not clear that a more attentive electorate would have had a different opinion. Since microtargeting increases polarization and elite affection (see Chapter 2 for the theory and Chapter 4 for the empirical evidence).4 Further Discussion: Party Structure and Alignment The cue-taking effect that causes large segments of the population to believe falsehoods also leads to closer party affiliation. the economy is probably more relevant to the voter. 11 170 .

which also increases political participation (Blais and Dobrzynska.g. economic conservatives). which can produce negative results if the parties abuse the voters’ affiliations by pursuing rent-seeking or ideologically extreme policies that are detrimental to the state of the nation (Sartori. the economy)...4 Conclusion The Personal Experience Model provides a theory explains and can inform campaigns’ microtargeting tactics. On the other hand. voters polarize when affiliating with parties. the more they can control their own destinies. The Personal Experience Model theorizes that many voters have stable opinions on a (potentially small) set of issues. Microtargeting on an issue a voter has a stable opinion about (e.12 6. The Cue-Taking Hypothesis requires that voters’ know the parties’ positions on issues. which is generally unknowable. This artifact of microtargeting has both positive and negative implications for democratic participation. Instead of simply hoping that an 12 Even with increased cue-taking with microtargeting. The more campaigns combine political science research with individuallevel data. thus increasing political participation (Dalton. would determine whether having a small number of big-tent parties is beneficial to voters. 2008). On one hand. this condition holds only when an issue has become heavily politicized. As the 2000 Social Security example shows. If voters had personal experience with every issue would they agree with their party on that issue? The answer to that question. parties are less representative of voters’ stable beliefs. Whether parties that comprise disparate groups are beneficial to voters depends on the ideological alignment of voters and parties. 1998). Converse (1964) finds that most citizens hold ideologically incoherent opinions and that these opinions are highly unstable. The number of parties is kept to a minimum (since parties can better keep factions under one roof). 171 .“types” of voters (e. 1976). voters may have ideologically incoherent issue opinions. social conservatives.g.. abortion) raises her opinion of her party and leads her to more easily accept the party’s platform on other issues (e.g.

as always. Which side has more weight. Microtargeting thus offers both positives and negatives for democracy. a campaign can plan well in advance which voters they will microtarget and when. those easy decisions can lead to voters being too trusting of elites.issue will resolve in the campaign’s favor a week before the election. These voters are also more likely to receive information about candidates’ stances that is relevant to their lives and that therefore makes their voting decisions easier. However. 172 . depends on one’s perspective. Microtargeting is a powerful tool that helps identify the voters who are most likely to change their vote preference as a result of a persuasive appeal.

g.g. Two times in the last decade. Knutson et al. Goren. The field of neuroscience has continued this trend.. Several presidential elections have yielded popular-vote results in which the margin of victory was closer to one percentage point. 2002). 2007).. 2007).. and cue-taking (e.. 1960.. Even small exceptions to the generally-accepted understanding of the role of partisan bi- 173 . exploring how the brains of Democrats and Republicans process information differently (Westen et al. Bartels. Zaller. control of the Senate has been determined in a single election.g. cognitive filters (e.. 1979.g. 2000). Gilens and Murakawa.. 2006.Chapter 7 Conclusion Political scientists spend more time explaining why campaigns do not matter (e. Lodge and Taber. motivated reasoning (e.. Markus and Converse. Gelman and King. research based on these paradigms cannot account explain why voters shift from one party to another—a phenomenon that has a tremendous impact on policy outcomes and the political trajectory of the country. 2006).. 2005) than why they do (Gerber et al. 1993. However. 2002.g. This perspective remains the norm for studies of micro-level behavior. Campbell. Campbell et al. 1992). Thousands of pages have been devoted to partisan bias (e.

and achieving that aim often necessitates persuading voters who would normally vote one way to vote another way. The goal of all campaign managers is to win elections. Fourth. the people whose job it is to counter political biases among the masses. This proximity is cast as the voter taking an active role in the experience. Research that demonstrates the pervasiveness of partisan biases is of little help to these campaign managers. Political science does identify several mechanisms for persuasion.e. in effect demonstrating 174 . a voter might evaluate politicians on what they know about an issue. notably issue publics and self-interest. the Personal Experience Model explains why an individual might join an issue public and why self-interest might not always dominate. a micro-theory formally describes the reasons why voters cue-take from politicians in general but judge these elites on issues when the voters have issue experience. personal experience is introduced as a mechanism that voters use to develop a stable opinion on an issue independent of their political predispositions. the Model is evaluated in terms of its usefulness to political practitioners. They need a way to counter these biases and persuade individuals to alter their predispositions. This dissertation’s primary purpose is to explore one of those exceptions: personal experience. In these cases. First. leads to stable opinions regarding that issue. Third. Personal experience with an issue. The dissertation’s second goal is to bridge the divide between researchers and political practitioners.” instead of uncritically adopting the politician’s opinion (i. whether it is a politicized issue or not.ases are extremely consequential. The proximity of the experience to the individual must be sufficient to circumvent the partisan filters that usually color voters’ perceptions.. This dissertation differs from the issue publics and self-interest threads of the literature in four ways. Second. cue-taking). a process here labeled political “ground truthing.

” The two hypotheses of the Personal Experience model help explain an incongruity in the literature on self-interest. This model formalizes Zaller’s (1992) RAS framework as a learning model. The explanation offered here is: if a voter does not have experience understanding the complexities of an issue. This cue-taking prevents voters from re-evaluating politicians on that issue. voters with unstable opinions on the issue— often individuals who lack personal experience—adopt the position of their preferred elite. 2006) while citing the same topic as the country’s “most important issue. Individuals who have experience with a nonpoliticized issue are likely candidates for the issue public if the issue becomes politicized. 1990). experienced voters who agree with the candidate’s issue positions are low-hanging fruit for persuasion while experiences voters with incongruent opinions are potential defectors. The Personal Experience Model demonstrates how personal experience forms the basis of political judgments. yet this effect is often unimportant in electoral contexts (Sears and Funk. This segment of the electorate is a key target for campaigns. which might explain why large portions of the public can belong to issue publics (Gershkoff.the phrase “seeing is believing. and one-sided also help individuals form stable opinions. Voters can be primed on self-interest in an artificial setting (Chong et al. These voters can ground truth politicians’ stances and alter their perceptions of these political elites. In contrast. 2001). When an issue becomes politicized. voters with experience have knowledge about the now-politicized issue apart from their political predispositions.. consistent in nature.” The experiential characteristics of frequent. The Personal Experience Model provides a richer understanding of voter-issue linkages such as issue publics and self-interest. The Personal Experience Model does not require an issue to be the most salient for an individual to drive vote changes. the voter’s initial opinion matters as she adopts 175 .

self-interest. The field experiments of Chapter 5 demonstrate that applying the lessons of the Personal Experience Model increases campaign efficiency. all of these analyses rely on survey responses to represent actual voting behaviors. Chapter 4 explores the effect of hypothetical candidates taking random stances on issues and finds an increased probability of three percentage points for experienced respondents. The price of these clean tests is external validity. 176 . on average. Chapter 5 examines the effect of campaign literature on voters in real campaigns and finds increased persuasion of about seven percentage points among experienced voters. And none of the analyses randomize the variable of interest. These shifts are particularly important when they counter existing partisan biases and lead to changes in vote choice. Each of these methods has its strengths and weaknesses. personal experience.the position of her preferred candidate. However. hypothetical candidates are a poor substitute for real-life politicians. this dissertation assesses the impact of campaign action on shifts in voter behavior. experienced voters increase their probability of defection by three to four percentage points. the data provide no information on voters’ issue opinions. making it difficult to test alternative hypotheses. 1991) that focuses on voter movement apart from campaign treatment. and political interest. The voter will have a stable opinion and be able to judge the candidates only on issues that offer clear benefits to the voter. The survey experiments of Chapter 4 discern the relative effects of personal experience. and it is difficult to separate out the effects of self-interest from those of personal experience. Also. Chapter 3 quotes the 2000 nomination speeches on Social Security and a Patients’ Bill of Rights. however. The data are noisy. finding that. Unlike recent research on voter persuasion (Hillygus and Shields. The observational data of Chapter 3 are useful because they demonstrate the effects of actual campaign activity during a very salient election.

Viewing this trend over time would most likely require consistent. 177 . This limited use mitigates the broad impact of increased campaign microtargeting and hides its macro effects. As campaigns generate more precise databases of voters. 2000-2008).. campaigns that are inherently disadvantaged and are trailing badly in the polls probably need a bigger boost than the single-digit returns that can be expected from microtargeting. 2006).1 American politics has experienced increased polarization since the advent of microtargeting tools. 1991) and independents are likely to be persuaded to one side or another depending on the issues that arise during a given campaign season. multi-election panel data. However. Campaigns would target voters with experiences congruent to their current party affiliation.thus. it is not always in the campaign’s best interest to segment the electorate and microtarget. Extensive use of microtargeting would probably lead to more voters identifying as strong partisans as well as to increased shifts in party affiliation in the middle of the partisan spectrum. Even with these caveats. 1 Parties might microtarget their own partisans to increase turnout in low-salience elections or to encourage increased participation in party politics. these “cross-pressured partisans” (Hillygus and Shields. this issue emphasis would lead voters to hold their party in even higher esteem. Notably. The Personal Experience Model presents a clear method for targeting voters. but other factors are certainly at play as well (McCarty et al. voters who have experience on an issue and hold a view at odds with their party are likely to defect if microtargeted. which is generally not available. In contrast. causal inference is challenging. Consistent with these two largescale implications is that the proportions of Americans identifying as strong partisans and independents are both above their historical averages (ANES data. the combined weight of the evidence supports both hypotheses of the Personal Experience Model.

If microtargeting dominated macro trends, then strong partisans would be more loyal on Election Day as parties would be able to find a congruent issue position to prevent defection. The data are also consistent with this prediction, as the effect of strong partisans on presidential vote choice is at a historical high (ANES data, 2000-2008, compared with Bartels, 2000). The implication for partisan leaners is less clear as their party affiliation might move in tandem with vote preference changes. These synchronized shifts are most likely during a presidential election, which is also when the most resources are available to microtarget, further reducing the probability of observing a macro trend. Overall, the macro data are consistent with the implications of increased microtargeting, but microtargeting’s contribution to these trends cannot be teased out and probably is small. As for democracy as an institution, the effects of microtargeting are mixed. Voters receive more information that they can use to evaluate politicians, but these targeted appeals might skew their understanding of other, potentially more important, issues. Microtargeting is neither good nor bad; rather, it is a tool that can be wielded for many purposes. The Personal Experience Model and subsequent studies quantify just how powerful a tool microtargeting is and why it works.

178

Appendix A

Appendix for Chapter 2
A.1 Details for Candidate Evaluation Simulation

The following tables display the initial parameter values of the simulation in Section 2.1.5. Voters’ issue opinions, beliefs of candidate stances, and candidate evaluation are updated each day of the hypothetical campaign using the formulae of the Personal Experience Model. Voters’ Prior Issue Opinions: Mean (Precision) Traits Health Care Education 0.95 (10,000) 0.95 (10,000) 0.95 (10,000) 0 (6.25) 0 (6.25) 0.5 (400) 0 (625) 0.5 (100) 0 (625)

Voter Alice the Architect Ted the Teacher Doris the Doctor

2 Table A.1: The mean (δ1 ) and precision (τ1 ), in parentheses, of voter’s prior beliefs on issues and traits.

The voters’ initial issue opinions are displayed in Table A.1. The precision on traits are extremely high because voters are sure that they want high valence (e.g., competent,

179

trustworthy, shared-values) politicians. When occupation and policy intersect (e.g., Ted the teacher and education) the voter’s precision on this issue is higher than otherwise. Voter’s Prior Beliefs about Candidate Mandy’s Positions Traits Health Care Education 0.4 (6.25) 0 (4) 0.3 (4) 0 (6.25) 0 (6.25) 0 (6.25) 0 (6.25) 0 (6.25) 0 (6.25)

Voter Alice the Architect Ted the Teacher Doris the Doctor bottomrule Voter Alice the Architect Ted the Teacher Doris the Doctor

Voter’s Prior Beliefs about Candidate Nathan’s Positions Traits Health Care Education 0 (4) 0.4 (6.25) 0.3 (4) 0 (6.25) 0 (6.25) 0.5 (6.25) 0 (6.25) 0 (6.25) 0 (6.25)

2 Table A.2: The mean (µ1 ) and precision (ν1 ), in parentheses, of voter’s prior beliefs on issues and traits.

Candidate Mandy Nathan

Traits

Candidate Signals Health Care Education -0.5 (1.5) 0.5 (1.5) 0.5 (1.5) -0.5 (1.5)

No signal No signal

2 Table A.3: The mean (γ1 ) and precision (ψ1 ), in parentheses, of the candidate’s signals about where they stand on the issues.

The precisions are low because I simulation that the topic of conversation about each issue lasts 10 days and that one signal is received by the voters each day from their preferred candidate. Alternatively, I could have simulated a higher precision and an additional parameter representing the chance that the voter receives a signal on a day that issue is discussed. (If a voter prefers the candidates equally, there is a 50% chance of receiving the signal from one candidate or the other. At the beginning of the campaign, voters know nothing about the candidates’ issue position and only care about traits. During the second phase of the campaign, a discussion 180

Campaign Phase Phase 1 (day 1) Phase 2 (days 2-10) Phase 3 (days 11-20)

Per-Day Change in Issue Weight Traits Health Care Education +1.0 0 No signal 0 +0.02 -0.02 0 -0.03 +0.03

Table A.4: The weight (δj ) placed on each issue. about health care begins, and voters add considerations about health care to their candidate evaluations. The third and final phase of the campaign sees the conversation shift to education to the detriment of health care. Weights are forced to be nonnegative.

181

Appendix B Appendix for Chapter 3 182 .

7) -0.4) -0. Sec. Dem Conv.4) 0.6 (0.7 (0.3) 0. Bolded coefficients are the coefficients of interest.4 (0.4 (0.4) 0.7) 0.7 (0.3) 0.2) -0.5) 2 (0.4) -0.3) -0.5 (0.7) 0.3 (0.3) -0. 183 .8 (0. Opinion Income Convention Watched Political Interest Party Affiliation Female Constant n Scale 2 Table B.1 (0.4) 2 (0.5) 0.5) 1 (0.2 (0. the cue-taking moderation ability of HMO experience does not extend to Social Security: the main experience coefficient is zero and the interaction term coefficient is negative.4) 0. For issues: Democratic positions is unity.7 (0.2) -2 (0.6) 0. Watch*Opinion Pol.2 (0.5) 0.8 (0.5) 1 (0.5) 1177 Soc.2 (0.5 (0. As mentioned in the text. The scale column indicates how many points are on the scale.5) -0.5) 2 (0. Sec.6) 1213 2 9 5 3 7 2 Variable Experience Experience*Opinion Income*Opinion Conv.B.3 (0.4 (0. Coefficients with standard errors in parentheses are reported.1 (0.3 (0. experience and interaction terms. Conv.8) -0.4 (0.5 (0.1 Regressions for the Cue-Taking Hypothesis Soc.4) 0. Int.7) 0.7) -0.6) -0. 1 (0. All variables are on a zeroto-one scale.6) -0. 0.7) 0.5) 0.4 (0.3) 0.5 (0.9 (0.2 (0.1: Evaluating the Cue-Taking Hypothesis with logistic regressions of postconvention opinion regressed on pre-convention opinion.4) 2 (0.6) -2 (0. -0.4) 1209 PBR Dem. GOp Conv.3 (0.5 (0.*Opinion Party*Opinion Pre-Conv.7 (0. the CueTaking Hypothesis predicts a negative coefficient for experience and a positive coefficient for the experience interaction.1 (0.2) -2 (0.3) 0.

9 (0.B.4) -0.5) (0. Int.7 0.003 Variable Experience Experience*Opinion Income*Opinion Conv. Regression restricted to voters who do not alter their issue opinion during the convention.1 (0.1 (0.3) 1.4) -0.4 (0.6) Soc.1) -1.9 (0.*Opinion Party*Opinion Pre-Conv.4) 0.1) 1. Opinion Income Convention Watched Political Interest Party Affiliation Female n -2. -0. 184 .3 (0.2) 0.2 (0.2) 888 Table B.6) 0.1 (0.2 (0. All variables are on a zero-to-one scale.2 (0. Conv.3) -0.6 (0.2: Evaluating the Candidate Evaluation Hypothesis with ordinal probit regressions of post-convention opinion regressed on pre-convention opinion.4 (0.2 Regressions for the Candidate Evaluation Hypothesis Soc.9) (0.1 (0.3) 0.5 (0.8) -0. Sec.3 (0.1 (0.1 (0.6) 0.4 (0.For issues: Democratic positions is unity.2) 0. Bolded coefficients are the coefficients of interest.9) (0. 0.8 (0.04 (0.5 (0.1) -0.2 (0.5) -0.3) 0.3) 0 (0.5) -0.5 (0.5) 0.2) 0.5) -0.5) -0. -0. the Candidate Evaluation Hypothesis predicts a negative coefficient for experience and a positive coefficient for the experience interaction.5 1.1) 0. Coefficients with standard errors in parentheses are reported.1) 917 PBR Dem.4) -0.2 (0.4) 0.8 (0.3) 0.3) 0.2) 0. GOp Conv.7 (0. Dem Conv.5 (0.5 (0.1 (0. Watch*Opinion Pol.5 (0.5) -0.3 (0.2 (0. Sec.4) 0. experience and interaction terms.2 0.

185 .3) 2. Ordinal probit regressions of voters who do not change their position on Social Security privatization during the convention.4) 220 Age ≥ 65 0.03) -0. so that the coefficients of interest are comparable. the na’s for two of the ordinal cut-offs.02 (0.3) -1.05) -1.05 (0.4) na 135 Table B. Dependent variable is change in vote (Bush.8 (0. The regression is restricted to those between the ages of (50.9 (0.3) 2.46) -0. Age is actual age minus 65.8 (0. No seniors switched their vote more than one scale-point. the cutoff for treatment.3) 1.54 (0.05 (0. Gore.02 (0.Variable Pre-convention Opinion Age Age*opinion DV DV DV DV n = = = = -2 | DV = -1 -1 | DV = 0 0 | DV = 1 1 | DV = 2 Age < 65 0. The coefficient of interest is the effect of support for privatization on change in vote. which is higher in the second regression of seniors.3: Regression discontinuity analysis: Social Security and the Republican convention. or undecided) before and after the Republican convention.45) 0.003 (0. Coefficient of interest bolded.1 (0.04) -0.85)—lack of vote switchers precludes further restriction of the sample closer to the cut point.1 (0.2 (0.05) na -1. hence.

1. 186 . For each.1 Question Wording in Nationwide Survey Candidate Descriptions First. on a second slider. Ratings between 0 and 49 degrees mean that you don’t feel favorable toward the person and that you don’t care too much for that person. move the slider upward toward 10. move the slider downward toward 0. I’d like to get your feelings toward people – some real and some hypothetical.Appendix C Appendix for Chapter 4 C. I’d like you to rate your certainty of that feeling from 0 to 10. please rate that candidate using something we call the feeling thermometer. Ratings between 51 degrees and 100 degrees mean that you feel favorable and warm toward the person. You would rate the person at 50 if you don’t feel particularly warm or cold toward the person. If you are very uncertain about your feeling. • Former Governor Richard Miller has been nominated as the Democratic United States Senate candidate for your state.1 C. If you are perfectly certain about your feeling. Use the slider (with either your mouse or keyboard) to indicate your rating. Then.

Republican leaders. is now 52. He is endorsed by the nurses’ unions and several environmental organizations. he easily won re-election. He is 62.Gov. however. Gov. he was elected Attorney General. Richard Miller grew up in a working-class family. he reduced violent crime in the state and cracked down on corrupt politicians in the state capital. claim he will raise taxes too much. • Attorney General Mark Jones has been nominated as the Republican United States Senate candidate for your state. however. Miller won praise for crafting an innovative health care policy. After rising quickly in the ranks to head that office. graduated top of his class in law school.S. Attorney General Mark Jones grew up in a small town. As Attorney General.1. • Immigrants who entered the country illegally clearly have no respect for the rule of law and must be sent back to their country of origin. is married and has two children. Jones is endorsed by the local farmers’ organization and the Fraternal Order of Police.2 Candidates’ Issue Signals • America has always been a nation of immigrants and today’s hard-working immigrants deserve the chance to achieve the American dream. District Attorney’s office. and began his career in the local U. 187 . during which time he won re-election by a large margin. has three children and five grandchildren. married. He served eight years as the state’s governor. promoting economic growth that outpaced the national average. and working well with state leaders of the opposite party. won a scholarship to a top college. Democratic leaders. C. claim that his policies are too biased toward big business.

• Not only does staying in Iraq provide America security against terrorists. Think about the proportion of important discussion you have with any person that fits the description below. and passing a constitutional amendment affirming that principle ensures that the courts cannot override the will of the people. leaving Iraq will finally force the Iraqi government to take responsibility for their county. military? – About what proportion (percent) of your important discussions is with a person who once served in the U. but America also has a duty to see the Iraq situation through to a peaceful conclusion.S. military? – About what proportion (percent) of your important discussions is with a person who is an immigrant to the United States? – About what proportion (percent) of your important discussions is with a person who is Hispanic? 188 . • We should recognize same-sex marriage so that all Americans are treated equally. C. – About what proportion (percent) of your important discussions is with a person who currently serves in the U. America has long outgrown its history of discrimination and we need to extend marriage rights to all citizens.• Not only is Iraq a distraction from dangerous terrorists in other countries.S. • Marriage has always been between one man and one woman.3 Experience Questions • Next I’m going to ask about the characteristics of these people with whom you discuss important matters. but the U.1.S.

• How often do you attend religious services.– About what proportion (percent) of your important discussions is with a person who is gay or lesbian? • Including yourself. has anyone living in your house or apartment ever served in the U. apart from special events like weddings and funerals? • If you are Christian: Would you say you have been ”born again” or have had a ”born again” experience. that is. do you ever think of yourself as part of a particular church or denomination? Which one? 189 .S. a turning point in your life when you committed yourself to Christ? • Regardless of whether you now attend any religious services. military? This includes the National Guard and the Reserves. Check all that apply.

2 Auxiliary Regressions Issue Iraq War Inter.22) 2.92 (.1) 2.9 (0.8) Self-Interest – -0.9 (0.09 (0.6 (1. Var.15 (.22 (0.4 (0.20) Political Attentiveness 0. meaning the expected sign is positive.7 (1.6 (1.7 (1.20) 0.74) Self-Interest – -1.6 (1.76) 1.C.22) 0.24 (0.28 (0. C.0) With Certainty Mediator Issue Experience 0.3 Issue Experience Measures and the Hypotheses 190 .4) 2.59 (0.0) 0.20) With Controls 0. Signal incongruity is the distance (magnitude) of the candidate’s signal (either -10 or +10) to the respondent’s wave one position on that issue.19 (0.27) 2.8) 1.03 (0.0 (1.0) Issue Certainty 0.7) No Controls 0.20) 0.8) Table C.7 (1. Dependent variable’s potential range is -100 to +100.4 (1.0) 3.15 (.13 (.52) 0. Self-interest is dichotomous.22 (0. they are insignificant for all regressions. Issue Experience Issue Experience Political Attentiveness Self-Interest Immigration Gay Marriage 0. Main effects are not shown.1: OLS regressions of shift in candidate evaluation.15 (.07 (0.74) – -1. N is about 270 for each regression.20) 0. Political attentiveness and issue certainty (from wave one) are measured on a 0 to 1 scale.20) 0. All variables listed are interacted with signal congruity.76) 1.95 (.58) Anti-Partisan Bias Signals Only Issue Experience 0. Top and bottom sets of coefficients represent two different regressions.2) -0.4 (0.20) -0.0 (0.28) 0.9 (0.22) Political Attentiveness 1.09 (0.8) 1.

42) 0.47 (0.2) Without Controls -0. Attentiveness Self interest Constant Immigration Issue Iraq War Gay Marriage -0. Issue Experience Constant Issue Experience From favored cand.02 (0.72 (0.41 (0.7) 0.15) Constant 0. Political attentiveness and issue certainty (from wave 1) are measured on a 0 to 1 scale.10 (0.26) 0.23) 1.35 (1.2 (0.34) -0.93 (0.13) 1.35) 0.63 (0.48 (0. “From favored cand.73) -0.26) – -0.74) Attentiveness -0.4 (1.47 (1.Exp.26) -0.26) Self interest – Certainty 0.” is a dichotomous variable indicating whether the signal is delivered by the candidate that the respondent favored in the first wave.24) 1. 0.36) 0.4 (0.21 (0.73) -0.47 (0.37) 0.34) -0.69) -0.2: OLS regressions of issue opinion shift in the direction of the delivered signal.35) From favored cand.41 (0.35) 0.39 (0.45 (0.89 (1.38 (0.84 (0.39 (0.26 (0.24 (0.94 (0.96) -0.26 (1.34 (0. Self-interest is dichotomous. Var.15) -0.36) With Controls -0.26 (0.54 (0.65 (0.56 (0.3 (1.4 (1.5) Table C.35) -0.74) 0.9) -0. 191 .83) -0. Top and bottom sets of coefficients represent two different regressions.7) 1.42) 0.9) 1.93) Mediator -0.74) -0. N is about 270 for each regression.09 (0.37 (0.5) With Certainty Issue Experience -0. Dependent variable’s potential range is -20 to +20.69) -0.50 (0.25 (0.

1 for regression details.17 (0.2) 0.46 (0.3 (1.2 (0.42 (0.61) -0.16 (.56) 0.39 (. with all respondents who received a signal on that issue included (n is about 270).57 (.63) 0.79 (0.. in workplace (26%) . HH Served(s) Talk to vets/mil (33%) Resp is GLBT Talks to GLBTs (30%) Congruity:Exp. (See Table C. in zipcode (31%) Talk to hisp (32%) Serve(d) in Mil.6 (0. The third column is the same regression as column two but only among respondents who received a signal that went against their partisan predisposition (n is about 135).0 (0. 192 .4) 0.75) 0.3: OLS regression coefficients for various experience measures.14 (0.4) 0. includes control for attentiveness.40 (0.44) -0.59) -.0) 0.. 0.5 (1.. The second column is the regression coefficient (and standard error).3) 0.59) 2.35 (0.) The first column lists the experience measure and the percentage of the population included in the measure if there is not a self-evident boundary.074) 0.5) Against Bias 0.32 (.Group Hispanic .33 (0.69) Table C.16) 0.80) -0..13 (.44) 1.

Group Hispanic ... in workplace (26%) ... in zipcode (31%) Talk to hisp (32%) Serve(d) in Mil. HH Served(s) Talk to vets/mil (33%) Resp is GLBT Talks to GLBTs

Shift in Direction of Signal -0.9 pts (n.s) -1.1 pts (n.q.s) no diff. -1.4 (p < .05) -0.1 (n.s.) -0.6 pts (n.q.s) -0.2 pts (n.s.) -1.1 pts (n.q.s) -1.1 pts (p < 0.1)

Table C.4: Number of points in the direction of the signal (or cue) respondents shifted when compared with individuals not in the group listed. The Cue-Taking hypothesis predicts these values to be negative. When the values are significant at conventional levels, pvalues are listed, n.s. means “not significant,” and n.q.s means “not quite significant at conventional levels.”

193

Thank you very much for taking my survey, which should only take 5 minutes of your time. Your participation is completely voluntary and you may withdraw your consent and discontinue participant of the project at any time. Your refusal to participate will not result in any penalty. This study has been approved by the University's Institutional Review Panel for Human Subjects. If you have questions about the project please email Aaron Strauss (abstraus@princeton.edu). For answers to any questions you may have about your rights as a research subject, contact Joseph Broderick, Secretary, Institutional Review Panel for Human Subjects (609.258.3976). If you complete the survey you will be entered in a random drawing for a $100 gift certificate to amazon.com or iTunes (your choice!).

The survey is on multiple pages; answer the first question below and click the "Next" button below to navigate through the pages. 1. What is your class year at Princeton? 2011 2010 2009 2008 (Freshman).......................................................... 19 (Sophomore) ........................................................ 20 (Junior) ............................................................... 25 (Senior) ............................................................... 36

[Candidates Rotated Below] Below are descriptions of two hypothetical candidates running for United States Senate. For each, please rate that candidate using something we call the feeling scale. Positive ratings between 1 and 10 mean that you feel favorable and warm toward the person, with 10 being the warmest rating. Negative ratings between -1 and -10 mean that you don't feel favorable toward the person and that you don't care too much for that person, with -10 the coolest rating. You would rate the person at 0 if you don't feel particularly warm or cold toward the person.

194

2.

Former Governor Richard Miller has been nominated as the Democratic United States Senate candidate for your state. Gov. Miller grew up in a working-class family, won a scholarship to a top college, is now 52, married and has two children. He served eight years as the state’s governor, during which time he won re-election by a large margin. Gov. Miller won praise for crafting an innovative health care policy, promoting economic growth that outpaced the national average, and working well with state leaders of the opposite party. He is endorsed by the state’s farmers organization as well as the National Organization for Women. 0 +1: 3 -10: most negative................................ ................................................................ -9: ................................ 0 +2: ................................................................ 6 -8: ................................ 0 +3: ................................................................ 6 -7: ................................ 0 +4: ................................................................ 6 -6: ................................ 0 +5: ................................................................ 15 -5: ................................ 0 +6: ................................................................ 11 -4: ................................ 1 +7: ................................................................ 15 -3: ................................ 1 +8: ................................................................ 14 -2: ................................ 1 +9: ................................................................ 7 -1: ................................ 2 7 +10: most positive................................ 7 89 0: neutral ................................ 4 Refused................................ 1

3.

Retired Admiral Mark Jones has been nominated as the Republican United States Senate candidate for your state. Admiral Jones grew up in a small town, graduated from the Naval Academy, and climbed the ranks of the Navy quickly. He is 62, married, has three children and five grandchildren. After retiring from service, he served as Secretary of State of a Republican presidential administration. During his tenure as Secretary, Admiral Jones won accolades for achieving peace in several longstanding conflicts. Admiral Jones is endorsed by several local newspapers and the Fraternal Order of Police. -10: most negative................................ ................................................................ 0 +1: 4 -9: ................................ 0 +2: ................................................................ 11 -8: ................................ 0 +3: ................................................................ 8 -7: ................................ 1 +4: ................................................................ 11 -6: ................................ 0 +5: ................................................................ 11 -5: ................................ 1 +6: ................................................................ 10 -4: ................................ 2 +7: ................................................................ 7 -3: ................................ 3 +8: ................................................................ 6 -2: ................................ 4 +9: ................................................................ 3 -1: ................................ 4 16 +10: most positive................................ 3 75 0: neutral ................................ 9 Refused................................ 0 195

Prefer Democratic Candidate....................................... Prefer Republican Candidate ....................................... Prefer Equally.............................................................. Refused on either evaluation........................................

63 20 15 1

Next, please consider the following political issues. Read the following two statements and again rate them on a scale similar to that of the last two questions. If you agree with Statement A you have ten choices, from "A:1" which means you slightly agree with Statement A all the way to "A:10" which means you completely agree with Statement A. Similarly, if you agree with Statement B pick a value between "B:1" (slightly agree with B) and "B:10" (completely agree with B). Choose "0, Neutral" if you are unsure or agree with the two statements equally. [Issue order rotated. A/B messages rotated] 4. With regard to U.S. foreign aid, do you think the federal government should: A) Cut foreign aid in half because it is ineffective and costs taxpayers over 25 billion dollars a year. B) Double foreign aid since it accounts for less than one percent of the federal budget and helps those in need. A:10, completely B:1, slightly agree w/ A ................................ 2 agree w/ B ................................ 5 A:9................................ 0 B:2................................................................ 7 A:8................................ 1 B:3................................................................ 9 A:7................................ 0 B:4................................................................ 8 A:6................................ 2 B:5................................................................ 6 A:5................................ 2 B:6................................................................ 7 A:4................................ 3 B:7................................................................ 5 A:3................................ 3 B:8................................................................ 4 A:2................................ 3 B:9................................................................ 2 A:1, slightly B:10, completely agree w/ A ................................ 5 21 agree w/ B ................................ 59 6 0: neutral ................................ 20 Refused................................ 0

196

........................................................................... 4 B:8...... 8 A:5.......... 4 A:7............................................................. B) Implement a cap-and-trade system that would restrict the amount of greenhouse gases released by businesses to a government-mandated level.... completely agree w/ A ......... 4 B:3................................................ slightly B:10.. completely B:1.................................................. 2 B:2.............................. 49 7 0: neutral .... slightly agree w/ A ........................................................................ 3 A:8............ 5 A:2................... 6 A:4.......5............... 5 B:7... A:10........................................................................................ 8 Refused.................................................................................................. 5 agree w/ B ............ 7 B:6.......... 2 A:1. do you think the federal government should: A) Encourage businesses to invest in new.......... 5 B:5.......... 4 42 agree w/ B .......................................................................................................... 5 A:9.......................................... 5 B:4.. 6 A:3.................. while letting the market shape the actions of businesses..................................................................... On a different issue... more energy efficient technology.......................................... 2 B:9................. 5 A:6........ 1 197 ........

.......... how many times a month have you attended a spiritual or religious event or service (e............ On average.............................................................................. B) Recognize gay marriage...............................................6...................... 4 36 10 0: neutral ........................................... A state's Supreme Court recently mandated that the state government must legalize same-sex unions................................................... mosque...................................... anything less A:10.......... either through civil unions or gay marriage................................................ meditation) during Fall term? Never ...................... synagogue........................... should the state: A) Implement civil unions--a less divisive solution that provides equal rights to same-sex couples....................................................................................................................... 4 B:2............ church...................... 4 B:9..... would be discriminatory against gays.... B:1.............................................................. 3 A:3........................... 11 A:9...................... 3 A:1... slightly agree w/ B ....... 4 A:2. completely agree w/ B .....g........................................................... 6 Four times a month .. 0 7. 3 B:5.... 5 B:6............ 2 B:7. 3 A:5... 40 Less than once a month ................................................................................. 4 A:4... 2 B:4........................................... 55 25 Refused....................................... slightly agree w/ A ............................................................... 2 A:6............................................ In your opinion........ 23 6 Once a month ....................... 3 B:3...................................... 15 More often than weekly .............................. 4 B:8. 198 ............................................................................... 2 A:7............. completely agree w/ A .................. 1 A:8.................................. 8 1 Refused................................................................. 3 B:10... Twice a month.................................

......................................................................................................................................................................................... 6 Three or four weeks .................................. Africa. ski cabin................... 12 Refused....... 15 Three to size months ..g.................... 1 In the past five years............................................ how many outdoor trips or excursions have you gone on (e............................ Mexico... 6 3 7: .............................................................................................................................................................. eastern Europe)? None .............................................. 10 17 2: ......................................................................................... 15 4: ................................................................ 10.......... 3: ............................................................................................ 199 ..................................................... 10 9 5: .................................................................................................. 6 Two weeks .......... 1 10 or more ............................................................. camping................................................. about how much cumulative time have you spent outside of the U......... South America.......S....... 6: ..... 19 Seven months to a year............... 8 One month.............. lesbian...................... bisexual or transgender? No............................. Over the past twelve months......................................................... 72 Refused.............. 7 Two months............................................ 21 One week ............15 Refused....................................................................................................................................... quiet beach)? 0: ...................................... 28 Yes ....................................................8.............g.. hiking....................................... Asia............ 0 9.................................................................................... and western Europe (e........... Canada.................. 1 Do you have friends or family members who are gay.... 8: ................................................................... 5 More than one year.. 10 1: .. 2 9: ........................

........................ do you usually think of yourself as: Extremely Conservative............................................................................. 12 Weekly.............................................................. an Independent.. 1 12................................................. lean Republican .............................................................................................................. lean to) Republican candidates or Democratic candidates? 5 Strong Republican ........... 21 Liberal ...................................................... 11 Two or three times a week.....................................) Never .................................. 0 13......... 16 Independent.................... 10 Slightly conservative . 1 On occasion .......11....................... Generally speaking............................. Not so strong Republican........................... 11 Moderate. a Democrat.............................................................. 24 Not so strong Democrat.......................... 17 Slightly liberal ......................... how often do you communicate with these individuals? (Feel free to check multiple frequencies below if they are applicable to different friends and/or family........................................................ 6 Other party ......................... do you tend to favor (i........................... middle of the road...................................................................... 0 Refused......... 16 Monthly................................... 2 Conservative ............................................................................................ If so........... 21 Answered “No” in previous question.......................................................... 10 Independent ..... 11 Daily..... do you think of yourself as a Republican...... 14 Other party ................. 28 Refused in either question..... 7 Independent..........................e............................................... or another party? And how strong is your affiliation? If you are an Independent........ lean Democrat ............................................................................... 0 200 .............. When it comes to politics............................................................. 0 Refused.......... 32 Extremely liberal ................................................................................................... 24 Strong Democrat..................................................

.................................. 0 201 ...................................................................................................................... 19: ......... 45 Refused.......................... 28 27 19 18 6 1 16...........................................................................20 Some........ Compared to other Princeton students................................ 1 Please Enter your age in the box to the right 17-18:.................................................................................................................... 22-24:......................................14....................................................................... 39 Very little ...................... 7 Quite a bit .. Please indicate your sex: Female............................................................................. 6 Refused...................................................................................... 15................... 21: ...................................................................... how much to do you pay attention to political news? A great deal................................... 54 Male ........................................................................ Refused..................................................................................... 28 Not at all ..... 20: .........................................................................................

] Signals: Gay Marriage • Marriage has always been between one man and one woman. America has long outgrown its era of discrimination and we need to extend marriage rights to all citizens. we need a cap-and-trade system that will restrict the amount of dangerous greenhouse gasses released into the atmosphere. without tight restrictions. Civil unions provide the same rights to same-sex couples without alienating the more traditional segments of the population. Foreign Aid • America has the world’s largest economy by far. • We should recognize same-sex marriage so that all Americans are treated equally.Wave Two Candidate Signals [Respondent receives two signals on distinct issues. Innovative. • Foreign aid is wasted on corrupt foreign governments. Businesses will always try to cut costs. eco-friendly businesses will prosper. 202 . Environment • To combat global warming. leading to a healthy environment and continued job growth. We should spent taxpayers money in America. • A market-based solution to climate change is the best way to ensure the health of the planet while maintaining a strong economy. where we can better evaluate its effectiveness. and as global leaders we have the responsibility to assist those less fortunate than us. our environment will continue to suffer.

Appendix D Appendix for Chapter 5 D. the exact solution of this linear programming problem is approximated by ordering the individual pairs by their maximum vote per dollar ratio and 203 . D.3. Following Dantzig (1957). In the knapsack problem.1.1 Nonpartisan Case: the Knapsack Problem To approximate the solution to the nonpartisan planner’s optimization problem defined in Section 5. with each object having its own value and size.1. one maximizes the total value of objects to be placed in a knapsack of fixed sized. the key is to notice that the linear optimization problem is identical to the canonical knapsack problem. The analogous case for the nonpartisan planner is maximizing the number of voters given a budget constraint where each individual-treatment pairing may be thought of as an object.1 Computational Appendices This appendix presents fast and approximate solutions to the planner’s optimization problem in both nonpartisan and partisan cases.

this individual is not treated. As an approximate solution to this problem. only the standard errors matter).3. this approximation yields solutions very close to the optimal result because the ratio of the per-use cost of the most expensive treatment (e. an algorithm based on Geoffrion (1967) is used. In most cases.. in which the probability is maximized that the total value of items in the knapsack equals or exceeds a target value where each object has a random value and a known size. If the ratio is non-positive (i. (For a discussion of when this approximation fails to yield the optimal result..000). Optimization is performed over the weight parameter.e.e. the key is to notice that this optimization problem is identical to the stochastic knapsack problem. each individual-treatment pair can be considered an item... As in the nonpartisan case. D. inefficiencies at the edge of the problem are of little importance.. Thus. only the means of the posteriors matter) and 0 (i. subgroups are ordered by the weighted combination of the mean and standard error of their posterior vote choice profile. $15 for a canvassing shift) is tiny compared to the overall budget (usually at least $10. π(ρ). when the addition of an expensive and efficient treatment runs just over budget and a cheaper yet less efficient tactic should be used in its place. The intuition behind this algorithm can be developed by considering the following scenarios. they would garner a majority of the vote without treatment) could further increase their probability of winning by contacting voters who are highly responsive on average 204 .e. see Henig (1990)).2 Partisan Case: the Stochastic Knapsack Problem To derive a fast and approximate solution to the partisan’s optimization problem defined in Section 5.4.treating the individuals with the highest such ratio first until the budget is exhausted. Campaigns with a natural advantage (i. which can take values between 1 (i.g.1.e. the best non-control treatment for an individual does not outperform the control).

the optimal subgroups to treat change depending on the outcome under the control. Thus. The algorithm finds an approximate solution by limiting its search to the subspace defined by the weight parameter. 205 . unlike in the nonpartisan case.and have a low variance of their treatment response. campaigns that are behind aim to treat segments of the population that are both highly responsive and have high variance. which makes optimization feasible when the dimension of δ is large. On the other hand.

4) 206 . then Var(X) = Var(E[X|Y ]) + E[Var(X|Y )] ¯ ¯ Var(E[X|Y ]) = p(µ1 − X)2 + (1 − p)(µ2 − X)2 2 2 E[Var(X|Y )] = pσ1 + (1 − p)σ2 2 2 ¯ ¯ Var(X) = p[(µ1 − X)2 + σ1 ] + (1 − p)[(µ2 − X)2 + σ2 ].1 Analytical Solution for Variance of Vote Share The general formula for the variance of a random variable X that is composed of two Gaussian distributions that occur with probability p and 1 − p respectively is derived. If the two normal distributions that compose X are N (µ1 . (E. σ1 ) 2 and N (µ2 . 0. let Y be a Bernoulli random variable that determines 2 which Gaussian produces x. σs ). To determine the variance of X.Appendix E Appendix for Chapter 6 E. 0. Figure E.16) and N (−1.1 shows the distribution with p = 0.16).1) (E. For an example.2) (E.3) (E.5 and sub-distributions of N (1.

4 0.Probability Density 0. in proportion to the probability that the distribution is activated by the Bernoulli probability.5 −2 −1 0 1 2 3 Hypothetical Random Variable X Figure E.3 0. as both the Democratic and Republican broadcasted issues have independent probabilities of either resolving or not resolving.0 −3 0. V¯ (δD . The mean of the distribution is.2 0.1 0.1: Hypothetical distribution of a variable composed of two Gaussian distributions with a Bernoulli process arbiter. Each sub-distribution contributes its own variance and the squared distance to the mean to the overall variance. δR ) = µ + m(δD − δR ) D 207 . ¯ where X = pµ1 + (1 − p)µ2 . There are four distributions because there are four possible outcomes. This derivation is important because the distribution of the vote function is the superposition of four normal curves. each of which occurs with a certain probability.

δR )) = pD (σ 2 + (q(2 − δD − δR ))2 ) +pR (σ 2 + (q(2 − δD − δR ))2 ) +(1 − pD − pR )(σ 2 ) combining terms and assuming pD = pR = p = σ 2 + 2pq 2 (2 − δD − δR )2 208 .The variance of Vi is a weighted average of the squared distance of the four normal ¯ distributions that constitute Vi to Vi . Using the derivation above. Var(VD (δD . the variance is.

.Bibliography Achen.. Achen. M. (2006). M. Iyengar. C. K. The strength of issues: Using multiple measures to gauge preference stability. S. It feels like we’re thinking: The rationalizing voter and electoral democracy. J. ideological constraint. 829–838. Partisan hearts and gall bladders: Retrospection and realignment in the wake of the great depression. and issue voting. J. S. H. Arceneaux. The American Political Science Review 87. Bartels.. M. Rodden. 215–232. L. (2008). (1994). 4. L. The American Political Science Review 102. 2. and Bartels. 2. L. and Bartels. In Annual Meeting of the American Political Science Association.. Philadelphia. A. N. (2009). Does attack advertising demobilize the electorate? The American Political Science Review 88. R. Messages received: The political impact of media exposure. Simon. 209 . (2005). (1993). and Valentino. 4. and Snyder. American Journal of Political Science 53. S.. Ansolabehere. 267–285. and Kolodny. Ansolabehere. C. Presented at 2005 the Annual Meeting of the Midwest Political Science Association . Educating the least informed: Group endorsements in a grassroots campaign.

Campbell. 239–262. A. Chapman & Hall. Miller. (2005). Campbell. American Journal of Political Science 44. 35–50. Political Behavior 24. E. and Stokes. 194–230. A. Public Opinion & Parties 15.Bartels. L. Bartels. Classification and Regression Trees. Brody. Turnout in electoral democracies. the economy and incumbency in the 2004 presidential election. Partisanship and voting behavior. (2005). Springer. 1. 1. C... Uninformed votes: Information effects in presidential elections. Breiman. (1985). 1952–1996. M. European Journal of Political Research 33. M. (1960). Comment: The assessment of policy voting. Berger. American Journal of Political Science 40. 2. 2nd edn. New York. J. J. L. Blais. The fundamentals in us presidential elections: Public opinion. R.. (2002). P. W. 2. 15–31. 117–150. Journal of Elections. Homer gets a tax cut: Inequality and public policy in the american mind. M. 1. 1. L. Stone.. Friedman.. I. J. and Page. New York. (1972). (2000). (1984). Statistical Decision Theory and Bayesian Analysis. D.. John Wiley & Sons Inc. L. A. 2.. 450–458. The American Voter. The American Political Science Review 66. R. 210 . Beyond the running tally: Partisan bias in political perceptions. Perspectives on Politics 3. and Olshen. L. Converse. B. London. J. 73–83. (1996). and Dobrzynska. O. (1998). Bartels. A. Bartels. M.

Apter. and Shotts. 3. J. D. J. Cohen. (2008). B. Free Press. J. C. (2008). 75–169. R. Caplan. Department of Statistics. 161–173. The Myth of the Rational Voter: Why Democracies Choose Bad Policies. Carraway. NJ. Available at http://arxiv. K. E. E. E. (1993). Citrin. and Weatherford... R. Princeton University Press. R. New York. Tech. Chipman. 1267–1282. New York. E.. D. E. L. R. John Wiley & Sons. An algorithm for maximizing target achievement in the stochastic knapsack problem with normal returns. Princeton. and King. 211 . Relative unemployment and gubernatorial popularity. P. J.Campell. Princeton. B. and Stimson. H. and McCulloch. Leadership and pandering: A theory of executive policymaking.. 2. Naval Research Logistics 40. J. Planning of Experiments. ed.. R. E. E. Issue Evolution. The Journal of Politics 66. Forecasting the presidential vote in the states. (1964).. 386–407. George. 3. 2. (2004). The nature of belief systems in mass publics. 541–570.. University of Pennsylvania. rep. A..3286v1. Chong. BART: Bayesian additive regression trees. American Journal of Political Science 36. 4.. D. American Journal of Political Science 45. (1990). (2001). NJ. Political Psychology 22. Canes-Wrone. Cox.. In D. L. W.org/abs/0806. When self–interest matters. M. I. (2001). (1958). P. L. Converse. (1992). Carmines. Herron. 532–550. Schmidt. and Conley. Ideology and Discontent. Princeton Univertsity Press.

A. 266–277. M. NJ. J. The Spatial Theory of Voting : An Introduction. Panagopoulos. Experimental propaganda techniques and voting behavior. Discrete–variable extremum problems. American Journal of Political Science 53. Eldersveld. B. Comparative Political Studies 41.. What Americans Know about Politics and Why It Matters. Delli Carpini. Rutgers University Press. J. Enelow. and Keeter. ed. M. New Brunswick. and its consequences. 4. A. Operations Research 5. (1996). On the predictive validity of attitudes: The roles of direct experience and confidence. (2008). Downs. 2. P. The quantity and the quality of party systems: Party system polarization. 1. R. S. Dehejia. Dantzig. Yale University Press. (1956). CT. Harper & Row. Journal of Personality 46. (2009). and Hinich. and Zanna. Mobilizing the mobiles: How text messaging can boost turnout. and Strauss.Dale. H. R. 7. (1957). A. R. Politicking Online: The Transformation of Election Campaign Communications. Dalton. Don’t forget to vote: Text message reminders as a mobilization tool. J. (1984). (2005). Journal of Econometrics 125. American Political Science Review 50. 899–920. 141–173. (1978). An Economic Theory of Democracy. its measurement. Cambridge University Press. Dale. S. M. and Strauss. 228–243. A. New York. Fazio. In C. New York. (1957). (2008). Program evaluation as a decision problem. 154–165. 212 . J. 2. M. A. G. New Haven.

409–451. (1981). Operations Research 15. T. Yale University Press. (1996). M. D. Stanford. J. Stanford University Press. P. Needham Heights. and Pesendorfer. Gelman. Direct experience and attitude-behavior consistency. J. 1. M. Divided Government. The influence of television and radio advertising on candidate evaluations: Results from a large scale randomized experiment. The swing voter’s curse. and King. Why are american presidential election campaign polls so variable when votes are so predictable? British Journal of Political Science 23. R. P. 213 . (1926). (1986). Retrospective Voting in American National Elections. Advances in Experimental Social Psychology 14. (1981). The arrangement of field experiments. New Haven. Feddersen. Fiorina. 2nd ed. G. Public Choice 50. 1. H. 408–424. CT.. G. Gerber. (2007). A. 5–25. Solving bicriterion mathematical programs. A. P. Allyn and Bacon. A. 161–202. Incumbent performance and electoral control.Fazio. Gimpel. R. J. (1957). 4. P. MA. 39–54.. D. L. 503–513. Unpublished manuscript . Green. (1996). The American Economic Review 86.. M. Fiorina. M. CA. Festinger. (1993). R. A theory of cognitive dissonance. (1967). Journal of the Ministry of Agriculture 33. Geoffrion. W. 3. and Shaw. and Zanna. Fisher. Ferejohn.

Rational learning and partisan attitudes. Does campaign spending work?: Field experiments provide evidence and suggest new theory. American Political Science Review 95. Getting–Out–the–Vote: An experiment in the stimulation of voting. M. 3. 541–574. Gerber.. A. P. H. and Participation (Vol. M. A. Princeton University. Delli Carpini. D. P. (2004). In M. P. Shapiro. D. P. Doctoral Disseration. A. eds. (1998). 2. and Murakawa. and Green. and Shachar. 15–49. 3. Green. Huddy. Goren. Gershkoff. New York. R. American Behavioral Scientist 47. Gilens. 6). Department of Politics . How issue interest can resue the american public. (2007). 794–818. American Journal of Political Science 42. Voting may be habit–forming: Evidence from a randomized field experiment. A.. 540–550.. F. and R. and Green. S. telephone calls. 305–325. Character weakness. Gerber. Research in Micropolitics: Political Decision–Making. S. Elite cues and political decision–making. 653–663. Chicago. (2002). Political Behavior 29. JAI Press. Gerber. S. D. (2003). and direct mail on voter turnout: A field experiment. A. Y. L. (2006). (1927). (2001). and presidential evaluation: Modiffcations and extensions.Gerber. 5. 214 . 379–396. Gilens. (2000). partisan bias. Political ignorance and collective policy preferences. The effects of canvassing. X. American Journal of Political Science 47. R. N. University of Chicago Press. American Political Science Review 94. Gosnell. Deliberation.

A. J. 5. A model of candidate location when one candidate has a valence advantage. (1991). D. (2007). 820–825. 215 .. G. D. 1. D. 4. Get Out the Vote: How to Increase Voter Turnout. 149–180. Groseclose. Public Choice 104. W. S. Hibbs Jr. (2008). University of Chicago. Journal of Abnormal Psychology 31. and Shields. Hartmann. Graduate School of Business. (1951). M.. Hake. 2nd edn. Bread and peace voting in u. 149–154. Zhu. Tech. and Murphy. Princeton. Journal of Experimental Psychology 41. (1936). J. 862–886. Poceedings of the 11th Conference on Artificial Intelligence in Medicine AIME 2007.. W. 99–114. Acquisition and extinction of a verbal conditioned response with differing percentages of reinforcement. I. Hill. S. Bayesian nonparametric modeling for causal inference. Brookings Institution.C. LNCS/LNAI 4594. S. D.. J. L.Grant. E. and McCulloch. Washington D. P. Risk criteria in a stochastic knapsack problem. (2008). The Persuadable Voter: Wedge Issues in Presidential Campaigns. 1. R.s. NJ. G.. Gunter. Henig. and Gerber. T. (2001). rep. Hillygus. Operations Research 38.. T. Princeton University Press. A field experiment on the comparative effectiveness of “emotional” and “rational” political leaflets in determining election results. and Hornseth. (2000).. presidential elections. Green. L. Variable selection for optimal decision making. 1. A. (1990). P. American Journal of Political Science 45. A. H.

945–960. Knutson. Planning the optimal get–out–the–vote campaign using randomized field experiments. Journal of the American Statistical Association 81. A decision–theoretic analysis of a problem in political campaigning. L. A. Political Behavior 28. D. Princeton University. K.princeton. Southern Methodist University.html. S. Statistics and causal inference. Priming presidential approval: The conditionality of issue effects. Harvard University Press. 1. Political Behavior 12. (1990). K.edu/research/choice. Government policy and citizen passion: A study of issue publics in contemporary america. N. V. NJ. 3. A. (1968). J.. (1929).. and Grafman. 153. 1. Key. Social neuroscience 1. IL. Spampinato. V. C. (1989). available at http://imai. Kramer. Imai. J. Politics on the brain: An fmri investigation. University Of Chicago Press. and Wolak. (2006). V. 137–160. Hotelling. News that matters : Television and American opinion. Bernd. (1966).. (2009). The Responsible Electorate: Rationality in Presidential Voting. Princeton University Press. Hutchings.. H. Mathematical Applications in Political Science. Chicago. J. Stability in competition. Krosnick. II. 216 . (2006). M. Public Opinion and Democratic Accountability: How Citizens Learn about Politics. M. 41–57. (1986). Tech. The Economic Journal 39. rep. ed. L..Holland. In J. Department of Politics. 1936–1960. Princeton. 25–40. Dallas. Kelleher. P. 193–210. 59–92. and Strauss. Iyengar. Wood. and Kinder. W. A. G. TX. Vol. J. (2003).

The American Political Science Review 88. (1994). D. S. 1. R. R. Choice. and Lepper. P. Decomposing the sources of incumbency advantage in the u. M. P. What Politics is About. K. C. thesis. (2008).. C. M. B.. D. (2001). Lupia. Lupia. McGraw. Popkin. 63–76. Lodge. The Democratic Dilemma: Can Citizens Learn What They Need To Know? Cambridge University Press. M. McCubbins. Levitt. New York. and Taber. Presented in the Princeton American Politics Graduate Research Seminar . Ross. 4. Cambridge. (1989). 45–60. A. Advantages and disadvantages of cognitive heuristics in political decision making. (1998). 399–419. Cambridge University Press. G. and Stroh. Lord. 2. Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence.Lau. Lenz. The American Political Science Review 83. Shortcuts versus encyclopedias: Information and voting behavior in california insurance reform elections. and S. G.D. D. 951–971. D. MA. and Redlawsk. Lodge. M. An impression–driven model of candidate evaluation. 11. house. Journal of Personality and Social Psychology 37. A. C. and McCubbins. 183–213. In A. American Journal of Political Science 45. M. (2000). (1997). (2006).. Legislative Studies Quarterly 22. Ph. Lupia. Princeton University.. (1979). Lauderdale. Three steps toward a theory of motivated political reasoning. Bayesian social learning: Theoretical implications for the measurement of projection phenomena. 217 . L. R. and Wolfram. Elements of Reason: Cognition. and the Bounds of Rationality. eds. M. s. 1. 2098–2109..

Mueller. C. 2. N. 483–508. MA. D. B. Markus. H. Princeton University Press. 179–199. N. (1979). In Annual Meeting of the Midwest Political Science Association. Poole. P. Manski. Campaigns & Elections. H. (1970). R. E. A dynamic simultaneous equation model of electoral choice. R. Nickerson. The American Political Science Review 64. G. McFadyen. P. C. (1992). Mitchella. 4. Social Choice with Partial Knowledge of Treatment Response. (2003). E. Washington. T. Polarized America: The Dance of Ideology and Unequal Riches. J. (2005).. J. (2006). American Journal of Political Science 36. Emotions as moderators of information cue use. Mass media and the depolitization of personal experience. 2. Nyea. (1998). and Isaacsa. The New Political Targeting.. E. M. D. K. MIT Press. The American Political Science Review 73. E. and Kuo. (2006). 2. K. 18–34. Human Relations 51. G. Mutz. McCarty. (2009). Malhotra.. and Arceneaux. Who is mobilized to vote? a meta–analysis of 7 field experiments. Robinsona. F. J. 1055–1070. G. 218 .C. and Rosenthal. Journal of Experimental Child Psychology 64.. (1997).Malchow. American Politics Research 37. A. Presidential popularity from truman to johnson. D. Cambridge. When speech conflicts with seeing: Young children’s understanding of informational priority. 1. and Converse. Attitudes toward the unemployed. 276–294. 301–326..

Assmann. American Journal of Political Science 40. 1. and Fazio. L. (2007). The Rational Public: Fifty Years of Trends in American’s Policy Preferences. E. (1994). Chicago. Attitude and Persuasion: Classic and Contemporary Approaches. 1. On the consistency between attitudes and behavior: Look to the method of attitude formation. (1977). S. Issue ownership in presidential elections. Pocock. Regan. J. S. University Of Chicago Press. 219 . 2. Enos. IA. and Cacioppo. Journal of Experimental Social Psychology 13. S. E. R. Petrocik. Wm. Page. W. Parker-Stephen. IL. 825–850. E. E. (2008). Is voting contagious?: Evidence from two field experiments. 2917–2930. T. (1996). B. with a 1980 case study. Nickerson. Quality is job one: Professional and volunteer voter mobilization calls. J. D. Brown. (1992). 49–57. E. H. (1981).. Statistics in Medicine 21. American Political Science Review 102. University of Chicago Press. Dubuque. In Annual Meeting of the Midwest Political Science Association. Popkin. D. Petty. W. American Journal of Political Science 51. (2002).. R. R. 28–45. When perceptions polarize: How motives and information shape partisan inference. (2007). and Kasten. The Reasoning Voter: Communication and Persuasion in Presidential Campaigns. covariate adjustment and baseline comparisons in clinical trial reporting: current practice and problems.Nickerson. T. J. C. Subgroup analysis.. Chicago. D. 269–282. and Shapiro. L. R. 3.

P. (1978). Cambridge University Press. Bayesian inference for causal effects: The role of randomization. B. Hagen. Statistical Science 5. Mansbridge. Uncertainty and turnout. The effects of repetition of refutational and supportive advertising appeals. 253. Sawyer. G. and Funk. Journal of Marketing Research 10. (1951). (1990). L. Self–interest in americans’ political opinions. M. Sartori. Dabrowska and T. O. S. G. 34–58. (2004). 23–33. The psychology of Social Norms. Harper and Brothers. Sherif. Rubin. Beyond self–interest. J. ed. Comments on “On the application of probability theory to agricultural experiments. (1936). (1990). J. Pattern Recognition and Neural Networks. Ripley. D. (1996). Savage. Sellers. J. In J. 147–170. J. D. B. The theory of statistical decision. Journal of the American Statistical Association 46. (1998).. 472–480. Strategy and background in congressional campaigns. Chicago. Cambridge University Press. New York. Splawa–Neyman translated from the Polish and edited by D. 1. The 2000 Presidential Election and the Foundations of Party Politics. Speed. 1. University Of Chicago Press. The American Political Science Review 92. Sanders. Parties and party systems. L. H. Sears. 159–171. Cambridge University Press. D. Political Analysis 9. Essay on principles. 1. 45–57. (1976). 220 . A. D. (2000). Cambridge. P. M. C. (1973). Rubin. M. Section 9” by J. B. New York. Michael G. K. 55–67. 1.Richard Johnston. Annals of Statistics 6.

Journal of Educational Psychology 51. (2009). 755–799. Effects of media use on complexity and extremity of attitudes toward the death penalty and prisoners’ rehabilitation. M.. and Religious Leaders Connect with the New American Community. 1–24. Minimax regret treatment choice with finite samples.Sherman.. 1. Sotirovic. D. B. L. (2006). Media Psychology 3.. M. Applebees America: How Successful Political. J. van Knippenberg. Presson. Pers Soc Psychol Bull 8. M.. and Lodge. (2004).s. D. and Fournier. Personality and Social Psychology Bulletin 17. presidential campaigns. R. (1960). (1991). and Buell. D. Emmett H. Stoye. C. Providing information in public opinion surveys: Motivation and ability effects in the information–and–choice questionnaire. Corty. E. Sosnick. 3. Journal of Econometrics Forthcoming. (1982). J. C. 650–661. S. M. (2001). 188–93. Regression discontinuity analysis: An alternative to the ex–post–factor experiment. New York. and Olshavsky. 221 . Verplanken. Bensenberg. and Daamen.. American Journal of Political Sciencey 50. W. 4. Chassin. International Journal of Public Opinion Research 8. 2.. Dowd. (1996). D. Motivated skepticism in the evaluation of political beliefs. 1. 2. Simon & Schuster. 1960–2000.. J. Smoking Intentions in Adolescents: Direct Experience and Predictability.. Thistlewaite. Taber. Persuasive communication of risk information: A test of cue versus message processing effects in a field experiment. J. Sigelman. S. 309–317. 376–383. D. Avoidance or engagement? issue convergence in u. and Campbell. American Journal of Political Science 48. C. B. (2006). Business. R. 70–82. L.

Statistical Decision Functions. Wood. (1950). 1947–1958. Public Affairs. Cambridge. A. New York. Weston. American Economic Review 98.. Personality & social psychology bulletin 42. Cambridge University Press. 11. W. (1982). S. 5. S. Zaller. (2006). (1992). 311–332. Female socialization: How daughters affect their legislator fathers’ voting on women’s issues. (2008).. J.. D. 798–810. Harenski. Retrieval of attitude–relevant information from memory: Effects on susceptibility to persuasion and on intrinsic motivation. MA. E. Blagov. K. Washington. Neural bases of motivated reasoning: An fmri study of emotional constraints on partisan political judgment in the 2004 u. The Nature and Origins of Mass Opinion.s. Wiley. New York. 222 . R. P. D. The Political Brain: The Role of Emotion in Deciding the Fate of the Nation. Kilts. presidential election. Journal of Cognitive Neuroscience 18. (2007). Westen..Wald. C. and Hamann.