The Effects of Partisanship on American Voter Turnout

Ashley G. Chico

Dr. Kenecia Wright

POS3703 – Scope and Methods

27 April 2020
 Chico 2

I. INTRODUCTION

Partisanship has historically been considered to not affect American voting behavior or

election results. However, since the resurgence of American political polarization, partisanship

has become a contributing factor to how voters vote from varying levels of elections. The

research question of this study aims to answer: to what extent does partisanship affect voter

turnout? While existing literature analyzes contributing factors that are more likely affecting

partisanship and voting behavior such as ethnicity, socio-economic level and geography, this

study highlights alternative variables from the literature: party identification, approval of

President handling Office, age and occupation. The ANES 2016 Time Series Study was used in

the STATA software to complete all tests and models for analysis.

For the following study, logit and correlation tests were completed in order to examine

the strength and types of relationship between each variable against the dependent variable. The

dependent variable of whether the respondent voted or not, “Covariate”, will be cross-examined

with any variable that is statistically significant. As a result, this will show which variables

caused voters to vote or not vote with age and occupation as control variables. The argument of

this study is if there are high levels of partisanship and elite behavior, voters are less likely to

vote in election. By analyzing these key variables, a possible explanation for contributing factors

to whether American voters decide to vote or not on the grounds of partisanship can be

concluded.

II. LITERATURE REVIEW

The research on voting behavior on the local and national level in political science has

been both extensive and present for decades. The correlation between voting behavior and

election results have particularly become of interest to scholars in the field of political science in
 Chico 3

hopes of predicting candidate results. Some factors have repeatedly been included in these

studies, of which most scholars attest hold quantifiable and significant values of influence in the

data. These traditionally tested variables include the various levels of socioeconomic status and

age ranges of voters (Hassel, 2017; Hetherington, 2001). However, additional factors have

caused a long-standing debate among scholars when determining the variables that repeatedly

and significantly influence election results. One of the factors that have been limited in testing is

that of partisanship. There have been few studies that have fully tested the extent to which

partisanship affects election results under varied voting circumstances. As a result, the current

literature lacks a strong consensus for the degree in which partisanship takes in voting behavior

and the electorate (Hassel, 2017; Warshaw, 2019). Therefore, there are several differences in

research opinions on the correlation between varying levels of socio-economic levels, age, and

partisanship.

First, the issue of the legitimacy of partisanship in effecting the electorate must be

addressed to better understand the available data. The understood perception of partisanship

concerning affecting voting behavior and the electorate is little to none by scholars (Bafumi and

Shapiro, 2009; Hetherington, 2001). Historically, partisan dealignment was apparent during the

mid-twentieth century after the New Deal had been in effect in America (Bafumi and Shapiro,

2009; Hansford and Gomez, 2010). The call for unification among Americans caused political

parties to decline in activism and diverging in ideology (Bafumi and Shapiro, 2009). What is

striking about the studies discussing the Post-New Deal effect on partisanship among Americans

is the open dialogue of possibilities explaining the cause of the resurgence of partisanship. One

view claims that the cause of the civil rights movement was responsible for an increase in

American partisanship (Bafumi and Shapiro, 2009). Others suspected that the Vietnam war
 Chico 4

protests and new ideologies between left and right issues allowed a new generation of voters to

take place (Hansford and Gomez, 2010). Most studies, however, at least acknowledge that

partisanship resurged after the mid-twentieth century, thus, a new shift in the political party

sphere was surfacing, unlike previous decades. While there is a difference of opinion on the

historical origin of the American resurgence of voter partisanship, there are also strong

differences in determining the value of partisanship on a local level.

In terms of instating partisanship legitimacy on the local level, the debate is fueled by

scholars finding it difficult to measure partisanship for several reasons. The first is argument

between scholars is determining whether there are any valid differences between Democratic and

Republican policies on the local level. Local elections are primarily non-party affiliated and

nonideological as local voters do not possess a need to conduct local affairs based on party

preferences (Bafumi and Shapiro, 2009; Warshaw, 2019). Another contributing factor to the

mass nonideological stance is the lack of resources local offices possess. To compare the needs

and ability to operate local affairs to that of state would fail to acknowledge the difference of

needs between the two levels. The third reason is the local election accountability due to

common local-level causes. In most cases, there is a lack of mass media coverage completed on a

local level to reach a large enough population to create a significant probability of effecting

elections based on partisanship. Yet, the largest responsible factor for affecting local election

accountability the timing of local elections (Bafumi and Shapiro, 2009; Warshaw, 2019).

Especially when taking place during national or state elections, the latter takes precedence over

local polls (Bafumi and Shapiro, 2009; Hassel, 2017). The final reason there is difficulty is the

number of local elections held to be tested and measured. As there is no consistent way to cover

multiple local elections, it poses an additional hindrance to scholars (Warshaw, 2019).

 Chico 5

Secondly, scholars differ on how much emphasis should be placed on the effects of elite

political polarization and partisanship on the mass electorate. By elite, this is in reference to

politicians, and by masses, referring to eligible voters (Warshaw, 2019). Recent studies on

elections have concluded that partisanship and ideology greatly influence voting behavior and

the electorate as is reflected by current incumbents (Warshaw, 2019). Evidence has also been

noted in the increased of intraparty divergence on the elite level (Bafumi and Shapiro, 2009).

One school of thought is that elite behavior, especially from distinct political parties, influences

behavior on the mass level (Branton, 2003; Hetherington, 2001). Additionally, this influence

could positively create a prediction between elite behavior and election results causing scholars

the need to add more to this research in the literature. Measurements in these studies have

included party engagement and the level of likeness towards a party (Branton, 2003;

Hetherington, 2001). The central point and theory behind these studies is that those on the elite

level, showcase political or ideological-centered cues that are reflected on the mass level causing

election results to skew in favor of one party over another (Branton, 2003; Hetherington, 2001).

The primary issue with these studies is that the data may be unreliable given the reliance

of individual responses to base the argument on. More so, there is little empirical insight offered

to substantially determine that behavior on the elite level has a direct effect on mass level voting

behavior. Another limitation is the lack of correlation defined between if an elite behavior only

affects mass behavior from the same party or those non-affiliated. A final limitation is that the

studies both addressed elite behavior as involving mostly Congress, thus, not accounting for

other potential combinations of elite influences such as involving cross-testing with other

branches of government. However, each addresses that individual behavior is what is influenced,

rather than a result of neighboring behavior being responsible for how voters respond
 Chico 6

ideologically at the polls. Additionally, the new American voter is more likely to favor one

political party of another on the grounds of likeliness which is statistically different than what

was the norm during the mid-twentieth century (Bowler, 2016; Hetherington, 2001).

Thirdly, there are arguments as to how conditional partisanship in correlation to elite and

local politicians. Some scholars establish that the average voter is more affluent, older in age and

more educated than the non-voter while others argue in favor of younger eligible voters

(Ansolabehere and Konisky, 2005; Branton, 2003). Most studies conclude that the Republican

party possesses the ideology most expected from this type of voter (Ansolabehere and Konisky,

2005; Bafumi and Shapiro 2009; Branton, 2003; Hassel, 2017). It can, therefore, be predicted

that there should be a surge in Republican-affiliated officials. However, the results in recent

studies have shown otherwise and there are varying data as to why. One theory discussed was the

Partisan Hypothesis in which increases in turnout should ultimately lead to increases in a

Democratic electorate (Hansford and Gomez, 2010). This is following the logic that if the

average turnout is of regular voters, then the results will confer Republican incumbent.

Inversely, if there are more non-regular voters present, then the results will favor a

Democratic incumbent. What should be noted is the possibility of conditionality an increase in

voter turnout gives to the Democratic win. According to the study conducted by Hansford and

Gomez, an increase in non-regular voters is favorable for the Democratic electorate if there is a

Republican incumbent (Hansford and Gomez, 2001). This was otherwise classified as the anti-

incumbent hypothesis. More so, the partisan of the current president also influenced how non-

traditional voters selected candidates, thus, higher voter turnout for the Democratic electorate is

dependent based on multi-level partisanship conditions (Hansford and Gomez, 2001).

 Chico 7

For scholars, understanding the number of ways in which partisanship can affect local

elections is key in grasping a better idea of how voter turnout affects the electorate and the

incumbents that follow. One view presented by scholars is understanding the volatility effect

non-voters have on the electorate. As a non-traditional voter will typically not be as educated as

to the counterpart, some scholars conclude that this type of voter will vote opposite of slight bias.

Should a non-voter retain even a weak bias towards one party, the non-voter will select a

candidate based on the preferability of the candidate to the voter (Frederick 2012; Hansford and

Gomez, 2010). This type of result would indicate that should the majority of minority typical

voters turnout on election day, the election results should be unpredictable and more likely to

favor Republican candidates.

The opposite effect can be viewed from another study in which a survey of partisanship

levels was conducted. The analysis conducted showed that a slight bias for one political party

will result in a “cheerleading” effect (Bullock and Lenz, 2019). Partisan bias was measured in

response to factual questions in which there was a correct response. The primary conclusion was

that the variance of accountability by each participant could potentially reflect the level of

accountability of voters at the voting poll (Bullock and Lenz, 2019). However, since most of the

research presented varies due to the confidence level of each individual in supporting the given

response, it is not as strong as other opposing views on partisanship are with more empirical and

consistent evidence.

Fourth, the lack of consistency that partisanship takes in soliciting how voters affect the

electorate leads the way for further analysis of voting behavior when there is a lack of

information. Whether it be from a lack of community engagement, media, campaigning, or due

to the type of election it is, such as local or national, the judgmental cues a voter first resorts to
 Chico 8

does affect in conjunction with partisanship on whom the non-traditional voter, especially,

selects as the candidate. There have been several studies focused on analyzing how the level of

attractiveness, perceived competence, snap judgments and candidate appearance positively or

negatively holds electorate advantage. According to these studies, when voters are presented

with less information about the candidate and are forced to select a candidate based on perceived

competence or attractiveness, voters tend to choose candidates with higher individual preferable

ratings first (Banducci et. al., 2008; Lawson, et. al., 2010; Praino et. al., 2014). The more

information voters were presented with, the lower the voters chose candidates based on

stereotypes and snap-judgments (Banducci et. al., 2008; Lawson, et. al., 2010; Praino et. al.,

2014).

It can, therefore, be observed, that if local elections possess fewer resources to provide

more information of the candidates, then the non-voters especially will not only select candidates

based on snap-judgments and stereotypes, but also a party candidate in which the voter does not

possess even a slight bias in. On the other hand, since the evidence has not been tested with

various imitations yet, it could also be true that non-voters who turnout on election say may

subdue the suppression of ideological bias and select the best candidate based on preferability

alone. The role of partisanship in voting behavior and election results is therefore present, but

diverse in scholarship.

III. THEORY AND HYPOTHESES

According to the existing literature, partisanship can have an effect on voting behavior

thereby suggesting a positive relationship between partisanship and voting behavior. This study

aims to test if partisanship, the independent variable, could be a contributing factor in whether

voters voted or did not vote, the dependent variable. The control variable of age will be assessed
 Chico 9

to determine whether age bears a statistically significant impact on the decision of party-

registered voters to vote. Previous studies have not focused on expanding the idea between

partisanship affecting voting on the factor of age, therefore, it is included in this study.

The expectation behind this theory is rooted in what previous scholars have noted in the

literature. The existing literature suggests non-party affiliated voters to be more likely to vote

than registered party voters (Ansolabehere and Konisky, 2005; Branton, 2003). In addition,

younger voters are more likely to be non-party affiliated than older voters (Ansolabehere and

Konisky, 2005; Branton, 2003). Therefore, if younger voters are less likely to be party-affiliated,

then I can expect age to be a factor in conjunction with partisanship to positively affect the

outcome of whether a voter voted or did not vote.

HYPOTHESIS I: I expect party identification to effect whether r voted or not.

The existing literature has suggested that elite behavior can have a level of influence on

mass behavior and ultimately thereby effecting the voting polls (Hetherington, 2001). As such,

the literature has provided some insight on when certain groups of politicians, such as in the

legislature, have affected the masses on a lower level such as with polarization in state elections.

However, one gap that could be further examined as the correlation between the approval of the

President handling Office and how that affects a voter voting or not. This relationship would

provide key information on how elite level behavior is not only the main contributor to lower

elections in large groups, but how smaller groups could pose effects on the mass level as well. In

doing so, this will also provide an opportunity to see how partisanship contributes to the type of

results of whether a respondent voted or not based on an effect of the party of the President.

Therefore, this study will also focus on measuring the correlation between the approval of the
 Chico 10

President handling Office and whether the respondent voted or not with occupation as a control

variable.

Previous literature has downsized the importance of the factor of the occupation of a

voter on if a voter will vote separate from having a level of partisanship (Bafumi, Shapiro, 2009).

The theory stems from the concept that if lower-income voters are more likely to be liberal and

Democrat party-affiliated, then the respondent will most likely have voted to increase the

positive effects of benefits (Bafumi, Shapiro, 2009; Warshaw, 2019). As higher-income voters

are less likely to utilize and need government benefits and more likely to be conservative and

Republican party-affiliated, then the drive to vote as a need depletes causing the type of

occupation to ultimately determine whether a voter will vote or not.

HYPOTHESIS II: I expect approval of President of the United States handling their job to

affect whether r voted or not.

IV. DATA AND METHODS

To assess the expectations outlined in the study, I am using the data from the ANES 2016

Time Series Study. The ANES 2012 Time Series was conducted beginning in September 2016

through January of 2017. All pre-election interviews occurred at least two months before the

2016 elections. The respondents were asked post-election questions beginning in November of

2016 while all face-to-face interviews were web-based. Data collection for the study was

collected in two separate modes with different samples with similar questionnaires. Any web-

administered cases were considered a representative sample apart from those face-to-face

collected via the internet. All results and models were performed using the STATA 16 software.

The dependent variable for this study is whether the respondent voted or not. The data

collected for the variable was carried out as a web-screener during a 2012 election. Respondents
 Chico 11

had five choices to select from: ‘Did not answer’, ‘Inapplicable question’, ‘Inapplicable, FTF

sample,’ ‘Yes, voted’, and ‘No, didn’t vote’. The frequency of distribution and number of

observations yielded for this variable was a total of 4,270. The five categories offered to

respondents to select from offers minimum insight on what the focus of the study is. If the type

of measure the dependent variable is offering is whether the respondent voted or not, then a

change in the coding scheme is required. In doing so, there is a better ability to compare the

variables is possible and measure the effect the other variables have on the primary variable of

interest in the study, that is, the dependent variable.

To provide the most insight on how to best evaluate the sample collected for the covariate

of interest, the dependent variable was operationalized as a dichotomous variable (Covariate).

The covariate is whether the respondent had voted or not voted. If the respondent had not voted,

the variable would hold a value of “0” under the value name of “Did vote”. Likewise, if the

respondent had voted, the variable would yield a value of “1” under the new name “Did not

vote”. All those who had selected options other than “Yes” and “No” were coded to fall under

the value of “0” for “No”. By coding the majority of responses under the value of 0, this allows a

clearer comparison in evaluating the effect other variables have on the dependent variable.

Additionally, by having the majority of responses coded under the result of “Did not vote”, this

follows with the theory of the study. If the theory is that the majority of voters will not vote due

to a sense of security from competent partisanship, then there will be less who vote at the polls.

If a majority had been coded under the positive “Did vote”, then the results would reflect a

converse theory in which more who voted indicated fewer partisanship ties.

As the primary independent variable, Party Identification needed to be recoded to best fit

the model (Party_ID). To accomplish this task, I recoded the original seven nominal categories
 Chico 12

into only two nominal categories: “Democrat” and “Republican”. All other observations that did

not fit into the new party codes were coded as “Missing”. The majority of the observations were

identified as “Democrat” and the frequency of distribution was 4,270. By allowing the two

dominant parties to remain, it lowers the number of comparisons drawn between party

identification and the covariate. The third variable in the experiment is the second independent

variable of approval of the President handling Office (Presidential_Approval). The variable was

recoded from the four categories of “refused”, “Don’t Know”, “Approve” and “Disapprove” to

two categories of “Approve” and “Disapprove”. With a total of 4,270 observations, the

“Approve” rank yielded 162 more observations with a frequency of 2,193 than “Disapprove”,

bearing a frequency of 2,031 respectively. By limiting the data to only provide a dichotomous

outcome, this provides an opportunity for a clearer comparison between the Covariate and

Approval or Disapproval of the President handling office.

For this experiment, two controls could affect the results of the study. The first control of

age (Age_) was recoded to ensure only those qualifying as eligible to vote showed in the data

ranging from “18” through “90 and older”. Insignificant observations were labeled “Missing” at

the beginning of the data set. Finally, the second control variable is that of the occupation of the

respondent (Work_Status). Out of the eight original categories, two were re-coded where the

values were held as “0” if “Not Working” and “1” if “Working”. The respondents that selected

the category “working” and “students” were considered working if more than two hours a week

were filed. Additionally, I wanted to include “student” within the considered “working”

category, regardless if respondents did have a job present, as I consider the time and environment

spent in this type of age group to be a significant factor in the study. If students are active in the

political sphere and a large component of potentially shifting voter turnout, then it should be
 Chico 13

included in the study as considered a part of the “working” population. The total observations for

both control variables was 4,270 observations.

Figure 1 displays a histogram of the results of the variable “Covariate”, the dependent

variable. The histogram shows the frequency occurrence of each value within the data set. In this

case, the variable is dichotomous,

therefore only two types of responses

are displayed. As shown in the

graph, approximately eighty percent

of respondents reported to not have

voted in the two-thousand-and-

twelve election web-based survey. In

calculating the data via percentage, it

FIGURE 1.0

provides a clearer understanding of how many of the respondents did not vote for the 2012

election as surveyed compared with those who did. The data then can be compared with other

key variables in the study such as the independent variable of partisanship, “Party_ID”.

A notable factor in this study

is the control variable of Age, coded

as “Age_” as it has a wide number of

responses. In Figure 2.0, a histogram

of the variable is shown with a

frequency distribution ranging from

0 through 0.025. The age of most

respondents in the survey was 48

FIGURE 2.0
 Chico 14

years of age as the mean is 48.17728. This indicates that the majority of respondents are non-

retired and potentially still working. When cross-examining partisanship with those who voted, I

can expect age to hold a role in effecting respondents voting.

V. FINDINGS

FIGURE 3.0

The best fit model for this experiment was a logit regression model. This would allow

the dichotomous variable to be cross-examined properly against all other variables. Figure

3.0 dictates the statistical significance within the variables of the experiment. By statistically

significant, this indicates that the variable was not correlated by chance, but has a significant

stance and contributes to the data set in a meaningful way. The p-values of all five variables

were compared to that of 0.05. Variables less than 0.05 are statistically significant while

others are not. According to the data, only two variables were statistically significant, and

these were “Party_ID” and “Work_Status”. Therefore, partisanship and the work status of the

respondent did positively affect the “Covariate”.

The Odds Ratio offered insight on how much of an impact the odds are of effecting

the dependent variable. Since “Party_ID” and “Age_” were higher than the value of “1”, then

the probability of the event happening is higher than the probability of the event not
 Chico 15

happening. The positive outcome is indicative of an increase in the relationship between the

variable and the dependent variable. Whereas, the two remaining variables, “Work_Status”

“Presidential_Approval” are indicative of a decrease. It can be then be interpreted as follows:

“Party_ID” increases the log odds ratio of voting by 1.151803 and “Age_” increases the log

odds ratio of voting by 1.001924. The data shows that both partisanship and age positively

effect the “Covariate” in the study.

In Figures 4.0 and 5.0, a

correlation was completed between

“Covariate” and each statistically

significant variable yielded from the

regression. In Figure 4.0, the results

showed that there was a positive

FIGURE 4.0
correlation between “Covariate” and “Party_ID”. It can be concluded, that partisanship has a

positive relationship in effecting whether a respondent votes. This can be determined as the

value of 0.0462 is less than 0.05, thus, there is a less than 5% chance of error in the

correlation between the two variables.

Therefore, the first hypothesis can be

accepted. For Figure 5.0, the results display

a negative relationship between “Covariate”

and “Work_Status” as the value of -0.0284

FIGURE 5.0 does not meet the standard of an absence of

error. In sum, the data shows that the relationship between a respondent voting or not voting

is not affected with certainty due to the work status of the respondent. Finally, to test the
 Chico 16

relationship between “Covariate” and

“Presidential_Approval”, a correlation test

was applied (Figure 6.0). If the results were

positive, then the second hypothesis was

correct. According to the results, there is a

FIGURE 6.0
negative relationship between “Covariate”

and “Presidential_Approval”. It can then be concluded, that whether a respondent votes or

not is not affected by Presidential Approval, therefore, disproving the second hypothesis.

VI. CONCLUSION

Party identification has long been overlooked as a non-determining factor when

evaluating voting behavior. The debate stemmed from how influenced Americans with long-term

partisan factors and short-term election-specific factors are (Bafumi and Shapiro, 2009). As a

result, limited long-term testing on the relationship between partisanship and the modern

American voter occurred. This study focused on examining alternative explanations for the

research question: to what extent does party identification affect voter turnout? In the data,

correlations were completed between the dependent variable (Covariate) and two statistically

significant variables (Party_ID and Work_Status).

The variables “Party_ID” and “Covariate” yielded a strong correlation outcome of .0462,

thus, suggesting that partisanship has a positive relationship with whether a voter voted or did

not vote. Given the results, American political parties could consider increasing activeness prior

to election terms in order to increase the likelihood of a candidate from a political party resulting

in holding office. The results of this study, however, are limited in that the application of the

scope requires more testing. For example, I suspect there are varying levels of how much
 Chico 17

partisanship effects voter turnout dependent on whether a local, state or national election is

tested.

The second statistically significant variable of “Work_Status” against a correlation test

with “Covariate” resulted in a negative relationship between the two. Consequently, future

studies examining the relationship between partisanship and voter turnout should hinder from

relying on occupation to effect alternative explanatory results. The difference of the results in

comparison to the existing literature is potentially caused by other variables bearing a higher

impact on the decision to vote or not vote such as awareness of election, type and level of

election and levels of party polarization.

A correlation test between the approval of the President handling Office and “Covariate”

was also conducted in order to better understand how elite behavior from the Presidential

affected voter turnout. An advantageous outcome in studying this relationship was the ability to

test the partisanship levels from a smaller elite poll and the effects on voting behavior in this

manner as the President will belong to a party. However, the results showed a negative

relationship between the two variables. Future studies should continue to analyze partisanship in

elite groups that are smaller as this could allow insight on what levels of effects to party-centered

elite groups have on voting or not voting.

In conclusion, further research on the relationship between partisanship and voting

behavior needs to be completed to offer higher qualitative results. While the existing literature

does propose valid arguments for the resurgence and effects of partisanship on the American

voter, there continues to be a gap in what variables could be responsible for how the modern

American votes. Additional variables should be tested against partisanship in hopes of finding

stronger significant data and completed on various election levels such as local, state and
 Chico 18

national. What would be compelling to view, is a cross-analysis between partisanship behavior in

the United States and with other similar democracies to widen the scope of application on this

important aspect of voting behavior. Until then, future studies should focus on finding alternative

courses of explanation between partisanship and voting behavior in hopes of grasping a better

understanding of what drives the American voter to vote.

 Chico 19

References

Ansolabehere, S., & Konisky, D. (2006). The Introduction of Voter Registration and Its Effect on

Turnout. Political Analysis, 14(1), 83-100. Retrieved March 2, 2020 from

www.jstor.org/stable/25791836.

Bafumi, J., & Shapiro, R. (2009). A New Partisan Voter. The Journal of Politics, 71(1), 1-24.

Retrieved on March 6, 2020 from DOI:10.1017/s0022381608090014.

Banducci, S., Karp, J., et. al. (2008). Ballot Photographs as Cues in Low-Information Elections.

Political Psychology, Vol. 29, No. 6, 903-917. Retrieved February 23, 2020 from

https://doi.org/10.1111/j.1467-9221.2008.00672.x.

Bowler, S., Donoval, T. (2016). A Partisan Model of Electoral Reform: Voter Identification

Laws and Conficen in State Elections. Sage Publications, Inc., Vol. 16, No. 3. 240-361.

Retreievd on Marcch 26, 2020 from : https://www.jstor.org/stable/44014940

Branton, R. P. (2003). Examining Individual-Level Voting Behavior on State Ballot

Propositions. Political Research Quarterly, 56(3), 367–377. Retrieved on March 2, 2020

on https://doi.org/10.1177/106591290305600311.

Bullock, J., & Lenz, G. (2019). Partisan Bias in Surveys. Annual Review of Political Science,

Vol. 22: 325-342. Retrieved March 1, 2020 from

https://doi.org/10.1146/annurev-polisci-051117-050904

Frederick, H., (2012). Reforming the Presidential Primary System: The Voter Turnout Initiative,

American Political Science Association, Vol. 45, No. 1: 51-57. Retrieved on March 26,

2020 from https://www.jstor.org/stable/41412721.

Hansford, T., & Gomez, B. (2010). Estimating the Electoral Effects of Voter
 Chico 20

Turnout. American Political Science Review, 104(2), 268-288. Retrieved March 3, 2020

from DOI:10.1017/S0003055410000109.

Hassell, H. J. G. (2017). Teaching Voters New Tricks: The effect of partisan absentee vote-by-

mail get-out-the-vote efforts. Research & Politics, Vol. 1-6. Retrieved on February 22,

2020 from https://doi.org/10.1177/2053168017694806.

Hetherington, M. (2001). Resurgent Mass Partisanship: The Role of Elite Polarization. American

Political Science Review, 95(3), 619-631. Retrieved February 22, 2020 from

doi:10.1017/S0003055401003045.

Lawson, Chappell et al. Looking Like a Winner: Candidate Appearance and Electoral

Success in New Democracies. World Politics, 62.04: 561–593. Retrieved February 19,

2020 from Cambridge University Press.

Praino, R., Stockemer, D., & Ratis, J. (2014). Looking Good or Looking Competent? Physical

Appearance and Electoral Success in the 2008 Congressional Elections. American

Political Research, Vol. 42(6) 1096-1117. Retrieved on February 22, 2020 from DOI:

10.1177/1532673X14532825.

Warshaw, C. (2019). Local Elections and Representation in the United States. Annual Review of

Political Science, Vol. 22:461-479. Retrieved on February 23, 2020 from

https://doi.org/10.1146/annurev-polisci-050317-071108
 Chico 21

Do-File STATA 16

tab Covariate

sum Covariate

*DV: Did r vote or not? 0 = No, 1 = Yes

**Dichotomous Variable

*Steps taken to recode DV:

**recode V165726 (-7=0) (-5=0) (-2=0) (1=1) (2=0), generate (Covariate)

**label define Covariate 0 "No, Did not vote" 1 "Yes, Did vote"

**label values Covariate Covariate

tab Party_ID

sum Party_ID

*IV: Party Identification? 0 = None, 1 = Dem, 2 = Rep

**Nominal Variable

*Steps taken to recode IV:

**recode V161019 (-9=0) (-8=0) (-1=0) (1=1) (2=2) (4=0) (5=0), generate (Party_ID)

**label define Party_ID 0 "Missing" 1 "Democrat" 2 "Republican"

**label values Party_ID Party_ID

tab Presidential_Approval

sum Presidential_Approval
 Chico 22

*IV: approval of POTUS handling their job? 0 = Disapprove, 1 = Approve

**Dichotomous Variable

*Steps to recode IV:

**recode V161082 (-9=0) (-8=0) (1=1) (2=0), generate (Presidential_Approval)

**label define Presidential_Approval 0 "Disapprove" 1 "Approve"

**label values Presidential_Approval Presidential_Approval

tab Age_

sum Age_

*CO: Age? 18 through 90+

**Ratio Variable

*Steps to recode CO:

**recode V161267 (-9=0) (-8=0) (18=18) (19=19) (20=20) (21=21) (22=22) (23=23) (24=24)

(25=25) (26=26) (27=27) (28=28) (29=29) (30=30) (31=31) (32=32) (33=33) (34=34) (35=35)

(36=36)(37=37) (38=38) (39=39) (40=40) (41=41) (42=42) (43=43) (44=44) (45=45) (46=46)

(47=47) (48=48) (49=49) (50=50) (51=51) (52=52) (53=53) (54=54) (55=55) (56=56) (57=57)

(58=58) (59=59) (60=60) (61=61) (62=62) (63=63) (64=64) (65=65) (66=66) (67=67) (68=68)

(69=69) (70=70) (71=71) (72=72) (73=73) (74=74) (75=75) (76=76) (77=77) (78=78) (79=79)

(80=80) (81=81)(82=82) (83=83) (84=84) (85=85) (86=86) (87=87) (88=88) (89=89) (90=90),

generate (Age_)

**label define Age_ 0 "Missing" 90 "90 and older"

 Chico 23

**label values Age_ Age_

tab Work_Status

sum Work_Status

*CO: Occupation? 0 = None, 1 = Working (including students 20+ hours a week), 2 = Not

working

**Ordinal Variable

***Note, working includes being a student whereas not working included disabled, retired,

homemaker etc.

*Steps to recode CO:

**recode V161276x (-9=0) (1=1) (2=2) (4=2) (5=2) (6=2) (7=2) (8=1), generate (Work_Status)

**label define Work_Status 0 "Status unknown" 1 "Working" 2 "Not Working"

**label values Work_Status Work_Status

*ALL NECESSARY VAR NOW REOCODED.

*CREATING GRAPHS/CHARTS

hist Covariate, percent

hist Age_

sum Age_

Commands and models:

 Chico 24

corr Covariate Party_ID

corr Covariate Work_Status

corr Covariate Age_

ologit Covariate Party_ID Presidential_Approval Age_ Work_Status

*Not Appropriate for Dichotomous DV

logit Covariate Party_ID Presidential_Approval Age_ Work_Status

*Appropriate, provides Coefficient NOT Odds Ratio

logit Covariate Party_ID Presidential_Approval Age_ Work_Status, or

*Appropriate for Dichotomous DV ; or DOES NOT give the coefficient, provides Odds Ratio

**logit fits a logit model for a binary response by maximum likelihood; it models the probability

of a positive outcome given a set of regressors.

****NOTE: When the odds are greater than one, the probability of the event happening is higher

than the probability of the event not happening ; An ODDS RATIO of less than 1 indicates a

Decrease.

****NOTE: the COEFFICIENTS indicate the amount of change expected in the log odds when

there is a one unit change in the predictor variable with all of the other variables in the model

held constant.

****NOTE: LOG ODDS = the natural log of the odds

*if using COEFFICIENTS: "interpretation"

 Chico 25

**(if IV/CO shows a negative reasult) This indicates that a decrease of (number, for

example,:1.78 [yes say it as a +]) is expected in the log odds of ()DV) with a one-unit increase in

(IV).

*if using ODDS RATIO: "interpretation"

**For DV "new1_V165726": We expect the respondent deemed who "did vote" to increase the

logs odds ratio of voting by 1.151803.