554756ANN The Annals of the American AcademyPartisanship and Attitudes Toward Science and Public Policy

research-article2014

Despite the apparent partisan divide over issues such as

global warming and hydraulic fracturing, little is known
about what shapes citizens’ willingness to accept scien-
tific recommendations on political issues. We examine
the extent to which Democrats, Republicans, and inde-
pendents are likely to defer to scientific expertise in
matters of policy. Our study draws on an October 2013
U.S. national survey of 2,000 respondents. We find that
partisan differences exist: our data show that most
Americans see science as relevant to policy, but that
Does their willingness to defer to science in policy matters
varies considerably across issues. While party, ideology,
Partisanship and religious beliefs clearly influence attitudes toward
science, Republicans are not notably skeptical about

Shape Attitudes accepting scientific recommendations. Rather, it seems

that Democrats are particularly receptive to the advice
and counsel of scientists, when compared to both inde-
toward Science pendents and Republicans.

and Public Keywords: public opinion; partisanship; science; ide-

ology; religion

Policy? The
Case for
Ideology and
S ince 2004, numerous academic and political
commentators have observed the increasing
difficulty with which common ground is found
between Republicans and Democrats. One
Religion might hope that empirical, objective scientific
inquiry could serve as an agreeable basis for
substantive policy action. After all, many
(most?) public policy issues consist of an array
of technical questions requiring, or at least
inviting, relevant data and informed analysis.
By But a slew of recent articles and books tell us
Joshua M. Blank that Americans are suspicious of science,
and
Daron Shaw Joshua M. Blank is manager of polling and research at
the Texas Politics Project and a PhD candidate at the
University of Texas at Austin, specializing in political
behavior, political psychology, and applied survey
research.
Daron Shaw is a distinguished teaching professor and
the Frank C. Erwin Jr. Chair of State Government at
the University of Texas at Austin, specializing in cam-
paigns and elections, public opinion and voting behav-
ior, political parties, and applied survey research.

DOI: 10.1177/0002716214554756

18 ANNALS, AAPSS, 658, March 2015

Partisanship and Attitudes Toward Science and Public Policy 19

especially when it comes to the political application of scientific research. If true,

this is indeed a disturbing fact. How are we to proceed as a democratic society if
we cannot even agree on matters “settled” through rigorous scientific inquiry?
Like many of you, we have read articles documenting conservative and
Republicans attitudes toward teaching creationism versus evolution in the public
schools, toward global warming, and toward embryonic stem cell research. We
have also read the books claiming that there is a more pervasive “antiscience”
tendency among those on the Right, with its roots in the physiology of the brain
(e.g., Mooney 2012). We have also read the many rebuttals, arguing that it is in
fact liberals and Democrats who are antiscience, as evidenced by their attitudes
toward nuclear power, fetal viability, genetically engineered crops, and vaccina-
tions (e.g., Berezow and Campbell 2012). Although both sides offer entertaining
(and occasionally compelling) arguments, we are struck by the lacuna of reliable,
high-quality public opinion data in this increasingly contentious debate; in short,
the existing literature on both sides lacks the data necessary to identify and test
the underlying propositions that they are all too willing to advance.
We rely on a 2,000-person national sample from October 2013 to ascertain
attitudes toward the use of scientific expertise on a wide range of public issues.
Moreover, we estimate the effects of several political attitudes and demographic
characteristics on deference toward science across these issues. We are particu-
larly interested in how political attitudes, along with religious faith and education,
impact views about the proper role of science in shaping public policy. The data
demonstrate that most Americans view scientific expertise as relevant—even on
the most politically charged issues. The willingness to defer to science varies
considerably across issue domains, however. And while party, ideology, and reli-
gious beliefs clearly influence attitudes toward science, Republicans are not
notably skeptical about accepting scientific recommendations. Rather, it seems
that Democrats are particularly receptive to the advice and counsel of scientists,
when compared to both independents and Republicans. Thus, as is often the
case, the question of antiscience attitudes is more complicated (but perhaps
more encouraging) than many have presumed.

Elites, Scientists, and Public Opinion

If we want to describe the attitudes of Americans toward science and public
policy, we need a more general theory (or, at the very least, a theoretical perspec-
tive) of attitudes. Toward this end, we freely beg, borrow, and steal from the
established literature. A great deal of public opinion research focuses on how
most Americans do not know very much about public affairs and are not very
interested in politics (e.g., Campbell et al. 1960; Converse 1964; Delli Carpini
and Keeter 1996). They develop attitudes—most notably, party identification—in
early childhood and throughout their adolescent years through the same sociali-
zation processes that drive attitude formation more generally. But these attitudes
consist mostly of broad notions of affect and identity; an individual identifies as a
20 THE ANNALS OF THE AMERICAN ACADEMY

“Republican” and feels positive toward other Republicans. For most Americans,
there is little substance or depth to these attitudes.
When faced with questions about politics and specific policies, most Americans
must therefore “construct” opinions based on a sampling of available considera-
tions (Zaller 1992). Which considerations are available depends on factors that
are sometimes random (weather events, popular culture, or news of the day) and
sometimes not so random (a story they saw on Fox News or MSNBC that morn-
ing, or “water-cooler” conversations with friends, coworkers, and associates). For
those Americans who are paying attention, party and other elite sources of “cred-
ible” information serve as useful guides for opinion. We know what to think when
“our” elites articulate an issue position and we happen to hear about it. Elites
thereby introduce “constraint” to public policy discussions, at least among those
who are following along.
To the extent that individuals have opinions or broader attitudes on political
questions, they seem to be more interested in reinforcing them than subjecting
them to rigorous deliberative scrutiny. Most notably, “motivated reasoning”
occurs when people encounter information that is inconsistent with their existing
attitudes or opinions on a subject, and then discount that information to mitigate
cognitive dissonance. This is “a form of implicit emotion regulation in which the
brain converges on judgments that minimize negative and maximize positive
affect states associated with threat to or attainment of motives” (Western et al.
2006, 1947). For our purposes here, it is important to observe that motivated
reasoning might cause individuals to reject scientific evidence if it conflicts with
policy judgments and opinions based on other attitudes or belief systems.
This is, obviously, an oversimplification of how opinion dynamics work. But it
seems a useful starting point for thinking about how citizens come to see scien-
tists and their role in public policy matters. From our perspective, scientists are
certainly elites. They are highly educated and often granted a platform for com-
municating with the public when they happen to produce research that is politi-
cally and/or personally relevant. And people think this is generally a good thing.
A July 2009 national survey conducted by Pew compared the attitudes of the
public and scientists and found that 84 percent of Americans think science’s
effect on society is “mostly positive” and 70 percent think scientists contribute “a
lot” to society’s well-being.1 Even those who hold opinions inconsistent with sci-
entific consensus rated scientists highly: 63 percent of those who think beings
were created in their present form (a creationist perspective) say scientists con-
tribute “a lot” to the well-being of society. Perhaps more to the point, both scien-
tists and the public overwhelmingly say it is appropriate for scientists to become
active in political debates about such issues as nuclear power or stem cell
research. Virtually all scientists (97 percent) endorse their participation in
debates on these issues, while 76 percent of the public agrees.
However, despite the classical notion of scientist as dispassionate researcher,
survey research also suggests that scientists have distinct political and ideological
preferences. In the Pew survey, 52 percent of scientists identified themselves as
“liberal” (compared with 20 percent of the public) and only 9 percent identified
as “conservative” (compared with 37 percent of the public). Fifty-five percent of
Partisanship and Attitudes Toward Science and Public Policy 21

scientists identified themselves as Democratic, compared to 6 percent who iden-

tified as Republican (the public was split 35 percent Democratic to 23 percent
Republican). Scientists were similarly much more sanguine than the general
public about the efficacy of government action.2
Do Americans, especially those on the Right, notice that scientists are dispro-
portionately liberal?3 And, more to the point, does this affect the willingness of
conservatives or Republicans to accept the recommendations of scientists on
public policy matters? The notion that some Americans might cast a skeptical eye
toward scientists is not new. To a degree, this skepticism makes sense. People
tend to have a perceptual problem with uncertainty, and since a principal com-
ponent of science is to be explicit about one’s level of uncertainty, this cognitive
state of affairs has the potential to lead to diminished ratings of credibility among
the mass public (Tenney et al. 2007). But, to the extent that this sentiment holds
sway, or that suspicions of scientific bias exist, are they disproportionately con-
centrated among those on the Right?

Research Hypotheses
We suspect deference to scientists varies across social and demographic groups,
but we are ambivalent with respect to its partisan tilt. We think it is likely that
skepticism toward science is greater on the political Right than on the Left,
mostly because (1) scientists are more likely to be liberal than moderate or con-
servative; (2) public policy recommendations from scientists typically call for
governmental action or regulation, something conservatives generally oppose;
and (3) political leaders on the Right have been more aggressive in “calling out”
scientists who advocate for liberal policies.4 But we also think that political differ-
ences with respect to skepticism are likely to be concentrated on issues on which
scientific recommendations push in a direction that conflicts with preferences
derived from alternative belief systems. Put plainly, we suspect that people are
less deferential toward science when scientists tell them that they need to do
something at odds with their ideological or religious beliefs.
Along these lines, we think that political ideology is a more promising explana-
tion for lack of deference toward science than is partisanship. Ideology in the
American context consists of one’s assessment of the extent to which government
ought to be involved in managing and regulating the lives of citizens. Conservatives
believe, as Ronald Reagan famously said, “government is the problem, not the
solution.” Thus, conservatives can conceivably have two problems with scientific
recommendations: the specific policy or regulation proposed, and the more gen-
eral proposition that government ought to be involved at all. Presumably, liberals
would only object to the specific nature of the proposed government action.
Since partisanship is a more particular attitude than ideology and is more closely
connected to electoral competition and the specific issue conflicts that animate
party competition, we are less convinced that it is strongly connected to science
attitudes.5
22 THE ANNALS OF THE AMERICAN ACADEMY

Our core assumption is embedded within the preceding paragraphs: we think

that skepticism toward science is caused by a conflict between a scientific view-
point and alternative points of view—most notably, political ideology and reli-
gious faith. As suggested earlier, we think individuals engage in motivated
reasoning when they encounter scientific evidence (or a policy recommendation)
that contradicts their belief systems. This assumption squares with a series of
recent studies that show that political ideology and moral beliefs influence peo-
ple’s judgments of facts. In an experimental setting, subjects who read an essay
arguing that the death penalty is barbaric and immoral estimated lower crime-
reducing benefits and greater costs for the death penalty than subjects who read
an essay arguing that the death penalty was morally justified and showed high
regard for human life. The authors found essentially similar results for several
morally charged political issues, including the use of torture, promotion of con-
dom use, and stem cell research (Liu and Ditto 2013). Studies have also showed
that these modest tendencies for distortion tended to be strongest among people
who (1) held strong moral convictions, (2) considered themselves most informed,
and (3) were conservatives. This last finding was quite modest, so that even if
conservatives’ views of science are a bit more distorted than liberals, there was
ample distortion among liberals.6
We have a small set of specific research hypotheses. First, we expect that the
American public continues to view scientists as credible experts. But we also
think that science is more obviously relevant on some issues than others, and that
political and religious ideology will come into play on certain issues. Therefore,
deference should decline for issues where other sources of opinion are available
and clearly applicable. In particular, issues such as mandatory health insurance,
regulation of coal production, and mandatory background checks for gun permits
are likely to bring political ideology into play; while fetal viability, embryonic stem
cell research, and the teaching of evolution and the origins of human life are
likely to bring religious beliefs into play. More formally:

Hypothesis 1 (H1): The public generally defers to scientific expertise on a

range of public policy issues, but deference will be lowest for issues on
which scientific advice contradicts political ideology and/or religious beliefs.

Given the tendency of scientists to prefer government action and to be agnos-

tic (so to speak) on matters of religious faith, we also expect that:

H2: Political conservatives and those who are religiously observant will be
more skeptical toward scientific policy recommendations.

But what about our expectations for the relationship between party identifica-
tion and attitudes toward science? As suggested earlier, once political ideology
and religious beliefs are taken into consideration, we believe that there will be no
strong differences between Republicans, Democrats, and independents on def-
erence to scientific recommendations. To the extent that differences do exist, we
think that Democrats are relatively more likely than Republicans to accept
Partisanship and Attitudes Toward Science and Public Policy 23

scientific advice. But given the strong overall support for scientists in previous
polls, we expect differences to be driven by Democrats expressing overwhelming
support for science. In this sense, we depart from the conventional wisdom—the
interesting thing is not antiscience attitudes among Republicans but proscience
attitudes among Democrats. This leads to:

H3: Democrats, Republicans, and independents will all report a willingness to

defer to science on policy issues, with Democrats being particularly defer-
ential.

While those on the political Right may face cognitive difficulty in dealing with
evidence that contradicts their policy preferences, or their more general prefer-
ences for limiting government size and scope, those on the Left should face no
such difficulty. Most individuals probably harbor few specific, complex thoughts
where policy and scientific recommendations intersect; and on the Left, elite
cues signaling skepticism or outright rejection of scientific recommendation
appear to be essentially absent. Therefore, more cognitive work would be
required to find reasons to reject scientific policy recommendations than would
be required to simply accept them based on the notion that scientists are them-
selves educated and highly esteemed elites. The theoretical rational laid out
above is, in a sense, the flip side of motivated reasoning. That specific counter-
vailing considerations to scientific recommendations do not exist on the Left is
why liberalism alone is unlikely to drive this relationship. It is the absence of a
countervailing viewpoint, not its presence, that makes those on the Left more
accepting of science. At any rate, this counterintuition is a testable proposition.
Finally, we also expect that individuals who have been exposed to more formal
education are especially likely to have knowledge of the scientific process and, in
turn, respect for scientific findings. This can be stated as:

H4: Education will be positively associated with deference to science.

As you can see, we prefer simple explanations to complex ones. We explicitly

reject the notion advanced in recent books that both conservatives and
Republicans are “hardwired” to reject science (Mooney 2012). It seems to us
much more plausible that conservative skepticism (to the extent that it exists) is
driven by the need to reconcile science with deeply held belief systems. At the
very least, this discussion of brain physiology puts the cart ahead of the horse. We
do not know very much about how much weight citizens think we should give to
scientists and scientific expertise on different public policy matters. Nor do we
know much about the extent to which citizens see scientists as biased or other-
wise potentially compromised as elite opinion leaders. And despite recent polls
exploring party differences on specific issues for which science is presumably
relevant, we do not know much about whether Republicans and Democrats dif-
fer in their views of the relevance and credibility of science for adjudicating
public policy debates. It is toward these questions that we now turn.
24 THE ANNALS OF THE AMERICAN ACADEMY

Research Design
To assess partisan attitudes toward science, we commissioned the company
YouGov to administer a U.S. national survey (see appendix). From October 24
through October 31, 2013, 2,000 registered voters were interviewed online about
their attitudes toward science and politics. For our purposes here, two batteries
of questions were relevant. The first battery posed the following frame:

Some people think that our elected officials and policy-makers should follow the advice
of scientists on many issues. Others think that scientists are often either biased or wrong
on many issues. For each of the following issues, please use a 0−10 scale to indicate how
much politicians and public officials should defer to scientists: 10 means to follow scien-
tific advice completely, 0 means to ignore scientific advice, and 5 means to weight it
equally with other factors.

We then presented respondents with a randomized list of sixteen policy areas

including AIDS prevention, mandatory childhood vaccination, regulation of
nuclear power, childhood obesity and diet restrictions, birth control education,
stem cell research, mandatory background checks for gun permits, fetus viability,
global warming/climate change, the regulation of coal production, the production
of biotech foods and crops, legalizing drug use, animal testing for medical
research, teaching evolution and the origins of humans, mandatory health insur-
ance, and gay adoption. After pilot testing a similar battery in an earlier study,7
we prioritized specific policies and maximized the span of policy items.8 In par-
ticular, we wanted to have some issues on which scientific recommendations are
clear and some on which they are unclear, some issues on which science is obvi-
ously relevant and some on which it is not so obviously relevant, and some issues
on which scientific policy recommendations please the Right and some on which
they please the Left. Our views on the issues are expressed in Table 1. These
sixteen items served as potential dependent variables, with the underlying simi-
larity (or dissimilarity) of issue responses a matter of empirical estimation.
The second battery of questions asked respondents about their political atti-
tudes and demographic characteristics. In particular, we asked people about their
ideological orientation, party identification, occupation, age, education, race and
ethnicity, church attendance, and their view on whether the Bible is the literal
word of God. This second set of questions provided us with a wide range of
explanatory variables with which we hope to model deference toward scientific
opinion.

Results
Table 2 shows the average rating for each of the policy areas we chose to examine.
Recall that a rating of 10 indicates that the respondent believes that policy-
makers should completely follow the advice of scientists on that particular policy
issue and 0 indicates that they should completely ignore the same advice. The
Partisanship and Attitudes Toward Science and Public Policy 25

Table 1
Policy Items Used to Test Attitudes toward Scientific Advice

Scientific Direction of Anticipated Objections

Consensus Is… Consensus… Mainly From…

Embryonic stem cell research Strong Anti-regulation Conservatives

Fetus viability Mild Anti-abortion Liberals
Global warming/climate change Strong Pro-regulation Conservatives
Gay adoption Mild Pro-adoption Conservatives
Childhood obesity and diet restric- Strong Pro-regulation Conservatives
tions
AIDS prevention Strong Pro-regulation Conservatives
Birth control education Mild Pro-education Conservatives
Legalizing drug use Unclear N/A Neither
Mandatory health insurance Unclear N/A Neither
Regulation of coal production Strong Anti-coal Conservatives
Mandatory background checks for Unclear N/A Neither
gun permits
Producing bio-tech food and crops Strong Pro-biotech Liberals
Regulation of nuclear power Mild Pro-nuclear Liberals
Animal testing for medical research Strong Pro-testing Liberals
Mandatory childhood vaccinations Strong Pro-vaccine Liberals
Teaching evolution and the origins Strong Pro-evolution Conservatives
of humans

first column shows the overall means ranked by level of deference to scientific
opinion. The highest deference to science score is for AIDS prevention (mean =
7.9) while the lowest is for gay adoption (4.9). These simple comparisons tell us
quite a lot about public attitudes toward science and policy. For starters, we see
that scientific opinion matters for most people. As noted above, the overall lowest
scientific deference rating is close to 5.0—on no issue do we see ratings fall any-
where near the “ignore scientific advice” endpoint. The overall average rating,
across all issues, is 6.4. So as we discuss variation with respect to scientific defer-
ence, it is important not to obscure the fact that most Americans care about sci-
entific opinion when it comes to most public policy issues.
There is, however, substantial variance in the degree to which people think we
ought to defer to science. This variation occurs across issue domains and across
different groups of respondents. As expected, willingness to defer to scientific
expertise decreases as we encounter issues that touch on matters of religious faith
(gay adoption, evolution) or political ideology (mandatory health insurance). The
six issues where people are most deferential to scientific advice are AIDS preven-
tion, mandatory vaccines, nuclear power, childhood obesity, birth control, and
stem cell research; although the last two touch on religious faith, the first four do
not. The average deference scores range from 7.9 to 6.8, and no political or
26 THE ANNALS OF THE AMERICAN ACADEMY

demographic subgroups score at or below the midpoint of 5.0. Conversely, the six
lowest-scoring issues are biotech foods, legalizing drug use, animal testing, evolu-
tion, mandatory health insurance, and gay adoption; the final three issues in this
set are obviously related to intense political or religious debate. The average
deference scores here range from 6.0 to 4.9, and several subgroups score below
the midpoint of 5.0. The data thus offer mixed support for our first hypothesis
(H1). Not all “political” or “religious” issues have relatively low deference scores;
but it does seem that on these issues the public’s willingness to defer to science
somewhat (though by no means completely) diminishes.
Table 2 also broaches the question of whether partisanship influences one’s
willingness to defer to scientific advice. Recall that H3 predicts (1) deference to
science to be high across the board and (2) Democrats to be relatively more sup-
portive of science. Looking across the means by party affiliation of the respond-
ents, we see that there are striking differences between Democrats and
Republicans—greater and more consistent than we expected. The Democrats in
our survey are more deferential to science than are the Republicans across every
issue domain, and these differences are all statistically significant at the .001
level. The average deference score across all sixteen items is 7.46 for Democratic
respondents, 5.58 for Republicans, and 5.84 for independents. This, of course, is
not necessarily surprising given all the talk about antiscience attitudes among the
GOP. What is surprising is that those lacking a partisan affiliation—the independ-
ents9—hold opinions closer to the Republicans. In other words, it is the relative
proscience attitudes of Democrats that stand in contrast to the rest of the popula-
tion and not the antiscience attitudes of Republicans. Indeed, Republicans rate
their own deference to science at or above the midpoint of 5.0 for all but four of
the sixteen issues: global warming (4.8), evolution (4.4), gay adoption (3.9), and
mandatory health insurance (3.5).
While suggestive, the comparison of means is hardly definitive. A more con-
clusive analysis involves estimating respondents’ deference to science with a
theoretically comprehensive set of explanatory variables. Table 2 offers models
for each of the sixteen policy areas. The specification of the models is substan-
tially more complete than any of the existing empirical analyses that we have
encountered: we allow self-rated political ideology (measured by a standard
5-point scale), party identification (with separate Democratic and Republican
variables), age (under 30 years of age, and 65 and over), education (college edu-
cated versus less than a college degree), religiosity (with a dummy variable for
those who attend church at least once a week and a 3-point scale indicating the
extent to which the respondent believes that the Bible is the literal word of
God), race (African American or not), and ethnicity (Latino or not) to predict
each rating.10
In deciding to estimate sixteen individual models, we do not wish to overlook
the fact that the different items may mask substantial similarity. To ascertain
whether there was any consistent structure to deference across the different issue
domains, we conducted several dimensional analyses. No matter what specifica-
tion or estimator we used, however, the results were remarkably consistent: there
is one factor that structures most of the variance between and across the issue
Partisanship and Attitudes Toward Science and Public Policy 27

Table 2
Attitudes toward Scientific Advice in Different Policy Issues

Republican
-Democratic
Overall Democratic Independent Republican Difference
Mean (SE) Mean (SE) Mean (SE) Mean (SE) (p-Value)

AIDS prevention 7.9 (0.08) 8.7 (0.11) 7.5 (0.13) 7.4 (0.17) −1.3 (p < .0001)
Mandatory childhood 7.2 (0.09) 8.1 (0.12) 6.7 (0.14) 6.8 (0.20) −1.3 (p < .0001)
vaccinations
Regulation of nuclear 7.2 (0.08) 7.9 (0.13) 6.7 (0.13) 6.9 (0.18) −1.0 (p < .0001)
power
Childhood obesity and 6.9 (0.08) 8.0 (0.12) 6.3 (0.13) 6.1 (0.18) −1.9 (p < .0001)
diet restriction
Birth control educa- 6.8 (0.09) 8.0 (0.12) 6.3 (0.15) 5.9 (0.22) −2.1 (p < .0001)
tion
Embryonic stem cell 6.8 (0.10) 8.0 (0.12) 6.3 (0.14) 5.9 (0.23) –2.1 (p < .0001)
research
Mandatory back- 6.5 (0.11) 7.8 (0.14) 5.9 (0.17) 5.4 (0.23) –2.4 (p < .0001)
ground checks for
gun permits
Fetal viability 6.4 (0.10) 7.3 (0.14) 6.1 (0.15) 5.8 (0.22) −1.5 (p < .0001)
Global warming/cli- 6.4 (0.10) 8.2 (0.12) 5.6 (0.17) 4.8 (0.23) −3.4 (p < .0001)
mate change
Regulation of coal 6.2 (0.10) 7.5 (0.13) 5.5 (0.15) 5.1 (0.20) −2.3 (p < .0001)
production
Producing biotech 6.0 (0.09) 6.6 (0.14) 5.5 (0.15) 5.6 (0.20) −1.0 (p < .0001)
food and crops
Legalizing drug use 5.8 (0.10) 6.6 (0.16) 5.6 (0.16) 5.2 (0.23) −1.4 (p < .0001)
Animal testing for 5.8 (0.09) 6.5 (0.15) 5.4 (0.13) 5.6 (0.22) −1.0 (p = .0003)
medical research
Teaching evolution 5.7 (0.11) 7.1 (0.17) 5.3 (0.17) 4.4 (0.23) −2.7 (p < .0001)
and the origins of
humans
Mandatory health 5.0 (0.11) 6.9 (0.15) 4.2 (0.17) 4.5 (0.24) −3.5 (p < .0001)
insurance
Gay adoption 4.9 (0.11) 6.1 (0.17) 4.6 (0.17) 3.8 (0.24) −2.2 (p < .0001)

SOURCE: Data are from a U.S. national survey of 2,000 registered voters conducted online
by YouGov (October 24–31, 2013).
NOTE: Cell entries are 0−10 means. Respondents were asked to use a 0−10 scale, where 0
means ignoring scientific advice and 10 means following scientific advice completely, to assess
their attitudes toward science on the specific issues.

items. We were open to the possibility that issues that invoke aspects of political
ideology or religious faith might evince distinct response structures. But this was
not the case.11 In the end, along with the sixteen individual models, we chose to
28 THE ANNALS OF THE AMERICAN ACADEMY

include a summary model, with the average rating across the sixteen items serv-
ing as the dependent variable.
We use a least squares estimator because it is simple, robust, and appropriate.
We have a continuous dependent variable and a reasonable expectation that the
explanatory variables are exogenous with limited multicollinearity.12 Preliminary
examinations of the errors suggest that they are homoscedastic and serially
uncorrelated. We did experiment with alternative estimators and none provided
estimates substantively distinct from those offered here.
The results of the least squares models are presented in Table 3. Let us begin
by noting the consistent and strong effect of ideological self-identification on
deference toward science. The negative coefficient indicates that those who iden-
tify as more conservative are less likely to rate themselves as deferential toward
scientific expertise. The effect is statistically significant at the .01 level for all
sixteen models. The largest coefficient is −0.84 for global warming. This means
that every unit increase in the 1–5 ideology rating scale is associated with a 0.84-
point decrease on the 0−10 deference to science scale. Put another way, moving
from “extremely liberal” to “extremely conservative” is associated with a 3.36-
point decrease in one’s deference toward science rating for global warming. The
smallest coefficient is −0.19 for animal testing, which means minimum-to-maxi-
mum movement on the ideology scale produces a 0.76-point decrease in one’s
deference toward science on that issue. On the summary dimension, the ideology
coefficient is −0.22, which means minimum (extremely liberal) to maximum
(extremely conservative) movement on the ideology scale produces a 0.88-point
decrease in one’s deference toward science. The models, therefore, unambigu-
ously support H2—that identifying as more conservative is associated with less
willingness to defer to scientific recommendations.
While ideology has a clear and substantive effect on attitudes toward science
and public policy, party identification has a more complex effect. Keeping in
mind that the baseline for the models is an “independent” respondent, Table 3
shows that identifying as a Democrat has a positive and statistically significant
effect (at the .05 level) for all sixteen models. The effect of being a Democrat on
the 0−10 scientific deference scale ranges from 0.4 for legalizing drug use to 1.6
for mandatory health insurance. (Since Democratic identification is measured
dichotomously, the coefficient estimate is the same as the effect size.) As hypoth-
esized (H3), Democrats seem to be relatively proscience even controlling for
political ideology and a host of demographic factors. Identifying as a Republican,
however, has a small and decidedly mixed effect on deference toward scientific
expertise. In fact, the Republican coefficient is positively signed for fourteen of
the sixteen models, with the positive impact reaching conventional levels of sta-
tistical significance for four issues: biotech foods, nuclear power, animal testing,
and mandatory vaccines. We do not wish to make much of these differences; by
and large, the relevant coefficients are tiny, and it is safer to say that Republicans
look a lot like our baseline group of independents. But we do wish to emphasize
that Republican identification does not appear to explain antiscience sentiment
once we control for other plausible explanations. The data largely bear out expec-
tations articulated in the third hypothesis (H3), although the magnitude of the
 Table 3
OLS Regression Models Predicting Deference toward Scientific Expertise

Mandatory Mandatory
Stem Cell Fetus Global Childhood AIDS Birth Legalizing Health Regulation Background Biotech Nuclear Animal Mandatory Gay
Overall Research Viability Warming Obesity Prevention Control Drug Use Insurance of Coal Check Foods Power Testing Vaccines Evolution Adoption

Republican 0.073 0.263 0.243 0.104 0.147 0.210 0.222 0.179 −0.037 0.211 0.010 0.575** 0.589*** 0.510** 0.485** 0.125 −0.063
0.065 0.181 0.194 0.188 0.170 0.149 0.176 0.201 0.206 0.185 0.209 0.195 0.168 0.194 0.173 0.197 0.224
Democratic 0.400*** 1.072*** 0.786*** 1.353*** 1.077*** 0.813*** 0.864*** 0.432* 1.597*** 0.978*** 1.018*** 0.583** 0.616*** 0.901*** 1.005*** 1.029*** 0.958***
0.060 0.170 0.185 0.175 0.156 0.138 0.164 0.188 0.193 0.174 0.195 0.184 0.156 0.181 0.161 0.185 0.207
Ideology −0.222*** −0.509*** −0.355*** −0.84*** −0.378*** −0.29*** −0.571*** −0.438*** −0.687*** −0.587*** −0.608*** −0.423*** −0.412*** −0.189** −0.391*** −0.616*** −0.495***
0.020 0.560 0.060 0.057 0.052 0.046 0.055 0.062 0.063 0.057 0.064 0.061 0.052 0.060 0.053 0.061 0.069
Female 0.066 −0.104 0.057 0.610*** 0.105 0.515*** 0.291* 0.132 −0.015 0.216 0.299 −0.459** 0.186 −0.583*** 0.074 −0.179 0.025
0.049 0.136 0.147 0.141 0.126 0.112 0.132 0.151 0.155 0.139 0.157 0.147 0.126 0.146 0.129 0.148 0.168
Under 30 years of 0.028 −0.04 −0.366* 0.644*** 0.171 −0.523*** 0.020 −0.406* 0.218 0.246 0.060 0.300 −0.528*** 0.270 −0.23 −0.08 −0.013
age 0.062 0.175 0.189 0.179 0.162 0.142 0.169 0.194 0.200 0.179 0.201 0.188 0.161 0.187 0.167 0.190 0.212
Over 64 years of −0.004 0.616*** −0.011 0.139 0.304 0.394** −0.211 −0.139 0.020 −0.294 0.287 0.055 0.340* 0.188 0.718*** −0.173 0.012
age 0.067 0.185 0.200 0.191 0.171 0.151 0.177 0.204 0.209 0.188 0.211 0.200 0.169 0.198 0.174 0.201 0.227
College degree 0.206*** 0.292 0.488** 0.105 0.368* 0.042* 0.151 0.234 0.580*** 0.300 −0.091 0.623*** 0.099 0.699*** 0.379** 0.649*** 0.648***
0.056 0.159 0.170 0.164 0.147 0.130 0.155 0.175 0.180 0.163 0.183 0.171 0.147 0.170 0.152 0.173 0.195
Attend church −0.045 −0.871*** −0.476** −0.453* 0.162 −0.011 −0.641*** −0.104 0.055 −0.016 0.100 −0.334 −0.067 −0.034 −0.247 −0.675*** −0.171
weekly or more
0.061 0.169 0.183 0.175 0.157 0.138 0.164 0.187 0.191 0.172 0.194 0.181 0.156 0.179 0.160 0.184 0.207
Bible is literal −0.175*** −0.518*** −0.541*** −0.301** −0.24** −0.308*** −0.453*** −0.628*** −0.469*** −0.536*** −0.174 −0.186 −0.286** −0.19 −0.087 −1.386*** −0.801***
word of God 0.037 0.102 0.111 0.106 0.095 0.084 0.100 0.113 0.116 0.105 0.117 0.110 0.094 0.109 0.097 0.112 0.127
African American 0.263** 0.313 0.175 1.107*** 0.707*** 0.121 0.387 0.554* 1.421*** 0.985*** 0.585* 0.131 0.318 0.671** 0.074 0.334 −0.16
0.086 0.230 0.248 0.236 0.210 0.187 0.224 0.254 0.257 0.237 0.261 0.249 0.211 0.245 0.216 0.252 0.287
Hispanic 0.279*** 0.501* 0.452* 1.158*** 0.644*** 0.312 0.738*** 0.382 1.185*** 0.771*** 0.731** 0.713*** 0.332 0.404 0.555 0.570* 0.253
0.071 0.205 0.218 0.213 0.190 0.166 0.199 0.226 0.231 0.210 0.235 0.219 0.187 0.219 0.195 0.223 0.252
Constant 0.802*** 10.009*** 9.065*** 10.118*** 8.856*** 9.823*** 10.197*** 9.004*** 8.350*** 9.413*** 9.586*** 8.564*** 9.729*** 7.276*** 9.273 10.281 8.321***
0.101 0.289 0.314 0.295 0.267 0.236 0.279 0.317 0.328 0.294 0.332 0.312 0.267 0.308 0.275 0.314 0.350
Valid cases 1,188 1,742 1,621 1,804 1,801 1,793 1,772 1,719 1,711 1,727 1,768 1,692 1,760 1,753 1,800 1,756 1,583
Adjusted 0.306 0.207 0.118 0.314 0.134 0.112 0.206 0.105 0.262 0.207 0.141 0.092 0.095 0.060 0.105 0.313 0.166
R-squared
Standard error 0.83 2.809 2.935 2.951 2.652 2.337 2.752 3.094 3.155 2.862 3.257 2.987 2.600 3.019 2.717 3.067 3.292

NOTE: Ideology is estimated on a 1–5 scale from extremely liberal to extremely conservative. Biblical literalism is estimated on a 1–3 scale from Bible is literal word of God and every word is true, to Bible is literal
word of God but stories are allegorical, to Bible is not literal word of God. The results display coefficients for the variables listed, with associated standard errors listed below.

29
*Significant at .05 level. **Significant at .01 level. ***Significant at .001 level.
30 THE ANNALS OF THE AMERICAN ACADEMY

positive effect associated with Democratic identification raises the possibility that
party “matters” more than we suggest as a predictor of scientific deference.
Beyond examining party and ideology, we also wanted to consider the impact
of religious belief on deference to scientific expertise (this was expressed in H2,
which posited that religiosity will decrease willingness to defer to science). The
two variables of interest in this regard are (1) biblical literalism and (2) church
attendance. Across our sixteen different issue areas, biblical literalism is consist-
ently negatively signed, meaning the belief that the Bible is the literal word of
God always reduces one’s self-rated deference toward scientific expertise.13
Moreover, this influence is statistically significant in all but four cases: mandatory
background checks, biotech foods, mandatory vaccines, and animal testing.
This effect is substantively impressive as well. The largest coefficient (by far)
is the −1.39 for evolution, which means minimum-to-maximum movement on
the three-category biblical literalism scale decreases deference to scientific rec-
ommendations on evolution 2.78 points on the 0−10 scale. The next greatest
impact is on gay adoption (a minimum-to-maximum shift across the literalism
scale decreases deference by 1.60 points), and then legalizing drug use (a mini-
mum-to-maximum shift in literalism decreases deference by 1.26 points). The
effects on less religiously salient issues are less dramatic. The coefficient for
biblical literalism in the overall deference model is −0.18, which produces a mod-
est minimum-to-maximum effect of −0.53 points.
The impact of religious faith is less powerful when we gauge it with self-
reported religious behavior. Here the variable of interest is the dummy for
church attendance—which is scored as 1 if the respondent goes to church once
a week or more—and it is negatively signed in all but three cases (childhood
obesity, health insurance, and mandatory background checks). However, while
the influence of church attendance on scientific deference is typically negative,
it reaches conventional levels of statistical significance only for the religious hot-
button issues of stem cell research, fetal viability, birth control education, and the
teaching of evolution. In substantive terms, church attendance elicits a maximum
effect of only −0.87 points on the 0−10 scale (for stem cell research).
Still, the overall models clearly support the hypothesis (H2) that religiously
observant individuals are less inclined to defer to science on public policy mat-
ters. The individual policy models suggest that this reticence is particularly strong
on issues such as evolution and gay adoption.
Other variables in the models also produced some intriguing effects. As men-
tioned earlier, we included dummy variables for age (one for those under
30 years of age and another for those 65 years of age and older) because it is
plausible that the younger, more tech-savvy generation might be more amenable
to deferring public policy decisions to scientists. As it turns out, age has inconsist-
ent effects on scientific deference. For example, compared to middle-aged peo-
ple, younger people are relatively likely to defer to science on global warming,
but are much less likely to do so for AIDS prevention, legalizing drug use,
nuclear power, and fetal viability. On the other side of the ledger, older people
are more likely than middle-aged people to defer to science for AIDS prevention,
stem cell research, nuclear power, and mandatory vaccines.
Partisanship and Attitudes Toward Science and Public Policy 31

Recall that we are particularly interested in the effects of education on atti-

tudes toward science and public policy. Indeed, we hypothesized (H4) that those
with a college degree are relatively likely to defer to science because they have
been socialized to the scientific method and the benefits of scientific inquiry. As
it turns out, those with a college degree are more likely to defer to scientific
expertise on fifteen of sixteen issues tested. More importantly, this tendency
holds at a statistically significant level (p < .05) for animal testing (coefficient of
0.699), evolution (0.649), gay adoption (0.648), biotech foods (0.623), mandatory
health insurance (0.580), fetal viability (0.488), mandatory vaccines (0.379), and
childhood obesity (0.368). The overall effect is positive and highly significant.
Finally, we controlled for the effects of race, ethnicity, and gender. More spe-
cifically, we find that African Americans are more deferential to science than
whites and Asians on every issue except for gay adoption. In fact, this tendency is
statistically significant in eight of fifteen positive cases. Most notably, the positive
impact of race on scientific deference is substantively large for mandatory health
insurance (+1.42) and global warming (+1.11). Similarly, Hispanics are more
deferential to scientific expertise than whites or Asians on all sixteen issues, with
enormous positive effects associated with global warming (+1.16) and mandatory
health insurance (+1.19). These findings are all the more impressive given that
we are controlling for both ideology and party identification, and indicate that
African Americans and Hispanics believe scientific opinion is critical, even on
issues that are commonly perceived as extremely political.
Although race and ethnicity produce consistent and larger-than-expected
effects, gender is not so strongly correlated with deference to science. For
instance, women are relatively more deferential to science on global warming,
AIDS prevention, and birth control. But women are relatively less deferential to
science on biotech foods and animal testing.

Conclusion
Recall that we posed four specific research hypotheses to guide this study. The
survey data provide strong support for three of the four. Namely, we find that (1)
the public generally defers to scientific expertise on a range of public policy
issues, (2) political ideology and religious beliefs influence one’s willingness to
defer to science, and (3) education is positively associated with deference to sci-
ence. The strength of support for these hypotheses does vary according to the
particular issue—with issues more directly conflicting with political or religious
belief systems eliciting decreased deference—as predicted.
The other hypothesis, that party identification does not have a significant inde-
pendent effect on deference to scientific expertise, has been the subject of most
popular inquiry and is the most interesting of our expectations in light of the data.
Both the simple means and the multivariate models demonstrate that Democrats
are relatively more likely to say they defer to scientific expertise. Contrary to
much speculation, this is true across all issue domains examined here, many of
32 THE ANNALS OF THE AMERICAN ACADEMY

which were included to parse out potential Democratic skepticism (mandatory

vaccines, nuclear power, etc.). The significance of Democratic identification as a
predictor of scientific deference, even controlling for other factors, is not entirely
consistent with our expectations.
On the other hand, the notion that Republicans are distinguished from others
by a tendency to reject scientific expertise seems far-fetched. Republicans are not
especially different from independents (and other non-Democrats) in their will-
ingness to defer to science. Indeed, controlling for ideology and a host of demo-
graphic factors, Republican identification is associated with an increase in the
deference to science rating for fourteen of sixteen issues. On only two issues,
mandatory health insurance and gay adoption, does being a Republican correlate
with a decreased willingness to defer to what science says. And even here the
effect is not statistically significant. This result comports with our expectation.
Our results suggest that scientific recommendations on public policy are taken
seriously by partisans of all stripes. Thus, what science has to say about an issue
appears to be a reasonable starting point for lawmakers or bureaucrats seeking to
forge consensus on a given issue. Of course, we are not arguing that everyone will
simply accept the edicts of science on every issue; in fact, it may be that defer-
ence erodes as scientists turn their attention to issues that require massive gov-
ernment regulation or challenge the edicts of faith. In a time of extreme
polarization, though, there is some reason to believe that science can offer com-
mon ground.
Our findings may also offer a specific path for strategists who are seeking to
mobilize or persuade Left-leaning voters: if you want to get Democrats on your
side, you may want to use scientific research to back up your policy positions. The
self-expressed willingness of those on the Left to defer to scientists indicates that
political arguments based on objective, scientific research might have a powerful
influence on opinion. Moreover, the opinions of scientists are not just important
to Democrats, per se; they are also important for key elements of the Democratic
coalition, such as blacks and Latinos.
As with any empirical study, the current study has limitations. First, these
results do not explain the causal pathway by which certain individuals reject sci-
entific advice for specific public policy prescriptions. It could be that their per-
ception of ideological bias on the part of scientists causes them to reject those
findings. It is also possible that a rejection of the scientific method in favor of
other frameworks (e.g. faith, gut reasoning) creates a strong negative perception
of scientific evidence, and opinions about the researchers are merely a consist-
ency-serving by-product. It could also be that the politicization of some of these
issues has led to a blanket degradation of scientists and scientific evidence that
stems beyond the issues that originally produced those attitudes. None of this can
be answered with a single survey, and it will require more explicit experimental
manipulation to tease out some of those answers.
Second, we did not test more basic aspects of self-interest that could affect
deference toward scientific recommendations. Most obviously, economic self-
interest may be an important factor. Someone working in the oil and gas industry,
for example, is likely to be relatively skeptical about scientific evidence on global
Partisanship and Attitudes Toward Science and Public Policy 33

warming or hydraulic fracturing. Self-interest could also come in the form of

being medically at-risk (for example, AIDS prevention or stem cell research),
which we assume would influence attitudes on scientific deference. We did
include items measuring income and occupation in this survey, but these offer
only broad gauges of economic self-interest and were excluded from the final
models. Alas, we did not include measures of individual health, which we deemed
too intrusive to be included. Perhaps we will take another run at these in future
research.
Third, despite the broad coverage of our policy battery, the issues were pre-
sented with an explicit lack of content in terms of what the specific scientific
policy prescriptions might be. There is no doubt that there is variance in respond-
ents’ knowledge of the scientific prescriptions regarding a particular policy area;
moreover, current research has not addressed whether one’s willingness to defer
to science declines once an individual knows that the scientific policy prescrip-
tion goes against his or her partisan, ideological, or religious conception of that
particular issue.
Despite these limitations, we think the range of specific issue items, political
attitudes, and demographic characteristics offered here represents a leap forward
in our measurement and understanding of public willingness to accept what sci-
entists have to say about public policy. Moreover, the data provide some encour-
aging signs, such as the finding that science remains credible in the eyes of the
public. Even on issues that touch on fundamental aspects of politics and morality,
science is a relevant touchstone for the vast majority of Americans.

Appendix
YouGov interviewed 2,153 respondents who were then matched down to a sam-
ple of 2,000 to produce the final dataset. The respondents were matched to a
sampling frame on gender, age, race, education, party identification, ideology,
and political interest. The frame was constructed by stratified sampling from the
full 2010 American Community Survey sample with selection within strata by
weighted sampling with replacements (using the person weights on the public
use file). Data on voter registration status and turnout were matched to this
frame using the November 2008 Current Population Survey. Data on interest in
politics and party identification were then matched to this frame from the 2007
Pew Religious Life Survey.
The matched cases were weighted to the sampling frame using propensity
scores. The matched cases and the frame were combined and a logistic regression
was estimated for inclusion in the frame. The propensity score function included
age, gender, race/ethnicity, years of education, and ideology. The propensity
scores were grouped into deciles of the estimated propensity score in the frame
and poststratified according to these deciles.
34 THE ANNALS OF THE AMERICAN ACADEMY

Notes
1. Scientists rank below member of the military (84 percent) and teachers (77 percent), but above
medical doctors (69 percent), engineers (64 percent), clergy (40 percent), journalists (38 percent), artists
(31 percent), lawyers (23 percent), and business executives (21 percent).
2. See Zipp and Fenwick (2006) for a contrary view.
3. In this study, we do not discriminate between different types of scientists (for example, engineers
versus university professors). This is undoubtedly an important distinction and one we would like to inves-
tigate in the future.
4. Furthermore, the rise of 24/7 cable news and the blogosphere has created an unending stream of
accessible means for skeptical (and even accusatory) viewpoints. Of course, history offers dozens of exam-
ples of science being met with forceful pushbacks. But the pervasiveness of media and Internet outlets
available today means that pushback is almost certain to occur and to receive lots of attention, especially
on issues where science crosses swords with religious faith or political ideology. Still, we doubt that the
status of scientists has been globally degraded, and instead expect this downgrading to be concentrated
within certain groups.
5. This is not to say that scientists do not ever have conflicting interests.
6. When Liu and Ditto’s (2013) findings are combined with the fact that liberals dominate the psycho-
logical sciences (the ratio of liberals to conservatives is about 10:1 in social psychology; Inbar and Lammers
2012), a serious question about “science” is raised: how much do ideological/moral beliefs of psychological
scientists color and distort the research questions posed and their eventual conclusions?
7. We conducted a pilot test in an October 2012 statewide survey of Texas.
8. For example, we did not include the generic and not terribly instructive “health care” item that ran
in the earlier study.
9. For the descriptives in Table 2 and the multivariate analysis in Table 3, we identify independents as
including leaners. Allowing leaners to be labeled as partisans does not substantively change the results.
10. We generally prefer dummy variables in our model because the posited relationships are not linear.
For example, an extra year of age or education is unlikely to produce a consistent effect on the deference
to science items. We are particularly interested in the impact of identifying as a Republican or Democrat
on deference toward science. The exception for us is ideology, which we believe is properly understood
here as a scaled attitude. Given our use of dummy variables, it is important to note that our baseline
respondent is independent, between 30 and 64 years of age, without a college degree, attends church less
than once a week, and is neither black nor Latino.
11. This principal components analysis relies on a varimax rotation method and extracts factors based
on eigenvalues greater than 1. Results are available from the authors upon request.
12. We are especially interested in whether the data supported our theoretical (but empirically crucial)
assumption that party and political ideology measure distinct things—the associative correlation, 0.39,
reassures us on this count.
13. The variable consists of three response categories: the Bible is the actual word of God and is to be
taken literally, the Bible is the word of God but not everything in it should be taken literally, and the Bible
is a book written by man and is not the word of God.

References
Berezow, Alex, and Hank Campbell. 2012. Science left behind: Feel-good fallacies and the rise of the anti-
scientific Left. New York, NY: Public Affairs.
Campbell, Angus, Philip Converse, Warren Miller, and Donald Stokes. 1960. The American voter. New
York, NY: John C. Wiley & Sons.
Converse, Philip E. 1964. The nature of belief systems in mass publics. In Ideology and discontent, ed.
David Apter, 206–61. New York, NY: Basic Books.
Delli Carpini, Michael X., and Scott Keeter. 1996. What Americans know about politics and why it mat-
ters. New Haven, CT: Yale University Press.
Partisanship and Attitudes Toward Science and Public Policy 35

Inbar, Yoel, and Joris Lammers. 2012. Political diversity in social and personality psychology. Perspectives
on Psychological Science 20 (10): 1–8.
Liu, Brittany S., and Peter H. Ditto. 2013. What dilemma? Moral evaluation shapes factual belief. Social
Psychological and Personality Science 4 (3): 316–23.
Mooney, Chris. 2012. The Republican brain: The science of why they don’t believe in science. New York,
NY: Wiley.
Tenney, Elizabeth R., Robert MacCoun, Barbara A. Spellman, and Reid Hastie. 2007. Calibration trumps
confidence as a basis for witness credibility. Psychological Science 18 (1): 46–50.
Westen, Drew, Pavel S. Blagov, Keith Harenski, Clint Kilts, and Stephan Hamann. 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): 1947–58.
Zaller, John R. 1992. The nature and origins of mass opinion. New York, NY: Cambridge University Press.
Zipp, John F., and Rudy Fenwick. 2006. Is the academy a liberal hegemony? The political orientations and
educational values of professors. Public Opinion Quarterly 70 (3): 304–26.