UNDERSTANDING THE DEBATE: THE LIMITS OF SCIENTIFIC KNOWLEDGE
By
Jason Anthony Kalmbach

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Political Science – Doctor of Philosophy
2013

ABSTRACT
UNDERSTANDING THE DEBATE: THE LIMITS OF SCIENTIFIC KNOWLEDGE
By
Jason Anthony Kalmbach
Policy debates feature conflicting causal arguments offered by political opponents in order
to explain ongoing events. This characteristic is present in the current policy debate over climate
change, with conflicting arguments emerging in an attempt to explain changes in global
temperatures. Despite assertions by scientists that climate change is anthropogenic, disagreement
remains within public opinion over the primary cause of climate change as well as the perceived
threat if change continues unchecked.
A two-step process is introduced as a way to understand the polarization within public
opinion. Utilizing a proprietary public opinion dataset, the analysis first considers how the public
comes to understand the causal arguments relative to climate change. The acceptance of causal
arguments, in turn, influence whether respondents are concerned about the phenomenon. In effect,
a two-step process exists, where the public must understand the causal arguments before
demonstrating an elevated level of concern for the problem.
The dissertation emphasizes the role of scientific knowledge and political values in shaping
public opinion. Results support the argument that scientific knowledge guides the public’s
understanding and acceptance of the causal arguments associated with climate change, with most
(but not all) high-knowledge individuals agreeing to even the most politically contested claims.
Political values, however, also guide the acceptance of causal arguments and moderate the effect
of knowledge. Once respondents accept the causal arguments offered by climatologists, however,
they demonstrate an elevated level of concern. The dissertation concludes with a discussion of the

implications of the analysis, with consideration given to whether the polarization within public
opinion can be mitigated through meaningful policy action.

ACKNOWLEDGEMENTS

I want to thank my dissertation committee, Chuck Ostrom, Richard Hula, Joshua Sapotichne, and
Sandy Marquart-Pyatt, for their feedback throughout the writing process. The dissertation has
evolved extensively, for the better, over the last two years with their feedback. Deserving particular
thanks as well is Petra Hendrickson and Jonah Ralston for their feedback and assistance in
developing various aspects of the argument enclosed in this manuscript.

iv

TABLE OF CONTENTS

LIST OF TABLES ...................................................................................................................... vii
LIST OF FIGURES ..................................................................................................................... ix
CHAPTER 1: INTRODUCTION ................................................................................................ 1
Climate Change Policy in the United States ........................................................................... 6
Public Opinion & Policy ......................................................................................................... 11
Sources of Climate Change Attitudes .................................................................................... 11
Media Effects ....................................................................................................................... 12
Self-Interest .......................................................................................................................... 15
Partisan Cues ....................................................................................................................... 19
Next Steps ................................................................................................................................. 21
CHAPTER 2: SCIENTIFIC ORIENTATIONS & THE LIMITS OF KNOWLEDGE ...... 22
A Framework of Public Opinion............................................................................................ 23
Sources of Belief Systems .................................................................................................... 26
Climate Change Context ..................................................................................................... 27
The Science Comprehension Thesis....................................................................................... 30
Development of Empirical Assessments ................................................................................ 34
Constructing the Modern Knowledge Assessment Questions ......................................... 35
Assessing Outcomes ............................................................................................................. 35
Predisposed to Science ............................................................................................................ 38
Values as an Alternative Framework .................................................................................... 39
Why Values Dominate Climate Change Models .................................................................. 41
Reconciling the Two Views ..................................................................................................... 43
A Moderating Effect ............................................................................................................ 44
Different Outcome Expectations ........................................................................................ 47
Summary .................................................................................................................................. 51
CHAPTER 3: DATA & OPERATIONALIZATION .............................................................. 52
Survey Data .............................................................................................................................. 52
Why Climate Change? ............................................................................................................ 54
Operationalizing Key Concepts ............................................................................................. 55
Scientific Knowledge ........................................................................................................... 55
Normative Science Views .................................................................................................... 59
Correlations.......................................................................................................................... 62
Means Analysis .................................................................................................................... 63
Political Ideology ................................................................................................................. 63
Outcome Measures .............................................................................................................. 65
Model Presentation & Next Steps .......................................................................................... 67
Summary .................................................................................................................................. 70
CHAPTER 4: ORIENTATIONS AND ACCEPTANCE OF CAUSAL ARGUMENTS ..... 72
v

Analysis: Greenhouse Effect .................................................................................................. 76
Substantive Effects .............................................................................................................. 80
Summary: Greenhouse Effect ............................................................................................ 86
Analysis: Fossil Fuels .............................................................................................................. 87
Substantive Effects .............................................................................................................. 91
Summary: Fossil Fuels ........................................................................................................ 96
Analysis: Solar Radiation ....................................................................................................... 97
Substantive Effects ............................................................................................................ 102
Summary: Solar Radiation ............................................................................................... 106
Discussion ............................................................................................................................... 107
Knowledge by Issue ........................................................................................................... 108
Normative Views ................................................................................................................ 110
Ideology .............................................................................................................................. 113
Ideology & Solar Radiation .............................................................................................. 115
Ideology & Interaction Effects ......................................................................................... 116
Summary ............................................................................................................................ 119
CHAPTER 5: A TWO-STEP CONNECTION TO CLIMATE CHANGE CONCERN ... 122
Re-Analysis of First Stage..................................................................................................... 126
Two-Step Model..................................................................................................................... 128
Summary ................................................................................................................................ 133
CHAPTER 6: DISCUSSION ................................................................................................... 136
Review of Empirical Results................................................................................................. 138
The Limits of Knowledge ...................................................................................................... 144
Additional Observations ....................................................................................................... 145
Connection to Literature ...................................................................................................... 147
Robustness.............................................................................................................................. 150
Temporal Component ........................................................................................................... 151
Policy Implications ................................................................................................................ 153
Generational Replacement ................................................................................................... 155
Next Steps ............................................................................................................................... 157
APPENDICES ........................................................................................................................... 161
Appendix A: Participation in the Climate Change Debate ............................................... 162
Substantive Effects ............................................................................................................ 166
Discussion ........................................................................................................................... 169
Appendix B: Interrelationship between Beliefs .................................................................. 172
Technology Defender Hypothesis ..................................................................................... 176
Summary ............................................................................................................................ 178
Appendix C: Toward a More Complete Model .................................................................. 181
Analysis............................................................................................................................... 184
Model Performance ........................................................................................................... 186
Summary ............................................................................................................................ 187
BIBLIOGRAPHY ..................................................................................................................... 190

vi

LIST OF TABLES

Table 2.1: Sample of the views associated with political and scientific orientations ............ 28
Table 2.2: Dimensions of culture ............................................................................................... 40
Table 3.1: Summary of Science News Survey .......................................................................... 53
Table 3.2: Variable coding & operationalization ..................................................................... 57
Table 3.3: Principal-component factor analysis results .......................................................... 60
Table 3.4: Cross-tabulation of normative science views.......................................................... 61
Table 3.5: Correlation matrix of science measures (Pearson’s r) .......................................... 62
Table 3.6: Demographic variation across measures of scientific orientation ........................ 64
Table 3.7: Hypotheses for first-stage analysis .......................................................................... 68
Table 3.8: Hypotheses for full two-step model ......................................................................... 70
Table 4.1: Acceptance of the greenhouse effect argument ...................................................... 77
Table 4.2: Percent correctly predicted for each model ........................................................... 78
Table 4.3: Probability of identifying greenhouse effect outcome by normative views of science
....................................................................................................................................................... 82
Table 4.4: Acceptance of the fossil fuels argument .................................................................. 88
Table 4.5: Percent correctly predicted for each model ........................................................... 89
Table 4.6: Probability agree with IPCC on fossil fuels by normative views of science ........ 93
Table 4.7: Rejection of the solar radiation counterargument ................................................ 98
Table 4.8: Percent correctly predicted for each model ......................................................... 100
Table 4.9: Probability of rejecting solar radiation counterargument by normative views of
science......................................................................................................................................... 104
Table 4.10: Summary of support for hypotheses ................................................................... 108
Table 4.11: Distribution of faith-in-science scores by outcome measure ............................. 112
Table 4.12: Faith-in-science by solar radiation response and level of knowledge .............. 113
vii

Table 4.13: Cross-tabulation of whether respondents reject solar radiation by ideology . 116
Table 4.14: Distribution of faith in science scores by outcome measure ............................. 118
Table 5.1: Total effects for the first stage of the two-step model .......................................... 127
Table 5.2: Direct, indirect, and total effects on climate change concern ............................. 129
Table 5.3: Summary of support for hypotheses ..................................................................... 134
Table 6.1: Differences between Hamilton and Kalmbach ..................................................... 149
Table 6.2: Age by select science measures .............................................................................. 156
Table A.1: Determinants of climate change debate participation ........................................ 165
Table A.2: Expected involvement in climate change debate by normative views ............... 168
Table A.3: Expected involvement in climate change debate by normative views ............... 169
Table B.1: Correlation matrix ................................................................................................. 173
Table B.2: Reliability coefficients ............................................................................................ 173
Table B.3: Distribution of additive index of climate change knowledge ............................. 173
Table B.4: Climate change knowledge by ideology................................................................ 174
Table B.5: Patterns in disagreement with climatologists ...................................................... 175
Table B.6: Climate change knowledge by science quiz score performance......................... 176
Table B.7: Determinates of individuals who believe in climate change but reject the solar
radiation argument ................................................................................................................... 178
Table C.1: Direct, indirect and total effects of climate change knowledge and concern ... 184
Table C.2: Further diagnostics ................................................................................................ 187

viii

LIST OF FIGURES

Figure 1.1: The two-step process ................................................................................................. 3
Figure 2.1: The two-step process, general framework ............................................................ 24
Figure 2.2: The two-step process for climate change beliefs ................................................... 29
Figure 2.3: The two-step process for climate change beliefs ................................................... 48
Figure 2.4: A two-step process for climate change beliefs, tested model ............................... 51
Figure 3.1: Distribution of science quiz scores ......................................................................... 56
Figure 3.2: Proposed relationships between selected orientations and acceptance of climate
change arguments ....................................................................................................................... 68
Figure 3.3: The two-step process ............................................................................................... 71
Figure 4.1: Proposed relationships between selected orientations and acceptance of climate
change arguments ....................................................................................................................... 73
Figure 4.2: Probability agree with IPCC by science knowledge for liberals (-5), moderates
(0), and conservatives (5) ............................................................................................................ 84
Figure 4.3: Average marginal effect of scientific knowledge by ideology .............................. 86
Figure 4.4: Probability agree with IPCC by science knowledge for liberals (-5), moderates
(0), and conservatives (5) ............................................................................................................ 94
Figure 4.5: Average marginal effect of scientific knowledge by ideology .............................. 95
Figure 4.6: Probability agree with IPCC by science knowledge for liberals (-5), moderates
(0), and conservatives (5) .......................................................................................................... 105
Figure 4.7: Average marginal effect of scientific knowledge by ideology ............................ 106
Figure 4.8: Probability of agreement with IPCC by science knowledge for all three outcome
measures..................................................................................................................................... 109
Figure 5.1: Tested two-stage model ......................................................................................... 124
Figure 6.1: A two-step process for climate change beliefs .................................................... 137
Figure 6.2: Replication of climate change beliefs from the Pew Research Center ............. 152

ix

Figure A.1: Proposed relationships between selected orientations and attention to climate
change (passive/active).............................................................................................................. 164
Figure A.2: Expected level of engagement in information acquisition activities ................ 168
Figure C.1: Basics of the science comprehension thesis ........................................................ 181
Figure C.2: Multi-stage model ................................................................................................. 183

x

CHAPTER 1: INTRODUCTION
In the fall of 2010, a group of leading scientists and climate change advocates met to
discuss the ramifications of “Climategate” for the scientific community in a webinar sponsored by
the American Association for the Advancement of Science (AAAS).1 During the discussion, the
panel referred to growing polarization within the American public about whether climate change
is occurring. While the public was already polarized on the issue to some extent, the gap between
“believers” and “contrarians” grew during the first decade of the 21

st

Century.2 Reports from

climatologists working with the Intergovernmental Panel on Climate Change (IPCC) suggest that
there is a “very high confidence that the net effect of human activities … has been one of warming”
– a claim that resonates with “believers,” who, for a variety of reasons to be discussed, agreed with
the scientific findings (IPCC 2007, p. 37). The “contrarians” did not respond to the information
within the IPCC report, with the allegations of fraud associated with “Climategate” perhaps
reinforcing skeptical attitudes toward climatologists working with the IPCC.
The AAAS panel was alarmed by the growing disconnect between the “best” scientific
information and beliefs of the public. In response, the panel discussed strategies to promote both
the science supporting arguments for anthropogenic climate change and remedial policies that
would mitigate greenhouse gas emissions. To generate increased policy support, the panel

1

“Climategate” involved the unauthorized release of electronic communications between
climatologists. In these communications, the climatologists referred to using statistical tricks in
their research while also criticizing climate change skeptics. For some, these communications
called into question the objectivity of climatologists. The event corresponded with a decrease in
public belief in climate change, as well as a loss of trust in climatologists (Leiserowitz et al. 2013).
2
“Believers” refer to individuals who readily accept climate change as ongoing and anthropogenic,
while “contrarians” is a term often used by climatologists to describe individuals who call the
scientific information into question.
1

discussed the importance of making climate change a real, meaningful concept to the American
population. This meant informing the public about the merits of scientific arguments and
connecting the effects of climate change to tangible phenomena in their everyday lives such as
pollution and air quality. Both actions, however, assume the growing disconnect between
climatologists and public opinion is the result of an unaware, scientifically unsophisticated public.
If the public were simply better informed so that they understood the effects of climate change,
then perhaps the polarization within public opinion could be reversed or at least minimized.
This argument is referred to in recent literature as the science comprehension thesis (Kahan
et al. 2012a).3 These arguments suggest the polarization within public opinion could be mitigated
by implementing policies to educate the public on the causes and consequences of climate change.
Emphasis, then, should be placed on presenting the science to the American public “in plain
English” to simplify the information and arguments presented by the scientific community. Recent
studies, however, view the thesis with a degree of skepticism. As an alternative, these studies favor
theoretical frameworks that emphasize values and risk perceptions as determinants of climate
change beliefs (e.g., Kahan et al. 2011). Collectively, these alternative frameworks that emphasize
values are referred to as the value-centered thesis.
In this dissertation, I argue that the relationship between scientific knowledge and climate
change beliefs is best conceptualized as a two-step process. This process envisions “problem
understanding” as an intermediate step between scientific knowledge and more specific beliefs
about climate change, such as concern and perceptions of a scientific consensus. The first stage of

3

The traditional science knowledge literature refers to the science comprehension thesis as the
“deficit model” (Irwin and Wynne 1996; Sturgis and Allum 2004). The core arguments, however,
are the same: a knowledge gap is responsible for public attitudes and beliefs that are inconsistent
with the scientific community.
2

the process entails a direct relationship between foundational understanding of science and climate
change beliefs (or what might be called climate change knowledge). That is, a basic understanding
of molecules, DNA and physics provides a foundation from which individuals can learn about
more complex scientific arguments such as the causes of climate change. Consistent with the
science comprehension thesis, the expectation is that high-knowledge individuals understand the
causal arguments about climate change in a manner consistent with the scientific community, or
more specifically, climatologists.
The second step extends the analysis to problem recognition. That is, is the public
concerned about climate change? Once individuals understand climate change in a manner
consistent with climatologists, then they can determine whether the problem constitutes a threat.
This second stage conceptualizes a direct relationship between problem understanding and
problem recognition.
Figure 1.1: The two-step process
Scientific
knowledge

Beliefs about
problem

Problem
recognition

This two-step conceptualization, diagramed in Figure 1.1, is offered as a means to reconcile
the science comprehension and value-centered theses. Prior research suggests each has a degree of
merit, but neither research stream has properly recognized differences within the analyses or
established the boundaries under which scientific knowledge correlates with climate change
beliefs. This is because proponents of the science comprehension thesis have not assessed the
thesis in a manner that is consistent with the causal mechanisms and expectations of the thesis
itself. Researchers that focus on values inappropriately use the science comprehension thesis as
part of a “straw man” argument in order to support arguments championing values as the primary

3

determinant of climate change opinions. The two-step process presented here is an attempt to
assess the core argument of the science comprehension thesis and balance those arguments against
research that emphasizes values.
Throughout the analysis, individuals are conceptualized as possessing a variety of
orientations that predispose individuals to accept or reject propositions offered by climatologists.
Two orientations of interest in this analysis are political- and science-based orientations, both of
which are seen as drivers of climate change beliefs. It will also consider interactive effects between
these orientations to assess what happens when individuals hold conflicting orientations. Conflict
is expected because scientific information is not always compatible with political predispositions.
If certain political orientations are given preference over science-based orientations, an interactive
effect should emerge – one that shows high-knowledge individuals agreeing with climatologists at
different rates, depending on value preferences.
The forthcoming analysis supports the argument that a quality scientific education (i.e., an
orientation toward science) leads the public to accept the causal arguments offered by the scientific
community, which is the key outcome measure that proponents of the science comprehension
thesis have overlooked. This relationship materializes because individuals have the capacity and
skillset to process and evaluate the competing causal arguments that emerge from both
climatologists and their most vocal critics. This is, arguably, the essence of the science
comprehension thesis. Qualifications are in order, however. Political ideology moderates the
relationship between high-knowledge individuals and beliefs about climate change, with highknowledge liberals accepting the causal arguments from climatologists at higher rates than highknowledge conservatives.

4

More importantly, the analysis supports the two-step conceptualization. Evidence of a
direct relationship between foundational scientific knowledge and specific climate change beliefs
emerges. Once individuals understand climate change in a context similar to what climatologists
working with the IPCC would suggest, a direct, positive relationship is observed between climate
change knowledge and elevated levels of concern. A “profile” of an individual concerned about
climate change is one who accepts the causal arguments from climatologists and is knowledgeable
about science. Given the two-step process, the relationship between basic scientific knowledge
and climate change concern, then, is perhaps best understood as an indirect relationship.
Assessing the merits of the science comprehension thesis is essential to understanding the
prospects for remedial policies, such as those noted by the AAAS panel. If public opinion is
dominated by political orientations, then the polarization within public opinion on climate change
might appropriately be construed as permanent. However, if multiple orientations emerge, and if
there is some “give and take” between an individual’s conflicting views, then perhaps the
polarization can be mediated through purposeful action.
The discussion proceeds as follows: the remainder of this introductory chapter will discuss
the status of climate change policy development within the United States, followed by a proposal
that the lack of policy advancement in the United States, compared to other countries, is the result
of polarization within American public opinion. Acknowledgement is then given to the larger
literature about attitude formation before turning attention in Chapter 2 toward both the science
comprehension and value-centered theses. The second chapter develops multiple hypotheses in an
attempt to identify a proper understanding under which foundational scientific knowledge might
lead to changes in public opinion. The third chapter summarizes the data and offers supplementary
details about the construction of the science knowledge index. The fourth chapter offers an

5

empirical assessment of the first stage of the two-step process. Knowledge should help the
citizenry understand the causal arguments coming from the scientific community. This relationship
materializes because knowledge provides a foundation and skillset to talk and acquire information
related to science issues (see Appendix A). The fifth chapter looks specifically at concern – the
typical outcome variable in numerous climate change studies – and specifies a two-step
relationship under which a basic foundational understanding of science might lead to higher levels
of concern. Finally, the last chapter reviews the empirical results and their implications for policy
change. It concludes by specifying a research agenda that extends the research presented here.
Climate Change Policy in the United States
The United States and other countries demonstrate a high degree of faith in the “miracle of
science” because of the perceived benefits derived from a scientifically knowledgeable public,
such as the discovery of advanced technologies that will improve society’s quality of life. The
perceived economic benefits of technological advancement are sufficient to drive developing
countries to focus their energy on building an internal capacity to pursue scientific advancements
(Alberts 2011; Lawler 2011). Given this reverence toward science, one might expect countries to
adopt policies designed to correct problems identified by scientists such as mercury-laced water
supplies, acid rain and even anthropogenic climate change. This is not the case, however: several
notable discrepancies exist. Despite being a global leader in science over the years, the United
States continually fails to pass meaningful legislation – in the eyes of climatologists – to mitigate
the effects of greenhouse gases in the eyes of climatologists (e.g., Schneider 2010). Meanwhile,

6

the United States’ counterparts in Asia and Western Europe ratified the Kyoto Protocol and have
taken steps to implement a cap-and-trade program for carbon emissions.4
These divergent policy responses are of note given the relative consensus among
climatologists that the Earth is indeed warming due to human activity. One study (Oreskes 2004)
suggests 75% of scholarly research either directly or indirectly supports the consensus view of the
IPCC that climate change is anthropogenic. The remaining portion of scientific research takes no
position on the issue of anthropogenic climate change. Another survey – directly of scientists and
their opinions – finds 84% of scientists see global warming as the result of human activity (Kohut
et al. 2009).5 The same study, however, found only 56% of the public claimed scientists agree
global warming is the result of human activity.6
This public-expert disconnect, however, should be expected in some situations, contingent
on the salience of the issue. When issues are complex and when the solution (or range of solutions)
threatens core political values, there should be little expectation that policy outputs match the best
available information. Because of value differences, competing interpretations of scientific

4

One related issue where the global community did pursue policy action is in the context of
chlorofluorocarbons (CFCs). The Montreal Protocol, ratified by all member states of the United
Nations, represented a collective attempt to reduce CFCs, which were blamed for destroying the
ozone layer. The issue is fundamentally different from climate change on two fronts. First, ozone
depletion has imminent health consequences. This increased the salience of the issue for the public,
making it an issue that demanded attention. Second, there were technological solutions available
or on the immediate horizon that would mitigate CFC emissions (Schneider 2010; Pielke 2009).
5
The Pew Research Center conducted the survey in cooperation with the AAAS. Pew drew 9,998
names from the AAAS membership rolls to solicit participation in the survey. Approximately
2,500 scientists took part in the survey.
6
This is not to say notable scientists do not disagree with the arguments of the IPCC. Richard
Lindzen (2008) and Nir Shaviv (2006) argue the IPCC has overestimated the Earth’s sensitivity to
greenhouse gases and conclude natural variations are responsible for any observed change.
7

information emerge. Political leaders and media personalities attempt to frame the information in
a manner consistent with their values (Jones and Baumgartner 2005). Because of this prolonged
struggle, a gap emerges between existing policies and the additional remedial policies that
advocates contend are necessary in order to mitigate the problem.
Conversely, when salience is low, information is more accurately used in policy debates
(Mucciaroni and Quirk 2006). These are largely issues where bipartisan support for an action
exists. Bipartisanship, in turn, exists when there is no value-conflict among partisans. Thus, issues
with low-salience are areas where policy change can be anticipated. In this sense, science and
policy are not entirely disconnected.
Climate change, however, is not a low-salience issue. A combination of interest group
competition, continuous political commentary, and a sequence of alarmed discoveries work to
keep the issue at the forefront of the public’s consciousness. The range of solutions, which involve
incentivizing or enforcing behavior change, lead to value-conflict. Many of the remedial policies
proposed (which focus on greenhouse gas reductions) would reshape the existing energy market
and are likely to have notable effects on both lifestyles and the price of consumer services (Brown
and Sovacool 2011). In short, extensive government regulations would be required to achieve
significant reductions in greenhouse gases. Historically, such extensive regulatory oversight is
often contested in the American political system.7 Individuals valuing limited government and
personal choice will view arguments for new regulatory policies with greater skepticism. However,
other individuals value environmental protection and embrace the precautionary principle. In this

7

The exception, perhaps, is during times of perceived crisis, such as the New Deal under the
President Roosevelt in the 1930s or the explosion of federal environmental regulations in the
1970s.
8

way, scientific information resonates with the ideological orientations of decision-makers and the
public. As such, the ideological divide that occurs in the context of climate change should not be
terribly surprising given the political polarization in America (Abramowitz and Saunders 2008;
but see also Fiorina et al. 2005).
Resistance to remedial policies can also occur because elected leaders will protect their
electoral district. If legislators wish to be re-elected, they must look out for the interests of their
district (Fenno 1977; Mayhew 2004). Many districts rely heavily on the fossil-fuel industry, such
as oil-, coal-, and gas-mining operations. These operations tend to overlap with conservative
districts (e.g., in Texas, Oklahoma, Wyoming, and North Dakota) or districts that prefer
conservative Democrats (e.g., in West Virginia). These districts may face economic hardships if
new environmental regulations require greater use of renewable technologies such as wind and
solar. What appears to be an ideological divide among decision-makers in Washington may
alternatively be the result of legislators defending the economies of their respective districts
(Roberts). Political tension and legislator self-interest, then, make it difficult to align policy with
what might be perceived as the best available information.
Because of these differences, there has been little advancement in climate change policy
since it became a prominent issue for climatologists in the 1980s. The last notable air policy passed
by the United States Congress occurred in 1990, in the form of amendments to the Clean Air Act.
This legislation implemented a cap-and-trade policy on sulfur dioxide emissions but did little to
address rising greenhouse gas emissions. The Environmental Protection Agency (EPA), however,
has considerable discretion to implement remedial policies, particularly after the United States
Supreme Court in Massachusetts vs. EPA (2007) interpreted the language of the Clean Air Act to
permit the EPA to regulate greenhouse gas emissions. At the time of this writing in mid-2013, the

9

EPA under President Obama has begun to initiate plans the regulate greenhouse gases more
stringently.
Over the past few years, the EPA has taken small steps toward enacting policies that would
mitigate greenhouse gas emissions. Two policies of note include an increase in fuel economy
standards (Corporate Average Fuel Economy standards, determined in cooperation between the
Department of Transportation and the EPA) and new regulations on power plants (Mercury and
Toxic Air Standards, or MATS). While MATS’ primary focus is on mercury emissions, it also
includes restrictions on particulate matters, which have varying degrees of impact on climate
change. The EPA’s cost-benefit analysis anticipates MATS will reduce carbon dioxide emissions
from coal-fired power plants by 1% annually (2012b) while the CAFE standards will lower the
United States’ total carbon emissions by 4% by 2030 (EPA 2012a). While these are significant
efforts, they fall short of the Kyoto Protocol’s call for an 18% reduction in carbon emissions from
a 1990 baseline.
The bulk of climate change policy in the United States, however, has occurred at the state
level (Gamkhar and Pickerill 2012; Rabe 2010). In the New England area, several states have
pursued a regional cap-and-trade program referred to as the Regional Greenhouse Gas Initiative
(RGGI) while 29 states have pursued renewable portfolio standards that require energy sold within
the state to be produced by varying mixtures of renewable technologies. Meanwhile, California
has used its exemption within the Clean Air Act to pursue higher, more rigid air quality standards
than those set by the EPA for the rest of the nation. One of the themes of state variation is that,
reflective of the political characteristics of the United States, liberal states often lead in terms of
adopting climate change policies (Liang and Fiorino 2013; Matisoff 2008).

10

Public Opinion & Policy
Multiple avenues exist to understand this lack of policy development within the United
States. Interest groups, mass media outlets, and political elites all receive and interpret scientific
information. An additional explanation that cannot be ignored is the role of public opinion in the
policy process. In a democratic society, elites who are elected by the public make decisions. If
decision-makers deviate too far from the preferences of their constituents, voters potentially hold
them accountable in the next election cycle. Indeed, if reelection is a primary goal of legislators as
Mayhew suggests, then decision-makers should approve policies that conform to the opinions of
the public. This link between public opinion and Congressional behavior has been termed
responsiveness (Achen 1978), with scholars tending to find support for the idea that policy outputs
follow public opinion (Page and Shapiro 1983; Wright et al. 1987). This relationship occurs
because both public opinion and elite behavior tend to move in the same direction (Ansolabehere
et al. 2001; Griffin 2006), with legislators adapting to changes in the opinions of their constituents
(Stratmann 2000). This relationship occurs both at the federal and state level, with state policy
scholars estimating that, on average, policy and public opinion are aligned 48% of the time across
all states (Lax and Phillips 2012). A meta-analysis of research assessing the connection between
public opinion and policy adoption suggests “public opinion affects public policy three-quarters
of the time” (Burstein 2003). Scrutinizing public opinion, then, is one way to understand the lack
of policy development within the United States.
Sources of Climate Change Attitudes
While the discussion here is focusing on science and political orientations, it is useful to
acknowledge that there other concepts said to influence public opinion – both in general and with
respect to climate change. What is important to note is that scientists are not alone in offering

11

information to the public. Media outlets, political parties and other trusted personalities assess the
information from scientists and offer their own interpretation of the information to the public.
Thus, it is difficult for new information to bring about belief change precisely because more often
than not, groups are consulting with others and offering their own interpretation on the
information. Thus, beliefs may or may not be consistent with the original information as offered
by scientists. Despite the prestige afforded to science, the ability of scientists to influence public
opinion (and subsequently policy) is not guaranteed. Although Miller (1983) reports the general
public views scientists as the group most capable of solving complex societal problems, there
appears to be a partisan dimension as well. Partisans are more likely to follow party leadership
than defer to experts, although non-partisans appear more likely to align their views with those of
experts (Kuklinski et al. 1982). While the purpose here is not to develop or test an exhaustive list
of factors that may influence opinions, a few are worth mentioning since there is an opportunity
for non-experts to mute the influence of scientists and thus increase the expert-public disconnect.
Media Effects
First, the mass media, whether through the Internet, newspapers, or radio talk shows, have
the capability of “priming” information. Under this conceptualization, the media are a powerful
influence that can shape beliefs by controlling the amount or degree of information made available
to the public. If a media outlet wants to make the public aware of an issue, it can bombard audiences
with news about the particular issue. If an issue is consistently discussed, then there is more
information available for the public to develop views based on that information (Krosnick and
Kinder 1990; Zaller 1992; Barker and Knight 2000). While advocates, such as some climatologists,
may be involved in “hammering home” their research to the public through the media, the above
studies suggest it is just as easy for a contrarian to fight against climatologists by delivering a

12

strong, consistent message that questions the scientific evidence. Generally, exposure to science
via media reports appears to correlate with positive feelings about scientists. However, exposure
to negative images of science via the media correlates with increased levels of reservation about
scientific advancements (Nisbet et al. 2002).
In addition to flooding the information environment with news and letting the public absorb
the content, the media (as well as other elites) can frame issues in a particular manner that stresses
certain characteristics of a story that can alter public support for a given issue. Under this
conceptualization, a single belief held by an individual is formed by multiple underlying beliefs.
These underlying views are weighted by some unknown factor. By emphasizing some themes over
others, media outlets can alter the weights associated with a given belief. This shift in weights
produces the observable changes in public opinion during framing experiments (Chong and
Druckman 2007).8
To illustrate in the context of climate change, individuals hold some baseline opinion about
the tradeoffs between environmental regulations and economic prosperity. Framing maneuvers by
media personalities emphasize one of these two aspects when discussing remedial policies that
would reduce greenhouse gas emissions. A frame that emphasizes the cost of a cap-and-trade
program on carbon emissions pushes receptive listeners to place considerably more weight on
economic prosperity, thus shifting some public support away from the cap-and-trade program.
Alternatively, emphasizing the dangers of climate change can push some (but not all) media
consumers toward support for improved environmental regulations. These frames thus alter the

8

The framing literature argues the context of the story influences attitudes, while the priming
literature refers to the frequency with which the story is presented to the public.
13

weights associated with the underlying beliefs used by individuals to make decisions about climate
change, such as whether climate change is anthropogenic.
Framing does have its limitations, especially when media consumers question the
credibility of the news source or the quality of the frames (Druckman 2001). If there is no trust in
the person doing the framing, the frame is likely to have a minimal impact. Generally, scientists
are seen as trusted sources of information, although even then that trust varies by who exactly is
carrying out the research. For instance, Critchley (2008) argues the public is more likely to trust
public scientists compared to corporate scientists. Greater trust in scientists leads to more support
for scientific research. In the context of climate change, there is considerable polarization in the
level of trust held for the IPCC – especially after “Climategate” (Leiserowitz et al. 2013). Some
individuals, then, will inevitably tune out the work of scientists, and the resulting arguments, from
the IPCC.
In fact, divergent messages have emerged from the media in the context of climate change
in an attempt to frame the issue in a preferred manner. Conservatives, especially, have made efforts
to frame the debate in a manner opposed to climatologists. A study of the divergent views offered
by conservative think tanks suggest a coordinated theme, consisting of two claims: (1) “the science
of global warming appears to be growing more and more uncertain; ” and (2) “the harmful effects
of global warming policy are becoming increasingly certain” (McCright and Dunlap 2000, p. 499).
These arguments are inconsistent with those of the IPCC, which emphasize areas of scientific
consensus and discuss the consequences of inaction.
The idea of trade-offs from this framing discussion is worth emphasizing. Pielke (2010)
refers to the perceived trade-off between economic prosperity and environmental protection as the
“iron law of climate change.” Whether the trade-off is real or not, advocates must be prepared to

14

respond to claims by opponents that a new policy will adversely affect the economy. This is even
though a majority of the public appears to think economic success and environmental protection
(broadly defined) are not mutually exclusive (Guber 2003). Still, a specific, targeted frame that
emphasizes economic costs does appear capable of swaying public opinion. In the context of
expanding nuclear power and wind energy, frames emphasizing economic costs have rallied the
public against the expansion of these alternative technologies (Pralle and Boscarino 2011).
Framing is also an issue in the context of the survey questions used to explore opinions of
climate change. One study (Schuldt et al. 2011) finds, through a survey experiment, that whether
the problem is referred to as “global warming” or “climate change” produces dramatically different
results. Republicans are more likely to agree that the problem is real if it is referred to as climate
change rather than global warming. This pattern was also found to reflect the way conservative
think tanks discuss the issue, with references to the problem as climate change rather than global
warming. No such differences emerge with Democrats. Thus, part (but not all) of the polarization
within public opinion might be the result of survey wording.
To reiterate, how issues are framed can shape how the public approaches the issue.
Conservatives are strategically framing the problem in a manner that is inconsistent with the IPCC,
which in turn can potentially shape broader public opinion. It is important to note that an
examination of the public’s level of concern over a nine-year period (2002 to 2010) does not find
a substantively powerful relationship between media activity and fluctuations in climate change
concern (Brulle et al. 2012).
Self-Interest
A second path worth noting is self-interest. Science can be complex, and something
“special,” such as self-interest, might be required in order for the public to align their views with

15

scientists. This is especially true when the problem is distant and invisible, as is climate change.
The political science literature has generally found little or weak evidence of self-interest
correlating with public attitudes (Sears and Funk 1990). Symbolic attitudes, such as partisanship,
are argued to be the root source of attitudes in all but a few cases. In their review of the literature,
Sears and Funk argue self-interest contributes to beliefs under only three circumstances: (1) when
high levels of magnitude and clarity make the personal stakes clear, (2) when the consequences of
the policy are ambiguous, and (3) when politicians press, or perhaps prime, the self-interest issue.
Furthermore, they suggest these relationships are minimal and might even be exaggerated due to
survey design issues. Follow-up work supported these conclusions, noting “neither greater political
sophistication or emotions, nor social identifications or political values, nor perceiving an issue to
be a serious national problem, have been sufficient to make self-interest salient” (Lau and Heldman
2009, p. 527-528).
The conditions of influence found by Sears and Funk (1990) suggest that if self-interest
has any potential to promote congruence between scientists and the general public, such effects
are most likely be found when considering issues where the health or safety of the population is at
stake. When scientists raise alarms about consumer products causing cancer or the dangers of toxic
fumes, they send a signal to the population that their well-being is endangered and that
precautionary actions should be taken. However, the political battle to interpret the information
can muddle the signal received by the population about whether a danger truly exists. Despite
scientific studies citing smoking tobacco as a cause of lung cancer, segments of the public have
bought into the counter messages from the tobacco industry, (which, naturally, has denied the
claim), and continue to smoke.

16

Some limited evidence suggests – at least in the context of environmental policy – that a
form of self-interest can alter individual beliefs. This evidence comes from a specific area of
research that looks at whether local environmental conditions correlate with environmental views.
Directly experiencing poor environmental conditions helps individuals recognize environmental
problems and align their beliefs with scientists. The experience makes the science “real” for them,
giving them a context that helps them adjust their beliefs accordingly.
When looking at climate change, both Hamilton and Keim (2009) and Egan and Mullin
(2012) find a pattern of regional variation in beliefs that correlates with the weather conditions
those regions experience. That is, warmer winters increase the probability of respondents
perceiving a personalized effect from climate change. In this situation, the observed temperature
provides a signal to individuals that climate change is a meaningful phenomenon with real
implications. The analysis is expanded in a separate article where the authors find that place effects
are not just limited to beliefs about climate change but also speak to larger environmental
perceptions (Hamilton et al. 2010). However, it has been suggested “experiential learning” is
limited to individuals not already engaged in the climate change debate or committed to beliefs on
climate change (Myers et al. 2013)
Borick and Rabe (2010) ask why individuals believe climate change is real. They find that
experiencing local weather phenomena may facilitate affirmative climate change beliefs. In this
instance, residents of Mississippi were more likely to report violent hurricanes as the source of
their beliefs regarding climate change. Furthermore, the occurrence of hurricanes seems to have
encouraged Republicans in the state to overcome their partisan predispositions by leading them to
believe in climate change at rates higher than Republicans in other states.

17

Further illustration comes from Brody et al. (2008), who find detectable variations by
community when considering perceptions of climate change risk. Localized measures, such as
residing in a floodplain and experiencing economic damage from natural events, produce
variations in the level of the public’s perception of environmental risks, even after controlling for
demographic and partisan measures.
Disputed evidence exists that the public is capable of identifying their own self-interest
when it comes to more subtle problems such as air quality.9 Work by Jacquemin et al. (2007) finds
a positive correlation between self-reported annoyance with outdoor air pollution and actual
measures of air quality (particulate matter and sulfur) in European communities. This suggests that
individuals are capable of gauging the quality of their air, whether through independent means or
relying on reports from European environmental regulatory agencies. Brody et al. (2004),
however, found no relationship between perceptions of air quality and actual pollution levels in a
study focusing on Texas. Furthermore, Brody and Zahran (2007) note the wording of the question
used by Jacquemin and colleagues adds a degree of bias to the analysis. Consequently, it remains
unclear as to whether individuals can identify whether there are pollution issues with their local
air quality.
Still, it is worth recognizing that individuals are not ignorant when it comes to the dangers
of burning fossil fuels. Outdoor, rural communities are generally associated with a pristine,
uncontaminated environment. Power plants, factories and congested streets – typical of urban
settings – are not associated with clean air. At the very least, individuals seem aware of the

9

Compared to thoroughly understanding all the risks associated with climate change, it might be
easier for individuals to identify areas with substandard air quality. Smokestacks, smog, and foul
odors provide simple cues that can be used to gauge local air quality.
18

potential negative effects associated with power plants operating on fossil fuels (Ansolabehere and
Konisky 2009). This suggests individuals maintain at least a minimal capacity to identify problems
in their community and recognize, based on self-interest, that such conditions are undesirable for
their health and well-being. It also underscores the importance of the ability of individuals to
personally experience environmental problems when evaluating the merits of specific causal
arguments (Stone 2002).
Partisan Cues
One final concept is political partisanship, which functions as a heuristic utilized by the
public when forming opinions on contemporary policy issues. Partisanship is a social identity, in
that individuals identify with the party that best matches their interests or “speaks” to them (Green
et al. 2002). One who places a premium on moral values or identifies with moral issues will favor
the Republican Party under this conceptualization. However, this does not mean that individuals
care about or are concerned with all the views of the party. In some areas, an individual may have
no initial preferences on a given issue. When they need to make decisions (or answer a survey
question), they look at the stance of their preferred party on the issue. Since the party leadership is
trusted, partisanship functions as a heuristic and guides individuals toward the views they “should”
adopt as a member of that party. They drift toward the party’s position because they trust the party
leaders to offer a valid interpretation. Self-identifying Republicans who are not concerned about
the environment, for instance, are not going to spend time attempting to understand climate change.
Instead, they look to the party and receive a reinforcing message that anthropogenic climate change
is either false or at least highly contested among scientists (Feldman et al. 2011; McCright and
Dunlap 2000). Conversely, Democrats with little knowledge or interest in science receive a
reinforcing message from party leaders that anthropogenic climate change is ongoing. Thus,

19

acceptance or rejection of climate change beliefs comes from a signal provided by trusted political
elites, not necessarily from one’s level of knowledge or value preferences.
Evidence seems to suggest that as political parties adopt clear positions on an issue, the
public molds their opinions to match their preferred party’s platform (Carmines and Stimson
1986). This is one potential explanation for the increased polarization between Republicans and
Democrats regarding beliefs about climate change (Dunlap et al. 2001; Dunlap and McCright
2008; McCright and Dunlap 2011). Both Republicans and Democrats battled for the “green vote”
in the 1970s, but ultimately President Nixon was not willing to go as far as some of the citizenry
desired. Over time, Republican elites developed an anti-regulatory approach to environmental
issues while Democrats incorporated environmentalists into their party’s base. Today, the
Republican Party rejects the idea that corrective policies, such as cap-and-trade, are required to
mitigate carbon emissions because the dangers of climate change are either false or exaggerated
by scientists. Meanwhile, the Democratic Party has long been seen as the party of environmental
concern.
In short, a strong Democrat does not need to be familiar with the science in order to support
the position of his or her party, nor does a Republican need to know what the IPCC says in order
to reject claims that climate change is anthropogenic. Partisans with no opinion or knowledge
regarding climate change can turn to their preferred party for signals about what position they
should adopt on the issue. The use of political heuristics is consistent with Zaller and colleagues’
(Zaller 1992; Zaller and Feldman 1992) conceptualization of public opinion, where individuals
have a tendency to mimic attitudes they are familiar with from the information environment. As
such, individuals are likely to rely on other sources of information besides expert commentary.

20

Next Steps
In the next chapter, I pursue a theoretical framework to identify a more accurate
understanding of the relationship between scientific knowledge and climate change beliefs. The
discussion will first offer a framework to understand public opinion. It then elaborates on the
historical development of the science comprehension thesis and the expected role of knowledge in
the public’s decision-making process. The second step explores the limits of scientific knowledge
and develops an understanding for why information – even information offered by the best and
brightest scientists – may be dismissed. The discussion suggests that while there is support for the
science comprehension thesis, there is also reason to suspect the relationship between science
knowledge and beliefs is not uniform for all ideological groups. The motivated reasoning theory
is used to reinforce the expectation that the effect of knowledge is conditional, based on one’s
political orientations. The two-step model is then proposed as a means to understand the
relationship between science knowledge and climate change beliefs. Chapter 3 presents the survey
data utilized in the analysis and discusses the operationalization of key concepts. Chapter 4
examines just the first stage of the two-step process while Chapter 5 assess the entire model. The
last chapter reviews the results and discusses the implications of the research.

21

CHAPTER 2: SCIENTIFIC ORIENTATIONS & THE LIMITS OF KNOWLEDGE
Over the last three decades, the scientific community has emphasized the value of a
scientifically knowledgeable public. By accumulating knowledge and understanding scientific
processes, the public will be in a position to understand complex policy debates because they have
the skill and capacity to evaluate competing causal arguments about alleged problems and, in the
end, make informed decisions about whether the problem merits attention.10 It is often presumed
knowledge ultimately leads the citizenry to accept the causal arguments offered by the scientific
community. For instance, a panel assembled by the AAAS (noted earlier) speculated that the
scientific community should work on increasing the public’s awareness of scientific information
as a means to persuade the public about anthropogenic climate change (2010). The panel’s
discussion occurred in the aftermath of “Climategate,” when doubts about the validity of climate
change rose noticeably. Discussion points included how the scientific community should respond
to increasing doubt as well as how policies should be framed in order to generate public support.
The assumption is that once the public understands the scientific merits of an argument, then the
citizenry will come around to the idea of anthropogenic climate change and accept remedial
policies designed to mitigate greenhouse gas emissions.11 Oddly, as will be noted in the
forthcoming section, this is a relationship in the literature rarely explored by proponents of the
science comprehension thesis.
10

Boosting scientific knowledge also creates an additional benefit in that exposure to science
inspires young minds, which will help draw promising students to STEM (Science, Technology,
Engineering, and Mathematics) careers.
11
These arguments are not exclusive to climate change. Evolutionary scientist Richard Leaky, for
instance, predicted an end to the public debate over evolution (Eltman 2012). The accumulating
scientific information is so convincing that in 15-30 years Leaky believes no one will be able to
reject Darwin’s basic theory of evolution given information derived from the fossil record.
22

Recent research, however, casts doubt on whether those possessing elevated levels of
scientific knowledge think differently about issues compared to their low-knowledge counterparts.
This evidence is particularly notable in the context of climate change. For instance, scholarship
suggests there is no relationship between one’s level of scientific knowledge and concern over
climate change (Kahan et al. 2012a). Instead, values – specifically cultural values – are the optimal
predictor of climate change beliefs (Kahan et al. 2011; Kahan et al. 2012a). With this
conceptualization, there is little to no opportunity for information to shape beliefs regarding
contemporary policy issues such as climate change. Thus, information offered by the IPCC or
other science agencies that details the extent of a problem or underscores the need for corrective
policies is unlikely to have a meaningful impact on how the public views the phenomenon. These
frameworks that emphasize values are collectively referred to as the value-centered thesis.
This discussion attempts to reconcile these competing views about the relationship between
scientific knowledge and climate change beliefs. While accepting the basic tenants of these two
theses, I argue neither conceptualizes the appropriate conditions under which scientific knowledge
might guide or otherwise influence public opinion. Both views are discussed in more detail in the
following sections, with hypotheses offered as part of an attempt to reconcile the two views. Both
views are understood as having a degree of merit, yet careful consideration is required toward the
causal mechanisms at work and the use of proper dependent variables given those causal
arguments. In the end, a two-step process is conceptualized as one way to reconcile these
competing theses.
A Framework of Public Opinion
The foundation for the proposed two-step process is rooted in a framework of public
opinion that conceptualizes a sequence of steps utilized by the public to understand policy debates.

23

The discussion focuses on three related aspects of public opinion: sources of beliefs, beliefs about
a problem, and problem recognition. Conceptually, they are diagramed in Figure 2.1, with a natural
progression between the three concepts. The multiple rings on the far left represent the various
sources of beliefs often discussed in the literature. As noted earlier, partisan cues are one source
of beliefs, as are self-interest and information from various media sources. Knowledge, cultural
values, and risk perceptions fall under this grouping as well.
Figure 2.1: The two-step process, general framework

Source of
beliefs

Beliefs about
problem

Problem
recognition

With this progression in mind, I start with the proposition that individuals possess a belief
system that they utilize to navigate contemporary policy debates. This belief system is similar to a
blueprint as it provides a guide to understanding the competing causal stories that are offered by
policy advocates, decision-makers, and media pundits. For purposes here, a belief system is a
collection of orientations individuals rely on when formulating specific beliefs that are relevant to
political discussions.
At any given time, individuals have basic beliefs and assumptions about the world. These
include views about appropriate government activity, the trustworthiness of select actors, how to
handle risk and uncertainty in decision-making, as well as the nature of the relationship between
humans and their environment. An orientation is a collection of beliefs and assumptions on a broad
subject matter. For instance, individuals can possess orientations toward government, science,

24

religion, and the environment. A political orientation might include views about proper
government activity, how collective decisions should be reached, the degree to which society
should defer to experts, and whether decision-makers are inherently trustworthy.
For some individuals, there might be patterns as to how these orientations are bundled
together. That is, a sizeable portion of the public may share a belief system with similar orientations
toward government and science. For others, their belief system might resemble a seemingly
random or complex mix of orientations. Sabatier and Jenkins-Smith (1993), as well as others, argue
belief systems are well organized, with a clear structure and interrelationship between the various
orientations. However, they argue higher levels of organization are mostly reserved for political
leaders and the well-educated. The belief systems held by the public, on the other hand, are not as
organized. However, others (Carmines and Stimson 1986; Zaller 1992; Zaller and Feldman 1992)
have pointed out that the public tends to mimic elite discourse, using it as a heuristic for the
formation of their own beliefs. A reasonable expectation exists that the belief systems of the public
should feature similar organizational patterns as opinion leaders, if they receive a clear signal.
Regardless of how well they are structured, belief systems are utilized to understand
contemporary policy debates. At any given time, an individual, advocacy group, or politician might
declare that a problem is at hand – something that requires a solution in order to improve society.
The public makes sense of these claims by deferring to their belief system. Decisions about
whether to accept anthropogenic climate change as fact, believe in evolutionary theory, or support
stem cell research, all flow from some mixture of orientations that assign a sense of “right” and
“wrong” to the propositions that work their way into policy debates.

25

Sources of Belief Systems
One’s belief system is the product of a socialization process. Developing and assimilating
multiple orientations to form a coherent belief system starts at an early age, where parents imbue
their children with a set of values and beliefs as part of the childrearing process. For children,
relying on parents is an easy, convenient heuristic to developing a belief system (Achen 2002). A
lack of parental guidance, however, is not an obstacle to developing a coherent belief system, as
other resources are also available to help guide youth. These other socialization processes include
primary and secondary education, social interaction with friends and teachers within the local
community, and even consumption of basic news. In this socialization process, adolescents are
also accumulating knowledge – from parents, family and teachers in the classroom. Adolescents
match newly obtained knowledge with the preexisting beliefs and values they learned at a younger
age. Sometimes the knowledge is incorporated into the belief system, stored for later use as part
of the master blueprint utilized for guidance. Other times, the knowledge is rejected.
This socialization process is ongoing. At an early age children develop beliefs as well as a
willingness to express adult-like opinions (Sears 1975), although those beliefs may or may not be
firmly held until overt political action such as voting is taken. By the time individuals can
participate formally in the political process, they have developed the beginnings of an ideological
framework that filters political events (Beck and Jennings 1982; Jennings and Markus 1984; Niemi
and Jennings 1991). This ideological framework is the primary lens utilized by the citizenry to
navigate policy debates (Converse 1964). Other concepts have been argued on occasion to guide
specific beliefs and policy attitudes, such as self-interest, personal experiences and knowledge, yet
ideology and partisanship are often found to be the strongest predictor of beliefs and policy
preferences (Sears and Funk 1990; Lau and Heldman 2009).

26

Once individuals have developed a belief system (the first column of Figure 2.1), they are
in a position to make decisions about specific beliefs that arise during policy debates. Are the poor
deserving of welfare? Does hydraulic fracturing contaminate the water supply? Do vaccinations
have deadly or life-altering side effects? Policy advocates offer competing causal stories to the
public in order to shift opinion toward the advocates’ preferred answer. Individuals rely on their
belief system in order to navigate the competing causal arguments that emerge. Once deciding on
what to believe about an alleged problem, the individual is in a position to decide whether the
problem warrants a degree of concern.
This conceptualization offers a logical progression from a starting point (belief system) to
an end point (problem recognition). Intuitively, this series of relationships is appealing. There is a
reasonable expectation that individuals who are concerned about climate change also possess
various beliefs about climate change. In turn, these specific beliefs about climate change come
from some belief system that might include one’s technical understanding of climate change,
ideological preferences for what constitutes appropriate government activity, or general beliefs
about the relationship between humanity and the environment.
Climate Change Context
In the context of climate change, there are at least two orientations that help individuals
navigate the policy debate. Political orientations guide beliefs, which involve ideas about the
proper role of government. Scientific orientations also provide a basis for belief formation, guiding
how information provided by scientists is interpreted. These orientations are used to form specific
beliefs about climate change. Figure 2.2 offers further details about the basic components
associated with these two orientations. How individuals respond to the items listed in Figure 2.2
determine their precise orientations toward politics and science, respectively.

27

Table 2.1: Sample of the views associated with political and scientific orientations
Political orientations
Scientific orientations
 the proper role of government;
 the veracity of claims from analyses
adhering to the scientific method;
 how collective decisions should be
reached;
 the trustworthiness of scientist;
 the degree to which society should defer
 the appropriateness of evidence required
to experts on complex decisions; and
to justify beliefs; and
 the definition of a public problem.
 the benefits of scientific advancement.
In the policy debate over climate change, there are at least three contested issues of interest
in this study. In each area, policy advocates offer conflicting causal arguments. Individuals must
rely on their belief system (that is, their mix of political and scientific orientations) to make sense
of the conflicting arguments. First, disagreement exists about whether the Earth is warming. Are
temperatures increasing? If individuals have researched the question, they might be familiar with
temperature records produced by the IPCC and other agencies. They also have their own
experiences with local weather conditions. Scientific orientations can guide individuals in terms
of understanding the systematic processes of collecting data, and whether they trust scientists to
accurately report the data. Political orientations can guide individuals by predisposing them to
either outright reject or accept the information.
Second, disagreement exists about the sensitivity of the Earth to human influence, and
whether society can even engage in activities that would contribute to a warming climate. In other
words, how sensitive is the Earth to the burning of fossil fuels, the release of toxic compounds, the
destruction of ecosystems, the harvesting of natural resources, and many of the other activities
society engages in on a daily basis? Again, one’s belief system provides a roadmap to answering
these questions. A scientific orientation might lead some to accept the physics behind the
arguments of climatologists while a political orientation may lead them to accept or reject the
arguments out of disdain for relying on experts.

28

Third, disagreement exists about the cause of changes in the Earth’s temperature. Since the
1970s, satellites have provided a modern temperature record for climatologists to analyze while
various other proxies such as tree rings and ice cores provide temperature readings going back
even further, to hundreds of thousands of years ago. These temperature readings can be used as a
dependent variable in elaborate time-series analyses in an attempt to parse out explanations for
changes in temperature. Disagreements occur over which explanatory concepts are driving
observed temperature change, or even whether the statistical models include appropriate
assumptions that fit the data. This is perhaps the most contested component of the climate change
policy debate. The two orientations guide individuals in terms of informing them of whom to trust
and what constitutes appropriate processes to justifying knowledge.
The model tested in the forthcoming analyses is consistent with Figure 2.3, where specific
beliefs about climate change derived from both scientific and political orientations. These
orientations guide individuals as they make decisions about whether to accept or reject the causal
arguments offered by climatologists. Once the problem is understood, individuals are in a position
to make an informed decision about whether climate change is a problem. In the following
sections, both science and political orientations are discussed in more detail, with specific
hypotheses offered in order to assess the science comprehension and value-centered theses.
Figure 2.2: The two-step process for climate change beliefs
Science
orientations
Beliefs about
problem
Political
orientations

29

Problem
recognition

The Science Comprehension Thesis
For some time, possessing various types of education or knowledge has been recognized
as a key characteristic of individuals demonstrating higher levels of civic responsibility. In the
political context, knowledge strongly correlates with the retention of political information (Price
and Zaller 1993), which in turn helps the citizenry identify preferred political candidates (Gelman
and King 1993), assess political performance via merit rather than personal factors (Popkin and
Dimock 1999), and participate in elections (Popkin and Dimock 1999). Those with higher levels
of knowledge also maintain more stable attitudes (Delli Carpini and Keeter 1996). Once
knowledge is obtained, the citizenry appears to retain that knowledge throughout the remainder of
their life (Jennings 1996). In short, political knowledge is a key component in understanding civic
behavior. Putnam referred to education as “one of the most important predictors – usually, in fact,
the most important predictor – of many forms of social participation – from voting to associational
membership, to chairing a local committee to hosting a dinner party to giving blood” (2000, p.
186) while Converse (1972) referred to education as a “universal solvent” that correlates with
socially desirable outcomes.
Similar arguments exist as to the value of scientific knowledge. The influence of science
specific knowledge on behavior and opinions has been referred to as either the deficit model (Irwin
and Wynne 1996) or science comprehension thesis (Kahan et al. 2012a). The premise, however, is
the same, in that variations within public opinion on science-based issues can be explained via a
lack of understanding scientific research. Just as political knowledge is required to assess the
success of incumbent candidates, scientific knowledge is required to assess claims regarding the
necessity of new public policies for issues such as climate change.

30

An early report by the Royal Society in London argued for the importance of public
understanding of basic scientific concepts. The report argued that a more scientifically
knowledgeable public “will not automatically lead to a consensus about the best answer, but it will
at least lead to more informed, and therefore better, decision-making” (Bodmer 1985, p. 10). The
report observed that science is a critical component of the policymaking process. In several areas,
scientists are offering critical information relevant to policy debates. Given their advisory role on
contemporary policy issues, scientist have been described as an additional branch of government
(Jasanoff 1990).
However, scientists cannot avoid the political conflict inherent to the policy process as they
fulfill their role as advisors to elected decision-makers. While some scientific domains such as
nanotechnology research may be relatively bipartisan, other domains – particularly in the
environmental context – possess clear partisan boundaries where Republicans and Democrats have
established opposing positions on issues. These positions, which are based on political
orientations, inform the range of policy options partisans are willing to consider. In the
environmental context, issues such as climate change raise questions about whether government
intervention is necessary and the extent to which regulations are required in order to minimize the
risk associated with greenhouse gas emissions.
Whatever information the scientific community offers, there is a strong likelihood those
findings will either support or contradict the established positions of the political parties. As noted
in the opening chapter, issue evolution provides at least one explanation of how the political parties
developed different positions on environmental issues. Because of the political dimension
embedded within contemporary policy issues such as a cap-and-trade program or a carbon tax,
partisan actors are expected to interpret and frame information about climate change in a context

31

compatible with the pre-existing values and beliefs of their respective political coalition (Jones
and Baumgartner 2005; Schattschneider 1960).
Thus, for the public to understand the policy debate between politicians (and sometimes
between politicians and scientists), the citizenry requires a minimal understanding of science
(Trefil 2008; Prewitt 1983). By possessing some common or basic understanding of the scientific
process, the citizenry can interpret scientific news reports and better evaluate the causal arguments
offered by partisan actors. This interpretation process is inherently a social process, given the
diverse range of backgrounds and preferences held by the public. Understanding of the scientific
process (via accumulated scientific knowledge) is one way for the public to interpret information
and understand the problem. Once the problem is understood, the public can then make informed
policy decisions and even participate in the debate itself.
Embedded within this process is the notion that knowledge is necessary in order for the
public to pay attention and mobilize around a given issue (Almond 1960; Miller 1983). To create
the momentum necessary to enact policies designed to correct a given problem such as climate
change, the citizenry needs to know something about the problem in the first place. That is, they
need to understand the alleged causes of a problem and the potential consequences if the problem
is left unchecked. Only after comprehending these two components can the citizenry mobilize and
place serious pressure on elected leaders for policy change. Once a critical level of awareness of a
problem is achieved within public opinion, the conditions become ripe for policy change (Page
and Shapiro 1983). With this conceptualization, how the public comprehends the cause of the
problem (i.e., are observed temperature changes the result of burning fossil fuels or some natural
cause?) and its consequences (i.e., are rising seas and stronger storms a main result?) will dictate

32

the pace of policy change. Of course, the scientific community and specifically climatologists have
information to offer on both of these fronts.
This discussion leads to one hypothesis that speaks directly to science comprehension
thesis. If the thesis has any merit, at a minimum those individuals possessing a foundational
understanding of science should think differently about the causal arguments related to climate
change compared to their low-knowledge counterparts. Specifically stated:
Hypothesis 1: Individuals with higher levels of scientific knowledge accept the causal
arguments from climatologists at higher rates than their low-knowledge counterparts,
holding all else constant.
This hypothesis refers to an expected relationship within the first stage of the two-step model. As
individuals are socialized to accept beliefs supported by scientific processes, they accumulate basic
scientific knowledge. During this process, individuals also develop attitudes about the reliability
of experiments and scientific analyses, as well as the trustworthiness of scientists executing the
research. Thus, scientific knowledge is an indicator of how individuals are orientated toward
science.
What makes high-knowledge individuals stand out in their beliefs? If individuals are going
to make sense of the causal arguments pushed by climatologists, they must engage in information
acquisition activities. To that end, knowledge is seen as providing the skills necessary to process
technical reports (Jerit et al. 2006) and ask critical questions of the claims that surface in the policy
debate. They also develop the confidence to engage others in the debate, as their familiarity with
science mitigates fears of sounding ignorant during interpersonal communications. In other words,
knowledge provides the capacity to engage in information acquisition activities, which in turn lead
to more refined and specific beliefs about issues such as climate change (Trefil 2008).

33

This argument forms what I call the core of the science comprehension thesis. The
arguments of climatologists (and scientists in general) are publically contested at various times.
To understand the public debate, the thesis suggests individuals need the capacity to engage in
information acquisition activities in order to evaluate scientific claims. If they are confident in the
work of scientists, they will accept their arguments. Possessing a foundational understanding of
scientific knowledge provides the capacity for this process to unfold. Note again that there is no
requirement that individuals believe the scientific information implies a problem. As specified in
the two-step model, problem definition and understanding scientific arguments are conceptually
different stages of public opinion. If the relationship specified in the first hypothesis does not
materialize through empirical analyses, then perhaps there is no merit to the arguments offered by
the scientific community. Education and knowledge would not be a “universal solvent” when it
comes to climate change beliefs and perhaps science or policy issues more broadly. This first
hypothesis, the relationship between knowledge and specific causal arguments, constitutes the first
step of the two-step process noted earlier.
Development of Empirical Assessments
The aforementioned Bodmer (1985) report occurred in the context of increasing skepticism
over science, with the hope that greater scientific understanding would translate to a greater
appreciation for scientists (Lewenstein 1992). That is, “science itself would be a beneficiary of
increased scientific literacy” through increased research support, newfound faith in technology,
etc. (Miller 2001, p. 116). In other words, training and educating adolescents about the scientific
process and the work of scientists would orient individuals to think favorably of the scientific
community. The first steps for scholars interested in these relationships, then, was to (1) develop

34

a battery of questions that accurately capture scientific knowledge and (2) execute a research
agenda to examine the outcomes associated with a scientifically knowledgeable public.
Constructing the Modern Knowledge Assessment Questions
In constructing a valid measure of science knowledge, researchers had to first ask what
type of knowledge informed individuals should possess. On this front, Miller (1998; 2004) argues
scientifically knowledgeable persons should be able to identify basic scientific terms and possess
an understanding of the scientific process. These are said to represent two dimensions of scientific
knowledge. Evans and Durant (1995) argue, however, that these two dimensions have merged
together and are, for all intents and purposes, a single dimension. Respondents who can identify
basic scientific constructs also understand scientific processes, such as probability and
experimental design.
The modern knowledge questions are the product of collaboration between Thomas and
Durant, and Miller (1998), who elected to focus on the core scientific constructs. Knowledge
should include an understanding of basic scientific concepts, such as DNA and molecules – terms
that apply to a wide range of research issues. These basic constructs are building blocks that
provide the capacity for individuals to read and comprehend more in-depth media reports. For
instance, it is difficult to comprehend an article about genetically modified food if one does not
understand basic concepts like DNA. The advantage of this conceptualization of knowledge is that
it focuses on basic scientific constructs, thus avoiding temporal science issues and allowing for a
comparison of science knowledge across generations.
Assessing Outcomes
The initial analyses of the science comprehension thesis were drawn towards empirical
tests examining the correlation between knowledge and generalized science attitudes (e.g., Bauer

35

et al. 1994; Sturgis and Allum 2004; Evans and Durant 1995; Sturgis et al. 2005; Hayes and Tariq
2000); however, the correlation between science knowledge and generalized attitudes is generally
considered weak (Allum et al. 2008). These studies with their focus on generalized attitudes as the
outcome measure formed the core tests of the science comprehension thesis up until the late 1990s
and early 2000s.
Over the last decade, alternative studies have looked beyond generalized science attitudes
towards specific policy domains with a focus on other concepts such as problem recognition,
perceptions of consensus, and policy preferences. To briefly illustrate, Hayes (2001) finds a
positive relationship between scientific knowledge and two environmental specific dependent
variables – attitudes favoring environmental protection and overall respect for nature.12 Meilby et
al. (2012) find a correlation between knowledge of basic biology and public support for genetically
modified products, with those demonstrating higher levels of knowledge also capable of
discriminating between support for the technology via application (e.g., medical versus
agricultural). Another study finds knowledge and attitudes towards space policy are positively
correlated (Cook et al. 2011).
Some studies, however, identify no meaningful relationship between scientific knowledge
and attitudes, such as when considering support for nanotechnology policies (Ho et al. 2010). In
some cases, the link between knowledge and opinions may even be negative. Evans and Durant
(1995) argue the relationship between knowledge and research support is positive for what they
call “useful” science research. The public, for instance, can see the value of publically funded
12

Note both Hayes (2001) and Hayes & Tariq (2000) argue the science comprehension thesis is
found wanting despite a significant relationship between science knowledge and selected
dependent variables. See also Sturgis and Allum (2001).

36

research targeting medical breakthroughs; there is a value associated with developing a cure to
cancer. As such, the citizenry support efforts to funnel public resources toward such ventures. If
the research may be considered non-useful (e.g., space exploration) or if it impinges on moral
issues (e.g., cloning), a negative relationship emerges. This suggests to some extent that highknowledge individuals do not blindly follow the scientific community’s lead, at least when it
comes to research support.
The relationships that emerge between scientific knowledge and specific climate change
attitudes are weak and, in some cases, puzzling. In the context of climate change, Hamilton (2012)
argues education (as a proxy for knowledge) leads individuals to correctly identify the scientific
community’s arguments about what is driving changes in the arctic region. Wood & Vedlitz (2007)
find that their measure of scientific knowledge does not explain concern about climate change.
The lack of a positive, significant relationship between science knowledge and climate change
concern has been observed elsewhere as well, with the relationship found to be negative (Kahan
et al. 2012a; Kellstedt et al. 2008). Looking specifically at environmental knowledge, however,
reveals a small but positive relationship with concern for global warming (Wood and Vedlitz
2007).
Others have observed that the role of knowledge, specifically knowledge of climate
change, may lead to support for remedial policies, but evidence also suggests knowledge does not
function equally for all individuals (Marquart-Pyatt et al. 2011). This inconsistency is perhaps due
in part to the use of self-reported knowledge of climate change in some studies (e.g., Malka et al.
2009), which is arguably an inferior measure when compared to actual knowledge of both science
and climate change.

37

The message at this point is that a chain of research contains some support for the idea that
possessing scientific knowledge leads individuals to different conclusions about science-related
policy issues. Furthermore, the direction of these attitudes appears to be in alignment with
positions shared by the scientific community. With only mixed success, though, other factors
might be contributing to science-related beliefs and attitudes. The upcoming discussion will
consider two alternatives: that there are additional orientations toward science that need to be
considered and that values might be supplanting the influence of knowledge in some cases.
Predisposed to Science
Recent scholarship has suggested attitudes toward science vary even among those who are
highly knowledgeable (e.g., Gauchat 2011). Understanding this divide relies on the idea that
science is a social construct (Wynne 1991; Ziman 1991; Locke 2002). It is important, then, to
consider how individuals think of scientists. Because of the transaction costs associated with
accumulating knowledge and understanding all aspects of science, alternative concepts such as
institutional alienation might play a role in belief formation as well. Just as individuals come to
distrust government operations or large bureaucracies (Giddens 1991), individuals may distrust or
become discontent with the whole scientific enterprise in general (Yearley 2000).
Specifically, how individuals feel about scientists might have some influence on how the
citizenry makes sense of the climate change debate. If citizens feel good about scientists and
believe they are working to better humanity’s condition, then they are possibly more likely to
accept the claims of scientists. Conversely, if one feels scientists threaten the status quo and may
be complicating society with unnecessary claims of a problem, then they may feel more inclined
to reject the arguments of scientists. This line of thought leads to additional hypotheses that
examine normative beliefs:

38

Hypothesis 2: Individuals perceiving benefits from scientific advancement are more likely
to accept causal arguments from climatologists, holding all else constant.
Hypothesis 3: Individuals skeptical of scientists and their work are less likely to accept the
causal arguments from climatologists, holding all else constant.
The normative views conceptualized in Hypotheses 2 and 3 constitute an orientation toward
science. Positive and negative feelings about scientists in general push the citizenry toward
accepting or rejecting causal arguments from climatologists. They can hear the debates about
climate change in the media, identify the position of climatologists, and then make decisions about
which causal arguments to accept based on their feeling of scientists. These hypotheses make no
requirement as to whether they have any factual knowledge about science.
It is important to note that these positive and negative normative views are seen as two
distinct concepts. An individual can see positive benefits coming from scientific adventures, yet
also be cautious about the pace of technological development or the ethics of select scientists.
Thus, two distinct hypotheses are tested. This distinction between positive and negative views is
explored further in the Chapter 3.
A second important note is that the science comprehension thesis says nothing about
normative science views. Normative views of science were the dependent variables in the original
analyses of the thesis. In this analysis, they are conceptualized as an independent variable, a
predictor of climate change beliefs. As described above, normative views are a heuristic that can
be utilized in the absence of knowledge.
Values as an Alternative Framework
The work by Evans and Durant (1995) noted earlier implies issue context matters, which
in turn suggests the link between knowledge and public opinion is not as simple as advocates of

39

the science comprehension thesis contend. This complexity does not imply that the science
comprehension thesis is invalid or somehow compromised; rather, it forces one to consider why
context matters. Knowledge may matter, but its influence is contingent on the context of the issue
and the degree to which the science supports one’s preexisting beliefs. The value-centered thesis
refers to any framework that emphasize values as providing heuristics to facilitate the public’s
decision-making process.
An emerging framework that falls under this grouping is cultural cognition (Kahan et al.
2011; Kahan et al. 2012a). Kahan and colleagues have developed a theory that focuses on four
cultural values (hierarchy, egalitarianism, fatalism, and individualism) originally identified by
Douglas (1983) and others (Douglas and Wildavsky 1983; Thompson et al. 1990; Scharwz and
Thompson 1990). The four values are arranged in a matrix (replicated in Figure 2.4) based on
views about collectivism and individual autonomy in society. Those holding hierarchical values
see the world as highly ordered with little individual autonomy in any decision-making process.
Egalitarians also see the world as highly ordered, but value individual thought and believe
everyone has an equal voice. Fatalists see the world as fairly random, with no collective
organization in the decision-making process and little role for individual choice. Lastly,
individualists see a less restrictive decision-making structure, one where it is the individuals who
should ultimately be the driver of decisions.
Table 2.2: Dimensions of culture
Low individual autonomy
Hierarchy
High collectivization
Fatalism
Low collectivization

High individual autonomy
Egalitarianism
Individualism

The important part of this values discussion is that each value is connected to varying
beliefs about the environment. Thompson has been instrumental on this front (Scharwz and
Thompson 1990; Thompson et al. 1990; Swedlow 2012). For these authors, cultural values shape
40

one’s view of environmental issues. Both hierarchical and individualistic personalities see nature
as resilient to the influence of human activity. For instance, they are likely to think that even if all
the trees are cut down, the forest can still grow back. Similarly, they might think increasing the
amount of total carbon dioxide in the atmosphere by only 0.5% cannot cause detrimental effects
because the Earth is resilient to what are perceived as minute human influences.
Conversely, personalities characterized by egalitarianism and fatalism see nature as more
sensitive to human activity, that slight disturbances might upset the natural equilibrium of the
world – perhaps permanently. For Kahan and his colleagues (2012), it is these cultural values and
associated risk perceptions that drive beliefs about climate change. Additional work by Jones
(2011) has argued that climate change deniers are largely individualists: personalities who do not
desire collective action and seek to preserve individual choice. This emphasis on cultural values
has risen in the literature over the last several years as an explanation for the polarization in public
opinion over climate change. In the process, their empirical work rejects the notion that scientific
knowledge has a role in the public’s decision-making calculus.
Why Values Dominate Climate Change Models
Value differences might explain why the public disagrees on issues such as climate change
and explain why inconsistent findings emerge when scholars look at relationships between
scientific knowledge and key constructs. Still, why, from a psychological perspective, do segments
of the public reject scientific arguments when science is often viewed as objective and beyond
reproach? One answer found within the theory of motivated reasoning (Kunda 1990; Kunda 1987)
considers the motivations of citizens. Individuals retain a preferred outcome in policy disputes and
seek a coherent, consistent story to explain the world. As a result, they look to preserve that
consistency as new policy information surfaces. A bias, of sorts, results, which affects the entire

41

decision (or reasoning) process by shaping how individuals reconsider their beliefs and evaluate
evidence.
Two types of motivation are considered. Naturally, the citizenry can be motivated by
accuracy in the sense that they want their beliefs to be right. There is little incentive to appear out
of touch or foolish in the face of undisputable evidence. However, there is also motivation to reach
specific conclusions, ones that match the initial predispositions dictated by one’s belief system.
The two motivations must strike a balance – “people are more likely to arrive at conclusions that
they want to arrive at, but their ability to do so is constrained by their ability to construct seemingly
reasonable justifications for these conclusions” (Kunda 1990, p. 480). When a high degree of
ambiguity exists over an issue, such as with climate change, individuals are prone to favor views
that match their initial preferences rather than strive for accuracy (Doherty and Wolak 2012).
As a result, persuading individuals to change their beliefs can be quite difficult. Initial
views are anchored to specific positions (Anderson 1981; Kahneman and Tversky 1982) because
“citizens are prone to overly accommodate supportive evidence while dismissing out-of-hand
evidence that challenges their prior attitudes” (Taber and Lodge 2006, p. 755-756). There also
appears to be a tendency to overestimate the support for one’s preferred position (Nir 2011;
Druckman and Bolsen 2011). New information may even have a “backfire effect,” where
“unwelcome information” leads individuals to support their original opinions with even stronger
conviction (Nyhan and Reifler 2010).
In the context of science-related issues, scholars have noted conservatives tend to dismiss
research findings that support liberal conclusions, contributing the results to the biases of liberal
researchers (MacCoun and Paletz 2009). In respect to climate change specifically, a boomerang
effect is observed where media reports discussing the consequences of climate change reduce

42

policy support among those predisposed to reject climate change (Hart and Nisbet 2012). These
results suggest it is difficult to alter the beliefs of individuals.
The possibility remains that new information is accepted despite value conflict. At some
point, the accuracy of new information can no longer be denied. However, the amount or quality
of information required can be quite large. Even when presented with a scenario where conclusive
evidence of the consequences of climate change is available, segments of the public still remained
unconcerned over the issue (Wood and Vedlitz 2007). It seems individuals with a developed belief
system are going to resist altering their views even if the scientific community suggests definitive
evidence exists of a problem.
Reconciling the Two Views
There appears to be only a small window of opportunity for information to influence
beliefs. The Kahan et al. (2012) work that specifies a limited relationship between basic scientific
understanding and climate change beliefs is reinforced in other research as well (e.g., Wood and
Vedlitz 2007; Malka et al. 2009). Based on their research, Kahan and his colleagues are
comfortable asserting that “[s]imply improving the clarity of scientific information will not dispel
public conflict so long as the climate change debate continues to feature cultural meanings that
divide citizens of opposing world-views” (2012, p. 734). Still, advocates of the science
comprehension thesis insist there must be something about knowledge that helps guide public
opinions on climate change (McCaffrey and Rosenau 2012).
Are these two theses mutually exclusive, or is there an opportunity to reconcile the two
views? Emphasis is placed on two points in an attempt to understand this apparent conflict. First,
the science comprehension thesis requires qualification, in that not all individuals respond to
scientific information. It is appropriate to model science knowledge as interacting with or being

43

moderated by values. Second, the focus on climate change concern as one of the dependent
variables is a misrepresentation of the science comprehension thesis. Nothing in the
aforementioned discussion of the thesis suggested knowledge makes individuals more risk averse.
The argument was that knowledge allows individuals to understand the causal arguments from
scientists and then, using that information, make informed decisions. While it is certainly difficult
to measure what constitutes an informed decision, efforts can be made to offer a more precise
assessment of the science comprehension thesis.
A Moderating Effect
The emphasis of the value-centered thesis was that values moderate or condition the
interpretation of information. Certainly, then, there is an expectation that the relationship between
science knowledge and climate change beliefs might be moderated by political values or other
theoretically important concepts. It does appear that partisans tend to be better educated (Taber
and Lodge 2006), and as suggested earlier, there is a partisan dimension to opinions about climate
change (Dunlap et al. 2001; McCright and Dunlap 2011). Research also suggests that the educated
are less likely to follow the advice of experts, electing to rely more on their predispositions and
political preferences (Kuklinski et al. 1982). As such, it should not be surprising if high-knowledge
individuals think differently about climate change, with liberals (Democrats) more likely to
embrace the scientific claims and conservatives (Republicans) more willing to call the information
into question.
There is support for this argument in the literature. For instance, Malka et al. (2009)
proposed only select individuals who possess higher levels of scientific knowledge are likely to
stand out in empirical analyses. Specifically, among those with knowledge, only Democrats were
more likely to perceive climate change as a serious problem. It was argued that climate change

44

research coincides with the environmental orientation of the party. This suggests that any potential
relationship between scientific knowledge and beliefs is likely to be conditional, based on prior
beliefs. Since ideologues receive reinforcing messages from ideological leaders that openly
question or support the scientific community’s claims, this ideological or partisan divide is not
surprising (McCright and Dunlap 2000). Conservatives, with their focus on fiscal restraint and
limited regulation (Zumbrunnen and Gangl 2008), tend to be more skeptical of climate change.
Meanwhile, liberals have been identified as demonstrating strong environmental support since the
1970s’ environmental movement (Jones and Dunlap 1992; Dunlap and McCright 2008) and tend
to receive reinforcing messages that the environment needs to be protected.
Further support for interactive effects occurs in Hamilton’s work when looking at climate
change concern (2011), but such effects are more limited when looking at climate change beliefs
(2012). This latter study comes the closest to the analysis proposed here. The key difference is in
how the interaction effect is operationalized. Hamilton utilized a measure of education and
partisanship while this study uses a more nuanced and precise measure of science knowledge. The
use of science knowledge in the interaction term offers a different conclusion than that found in
Hamilton’s work – an issue that is taken up further in the discussion.13
The justification for including an interaction term in the analysis is based on the valuecentered thesis, which creates an expectation that a subclass of high-knowledge individuals will
think about climate change in unique ways compared to their other high-knowledge counterparts.
13

Also see McCright (2011) for additional commentary on the moderating effect of political
ideology and partisanship. Society is divided into groups that either “attack” or “defend” industrial
capitalism, according to the argument. These groups map along current political cleavages in the
United States, which might help to understand why conservatives think differently about climate
change compared to other ideologues.

45

Three specific expectations, then, are anticipated in the analysis, depending on the ideological
beliefs of the survey respondents:
(1) Moderates without a firmly held political ideology rely on knowledge to guide beliefs
when ideas are contested. With, presumably, no strong value bias, they are more willing to
accept information from scientists and incorporate that information into their existing
beliefs.
(2) If new information supports an ideological position, this ideological partisan will more
readily accept new beliefs into his or her pre-existing belief structure (compared to
moderates). That is, the information is a tool to bolster or support one’s original position.
(3) If new information threatens an ideological position, this ideological partisan will more
readily reject the information and defend those values. This ideologue will more likely
question the accuracy of the information and rally around his or her core beliefs.
This discussion suggests a hierarchy should emerge in the data when applying this
conceptualization to climate change. Liberals should be more likely to agree with climatologists,
conservatives the least likely to agree, and moderates should be somewhere between the two
ideological extremes. From this breakdown, the following hypotheses are proposed:
Hypothesis 4: Liberals (conservatives) are more likely to agree (disagree) with causal
arguments consistent with the IPCC, holding all else constant
Hypothesis 5: Ideology moderates the relationship between scientific knowledge and
climate change beliefs.
Evidence of a moderating effect would include differences between high-knowledge liberals and
conservatives emerging with respect to the rate at which they accept the causal arguments of
climatologists. This observation would also support the value-centered thesis. The evidence would

46

not suggest, however, that the science comprehension thesis is somehow compromised. Rather,
the moderating effect reveals the conditional nature of the science comprehension thesis as it only
applies to a targeted group of the population. Whether high-knowledge moderates are more or less
likely to accept causal arguments about climate change will be critical to assessing the veracity of
the science comprehension thesis. Given the discussion of ideological differences with respect to
climate change, it is necessary for models to account for potential interactive effects between
knowledge and beliefs. Failure to do so may lead to inappropriate conclusions about the science
comprehension thesis. It also potentially explains why there is disagreement in the literature about
the veracity of the thesis.
Different Outcome Expectations
An additional measure to understand the conflict in the literature is the result of a focus on
climate change concern as the dependent variable of interest (e.g., Kahan et al. 2012b; Kellstedt et
al. 2008; Wood and Vedlitz 2007). Problem recognition, however, is rooted, in part, in individual
risk perceptions (Leiserowitz 2005; Lorenzoni et al. 2005). It is difficult to imagine a causal
process where simply knowing basic foundational science leads individuals to change their level
of risk aversion, although new information might make the risks clearer to individuals, thereby
causing changes in risk perceptions. Still, it is likely individuals possess varying degrees of risk
acceptance or aversion before any new information is received. There are valid reasons individuals
might acknowledge a phenomenon such as climate change yet not be concerned with the problem.
Perhaps the consequences are not viewed as entirely negative, or perhaps they feel scientists will
eventually develop technological solutions before severe problems emerge. However, nothing
about the science comprehension thesis (at its core) suggests higher levels of knowledge should
lead to greater recognition of a problem. Rather, as indicated earlier, knowledge should help

47

individuals understand the problem, from which point they can decide on their own whether they
should be concerned about the alleged problem.
In a sense, prior scholarship has misconstrued the essence of problem definition. One has
to understand the problem in a manner similar to climatologists before exhibiting the scientific
community’s expected level of concern. Wood and Vedlitz (2007) hint at (but do not develop) this
distinction when they briefly observe that basic scientific knowledge does not correlate with
concern over claim change, but a domain specific construction of environmental knowledge does
positively correlate with concern. The Wood and Vedlitz observation provides a basis for
conceptualizing the two-step processed noted at the beginning of the chapter.
Figure 2.3: The two-step process for climate change beliefs
Science
orientations
Beliefs about
problem

Problem
recognition

Political
orientations

Figure 2.5 presents the two-step process introduced earlier. The figure shows a proper
conceptualization of the science comprehension thesis and how it is expected to “connect” with
climate change concern (i.e., problem recognition). Individuals are oriented toward science, a
process that includes exposure to science and the scientific process. This includes accumulating
basic scientific knowledge and developing the skills to properly evaluate scientific arguments.
Once this knowledge accumulates, individuals are in a position to evaluate causal arguments
offered by the IPCC, political leaders and other policy advocates. The thesis suggests possessing
higher levels of scientific knowledge provides the skill set and capacity to make sense of the causal
48

arguments offered as part of the climate change debate. Basic scientific knowledge, then, facilitates
the understanding of contemporary problems like climate change.
With an understanding of climate change, individuals are then in a position to make
informed decisions about whether climate change is a threat. Again, the science comprehension
thesis does not specify whether understanding climate change in a manner similar to climatologists
should increase concern. It simply says individuals are in a position to make an informed decision
about the accuracy of the claim. However, a link is expected between climate change concern and
how one understands the problem. Climatologists, for one, believe remedial policies are required,
and the public can certainly reach similar conclusions.
Properly understood, then, the role of scientific knowledge in this conceptualization and
its relationship with climate change concern is best understood as an indirect effect. Knowledge
helps individuals understand what climatologists are saying, and if they reach this point they are
then more likely to be concerned about the phenomenon. This discussion leads to two additional
hypotheses as part of an assessment of the two-step process just described:
Hypothesis 6: Individuals who understand climate change in a manner consistent with
climatologists are more likely to demonstrate higher levels of concern.
Hypothesis 7: Individuals concerned about climate change are knowledgeable about basic
scientific principles.
These last two hypotheses are assessed through path analysis to show the causal relationship
developed in the two-step approach. If scientific knowledge possesses an indirect effect on climate
change concern when assessing the causal argument in Figure 2.2, then there is arguably support
for Hypothesis 7. That is, those with scientific knowledge are more likely to be concerned about

49

climate change, but this relationship emerges due to the prior effect of science knowledge on
problem understanding and acceptance of arguments from climatologists.
These hypotheses outline the expected relationships of the two-step process. Again, the
suggestion is that individuals have some understanding of climate change, which in turn directly
influences problem recognition. That understanding, however, hinges on the belief systems of
individuals. This discussion is primarily interested in whether science knowledge can guide
individuals toward understanding climate change in a manner consistent with climatologists, but
at the same time, other orientations are important as well in determining climate change beliefs.
To be clear, this is not an argument that cultural values or other factors such as affect (e.g.,
Leiserowitz 2006) and risk perceptions (e.g., Lorenzoni et al. 2005) are not essential to the public’s
decision-making calculus. Indeed, concern over an issue is an inherently subjective process and a
variety of non-objective factors (i.e., those not related directly to knowledge) should be expected
to problem understanding and problem recognition as well.
As noted earlier, however, there is the possibility that individuals accept the arguments of
climatologists but may simply not be concerned about the problem. This is because they have risk
accepting personalities (Brody et al. 2008; Leiserowitz 2005; Lorenzoni et al. 2005; Pidgeon and
Butler 2009; Searle and Gow 2010; Spence et al. 2012) or perhaps feel technological solutions will
be available if the problem does get out of hand many years down the road. These potential
relationships are consistent with the value-centered thesis in that predispositions and not
knowledge are utilized to assess the severity of a problem. Consequently, the two-step process
does not explain all the variation in public opinion. As such, the proposed diagram presented earlier
should be modified as seen in Figure 2.6, with direct connections between orientations and concern

50

to represent these other unexplored relationships. This figure provides the final model to be tested
in the forthcoming chapters.
Figure 2.4: A two-step process for climate change beliefs, tested model
Science
orientations
Understanding
of the problem

Problem
recognition

Political
orientations
Summary
The preceding discussion attempts to reconcile conflicting results in the literature by paying closer
attention to the core expectations of the science comprehension thesis. In the process, careful
attention is given to the dependent variable selection and the potential for moderating effects. The
hypotheses are tested with a public opinion dataset that features an extensive battery of questions
that allow an assessment of the science-orientated concepts discussed above. The next chapter will
speak more about the data to be utilized in the analysis, how the variables are operationalized, and
the methodologies to be incorporated in future chapters.

51

CHAPTER 3: DATA & OPERATIONALIZATION
Assessing the relationships identified in the previous chapter requires a public opinion
dataset that no only inquires about climate change attitudes, but also includes detailed questions
about science attitudes. The forthcoming section discusses such a dataset and offers details about
how the operationalization of the key concepts identified in Chapter 2. A descriptive analysis of
the science concepts is offered, followed by an outline of the planned approach for the empirical
chapters.
Survey Data
The survey utilized in the analysis was collected by Dr. Jon D. Miller and is referred to as
the Science News Survey. The survey is a nationwide three-wave panel survey that was originally
designed to assess whether exposure to local television news reports of science-related issues
influenced public opinion. Throughout the survey, questions were asked about attitudes toward
science, including specific inquiries into the beliefs and policy preferences of the respondents.
The first wave of the survey was conducted in September 2007. During this wave, the
survey focused on the beliefs and attitudes of respondents on science issues. Included in this wave
was a self-assessment of how informed respondents claimed to be on a variety of science issues,
including global climate change and space exploration. Also included was a battery of questions
to assess scientific knowledge. The questions were in true or false, as well as multiple choice,
formats. The content of the questions covered a range of topics. Several concepts utilized in the
analysis come from the first wave of the survey, including science knowledge, generalized science
attitudes, and beliefs about climate change. The second wave of data collection occurred in
November 2007 and asked about scientific concepts that appeared on local television news

52

broadcasts over the preceding month. The forthcoming analysis does not utilize data from the
second wave of the survey.
The third wave occurred in March 2008, with the purpose of assessing respondents’
political attitudes during the 2008 presidential primary campaign season. Included in the third
wave was a battery of declarative policy statements that asserted a clear position on contemporary
political issues. Respondents were asked about their level of agreement with each policy statement.
This wave of the survey also included questions specific to climate change, such as one’s level of
concern, how often they interacted and discussed the issue with others, and whether they sought
out additional information about the issue. This information is summarized in Table 3.1.
Table 3.1: Summary of Science News Survey
Wave 1
Wave 2
Wave 3
(September 2007)
(November 2007)
(March 2008)
N=
1407
1166
960
Notes: The first stage of the analysis (Chapter 4) relies primarily on survey questions from the
first wave of the survey. No analyses utilize questions from the second wave. The third wave of
the survey is utilized in analyses presented in Chapter 5, as well as Appendices A and C.
Knowledge Networks, a polling service that utilizes a random sample of the United States
adult population to distribute online surveys, collected the data. Once respondents are in the
Knowledge Network survey population (the so-called “Knowledge Panel”), they complete two to
four surveys per month, delivered by Knowledge Networks through the internet. 14 The Science
News Survey was randomly assigned to 1407 individuals comprising the Knowledge Panel. The
14

Knowledge Networks built a probability-based random sample of the U.S. population.
Households were strategically recruited to participate in periodic monthly surveys, with provisions
to account for non-internet populations. They claim the Knowledge Panel accurately represents
97% of U.S. households and argue results are comparable to random-digit dialing procedures.
Scholars have argued probability-based online surveys are an optimal form of data collection, as
procedures like those used by Knowledge Network reduce error and provide more accurate results
(Chang and Krosnick 2009).

53

size of the sample did fluctuate by wave, with the sample size dropping to 1116 and 960 on the
second and third waves, respectively. A weight is included in the dataset to account for the loss of
survey respondents over the seven-month timeframe, which also serves to keep the survey sample
weighted to the national adult population.
Why Climate Change?
The advantage of the Science News Survey is its wealth of science questions, including
specific inquiries about climate change beliefs and attitudes. Climate change is an ideal context
for an analysis assessing the science comprehension thesis for multiple reasons. First, as noted
earlier, a high degree of consensus exists among scientists that the Earth is warming and that these
changes are the result of anthropogenic activities. The idea of a consensus is essential to the science
comprehension thesis as those familiar with the scientific process and the general work of scientists
should be able to identify the causal arguments made by scientists. If there is genuine disagreement
within the scientific community, then it is doubtful those with elevated levels of knowledge should
stand out in analyses. Thus, the expectations outlined in Chapter 2 should materialize in this
context.
Second, despite the scientific consensus on climate change (see the first chapter), it is an
issue where non-scientists contest the causal arguments from climatologists. Senator James Inhofe,
a Republican from Oklahoma, argues that the idea of anthropogenic climate change is a hoax
(Inhofe 2012) while the broader Republican Party and its allies coordinate a counter-narrative to
the work of climatologists that emphasizes growing uncertainty about the scientific community’s
understanding of the problem (McCright and Dunlap 2000). These counterarguments from nonscientists might appeal to segments of the public, depending on the mix of one’s political and
cultural orientations. These counterarguments provide another voice in the public debate, and a

54

reason to suspect the relationship between knowledge and beliefs is weakened by ideological
predispositions.
Given the above, there is reason to suspect the relationships outlined in the Chapter 2
should emerge in the context of climate change. It is important to note, however, that the
relationships outlined in the Chapter 2 may not emerge in other contexts. If there is no strong
counterargument by an organized interest to a scientific issue, then (1) the relationship between
science knowledge and specific beliefs should be empirically stronger, and (2) no moderating
effects (whether by ideology or some other factor) should emerge. However, climate change
represents an issue where counterarguments emerge, which means the context offers a challenging
test for the science comprehension thesis, especially given the prior research noted earlier.
Operationalizing Key Concepts
The main theoretical concepts of interest involve the orientation of survey respondents
toward the scientific enterprise. These include both one’s foundational understanding of science
and normative science views. The operationalization of these concepts is discussed below.
Scientific Knowledge
The key concept that predicts agreement with the work of scientists is one’s understanding
of foundational scientific concepts. This is traditionally captured in the scholarship with some
variant of a science quiz. The Science News Survey includes a series of quiz-style questions that
one might expect to see on a comprehensive science exam, with each question speaking to the core
building blocks of science. Scientific knowledge is operationalized as an additive index of
responses to ten science questions. Table 3.2 displays the questions utilized to construct the
knowledge scale, including eight true or false questions and two multiple-choice questions. These
questions were selected due to the lack of controversy surrounding the veracity of the scientific

55

claims. For each question answered in a manner consistent with the scientific community’s
position, respondents were awarded one point. Each respondent’s score was totaled, producing an
index that ranges from zero to ten, with a median score of seven (i.e., respondents correctly
answered 70% of the questions). The distribution of the quiz scores is plotted in Figure 3.1.
Figure 3.1: Distribution of science quiz scores
0.2
0.15
Percent of
0.1
survey sample
0.05
0

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Science quiz score (percent correct)
Before proceeding, three points are worth emphasizing when discussing the
operationalization of scientific knowledge. First, with respect to question selection, the overall
results presented in the following empirical chapters remain consistent when utilizing alternative
knowledge questions from the survey or varying the number of questions included in the index.
Second, the data were conditioned to remove respondents who completely ignored or
refused to answer the battery of science questions as well as respondents who answered “Don’t
Know” (DK) to all 19 of the science knowledge questions. This effectively removed 14 (refused
all questions) and 22 (responded DK to all questions) respondents from the analysis, respectively.
The rationale was that these respondents were not taking the survey seriously, so their score on the
science quiz did not reflect true ignorance to science. Removing these initial respondents did not
alter any of the analyses provided in the forthcoming chapters.

56

Table 3.2: Variable coding & operationalization
Variable
Question wording
Operationalization
Science
o “The center of the Earth is very hot” (true/false)
Additive index of the number of
knowledge
o “Antibiotics kill viruses as well as bacteria”
correct answers ranging from 0 to
(true/false)
10
o “Ordinary tomatoes, the ones we normally eat, do not
have genes, whereas genetically modified tomatoes
do” (true/false)
o “Lasers work by focusing sound waves” (true/false)
o “Stem cells occur only in plants” (true/false)
o “For the first time in recorded history, some species of
plants and animals are dying out and becoming
extinct” (true/false)
o “Electrons are smaller than atoms” (true/false)
o “All plants and animals have DNA” (true/false)
o “Which travels faster: light or sound?” (multiple
choice; answer options light, sound, and both the
same)
o “Would you say that the Sun is a planet, a star, or
something else?” (multiple choice; answer options a
planet, a star, something else)
Faith-inOrdinal response (strongly agree, agree, disagree, strongly 0 to 4 additive index, averaged;
science index disagree) to the following propositions:
high value equals strongly agree
o “Science and technology make lives healthier, easier,
across all issues
and more comfortable;
o Because of science and technology, there will be more
opportunities for the next generation;
o Most scientists want to work on things that will make
life better for the average person; and
o Even if it brings no immediate benefits, scientific
research which advances the frontiers of knowledge is
necessary and should be supported by the federal
government”
57

Mean
6.5

Std. dev.
2.6

3.2

0.6

Table 3.2 (cont’d)
Skeptical-of- Ordinal response (strongly agree, agree, disagree, strongly 0 to 4 additive index, averaged;
science index disagree) to the following propositions:
high value equals strongly agree
o “One of the bad effects of science is that it breaks
across all issues
down people’s ideas of right and wrong;
o The growth of science means that a few people could
control our lives;
o Science makes our way of life change too fast; and
o We depend too much on science and not enough on
faith”
Ideology
“In talking about politics, the expressions "liberal" and
Recoded on -5 to 5 scale;
"conservative" are often used. Please think of a scale from
“5” indicates most
0 to 10 where 0 means very liberal and 10 means very
conservative position, “-5”
conservative. Where would you locate yourself? If you are
indicates most liberal
not sure, you may check the "Not Sure" box.”
Democrat
“Generally speaking, do you usually think of yourself as a
Dichotomous 1/0 identifier;
Democrat, a Republican, an Independent, or what?”
“1” indicates Democrat
Republican
“Generally speaking, do you usually think of yourself as a
Dichotomous 1/0 identifier;
Democrat, a Republican, an Independent, or what?”
“1” indicates Republican
Age
Not available, information from Knowledge Networks
Continuous measure
Gender
Not available, information from Knowledge Networks
Dichotomous 1/0 identifier;
“1” indicates female
Four-year
“What is the highest degree or level of education that you
Dichotomous 1/0 identifier;
degree
have completed?”
“1” indicates bachelor’s
degree received
Advanced
“What is the highest degree or level of education that you
Dichotomous 1/0 identifier;
degree
have completed?”
“1” indicates advanced
degree (master’s, doctoral,
professional)

58

2.3

0.6

0.5

2.4

0.3

0.5

0.3

0.4

46.8
0.5

17.2
0.5

0.2

0.4

0.1

0.3

Lastly, it is beneficial to emphasize treatment of the DK responses. Mondak (2001) notes
that different results emerge when DK responses are encouraged in questions designed to assess
the knowledge of survey respondents. Inviting DK response creates a degree of “noise” in the
measure, as respondents who might be able to make informed guesses are more likely to be
discouraged from indicating an answer. Individuals who are lacking knowledge, however, are
more likely to guess (and guess correctly). Encouraging DK responses potentially creates a bias in
a knowledge scale, although there is disagreement on this front (e.g., Luskin and Bullock 2011).
If there is a bias, however, it appears that the bias is in the direction against the expected
relationships outlined in the second chapter. Given that respondents were offered a DK option on
the Science News Survey, a significant relationship would suggest that the science knowledge scale
sufficiently captures a systematic pattern despite the propensity of some groups to guess more than
others.
Normative Science Views
The last two measures proposed in the second chapter are the normative views of science. That
is, are scientists working for the common good and will society be better off with technological
advancements, or are there negative consequences for society’s deference to scientists?
Respondents were presented with a variety of questions and asked about their level of agreement
on a four-point ordinal scale, ranging from strongly disagree to disagree to agree to strongly agree.
To construct these measures, a principal-component factor analysis was performed on a range
of questions assessing normative attitudes about the scientific enterprise (presented in Table 3.2).
The factor loadings are presented in Table 3.3 and suggest that the two concepts load on separate
dimensions rather than falling along a single continuum. Factor one, which features positive views
about science and technological advancements, accounts for 32% of the variance. Factor two,

59

which includes four questions that express reservations about the scientific process, accounts for
29% of the variance. The two factors cumulatively explain 61% of the variance.
Table 3.3: Principal-component factor analysis results
Questions (from Table 3.2)
Factor 1
Factor 2
#1: Healthier, easier and comfortable
0.84
-0.07
#2: More opportunities
0.84
-0.00
#3: Science advances frontiers
0.70
-0.03
#4: Scientists want to better world
0.75
0.09
#5: Makes life change too fast
0.00
0.77
#6: Breaks down right/wrong
0.04
0.72
#7: Means few control lives
-0.08
0.66
#8: Not enough faith
0.02
0.78
Notes: N=1315. No weights were utilized in the analysis. An oblique promax rotation is utilized
due to correlation between the two factors.
Two indices were created, one for each factor. The first factor, which is labeled the “faithin-science index,” is an average score (rounded) of respondent’s answers to questions one through
four. A high score indicates strong agreement with the selected questions. Factor 2 is referred to
as the “skeptical-of-science index” and consists of an average score on the last four questions. A
high score indicates more skepticism or ambivalence over the effects of science and technological
advancement.
The dimensions are likely unique due to a degree ambivalence or conflict among the
respondents. That is, one can think science makes life healthier and more comfortable for society
but also be skeptical about science due to its tendency to facilitate change or threaten spiritual
beliefs. The distributions of these indices are presented in Table 3.4. The patterns suggest that
respondents generally look positively on the contributions of science and technological
advancements for society as 93% of the survey respondents indicated they agree or strongly agree
with the positive statements about science. Meanwhile, only 39% of the sample agreed or strongly
agreed with statements expressing skepticism about the actions of scientists. Only a small portion

60

of the survey respondents, 4.3%, may be considered anti-science in that they agree there are
reasons to be skeptical about science while also disagreeing that society has benefited from
scientific advancements. The table reveals, however, a degree of ambivalence. A notable number
of respondents found a reason to be skeptical about the work of scientists but also held generally
positive views about science. Almost 37% of the sample is in the bottom-right quadrant of the
table, the area where skeptical and favorable views of science are both expressed. This
ambivalence demonstrates the need to break the generalized science attitude measures into two
dimensions – faith and skepticism.
Table 3.4: Cross-tabulation of normative science views
Skeptical-of-science index
S. Disagree

Agree

S. Agree

S. Disagree
Faith-in
-science
index

Disagree

0.3%

-

-

0.1%

Disagree

0.1%

2.1%

3.7%

0.5%

Agree

1.0%

33.2%

27.8%

1.1%

S. Agree
3.4%
18.7%
6.6%
1.3%
Notes: N=1315. Reporting total percentage of the survey population within each cell. Indices
constructed using the questions outlined in Table 3.2. Distributions are weighted to the U.S. adult
population.
In order to discuss the relationships between normative science views and the outcome
measures in the forthcoming analyses, three profiles of individuals have been identified based of
the most populated cells in Table 3.4. These three profiles are:
Strong Pro-Science, shaded [ ]: These are individuals who strongly agree that society
benefits from scientific advancements (faith-in-science index score of four) and disagree
with propositions that there are reasons to be skeptical about the work of scientists
(skeptical-of-science index score of two).

61

Weak Pro-Science shaded [ ]: These are individuals who agree that society benefits from
scientific advancements (faith-in-science index score of three) and disagree with
propositions that there are reasons to be skeptical about the work of scientists (skepticalof-science index score of two).
Conflicted, shaded [ ]: These are individuals who agree with propositions that there are
reasons to be skeptical of the work of scientists (skeptical-of-science index score of three)
but also agree with propositions that there are benefits to scientific advancement (faith-inscience index score of three).
These profiles are shaded in Table 3.4 and represent the three most populated cells from the table.
When discussing the substantive effects of these indices, the analysis discusses differences in the
predicted probabilities of these three profiles. All other measures are held at their means during
this discussion.
Correlations
These three measures (faith-in-science index, skeptical-of-science index, and scientific
knowledge) represent components of the scientific orientations individuals possess. A correlation
matrix is presented in Table 3.5 and suggests that while there are some interrelationships between
the main concepts identified here, none are strong enough to suggest multicollinearity issues will
arise, at least when these three measures are considered.
Table 3.5: Correlation matrix of science measures (Pearson’s r)
Knowledge
Faith
Skepticism
1
Knowledge
0.1831
1
Faith
-0.2061
-0.1869
1
Skepticism

62

Means Analysis
Table 3.6 reveals the difference of means of the three science orientations by select
demographic measures. These bivariate relationships suggest there is a knowledge gap between
men and women. Males, on average, correctly answer one more question than their female
counterparts. This is consistent with prior literature examining gender differences, where this gap
is also noted (Hayes 2001; Hayes and Tariq 2000), although females do appear to know more about
climate change than their male counterparts despite this gap (McCright 2010). There is an age gap
as well, where the elderly (aged 66 and up) average one point lower on the quiz. Also of note is
the lack of variation between ideological groups. Conservatives (defined as those selecting the
three most conservative positions on an eleven-point ideological scale) are no different from
liberals (those selecting the three most liberal positions on the ideological scale). Lastly, there is
some overlap between the educational measures and science knowledge, which is expected, given
that participation in college science courses is required for a majority of degrees in the United
States. With respect to one’s level of faith in science, there is little difference in the mean scores
of demographic groups. A similar interpretation emerges when considering one’s level of
skepticism about science, although conservatives are a bit more likely to agree with cautious
statements by about one-half of one point on the four-point scale.
Political Ideology
As noted earlier, the citizenry’s orientation toward science is not the only dimension within
a belief system. The public utilizes other orientations as well. While there are no questions within
the Science News Survey to assess cultural values, there is a degree of overlap between cultural
and political values (Michaud et al. 2009; but see Ripberger et al. 2012). A focus on ideological
values is a valid analytical approach given this overlap as well as the political dynamics of climate

63

change. Policy solutions that seek to mitigate greenhouse gas emissions involve increased
government regulation, which may or may not clash with the political orientations of the citizenry.
Table 3.6: Demographic variation across measures of scientific orientation
Science
Skeptical-ofFaith-in-science
knowledge
science
Category
All
Gender

N
mean
1371 6.6

s.d.
2.6

N
mean
1353 3.2

s.d.
0.6

N
1335

mean
2.4

s.d.
0.6

Male
675
7.1
2.5
666
3.3
0.6
664
2.3
0.6
Female
696
6.1
2.5
687
3.2
0.6
671
2.4
0.6
Education
No Degree
952
6.0
2.5
943
3.2
0.6
930
2.4
0.6
Bachelors
280
7.8
2.1
274
3.3
0.6
275
2.2
0.6
Advanced
139
8.2
2.2
136
3.4
0.6
130
2.2
0.6
Age Cohort
18-25
154
6.8
2.4
154
3.3
0.6
156
2.4
0.7
26-35
241
7.0
2.4
242
3.2
0.6
240
2.4
0.6
36-45
288
6.9
2.5
287
3.2
0.6
284
2.4
0.7
46-55
242
6.8
2.7
240
3.2
0.6
235
2.3
0.6
56-65
226
6.5
2.5
218
3.2
0.6
217
2.4
0.6
66+
220
5.5
2.6
212
3.3
0.5
203
2.3
0.6
Ideology
Conservative
(Top 3)
266
7.1
2.4
257
3.2
0.6
252
2.5
0.7
Moderate
(Middle 5)
743
6.8
2.4
733
3.3
0.6
724
2.3
0.6
Liberal (Top 3)
130
7.3
2.6
126
3.4
0.6
129
2.0
0.7
Notes: Values reported in the table, from left to right, are number of observations, mean score
on the science knowledge quiz, and the standard deviation (s.d.).
A long line of literature has documented the citizenry view the world through a liberalconservative ideological lens (e.g., Converse 1964). A measure of political ideology is leveraged
to account for this alternative orientation utilized by the public to interpret policy debates. The
measure was recoded to a -5 (liberal) to +5 (conservative) scale in order to ease the interpretation
of the coefficients in the forthcoming regression. Respondents unsure of their placement on the
scale (236 respondents) were removed from the analysis, although supplementary analyses suggest

64

there are no broad-level interpretative differences if these respondents were recoded as moderates.
In fact, in some cases, the inclusion of these additional survey respondents leads to even stronger
support for the hypotheses, suggesting that the direction of any omission bias is against the
specified hypotheses.
In addition to ideology, a battery of demographic figures is included at times in the
analyses. These include age (continuous), gender (dichotomous, female high value), partisanship
(two separate dichotomous measures for Republican and Democrat, independents as the reference
group), and two alternative education measures (two dichotomous measures for the type of college
degree earned, four-year and advanced). These measures are also included in Table 3.2. These
measures are included to account for additional explanatory power no captured by orientations.
Given the science knowledge gaps noted earlier for age and gender, women and the elderly
population may reach conclusions about climate change due to unique viewpoints not captured by
science and political orientations. With respect to the education metrics, other educational
processes, outside exposure to science courses, may provide the skills necessary for the public to
evaluate causal arguments about climate change. The forthcoming analyses considers multiple
models that test whether adding these socio-demographic controls account for any unique variation
left unexplained by science and political orientations.
Outcome Measures
The two-step model described in Chapter 2 conceptualized two types of dependent
variables. The first stage requires measures that represent an individual’s understanding of climate
change. Three outcome variables were identified within the Science News Survey that meet this
requirement. The dependent variables selected for the first-stage of the analysis include the
following true or false quiz-style questions, all of which come from the first wave of the survey:

65



The greenhouse effect causes the Earth's temperature to rise;



The primary human activity that causes global warming is the burning of fossil fuels
such as coal and oil; and



Global warming is increasing primarily because the level of direct radiation from the
Sun is increasing.

Each question represents various stages of the public debate on climate change. The first question
covers the basic outcome associated with the greenhouse effect while the second addresses the
possibility that human activity can influence the environment. On both issues, there is considerable
certainty among climatologists that the greenhouse effect causes warming and burning fossil fuels
emits greenhouse gases. The greenhouse effect theory is supported by basic physics, as is the
argument that burning fossil fuels can release greenhouse gases and contribute to global warming.
The third question taps a more politically- and scientifically-contested issue: whether solar
radiation (a natural force) is primarily responsible for global warming. Climatologists at the IPCC
do not dismiss the role of solar radiation, but the 2007 report (IPCC 2007) references
anthropogenic causes as the primary cause of climate change (but see Shaviv 2006). As such, an
answer consistent with the majority of climatologists would be FALSE for the third question and
TRUE for the first and second questions.15 Each question featured three answer options – true,
false, and do not know. Consistent with prior studies (Durant et al. 1989; Miller 1998), each
measure was recoded into a dichotomous measure so that agreement with the consensus position
15

A careful interpretation of the question wording is appropriate on this front. The third question
only speaks to whether solar radiation is the primary driver of global warming. For those answering
FALSE, it cannot be definitively determined whether they view anthropogenic forces as the
primary driver or perhaps some other natural variation. It is also possible respondents answered
FALSE because they completely reject the idea of global warming.

66

of scientists is the high value while the incorrect and do not know responses represent a zero
value.16
The second stage of the analysis requires a measure that represents the perceived threat of
climate change. In the third wave of the survey, respondents were asked the following question:
“How concerned are you about global climate change?” Survey respondents were presented with
a five-point scale to indicate their level of concern. The options included: very concerned,
concerned, mildly concerned, largely unconcerned, and totally unconcerned. In Chapter 5, this
measure of concern is used as an additional dependent variable when testing the two-step process.
Model Presentation & Next Steps
The first empirical chapter considers only the first stage of the two-step process. This initial
test focuses on the relationship between orientations and acceptance of causal arguments offered
by climatologists. The hypotheses presented in Table 3.7 are tested in this analysis. Each
hypothesis is restated and accompanied by the independent variable relevant to assessing the
relationship. Figure 3.2 diagrams the expected relationship between these key variables and the
three outcome measures note earlier. The outcome measures, again, represent the causal arguments
offered by climatologists.

16

This coding has potential consequences for the analysis. The rationale for the coding is that the
assessment seeks to identify those who agree with scientists. For the science comprehension thesis,
there is no fundamental difference between DKs and incorrect answers. Commentary in the
literature suggests some DK responses could be randomly recoded with true values to offset any
biases due to the propensity of some groups to guess more than others (Mondak 2001). Recent
literature has operationalized the DKs as a middle category residing between clear assertions of
agreement/disagreement (e.g., Kahan et al. 2011). Other work, however, suggests DK responses
do not reveal very much missing knowledge (Luskin and Bullock 2011).
67

Table 3.7: Hypotheses for first-stage analysis
Hypothesis
Measure
Hypothesis 1: Individuals with higher levels of scientific knowledge Science knowledge
accept the causal arguments from climatologists at higher rates than their quiz score
low-knowledge counterparts, holding all else constant.
Hypothesis 2: Individuals perceiving benefits from scientific
advancement are more likely to accept causal arguments from
climatologists, holding all else constant.
.
Hypothesis 3: Individuals skeptical of scientists and their work are less
likely to accept the causal arguments from climatologists, holding all
else constant

Faith-in-science
index

Skeptical-of-science
index

Hypothesis 4: Liberals (conservatives) are more likely to agree Ideology
(disagree) with causal arguments consistent with the IPCC, holding all
else constant.
Hypothesis 5: Political ideology conditions the relationship between Interaction: Science
scientific knowledge and climate change beliefs, holding all else knowledge *
constant.
ideology

Figure 3.2: Proposed relationships between selected orientations and acceptance of climate
change arguments
Science orientations
H1 Science knowledge (+)
H2 Faith-in-science index (+)
H3 Skeptical-of-science index (-)

Understanding of climate change
1) The greenhouse effect causes the
Earth's temperature to rise;
2) The primary human activity that
causes global warming is the
burning of fossil fuels such as coal
and oil; and
3) Global warming is increasing
primarily because the level of
direct radiation from the Sun is
increasing.

Political orientations
H4 Political ideology (+)
Orientation interaction
H5 Science knowledge by ideology (-)
Socio-demographic controls
Age (-)
Gender (+)
Democrat (+)
Republican (-)
Four-year degree (+)
Advanced degree (+)
68

Chapter 4 assesses Hypotheses 1 through 5. Each outcome measure – greenhouse effect,
fossil fuels, and solar radiation – is considered separately. The analysis presents three logistic
regression models for each outcome measure. These models take the following form:

 Equation 1: ln (

𝑝

1−𝑝

) = 𝑏0 + 𝑏1 ∗ 𝑠𝑐𝑖𝑒𝑛𝑡𝑖𝑓𝑖𝑐 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒 + 𝑏2 ∗

𝑓𝑎𝑖𝑡ℎ 𝑖𝑛 𝑠𝑐𝑖𝑒𝑛𝑐𝑒 + 𝑏3 ∗ 𝑠𝑘𝑒𝑝𝑡𝑖𝑐𝑎𝑙 𝑜𝑓 𝑠𝑐𝑖𝑒𝑛𝑐𝑒 + 𝑏4 ∗
𝑝𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦 + 𝜀
 Equation 2: ln (

p

) = 𝑏0 + 𝑏1 ∗ 𝑠𝑐𝑖𝑒𝑛𝑡𝑖𝑓𝑖𝑐 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒 + 𝑏2 ∗

1−p

𝑓𝑎𝑖𝑡ℎ 𝑖𝑛 𝑠𝑐𝑖𝑒𝑛𝑐𝑒 + 𝑏3 ∗ 𝑠𝑘𝑒𝑝𝑡𝑖𝑐𝑎𝑙 𝑜𝑓 𝑠𝑐𝑖𝑒𝑛𝑐𝑒 + 𝑏4 ∗
𝑝𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦 + 𝑏5 ∗ 𝑠𝑐𝑖𝑒𝑛𝑡𝑖𝑓𝑖𝑐 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒 ∗
𝑝𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦 + 𝜀
 Equation 3: ln (

𝑝

1−𝑝

) = 𝑏0 + 𝑏1 ∗ 𝑠𝑐𝑖𝑒𝑛𝑡𝑖𝑓𝑖𝑐 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒 + 𝑏2 ∗

𝑓𝑎𝑖𝑡ℎ 𝑖𝑛 𝑠𝑐𝑖𝑒𝑛𝑐𝑒 + 𝑏3 ∗ 𝑠𝑘𝑒𝑝𝑡𝑖𝑐𝑎𝑙 𝑜𝑓 𝑠𝑐𝑖𝑒𝑛𝑐𝑒 + 𝑏4 ∗
𝑝𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦 + 𝑏5 ∗ 𝑠𝑐𝑖𝑒𝑛𝑡𝑖𝑓𝑖𝑐 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒 ∗
𝑝𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦 + 𝑏6 ∗ 𝑑𝑒𝑚𝑜𝑝𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑠 + 𝜀
The first equation provides a means to test Hypotheses 1 through 4. The second equation adds the
interactive measure, allowing for an assessment of Hypothesis 5. The third equation includes the
socio-demographic controls. Dividing the analysis into multiple components allows a test of
whether the interaction term and socio-demographic measures substantively contribute additional
69

explanatory power to the core model. Specifically, considering the first and second models
separately assesses whether conceptualizing the science comprehension thesis as conditional on
ideological values significantly improves the explanatory power of the overall model. Model 3
provides additional analysis as to whether the socio-demographic controls capture significant
amounts of unexplained variation, whether through the gender, age, or education gaps noted above.
The last empirical chapter utilizes path analysis to test the causal process outlined in the
two-step model. This analysis includes two additional hypotheses, re-stated in Table 3.7. Assessing
the full model requires taking the dependent variables of the first stage and using them as
independent variables that predict climate change concern (see Figure 3.3). Path analysis is
selected because the technique was designed to assess the structural relationships specified within
causal processes such as the two-step process described here (Kline 2005). The intent of the
analysis is to only show support for the two-step process – not compare a range of competing
causal models. The technique will assess whether scientific knowledge has a prior effect on other
endogenous measures that, in turn, correlate with higher levels of concern about climate change.
Table 3.8: Hypotheses for full two-step model
Hypothesis
Measure(s)
Hypothesis 6: Individuals who understand climate change in a
manner consistent with climatologists are more likely to Solar radiation and fossil
demonstrate higher levels of concern.
fuel beliefs
Hypothesis 7: Individuals concerned about climate change are Climate change knowledge,
knowledgeable about basic scientific principles.
scientific knowledge
Summary
The discussion now turns toward assessing the first step of the two-step process developed
in Chapter 2. To reiterate, the two-step process is a conceptualization of public opinion that is
utilized to assess both the science comprehension and value-centered theses. The two theses have
led to inconsistent, and in some cases conflicting, results. The arguments, however, are not
70

mutually exclusive. Rather, the inconsistencies that have emerged when analyzing climate change
attitudes are the result of improper model specification. The two-step process is an attempt at
addressing this inconsistency in the literature.
Figure 3.3: The two-step process
Science
orientations
Understanding
of the problem
Political
orientations

71

Problem
recognition

CHAPTER 4: ORIENTATIONS AND ACCEPTANCE OF CAUSAL ARGUMENTS
This opening empirical chapter assesses the first stage of the two-step model. The model
conceptualizes “problem recognition” as a bridge between basic science knowledge and climate
change concern. That is, individuals need to recognize the problem before deciding whether they
are concerned about the problem. As policy actors attempt to frame the problem to the public, they
offer different causal arguments that can lead to divergent understandings of the problem. This
chapter considers, then, the direct relationship between both political and scientific orientations
with the acceptance of three specific arguments relevant to climate change: whether the greenhouse
effect produces global warming, if burning fossil fuels is the primary mechanism by which humans
warm the Earth, and whether warming is the result of increased solar radiation.
The proposed relationships discussed in the prior chapters are diagramed in Figure 4.1. The
first hypothesis expects a positive relationship between scientific knowledge and climate change
beliefs. Individuals who score higher on the ten-point knowledge scale should demonstrate an
increased likelihood of agreeing with the three propositions about climate change. Knowledge,
however, only captures one dimension of an individual’s scientific orientation. Predispositions
about scientists are potentially important as well. Hypothesis 2 captures the expected relationship
between positive normative views of science and climate change beliefs. Those who view society
as better off due to scientific advancements and view scientists themselves as working to benefit
society are going to demonstrate a greater tendency to agree with scientists on climate change. The
third hypothesis assesses relationships related to skeptical normative views, such as those of
individuals who question the work of scientists and are open to other methods of justifying beliefs.
Such individuals view science skeptically, and as such, they are less likely to agree with scientists
in general and climatologists in particular. As shown in Table 3.3, the bulk of the population (94%)

72

perceives some level of benefits from science and technological advancement. Roughly 35% of
these respondents, however, also find reasons to doubt the work of scientists. This group expresses
an ambivalent attitude toward the contributions of scientists. In short, individuals can score high
on both indices as noted in chapter three. The second and third hypotheses, then, are considered
separately due the multi-dimensionality of the two constructs.
Political orientations are the focus of the fourth hypothesis. Due to policy preferences,
views of appropriate government activity, as well as a tendency to utilize heuristics offered by
ideological leaders, the fourth hypothesis expects conservatives to reject the views of
climatologists at greater rates. Meanwhile, liberals are more likely to share the climatologists’
beliefs. Moderates will fall somewhere between liberals and conservatives when it comes to their
support for a given proposition.
Figure 4.1: Proposed relationships between selected orientations and acceptance of climate
change arguments
Science orientations
H1 Science knowledge (+)
H2 Faith-in-science index (+)
H3 Skeptical-of-science index (-)

Understanding of climate change
1) The greenhouse effect causes the
Earth's temperature to rise;
2) The primary human activity that
causes global warming is the
burning of fossil fuels such as coal
and oil; and
3) Global warming is increasing
primarily because the level of
direct radiation from the Sun is
increasing.

Political orientations
H4 Political ideology (+)
Orientation interaction
H5 Science knowledge by ideology (-)
Socio-demographic controls
Age (-)
Gender (+)
Democrat (+)
Republican (-)
Four-year degree (+)
Advanced degree (+)

73

The next point of inquiry (Hypothesis 5) considers the possibility of an interactive effect
between scientific knowledge and political ideology. The presumption noted earlier was that highknowledge individuals are more likely to follow ideological cues and reject expert positions. This
last hypothesis considers whether the science comprehension thesis is affected by political values.
If individuals rely on values and predispositions as expected, high-knowledge conservatives and
liberals should view climate change differently when compared to ideologically neutral moderates.
The earlier descriptive analysis suggested that gender and age gaps exist in respect to
science knowledge. Additional literature also notes differences in environmental attitudes
according to gender, with females often more concerned about climate change and more
knowledgeable about climate change facts (McCright 2010, but also see Hayes 2001). Including
age and gender as socio-demographic controls offers additional information as to whether there is
something unique about women or age that cannot be explained via scientific knowledge or
political values. If scientific knowledge is significant while age and gender both remain
insignificant, then perhaps the science comprehension thesis is one means of understanding these
demographic gaps in environmental attitudes.
The educational achievement and partisan control measures are included for comparison
purposes. Reasonable counterarguments to the model specification might emphasize partisanship
and educational values as more dynamic predictors of opinions. Experience with science may not
be a necessary condition to sort through the climate change debate. In the absence of science
knowledge, individuals with non-scientific knowledge may have sufficient skills to decipher the
political debate and evaluate the claims of scientists. Similarly, partisanship may function as the
optimal predictor of attitudes – especially if survey respondents were unable to understand the
meaning or intent of the ideology question on the survey. The expected directions of the

74

coefficients for these additional relationships are noted in the diagram, although the science and
political orientation measures are expected to explain the majority of the variation.
Climate change beliefs are operationalized with the three previously noted survey questions
that speak to different aspects of the causal argument. To reiterate, the first outcome measure asks
respondents about the outcomes associated with the greenhouse effect while the second outcome
measure considers the relationship between humans and the damage caused by burning fossil fuels.
The last outcome measure considers a specific counterargument to the scientific community – that
solar radiation is primarily responsible for global warming. These three questions constitute the
dependent variables to be considered here. As noted earlier, there is more certainty among
climatologists in respect to the first two measures. However, the solar radiation proposition is more
complicated and contested, with perhaps less certainty among select scientists. Despite these
differences, the initial expectation is that the hypothesized relationships will emerge throughout
the analysis, regardless of the outcome measure.
Each measure is coded on a dichotomous scale, with high values (coded as “1”) indicating
agreement with the causal arguments from the climatologists working with the IPCC. Low values
(coded as “0”) indicate a failure to acknowledge these positions – either an incorrect answer or a
“Don’t Know” response. This operationalization tests for explicit acknowledgement by
respondents to the arguments of climatologists.
The forthcoming analysis will consider each outcome variable separately. For each
outcome, the discussion will first note the extent to which support for the hypotheses is found
within the models. This is followed by a comparison of the three equations noted earlier, including
how well each explains the variation within climate change beliefs. The substantive effects of key

75

variables are then discussed. The chapter will close with a comparison of all three issues and a
discussion of the implications associated with the results.
Analysis: Greenhouse Effect
The first outcome measure introduced here addresses a core scientific concept in climate
change discussions: the greenhouse effect. The question asked respondents whether the following
statement is true or false: “The greenhouse effect causes the Earth's temperature to rise.” Table 4.1
presents the logistic regression coefficients and z-score for models one through three. The first
model offers an assessment of the first four hypotheses, with the model supporting three of the
four expectations. First, a statistically significant relationship emerged between science knowledge
and responses to the question. Individuals with elevated levels of scientific knowledge are more
likely to identify the outcomes of the greenhouse effect (Hypothesis 1). Second, individuals who
scored higher on the faith-in-science index are also significantly more likely to support for the
proposition (Hypothesis 2). Third, an ideological divide is observed. The coefficient for political
ideology is statistically significant and suggests that conservatives are more likely to reject the
proposition compared to moderates while liberals are more likely than moderates to accept the
proposition (Hypothesis 4). Lastly, the direction of the relationship between those skeptical of
science and the proposition in question is negative (Hypothesis 3). More skepticism corresponds
with a lower likelihood of identifying outcomes associated with the greenhouse effect. However,
the relationship itself is not significant. The magnitude of these relationships will be taken up later,
but for now, there is support for both the science comprehension thesis as well as the valuecentered thesis.

76

Table 4.1: Acceptance of the greenhouse effect argument
Model 1
Model 2
0.17
0.26
Scientific knowledge
(4.59)
(6.34)
0.42
0.42
Faith in science
(2.48)
(2.50)
-0.08
-0.05
Skeptical of science
(-0.52)
(-0.32)
-0.23
0.26
Ideology
(-4.87)
(2.48)
Ideology X science
-0.08
knowledge
(-5.16)
Republican

-

-

Democrat

-

-

Age

-

-

Female

-

-

4-yr degree

-

-

Advanced degree

-

-

Model 1: N=1082

-1.22
(-1.63)
Log-likelihood -578

Wald Chi2

Model 2: N=1082

Log-likelihood -559

Wald Chi2

Constant

-1.75
(-2.43)

Model 3
0.24
(5.55)
0.42
(2.47)
-0.06
(-0.39)
0.28
(2.78)
-0.08
(-4.97)
-0.02
(-0.09)
0.29
(1.13)
-0.01
(-1.67)
0.09
(0.49)
0.22
(0.91)
0.49
(1.45)
-1.32
(-1.62)
70.0 PCP 72.5%
83.7 PCP 72.7%

Model 3: N=1082
Log-likelihood -553 Wald Chi2 91.6 PCP 73.1%
Notes: The first number represents the coefficient derived from a logistic regression. The
number in parenthesis indicates the z-score from the significance test. In all models, a weight
provided by Knowledge Networks was utilized in order to make inferences from the results back
to the U.S. adult population.
The percent correctly predicted by each model was also calculated. This analysis compares
the predicted probability that respondents agree with a proposition with the actual observed answer
for each respondent. If the model calculates a predicted probability that a respondent agrees with
the proposition at a rate greater than 50% and if the responded actually agreed with the proposition,
then it is a successful prediction. Conversely, if the predicted probability was less than 50% and

77

the respondent did not indeed support the proposition, then it is also a successful prediction. The
results of this analysis are displayed in Table 4.2. Overall, the model successfully predicts answers
for 72.5% of the respondents. A more nuanced look is obtained by looking at the sensitivity (how
many of those who agreed were predicted successfully) and the specificity (how many of those
who disagreed were predicted successfully). The sensitivity was 94.6% while the specificity was
16.4%. Many of the incorrect predictions, then, come from people who answered false yet for
whom the model predicted a true value. These calculations suggest there is perhaps a missing
measure from the model, one that would facilitate better predictions of those disagreeing with the
proposition.
Table 4.2: Percent correctly predicted for each model
Model 1
Model 2
Percent correctly predicted
72.5%
72.7%
Sensitivity
94.6%
94.2%
Specificity
16.4%
18.4%

Model 3
73.1%
94.7%
18.4%

Model 2 in Table 4.1 adds an interactive term to assess the fifth hypothesis, that ideology
moderates the relationship between knowledge and beliefs. Support for an interactive effect is
found on multiple fronts. First, the interactive term and its two components are all significant. The
coefficients suggest a more nuanced interpretation of the science comprehension thesis is
warranted. A high-knowledge moderate (when ideology and the interaction equal zero) is more
likely to agree with climatologists (coefficient of 0.26). With the ideology coefficient, there is a
sign reversal from negative to positive. This measure is interpreted as representing low-knowledge
ideologues (Brambor et al. 2006). Among those with no knowledge (a zero on the quiz score), the
coefficients suggest it is conservatives who are more likely to agree with the proposition than both
moderates and liberals.

78

The interactive term is interpreted as high-knowledge ideologues. The negative coefficient
suggests high-knowledge conservatives are less likely to agree with the proposition compared to
their high-knowledge counterparts while high-knowledge liberals are more likely agree with the
proposition at greater rates. The overall effect of ideology remains negative in the second model.
The substantive details of this interaction will be taken up shortly after discussing the other models.
Also supporting the fifth hypothesis is the overall improvement in the model’s fit. A Wald
test was utilized to test whether constraining the interaction term to zero significantly reduces the
overall fit of the model. The test indicates that the fit of the model is improved by including the
interactive term (p < 0.001). A similar analysis was performed with a likelihood ratio test, which
compares whether the difference in the log-likelihood between the two models is significant. This
statistic is two times the difference in the log-likelihood for each model. The derived statistic is
38.8 with one degree of freedom, which is significant according to the chi-square distribution
matrix (p<0.001).
While the Wald and log-likelihood ratio test statistics suggest the second model improves
on the first, the overall predictive power of the model is not improved by adding the interactive
term. Referring back to Table 4.2, the second model’s prediction success rate is only 72.7% - an
increase of 0.2%. The second model performs equally well when looking at the sensitivity and
specificity. The consistency of this imbalance suggests identifying high-knowledge conservatives
as less likely to agree with the proposition does not significantly improve the predictive capabilities
of the model.
The third model adds the socio-demographic controls. With respect to the core hypotheses,
there is no change in the interpretation compared to model two. In respect to the socio-demographic
controls, none of the measures achieve significance in the model. The direction of the relationships,

79

however, fits with expectations. The strength of the science knowledge coefficient and the lack of
a significant age and gender relationship suggests differences along these socio-demographic
identities might indeed by the result of differing levels of scientific knowledge. Furthermore, a
Wald test of the socio-demographic controls as a block suggests that while the Wald statistic
increases with the third model, it did not do so significantly. A similar interpretation is found with
the likelihood ratio test. A statistic of 11.8 is derived with the previously noted formula. With six
degrees of freedom, the additional variables are not significant at the 95% confidence level. With
respect to the model’s predictive power, the socio-demographic measures offer little additional
explanatory power. The overall predictive power does improve slightly (73.1% correctly
predicted), but there is little change in the sensitivity or specificity.
Substantive Effects
The preceding discussion identifies statistically significant relationships within the data,
but statistical relationships do not tell the full story (Kline 2013). The next section explores the
substantive effects using the coefficients from Model 3 of Table 4.1. Although Model 3 offered no
additional explanatory power in this context, the controls do increase performance of the model in
later analyses. For consistency purposes, the coefficients from the third model will be utilized
throughout all substantive analyses.
As knowledge accumulates, the likelihood that respondents agree with the scientific
community increases. The predicted probability of agreement with climatologists for those with
medium knowledge (score of 7 on the quiz) is 76.9% of the time – holding all other variables in
Model 3 at their respective means.17 A move to full knowledge increases the predict probability

17

All predicted probabilities are derived from Model 3 by using the Delta method, holding all
other independent variables at their respective means unless otherwise noted.
80

of agreement with climatologists to 85.8%. While the nine-point improvement is notable and
statistically significant, it is important to stress that the model accurately predicts responses for
individuals even at medium levels of knowledge. In fact, an individual who hypothetically guesses
on the true and false questions to achieve a 50% knowledge-quiz score would still be predicted to
agree with the question 69% of the time. The average marginal effect, which represents the average
change in the predicted probability for a one-unit change in knowledge, is 2.6%. On average, then,
answering one additional question correctly on the quiz corresponds to a 2.6% increase in the
probability that a respondent agrees with the proposition.
Turning attention toward the normative science views, recall there was mixed support for
hypotheses two and three. As seen in Table 4.1, those with higher levels of faith in science are
more likely to identify the outcome of the greenhouse effect, consistent with Hypothesis 2.
However, those who question science do not appear to stand out in their survey responses. To
explore this relationship further, the three profiles of normative views noted in Chapter 3 are
considered in Table 4.3. The weak pro-science profile provides a baseline to examine an increase
on the faith-in-science index from “agree” to “strongly agree” (the strong pro-science profile). A
move along this index, holding skeptical science views at “disagree” and all others measures at
their respective means, increases the probability of agreement by 8% - an increase that is roughly
equivalent to a move from medium to full knowledge on the science knowledge quiz. There is
little noticeable difference between the weak pro-science and conflicted individuals, where there
is just over a 2% difference in the probability of agreeing with climatologists. Regardless of which
profile respondents fall under, there is still a high probability of agreement with climatologists
when looking at normative science views.

81

Table 4.3: Probability of identifying greenhouse effect outcome by normative views of science
Strong pro-science:
Weak pro-science:
Conflicted:
Faith science (s.
Faith science (Agree), Faith science (agree)
agree), skeptical
skeptical science
skeptical science
science (disagree)
(disagree)
(agree)
Greenhouse Effect
82.1%
75.1%
72.9%
Notes: Probabilities calculated with coefficients from Model 3 of Table 4.1. All other variables
outside those indicated here are held at their respective means.
Also consistent with expectations, conservatives are less likely to identify the outcome
associated with the greenhouse effect. A strong conservative (at the “5” position on the elevenpoint ideological scale) retains a 49.4% probability of accepting the proposition, all other variables
held at their respective means. Liberals at the opposite end of the scale retain a predicted
probability of 93.4%. Moderates appear closer to liberals than conservatives, all things equal, with
a calculated probability of 78.8%. This ideological difference is expected given the divergent
political messages offered by political leaders about climate change. Unlike the other hypotheses,
the fourth hypothesis that focuses on ideology is particularly notable given the 40-point gap in
predicted probabilities for conservatives and liberals (all things equal). The average marginal effect
for ideology is approximately 50% larger than that calculated for science knowledge. A one-point
movement along the ideological continuum results, on average, in a 4.0% change in probability of
agreeing with the proposition. The substantive strength of the relationship suggests political
ideology is a potent factor utilized by the public to interpret policy debates that involve scientific
information.
While there is support for the science comprehension thesis here, the interactive effects
suggest the thesis requires some qualification. The models support the interactive relationship
between scientific knowledge and political ideology (Hypothesis 5). The interactive nature of the
relationship is displayed in Figures 4.2 and 4.3. The first figure shows the probability respondents
agree with the proposition for three types of ideologues – strong conservatives (“5” position on
82

the ideology scale), strong liberals (“-5” position) and moderates (“0” position). The 95%
confidence intervals for each of the groups is included as well. Consistent with the prior
interpretation, the figure suggests low-knowledge conservatives are more likely to identify the
outcome associated with the greenhouse effect compared to moderates and liberals possessing
similar levels of knowledge. However, the slope is at a downward angle, with conservatives
becoming less likely to share the beliefs of climatologists as they accumulate scientific knowledge.
Meanwhile, as liberals and moderates accumulate knowledge, the probability they associate the
greenhouse effect with warming increases.
At low levels of knowledge, the chart suggests there is little difference between ideologues
at opposite ends of the scale. Respondents who can correctly answer 30 to 40% of the questions
possess approximately the same predicted probability of agreeing with climatologists (around
60%). The ideological differences, however, start to emerge once respondents pass the halfway
point on the quiz. For liberals, once medium levels of knowledge are obtained, there is a
remarkably high probability that respondents will agree with the proposition. A move from
medium to full knowledge increases the probability by almost six points (93.6% to 99%).
Moderates who move from medium to full knowledge experience a bit more change in prediction
capabilities, moving from a 79.0% to 88.4% probability of agreement. For both moderates and
liberals, there is support for the science comprehension thesis.

83

Figure 4.2: Probability agree with IPCC by science knowledge for liberals (-5), moderates
(0), and conservatives (5)
1
0.9

0.8
0.7
0.6
Probability
Agree with 0.5
IPCC
0.4
0.3
0.2
0.1
0
0%

10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Conservative
Liberal

Moderate
Confidence Intervals

As shown in the figure, the probability of conservatives agreeing with the proposition
slowly declines as knowledge is accumulated. There is a double-digit drop in the predicted
probability of agreement with the proposition for strong conservatives. A move from medium to
full knowledge (a drop from 49.2% to 37.8%). A high-knowledge conservative is almost 50 points
away from a moderate and 60 points away from a liberal.
As a whole, this is strong support for Hypothesis 5, as the science comprehension thesis
appears to explain climate change beliefs for only two of the three ideological groups. The valuecentered thesis finds support due to the apparent reliance of conservatives on other views besides
those from climatologists. That high-knowledge conservatives were less likely to agree with lowknowledge conservatives was not hypothesized in Chapter 2, although the finding is consistent
with the observation that ideologues are more educated (Taber and Lodge 2006) and that these

84

ideologues are likely to adopt positions opposite their opponents in policy debates (Nicholson
2012). Furthermore, the s-shaped pattern for strong liberals is of note. They more readily embrace
the position of climatologists as scientific knowledge accumulates because they are less likely to
question the science since the findings and conclusions generally conform to their predispositions.
This pattern supports both the value-centered and science comprehension theses.
Figure 4.3 above shows the average marginal effects of science knowledge by ideology.
The figure reveals an increase in the predicted probability of liberals and moderates identifying
the outcome of the greenhouse effect as they accumulate knowledge. For each question answered
correctly by a strong liberal, the predicted probability increases by 3.6% on average, all things
equal. A move from a medium score of seven to maximum knowledge at 10 would produce, on
average, a 10.8% increase in the predicted probability. This improvement in prediction increased
by an average of 4.3% for moderates, or about 12.9% as one moves from medium to full
knowledge. However, all things equal, the probability of predicting the beliefs of conservatives
does not improve as they accumulate scientific knowledge. The probability of successfully
predicting strong conservatives (ideology=5) drops by an average of 3.4% for each knowledge
question answered successfully. A move from medium to full knowledge would result in a 10.2%
drop in predictive capability, all things equal. There appears to be no significant benefit from
accumulating knowledge for more conservatives in the “3” and “4” positions on the ideology scale.
These results from Figure 4.3 are interpreted as continued support for the both theses.

85

Figure 4.3: Average marginal effect of scientific knowledge by ideology
0.20
0.15
0.10
0.05
Average
0.00
Marginal Effect
-0.05
-0.10
-0.15
-0.20

Liberal

Conservative

Summary: Greenhouse Effect
When it comes to understanding the outcomes associated with the greenhouse effect, there
is support for four of the five hypotheses. Consistent with the first hypothesis, high-knowledge
individuals answered the question differently from their low-knowledge counterparts, suggesting
support for the science comprehension thesis. The thesis, however, requires qualifications. The
fifth hypothesis, that ideology moderates the relationship between knowledge and beliefs, is also
supported. The analysis suggests the science comprehension thesis can explain beliefs of liberals
and moderates but not conservatives. This suggests there are some limits to knowledge, given that
knowledgeable individuals will defer to their predispositions and values.
The substantive strength of the relationship seems notable at first, but it is important to
remember the models retain considerable predictive power regardless of whether individuals
possess average or high knowledge. In this sense, there does not appear to be much value
86

associated with taking extra steps to accumulate additional scientific knowledge. Still, individuals
with low-knowledge retained predicted probabilities below 50%, so there is a notable change as
knowledge accumulates. The effect of accumulating knowledge occurs as respondents move from
low to middle levels of knowledge, rather than from average to high levels of knowledge.
Comparatively, a move in ideological views corresponds with a greater change in predicted
probability. This suggests that while knowledge is one explanatory concept, ideology is a more
powerful predictor of beliefs about the greenhouse effect.
Furthermore, individuals retaining a degree of faith in science were more likely to answer
the question in a manner consistent with climatologists. The substantive effects, however, are
limited. As noted in Chapter 3, most respondents either agree or strongly agree with the positive
benefits of science. A move from “agree” to “strongly agree” improved the probability of
agreement by 7%. The substantive effect of such a move is similar to a move on the scienceknowledge scale from medium to full knowledge. Those cautious of scientific advancements,
however, did not stand out in their beliefs as expected, suggesting no support for the third
hypothesis in the context of the greenhouse effect. These results will be compared to the other two
issues toward the end of the chapter. For now, the analysis turns to propositions about fossil fuels,
followed by solar radiation.
Analysis: Fossil Fuels
The second analysis focuses on whether the burning of fossil fuels is the primary way in
which society can cause global warming. Specifically, the question asked the following true and
false question: “The primary human activity that causes global warming is the burning of fossil
fuels such as coal and oil.” Table 4.4 presents the results from the logistic regression using
equations one through three specified earlier.

87

Table 4.4: Acceptance of the fossil fuels argument
Model 1
Model 2
0.14
0.19
Scientific knowledge
(3.95)
(5.25)
0.39
0.41
Faith in science
(2.75)
(2.83)
-0.04
0.00
Skeptical of science
(-0.27)
(0.01)
-0.19
0.24
Ideology
(-5.14)
(2.56)
Ideology X science
-0.07
knowledge
(-4.89)
Republican

-

-

Democrat

-

-

Age

-

-

Female

-

-

4-yr degree

-

-

Advanced degree

-

-

Model 1: N=1082

-1.70
(-2.44)
Log-likelihood -671.3

Wald Chi2

Model 2: N=1082

Log-likelihood -654.8

Wald Chi2

Constant

-2.12
(-3.00)

Model 3
0.19
(5.07)
0.35
(2.44)
-0.00
(-0.00)
0.24
(2.62)
-0.06
(-4.52)
-0.26
(-1.27)
0.33
(1.54)
0.00
(0.32)
-0.32
(-1.97)
0.08
(0.42)
0.03
(0.12)
-1.93
(-2.45)
56.6 PCP 64.1%
76.3 PCP 64.9%

Model 3: N=1082
Log-likelihood -647.6 Wald Chi2 89.2 PCP 66.8%
Notes: The first number represents the coefficient derived from a logistic regression. The
number in parenthesis indicates the z-score from the significance test. In all models, a weight
provided by Knowledge Networks was utilized in order to make inferences from the results back
to the U.S. adult population.
Within the first model, there is again support for three of the first four tested hypotheses.
Individuals with higher performance on the science knowledge quiz were significantly more likely
to agree with the proposition (Hypothesis 1). Similarly, individuals who agreed with the positive
statements about science were also more likely to express opinions similar to scientists (Hypothesis
2). Empirical support for an ideological gap also emerges. Again, conservatives were significantly

88

less likely to support the proposition than liberals and moderates while liberals agreed with the
proposition at higher rates than moderates (Hypothesis 4). Just as before, the third hypothesis finds
no support. Those holding more skeptical attitudes about science were less likely to agree that
burning fossil fuels is the primary mechanism by which humans cause global warming.
The proportions of correct predictions are displayed below in Table 4.5. For the first model,
64.1% of the respondents were successfully classified. This is a drop in successful predictions
compared to the greenhouse gas analyses. The model does a better job successfully predicting
responses for individuals who agreed with the proposition (81.6%) rather than for those who
disagreed (39.4%). Compared to the greenhouse effect (specificity below 20%), the models here
better predict non-agreement.
Table 4.5: Percent correctly predicted for each model
Model 1
Model 2
Percent correctly predicted
64.1%
64.9%
Sensitivity
81.6%
78.0%
Specificity
39.4%
46.3%

Model 3
66.8%
77.6%
51.4%

As before, the second model adds an interactive effect between science knowledge and
political ideology to assess the fifth hypothesis. Support is once again found on multiple fronts.
First, the interactive components as well as the interactive term are all significant. High-knowledge
moderates are more likely to agree with climatologists (science knowledge coefficient of 0.19).
Meanwhile, high-knowledge conservatives (the interaction term; coefficient of -0.07) are less
likely than high-knowledge moderates and liberals to agree with the proposition. As before, there
is a change in the direction of the relations between ideology and the response, suggesting lowknowledge conservatives are more likely than low-knowledge moderates and liberals to accept the
proposition (ideology coefficient of 0.24). The overall effect of ideology remains negative in the
second model as well.

89

Including the interaction term also significantly improves the overall fit of the model. A
Wald test suggests the interactive term does significantly improve the fit of the model (p<.001)
while a likelihood ratio test also supports this conclusion. The value of the test statistic is 33, with
one degree of freedom – suggesting the improvement is significant (p<0.001) utilizing a chisquared distribution matrix. The substantive details of this interaction are reported below, but for
now, there is continued support for this more nuanced understanding of the relationship between
science knowledge and political ideology.
In terms of the model’s ability to correctly predict responses, there is little overall
improvement between the first and the second model (see Table 4.5). The overall percent correctly
predicted remains roughly the same at 64.9% - an improvement of only 0.8%. The sensitivity and
specificity statistics suggest the interaction term improves the predictive capabilities for those who
disagree with the proposition. For those answering false, the percent correctly predicted improves
by nearly 7% (from 39.4% to 46.3%). There is a slight trade-off, however, as the ability to predict
agreement with climatologists drops by 3.6%. As before, these differences in Table 4.5 suggest an
additional explanatory concept is needed to help better understand why some respondents reject
the proposition in question.
The third model adds the same socio-demographic controls noted earlier. The results
suggest no change to the interpretations of the core hypotheses offered for the second model. Of
the socio-demographic controls themselves, gender is perhaps of the most interest. The
relationship is negative and significant at the 95% confidence level. This relationship suggests that
the knowledge gap is not entirely responsible for gender differences when it comes to this
proposition about fossil fuels. After controlling for scientific knowledge and the other educational
measures, there is still something unique about females and their response to the proposition on

90

the survey. The remaining socio-demographic controls are in the expected direction but do not
achieve statistical significance.
As a block, however, there is mixed support as to whether Model 3 is a better fit to the
data. A Wald test suggests no additional explanatory is offered by the inclusion of these additional
variables as a group. However, the likelihood ratio test suggests the block of socio-demographic
controls does improve the overall model fit, with a test statistic of 14.4 with six degrees of freedom
– significant at the 95% confidence level. This difference is likely due to the minimal explanatory
power offered by the socio-demographic measures. Only gender was significant. The likelihood
ratio test indicated the gender relationship was powerful enough to improve the fit of the entire
block of demographic controls while the Wald test offered no such indication. An additional Wald
test considered gender alone and finds its inclusion significantly improved the fit of the model
(p<0.05).
Compared with Models 1 to 2, the third model does slightly improve the overall predictive
capability of the model by 2%. This gain is among those who answered false to the proposition. It
appears that adding gender to the model improves the predictive capability of false responses by
5% compared to the second model. All together, the third model offers nearly a fifteen-point
improvement in predictive capability of false responses by including the interactive term and other
socio-demographic controls (notably gender). The discussion now turns to the substantive effects
of these identified relationships.
Substantive Effects
With this second analysis, there was once again support for the science comprehension
thesis. As before, coefficients from Model 3 are utilized to explore the size of the effects. This is
due to the improvement in model fit by including the socio-demographic control (particularly

91

gender in this case). Focusing in on scientific knowledge, the coefficients suggest that as
respondents accumulate more scientific knowledge, the likelihood that they agree society can
influence global warming via burning of fossil fuels increases. All things equal, a respondent with
average knowledge of science (a score of seven on the quiz) is predicted to agree with the claim
60.9% of the time. As knowledge accumulates to full knowledge, the probability rises to 71.8% over a ten-point improvement in predictive power. The average marginal effect of science
knowledge is 3.1%, which is slightly elevated compared to the marginal effect calculated in the
greenhouse effect analysis.
As before, the proposition that normative views of science correlate with beliefs draws
mixed support. In the context of fossil fuels, there is again support for Hypothesis 2 but not for the
third hypothesis. Individuals who agreed with basic propositions about the merits of scientific
advancement were more likely to assert beliefs similar to climatologists (Hypothesis 2). The
probabilities based on the three normative profiles (see Chapter 3) are offered in Table 4.6. The
results suggest that a move from “agree” to “strongly agree” on the faith-in-science index (holding
the skeptical of science index at “disagree” with all other values in Model 3 at their respective
means) increases the probability of agreeing with the proposition by eight points – similar the
change found when considering the greenhouse effect proposition earlier. As before, there is no
difference in probabilities based on whether skeptical views of science are held. The profiles for
weak pro-science and conflicted individuals reveal the same probability of agreement with the
proposition. The eight-point improvement in the predicted probability is roughly equal to a move
from medium to full knowledge on the science quiz score noted earlier.

92

Table 4.6: Probability agree with IPCC on fossil fuels by normative views of science
Strong pro-science:
Weak pro-science:
Conflicted:
Faith science (s.
Faith science (agree), Faith science (agree)
agree), skeptical
skeptical science
skeptical science
science (disagree)
(disagree)
(agree)
Fossil Fuels
66.7%
58.5%
58.5%
Notes: Probabilities calculated with coefficients from Model 3 of Table 4.4. All other variables
held at respective means unless specified in column header.
Again, there was support for the fifth hypothesis – the suggestion that political ideology
influences beliefs as well. As respondents grow more conservative in their ideological views, the
probability they agree with the fossil fuel proposition declines. A strong conservative is predicted
to agree with the proposition 40.7% of the time, all other variables held at their respective means.
On the opposite end of the scale, a strong liberal retains a predicted probability of 80.4%.
Moderates once again are slightly closer to liberals than to conservatives – agreeing with the
proposition 62.6% of the time. These results are consistent with expectations outlined in
Hypothesis 4 and are clearly indicative of the well-documented ideological gap within respect to
climate change beliefs. There is nearly a 40% swing as one moves across the ideological scale,
although the average marginal effect for ideology is approximately -3.5%. This movement is on
par with the average marginal effect for science knowledge, but slightly reduced compared to the
greenhouse effect analysis earlier.
While the first two models offer support for both the science comprehension and valuecentered theses, recall that the interaction term once again suggests qualifications are in order. The
second and third models both support the argument for an interactive relationship between
scientific knowledge and political ideology (Hypothesis 5). The interactive nature of the
relationship is displayed in Figures 4.4 and 4.5. The first figure shows the probability of agreement
with the proposition for three types of ideologues – strong conservatives (“5” position on the
ideology scale), strong liberals (“-5” position) and moderates (“0” position). The 95% confidence
93

interval is once again included for each group. As before, at low levels of knowledge there is little
difference between ideologues. Respondents who score roughly 40% on the science knowledge
quiz possess the same predicted probability of agreeing with climatologists (around 50%). For
these low-knowledge individuals, agreement with climatologists is approximately a 50-50
proposition. The ideological gap starts to emerge once respondents can answer at least 50% of the
questions. The probability that strong liberals agree with the proposition escalates rapidly but
levels off once a medium level of knowledge is reached. A move from average to complete
knowledge for liberals increases the probability by just over 14% (80.8% to 94.4%). A similar
shift among moderates produces a change from 62.9% to 75.1% in the predicted probability of
agreement. For both moderates and liberals, there is once again support for the science
comprehension thesis.
Figure 4.4: Probability agree with IPCC by science knowledge for liberals (-5), moderates
(0), and conservatives (5)
1
0.9
0.8
0.7
0.6
Probability
Agree with 0.5
IPCC
0.4
0.3
0.2
0.1
0
0%

10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Conservative
Liberal

94

Moderate
Confidence Intervals

Conversely, the probability of conservatives agreeing with the proposition slowly declines
as knowledge accumulates. There is a drop in the probability of agreement with the proposition
for strong conservatives. The probability of agreement drops from 40.5% to 32.9% for strong
conservatives. A high-knowledge conservative is almost 40 points away from a moderate and 60
points away from a liberal.
These relationships offer strong support for Hypothesis 5. The science comprehension
thesis finds support when only looking at two of the three ideological groups. Conservatives,
meanwhile, appear to rely on other sources of beliefs, which is strong support for the valuecentered thesis. The s-shape of the liberal line also suggests support for the value-centered thesis,
as liberals appear to more readily accept the proposition as knowledge accumulates due to their
environmental predispositions. The negative slope associated with conservatives is consistent with
observations that strong ideologues take positions opposite of their political opponents.
Figure 4.5: Average marginal effect of scientific knowledge by ideology
0.20
0.15
0.10
0.05
Average
0.00
Marginal Effect
-0.05
-0.10

-0.15
-0.20

Liberal

Conservative

95

The average marginal effect of science knowledge by ideology is displayed in Figure 4.5,
which includes the 95% confidence interval. Knowledge has a positive, significant effect for
liberals and moderates. For each question answered correctly, the probability a strong liberal
agrees with climatologists increases by 7.6%, on average. A move from average to complete
knowledge, then, would increase the probability of agreement by 22.8% on average. The increase
in probability for moderates is 4.5%, which corresponds to a 13.5% increase in the probability of
agreement with the proposition. For moderate to strong conservatives, scientific knowledge has no
significant effect on the probability of accepting the proposition. While the marginal effect is
negative, the derived coefficient is not significant. These results support the proposition that the
science comprehension thesis offers a means to understand the beliefs of liberals and moderates,
but not conservatives.
Summary: Fossil Fuels
The analysis of a second contested component of the climate change debate provides
further support for four of the five hypotheses. As before, as individuals accumulate scientific
knowledge they demonstrate an increased likelihood of accepting the proposition that fossil fuels
cause global warming – a relationship expected according to the first hypothesis. However,
qualifications are in order. Political ideology provides an alternative mechanism to understanding
climate change beliefs as expected by Hypothesis 4. Furthermore, the interactive effects expected
by Hypothesis 5 also emerged. The science comprehension thesis is one mechanism to understand
beliefs for liberals and moderates, but it cannot explain the beliefs of conservatives. Highknowledge liberals and conservatives view the proposition differently. Again, there are limits to
knowledge – limits that are dictated by the political orientations of individuals.

96

Substantively, the probability of accepting the proposition is relatively high whether
individuals possess medium or elevated levels of knowledge. Those possessing average to full
knowledge already possess a high probability of agreeing with climatologists – 60.1% and 71.8%,
respectively. Once individuals achieve a quiz score of over 40%, they have a 50-50 chance of
accepting the proposition. It is again the move from low to middle levels of knowledge that perhaps
bring public opinion into congruence with scientists. Furthermore, in this analysis the average
marginal effect of science knowledge and ideology is roughly the same, suggesting similar
explanatory power for both concepts.
Expectations about normative scientific views once again find mixed supported. The
second hypothesis finds support. Individuals who “strongly agree” with the positive statements
were 8% more likely to accept the proposition – all else equal. Substantively, this increase is just
behind a move from average to full knowledge on the science quiz. However, those skeptical of
science were not significantly less likely to accept the proposition, providing no support for the
third hypothesis. The pattern at this point is continued support for four of the five hypotheses
specified earlier in the discussion. The analysis will now shift toward solar radiation to examine
whether these patterns remain.
Analysis: Solar Radiation
The last outcome measure taps into beliefs about the primary driver of global warming,
which again is perhaps the most contested issue within the policy debate. One of the often cited
counterarguments to the IPCC is that recent warming is the result of increased solar radiation. Heat
output from the sun as well as other cosmic forces are argued to be the cause of warming. The
question on the survey asked respondents whether they accept this proposition. Specifically, they
were asked the following question: “Global warming is increasing primarily because the level of

97

direct radiation from the Sun is increasing.” Unlike the prior questions, the answer consistent with
climatologists is false. The measure was recoded so that high values are consistent with what the
IPCC might advise.
Table 4.7: Rejection of the solar radiation counterargument
Model 1
Model 2
0.24
0.26
Scientific knowledge
(6.70)
(7.12)
-0.33
-0.32
Faith in science
(-2.16)
(-2.14)
-0.38
-0.35
Skeptical of science
(-2.73)
(-2.50)
-0.04
0.22
Ideology
(-1.26)
(2.10)
Ideology X science
-0.04
knowledge
(-2.64)
Republican

-

-

Democrat

-

-

Age

-

-

Female

-

-

4-yr degree

-

-

Advanced degree

-

-

Model 1: N=1082

-0.04
(-0.05)
Log-likelihood -669

Wald Chi2

Model 2: N=1082

Log-likelihood -665

Wald Chi2

Constant

-0.27
(-0.38)

Model 3
0.24
(6.23)
-0.31
(-2.05)
-0.34
(-2.26)
0.23
(2.15)
-0.04
(-2.76)
0.46
(2.16)
0.34
(1.65)
-0.01
(-1.24)
0.15
(0.91)
0.30
(1.57)
0.63
(2.33)
-0.33
(-0.40)
64.9 PCP 65.3%
70.9 PCP 65.4%

Model 3: N=1082
Log-likelihood -653 Wald Chi2 83.1 PCP 65.9%
Notes: The first number represents the coefficient derived from a logistic regression. The
number in parenthesis indicates the z-score from the significance test. In all models, a weight
provided by Knowledge Networks was utilized in order to make inferences from the results back
to the U.S. adult population.
The three models for the analysis are presented in Table 4.7. As seen in the first model, the
degree of support for the first four hypotheses is noticeably altered. Consistent with prior analyses,

98

there is a positive, significant relationship between scientific knowledge and the outcome measure.
Individuals who have accumulated greater levels of scientific knowledge are more likely to reject
solar radiation as an explanation for global warming – an action consistent with the arguments
coming from climatologists working with the IPCC. This is the only one of the four hypotheses
supported in the first model.
Shifting focus to normative views reveals support for the third hypothesis, which had not
materialized in the prior analyses. Those more skeptical of scientists and their work fail to express
positions similar to those of climatologists working with the IPCC, which is consistent with the
third hypothesis. Meanwhile, those who view science more favorably are also less likely to assert
beliefs consistent with climatologists. This relationship reflects a reversal in the sign of the
relationship between favorable views of science and climate change beliefs when compared to the
prior analyses. The second hypothesis, then, fails to find support. This suggests that a more
sophisticated understanding of climate change deniers is warranted given what appears to be
selective rejection of a particular component of the climate change debate. The change in direction
is explored in greater detail in the discussion.
Lastly, the first model reveals no support for the fourth hypothesis. The ideological gap
does not emerge as in the prior analyses, although the direction of the relationship is consistent
with expectations. The discussion section will explore this relationship further by looking at the
differences within beliefs along the eleven-point ideological scale. Currently, conservatives are
not significantly less likely to accept the argument of climatologists compared to their moderate
and liberal counterparts.
The percent correctly predicted metric is presented in Table 4.8. For the first model, the
model correctly predicts responses for over 65% of the survey respondents. Compared to the prior

99

models, the first model performs better predicting respondents who disagree with climatologists
working with the IPCC (76%) but only correctly predicts approximately half of those agreeing
with climatologists.
Table 4.8: Percent correctly predicted for each model
Model 1
Model 2
Percent correctly predicted
65.3%
65.4%
Sensitivity
50.8%
46.0%
Specificity
76.0%
79.8%

Model 3
65.9%
48.9%
78.7%

The second model includes an interactive term to assess the proposition that ideology
moderates the relationship between knowledge and beliefs. The results suggest that while ideology
was not a significant predictor of beliefs in Model 1, there is still continued support for the more
nuanced expectations of the fifth hypothesis. Moderates with elevated levels of knowledge were
indeed more likely to reject the solar radiation counterargument (knowledge coefficient of 0.26).
High-knowledge conservatives were less likely to reject the solar radiation proposition compared
to their high-knowledge counterparts (interacted coefficient of -0.04) while low-knowledge
conservatives were more likely to accept the causal arguments from climatologists (ideology
coefficient of 0.22). These significant relationships support the expectations of the fifth hypothesis.
The interpretation of the other coefficients remains the same as in the first model.
The interactive effect does provide a slight variation, however. While high-knowledge
conservatives are less likely to reject the solar radiation counterargument compared to moderates
and liberals, the coefficients suggest all ideologues are likely to accept the proposition as
knowledge accumulates (a positive relationship). For instance, the coefficient for a highknowledge strong conservative is 0.04 (derived from adding the ideology and interaction
coefficients). Recalling the discussion earlier, the difference with this issue is that there is more
uncertainty and confusing rhetoric about what is driving global warming. This effect along with

100

the overall interaction between ideology and science knowledge is taken up further in the
substantive effects discussion.
In addition, model fit diagnostics support the proposition that this more nuanced
understanding of the relationship between scientific knowledge and ideology is appropriate. A
Wald test suggests the interaction term adds explanatory power to the second model (p<0.01). A
likelihood ratio test supports this conclusion as well. With a test statistic of nine and one degree of
freedom, the difference between the first and the second model is significant (p<0.01).
The interaction term does not improve the overall predictive capabilities of the model. The
overall percent correctly predicted is the same between Model 1 and Model 2 (approximately
65%). As with the other outcomes measures, there is a slight shift in the model’s ability to predict
respondents who answered true and those who answered false. Among those who disagreed with
the claim that solar radiation drives climate change, there is a five-point drop in the percent of
respondents correctly classified. For those agreeing with the proposition, there is a four-point
increase in the model’s predictive capabilities. This trade-off between the sensitivity and
specificity metrics is consistent with the prior analyses.
The third model includes socio-demographic controls as a means to assess whether there
is anything unique about gender, age, and education not captured by values and knowledge. The
age and the gender gaps in environmental beliefs disappear in this context after controlling for
scientific knowledge. However, the educational achievement measures suggest science knowledge
is not the lone educational metric with predictive capabilities. Recall a portion of respondents
reported obtaining college degrees yet performed poorly on the science knowledge quiz (see
Chapter 3). The coefficients for Model 3 suggest those obtaining advanced degrees are
significantly more likely to reject the solar radiation counter argument compared to those without

101

a college degree. There is also a similar but not significant relationship between those achieving
at least a four-year degree and those who have not.
Lastly, the partisan metrics are of interest. Both Republicans and Democrats were more
likely to identify the argument from climatologists than Independents. The relationship for
Democrats makes sense as partisan cues provide a basis for beliefs. The direction and significance
level (p<0.05) was unexpected for Republicans, however. The reference group is perhaps key to
understanding this relationship. If partisans are more educated and education partially explains the
response of survey respondents to the question, then the partisan metrics may be capturing some
of this variation.
Turning to the goodness of fit statistics, the block of socio-demographic controls
significantly improves the explanatory power in the third model. The Wald test suggests the
controls explain additional variance (p<0.05). The likelihood ratio test produces a similar
conclusion. The test statistic of 23.2, with six degrees of freedom, is significant (p<0.001). With
respect to the predictive performance of the model, the third model correctly predicts responses
for 65.9% of the respondents – less than a 1% improvement compared to the other two models.
There is also only slight movement in the specificity and sensitivity. Overall, then, there is no
dramatic improvement in the success rate of prediction by adding the socio-demographic controls.
Substantive Effects
To examine the substantive effects, the analysis moves to a discussion of the predicted
probabilities for the independent variables of interest in the hypotheses. Coefficients from the third
model are utilized for this analysis given the improved explanatory power achieved by adding the
socio-demographic controls. Turning attention first toward science knowledge, the analysis
suggests that individuals possessing an average knowledge of basic scientific concepts (average

102

score of seven on the quiz) agree with climatologists roughly 42.1% of the time. Individuals with
full knowledge retain a predicted probability of 58.8%. The average marginal effect suggests each
additional question answered correctly increases the probability of accepting the proposition by
4.7% - almost double the effect found in the first analysis. This change in predictive capabilities
merits further commentary later in the discussion.
Shifting focus toward normative views of science, recall the analysis found support for the
third hypothesis yet no support for the second. The predicted probabilities for the three main
profiles are presented in Table 4.9. Holding all other variables at their respective means, those
demonstrating weak pro-science attitudes were the most likely to agree with scientists at 46.9%.
This profile can provide a baseline for comparison to the other two profiles. The pro-science profile
represents a one-unit increase in the faith-in-science index compared to the weak science profile.
The analysis revealed a negative relationship between faith in science and beliefs about solar
radiation. Consequently, a one unit change from “agree” to “strongly agree” – holding skeptical
science views at “disagree” and all other values at their means – corresponds to a decrease in the
predicted probability by nearly eight points. A similar drop in probability occurs when considering
the skeptical-of-science index. Here, a one-unit increase from disagree to agree on the skepticism
scale corresponds to an eight-point drop in the probability of agreement with climatologists.
Taking the analysis a step further in order to explore this counterintuitive result, an additional
profile (strongly conflicted) was added to Table 4.9. This position represents individuals who
average strong agreement with the science attitude questions for both indices. Such individuals,
while only a small portion of the overall sample, possess a 24.8% probability of agreeing with
climatologists.

103

Table 4.9: Probability of rejecting solar radiation counterargument by normative views of
science
Strong proWeak pro-science:
Conflicted:
Strongly
science:
Faith science
Faith science
conflicted: S.
Faith science (s.
(agree),
(agree),
agree on both
agree), skeptical
skeptical science
skeptical science
indices
science (disagree)
(disagree)
(agree)
Solar
39.3%
46.9%
38.7%
24.8%
Radiation
Notes: Probabilities calculated with coefficients from Model 3 of Table 4.7. All other variables
held at respective means unless noted in column heading.
The fourth hypothesis, again, finds no support in the models although there is slight
movement in the predicted probabilities. Holding all other measures at their respective means, a
strong conservative has a 37.7% probability of agreeing with climatologists on the issue. A
moderate possess at 42.3% probability, and a strong liberal retains a 47.0% probability. All
ideologues, then, are within a ten-point range of each other in respect to their predicted
probabilities. The average marginal effect is -1%, which is the weakest effect for ideology across
the three outcome measures.
A key difference in the solar radiation analysis is the interactive effect. Figure 4.6 below
illustrates the change in predicted probabilities for the previously used profiles of strong liberals,
moderates, and conservatives. The figure reveals similar conclusions to the prior analyses with
respect to liberals and moderates. The probability of both groups rejecting the solar radiation
counterargument increases dramatically as knowledge accumulates. As noted earlier, the
coefficients from Model 3 suggest high-knowledge conservatives are also more likely to reject the
solar radiation counterargument, although compared to moderates and liberals the probability
increases at a much slower rate as knowledge accumulates. The relationships suggest there are still
indeed differences between high-knowledge ideologues: high-knowledge conservatives think
differently about science compared to their colleagues. This is consistent with the fifth hypothesis

104

despite the positive slope calculated for strong conservatives in the figure. The fifth hypothesis
speculated only that differences should emerge between ideological groups, but how highknowledge conservatives compare to low-knowledge conservatives was not specified.
Figure 4.6: Probability agree with IPCC by science knowledge for liberals (-5), moderates
(0), and conservatives (5)
1.0
0.9
0.8
0.7
0.6
Probability
Agree with 0.5
IPCC
0.4
0.3
0.2

0.1
0.0
0%

10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Conservative
Liberal

Moderate
Confidence Intervals

It is important to emphasize, however, that for strong conservatives knowledge has no
significant effect. Figure 4.7 displays the marginal effects of knowledge by ideology. For strong
liberals (ideology=-5), the average change in probability for each accumulated level of knowledge
is 10.8%. A move from average to full knowledge, then, would on average increase the probability
of rejecting solar radiation by 32.4%. For moderates (ideology=0), there is an average change of
5.9% in the predictability for each quiz score answered correctly, corresponding to a 17.7% change
in predicted probability on average. The average marginal effect remains positive for moderate
conservatives (ideology=3) while knowledge has no significant effect for strong conservatives

105

(ideology=5). This is taken as support for the fifth hypothesis and suggests a notable, significant
relationship between science knowledge and beliefs.
Figure 4.7: Average marginal effect of scientific knowledge by ideology
0.20
0.15
0.10
0.05

Average
0.00
Marginal Effect
-0.05
-0.10
-0.15
-0.20

Liberal

Conservative

Summary: Solar Radiation
This last analysis further suggests support for the science comprehension thesis. The first
hypothesis is again supported by the analysis as high-knowledge individuals are more like to share
the views of climatologists even when considering the most complex, divisive issue in the climate
change debate tested here. The improvement in prediction capabilities with a move from average
to full knowledge was the highest out of all three outcome measures. Meanwhile, ideology was
not a significant predictor of beliefs as proposed in Hypothesis 4. The interactive effect expected
by Hypothesis 5, however, did emerge. High-knowledge liberals, moderates and conservatives all
accepted the proposition at different rates. The unexpected result in the third analysis was that
high-knowledge conservatives were more likely to accept the proposition compared to lowknowledge conservatives, although the relationship is not significant. The strength of not only

106

scientific knowledge but also the other education measures suggest that educational metrics are a
powerful predictor of beliefs when the issue is highly contested, as is the case here.
For the first time, a significant negative relationship emerged between those skeptical of
science and a specific climate change belief. A move from the weak science to conflicted profile
produced an 8% change in the probability of agreement with climatologists. This change is on par
with a move from average to full knowledge as well as a move from the weak to strong pro-science
profile in prior analyses. In this context, the magnitude is roughly half that of the approximate
seventeen-point movement caused by a similar move. This disparity reinforces the importance of
knowledge in guiding beliefs, but also suggests scientists need to tread carefully and not give the
public additional reasons to doubt their work (as was the case with “Climategate”).
The lack of an ideological relationship is somewhat curious, as is the sign reversal in the
faith-in-science index. Both findings are taken up in more detail in the discussion since the
identified relationships do not match the pattern of the prior analyses. Final judgment on the second
and fourth hypotheses, then, are reserved for the discussion in the following section as there is
some suggestion that both still might have support.
Discussion
The regression analyses revealed strong support for four of the five hypotheses. The results are
quickly summarized in Table 4.10. The following section discusses the key observations for the
three analyses above, as well as notes the implications of the results.

107

Table 4.10: Summary of support for hypotheses
Greenhouse
effect

Hypothesis
Hypothesis 1: Individuals with higher levels of
scientific knowledge accept the causal arguments
from climatologists at higher rates than their lowknowledge counterparts, holding all else constant.
Hypothesis 2: Individuals perceiving benefits from
scientific advancement are more likely to accept
causal arguments from climatologists, holding all
else constant.
.
Hypothesis 3: Individuals skeptical of scientists and
their work are less likely to accept the causal
arguments from climatologists, holding all else
constant
Hypothesis 4: Liberals (conservatives) are more
likely to agree (disagree) with causal arguments
consistent with the IPCC, holding all else constant.
Hypothesis 5: Political ideology conditions the
relationship between scientific knowledge and
climate change beliefs, holding all else constant.

Fossil fuels

Solar
radiation

Yes

Yes

Yes

Yes

Yes

No

No

No

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Knowledge by Issue
Across all the three issues, there is solid support for the science comprehension thesis. In
all cases, individuals with higher levels of scientific knowledge were more likely to offer answers
that are consistent with the causal arguments offered by climatologists working with the IPCC.
There are some differences across issues, however. The predicted probabilities by science
knowledge are offered in Figure 4.8, again using coefficients from Model 3 of the prior analyses
and holding all other values at their means. The substantive effect of knowledge was perhaps the
most notable for the first two issues. In both cases, the beliefs of individuals mastering the science
knowledge quiz could be predicted with a high degree of power – 85.8% (greenhouse effect) and

108

71.8% (fossil fuels). In each case, a move from average knowledge to full knowledge increased
the predicted probability by between 9 and 11%.
Figure 4.8: Probability of agreement with IPCC by science knowledge for all three outcome
measures
1
0.9
0.8
0.7
Probability 0.6
agree with 0.5
IPCC
0.4
0.3
0.2
0.1
0
Low Knowledge
Greenhouse

High Knowledge
Fossil Fuels

Solar Radiation

It is this last issue, solar radiation, which is perhaps of the most interest to proponents of
the science comprehension thesis. On one hand, scientific knowledge does not do as well of a job
explaining the variance within public opinion. The overall predicted probabilities for lowknowledge individuals do not standout. This makes sense given that the political debate over
climate change often includes discussions of whether warming is the result of anthropogenic or
natural causes. The outcome measure in the survey taps into this debate, and it seems intuitive that
the relationship between scientific knowledge and beliefs would be reduced as political claims are
escalated. Non-scientific actors also attempt to influence opinions and offer alternative causal
arguments. The result is the growth of polarization within public opinion since some highknowledge individuals are motivated to reject scientific information that threatens their work.

109

On the other hand, solar radiation is also the issue where a move from average to full
knowledge reveals considerable improvement in predictive power – a seventeen-point increase in
overall predictive capabilities. Of the three issues, solar radiation is the issue where a move from
average to full knowledge produces the most value in terms of changing the predicted probabilities
of agreement with climatologists. This jump speaks to the value advocates of science education
expect to realize by educating the public on basics scientific concepts. In this case, the scientifically
knowledgeable appear able to sort through the noise and identify the core scientific arguments in
respect to the causes of climates change.
This is robust support for the first hypothesis. Science benefits from a perception that it is
an objective process that can be used to justify knowledge. Once individuals accept the scientific
process as a means to justify their beliefs, they begin to develop a pro-science framework that
guides their approach to understanding and interpreting policy debates. At a general level, this
framework seems to lead the population to accept the causal arguments offered by climatologists.
Claims by advocates that knowledge does lead individuals to accept basic scientific findings do
find support in this analysis. Specifically, they come to understand climate change in a manner
similar to climatologists.
Normative Views
Possessing a foundational understanding of science, however, is only one component of
understanding the climate change debate. As noted earlier, there has been some emphasis in the
literature that science is a social construct and normative views of science require consideration.
This analysis found mixed support for generalized science attitudes as a predictor of climate
change beliefs. The third hypothesis, which focused on individuals with a skeptical view of
science, did not find support in two of the three models. Only in the most contested issue – solar

110

radiation – did a relationship with beliefs emerge. The overall effect is rather limited, however.
There was only a net decrease of 8% in the probability of agreeing with climatologists as skeptical
views of science increase (a move from the “weak science” to “conflicted” profile). Given that this
move between these two points represents a sizeable majority of survey respondents, the overall
influence of cautious science views in the models seems limited.
More noteworthy is the second hypothesis, which focused on the level of faith respondents
place in science. The expected relationship materialized in the first two analyses, although the
overall effect was limited. The change in probability of agreeing with climatologists did not exceed
double-digits for individuals moving from the “weak science” to “strong science” profiles, which
again are among the three most common positions on the normative scale matrix (see Table 3.3).
Across all issues, the changes that did occur as respondents moved between the three profiles was
comparable to changes in science knowledge from medium to full knowledge on the scale (with
the exception being solar radiation).
Also of interest is the result of the third analysis, where a significant negative relationship
emerged between favorable science views and the proposition about solar radiation. The switch in
sign suggests a more nuanced understanding of climate change deniers might be in order.
The first point of inquiry is to assess whether a bivariate relationship exists between the
faith in science index and the three outcome measures (presented in Table 4.10). In respect to solar
radiation (last two columns), there is a slow, minor increase in the proportion of respondents who
agree with the IPCC as positive feelings of science increase. This increase is minimal compared
to the first two outcome measures, where a move from “agree” to “strongly agree” corresponds to
an increase of 9.7% (greenhouse effect) and 7.5% (fossil fuels) in the proportion of respondents

111

agreeing with climatologists. With this bivariate approach, there is no indication that a sign
reversal should have been expected in the analysis.
Table 4.11: Distribution of faith-in-science scores by outcome measure
Faith-inscience
Greenhouse effect
Fossil fuels
Solar radiation
score
Agree
Agree
Agree
Disagree
Disagree
Disagree
IPCC
IPCC
IPCC
Strongly
36.4%
63.6%
27.3%
72.7%
27.3%
72.7%
Disagree
(4)
(7)
(3)
(8)
(3)
(8)
(N=11)
Disagree
(N=72)

61.1%
(44)

38.9%
(44)

30.6%
(22)

69.4%
(50)

34.7%
(25)

65.3%
(47)

Agree
(N=855)

66.2%
(289)

33.8%
(566)

54.4%
(465)

45.6%
(390)

39.1%
(334)

60.9%
(521)

Strongly
Agree
(N=415)

75.9%
(315)

24.1%
(100)

61.9%
(257)

38.1%
(158)

40.5%
(168)

59.5%
(247)

To investigate further, a series of bivariate logistic regressions were calculated (not shown).
A bivariate regression suggests a positive, insignificant relationship exists between the faith-inscience index and solar radiation beliefs. It is not until science knowledge is added to the model
that the direction of the relationship becomes negative. With this in mind, a cross-tabulation was
performed between the three measures: the faith in science index, responses to the solar radiation
question, and performance on the scientific knowledge quiz. The results are presented in Table
4.11. For the purposes of the table, respondents with an above average score on the science quiz
(70% or above) were coded as possessing high levels of scientific knowledge while those below
the threshold are placed in the low-knowledge category. The results offer a different view from
those in the prior table. After controlling for high- and low-knowledge individuals, the negative
relationship starts to emerge. The trend is most distinct among low-knowledge individuals, where
there is a clear drop in agreement with climatologists among those ranking higher on the index.
112

For high-knowledge individuals, the relationship is a bit more mixed. Moves from “strongly
disagree” to “disagree” to “agree” on the index correspond to an increase in agreement with
climatologists – consistent with the expectations of the second hypothesis. However, the move
from “agree” to “strongly agree” results in a drop in the level of agreement with climatologists.
From the table, it appears the sign reversal is best understood as individuals with low-knowledge
yet high levels of faith in science accepting the solar radiation counterargument in sufficient
numbers to create the observed sign change.
Table 4.12: Faith-in-science by solar radiation response and level of knowledge
Faith-inAgree IPCC,
Disagree
High Agree IPCC,
Disagree
Low
science
High
IPCC, High Know.
Low
IPCC, Low
Know.
score
Knowledge
Knowledge
N
Knowledge
Knowledge
N
Strongly
25.0%
75.0%
33.3%
66.7%
Disagree
8
3
(2)
(6)
(1)
(2)
(N=11)
Disagree
40.0%
60.0%
31.0%
69.0%
30
42
(N=72)
(12)
(18)
(13)
(29)
Agree
52.2%
47.8%
24.6%
75.4%
448
407
(N=855)
(234)
(214)
(100)
(307)
Strongly
50.6%
49.4%
22.3%
77.7%
Agree
267
148
(135)
(132)
(33)
(115)
(N=415)
Notes: Percentages reflect the distribution of high knowledge and low knowledge responses by
agreement with the IPCC and the faith in science index. The number in parenthesis indicates the
sample size.
Ideology
Scientists are not alone, however, in offering causal arguments to the public. Political
ideology as an alternative framework for understanding climate change found support in two of
three analyses. This was especially true as to whether individuals understood the greenhouse effect,
where liberals were more likely to recognize the outcomes of the scientific theory than both
moderates and conservatives. For conservatives and others who failed to acknowledge the
outcomes associated with the greenhouse effect theory, it is unclear whether that negative

113

correlation results from survey respondents not being familiar with the concept or whether it
represents an outright rejection the theory. Some respondents could have approached the question
as if it were asking whether certain human activities might induce increased temperatures,
doubting whether greenhouse gases actually trap heat. Others might simply not have been familiar
with the concept. Due to question wording, it is tempting to think the former interpretation is
appropriate. The wording hints at a logical answer, as the term greenhouse implies a process of
trapping heat. As such, it would seemingly be plausible for respondents to relate the outcome
(rising temperatures) to the term greenhouse effect. Given this deductive process to answer the
question, it possible answering “false” might indicate a rejection of the proposition that humankind
can heat the Earth.
A similar question is raised about the fossil fuel questions. Did respondents interpret the
question strictly as asking whether the burning of fossil fuels might cause global warming, or
whether global warming is certainly occurring because of the burning of fossil fuels? Survey
respondents with a limited amount of time might not have accurately interpreted the question.
There are no supplementary questions, however, to help piece together an explanation why
respondents answered as they did. However the questions were interpreted, ideological differences
did emerge, suggesting the political framework individuals are socialized toward as youths
influence interpretations of policy debates later in life as well.
One possible explanation for why conservatives reject arguments from the IPCC about
climate change is the potential consequences of acknowledging its occurrence. That is, admitting
to the problem opens up the possibility of eventual environmental regulation. If both political
parties agree that anthropogenic climate change is a possibility, enough public support might
develop to push for regulations. Government regulations, of course, are seen as undesirable to

114

conservatives. Thus, denying climate change may be a strategic decision. In a supplementary
analysis, a measure of regulatory attitudes was developed, assessing how the survey respondents
feel about tax, regulations, and healthcare regulations. Substituting this anti-regulatory index into
the models as a replacement for ideology produced similar regression results. This anti-regulation
index also correlates highly with ideology, suggesting the motivation for denying climate change
might be out of a desire to avoid regulation.
Ideology & Solar Radiation
The lack of a significant relationship between ideology and solar radiation beliefs warrants
further discussion. The first analysis was relatively uncontroversial among climatologists, at least
compared to the other two outcome measures. Again, the first question was simply asking about
the outcome associated with a scientific concept. Still, despite what seems like a relatively
apolitical question, an ideological dimension was observed. The strength of the relationship
between ideology and beliefs, however, diminishes as the outcome measure becomes more
controversial. The average marginal effect of ideology in the first model was -0.04, decreasing
slightly 0.005 points when considering the proposition on fossil fuels. This suggests the
relationship between ideology and acceptance of arguments from climatologists is consistent. As
noted earlier, the relationship is insignificant in the context of solar radiation, with the marginal
effect declining to -0.01.
To explore the relationship further, a cross-tabulation is presented as Table 4.12. The table
suggests there are considerable fluctuations in the rate at which ideologues identify the arguments
of climatologists. The least likely to accept the arguments are moderates and moderate
conservatives. Conversely, moderate liberals are more likely to accept the proposition. When it
comes to strong liberals and strong conservatives, both groups are likely to express similar

115

opinions. This seemingly random distribution helps explain why there was no relationship between
ideology and solar radiation beliefs.
Table 4.13: Cross-tabulation of whether respondents reject solar radiation by ideology
Ideology
Fail to reject
Reject
Total
Liberal (-5)
53.1%
46.9%
51
-4
59.0%
41.0%
25
-3
45.2%
54.8%
56
-2
46.2%
53.8%
86
-1
53.8%
46.2%
82
0
64.9%
35.1%
325
1
58.7%
41.3%
110
2
55.7%
44.3%
145
3
62.1%
37.9%
146
4
62.3%
37.7%
58
Conservative (5)
56.3%
43.7%
48
Notes: Percentages reflect weighted distributions.
This lack of a systematic pattern may occur because of the nature of the question. Again,
the cause of rising temperatures is arguably the most contested claim in the policy debate.
Individuals are embracing the counterargument perhaps because of their predispositions and own
personal preferences. Others, however, may be rejecting the proposition because it is genuinely a
confusing topic. The solar radiation argument has scientific roots, and the causal mechanism
appears valid. Climatologists working with the IPCC argue they can reject this alternative
hypothesis as an explanation for recent warming over the last few decades, but this does not
prevent the counterargument from re-occurring, or a minority of climatologists insisting
temperature change is largely a function of solar radiation (see Shaviv 2006).
Ideology & Interaction Effects
Whether from the ideological lens or other mechanisms, subgroups accept the arguments
of climatologists at different rates. Of note is the within-group gap observed when the interactive
effects are considered. It is notable that ideological differences emerge even when considering a

116

seemingly apolitical concept such as the greenhouse effect. Even basic scientific concepts can
invoke ideological differences. Across all outcome measures, there is a gap in the beliefs between
the “low knowledge” and “high knowledge” populations within each ideological groups. As noted
earlier, however, that gap is significantly larger for liberals who “benefit” the most from
accumulating knowledge. Conservatives, on the other hand, are less influenced by the foundational
scientific knowledge that they have accumulated. In fact, in two out of three cases strong
conservatives with elevated levels of knowledge were more likely to disagree with the proposition
than their low-knowledge counterparts. This is consistent with the political polarization literature,
which suggests partisans will seek out positions or beliefs that are inconsistent with their political
counterparts (Nicholson 2012). It is also consistent with the idea of motivated reasoning, which
was described in Chapter 2. Conservatives simply have no reason to accept the arguments from
climatologists, as acknowledging their positions create an opportunity for future environmental
regulations.
Support for the interaction hypothesis is found on multiple fronts, including the
significance of the measures (linear log-odds coefficients), the probabilistic component (calculated
by the Delta method in previous figures), and the overall improvement in the model fit once the
interaction term is included. One finding in the interaction models is that low-knowledge
conservatives tend to think differently about the outcome measures compared to moderates and
liberals. However, this interpretation should be viewed cautiously due to few respondents
populating the low end of the knowledge scale. At quiz scores of 0% and 10%, there are only a
small number of respondents. It is not until quiz scores in the 30%-40% range that the frequencies
of respondents increases. This is the area where there was little to no differences in the predicted
probability of accepting the arguments of climatologists. Simplified cross-tabulations like below

117

suggest low-knowledge (score of 0 to 6 on the science quiz) conservatives are not more likely to
identify the arguments asserted by climatologists, contrary to what the regression coefficients
suggest. Attention is drawn to this relationship only to caution readers that the limited observations
on the low end of the knowledge scale may be an anomaly.
Table 4.14: Distribution of faith in science scores by outcome measure
Greenhouse effect
Fossil fuels
Solar radiation
Agree
Agree
Agree
Disagree
Disagree
Disagree
IPCC
IPCC
IPCC
Low-knowledge
48.1%
51.9%
25.9%
74.1%
22.2%
77.8%
conservative
(13)
(14)
(7)
(20)
(6)
(21)
(N=27)
Low-knowledge
64.0%
36.0%
54.0%
46.0%
27.3%
72.7%
moderate
(103)
(58)
(87)
(74)
(44)
(117)
(N=161)
Low-knowledge
60.0%
33.8%
33.8%
60.0%
39.1%
60.9%
liberal (N=15)
(9)
(6)
(6)
(9)
(4)
(11)
Overall, the interactive effects suggest values and predispositions play a strong role in
guiding the beliefs of ideologues. However, the fact that a positive, strong relationship occurred
between scientific knowledge and climate change beliefs despite the role of values suggests
knowledge is also a significant driver. In all three outcome measures, a sufficient number of highknowledge moderates agreed with high-knowledge liberals and conservatives maintain this
positive relationship.
These findings are consistent with the idea that science is partly an information tool utilized
to strengthen one’s own ideological argument. For liberals in this context, science is a tool used to
justify rallying around an issue that aligns with prior predispositions about the necessity of
environmental protection. Knowing nothing else, they are less likely than moderates to accept the
causal arguments of climatologists, but once they have a basic foundational understanding of
science, they rally behind those causal stories in greater numbers. When science coincides with

118

values, the information provides a strong reason to believe. When information is at odds with
predispositions, it is more tempting to call information into question or find a reason to disagree.
Summary
The analyses presented here supports the proposition that individuals possessing a
foundational understanding of science are more likely to accept the causal arguments offered by
climatologists. That is, they are more likely to understand and interpret the policy debate in a
manner consistent with what the majority of scientists would advise. The strength of the
relationship is best exemplified by the solar radiation analysis, where there was a notable increase
in the probability of respondents agreeing with climatologists as they mastered the questions on
the science knowledge quiz. There was a substantively meaningful increase in the probability of
agreement with climatologists as one moved from the medium position on the quiz to full
knowledge. A similar move did not produce as powerful of a change in prediction rates in the first
two analyses, which is perhaps in part due to survey respondents generally answering the questions
in a manner consistent with climatologists.
While there is support for the science comprehension thesis, there is also support for the
other hypotheses. In the fossil fuel and the greenhouse effect analyses, individuals with favorable
views of science were more likely to accept the propositions of scientists. Again, the substantive
effect of an increase on the faith-in-science index corresponds to change from average to full
knowledge on the science knowledge scale. This supports the second hypothesis. There was no
support for the second hypothesis, however, when looking at solar radiation. In fact, individuals
with elevated levels of faith in science were significantly less likely to accept the arguments of
climatologists on the cause of global warming. Supplementary analyses suggest this change in sign

119

is likely the result of predominantly low-knowledge individuals disagreeing with climatologists at
greater rates. A more detailed explanation for this relationship is offered in Appendix B.
The first and second analyses failed to support the third hypothesis. However, support for
the proposition that those skeptical of science will disagree with climatologists was found in the
context of solar radiation. This relationship only emerged from the most controversial question,
where a change in the skeptical-of-science index roughly corresponded with a similar change in
the faith-in-science index.
Support was also found for the value-centered thesis in two of the three analyses. With
respect to the greenhouse effect and fossil fuel discussions, ideology could be considered the
dominant predictor given the size of the average marginal effect. The relationship did not
materialize, however, with respect to solar radiation – probably due to the contested nature of the
question and the plausibility of the counterargument. Furthermore, supplementary analyses that
operationalized ideology based on policy preferences rather than on the self-reported scale
produced similar results, suggesting preferences for limited regulation are one reason why
conservatives were on occasion more likely to disagree with the arguments of climatologists.
The fifth hypothesis tested for interactive effects, with results supporting the argument that
the science comprehension thesis requires qualifications. On all three issues, high-knowledge
liberals were more likely to accept the propositions from climatologists than were high-knowledge
moderates, who in turn were more likely to accept the claims than high-knowledge conservatives.
However, liberals were more likely than moderates to accept the proposition as well. This result
can be taken two ways. While the observations support the science comprehension thesis for both
moderates and liberals, the rapid rate of acceptance by liberals also supports the value-centered
thesis. Recall the observed s-shaped pattern for the predicted probabilities of liberals. The evidence

120

that liberals rally around the scientific information on climate change while conservatives reject
the same information is consistent with the value-centered approach. The positive relationship for
high-knowledge moderates, however, cannot be dismissed as it offers clear support for the science
comprehension thesis.
The socio-demographic measures correlated with beliefs only occasionally in the three
analyses. There were no noteworthy relationships in the greenhouse effect analysis while the fossil
fuel analysis revealed only a statistically significant relationship between gender and acceptance
of the proposition. In both cases, the socio-demographic controls did little to improve the overall
fit of the model. In these two areas that feature a high-level of certainty among climatologists, the
dominant predictors were ideology and science knowledge. Conversely, the demographic controls
were more noteworthy in the solar radiation analysis. In this area featuring reduced certainty
among climatologists, there was more “noise” in that other relationships emerged – particularly
with the partisan and education measures. In this situation, the measures did significantly improve
the overall fit of the model.
These five hypotheses offered an initial test to the two-step process of understanding
problem recognition. Appendix A offers are more detailed look at why high-knowledge individuals
stood out in the analyses. The science comprehension thesis suggests high-knowledge individuals
are more likely to engage in the debate as a means to vet information and make informed decisions,
which is supported by additional analyses in the appendix. In addition, Appendix B examines the
interrelationships between the three outcome measures discussed here. The fifth chapter goes
beyond the first stage of the process by extending the analysis to climate change concern,
examining the merits of the causal argument enunciated in Chapter 2.

121

CHAPTER 5: A TWO-STEP CONNECTION TO CLIMATE CHANGE CONCERN
The preceding chapter explored the first step of the proposed two-step relationship between
scientific knowledge and climate change concern. In this next section, path analysis is utilized to
further evaluate the first five hypotheses, as well as consider two additional hypotheses related to
the two-step process. The argument has been that possessing higher levels of scientific knowledge
provides the skill set and capacity to make sense of the causal arguments offered as part of the
climate change debate. High-knowledge individuals demonstrate an increased likelihood of
engaging in behavioral activities that lead to the acquisition of information about climate change
(see Appendix A). This activity can be either passive (talk with acquaintances) or active (seek out
information from books and websites). With this information, individuals can then make informed
decisions that may or may not fully align with the causal arguments from climatologists working
with the IPCC. The analyses presented in Chapter 4 support this conceptualization (see also
Appendix A).
The next step looks at whether scientific knowledge and climate change knowledge
correlate, either directly or indirectly, with climate change concern. Understanding climate change
in a manner similar to climatologists (i.e., climate change knowledge) is expected to be associated
with elevated levels of concern (Hypothesis 6). Since scientific knowledge correlates with the
acceptance of causal arguments from climatologists, an indirect relationship between scientific
knowledge and concern is expected. That is, basic scientific knowledge correlates with higher
levels of concern through the prior effect of scientific knowledge on climate change knowledge.
As indirect relationship would support Hypothesis 7, which suggests knowledgeable individuals
are, in fact, part of a larger group demonstrating higher levels of concern about climate change.
The role of basic scientific knowledge in this conceptualization suggests that the relationship

122

between science knowledge and concern is best understood as an indirect effect (see Figure 5.1).
Scientific knowledge helps individuals understand arguments from climatologists. If the
individuals reach this stage, then they are then more likely to be concerned about the phenomenon.
However, basic science knowledge is not a necessary component to understanding climate
change or demonstrating elevated levels of concern. Messages from ideological leaders also
reinforce predispositions and provides a heuristic for ideologues to navigate the climate change
debate. As such, belief systems are likely to have effects on both how climate change is understood
but also on whether climate change constitutes a problem for society. Ideology and science
knowledge, as well as the faith-in-science and skeptical-of-science indices, all correlate with
climate change knowledge to varying degrees. However, it is expected that normative views
correlate with concern as well, given that their belief systems and values lead them to conclusions
regardless of whether they are aware of the available scientific information.
Path analysis provides a method to assess the causal relationship specified in this two-step
approach (Kline 2005). The technique compares structural relationships between multiple
exogenous and endogenous measures by calculating and comparing the observed covariance for
the relationships specified by the researcher. Fit statistics can then be utilized to assess which
structural design best fits the data. The emphasis of this chapter is to assess the merits of the twostep process. The intent is not to compare competing alternative model specifications, but rather
to assess the merits of the two-step process emphasized in Chapter 2.
Figure 5.1 presents a fully developed path model consistent with the two-step process.18
The diagram captures the essence of the process enunciated earlier The diagram accounts for the
18

Note that Figure 5.1, as written, assumes that all measures except climate change beliefs and
climate change concern are exogenous. Not including the socio-demographic controls, the model
123

heuristics and predispositions that individuals inherently rely upon when developing opinions. It
also shows the expected indirect relationship between science knowledge and problem recognition.
Thus, elements of both the science comprehension thesis and value-centered thesis are embedded
within Figure 5.1.
Figure 5.1: Tested two-stage model

Science
knowledge
Knowledge
* ideology
Ideology

Problem
understanding

Problem
recognition

Faith in
science
Skeptical
of science
Sociodemographics
The analysis in this chapter proceeds as follows. The first stage of the model is briefly
retested with path analysis. The independent variables from Model 3 of Chapter 4, which included
the socio-demographic controls and the interaction between science knowledge and ideology, is

features 28 parameters: 10 covariances, 7 variances and 11 direct relationships. Because the model
is recursive and all relationships are drawn, the model is considered just-identified. The model fits
the data perfectly, which is not necessarily desirable for two reasons. First, the model may be
fitting sampling error from the survey, which could lead to imprecise estimates. Second, there is
an inability to compare alternative model specifications (Kline 2005). The “just-identified” model
is utilized in this analysis, however, simply for hypotheses testing.

124

utilized due to the improved fit offered by the measures (see Chapter 4 discussion). This initial
section notes whether there are any changes in support for the first five hypotheses when analyzing
the covariance between the measures. This analysis functions as a further test of Hypotheses 1
through 5.
The second analysis expands on the model and tests the entirety of Figure 5.1, testing
Hypotheses 6 and 7. Because of the prior analyses, it is assumed at this point that the interactive
effect is a desirable aspect in modeling climate change beliefs and scientific knowledge. It is also
assumed that the socio-demographic controls account for variations within the outcome measures
not captured by the orientations identified for this analysis. Therefore, the second analysis includes
the interaction and socio-demographic controls utilized in Chapter 4. In this setup, concern about
climate change is used to operationalize problem recognition. Since the reliability analysis
suggested the three outcome measures featured in Chapter 4 capture distinct views about climate
change, problem understanding is operationalized with how respondents answered questions about
fossil fuels and solar radiation. The greenhouse gas measure is omitted in this discussion in order
to streamline the analysis and focus on these two more contested views about climate change.19
In the forthcoming analyses, the methodological techniques will utilize maximum
likelihood estimation. The tables will present the direct, indirect, and total effects of each variable
on the endogenous measure(s). Furthermore, all analyses utilize a weighting mechanism that
weights the third wave of the Science News Survey back to the United States adult population.
Utilizing this weighting mechanism will help control for any bias that might result from the loss
19

Recall from Appendix B that most respondents did correctly answer the greenhouse effect
question. There is also a high degree of certainty among scientists that greenhouse gases do warm
the planet. On the other hand, the fossil fuels and solar radiation question are more divisive and
offer a stronger test of the science comprehension thesis.
125

of survey respondents between the first and third wave of the survey. The models are “just
identified” in order to test the hypotheses, though this means there are limited opportunities to
discuss goodness-of-fit metrics.
Re-Analysis of First Stage
The first analysis reconsiders the first step of the two-step process using path analysis
techniques. Again, this analysis is offered to note differences within the analysis when considering
the covariance between the included measures. Table 5.1 replicates the Chapter 4 analysis. The
table presents standardized coefficients, with the z-statistic presented in parenthesis. Coefficients
in bold indicate there was a change in either the sign or significance of the measures compared to
the logistic regressions presented in Chapter 4.
As before, support for both the science comprehension and value-centered theses is
observed in the analysis across all three questions. Those possessing higher levels of scientific
knowledge were more likely to identify the position of scientists, but that relationship requires
qualification: high-knowledge conservatives are less likely than high-knowledge liberals and
moderates to accept the propositions. The increase in the size of the standardized coefficient for
science knowledge in Model 3 (solar radiation) underscores the importance of knowledge in
guiding beliefs on the most contested of the three issues.
One difference is the lack of a relationship between ideology and greenhouse effect beliefs
(see Model 1). In this case, the interpretation is that low-knowledge conservatives do not possess
unique beliefs about the greenhouse effect compared to low-knowledge liberals and moderates.
The interaction effect, however, remains significant and in the expected direction.

126

Table 5.1: Total effects for the first stage of the two-step model
Greenhouse effect
Fossil fuels
Solar radiation
(Model 1)
(Model 2)
(Model 3)
Scientific
0.19
0.15
0.25
knowledge
(4.15)
(3.42)
(5.76)
0.09
0.10
-0.07
Faith-in-science
(2.18)
(2.42)
(-1.48)
0.01
0.08
-0.09
Skeptical-of-science
(0.22)
(1.62)
(-1.90)
0.26
0.28
0.20
Ideology
(2.38)
(2.76)
(1.43)
Ideology * science
-0.44
-0.48
-0.39
knowledge
(-3.49)
(-4.58)
(-3.83)
-0.03
0.10
0.01
Republican
(-0.63)
(1.98)
(0.22)
0.03
0.08
0.04
Democrat
(0.25)
(1.70)
(0.70)
-0.02
0.05
-0.01
Age
(-0.40)
(1.06)
(-0.24)
-0.13
-0.00
-0.10
Female
(-2.92)
(-0.10)
(-0.31)
0.05
0.05
0.07
4-yr degree
(1.21)
(1.20)
(1.57)
0.05
0.11
-0.01
Advanced degree
(1.37)
(2.45)
(-0.13)
0.46
-0.17
0.77
Constant
(1.12)
(-0.42)
(1.91)
Sample Size
758
758
758
Log-likelihood
-12290
-12346
-12343
2
0.12
0.13
0.12
R
Notes: The first number represents the standardized coefficient. The number in parentheses
indicates the z-score from the significance test. A weight provided by Knowledge Networks was
utilized in order to make inferences from the results back to the U.S. adult population. Bolded
coefficients indicate there was either a change in the sign or significance of the measures
compared to the logistic regressions presented in Chapter 4.
The relationship between normative views and climate change understanding remains for
both the greenhouse effect and fossil fuel analyses. However, differences emerge in the context of
solar radiation. There is no statistically significant relationship between normative views and solar
radiation beliefs. The direction of the relationships, however, remains the same. Those more

127

skeptical of science as well as those more likely to perceive benefits from scientific advancement
are less likely to agree with climatologists.
With respect to the socio-demographic controls, there are some changes in the sign of the
measures, which is expected given the low-significance for many of the measures in Chapter 4.
The most noticeable measure to change signs was gender, which moved from a positive to a
negative relationship when considering propositions about the greenhouse effect and solar
radiation. The negative, statistically significant relationship between gender and fossil fuel beliefs
remained the same, however. The persistent negative sign is inconsistent with literature that
suggests females are more knowledgeable about climate change (McCright 2010). There were two
other sign reversals: Republicans switched from negative to positive in Model 1 (greenhouse
effect) and those with advanced degrees switched from a positive to a negative relationship in
Model 2 (fossil fuels). Neither relationship, however, is statistically significant here or in Chapter
4.
Two-Step Model
The next section shifts the analysis toward expanding the model with a multi-staged
approach. Table 5.2 below assesses the relationships hypothesized in Figure 5.1 in two separate
analyses – one using fossil fuels as the bridge between the exogenous measures and climate change
concern, the other using solar radiation. In both models, the direct, indirect, and total effects of the
measures on climate change concern are presented. The table suggests some common themes.

128

Table 5.2: Direct, indirect, and total effects on climate change concern
With fossil fuel response
With solar radiation response
Direct
Indirect
Total
Direct
Indirect
Total
effect
effect
effect
effect
effect
effect
Scientific
-0.00
0.02
0.02
0.01
0.01
0.02
knowledge
(-0.10)
(2.84)
(0.69)
(0.23)
(1.96)
(0.69)
Agree on
0.60
0.60
fossil fuels
(5.77)
(5.77)
Agree on
0.22
0.22
solar rad.
(2.13)
(2.13)
Faith-in0.26
0.05
0.31
0.32
-0.01
0.31
science
(2.78)
(2.18)
(3.31)
(3.36)
(-1.19)
(3.31)
Skeptical-of- 0.06
0.04
0.10
0.11
-0.02
0.10
science
(0.74)
(1.55)
(1.18)
(1.37)
(-1.51)
(1.18)
0.06
0.03
0.08
0.08
0.01
0.08
Ideology
(0.71)
(2.08)
(0.98)
(0.84)
(1.76)
(0.98)
Ideology *
-0.02
-0.01
-0.03
-0.03
-0.00
-0.03
Sci. Know.
(-2.09)
(-3.44)
(-2.66)
(-2.43)
(-1.91)
(-2.66)
-0.34
-0.02
-0.36
-0.39
0.03
-0.36
Republican
(-2.95)
(-0.68)
(-3.02)
(-3.23)
(1.45)
(-3.02)
0.39
0.05
0.44
0.43
0.01
0.44
Democrat
(0.12)
(1.58)
(3.40)
(3.36)
(0.62)
(3.40)
0.00
0.00
0.00
0.00
-0.00
0.00
Age
(1.11)
(0.94)
(1.26)
(1.29)
(-0.22)
(1.26)
0.08
-0.07
0.01
0.01
-0.00
0.01
Female
(0.87)
(-2.63)
(0.10)
(0.11)
(-0.15)
(0.10)
0.27
0.03
0.30
0.28
0.02
0.30
4-yr degree
(2.86)
(1.13)
(3.09)
(2.85)
(1.31)
(3.09)
Advanced
0.30
-0.01
0.29
0.25
-0.04
0.29
degree
(2.56)
(-0.14)
(2.29)
(2.02)
(1.63)
(2.29)
2
Fossil fuels
N=753
Log-pseudo likelihood -13311 R 0.29 (concern)
Solar radiation
N=753
Log-pseudo likelihood -13333 R2 0.24 (concern)
Notes: The first number represents the standardized coefficient. The number in parentheses
indicates z-score. A weight provided by Knowledge Networks was utilized in order to make
inferences from the results back to the U.S. adult population.
First, how one understands the problem correlates with greater levels of concern.
Respondents who agree that burning fossil fuels causes global warming were more likely to
demonstrate elevated levels of concern (standardized coefficient of 0.60). Similarly, if one agrees
with climatologists working with the IPCC about solar radiation, then they are also more likely to

129

demonstrate higher levels of concern (standardized coefficient of 0.22). Note the difference in the
size of the coefficients. While both measures are statistically significant, it appears the fossil fuel
proposition maintains are stronger relationship with climate change concern. Examining at the R2
values for each analysis suggests that utilizing the fossil fuel proposition explains 5% more of the
variance. Despite these differences, both relationships provide support for Hypothesis 6.
Understanding climate change in a context similar to climatologists correlates with higher levels
of concern. Note, however, that this is not consistent with the science comprehension thesis as the
thesis only states knowledge allows citizens to make informed decisions about whether climate
change poses a threat. It appears that obtaining knowledge about climate change is sufficient to
make individuals attuned to the potential threat of climate change.
Because of the interaction, the science knowledge measure is interpreted as moderates with
high levels of knowledge. For both models, the total effect of science knowledge for this
population is minimal and insignificant. There are, however, distinct indirect effects. Scientific
knowledge has a positive, significant indirect effect on climate change concern. The relationship
is strongest in the fossil fuels model, where concern is predicted to increase by 0.02 standard
deviations for a one standard deviation change of scientific knowledge. This relationship is
statistically significant and substantively similar compared to the other indirect effects. The
standardized coefficient is reduced by half in the solar radiation model, but it is still significant at
the 95% confidence threshold. Both results suggest science knowledge has an effect on concern,
but this relationship occurs through the prior effect of science knowledge on how the public
understands climate change. This is taken as support for Hypothesis 7, which suggested knowledge
is a key characteristic of those demonstrating more concern about climate change. Between
Hypothesis 6 and Hypothesis 7, there is support for the two-step process enunciated in Chapter 2.
130

Basic scientific knowledge leads individuals to understand climate change in a manner similar to
climatologists. Individuals who reach this stage are, in turn, more concerned about climate change.
Other orientations are found to correlate with concern. The interaction term between
scientific knowledge and ideology speaks to high-knowledge ideologues. A direct, negative
relationship is observed between the interaction term and climate change concern. Highknowledge conservatives are less likely to perceive climate change as a threat compared to their
high-knowledge liberal and moderate counterparts. This direct effect suggests that regardless of
knowledge, individuals will utilize their predispositions and values in establishing levels of
concern. Some of this relationship, however, occurs through the prior relationship between the
ideology/science knowledge interaction term and how respondents understand climate change.
Because conservatives do not accept propositions about the dangers of burning fossil fuels (or
accept the solar radiation counterargument), they are less likely to be concerned about climate
change. The opposite pattern is observed for high-knowledge liberals.
The political ideology measure speaks to ideologues with low-levels of knowledge. The
total effect of ideology in both models suggests no statistically significant direct relationship with
concern in either model. The significant indirect effect in the fossil fuel analysis suggests there is
a measurable number of low-knowledge conservatives who accept the arguments from
climatologists about climate change and demonstrate slightly higher levels of concern compared
to low-knowledge liberals and moderates, all else equal. This relationship is not observed when
views about solar radiation is used in the model.
The faith-in-science index also maintains a positive, direct relationship with concern. Even
after controlling for knowledge and ideological values, individuals perceiving benefits from
scientific advancement were more likely to assert greater levels of concern about climate change.

131

There was a significant, sizeable total effect of 0.31 in both models. An indirect effect also
emerges. The results suggest this relationship emerges because having faith in science correlates
with acceptance of arguments about fossil fuels, although a similar pattern does not occur in the
context of solar radNo table of figures entries found.iation. Lastly, there is also a significant
direct effect, which supports the argument that normative views lead individuals to make decisions
about a problem regardless of what they might know about climate change. This is consistent with
the value-centered thesis and the expectation that individuals utilize heuristics.
Those skeptical of science, however, are not significantly more or less likely to perceive
climate change as a threat. Neither the direct, the indirect nor the total effect is significant. The
direction of the relationship is surprising, however. After accounting for climate change
knowledge, basic science knowledge, ideology, and all the other factors, the leftover variance to
be explained by the index suggests a positive relationship.
The model also considers the socio-demographic controls that test for additional
relationships not identified by the four measures of orientations used in the analysis. There are two
sets of direct effects worth noting. First, partisanship maintains a strong, direct relationship with
climate change concern. Compared to Independents, Republicans are less likely to be concerned
about the alleged problem while Democrats are more likely to express greater levels of concern.
The size of the standardized coefficients for these partisan dichotomous variables is notable,
suggesting they are, comparatively, powerful predictors of climate change concern.
The second set of relationships worth noting is between concern and the additional
measures of educational achievement. Obtaining a college education, whether a four-year or
advanced degree, is an alternative path that leads respondents to demonstrate elevated levels of
concern. What is noteworthy in this case is that the two measures outperform science knowledge,

132

in that they are both positive and have a sizeable effect on concern. This observation is taken up
in the next chapter, as it appears science knowledge may not always be the best predictor of all
types of climate change beliefs.
Since the models presented here are just-identified, there are a limited number of fit
statistics to discuss. One metric, however, that increased notably between the models is the portion
of the variance explained. As noted earlier, the R2 statistic was 5% higher when considering fossil
fuels as the causal driver of the two-step model rather than views about solar radiation. As noted
throughout the dissertation, there is considerably more noise, more counterarguments, and more
uncertainty about the solar radiation argument. While individuals might not have a complete
understanding of the cause of climate change, it appears that understanding basic technical
arguments (as with the fossil fuel proposition) is sufficient to guide individuals to demonstrate
greater levels of concern about climate change.
Summary
The intent of this chapter was to assess the proposed two-step model. The model proposes
a causal process that envisions basic scientific knowledge helping individuals to understand
climate change. If climate change is understood in the same manner as climatologists working with
the IPCC, then the public was expected to demonstrate elevated levels of concern. This causal
process materialized in the analysis.
First, there was support for Hypothesis 6. Specific types of climate change knowledge
correlated with increased concern about climate change. As individuals come to understand
climate change in a manner similar to climatologists (acknowledging that burning fossil fuels is a
cause of global warming, and rejecting the solar radiation counterargument), they become more
likely to express higher levels of concern over the phenomenon. Again, this relationship is not
133

specified by the science comprehension thesis. The thesis, at least as expressed here, only goes as
far as saying information and knowledge allow the public to make informed decisions about
whether a problem exists and whether remedial policies are required. It does appear, however, that
individuals who understand climate change in a manner similar to climatologists also share their
elevated level of concern.
Second, the model also revealed support for the Hypothesis 7, which suggested individuals
who are concerned about climate change are knowledgeable about science. From Hypothesis 6,
these individuals with elevated levels of concern are knowledgeable about climate change.
However, Chapter 4 established that such individuals are also knowledgeable about basic scientific
concepts. The indirect effect between scientific knowledge and climate change concern in the
analysis above suggests basic, foundational knowledge does have an effect on concern, but that
influence works through helping these high-knowledge individuals come to understand the
technical components of climate change. Table 5.3 restates the hypothesis.
Table 5.3: Summary of support for hypotheses
Hypothesis
Hypothesis 6: Individuals who understand climate change in a manner
consistent with climatologists are more likely to demonstrate higher levels of
concern.
Hypothesis 7: Individuals concerned about climate change are knowledgeable
about basic scientific principles.

Support?
Yes
Yes

As before, qualifications are in order. The use of an interaction term suggests the
relationships identified above also hold for moderates with high-levels of knowledge. The
interaction term suggests a familiar interpretation at this point, where high-knowledge
conservatives demonstrate lower levels of concern than high-knowledge liberals. There were no
significant differences among low-knowledge ideologues, however, in this analysis. The evidence
that there were direct relationships between high-knowledge ideologues and concern suggests that

134

values and predispositions guide the public’s level of concern regardless of climate change
knowledge. This is consistent with the value-centered thesis.20
Also consistent with the value-centered thesis are the direct relationships between the
exogenous measures and climate change concern. In both models, five of the eleven exogenous
measures were significantly correlated with climate change concern.
Individuals also appear to use heuristics provided by their orientation toward science.
Individuals with more faith in science are also more likely to assert higher levels of concern.
Holding skeptical views about scientists, however, does not correlate with higher levels of concern.
Just as with ideological values, this use of science-oriented heuristics is consistent with the valuecentered thesis.
For supplementary analysis, Appendix C is offered below as part of an attempt to identify
a multi-stage model that offers a more nuanced and elaborated understanding of the science
comprehension thesis. Afterward, the discussion shifts toward reviewing the results discussed
throughout the dissertation, identifying next steps, and considering the conditions under which the
polarization within public opinion over climate change might be reduced or mitigated.

20

Future analyses might also model these relationships through a multi-group analysis. This
technique would partition the ideologues into groups, allowing for a comparison of the knowledge
coefficients for different groupings.
135

CHAPTER 6: DISCUSSION
In this chapter, the discussion will reiterate the science comprehension and value-centered
theses and review the support for each thesis when assessed through to the proposed two-step
model. The discussion then turns toward speculating on the conditions under which the
polarization in public opinion may be mitigated, as well as the policy implications for science
advocates who emphasize science education policies. It concludes with an outline of additional
research that branches off from this discussion.
This dissertation set out as an attempt to sort through competing claims from environmental
scholars about why the public is polarized in their opinions about climate change, despite the fact
that climatologists working with the IPCC are certain that temperatures are rising, the climate is
changing, and the burning of fossil fuels is the leading cause of these changes. It acknowledged
from the start that values lead individuals to hold certain predispositions. These predispositions
lead individuals to accept or reject specific causal arguments from climatologist if the issue in
question somehow resonates for or against those predispositions.
One of the differences in the argument presented earlier is that orientations toward science
are said to provide another type of predisposition. Science provides a systematic approach to help
individuals understand the world. If one is committed to the scientific approach, or at least thinks
highly of the method, then high-knowledge individuals should stand out in empirical assessments.
The science comprehension thesis proposes that exposure to science trains individuals to think
systematically about the world. Once individuals begin thinking systematically about events such
as climate change, they become more willing to accept the arguments from climatologists because
(1) they believe in the validity of the process and (2) they have the capacity to understand technical,
complex arguments coming from the IPCC and other scientists.

136

The precise relationship between scientific knowledge, a measure of one’s orientation
toward science, and climate change beliefs is unclear given prior research. The lack of a clear,
precise relationship was argued to be the result of improper model specification. When considering
knowledge, it is essential to recognize the multiple types of knowledge individuals may possess.
That is, there is a two-step process at work. Two types of knowledge are conceptualized in this
two-step model – foundational scientific knowledge and domain-specific knowledge. The former
type of knowledge is relatively uncontroversial and covers only core scientific concepts such as
DNA, molecular structures, and basic geology. Domain-specific knowledge is more specialized
and, in the case of climate change, a small minority of scientists as well as ideological leaders
contests the science. The science comprehension thesis would suggest that foundational scientific
knowledge allows individuals to make sense of the contested causal arguments about climate
change. That is, these individuals will have an appreciation for the scientific method and have the
skills to sort through the public debate in order to identify the causal arguments from
climatologists. At that point, they can then decide whether the information merits any type of
concern. This two-step process is reproduced below in Figure 6.1, which is a replication of the
model presented in Figure 2.2.
Figure 6.1: A two-step process for climate change beliefs
Science
orientations
Climate change
knowledge
Political
orientations

137

Climate change
concern

This two-step process offers a framework of public opinion that can assess both the science
comprehension thesis and the value-centered thesis. The latter thesis emphasizes values as guiding
interpretations of policy debates. Individuals are motivated to preserve their values, which leads
them to rely on biases and predispositions when interpreting information. New information is
interpreted in a manner consistent with those predispositions or otherwise outright rejected. These
two theses are not mutually exclusive, however; individuals friendly to science may very well be
motivated to align beliefs behind those of scientists, whether those beliefs are right or wrong. With
this review of the theses in mind, the support for the previously identified hypotheses is now
discussed.
Review of Empirical Results
Hypothesis 1 expected individuals with accumulated levels of scientific knowledge to think
differently about climate change compared to their low-knowledge counterparts. This relationship
is expected because knowledge of basic science reflects one type of orientation that can be utilized
by individuals to guide interpretations of policy debates. Support for the first hypothesis emerged
on several fronts. First, throughout Chapter 4 there was a positive, significant relationship between
high-knowledge individuals and specific climate change beliefs. In terms of predictive capabilities,
the models for the greenhouse effect and fossil fuel question produced prediction rates well above
50% for most levels of knowledge. Whether one scored average on the quiz or mastered the quiz,
the predicted probability that one would accept the causal arguments of climatologists was notable.
This indicates the possibility of a low threshold, in that individuals do not need to master the
science quiz in order to accept the first two propositions offered by climatologists.
On the other hand, it was more difficult for the specified model to predict beliefs relying
on responses to the solar radiation question, at least based on scientific knowledge. High-

138

knowledge individuals did not retain high probabilities of agreement with climatologists when
compared to the prior two analyses that considered fossil fuels and the greenhouse effect. This
might seem detrimental to the science comprehension thesis until the change from median
knowledge to full knowledge is considered. There was a seventeen-point increase in the probability
of agreement with climatologists based on a move from average to full knowledge. This
improvement in predictive capabilities is almost double the improvement observed in the other
two analyses. The threshold is higher in the solar radiation analysis in that there were more notable
differences between average and high-knowledge individuals, compared to the first two issues. On
one of the most politically contested aspects of the climate change debate, high-knowledge
individuals were still more likely to stand out in terms of how they understand the problem.
As would be expected, individuals with higher levels of science knowledge were also more
likely to accept all three positions. In Table B.6, over 40% of the individuals who scored 90-100%
on the science quiz accepted all three positions. No other knowledge group accepted the
proposition at rates greater than 31%. These relationships hold in Table 5.1, where path analysis
was utilized to re-assess the findings of Chapter 4. The conclusion is that there is a positive,
significant effect between high-knowledge individuals and climate change beliefs.
Hypothesis 2 assessed one dimension of the normative views individuals hold toward
scientists, specifically the faith-in-science index. General feelings about science were argued to be
just as important of a predictor of climate change beliefs as scientific knowledge. Hypothesis 2 is
not part of the science comprehension thesis. Rather, it is considered part of the value-centered
thesis, where predispositions and biases are utilized as heuristics to guide beliefs on a given issue.
The faith-in-science index represents whether individuals perceive benefits from science and see
society as better off because of technological advancements. Positive feelings about science were

139

expected to correlate with climate change beliefs. In the first two analyses of Chapter 4, which
focused on the greenhouse effect and fossil fuel propositions, there was a positive, significant
relationship between the faith-in-science index and climate change beliefs. A move from the
“agree” to “strongly agree” position on the index was roughly equivalent in influence to a move
on the science knowledge index from 70% to 100%.
The exception to the expectation of Hypothesis 2 occurred in the solar radiation analysis.
A negative, significant relationship emerged. Individuals perceiving more benefits from science
were less likely to reject the solar radiation counterargument. Supplementary analyses in Appendix
B suggest this sign reversal likely occurred because low-knowledge individuals were less likely to
reject the counterargument as they demonstrated higher scores on the faith-in-science scale. In
bivariate analyses, it is only after controlling for scientific knowledge that the relationship between
solar radiation beliefs and the faith-in-science index turns negative. This sign reversal remained in
the path analysis provided in Table 5.1, which accounted for the error correlation between the
independent variables. However, it did not achieve traditional levels of statistical significance. In
two of the three cases, then, there was support for Hypothesis 2.21
Hypothesis 3, that those skeptical of science are more likely to reject the arguments of
climatologists, found little support in the analyses. A significant relationship did not materialize in
analyses that focused on fossil fuel usage or the greenhouse effect, although the direction of the
relationship was in the expected direction. However, the relationship did materialize in the solar
21

The relationship may be an artifact of the analysis. Note that the constant term in the solar
radiation analysis (Table 4.7) is significantly reduced compared to the other models. This suggests
model specification and interrelationships between the independent variables, then, might also be
responsible for the sign switch.

140

radiation analysis. Those more skeptical of scientists were less likely to reject the solar radiation
counterargument. This relationship, however, was outside the 95% confidence interval when
utilizing path analysis in Table 5.1. The evidence for skeptical science views significantly
correlating with climate change beliefs, then, is weak a best, although alternative question wording
or perhaps re-conceptualizing the questions utilized in constructing the skepticism index might
produce different results.22 Conceptualizing normative views as a heuristic utilized by the public
to make sense of climate change debates finds mixed support, although the relationships that do
emerge are consistent with the value-centered thesis.
Hypothesis 4 focused explicitly on political ideology as an additional method of testing the
value-centered theses. There is a well-developed literature pointing to ideological and partisan
differences in the context of climate change, and those findings materialize in the Chapter 4
analysis. Conservatives were more likely to reject the positions of climatologists while liberals
were more likely to embrace the arguments. Supplementary analysis (not shown) employed a
measure of regulatory opinions as opposed to self-identified ideology in order to offer a more
precise understanding of why this relationship occurred. Results were similar, suggesting the
ideological divide occurs because conservatives express more aversion for government regulations
while liberals are more likely to conceptualize government as a problem solver. A significant
relationship between ideology and climate beliefs emerged across the first two analysis, but not in
the solar radiation analysis. Additional analyses suggested the relationship did not materialize
because of fluctuations within responses by ideological groups. Moderate liberals (at the “-3” and
22

Some surveys, for instance, specifically ask about the level of trust in what environmental
scientists say. Drawing this distinction might offer a better explanation of negative attitudes toward
science.

141

“-4” positions) were more likely to reject the counterargument as were (to a lesser extent) strong
liberals (at the “-5” position) and strong conservatives (at the “5” position). It was the moderate
conservatives (at the “3” and “4” position) and moderates (at the “0” position) who were less likely
to identify with the position of climatologists. This fluctuation in the responses across the
ideological scale did not lead to a detectable pattern in the survey responses, however.
The discussed relationships between climate change beliefs and both ideology and science
knowledge deserve qualification, however, due to the interactive effects specified in Hypothesis
5. Throughout the analysis, there was a common theme when considering interactive effects. Highknowledge conservatives think differently about climate change compared to high-knowledge
liberals and moderates. On all dependent variables in Chapter 4, high-knowledge conservatives
were significantly less likely to accept the three propositions. In two of the three cases, as
conservatives grew more knowledgeable, their predicted probability of agreement with
climatologists actually decreased. The exception was solar radiation, where high-knowledge
conservatives were more likely to reject the proposition than low-knowledge conservatives;
however, they were still less likely to agree than high-knowledge liberals and moderates.23
However, recall the average marginal effect was insignificant for strong conservatives (at the “5”
position), but significant for more moderate conservatives. While it is clear knowledge has an
23

One reason for this difference could be that the three questions represent different stages of the
policy debate. The precise cause of climate change is a more complex topic, one where there is
more uncertainty as to the primary driver of observed weather changes. However, as noted in
Appendix B, the observed difference in responses of conservatives to the solar radiation question
may be due to survey question wording. A notable number of conservatives only agreed with
climatologists on the solar radiation proposition, whereas moderates and liberals who only agreed
with climatologists once tended to agree just on the greenhouse effect question. If these
conservatives were confused by the question wording and meant to disagree with climatologists,
then the regression results for the interaction term would be similar across all three issues.

142

effect on beliefs, the effect is moderated by ideological predispositions as expected in the
Hypothesis 5.
Furthermore, high-knowledge liberals were more likely to rally around the information
when compared to moderates. The s-shaped nature of the predicted probabilities for liberals as
they accumulated knowledge suggests that they are quick to rally around the information offered
by scientists. This is because their predispositions favor environmental protection. These
relationships materialized in Table 5.1 as well. The relationship between high-knowledge liberals
and climate change beliefs can be interpreted as support for both the science comprehension and
value-centered theses. On one hand, high-knowledge liberals think differently about climate
change compared to low-knowledge liberals, as the science comprehension thesis would suggest.
However, the s-shaped change in the predicted probability of liberals when compared to moderates
suggests knowledge is interacting with values. This is consistent with the expectations of the valuecentered thesis as well. Proponents of both theses can claim support.
Hypotheses 6 and 7 shifted focus toward the two-step process. Those who agreed with
climatologists in respect to the solar radiation and fossil fuel propositions were more likely to
demonstrate elevated levels of concern for climate change (Hypothesis 6). Those individuals
demonstrating a high-level of climate change knowledge were also highly knowledge about
science in general (see Chapter 4 and Appendix B). The path analysis revealed a significant indirect
effect, where scientifically knowledgeable individuals were more likely to demonstrate elevated
levels of concern. This relationship, however, occurs through the prior effect of scientific
knowledge on climate change specific knowledge. This is taken as support for Hypothesis 7.

143

The Limits of Knowledge
The analysis provides support for both the science comprehension and value-centered
theses. Scholarship (e.g., Kahan et al. 2012) has dismissed the role of knowledge in guiding beliefs
about climate change. This conclusion is incorrect if climate change beliefs are modeled using the
two-step processed introduced here. The two-step model is consistent with the expectations of the
science comprehension thesis. Basic scientific knowledge allows individuals to make sense of the
causal arguments presented by scientists and other policy advocates. At that point, the science
comprehension thesis has no expectation as to whether individuals demonstrate greater levels of
concern for an alleged problem. The fact that there was a positive, significant relationship between
climate change knowledge and concern simply suggests these individuals are reaching conclusions
that climate change is a potential threat.
However, there are limits. There is no direct relationship between scientific knowledge and
climate change concern. It does affect, however, climate change concern indirectly by guiding
individuals in terms of how they come to understand climate change. As conceptualized here,
scientific knowledge is limited to helping the public understand the problem. Even then, that
relationship is conditional on ideological values. There is a window that provides an opportunity
for scientific knowledge to correlate with climate change beliefs, but that window is small. There
is the potential for basic scientific knowledge to correlate with other beliefs, such as problem
recognition and maybe even policy preferences, but that relationship is best understood as indirect
and minimal.
With this said, two caveats are in order. First, recall that question wording can exacerbate
the degree of polarization in public opinion over climate change. Questions in the Science News
Survey referred to the problem as global warming rather than climate change. Prior work has noted

144

that conservatives are more likely to reject the idea of global warming, but they are more agreeable
when the problem is referred to as climate change (Schuldt et al. 2011). As such, the ideological
gap noted throughout this analysis may be the result of question wording. Questions asking about
climate change might produce different results.
Second, the survey was carried out online via Knowledge Networks. There are noticeable
differences between probability-based online surveys and random-digit dialing data collection
methods (Chang and Krosnick 2009). These differences could explain why these results vary
compared to studies by Hamilton (2012). Although probability-based online surveys are argued to
offer more representative samples, there could still be underlying biases within the survey
population.
Additional Observations
Through the analyses presented in Chapter 4 and 5, there were several observations that
did not speak to the hypotheses, but are worth reiterating. The construct validity of the science
knowledge index may have been compromised by the true or false nature of the questions, with
some respondents easily guessing answers to some of the questions. The opportunity for
respondents to indicate “Don’t Know” on the survey may also have created a degree of noise in
the measure. In Appendix C, participating in college science courses is a powerful predictor
(comparatively) of both climate change beliefs and climate change concern. The notable effect of
taking a college-level science course further underscores support for the science comprehension
thesis. It also suggests a solution for public opinion scholars who are limited in the number of
questions they can ask on a survey. If researchers do not have an opportunity to ask a battery of
science knowledge questions, inquiring about whether science courses were completed in college
might provide a useful measure of science knowledge for empirical analyses.

145

Including socio-demographic controls in Chapter 5 suggested partisanship, rather than
ideology, might offer more explanatory power in empirical analyses when analyzing climate
change concern. The powerful, direct effects between both the Democrat and Republican
dichotomous measures and concern suggest devotion to the party message, rather than ideological
values, might be a more appropriate understanding of problem recognition. An alternative
viewpoint might conceptualize individuals as viewing political debates as a “team sport.”
Individuals will support their preferred team, with more weight given to supporting the team than
ideological values.
In several cases in Chapter 4, there was an observation that the models had room to improve
their predictive capabilities of climate change beliefs. Other concepts that could add additional
explanatory power include trust in scientists and risk perceptions. The analysis did not account for
whether high-knowledge individuals trust the IPCC or the scientists researching climate change.
There is an opportunity to specify additional models that introduce trust as an independent variable.
The Science News Survey included several questions about the level of trust held for sources of
climate change information. This included trust in the IPCC, NASA, various media outlets, and
friends/family members. These measures of trust were excluded from the analysis given the subtle
nuances within the measures. For instance, high-knowledge individuals do not trust the IPCC, but
they do trust NASA and FOX News. They do trust information from interest groups like the Sierra
Club and distrust information from energy companies. These relationships seem a bit
counterintuitive (i.e., NASA scientists work with the IPCC) and require supplementary analyses
that go beyond the scope of the two-step argument presented here. They were consequently left
out of the analysis and reserved for future studies.

146

There also was no opportunity to compare political ideology with the cultural values, which
were, again, the metric utilized by Kahan et al. (2012) and other proponents of the cultural
cognition thesis. Ideally, the analysis here would compare science knowledge directly with cultural
values in order to better respond to this work. At the same time, the cultural cognition theory
emphasizes risk perceptions. Unfortunately, risk perceptions are not included in this study either.
Their inclusion might improve the overall fit of the model, given other work in the literature
(Leiserowitz 2006; Leiserowitz 2005; Lorenzoni et al. 2005).
Connection to Literature
The two-step process developed here informs the literature on three fronts. First, compared
to prior studies it offers a more nuanced and accurate assessment of the science comprehension
thesis. As noted earlier, Sturgis and Allum (2008) observed that the bulk of empirical analyses
have considered generalized science attitudes as the outcome variable. The analysis offered here,
with the focus on the acceptance of causal arguments about climate change, offers a more nuanced
and complete assessment of the thesis.
This analysis is somewhat similar to Lawrence Hamilton’s work, which utilizes interactive
effects between a general measure of educational achievement and partisanship to understand
beliefs about changes. In one paper, Hamilton (2011) finds evidence of interactive effects when
looking at threat perceptions while in a second article (2012) he considers specific beliefs about
climate change (similar to the questions in Chapter 4). In this latter analysis, Hamilton finds the
interactive effect is largely insignificant. The strongest relationship observed is with respect to
whether respondents believe in climate change. A multinomial regression revealed more educated
Republicans were more likely to say global warming is ongoing and natural (compared to ongoing
and anthropogenic). They were also more likely to say climate change is not occurring. Questions

147

about (1) changes in the Arctic Ocean, (2) the concentration of CO2 in the atmosphere, or (3) the
greenhouse effect produced no significant relationships between the interaction and expressed
beliefs by the respondents. In most cases, partisanship and a measure of education (a four-point
categorical variable) were insignificant.
The closest overlap between this dissertation and Hamilton’s work is with respect to the
greenhouse effect. The survey question in the Science News Survey was true/false in nature, asking
whether global warming is the outcome associated with the greenhouse effect. Hamilton asked a
multiple-choice question: “Scientists use the term ‘greenhouse effect’ to describe” with the answer
options consisting of the following:


“The heat-trapping properties of certain gases, such as carbon dioxide or CO2;



A hole in the Earth’s ozone layer, which allows more sunlight to get through;



The warming effect of pavement and cities; and



Don’t know/no answer.”

In Hamilton’s multinomial regression, educated Independents were more likely to select an
incorrect answer, but the interaction variable was largely insignificant. Compared to the analysis
offered in Chapter 4, the results differ in that low-educated populations in the middle of the
political spectrum were less likely to identify with the position of climatologists. Furthermore,
Chapter 4 suggested a strong, significant interaction effect between science knowledge and
ideology.
In supplementary analyses not shown here, the greenhouse effect analysis was replicated
using the variables utilized by Hamilton. The intent was to identify whether the differences in
results were due to question wording in the dependent variable or the choice of variables for the
interaction term. Utilizing a categorical measure of education and partisanship as the interaction

148

term (as well as using controls similar to Hamilton), Hamilton’s (2012) work was replicated using
the Science News Survey. In short, the differences between Hamilton’s analysis and the findings
here are largely due to how education achievement is measured. These differences are noted in
Table 6.1. Science knowledge appears to be a superior and more nuanced measurement for
predicting belief when compared to other, simpler education measures.
Table 6.1: Differences between Hamilton and Kalmbach
Hamilton (2012)
Interaction: Science knowledge * ideology
-

Kalmbach (2013)
Significant

Interaction: Education * partisanship

Not significant

Not significant

Interaction: Science knowledge * partisanship

-

Significant

The second contribution of this dissertation is the use of the generalized science attitudes
as normative views that may also shape climate change beliefs. While there was little support that
those skeptical of science reject arguments from climatologists, there was typically a strong,
positive correlation between beliefs and the faith-in-science index (although see the solar radiation
analysis discussion above). This relationship also materialized when looking at climate change
concern. In fact, in terms of standardized coefficients, having favorable views of science was a
powerful predictor of climate change concern. This finding contributes to the science literacy
literature by utilizing a traditional dependent variable and moving it to the right-hand side of the
equation. It also reinforces the role of predispositions and heuristics in shaping beliefs about
climate change.
The third and main contribution of the dissertation is the conceptualization of knowledge
as a two-step process. This is noteworthy given the recent work of Kahan et al. (2012) that has
increasingly emphasized cultural values as the dominant explanatory force behind climate change
beliefs. The emphasis on values suggests there is only a small opportunity for objective

149

information to guide beliefs. Indeed, the analyses presented here did support conclusions that
predispositions and values are a primary predictor of climate change beliefs. However, the analyses
also suggest it is inappropriate to contend there is no role for a basic, foundational understanding
of science to help individuals navigate the climate change debate. Such individuals who master
the science knowledge quiz are more likely to understand the technical arguments about climate
change, which is essential for making informed decisions about whether climate change poses a
threat. Given the discussions in the Chapters 4 and 5, possessing a familiarity with science is highly
correlated with understanding and agreeing with the arguments from climatologists.
In short, the science comprehension thesis should not be as quickly dismissed as others
have suggested. Are scientifically knowledgeable individuals more concerned about climate
change? No. Are individuals who understand the causal arguments from climatologists more likely
to share the views of climatologists? Yes, and the people who understand these technical
arguments from climatologists also tend to be scientifically knowledgeable. That relationship
deserves emphasis as scholars search for additional biases and predisposition.
Robustness
The use of weights throughout the analysis does substantively alter the strength of the
coefficients on occasions. Thus, conclusions that there are differences between the survey
population and the random adult population is appropriate. However, the influence of the weights
is in the direction against the hypotheses. That is, the strength of the relationship for key measures
such as science knowledge and ideology is increased, more often than not, in analyses that do not
utilize the weights from Knowledge Networks.
These relationships discussed above also emerge in other issues outside climate change.
Analyses not included here found similar relationships in the context of evolution and vaccination

150

safety. Understanding foundational aspects of science led individuals to accept contested
propositions about whether humans evolved from lesser species while high-knowledge individuals
are also more likely to identify with arguments from the Centers for Disease Control and
Prevention about vaccination safety. Although there is no equivalent question on the survey to
assess problem recognition in the context of these alternative issues, the first-stage of the analysis
appears robust and is applicable to non-environmental issues.
Temporal Component
One theme in the climate change literature is the instability of public opinion. For instance,
the Pew Research Center regularly tracks public sentiments related to climate change. From 2006
to 2012, they note a shift within public opinion.24 These survey findings are replicated below in
Figure 6.2. Beliefs about the strength of evidence supporting global warming dropped 20-points
while there was an approximate 10-point drop in beliefs that warming can be attributed to
anthropogenic causes. In both cases, the numbers have rebounded in recent years, although they
have not reached the level of support prior to the 2009 marked decline.
The drop in 2009 coincides with two events. First, the “Great Recession” was peaking in
terms of its influence on the economy. At the time of the Science News Survey (March 2008), the
United States economy had not yet been declared to be in a state of recession. That label was not
applied until December 2008, which was after the Pew Research Center’s October 2008 survey as
well. By the 2009 cycle of the survey, numerous countries had declared their economies to be in a
state of recession. Meanwhile, public discourse centered on the necessity of various stimulus
packages to boost the economy, as well as how to regulate (if at all) industries which were labeled
24

Additional work by Marquart-Pyatt et al. (2011) suggests this movement in public opinion was
observed for only a limited number of survey items.
151

“too big to fail.” During this time of economic trial, there is said to have been a drop in public
support for environmental issues.
Figure 6.2: Replication of climate change beliefs from the Pew Research Center
90%
80%
70%
60%
50%

40%
30%
20%
10%
0%
2006

2008

2010

2012

Yes, solid evidence the Earth is warming
Warming mostly because of human activity
A second event coinciding with the drop in public opinion was the passage of the American
Clean Energy & Security Act in the U.S. House of Representatives. The legislation would have
implemented a cap-and-trade program on carbon dioxide. The public debate over the legislation
featured the expected partisan divide, with Republicans and Democrats split on the legislation.
Once Republicans regained control of the House in the 2010 elections, new legislation – Stop the
War on Coal Act – was proposed to counter President Obama’s attempt to regulate greenhouse
gases through new rules from the Environmental Protection Agency. It is perhaps this polarized
political setting, rather than economic conditions, that are driving the fluctuations observed in
Figure 6.2. A time-series analysis suggests partisan rhetoric, as well as media attention, correlate
with changes to the public’s perceived threat of climate change (Brulle et al. 2012). Changes in
macroeconomic indicators did not appear to correlate with the observed movement in the figure.
152

Regardless of what is driving these fluctuations, do the fluctuations themselves have
implications for the science comprehension thesis? That is, are the results observed in Chapters 4
and 5 conditional on exogenous political events? The question, perhaps, is which segment of the
population is moving its opinion as a response to events within the political system. A
supplementary analysis was performed utilizing survey data from a summer 2009 Pew Research
Center study that featured questions emphasizing science knowledge and questions about climate
change. This analysis focused on whether individuals perceived a scientific consensus existing on
climate change. In this analysis, those with higher levels of science knowledge were more likely
to acknowledge the existence of a consensus. The analysis demonstrated support for the science
comprehension thesis even when utilizing an alternative dataset, with different science knowledge
questions, and in a timeframe featuring declining beliefs in climate change. A more robust analysis
would require a longitudinal study of individuals in order to track changes in public opinion as
various factors became relevant. The key question, in terms of the science comprehension thesis,
is which respondents change their beliefs given external events.
Policy Implications
In terms of policies, the AAAS spends considerable time and resources focusing on how
well the public understands science. Special issues of Science are devoted to better understanding
how to teach and communicate science. Policies to enhance educational opportunities –
particularly when it comes to teaching climate change – are not uncontroversial. Although these
analyses suggest science education programs potentially decrease polarization within public
opinion, there has been a backlash to teaching climate change in some communities (Reardon
2011). Observers have noted similarities between teaching climate change and evolution, and some
teachers simply elect to skip the discussion altogether.

153

Based on the analyses presented earlier, the “good news” for advocates is that they do not
necessarily need to teach climate change in the classroom. It appears a focus on basic scientific
concepts, as well as participation in college science courses, would be sufficient to create
movement in public opinion. In fact, efforts to increase the public’s level of scientific knowledge
would increase the public’s understanding and interpretation of the climate change debate through
multiple paths. A more knowledgeable public would increase the rate at which causal arguments
from the scientific community are accepted. The greatest change in beliefs is likely to occur among
low-knowledge liberals. Compared to moderates and conservatives, liberals were more likely to
rally around the information and accept the arguments of climatologists. Providing the necessary
foundational knowledge to understand debates would potentially drive more liberals to rally
around the information offered by climatologists. The effect would not be as large for moderates,
but there would still be gains. Conservatives, on the other hand, do not appear to align beliefs with
scientists, even if they accumulate more foundational knowledge (although see the solar radiation
argument above). In fact, educational programs, if not implemented carefully, might spark further
resistance among some conservatives, as they are likely to rally against the threatening
information.
Of course, any policy prescriptions designed to help the population understand complex
policy debates and make sense of causal arguments must acknowledge the influence of normative
views. Being overly optimistic about science and technology might have unexpected
consequences, as it can lead to respondents to both accept and reject different causal arguments
offered by scientists. As long as the story is technical in nature (greenhouse effect), it appears
favorable normative views facilitate acceptance of the causal story. However, with more complex
arguments about the cause of climate change, normative views might have the opposite effect. If

154

students are socialized to think more positively about science, the solar radiation analysis suggests
a backfire effect might occur (see technology defender hypothesis in Appendix B). Still, it does
not appear this backlash observed in the solar radiation context drives individuals away from being
concerned about climate change as indicated in Chapter 5.
More importantly, though, is the recognition that values cannot easily be altered. A change
from average to full knowledge on a science quiz score is a change that can be manufactured
through education efforts. It is more difficult to change the ideological positions of individuals.
Both ideology and partisanship are seen as stable (Jennings 1996; Jennings and Markus 1984;
Abramson and Ostrom 1991). From this standpoint, there are few policy options, aside from those
at the margins, which may significantly alter climate change beliefs for these partisans. Except for
in the context of solar radiation, ideology was a more dominant predictor of beliefs than science
knowledge in Chapter 4. Meanwhile, when considering concern in Chapter 5, partisanship was the
identified as the dominant predictor, although knowledge measures did not lag considerably.
Alleviating the polarization within public opinion may require some sort of an agreement between
ideological leaders, such as the 2007 television commercial featuring then-Speaker Pelosi and
former Speaker Gingrich discussing the need to address climate change.
Generational Replacement
Still, the generational gap noted earlier suggests some of the polarization within public
opinion can be ameliorated through generational replacement. This gap can be understood as the
byproduct of lower college attendance rates for the older generation. Could generational
replacement possibly reduce the degree of polarization in public opinion? Table 6.3 looks at the
outcome measures of Chapter 4 by age cohort. The table suggests elderly segments of the
population are less likely to have taken science college courses, which appears to correlate with

155

lower performance on the science quiz. Note, however, that while the 26-35 age group is the most
likely to have taken a science college course (52.6%), that increased exposure does not appear to
translate into a significantly improved quiz score when compared to other groups.
With respect to climate change beliefs, divergent patterns emerge, especially for the
greenhouse effect. Those 66 and older are anywhere from ten to 20 points less likely to identify
the outcome associated with the greenhouse effect (depending on the comparison group). That gap
disappears when looking at fossil fuels. Younger generations were only two to six points higher in
terms of their agreement with the fossil fuel proposition. The gap reopens, however, when
considering solar radiation, although it is not as large when compared to responses about the
greenhouse effect. Those 26-35 years old are thirteen points higher in their level of agreement with
the proposition compared to those 66 and older. The gap is less noticeable between other groups,
however – ranging from a four to eight point difference.
Table 6.2: Age by select science measures
Age

College
science

Average
quiz score

18-25
26-35
36-45
46-55
56-65
66+

40.9%
52.6%
41.8%
37.6%
27.3%
15.1%

6.8
7.0
6.9
6.8
6.5
5.5

Agree
greenhouse
effect
74.9%
66.8%
66.9%
74.4%
62.8%
56.7%

Agree
fossil fuels

Reject solar
radiation

N

54.3%
58.4%
53.4%
54.2%
51.1%
52.1%

35.6%
44.5%
38.7%
39.3%
35.9%
31.1%

149
280
299
267
201
203

Despite some noticeable differences, the prospects of alleviating polarization via
generational replacement appear limited. At most, generational replacement would likely improve
scores on the science knowledge quiz. However, a high quiz score does not guarantee uniform
views on climate change. There is considerable variation among the non-retired population despite
having similar quiz scores. For instance, the 18-25 year old cohort scores similarly on the quiz
compared to the 26-35 cohort, yet there is a nine-point difference with respect to how these two
156

cohorts approach the solar radiation proposition. Participating in college science courses might
help explain this gap between the 26-35 year olds, but only with respect to solar radiation. It cannot
explain, for instance, differences in beliefs about the greenhouse effect. In short, it is difficult to
see a clear process under which generational replacement would reduce polarization within the
public. What change in beliefs would occur appear minor.
Next Steps
The preceding discussion and analyses specified and found a set of conditions under which
scientifically knowledge individuals think differently about climate change compared to their lowknowledge counterparts. According to the science comprehension thesis, high-knowledge
individuals should accept the causal arguments from climatologists and, as expected, this
relationship materialized. However, this relationship is conditional on political values and does
not extend (directly) to threat perceptions. This conceptualization of the science comprehension
thesis can be expanded through additional research. Some targeted opportunities for additional
research are noted below.
Besides comparing alternative path models (see Appendix C summary), there is an
opportunity to include policy preferences in the discussion. The shorter, more parsimonious twostep model discussed here could easily be expanded to a three-step process, where policy questions
hinge on how the problem is understood and whether a threat is perceived if the problem goes
unchecked. Preliminary analyses have suggested support for this third step. The Science News
Survey was limited in terms of the number of environmental policy questions presented to
respondents. There is only such one policy question, about fuel economy standards for
automobiles. There are additional opportunities, however, to pursue similar analyses as those here

157

through other surveys by Dr. Jon Miller that include not only the science questions discussed here
but also questions about policy preferences.
Within the Science News Survey, there are also opportunities to apply the science
comprehension thesis to other issues. The survey asks about evolution and religious beliefs, which
would provide an excellent comparison to the discussion here on climate change. Preliminary
analyses suggest the first stage of the model can be reproduced, with the substantive effect of
science knowledge similar to the effects found in Chapter 4. Similar analyses can also be carried
out in the context of vaccination safety via a different survey. Like evolution and climate change,
competing causal arguments emerge about the safety of vaccinations. There is a possibility to run
a similar two-step model in this situation: science knowledge leading to vaccination knowledge,
which leads in turn to decisions to get an influenza vaccination.
Regarding partisan differences, there is a suggestion in the literature that Republicans are
anti-science. This argument is at odds with the emphasis here on value differences. As emphasized
throughout the text, it is not that conservatives are anti-science; rather they are likely to rally
around their core values when those values are threatened by unwelcome information. When
exploring the survey data, there were areas where both liberals and conservatives shared similar
attitudes about science. For instance, both liberals and conservatives scored high on the faith-inscience index. Elsewhere, analyses were performed of legislators and their voting habits on science
issues. The results were similar to the patterns and views discussed in this dissertation. When
values are not in conflict with science, there is bipartisan support for science issues. When values
are in conflict, some partisan group will rally against the scientific information. Additional
research that merges individual survey data with legislative behavior at the state or federal level

158

would be beneficial, given continued scholarship that insists Republicans are anti-science, which
is argued here not to be the case.
Lastly, the framing of science and the battle to interpret information is relevant to this
discussion. Strategic framing of science occurs throughout the policy process. Policy entrepreneurs
call scientific information into question as part of an attempt to focus the debate on their preferred
alternative solutions. Potential strategies include (1) outright denial of the causal argument
suggested by the information, (2) acceptance of the causal story but insistence the information is
not complete enough to produce informed policies, and (3) acceptance of causal stories and
thoroughness of information, yet assert reluctance to pursue ambitious policy solutions.
Based on these strategies, a survey experiment would allow for an assessment of how
scientifically knowledgeable individuals respond to perceptions of scientific and political conflict
on select issues. Such a survey experiment could include multiple issues that have varying degrees
of presence in public debates. That is, not all issues have clear, partisan boundaries. Of interest is
the role of both ideological and science-based orientations in processing the presented information.
The expectation is that frames that emphasize conflict and disagreement reduce both trust in
scientists and support for relevant policies. However, orientations toward science such as scientific
knowledge and faith in the scientific process are expected to lessen the influence of the partisan
frames and perhaps even produce a backlash effect. This effect would occur if those with proscience orientations ended up supporting and trusting scientists even more when conflict erupts.
Also of interest is whether individuals rely on ideological beliefs when the issues are far removed
from public discourse, as there would be less commentary from ideological leaders to guide views.
Such an experiment would offer insights toward further mapping out the “vulnerable points” at
which advocates can sway opinion, and thus accelerate or delay policy change.

159

The research outlined above provides a way to assess the degree to which scientifically
knowledgeable individuals differ from the rest of the population. The framing experiment is
particularly worthwhile, given the degree to which scientists are involved in offering information
to decision-makers and the willingness of partisans to call into question scientists’ assertions. As
science advocates continue to push for a more scientifically aware public, it is essential that society
understands what a high-knowledge public “looks like” – both in terms of what a scientifically
aware public believes and how they behave.

160

APPENDICES

161

Appendix A: Participation in the Climate Change Debate
With some evidence for the science comprehension thesis in the analysis, an unanswered
question is how knowledge promotes congruence between public opinion and science.
Specifically, the thesis suggests that knowledgeable individuals have the capacity to engage in the
climate change debate. This follow-up to the prior analyses considers whether individuals with
higher levels of knowledge engage in climate change discussions and seek out information more
frequently than their low-knowledge counterparts. These two activities represent different levels
of engagement in the policy debate. The latter activity is more passive. Conversations can occur
while in the car, over dinner, or while simply hanging out over the weekend. Individuals who
engage in these activities are effectively paying attention to the issue, but it is more passive in
nature as the activity is contingent on the presence and willingness of others to engage in these
conversations. The second activity represents a more active form of attention as individuals have
to devote considerable time and cognitive effort to seek out and consume additional information
about climate change.
The third wave of the Science News Survey included a battery of activities that represent
types of participation, with respondents reporting how many times they engaged in various
information acquisition activities over the last year. Specifically, the question asked the following
question: “Thinking about the global climate change issue, how many times have you done each
of the following activities during the last 12 months. If you have not done an activity, please enter
zero and go to the next item.” Respondents were then presented with a series of activities, which
speak to the level of attention respondents devoted to the climate change discussions. These
activities have been grouped into a “passive” and “active” form of attention.

162

The measure of passive attention includes the following activities, which were listed after
the above noted question:


Talked to my friends or co-workers about this issue; and



Talked to other members of my family about this issue.

Responses to the two questions were simply added together to create a combined score. The
average reported number of conversations in a given year was approximately 11, with a median
occurrence of four (N=926). Approximately one-third of the survey respondents indicated that they
did not talk about climate change with friends, co-workers or family members at all over the course
of the last year.
The measure of active attention includes the following three activities listed after the above
noted question, with respondents again allowed to indicate their level of engagement:


Read a newspaper or magazine article about this issue;



Looked for information about climate change on the Internet; and



Read a book about climate change.

Responses to the three questions were added together. The average reported number of times
information was actively consumed was approximately nine, with a median occurrence of three
(N=925). Approximately 34% of the population indicated no activity on this front.
A negative binomial regression was selected to model the relationship between knowledge
and activity due to the nature of the dependent variable as it represents a count process where
individuals have a propensity to engage in these select activities once they reach an unspecified
threshold. That is, respondents who engage in an activity once or twice possess increased
likelihoods that they will continue to engage in such activities. This conceptualization is
inappropriate for ordinary least squares regression, which assumes a linear relationship and no
163

threshold effects. Instead, a negative binomial regression is utilized because of the non-linear
nature of the dependent variable. Furthermore, a negative binomial is favored over a Poisson model
because the conditional mean and variance fluctuate based on performance on the science
knowledge quiz. The analysis utilizes equations one and three from Chapter 4. There was no
evidence of an interactive effect, so the following table presents a core model followed by a
secondary model that includes socio-demographic controls. The anticipated directions of the
relationships are diagramed in Figure A.1.
Figure A.1: Proposed relationships between selected orientations and attention to climate
change (passive/active)
Science orientations
Science Knowledge (+)
Faith in science (+)
Skeptical of science (-)
Political orientations
Political ideology (+)

Active attention (additive index)
1) Read newspaper/magazine
2) Research on Internet
3) Read book

Socio-demographic controls
Age (-)
Gender (+)
Democrat (+)
Republican (-)
Four-year degree (+)
Advanced degree (+)

Passive attention (additive index)
4) Talked with friends/co-worker
5) Talked with family

Results from the negative binomial regressions are presented in Table A.1. Focusing first
on the willingness of respondents to talk about climate change, Model 1 suggests those with
elevated levels of scientific knowledge are more likely to talk about climate change with their
friends and family members. However, the level of significance is outside the 95% confidence
level. Still, 94 out of 100 times the relationship is expected to match that identified here. When it
comes to looking up information (Model 3), those who possess elevated levels of scientific

164

knowledge are indeed more likely to read newspapers and books, as well as carry out research on
the internet. This relationship is significant at traditional levels of significance. These relationships
do not change in the expanded models (Model 2 and 4). The substantive effect of science
knowledge on attention will be discussed more below, but at this stage, the relationships are
consistent with what proponents of the science comprehension thesis would suggest.
Table A.1: Determinants of climate change debate participation
Discuss climate change
Seek information
Model 1
Model 2
Model 3
Model 4
0.11
0.11
0.19
0.189
Scientific knowledge
(1.87)
(2.08)
(5.11)
(4.75)
0.28
0.35
0.35
0.35
Faith-in-science
(1.88)
(2.60)
(2.01)
(2.27)
-0.34
-0.29
-0.11
-0.11
Skeptical-of-science
(-2.04)
(-2.00)
(-0.81)
(-0.83)
-0.00
-0.05
-0.01
-0.04
Ideology
(-0.09)
(-1.38)
(-0.36)
(-1.17)
0.40
0.09
Republican
(1.49)
(0.37)
0.07
-0.09
Democrat
(0.40)
(-0.47)
0.01
0.01
Age
(1.70)
(2.47)
-0.48
-0.10
Female
(-2.72)
(-0.54)
-0.01
0.09
Education (4 year)
(-0.04)
(0.36)
0.46
0.40
Education (advance)
(2.00)
(1.80)
1.46
0.81
-0.03
-0.53
Constant
(1.49)
(0.92)
(0.04)
(-0.67)
Log-likelihood -2289
-2275
-2188
-2181
N= 743
743
742
742
2 0.012
0.018
0.017
0.020
Pseudo R
Notes: The first number represents the coefficient derived from a logistic regression. The
number in parenthesis indicates the z-score from the significance test. In all models, a weight
provided by Knowledge Networks was utilized in order to make inferences from the results back
to the U.S. adult population.

165

Unlike in previous analyses that examined climate change beliefs, ideology retains little
influence on the level of attention respondents devote to climate change throughout the models.
Liberals and conservatives do not differ in their tendencies to talk about climate change or seek
out information. This interpretation remains after including the socio-demographic controls.
Normative views of science reveal inconsistent relationships. Individuals who perceive
more benefits from science are more likely to talk about science and look up information. The
relationship is significant at the 95% confidence level in all models except Model 1. However,
those more skeptical of scientific endeavors are less likely to talk about climate change, but they
are not significantly less likely to carry out research on the issue by looking up information.
A Wald test suggests the socio-demographic controls increase the explanatory power of
the model (p<0.01). This explanatory power comes from accounting for females, who were less
likely to talk about climate change with their acquaintances. Furthermore, individuals with
advanced degrees were also more likely to talk about climate change – suggesting science
knowledge is not the only education metric capable of predicting climate change behavior. These
two measures are the primary drivers of Model 2’s improved explanatory power. A similar Wald
test, however, suggests the socio-demographic controls add no explanatory power when it comes
to seeking out information about climate change (Model 4). This is expected, given the poor
performance of the controls in predicting behavior.
Substantive Effects
Given the additional explanatory power offered by the socio-demographic controls in
Model 2, the expanded models (2 and 4) will be used to look at the substantive effects of the four
core hypotheses. Figure A.1 displays the predicted level of activity for survey respondents based
on the coefficients derived from Models 2 and 4 of Table A.1. Looking specifically at

166

conversations with friends, family, and co-workers suggests that individuals with no basic
foundational understanding of science talk about climate change approximately five times each
year. As knowledge accumulates, the number of times respondents engage in conversations
increases steadily. For respondents with a median score on the science quiz (70% correct), the
predicted number of conversations is just over ten times per year. This increases to fourteen
discussions per year for those mastering the science quiz. In this context, those with scientific
knowledge demonstrate an increased willingness to engage in climate change conversations.
A similar pattern emerges when looking at the tendencies of individuals to seek out
information on climate change, which are also charted in Figure A.1. Those who fail the quiz are
likely to look up information on climate change approximately twice a year. At the median science
quiz score of 70%, an individual is likely to look up climate change information roughly nine times
per year. At full knowledge, individuals are likely to pursue additional information almost sixteen
times per year. That is, those possessing higher levels of scientific knowledge are more likely to
engage in both discussions about climate change and in seeking additional information, although
the relationship between knowledge and information gathering is stronger.
Table A.2 focuses on normative science views, utilizing the three main profiles noted
earlier. The weak pro-science profile again provides the baseline for the discussion. These
individuals agree with propositions about the benefits of scientists and disagree with positions that
express skepticism about science. A move from weak to strong pro-science views corresponds to
four additional climate change discussions each year and approximately four more information
acquisition activities. An increase in skepticism, however, corresponds to a three-point decline in
climate change conversations, but only a one-point decrease in the pursuit of additional
information.

167

Figure A.2: Expected level of engagement in information acquisition activities
20
18
16
14
12
Predicted count 10
8
6
4
2
0
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Percent correct on science quiz
Discuss CC

Seek info

Notes: Predicted counts calculated with coefficients from models two and four in Table A.1. All
other variables held at their respective means.

Table A.2: Expected involvement in climate change debate by normative views
Strong pro-science:
Weak pro-science:
Conflicted:
Faith science (s.
Faith science (agree),
Faith science (agree)
agree), skeptical
skeptical science
skeptical science
science (disagree)
(disagree)
(agree)
Discuss climate
14.6
10.3
7.7
change
Seek information

12.1

8.5

7.6

Notes: Probabilities calculated with coefficients from Models 2 and 4 of Table A.1. All other
variables held at respective means, unless noted in column heading.
To see whether attention corresponds to specific climate change beliefs, mean scores for
the attention metrics are presented in Table A.3, which are then cross-tabulated with answers to
the climate change questions featured in Chapter 4. The difference of means was then calculated
to see if those agreeing with climatologists were more likely to pay attention to climate change.
The results suggest individuals agreeing with climatologists are indeed more likely to engage in

168

these identified behaviors. However, the difference is only significant in two of the six situations.
Individuals agreeing with climatologists on the greenhouse effect and solar radiation were more
likely to seek out additional information. There were no significant relationships when considering
an individual’s propensity to talk about climate change.
Table A.3: Expected involvement in climate change debate by normative views
Greenhouse effect
Fossil fuels
Solar radiation
Agree
Disagree
Agree
Disagree
Agree
Disagree
Discuss climate change
11.7
8.5
11.3
9.9
12.8#
9.4#
Seek information
10.2*
7.1*
10.2
7.9
12.1*
7.3*
Notes: Values reflect mean level of activity for each index. Calculations reflect weighting
mechanism provided by Knowledge Networks. An * indicates the means are significantly different
at the 95% confidence level, while a # indicates significance at the 90% threshold.
This secondary analysis provides a preliminary explanation for the results uncovered when
exploring climate change beliefs. It appears possessing a higher level of scientific knowledge
provides a foundation for the public to talk about climate change and seek out additional
information. That is, they have the capacity to vet the claims of both climatologists and partisan
commentators. This is especially true in the context of seeking out additional information via web
searches, reading books, or other information acquisition activities, where the relationship is
stronger and the results suggest greater confidence in the results. This does not always translate,
however, into the acceptance of causal arguments. The proposition that foundational knowledge
increases the capacity to consume information, which then allows individuals to make informed
decisions about what to believe finds the most support in the context of solar radiation. The fact
that the relationship materializes for this most contested and controversial take on climate change
reinforces support for the science comprehension thesis.
Discussion
This supplementary analysis informs why some of the relationships noted earlier in Chapter
4 materialized. The models suggest scientific knowledge maintains a positive overall relationship
169

with behavior. Knowledgeable individuals are more likely to talk about climate change with
friends and family. Knowledge also appears to provide the skills and intellectual capacity for
respondents to seek out additional climate change information. The results suggest scientific
knowledge provides an essential foundation that guides the citizenry to engage in policy debates
by providing the capabilities to assess the information and causal arguments from climatologists.
This relationship is especially true in the context of solar radiation, where seeking out additional
information is associated with rejecting the solar radiation counterarguments. This is consistent
with the science comprehension thesis, which speculates knowledge helps provide the intellectual
capacity to engage in discussions with friends and family members.
It is not clear who is actually initiating the conversations reported by respondents. Is it the
respondent or the family that initiates the discussions? It might be the case that those who are less
knowledgeable of science turn to family members (or friends) they deem more knowledgeable on
these matters for advice. As such, it is not that knowledge drives individuals to talk about climate
change and analyze ideas, but rather knowledge makes individuals “beacons” that attract lessknowledgeable acquaintances.
Those possessing ideological predispositions did not appear more or less likely to engage
in specific behavior. This can be expected given the attention climate change has received at times
on the systemic agenda. The fact that climate change has gathered the attention of political leaders
and policy entrepreneurs on both sides of the political spectrum increases the likelihood that both
conservatives and liberals will engage in the selected behavior. As such, the results suggest that
just because an issue does not align with core beliefs or a base set of values, respondents are still
willing to engage in discussions and seek out information. That is, they do not ignore the issue just

170

because it does not resonate with them. It is important for the ideologues to keep pace with their
counterparts to maintain a level of competence in the policy discussions.
The difficultly in further assessing the ideological relationships within the data is that the
content of the discussions is unknown. For instance, it is not known what specific information is
being gathered when conservatives research climate change information. A follow-up question
might consider what information such individuals are relying on or trust when they engage in this
activity. Are they turning to an official report from the IPCC (IPCC), or are they turning to
conservative blogs that offer an alternative message? In the former instance, it is possible they are
assessing the claims of political leaders that climate change is anthropogenic, but in the latter case,
they might be looking for a reason to remain skeptical. If they do turn to the IPCC, then it would
further support the science comprehension thesis. Respondents are using their scientific knowledge
as a foundation to understand the complex arguments coming from climatologists.
This appendix was offered to help further understand why high-knowledge individuals
appear more likely to accept the ideas coming from climatologists. The science comprehension
thesis suggests individuals have the capacity to engage in the debate by engaging in information
acquisition activities and discussing the information with others. Support for this component of
the thesis was found in Appendix A. With this understanding, the discussion turns to the second
appendix to look at the interrelationship between the three belief measures utilized throughout
Chapter 4.

171

Appendix B: Interrelationship between Beliefs
The three outcome measures discussed in Chapter 4 represent various stages of the climate
change debate and are interrelated. To support the position that humans are the cause of escalating
temperatures seemingly requires someone to understand the greenhouse effect and acknowledge
that the burning of fossil fuels warms the Earth. In this supplementary analysis, the relationship
between these three outcome measures is further explored in an attempt to identify whether any
systematic differences exist within the interrelationship between responses to these three
questions.
A correlation matrix is presented in Table B.1. The matrix suggests a moderate correlation
between beliefs about the greenhouse effect and fossil fuels, but a weaker correlation between
solar radiation beliefs and the other two measures. For an additional look at the relationship
between the measures, a reliability analysis was also performed. Table B.2 presents the Cronbach’s
alpha for the interrelationships between the three outcome measures. The results suggest the survey
respondents consider each question in a unique way. Responses to one component of the climate
change debate do not appear to correlate with agreement elsewhere. As with the correlation matrix,
the greatest similarity is between the greenhouse effect and fossil fuel questions. The coefficient
(0.59), however, suggests they do not align along a single dimension. The coefficient is reduced
noticeably when considering the relationship between solar radiation and the other two measures.
Combined, the Cronbach’s alpha is only 0.54 between the three measures. This “unreliability” is
expected given that each question represents different components of the policy debate and invokes
different reactions from survey respondents.

172

Table B.1: Correlation matrix
Green.
Fossil
effect
fuels
Greenhouse
1.0000
effect
Fossil
0.4153
1.000
fuels
Solar
0.2544
0.2557
radiation

Table B.2: Reliability coefficients
Green.
Fossil
effect
fuels
Greenhouse
.5860
effect
Fossil
fuels

Solar
rad.

1.000

Notes: Coefficients represent Pearson’s
calculations between the indicated measures.

All three

Solar
rad.
.3512
.3649

.5425

r Notes: Coefficients represent Cronbach’s alpha
calculations between the indicated measures.

Given the degree of controversy and scientific certainty found within each component of
the climate change debate, it is likely systematic patterns exist within the additive index. To
explore this further, Table B.3 presents an additive index of the three questions. The distribution
indicates 25% of the sample agrees with all three measures. The majority of respondents agree
with at least two of the environmental questions. Only 16% of the sample rejected all three
propositions. Collectively, this additive index (referred to as climate change knowledge) represents
the degree to which respondents agree with climatologists.
Table B.3: Distribution of additive index
of climate change knowledge
N
%
None correct
222
16.2%
One correct
386
28.3%
Two correct
414
30.3%
Three correct
343
25.1%
1364
Note: Distribution weighted to U.S. adult
population
Given the argued progression between the three questions, it is expected that ideological
players might “drop out” along the path to full agreement with the IPCC. That is, with more
certainty about the first two propositions and less certainty as to the third, it might be expected that
individuals in the “2” category agree in respect to questions about fossil fuels and the greenhouse

173

effect, but fail to reject the solar radiation counterargument as the IPCC suggests. Table B.4 charts
how ideologues perform across the climate change knowledge scale. Three different ideological
patterns emerge when looking at the table. Liberals are more likely to agree with all three
propositions compared to conservatives, with 43% of all strong liberals agreeing with
climatologists working with the IPCC. Conservatives were more likely to disagree with all three
questions as over 20% of moderate and strong conservatives (value of “3” or greater on the
ideology scale) received a score of zero on the knowledge index. Moderates are more uniform in
their distribution across the index compared to liberals and conservatives.
Table B.4: Climate change knowledge by ideology
Ideology
0
1
Liberal (-5)
27.5%
7.3%
-4
12.0%
7.3%
-3
0.0%
20.5%
-2
3.6%
25.6%
-1
4.8%
24.6%
0
14.6%
25.7%
1
6.1%
35.0%
2
16.8%
27.0%
3
23.8%
30.3%
4
23.4%
42.6%
Conservative (5)
21.6%
40.1%
Note: Distribution weighted to U.S. adult population

2
22.2%
39.7%
34.0%
27.9%
36.6%
35.6%
36.3%
28.0%
30.7%
20.5%
25.2%

3
43.0%
41.0%
45.4%
43.0%
34.0%
24.2%
22.6%
28.3%
15.2%
13.4%
13.1%

Total
48
25
56
86
80
317
110
145
146
58
48

Focusing on which specific questions respondents answered incorrectly reveals a common
theme across ideology, however. As shown above, conservatives are more likely to agree with
only one claim (Category 1). Conservatives (defined by those with a score of “3” or higher on the
ideology scale) agreed with just the greenhouse effect question at a rate of 40%; 19% agreed with
only the fossil fuels proposition; and 40% rejected only the solar radiation argument. Of the few
liberals agreeing with only one claim, the large majority, 87.5%, agreed with the question about
the greenhouse effect. Of the moderates agreeing with only one of the arguments, the greenhouse

174

effect was again the question most commonly answered in a manner consistent with the IPCC, at
56.7%. Shifting focus to those who agreed with two questions, liberals (defined as those with
scores of “-3” or lower on the ideology scale) most often disagreed with climatologists about solar
radiation (89.5%). Conservatives in the second category (agreeing with two climate change claims)
also largely disagreed with climatologists on this issue (63.2%). Moderates were also more likely
to disagree with the IPCC on solar radiation (73.7%). Individuals who answered only two
questions consistently with climatologists, then, largely disagreed with the IPCC on whether solar
radiation explains rising temperatures. These distributions are found in Table B.5.
Table B.5: Patterns in disagreement with climatologists
Agree with IPCC once, Only
Ideology
Agreeing with:
Green.
Fossil
Solar
effect
fuels
rad.
Conservatives
40.0%
18.8%
41.2%
(35)
(17)
(36)
Moderates
56.6%
25.0%
18.3%
(114)
(50)
(37)
Liberals
85.2%
14.8%
0.0%
(14)
(2)
(0)
Note: Distribution weighted to U.S. adult population

Agree with IPCC twice,
Only disagreeing with:
Green.
Fossil
Solar
effect
fuels
rad.
13.1%
24.0%
62.9%
(9)
(16)
(43)
5.1%
21.0%
73.9%
(13)
(52)
(182)
8.6%
12.5%
78.9%
(3)
(5)
(31)

Lastly, the relationship between science knowledge and climate change knowledge is
presented in Table B.6. Those with higher quiz scores are more likely to agree with all three climate
change questions. The jump is particularly noticeable for those who achieve quiz scores of 90%
and 100%. Of those with above-average performance on the science quiz, more than two-fifths
agreed to all three of the climate change arguments. Less than one-third of the respondents with
medium levels of knowledge (70%) were likely to agree with all three propositions. This
distribution meet the expectations of the science comprehension thesis, as those with higher quiz
scores were noticeably more likely to express beliefs consistent with mainstream climatologists.

175

Table B.6: Climate change knowledge by science quiz score performance
Climate change knowledge score
Science quiz
0
1
2
3
Total
score
0%
47.5%
52.5%
0.0%
0.0%
15
10%
64.5%
21.6%
0.0%
13.9%
22
20%
44.9%
40.4%
11.0%
3.7%
65
30%
25.6%
45.8%
23.7%
4.9%
97
40%
15.3%
40.1%
32.7%
11.9%
125
50%
18.9%
28.2%
33.7%
19.2%
167
60%
10.5%
35.0%
39.0%
15.5%
151
70%
15.2%
23.1%
30.5%
31.3%
154
80%
13.1%
18.3%
42.4%
26.1%
193
90%
4.9%
20.2%
28.4%
46.4%
170
100%
11.0%
23.0%
24.6%
41.5%
204
Mean science
51.6%*
58.5%*
67.7%*
78.2%*
quiz score
Note: Distribution weighted to U.S. adult population. An * indicates whether the mean quiz scores
are significantly different at the 95% confidence level from other quiz scores.
Technology Defender Hypothesis
The combined climate change knowledge index provides another opportunity to explore
the relationship between the faith-in-science index and solar radiation beliefs. Recall there was a
sign reversal in Table 4.7. Individuals may understand the greenhouse effect and agree that burning
fossil fuels damages the environment, but still view those technological advancements – such as
the combustion engine and the development of “power-hungry” consumer products – that emit
more greenhouse gases into the atmosphere as beneficial to society. Some scholars have noted
technological advancement comes at a price (Brown and Sovacool 2011). One potential response
to this trade-off is to embrace natural forces as an alternative causal explanation for global warming
rather than accept the environmental costs of technological advancement. If this interpretation
holds, this finding raises the possibility of a “backfire” effect. If respondents think too highly of
science and technological advancements, they might become blind to the negative consequences
associated with these developments. Thus, positive orientations toward science may actually
176

exacerbate the polarization within public opinion. This is one alternative explanation to the change
in sign associated with the faith-in-science index. This possibility, referred to as the technology
defender hypothesis, is explored further below.
In supplementary analyses, a dichotomous measure was constructed to assess who belongs
in this profile of individuals, those who accept propositions about the greenhouse effect and fossil
fuels yet fail to reject the solar radiation argument. Survey respondents were assigned a score of
“1” if they agreed with all the propositions except solar radiation, 0 for all others. A regression
analysis is presented in Table B.7. Model 1 suggests that liberals and those with higher degrees of
faith in technology comprise this group. These ideological results are consistent with prior
analyses, as conservatives were less likely to agree with two of the three arguments. As such, this
grouping contained more liberals and moderates than the rest of the survey population. Just as
importantly, those demonstrating elevated levels of faith in science were more likely to dominate
this category of individuals. Model 2, with demographic controls, offers slightly different results.
The coefficients suggest the Republican dichotomous measure does a better job explaining
variation than ideology. In fact, Republicans are less likely to fall into Category 2 compared to
others (which is still consistent with Table B.4). Females also fall into this group more frequently
than males. However, an additional difference of means analysis suggests men and women who
fit this technology defender profile perform equally well on the science knowledge quiz.
The coefficients from Model 2 were utilized to calculate the substantive effects of a change
in the normative science profiles. As before, Model 2 is utilized because the socio-demographic
controls increases the explanatory power of the model. Substantively, a move from the earlier
noted “weak pro-science” profile to the “strong pro-science” profile produced a change of 7.2% in

177

the probability respondents fall within this profile of individuals. This change is similar in
magnitude to other analyses performed earlier.
Table B.7: Determinates of individuals who believe in climate
change but reject the solar radiation argument
Model 1
Model 2
-0.00
0.02
Scientific knowledge
(-0.03)
(0.46)
Faith-in-science
0.47
0.41
(3.02)
(2.64)
0.21
0.15
Skeptical-of-science
(1.34)
(0.94)
-0.07
-0.04
Ideology
(-2.02)
(-0.85)
-0.73
Republican
(-3.02)
-0.13
Democrat
(-0.61)
0.00
Age
(0.45)
-0.38
Female
(-2.01)
-0.39
4-yr degree
(-1.60)
-0.45
Advanced degree
(-1.45)
-3.22
-2.61
Constant
(-4.21)
(-3.03)
Model 2: N=1082
Log-likelihood -563.3 Wald Chi2 15.2
Model 2: N=1082
Log-likelihood -549.3 Wald Chi2 35.3
Notes: The first number represents the coefficient derived from a
logistic regression. The number in parenthesis indicates the z-score
from the significance test. In all models, a weight provided by
Knowledge Networks was utilized in order to make inferences from
the results back to the U.S. adult population.
Summary
The analyses here support an interpretation of conservatives as growing increasingly
resistant to arguments from the IPCC as the issues become more technical and specific about the
relationship between humans and the environment. They are less likely to agree with the fossil fuel

178

and solar radiation propositions. The claims they do agree with tend to be of low controversy, such
as the basic facts about the outcomes associated with the greenhouse effect. Meanwhile, liberals
are more likely to agree with all three propositions, although they also dominate Category 2.
It is puzzling as to why 41% of conservatives rejected propositions about the greenhouse
effect and fossil fuels, yet aligned their views with the IPCC in respect to climate change. It is
possible that they found the question wording confusing, and that their true intent was to deny
anthropogenic climate change. Supplementary analysis, not shown here, recoded conservatives
who only agreed only with climatologists on solar radiation as, in fact, disagreeing with
climatologists. Recalculations of Model 3 from Table 4.7 produced an interactive effect similar to
the first two issues. That is, strong conservatives were increasingly less likely to agree with
climatologists as they accumulated knowledge. This is, again, consistent with arguments that
ideologues adopt positions opposite their political opponents.
Still, there are some commonalities regardless of ideology. Of those who only agree with
one proposition, all ideological groups appear most likely to agree with the greenhouse effect
argument. Of those who agree with two propositions, they are likely to agree that the greenhouse
effect is associated with warming and that humans can influence the climate by burning fossil
fuels. Regardless of ideology, respondents in Category 2 are not likely to accept the IPCC’s
narrative on solar radiation.
There was also support for the technology defender hypothesis. Those who accept
arguments from climatologists but still reject the solar radiation counterargument have elevated
levels of faith in science. This effect is likely understated, as individuals – notably conservatives
– might be defending their way of life when they deny the other propositions as well. Most
importantly, perhaps, is that some of these individuals who reject the arguments of climatologists

179

– particular on solar radiation – do have positive feelings about scientists and their work. This
creates a potential “backfire” effect, where efforts to generate goodwill for scientists may make it
difficult for those same scientists to convince a meaningful segment of the population there is a
problem.

180

Appendix C: Toward a More Complete Model
The prior path models were just-identified, meaning that all potential relationships among
the variables were tested. Outside the R2 statistic, there was little ability to talk about the goodnessof-fit for the models. The style of analysis was chosen, however, in order to test the two-step
process enunciated in the theory. Typically, it is beneficial to compare and contrast models and
look at the fit statistics to identify a superior conceptualization of the relationships. A more
elaborate, multi-staged model of public opinion is discussed below as a way to offer an expanded
test of the science comprehension thesis.
This conceptualization relies on the basic arguments discussed throughout the dissertation.
At a basic level, Figure C.1 provides a more elaborate diagram of the science comprehension
thesis. This process starts with the socialization process, where individuals are exposed to science
in their educational careers (Step 1). That process is the source of one’s foundational understanding
of science (Step 2). Science knowledge provides the skills to engage in information acquisition
activities (Step 3), which in turn lead individuals to understand and evaluate the claims of scientists
(Step 4). From that point, individuals can make informed decisions about whether they need to be
concerned about climate change (Step 5).
Figure C.1: Basics of the science comprehension thesis
College
science

Science
knowledge

Information
acquisition

CC
knowledge

CC
concern

Again, scientific knowledge is simply one factor that helps guide public opinion. Figure
C.2 populates the relationships found within Figure C.1 with additional measures. Going from left
to right, the model starts with two exogenous variables – age and gender. These two independent
variables are not contingent on any other factor. Once individuals are born, they begin the
181

socialization process. Unfortunately, the only question in the Science News Survey that might
account for this socialization processes is participation in college science courses. There are no
measures of political socialization that assess how parents and communities socialized their youth.
Once they go through this socialization process, individuals have developed a belief system. These
belief systems feature a series of orientations toward specific viewpoints. The science knowledge
quiz score represents an individual’s familiarity with science and whether they accept basic,
fundamental arguments from scientists. As noted in Chapter 2, individuals who score high on the
science knowledge quiz are also familiar with the scientific method. There are other orientations
as well, such as ideology and normative views of science. These additional orientations are
included alongside scientific knowledge. Individuals use these orientations to gather information
about climate change. These activities correspond to the information acquisition activities
discussed in Appendix A. They use this information, in turn, to develop specific beliefs about the
phenomenon. After developing these beliefs, they are in a position to determine whether climate
change constitutes a threat.
This description constitutes a multi-stage conceptualization of public opinion. While there
are elements of the parsimonious two-step model discussed earlier, the model presented in Figure
C.2 offers a more nuanced understanding of public opinion. The solid lines in Figure C.2 indicate
relationships consistent with the empirical results throughout the first two empirical chapters,
although some relationships come from unreported auxiliary research. The circular symbols
around the endogenous measures indicate the errors for the respective measures, with the dotted
lines indicating the anticipated correlation between the designated error terms. The results of
running Figure C.2 through a path analysis are presented in Table C.1. The analysis will focus on
the model’s explanatory power of climate change knowledge and climate change concern.

182

Figure C.2: Multi-stage model

Age

College
science

Science
knowledge

ɛ7

ɛ8

Seek out
information

ɛ3

Ideology

ɛ4

Gender
Interaction:
knowledge * ideology

ɛ5
ɛ6

Faith in
science
Skeptical
of science

183

ɛ10

ɛ9

Discuss CC

ɛ2

ɛ1

CC
knowledge

CC
concern

Table C.1: Direct, indirect and total effects of climate change knowledge and concern
Climate change knowledge
Climate change concern
Direct
Indirect
Direct
Indirect
Total effect
Total effect
effect
effect
effect
effect
Age
-0.00
-0.00
-0.00
0.00
(-0.71)
(-0.71)
(-0.10)
(-0.10)
Gender
-0.13
-0.13
-0.08
-0.07
(-4.29)
(-4.29)
(-3.02)
(-3.02)
College
0.25
0.25
0.12
0.12
Science
(11.61)
(11.61)
(10.35)
(10.35)
Science
0.12
0.00
0.13
0.05
0.05
Knowledge
(8.14)
(6.26)
(8.41)
(10.17)
(10.17)
Ideology
0.05
0.05
0.02
0.01
0.03
(1.02)
(1.02)
(0.39)
(1.02)
(0.65)
Interaction
-0.02
-0.02
-0.02
-0.01
-0.03
(-4.06)
(-4.06)
(-3.76)
(-3.59)
(-4.72)
Faith
0.08
0.08
0.14
0.02
0.16
Science
(1.23)
(1.23)
(2.07)
(1.23)
(2.34)
Skeptical
-0.00
-0.00
-0.00
0.00
Science
(-0.01)
(-0.01)
(-0.01)
(-0.01)
Talk CC
-0.00
-0.00
-0.00
0.00
(-0.14)
(-0.14)
(-0.14)
(0.30)
Seek
0.00
0.00
0.01
0.00
0.01
Information
(1.00)
(1.00)
(3.77)
(1.00)
(3.87)
Climate Ch.
0.30
0.30
Knowledge
(7.68)
(7.68)
Model Details: N = 737 Log-likelihood = -20249 R2: 0.27(c.c. know.) R2: 0.31 (concern)
Notes: The first number represents the standardized coefficient. The number in parentheses indicates
the z-score from the significance test. All covariances between the error terms drawn in Figure C.2
are significant (p<0.01). Results are not weighted.25
Analysis
Table C.1 displays the results from the path analysis, which reveal the direct, indirect, and
total effects for the two key endogenous measures for this discussion – climate change knowledge
and climate change concern. There are several notable observations in the model. First, consistent
25

The analysis is not weighted, in order to offer a discussion of the models’ goodness-of-fit
measures. Stata utilizes robust standard errors when applying the weight to the data, which
invalidates the goodness-of-fit statistics. Weighting the analysis to the third wave of the survey
increases the size of the coefficients, but the significance remains the same. Thus, this analysis
may understate the substantive strength of the relationships.
184

with the science comprehension thesis, there is continued support for the Hypothesis 1 and 7.
Moderates with higher levels of science knowledge are more likely to score higher on the climate
change knowledge scale while an indirect, significant relationship between knowledge and
concern. There is also support for Hypothesis 5. The interaction term suggests the science
comprehension deserves qualifications, as high-know moderates, liberals, and conservatives all
think differently about climate change, with similar interpretations to those made in Chapter 4 and
5. As in Chapter 5, the substantive effect of ideology is limited. Absent in this model compared to
others in Chapter 5 are partisan measures, which appear to be a more dominant predictor of climate
change concern compared to ideology.
There is still evidence of support for Hypothesis 2. Those scoring higher on the faith-inscience index were more likely to identify with the arguments of climatologists. However, this
relationship is not significant when observing the total effect between the index and climate change
knowledge. There is a direct, significant effect, however, between the index and concern.
Compared to prior analyses, the strength of the direct effect is noticeably weaker – suggesting
some of the variance between the faith-in-science index and concern is explained by other
relationships specified in the model. As before, there continues to be no support for Hypothesis 3.
Those skeptical of science have no statistically significant relationship with the key dependent
variables in the analysis.
The direct effect between climate change knowledge and concern (Hypothesis 6) remains
strong. In terms of total effects, understanding climate change in a manner similar to climatologists
is substantively the most powerful predictor of climate change concern. Understanding climate
change, as expected, comes from one’s foundational understanding of science. Note, however, the
total effect of participating in a science college course. The standardized coefficient is nearly

185

double that of scientific knowledge. In the context of climate change concern, the total effect of
college science is double the total effect of science knowledge. In terms of the science
comprehension thesis, this observation speaks to the effects of formally socializing students toward
science. It appears in this analysis to be just as strong, if not stronger, of a predictor of beliefs than
scientific knowledge.
Other components of the science comprehension thesis, as specified in Figure C.1, do not
materialize. Those who seek out additional information, presumably to evaluate the arguments of
scientists, are not more likely to accept the causal arguments from climatologists. Given
observations in Appendix A, this lack of a relationship is not surprising. The measures of attention
do not appear to mediate the relationship between science knowledge and climate change
knowledge, likely because it is unknown which sources individuals are turning to for information.
In terms of the exogenous measures and their total effects, there is no detectable
relationship between age and climate change beliefs. However, the gender gap does emerge in the
model, with females less likely to accept the arguments from climatologists and demonstrate a
lower level of concern compared to males. Again, this is inconsistent with literature that suggests
females are more knowledgeable about climate change (McCright 2010).
Model Performance
The model itself performs fairly well when considering the portion of the variance in
climate change beliefs it can explain. The R2 statistic for climate change concern is 31%.
Compared to 5.2, the R2 statistic increases slightly for climate change concern.26 The other

26

With the survey weights applied to the model, there is a 10% improvement in the R2 statistic
for concern and a 20% improvement for climate change knowledge.

186

goodness-of-fit metrics, however, do not suggest the model explained a unique amount of variance.
That is, the model is not significantly different from a saturated model that simply connected all
the measures. The chi-squared assessment compares the difference between an over-identified and
just-identified model (with all lines connected). Results from the test suggest that the model
specification offered above is not significantly different than if all paths were drawn. That is, no
theoretical insights were gained by the researcher’s model. An alternative evaluation metric is the
Root Mean Square Error of Approximation (RMSEA), which accounts for parsimony in
calculating a goodness-of-fit statistic by controlling for sample size.27 The RMSEA for the
proposed model here is above 0.05, further indicating no unique insights were gained from the
model. Other fit statistics suggest the model is not acceptable and are reported in the Table C.2.
Table C.2: Further diagnostics
Goodness-of-fit metric
Likelihood ratio (model vs. saturated)
RMSEA
Comparative Fit Index28

Result
1791, reject unique difference
0.26, reject unique difference
0.43, reject unique difference

Standardized Root Mean Squared Residual29

0.13, reasonable

Summary
While the model performs poorly, there are several observations that can be utilized to
improve the model fit in future analyses. The model suggests mixed support for the science
comprehension thesis. With respect to the Hypotheses 1 and 6, the anticipated relationship between
science knowledge, climate change beliefs, and concern materialized. Those with elevated levels
27

Formally, the metric is calculated by taking the difference between the chi-squared value and
degrees of freedom, divided that by the degrees of freedom multiplied by sample size minus one.
28
This statistic compares noncentrality parameters for a baseline and fuller model. If there is little
difference, it can be seen as a good fit.
29
Values less than 0.1 are considered favorable. The calculation looks at the mean absolute value
of the covariance residuals.
187

of knowledge were capable of identifying the arguments from climatologists, which in turn led
them to perceive climate change as a greater threat. To be clear, again, the science comprehension
thesis does not suggest high-knowledge individuals should be more concerned: it simply specifies
that individuals can better understanding the problem. Whether high-knowledge individuals
perceive more of a threat from a changing climate depends on risk attitudes and other factors. It
appears from the relationship above, however, that these individuals are more likely to perceive a
threat.
For the science comprehension thesis, there is no support for the proposition that
respondents who understand climate change are also engaging in specific activities such as talking
about climate change and seeking out information. While Appendix A suggested it is the highknowledge individuals engaging in these information acquisition activities, they are not
necessarily reaching unique conclusions. Further insight might be gained if information about
whom these respondents are consulting for additional information might improve the analysis.
The model could also be significantly improved, perhaps, by specifying a different order
to the relationships. For instance, concern might lead individuals to consume information.
Alternatively, concern might lead individuals to know something about climate change in the first
place, regardless of one’s level of basic science knowledge. Alternative models – perhaps a nonrecursive model – could be specified and compared in order to parse out a superior framework.
The possibility of additional direct relationships could also be explored. There were
additional relationships that could have been drawn in the model, but were not because they were
outside the expected relationships of the hypotheses. Exploratory analyses might suggest
additional relationships that were not considered here.

188

Lastly, including additional variables might also provide further insight. The strong effect
observed for the partisan metrics in Chapter 5, as well as the strength of the college science
measure, suggests it might be appropriate to consider mediating effects. The Science News Survey
also has questions about trust in science organizations that theoretically could have a place in the
model.

189

BIBLIOGRAPHY

190

BIBLIOGRAPHY

Abramowitz, Alan I., and Kyle L. Saunders. 2008. "Is Polarization a Myth?" The Journal of
Politics 70 (02):542-55.
Abramson, Paul R., and Charles W. Ostrom, Jr. 1991. "Macropartisanship: An Empirical
Reassessment." The American Political Science Review 85 (1):181-92.
Achen, Christopher H. 1978. "Measuring Representation." American Journal of Political Science
22 (3):475-510.
———. 2002. "Parental Socialization and Rational Party Identification." Political Behavior 24
(2):151-70.
Alberts, Bruce. 2011. "The New Egypt." Science 332 (6029):513.
Allum, Nick, Patrick Sturgis, Dimitra Tabourazi, and Ian Brunton-Smith. 2008. "Science
knowledge and attitudes across cultures: a meta-analysis." Public Understanding of
Science 17:35-54.
Almond, Gabriel A. 1960. The American People and Foreign Policy. New York: Praeger.
American Association for the Advancement of Science (AAAS). 2010. "Climate change and the
public: Overcoming skepticism after Climategate."
Anderson, Norman H. 1981. Foundations of Information Integration Theory. San Diego:
Academic Press.
Ansolabehere, Stephen, James M. Snyder, Jr., and Charles Stewart, III. 2001. "Candidate
Positioning in U.S. House Elections." American Journal of Political Science 45 (1):13659.
Ansolabehere, Stephen., and David. Konisky. 2009. "Public Attitudes Toward Construction of
New Power Plants." Public Opinion Quarterly 73 (3):566-77.
Barker, David, and Kathleen Knight. 2000. "Political Talk Radio and Public Opinion." The Public
Opinion Quarterly 64 (2):149-70.
Bauer, Martin, John Durant, and Geoffrey Evans. 1994. "European Public Perceptions of Science."
International Journal of Public Opinion Research 6 (2):163-86.
Beck, Paul Allen, and M. Kent Jennings. 1982. "Pathways to Participation." The American
Political Science Review 76 (1):94-108.
Bodmer, Walter F. 1985. "The public understanding of science." London: Royal Society.

191

Borick, Christopher P., and Barry G. Rabe. 2010. "A Reason to Believe: Examining the Factors
that Determine Individual Views on Global Warming*." Social Science Quarterly 91
(3):777-800.
Brambor, Thomas, Williams R Clark, and Matt Golder. 2006. "Understanding Interaction Models:
Improving Empirical Analyses." Political Analysis 14:63-85.
Brody, Samuel D, and Sammy Zahran. 2007. "Commentary: Linking particulate matter and
sulphur concentrations to air pollution annoyance: problems of measurement, scale and
control." International Journal of Epidemiology 36 (4):820-3.
Brody, Samuel D., B. Mitchell Peck, and Wesley E. Highfield. 2004. "Examining Localized
Patterns of Air Quality Perception in Texas: A Spatial and Statistical Analysis." Risk
Analysis 24 (6):1561-74.
Brody, Samuel D., Sammy Zahran, Arnold Vedlitz, and Himanshu Grover. 2008. "Examining the
Relationship Between Physical Vulnerability and Public Perceptions of Global Climate
Change in the United States." Environment and Behavior 40 (1):72-95.
Brown, Marilyn A., and Benjamin K. Sovacool. 2011. Climate Change and Global Energy
Security: Technology and Policy Options. Cambridge, MA: The MIT Press.
Brulle, Robert J., Jason Carmichael, and J. Craig Jenkins. 2012. "Shifting public opinion on climate
change: an empirical assessment of factors influencing concern over climate change in the
U.S., 2002-2010." Climatic Change 114 (2):169-88.
Burstein, Paul. 2003. "The Impact of Public Opinion on Public Policy: A Review and an Agenda."
Political Research Quarterly 56 (1):29-40.
Carmines, Edward G., and James A. Stimson. 1986. "On the Structure and Sequence of Issue
Evolution." The American Political Science Review 80 (3):901-20.
Chang, Linchiat, and Jon A. Krosnick. 2009. "National Surveys Via Rdd Telephone Interviewing
Versus the Internet: Comparing Sample Representativeness and Response Quality." Public
Opinion Quarterly 73 (4):641-78.
Chong, Dennis, and James N. Druckman. 2007. "Framing Theory." Annual Review of Political
Science 10 (1):103-26.
Converse, Philip E. 1964. "The Nature of Belief Systems in Mass Publics." In Ideology and
Discontent, ed. D. E. Apter. Reprinted in Critical Review 19 (2006): 1-74.
———. 1972. "Change in the American Electorate." In The Human Meaning of Social Change,
ed. A. Campbell and P. E. Converse. New York: Russell Sage.
Cook, Summer B., Marvin Druger, and Lori L. Ploutz-Snyder. 2011. "Scientific literacy and
attitudes towards American space exploration among college undergraduates." Space
Policy 27 (1):48-52.
192

Critchley, Christine R. 2008. "Public Opinion and Trust in Scientists: The Role of the Research
Context, and the Perceived Motivation of Stem Cell Researchers." Public Understanding
of Science 17 (3):309-27.
Delli Carpini, Michael X., and Scott Keeter. 1996. What Americans Know about Politics and Why
It Matters New Haven: Yale University Press.
Doherty, David, and Jennifer Wolak. 2012. "When Do the Ends Justify the Means? Evaluating
Procedural Fairness." Political Behavior 34 (2):301-23.
Douglas, Mary. 1983. How Institutions Think. Syracuse, NY: Syracuse University Press.
Douglas, Mary, and Aaron Wildavsky. 1983. Risk and Culture: An Essay on the Selection of
Technical and Environmental Dangers. Berkeley, CA: University of California Press.
Druckman, James N. 2001. "On the Limits of Framing Effects: Who Can Frame?" The Journal of
Politics 63 (4):1041-66.
Druckman, James N., and Toby Bolsen. 2011. "Framing, Motivated Reasoning, and Opinions
About Emergent Technologies." Journal of Communication 61 (4):659-88.
Dunlap, Riley E., and Araon M. McCright. 2008. "A Widening Gap: Republican and Democratic
Views on Climate Change." Environment: Science and Policy for Sustainable
Development 50 (5):26 - 35.
Dunlap, Riley., Chenyang. Xiao, and Aaron McCright. 2001. "Politics and Environment in
America: Partisan and Ideological Cleavages in Public Support for Environmentalism "
Environmental Politics 10 (4):23-48.
Durant, John, Geoffrey A. Evans, and Geoffrey P. Thomas. 1989. "The public understanding of
science." Nature 340:11-4.
Egan, Patrick J., and Megan Mullin. 2012. "Turning Personal Experience into Political Attitudes:
The Effect of Local Weather on Americans' Peceptions about Global Warming." Journal
of Politics 74 (3):796-809.
EPA. 2012a. "2017 and Later Model Year Light-Duty Vehicle Greenhouse Gas Emissions and
Corporate Average Fuel Economy Standards; Final Rule."
———. 2012b. "National Emission Standards for Hazardous Air Pollutants From Coal-and OilFired Electric Utility Steam Generating Units and Standards of Performance for FossilFuel-Fired Electric Utility, Industrial-Commercial-Institutional, and Small IndustrialCommercial-Institutional Steam Generating Units ": Federal Registrar.
Evans, Geoffrey, and John Durant. 1995. "The relationship between knowledge and attitudes in
the public understanding of science in Britain." Public Understanding of Science 4 (1):5774.

193

Feldman, Lauren, Edward W. Maibach, Connie Roser-Renouf, and Anthony Leiserowitz. 2011.
"Climate on Cable: The Nature and Impact of Global Warming Coverage on Fox News,
CNN, and MSNBC." The International Journal of Press/Politics.
Fenno, Richard F., Jr. 1977. "U.S. House Members in Their Constituencies: An Exploration." The
American Political Science Review 71 (3):883-917.
Fiorina, Morris P., Samuel A. Abrams, and Jeremy C. Pope. 2005. Culture War? The Myth of a
Polarized America: Pearson Longman.
Gamkhar, Shama, and J. Mitchell Pickerill. 2012. "The State of American Federalism 2011-2012:
A Fend for Yourself and Activist Form of Bottom-Up Federalism." Publius-the Journal of
Federalism 42 (3):357-86.
Gauchat, Gordon. 2011. "The cultural authority of science: Public trust and acceptance of
organized science." Public Understanding of Science 20 (6):751-70.
Gelman, Andrew, and Gary King. 1993. "Why Are American Presidential Election Campaign
Polls so Variable When Votes Are so Predictable?" British Journal of Political Science 23
(4):409-51.
Giddens, Anthony. 1991. Modernity and Self-Identity: Self and Society in the Late Modern Age.
Stanford, CA: Stanford University Press.
Green, Donald., Bardley. Palmquist, and Eric. Schickler. 2002. Partisan Hearts & Minds. New
Haven & London: Yale University Press.
Griffin, John D. 2006. "Electoral Competition and Democratic Responsiveness: A Defense of the
Marginality Hypothesis." The Journal of Politics 68 (4):911-21.
Guber, Deborah Lynn. 2003. The Grassroots of a Green Revolution: Polling American on the
Environment. Cambridge, Massachusettes: The MIT Press.
Hamilton, Lawrence. 2011. "Education, politics and opinions about climate change evidence for
interaction effects." Climatic Change 104 (2):231-42.
Hamilton, Lawrence C. 2012. "Did the Arctic Ice Recover? Demographics of True and False
Climate Facts." Weather, Climate, and Society 4 (4):236-49.
Hamilton, Lawrence C., Chris R. Colocousis, and Cynthia M. Duncan. 2010. "Place Effects on
Environmental Views." Rural Sociology 75 (2):326-47.
Hamilton, Lawrence C., and Barry D. Keim. 2009. "Regional variation in perceptions about
climate change." International Journal of Climatology 29 (15):2348-52.
Hart, P. Sol, and Erik C. Nisbet. 2012. "Boomerang Effects in Science Communication: How
Motivated Reasoning and Identity Cues Amplify Opinion Polarization About Climate
Mitigation Policies." Communication Research 39 (6):701-23.
194

Hayes, Bernadette C. 2001. "Gender, Scientific Knowledge, and Attitudes toward the
Environment: A Cross-National Analysis." Political Research Quarterly 54 (3):657-71.
Hayes, Bernadette C., and Vicki N. Tariq. 2000. "Gender differences in scientific knowledge and
attitudes toward science: a comparative study of four Anglo-American nations." Public
Understanding of Science 9 (4):433-47.
Ho, Shirley, Dietram Scheufele, and Elizabeth Corley. 2010. "Making sense of policy choices:
understanding the roles of value predispositions, mass media, and cognitive processing in
public attitudes toward nanotechnology." Journal of Nanoparticle Research 12 (8):270315.
Inhofe, James. 2012. The Greatest Hoax: How the Global Warming Conspiracy Threatens Your
Future: WND Books.
IPCC. 2007. "Climate Change 2007: Syntehsis Report." ed. R. K. Pachauri and A. Reisinger:
Intergovernmental Panel on Climate Change.
Irwin, Alan, and Brian Wynne. 1996. Misunderstanding science? The public reconstruction of
science and technology: Cambridge University Press.
Jacquemin, Bénédicte, Jordi Sunyer, Bertil Forsberg, Thomas Götschi, Lucy Bayer-Oglesby,
Ursula Ackermann-Liebrich, Roberto de Marco, Joachim Heinrich, Deborah Jarvis, Kjell
Torén, and Nino Künzli. 2007. "Annoyance due to air pollution in Europe." International
Journal of Epidemiology 36 (4):809-20.
Jasanoff, Sheila. 1990. The Fifth Branch: Science Advisers as Policymakers. Cambridge,
Massachusetts: Harvard University Press.
Jennings, M. Kent. 1996. "Political Knowledge Over Time and Across Generations." The Public
Opinion Quarterly 60 (2):228-52.
Jennings, M. Kent, and Gregory B. Markus. 1984. "Partisan Orientations over the Long Haul:
Results from the Three-Wave Political Socialization Panel Study." The American Political
Science Review 78 (4):1000-18.
Jerit, Jennifer, Barabas Jason, and Bolsen Toby. 2006. "Citizens, Knowledge, and the Information
Environment." American Journal of Political Science 50 (2):266-82.
Jones, Bryan D., and Frank R. Baumgartner. 2005. The Politics of Attention: How Government
Prioritizes Problems: The University of Chicago Press.
Jones, Michael D. 2011. "Leading the Way to Compromise? Cultural Theory and Climate Change
Opinion." Ps-Political Science & Politics 44 (4):720-5.
Jones, Robert Emmet, and Riley E. Dunlap. 1992. "The Social Bases of Environmental Concern:
Have They Changed Over Time?1." Rural Sociology 57 (1):28-47.

195

Kahan, Dan M., Hank Jenkins-Smith, and Donald Braman. 2011. "Cultural cognition of scientific
consensus." Journal of Risk Research 14 (2):147-74.
Kahan, Dan M., Ellen Peters, Maggie Wittlin, Paul Slovic, Lisa Larrimore Ouellette, Donald
Braman, and Gregory Mandel. 2012a. "The polarizing impact of science literacy and
numeracy on perceived climate change risks." Nature Climate Change advance online
publication.
Kahan, Dan M., Ellen. Peters, Donald. Braman, Paul. Slovic, Maggie. Wittlin, Lisa Larrimore.
Ouellette, and Gregory. Mandel. 2012b. "The Tragedy of the Risk-Perception Commons:
Culture Conflict, Rationality Conflict, and Climate Change." In The 70th Annual
Conference of the Midwest Political Science Association. Chicago, Illinois.
Kahneman, Daniel, and Amos Tversky. 1982. "Judgment under uncertainty: Heuristics and
biases." In Judgment under Uncertainty: Heuristics and Biases. New York: Cambridge
University Press.
Kellstedt, Paul M., Sammy Zahran, and Arnold Vedlitz. 2008. "Personal Efficacy, the Information
Environment, and Attitudes toward Global Warming and Climate Change in the United
States." Risk Analysis 28 (1):113-26.
Kline, Rex B. 2005. Principles and Practice of Structural Equation Modeling: The Guilford Press.
———. 2013. Beyond Significance Testing: Statistics Reform in the Behavioral Sciences. Second
ed: American Psychological Association.
Kohut, Andrew, Scott Keeter, Carroll Doherty, and Michael Dimock. 2009. "Scientific
Achievements Less Prominent Than a Decade Ago: Public Praises Science; Scientists Fault
Public, Media." Washington, D.C. : The Pew Research Center.
Krosnick, Jon A., and Donald R. Kinder. 1990. "Altering the Foundations of Support for the
President Through Priming." The American Political Science Review 84 (2):497-512.
Kuklinski, James H., Daniel S. Metlay, and W. D. Kay. 1982. "Citizen Knowledge and Choices
on the Complex Issue of Nuclear Energy." American Journal of Political Science 26
(4):615-42.
Kunda, Z. 1990. "The Case for Motivated Reasoning." Psychological Bulletin 108 (3):480-98.
Kunda, Ziva. 1987. "Motivated Inference: Self-Serving Generation and Evaluation of Causal
Theories." Journal of Personality and Social Psychology 53 (4):636-.
Lau, Richard R., and Caroline. Heldman. 2009. "Self-Interest, Symbolic Attitudes, and Support
for Public Policy: A Multilevel Analysis." Political Psychology 30 (4):513-37.
Lawler, Andrew. 2011. "A New Day for Egyptian Science? ." Science 333 (6040):278-84.

196

Lax, Jeffrey R., and Justin H. Phillips. 2012. "The Democratic Deficit in the States." American
Journal of Political Science 56 (1):148-66.
Leiserowitz, Anthony. 2006. "Climate Change Risk Perception and Policy Preferences: The Role
of Affect, Imagery, and Values." Climatic Change 77 (1-2):45-72.
Leiserowitz, Anthony A. 2005. "American Risk Perceptions: Is Climate Change Dangerous?" Risk
Analysis 25 (6):1433-42.
Leiserowitz, Anthony A., Edward W. Maibach, Connie Roser-Renouf, Nicholas Smith, and Erica
Dawson. 2013. "Climategate, Public Opinion, and the Loss of Trust." American Behavioral
Scientist 57 (6):818-37.
Lewenstein, Bruce V. 1992. "The meaning of `public understanding of science' in the United States
after World War II." Public Understanding of Science 1 (1):45-68.
Liang, Jiaqi, and Daniel J. Fiorino. 2013. "The Implications of Policy Stability for Renewable
Energy Innovation in the United States, 1974–2009." Policy Studies Journal 41 (1):97-118.
Lindzen, Richard S. 2008. "Is the Global Warming Alarm Founded on Fact?" In Global Warming:
Looking Beyond Kyoto, ed. E. Zedillo: Center for the Study of Globalization, Yale
University.
Locke, Simon. 2002. "The Public Understanding of Science—A Rhetorical Invention." Science,
Technology & Human Values 27 (1):87-111.
Lorenzoni, Irene, Nick F. Pidgeon, and Robert E. O'Connor. 2005. "Dangerous Climate Change:
The Role for Risk Research." Risk Analysis 25 (6):1387-98.
Luskin, Robert C., and John G. Bullock. 2011. "‘‘Don’t Know’’ Means ‘‘Don’t Know’’: DK
Responses and the Public’s Level of Political Knowledge." Journal of Politics 73 (2):54757.
MacCoun, Robert J., and Susannah Paletz. 2009. "Citizens' Perceptions of Ideological Bias in
Research on Public Policy Controversies." Political Psychology 30 (1):43-65.
Malka, Ariel, Jon A. Krosnick, and Gary Langer. 2009. "The Association of Knowledge with
Concern About Global Warming: Trusted Information Sources Shape Public Thinking."
Risk Analysis 29 (5):633-47.
Marquart-Pyatt, Sandra T., Rachael L. Shwom, Thomas Dietz, Riley E. Dunlap, Stan A.
Kaplowitz, Aaron M. McCright, and Sammy Zahran. 2011. "Understanding Public
Opinion on Climate Change: A Call for Research." Environment: Science and Policy for
Sustainable Development 53 (4):38-42.
Matisoff, Daniel. 2008. "The Adoption of State Climate Change Policies and Renewable Portfolio
Standards: Regional Diffusion or Internal Determinants." The Review of Policy Research
25 (6):527-46.
197

Mayhew, David R. 2004. Congress: The Electoral Connection. Second ed: Yale University Press.
McCaffrey, Mark, and Joshua Rosenau. 2012. "Science literacy still matters." Nature Climate
Change 2 (9):636-.
McCright, Aaron. 2010. "The effects of gender on climate change knowledge and concern in the
American public." Population and Environment 32 (1):66-87.
———. 2011. "Political orientation moderates Americans’ beliefs and concern about climate
change." Climatic Change 104 (2):243-53.
McCright, Aaron, M. , and Riley E. Dunlap. 2000. "Challenging Global Warming as a Social
Problem: An Analysis of the Conservative Movement's Counter-Claims." Social Problems
47 (4):499-522.
McCright, Aaron M., and Riley E. Dunlap. 2011. "The Politicalization of Climate Change and
Polarization in the American Public's Views of Global Warming, 2001-2010." Sociological
Quarterly 52 (2):155-94.
Michaud, Kristy E. H., Juliet E. Carlisle, and Eric R.A.N. Smith. 2009. "The Relationship between
Cultural Values and Political Ideology, and the Role of Political Knowledge." Political
Psychology 30 (1):27-42.
Mielby, Henrik, Peter Sandøe, and Jesper Lassen. 2012. "The role of scientific knowledge in
shaping public attitudes to GM technologies." Public Understanding of Science.
Miller, Jon D. 1983. The American People and Science Policy: Pergamon Press.
———. 1998. "The measurement of civic scientific literacy." Public Understanding of Science 7
(3):203-23.
———. 2001. "Public Understanding of Science at the Crossroads." Public Understanding of
Science 10:115-20.
Miller, Jon D. 2004. "Public understanding of, and attitudes toward, scientific research: what we
know and what we need to know." Public Understanding of Science 13:273-94.
Mondak, Jeffrey J. 2001. "Developing Valid Knowledge Scales." American Journal of Political
Science 45 (1):224-38.
Mucciaroni, Gary, and Paul J. Quirk. 2006. Deliberative Choices: Debating Public Policy in
Congress. Chicago, IL: University of Chicago Press.
Myers, T. A., E. W. Maibach, C. Roser-Renouf, K. Akerlof, and A. A. Leiserowitz. 2013. "The
relationship between personal experience and belief in the reality of global warming."
Nature Climate Change 3 (4):343-7.

198

Nicholson, Stephen P. 2012. "Polarizing Cues." American Journal of Political Science 56 (1):5266.
Niemi, Richard G., and M. Kent Jennings. 1991. "Issues and Inheritance in the Formation of Party
Identification." American Journal of Political Science 35 (4):970-88.
Nir, Lilach. 2011. "Motivated Reasoning and Public Opinion Perception." Public Opinion
Quarterly 75 (3):504-32.
Nisbet, Matthew C., Dietram A. Scheufele, James Shanahan, Patricia Moy, Dominique Brossard, and
Bruce V. Lewenstein. 2002. "Knowledge, Reservations, or Promise?: A Media Effects Model for
Public Perceptions of Science and Technology." Communication Research 29 (5):584-608.

Nyhan, Brendan, and Jason Reifler. 2010. "When Corrections Fail: The Persistence of Political
Misperceptions." Political Behavior 32 (2):303-30.
Oreskes, Naomi. 2004. "The Scientific Consensus on Climate Change " Science 306 (5702):1686.
Page, Benjamin I., and Robert Y. Shapiro. 1983. "Effects of Public Opinion on Policy." The
American Political Science Review 77 (1):175-90.
Pidgeon, Nick, and Catherine Butler. 2009. "Risk analysis and climate change." Environmental
Politics 18 (5):670-88.
Pielke, Robert A. 2010. "The Climate Fix: What Scientists and Politicians Won't Tell You About
Global Warming." Basic Books.
Popkin, Samuel L., and Michael A. Dimock. 1999. "Political Knowledge and Citizen
Competence." In Citizen Competence and Democratic Institutions, ed. S. L. Elkin and K.
E. Soltan. University Park: Pennsylvania State University Press.
Pralle, Sarah, and Jessica Boscarino. 2011. "Framing Trade-offs: The Politics of Nuclear Power
and Wind Energy in the Age of Global Climate Change." Review of Policy Research 28
(4):323-46.
Prewitt, Kenneth. 1983. "Scientific Illiteracy and Democratic Theory." Daedalus 112 (2):49-64.
Price, Vincent, and John Zaller. 1993. "Who Gets the News? Alternative Measures of News
Reception and Their Implications for Research." The Public Opinion Quarterly 57 (2):13364.
Putnam, Robert D. 2000. Bowling Alone: Simon & Schuster Paperbacks.
Rabe, Barry G. 2010. "Racing to the Top, the Bottom, or the Middle of the Pack? The Evolving
State Government Role in Environmental Protection." In Environmental Policy: New
Directions for the Twenty-First Century, ed. N. J. Vig and M. E. Kraft. Washington, D.C.:
CQ Press.

199

Reardon, Sara. 2011. "Climate Change Sparks Battles in Classroom " Science 333 (6043):688-9.
Ripberger, Joseph T., Geoboo Song, Matthew C. Nowlin, Michael D. Jones, and Hank C. JenkinsSmith. 2012. "Reconsidering the Relationship Between Cultural Theory, Political
Ideology, and Political Knowledge." Social Science Quarterly 93 (3):713-31.
Roberts, Dave. 2012. What ‘left’ and ‘right’ really mean on climate change [cited Dec. 3 2012].
Available
from
http://grist.org/climate-change/what-left-and-right-really-mean-onclimate-change-hint-nothing/.
Sabatier, Paul A., and Hank C. Jenkins-Smith, eds. 1993. Policy Change and Learning: An
Advocacy Coalition Approach: Westview Press.
Scharwz, Michiel, and Michael Thompson. 1990. Divided We Stand: Redefining Politics,
Technology, and Social Choice. Philadelphia, PA: University of Pennsylvania Press.
Schattschneider, E. E. 1960. The Semisovereign People: A Realist's View of Democracy in
America. 1975 ed. Hindale, Illinois: The Dryden Press.
Schneider, Stephen H. 2010. Science as a Contact Sport: Inside the Battle to Save Earth's Climate.
Washington, D.C. : National Geographic.
Schuldt, Jonathon P., Sara H. Konrath, and Norbert Schwarz. 2011. ""Global warming" or "climate
change"?" Public Opinion Quarterly 75 (1):115-24.
Searle, Kristina, and Kathryn Gow. 2010. "Do concerns about climate change lead to distress?"
International Journal of Climate Change Strategies and Management 2 (4):362-79.
Sears, David O. 1975. "Political Socialization." In Handbook of Political Science, ed. F. I.
Greenstein and N. Polsby. Reading, MA: Addison-Wesley.
Sears, David O., and Carolyn. Funk. 1990. "Self-Interest in Americans' Political Opinions." In
Beyond Self-Interest, ed. J. Mansbridge.
Shaviv, Nir. 2006. "On Climate Sensitivity and why it is probably small." In ScienceBits.
Spence, Alexa, Wouter Poortinga, and Nick Pidgeon. 2012. "The Psychological Distance of
Climate Change." Risk Analysis 32 (6):957-72.
Stone, Deborah A. 2002. Policy Paradox: The Art of Political Decision Making. Revised ed. New
York: W.W. Norton & Company.
Stratmann, Thomas. 2000. "Congressional Voting over Legislative Careers: Shifting Positions and
Changing Constraints." The American Political Science Review 94 (3):665-76.
Sturgis, Patrick, and Nick Allum. 2004. "Science in Society: Re-Evaluating the Deficit Model of
Public Attitudes." Public Understanding of Science 13:55-74.

200

Sturgis, Patrick, Helen Cooper, and Chris Fife-Schaw. 2005. "Attitudes to biotechnology:
estimating the opinions of a better-informed public." New Genetics and Society 24 (1):3156.
Sturgis, Patrick J., and Nick Allum. 2001. "Gender differences in scientific knowledge and
attitudes toward science: Reply to Hayes and Tariq." Public Understanding of Science 10
(4):427-30.
Swedlow, Brendon. 2012. "Cultural Coproduction of Four States of Knowledge." Science,
Technology & Human Values 37 (3):151-79.
Taber, Charles S., and Milton Lodge. 2006. "Motivated Skepticism in the Evaluation of Political
Beliefs." American Journal of Political Science 50 (3):755-69.
Thompson, Michael, Richard Ellis, and Aaron Wildavsky. 1990. Cultural Theory. Boulder, CO:
Westview Press.
Trefil, James. 2008. Why Science?: Teachers College Press & National Science Teachers
Association.
Wood, B. Dan, and Arnold Vedlitz. 2007. "Issue Definition, Information Processing, and the
Politics of Global Warming." American Journal of Political Science 51 (3):552-68.
Wright, Gerald C., Jr., Robert S. Erikson, and John P. McIver. 1987. "Public Opinion and Policy
Liberalism in the American States." American Journal of Political Science 31 (4):9801001.
Wynne, Brian. 1991. "Knowledges in Context." Science, Technology & Human Values 16 (1):11121.
Yearley, Steven. 2000. "Making systematic sense of public discontents with expert knowledge:
two analytical approaches and a case study." Public Understanding of Science 9 (2):10522.
Zaller, John, and Stanley Feldman. 1992. "A Simple Theory of the Survey Response: Answering
Questions versus Revealing Preferences." American Journal of Political Science 36
(3):579-616.
Zaller, John R. 1992. The Nature of Origins of Mass Opinion: Cambridge University Press.
Ziman, John. 1991. "Public Understanding of Science." Science, Technology & Human Values 16
(1):99-105.
Zumbrunnen, John, and Amy Gangl. 2008. "Conflict, Fusion, or Coexistence? The Complexity of
Contemporary American Conservatism." Political Behavior 30 (2):199-221.

201