Letter | Published:

The enduring effect of scientific interest on trust in climate scientists in the United States

Nature Climate Changevolume 8pages485488 (2018) | Download Citation


People who distrust scientists are more likely to reject scientific consensus, and are more likely to support politicians who are sceptical of scientific research1. Consequently, boosting Americans’ trust in scientists is a central goal of science communication2. However, while previous research has identified several correlates of distrust in climate scientists3 and scientists more broadly4, far less is known about potential long-term influences taking root in young adulthood. This omission is notable, as previous research suggests that attitudes towards science formulated in pre-teenage years play a key role in shaping attitudes in adulthood5. Using data from the Longitudinal Study of American Youth, I find that interest in science at age 12–14 years is associated with increased trust in climate scientists in adulthood (mid thirties), irrespective of Americans’ political ideology. The enduring and bipartisan effects of scientific interest at young ages suggest a potential direction for future efforts to boost mass trust in climate scientists.


Climate scientists are amongst the least trusted scientific experts in the United States; especially on the ideological right6,7. Recent public opinion research finds that only about one-third of Americans think that climate scientists’ research represents the best available scientific evidence. Similar numbers of Americans believe that climate scientists are mostly influenced by career advancement or their political leanings8.

This lack of trust in climate scientists has important environmental, social and political consequences. Today, less than half of American adults accept that climate change results from human activity, a trend that can be traced in part to Americans’ general distrust in scientists1,8. Americans’ distrust in the scientific community has also been shown to increase support for politicians that are hostile to scientific research1.

Consequently, scientists care about improving public trust in themselves and their research2. Understanding the origins of public trust in climate scientists is therefore an important question. If researchers can identify when and why people distrust scientists, it may be possible to design interventions and communication strategies that improve public trust.

Young adulthood may prove to be a critical period in the development of adult Americans’ attitudes towards climate scientists. Previous longitudinal work has found that positive affect towards science (as a discipline) tends to take shape when people are 12–14 years old5. According to this research, these ages are critical because they follow students’ first experiences with the study of scientific subjects in school, which play a major role in the development of attitudes towards science.

However, whether or not this reasoning applies to attitudes towards climate scientists specifically is uncertain. Although the development of attitudes towards climate scientists is an understudied topic in the literature, general positivity towards science is both conceptually9 and empirically10,11 associated with positive attitudes towards scientists. Thus, there is reason to suspect that attitudes towards climate scientists in adulthood result from factors taking root at young ages.

I hypothesize that two sets of factors in young adulthood may shape trust in climate scientists in adulthood. The first is young adults’ general interest in science. Pre-teens who are more interested in science should be more likely to trust climate scientists as adults. For example, previous work suggests that science curiosity—a specific form of scientific interest, focused on the enjoyment of scientific media and science fiction—leads people to consume information about scientific research in an open-minded way12. While it has not been linked to climate scientist attitudes directly, science curiosity is associated with acceptance of scientific consensus on anthropogenic climate change13.

Second, young adults with high levels of science comprehension may also be more likely to trust climate scientists as adults. Being able to ‘think the way scientists think’ has been theorized to boost support for scientific consensus, although this is not always borne out empirically14. Quantitative reasoning skills15 and knowledge of basic scientific facts16,17 have both been studied as measures of science comprehension. Neither has been shown to correlate with attitudes towards climate scientists specifically, but both have been linked to increased belief in anthropogenic climate change. However, unlike scientific interest, these effects may be conditional on adults’ prior political preferences; such that ideological conservatives who ‘think like scientists’ are more likely to reject scientific consensus14,16,17,18.

To address these questions, I analysed two waves of survey data on Americans’ trust in climate scientists from the Longitudinal Study of American Youth (LSAY)19; a large and representative survey of young adults in the United States (N = 3,116). As I described above, prior research points to quantitative ability, scientific knowledge, and interest in science as having the potential to increase trust in climate scientists. LSAY contained the data necessary to test whether these three factors at ages 12–14 (measured in 1987) were associated with increased trust in climate scientists in adulthood (measured in 2011).

I tested this in the LSAY data by modelling trust in climate scientists (in adulthood) as a function of scientific interest and the two science comprehension variables (quantitative ability and science knowledge) in junior high school. These models controlled for several known correlates of attitudes towards scientists measured in adulthood; including political ideology and religiosity3,4. I estimated the effect of scientific interest, controlling for an average of two comprehension variables (as the two are highly correlated)20, and modelled each comprehension variable separately (controlling for scientific interest). More information about how these models were constructed can be found in the Methods.

Adulthood trust in climate scientists was measured using four items; all were 11-point scales recoded to range from zero (low trust) to one (high trust). They included measures of how much Americans trust information about ‘global climate change’ from the National Aeronautics and Space Administration/National Oceanic and Atmospheric Administration (NASA/NOAA) (Mean M = 0.65, s.d. = 0.26), state environmental agencies (M = 0.57, s.d. = 0.26), State environmental departments (M = 0.61, s.d. = 0.25) and the International Panel on Climate Change (IPCC; M = 0.56, s.d. = 0.29). To guard against random measurement error21, and to provide a more-general measure of trust, I average all four into a summary index (α = 0.92). Additional information about these variables can be found in the Methods and Supplementary Fig. 1.

Quantitative skills (M = 0.40, s.d. = 0.17) and science knowledge (M = 0.39, s.d. = 0.39) in young adulthood were measured using short standardized tests (see Methods for sample items and links to questionnaires). Finally, scientific interest in young adulthood was measured as respondents’ self-reported interest in ‘scientific issues’ on a three-point scale (M = 0.43, s.d. = 0.33). These too were scaled to range from zero to one.

Figure 1 displays the predicted effect of scientific interest and science comprehension at ages 12–14 on trust in climate scientists in adulthood. Scientific interest was positively and significantly (at the P ≤ 0.05 level, two-tailed) associated with increased trust for all five outcomes. Specifically, it was associated with a 5% increase in trust in science professors (P = 0.05), an 8% increase for state environmental departments (P = 0.02), a 7% increase for NASA/NOAA (P = 0.02), an 8% increase for the IPPC (P = 0.02) and a 6% increase for the overall index (P = 0.02).

Fig. 1: The effect of scientific interest, quantitative ability and science knowledge (ages 12–14) on adulthood trust in climate scientists.
Fig. 1

Scientific interest in young adulthood is a positive and statistically significant predictor of trust in climate scientists in adulthood in all five cases. The change in trust for each outcome is determined by the movement from the minimum to maximum value of the interest and comprehension variables. Error bars represent 95% confidence intervals from each point. Estimates are derived from Supplementary Table 1.

Quantitative skills and scientific knowledge exhibited a less consistent pattern of effects. At times, they were even negatively associated with trust in climate scientists. Only one comprehension measure (quantitative skills) was both positively and significantly associated with trust, and only in one case; trust in science professors (a 9% increase, P = 0.02).

Figure 1 demonstrates that scientific interest in young adulthood is associated with increased trust in climate scientists. However, it does not provide a sense of how these effects compare to those of scientific interest in adulthood. I re-estimated the models displayed in Fig. 1, including both lagged and contemporaneous scientific interest measures. Figure 2 decomposes the cumulative predicted effect of scientific interest on trust in climate scientists into two parts; young adulthood effects (black bars) and adulthood effects (white bars).

Fig. 2: Decomposition of the cumulative effect of scientific interest on trust in climate scientists.
Fig. 2

A substantial portion of the cumulative effect of scientific interest on climate scientist trust can be attributed to scientific interest in young adulthood. Bars represent the percentage increase in trust attributable to movement from the minimum to maximum value of scientific interest at each age (black bars correspond to young adulthood; white bars to adulthood). Estimates derived from Supplementary Table 2.

The results show that the cumulative effect of scientific interest on trust in climate scientists is substantively large, ranging from 12 to 17%. Young adulthood effects were associated with 4–7% total increases in trust. Specifically, scientific interest in young adulthood contributed to 29% of the total effect of interest on trust in science professors (P = 0.09), 58% for state environmental departments (P = 0.03), 40% for NASA/NOAA (P = 0.07), 41% for the IPCC (P = 0.06), and 36% (P = 0.06) for the index.

Scientific interest in young adulthood is associated with increased trust in climate scientists and explains a substantial portion of the cumulative effect of interest on trust. Still, these results would require a major caveat if they could only be attributed to adults on one side of the partisan aisle. As reviewed earlier, the effects of science comprehension—but not scientific interest—on the acceptance of scientific consensus has been shown to be sensitive to partisan and cultural preferences. Whether or not this pattern holds with respect to attitudes towards climate scientists is an open question.

To test this, I re-ran all models presented in Fig. 1 by interacting the interest and comprehension measures (in young adulthood) with respondents’ self-reported political ideology (a ten-point scale ranging from 0 (liberal) to 1 (conservative)) in adulthood. The results are reported in Table 1. Scientific interest produced no significant interactions with political ideology, suggesting that insights from prior research can probably be extended to climate scientist attitudes.

Table 1 Conditional effects of scientific interest, quantitative skills, and science knowledge on climate scientist trust (by political ideology)

I also find evidence consistent with the idea that the effects of science comprehension on trust are conditional on political ideology. Science knowledge produced negative and statistically significant (P ≤ 0.05, two-tailed) interactions in all models, indicating that the effect of knowledge on increased trust in climate scientists was conditional on being politically liberal, relative to conservative. The models for quantitative ability were also conditioned by ideology, although the results were less consistent. Although additional research is necessary, this finding is consistent with the idea that ability-related traits tend to be associated with biased reasoning in various scientific domains, while motivational traits (that is, curiosity and interest) tend to have de-biasing effects13,14,15.

However, these results speak to only one measure of Americans’ ideological preferences; their symbolic identification with liberal and conservative labels. Measures such as this are commonly used in the literature, and have been shown to correlate with attitudes about science and trust in scientists1,3,5,6. Still, whether or not these results hold using alternate measures of ideology (for example, Americans’ stances on policy issues) is an interesting avenue for future research.

Together, the results offer three important conclusions for efforts to improve the public’s trust in climate scientists. First, scientific interest at young ages is associated with increased trust in climate scientists; more so than measures of science comprehension such as quantitative skills or science knowledge. Second, the effects of interest at young ages contribute to a substantively large portion of the total effect of interest in science on trust in climate scientists. Finally, the effects of scientific interest occur independently of political ideology.

Some might wonder about the extent to which the outcome variables studied here are sensitive to familiarity with different types of climate scientists. For example, respondents may be less familiar with the IPCC than state environmental agencies, science professors, or NASA/NOAA. Although I find no relationship between science knowledge and these outcomes (Fig. 1), I also assess the psychometric properties of these variables using graded response modelling (an application of item response theory). All four outcome variables track strongly—both substantively and statistically—onto latent trust (Supplementary Fig. 2; Supplementary Table 5). I also recover all results presented above when replacing the index used in the main text with the derived latent trust measure (Supplementary Table 6).

The results also suggest several avenues for future research. For example, while scientific interest is associated with increased trust in climate scientists, the precise mechanism by which this trust goes on to shape the acceptance of scientific consensus is an open question. Recent research suggests, for example, that the relationship between trust and acceptance may be mutually reinforcing22. Future research should consider this dynamic when investigating the downstream consequences of scientific interest and trust.

Second, the results presented here are based on data from a cohort of individuals born in the mid-to-late 1970s. Because younger Americans tend to be more accepting of climate science (in general) than older generations23, some might question if the pattern of effects observed here holds for Millennials and Generation Z. Recent research has found little evidence of generational differences in attitudes towards climate scientists specifically8, leading us to suspect that the pattern of results observed here would not be substantially different for younger generations. However, additional research in this area is necessary.

Finally, and perhaps most importantly, these results suggest a potential path forward for boosting Americans’ trust in climate scientists. Because young adults’ interest in science has long-lasting effects on trust in climate scientists, across the ideological spectrum, efforts to get pre-teens interested in scientific topics may be an effective trust-building strategy.

For example, the attention-grabbing and engaging content24 of board and video games offers an intriguing medium by which this goal might be accomplished. Attempts to educate children and young adults about scientific concepts through games have become increasingly common in recent years24,25. Of course, many of these games may not be expressly designed to get young adults interested in science; they may have other goals (for example, education) or targets (younger children, adults). More broadly, however, our research makes the case that efforts to boost pre-teen scientific interest may be effective at boosting trust in climate scientists later in life, and warrant future empirical investigation.



Data for this study come from the publicly available LSAY, which sampled middle and high school students (total NSSU = 5,945) from schools (NPSU = 109) across the United States (stratified by region and urban development). Data collection began in 1987, and has followed up with students several times in adulthood; most recently in 2011.


Respondents were sampled in one of two cohorts, one beginning at grade 7, and a second beginning at grade 9. To create the longest lag possible between young adulthood and adulthood, I look only at the youngest cohort (N = 3,116) and their responses in 2011 (N = 1,524). I weight all data to known population benchmarks and adjust for non-response (using the WEIGHTV variable available in the LSAY data).

Outcome variables

The four outcome variables presented in this letter are part of a larger battery of questions about who adults trust to provide them with accurate information from several sources about ‘global climate change’. Responses were recorded on a 0 to 10 scale, where 0 indicates low levels of trust and 10 represents high levels of trust. I selected only those questions pertaining to climate scientists. A full list (with unabridged question wording) can be found on the LSAY website (http://www.lsay.org/instruments/LSAY%202011%20paper%20v1#page= 12.pdf) or by consulting Supplementary Fig. 1.

I considered an item to pertain to climate scientists if the question’s subject is either a reference to a climate scientist, or an organization that necessarily produces (or disseminates) information from climate scientists. I identified four variables that satisfied this criterion. In order to ensure that results are consistent across each operationalization of the broader concept (trust in climate scientists), I ran separate models for each outcome variable. To reduce the possibility of random measurement error, and to provide a more-general measure of trust, I also re-ran all analyses using an index of the four (α = 0.92).

Outcome variables consist of the following four items. First, respondents were asked whether or not they would trust information about global climate change from ‘a NASA or NOAA website’. This is a clear example of a question pertaining to climate scientists. News features on the NOAA climate change website (climate.gov) summarize recent scientific research from climate scientists, and are peer-reviewed by experts in the field (for example, see this recent piece26). For the same reason, I also consider reports from the IPCC (the second outcome variable) and state environmental agencies (the third outcome variable) to satisfy this criterion. Both summarize, study and/or report evidence from organizations that are comprised of climate scientists.

Finally, I consider an interview with ‘a science professor at a university in [the respondent’s] state’ to be a reference to climate scientists. Although this question uses a more-general terminology (‘science professor’) than the other items, I think it is reasonable to assume that an interview with an expert on climate matters would probably come from a field related to the climate and environmental science more generally. Still, I urge caution with interpreting results from this particular model, and encourage turning to the broader trust in climate scientists index for more general assessments.

There are also several information sources that do not satisfy this criterion. For example, I do not consider Wikipedia articles or stories on local/national news outlets (Internet, television, and/or radio) to be references to climate scientists. While climate scientists may be quoted in the news, or have their research cited in a Wikipedia entry, they are not necessarily featured in these sources, and therefore do not qualify.

Independent variables

Independent variables were operationalized as follows, all scored to range from 0 to 1. First, I measure scientific interest using an item administered throughout the middle and high school waves, asking respondents to report their levels of interest in ‘science issues’ on a three-point scale ranging from ‘not at all interested’ to ‘very interested’. Because this is a three-point ordinal measure, I address concerns of low response option variability by averaging seventh and eighth grade scores. The resulting measure is a five-point scale ranging from those who exhibited low levels of interest in both seventh and eighth grade, to those highly interested at both time points.

Second, I measured quantitative ability using scores on a short multiple-choice mathematics test embedded in every childhood wave of the LSAY study. The skills that students were asked to demonstrate on these tests were fairly general in content; focusing not on discrete facts (for example, the regurgitation of formulae), but on basic algebraic, geometric and statistical reasoning. For example, students might be asked to calculate the probability of drawing a particular type of coin from one’s pocket, or to calculate distances between lines in a schematic. Pursuant with LSAY’s recommendations, missing data on the tests were imputed on basic demographic factors and prior test performance. These variables were calculated by LSAY’s principal investigators in the study’s public data release. Sample questionnaires can be found on the LSAY website (http://www.lsay.org/resources_questionnaires.html).

Likewise, scientific knowledge consists of respondents’ scores on short multiple-choice tests. Unlike the quantitative and verbal tests, which attempted to assess students’ more-general abilities in each domain, the scientific knowledge tests primarily focused on the knowledge of declarative scientific facts (for example, the types of compounds emitted by smokestacks, explaining curvature in plate tectonics). Again, missing data were imputed.


In all multivariate models, I also account for factors known to exert contemporaneous influences on anti-science attitudes, including respondents’ political ideology (coded such that movement from 0 to 1 reflects an increase in conservatism) and frequency of religious service attendance (religiosity). Models also control for respondents’ gender (1 if female, 0 if male), race (indicators of whether or not respondents are Black or Hispanic), educational attainment (measured in adulthood), and parental educational attainment.

Model procedure

The basic procedure for replicating the estimates found in Fig. 1 is as follows. Because students were cluster-sampled from 109 schools, it is important to account for the possibility-correlated error variance at the school level. To do this, I structure all models hierarchically; estimating school-level random effects. I do this using the xtmixed command in Stata 13, applying sampling weights to the fixed effects portion of the model.

I model each outcome measuring trust in climate scientists as a function of the set of young adulthood factors hypothesized to influence these attitudes as well as known contemporaneous influences; controlling for standard demographics. I do this by running three sets of five models; one set (Fig. 1, column 1) models trust as a function of scientific interest, controlling for respondents’ average scores on the science knowledge and quantitative ability variables combined; a second set (Fig. 1, column 2) looks just at quantitative skills, controlling for scientific interest; and a third (Fig. 1, column 3) does this for just scientific knowledge, controlling for scientific interest.

I chose this approach, because science knowledge and quantitative skills are fairly well correlated with one another (ρ = 0.71). This raises potential modelling concerns. Including two highly correlated terms in the same model poses collinearity issues, and may decrease the likelihood of finding evidence in favour of a link between quantitative ability or science knowledge on trust. I include a visual summary of these specification differences in Supplementary Table 4. This process creates fifteen different models, each one contributing a different ‘point’ to Fig. 1. Full results for the models used to produce Fig. 1 can be found in Supplementary Table 1.

The results presented in Fig. 2 amend this equation to include a term for contemporaneous scientific interest, measured using the same items administered in childhood, in 2011. For consistency with how interest was measured in childhood, I average together both available measures in adulthood (in 2008 and 2011). Because these models introduce two measures of the same construct (at different points in time), I simplify the models by dropping the averaged quantitative and science knowledge term (a move further justified by the variables’ weak predictive performance in Fig. 1). All models used to produce Fig. 2 can be found in Supplementary Table 2.

The results presented in Table 1 again alter this basic equation. This time, however, they include an interaction between the contemporaneous measure of self-reported ideology and each of the interest and comprehension variables.

Item response theory robustness checks

Finally, the procedure for replicating the item response theory robustness checks referenced in the main text is as follows. First, I estimate a graded response model using the irt grm command available for Stata 14; based on responses to the science professor, state environmental department, NASA/NOAA and IPCC outcome variables. Based on these results, I then calculate each item’s respective category characteristic curves, estimated with respect to offering high-trust answers on each of the four items (that is, earning a score of 10 on the scale) as well as offering low-trust answers (that is, earning a score of 0 on the scale). These analyses can be found in Supplementary Fig. 2, and can be reproduced using the irtgraph command.

The discrimination parameter analyses, presented in Supplementary Table 5, are also calculated using the irt grm command. Entries in the table correspond to the discrimination parameter estimates and their corresponding 95% confidence intervals produced via the graded response model.

Last, the robustness check summary presented in Supplementary Table 6 uses the predict option and latent sub-option available for the irt grm command to extract an estimate of latent trust; using an empirical Bayes estimator. For consistency with the main text, I recode the resulting variable to range from 0 to 1 (such that 0 indicates a low level of latent trust, and 1 indicates a high level). I then repeat all procedures described above in the Methods section to re-estimate the parameter estimates and two-tailed P values originally presented in Figs. 1, 2, and Table 1. Results were considered to be robust if they retained statistical significance and correct coefficient direction, unless the original result failed to reach conventional standards of significance. In that case, a test was considered robust if it retained statistical insignificance.

Ethics statement

This research complies with all relevant ethical regulations. Informed consent for human subjects was originally obtained from all participants by the providers of the external data used in this study.

Data availability

The longitudinal public opinion data that support the findings of this study are available at the Inter-university Consortium for Political and Social Research at the University of Michigan (ref. 19), with the identifier: https://doi.org/10.3886/ICPSR30263.v6 . Syntax (using Stata 13) to re-create tables and figures is available in the Supplementary Information.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.


  1. 1.

    Motta, M. The dynamics and political implications of anti-intellectualism in the United States. Am. Politics Res. https://doi.org/10.1177/1532673X17719507 (2017).

  2. 2.

    Besley, J. & Dudo, A. in Oxford Research Encyclopedia of Climate Science 10.1093/acrefore/9780190228620.013.380 (Oxford Univ. Press, Oxford, 2017).

  3. 3.

    Gordon, G. Politicization of science in the public sphere: A study of public trust in the United States, 1974 to 2010. Am. Sociol. Rev. 77, 167–187 (2012).

  4. 4.

    Gauchat, G., O'Brien, T. & Mirosa, O. The legitimacy of environmental scientists in the public sphere. Climatic Change 143, 297–306 (2017).

  5. 5.

    Simpson, R. D. & Oliver, J. S. A summary of major influences on attitude toward and achievement in science among adolescent students. Sci. Educ. 74, 1–18 (1990).

  6. 6.

    Blank, J. M. & Shaw, D. Does partisanship shape attitudes toward science and public policy? The case for ideology and religion. Ann. Am. Acad. Pol. Soc. Sci. 658, 18–35 (2015).

  7. 7.

    McCright, A. M., Dentzman, K., Charters, M. & Dietz, T. The influence of political ideology on trust in science. Environ. Res. Lett. 8, 044029 (2013).

  8. 8.

    Funk, C. & Kennedy, B. The Politics of Climate (Pew Research Center, 2016).

  9. 9.

    Klopfer, L. E. in Handbook on Formative and Summative Evaluation of Student Learning (eds Bloom, B. S. et al) 559–641 (McGraw-Hill, London, 1971).

  10. 10.

    Osborne, J., Shirley, S. & Collins, S. Attitudes towards science: A review of the literature and its implications. Int. J. Sci. Educ. 25, 1049–1079 (2003).

  11. 11.

    Gardner, P.L. Attitudes to science: A review. Stud. Sci. Educ. 2, 1–41 (1975).

  12. 12.

    Kahan, D. M., Carpenter K. & Landrum, A. Evidence-Based Science Filmmaking Initiative: Study No. 1 (The Cultural Cognition Project at Yale Law School, New Haven, 2015); https://doi.org/10.2139/ssrn.2713563

  13. 13.

    Kahan, D. M., Landrum, A., Carpenter, K., Helft, L. & Jamieson, K. H. Science curiosity and political information processing. Political Psychol. 38, 179–199 (2017).

  14. 14.

    Kahan, D. M. & et al The polarizing impact of science literacy and numeracy on perceived climate change risks. Nat. Clim. Change 2, 732–735 2012).

  15. 15.

    Kahan, D. M., Peters, E., Cantrell Dawson, E. & Slovic, P. Motivated numeracy and enlightened self-government. Behav. Public Poliÿcy 1, 54–86 (2017).

  16. 16.

    Gauchat, G. W. A test of three theories of anti-science attitudes. Sociol. Focus 41, 337–357 (2008).

  17. 17.

    Sturgis, P. & Allum, N. Science in society: Re-evaluating the deficit model of public attitudes. Public Underst. Sci. 13, 55–74 (2004).

  18. 18.

    Lewandowsky, S. & Oberauer, K. Motivated rejection of science. Curr. Dir. Psychol. Sci. 25, 217–222 (2016).

  19. 19.

    Miller, J. D. Longitudinal Study of American Youth, 1987-1994, 2007-2011 ICPSR30263 (Inter-university Consortium for Political and Social Research, 2016); https://doi.org/10.3886/ICPSR30263.v6

  20. 20.

    Kahan, D. M. Ordinary science intelligence: A science-comprehension measure for study of risk and science communication, with notes on evolution and climate change. J. Risk Res. 20, 995–1016 (2017).

  21. 21.

    Ansolabehere, S., Rodden, J. & Snyder, J. M. The strength of issues: Using multiple measures to gauge preference stability, ideological constraint, and issue voting. Am. Political Sci. Rev. 102, 215–232 (2008).

  22. 22.

    Landrum, A. R., Eaves, B. S. & Shafto, P. Learning to trust and trusting to learn: A theoretical framework. Trends Cogn. Sci. 19, 109–111 (2015).

  23. 23.

    Nuccitelli, D. The climate change generation gap. Bulletin of the Atomic Scientists (21 April 2016); https://thebulletin.org/climate-change-generation-gap9351

  24. 24.

    Wu, J. S. & Lee, J. J. Climate change games as tools for education and engagement. Nat. Clim. Change 5, 413–418 (2015).

  25. 25.

    Kwok, R. Enterprise: Game on. Nature 547, 369–371 (2017).

  26. 26.

    Lindsey, R. Record low sea ice in Alaska waters in fall 2017. ClimateWatch Magazine (21 November 2017); https://www.climate.gov/news-features/featured-images/record-low-sea-ice-alaska-waters-fall-2017

Download references


Funding for this research was provided by the National Science Foundation (grant no. 00039202).

Author information


  1. Department of Political Science, University of Minnesota, Minneapolis, MN, USA

    • Matthew Motta


  1. Search for Matthew Motta in:


M.M. contributed to all aspects of this paper, including statistical analysis, writing and revisions.

Competing interests

The author declares no competing interests.

Corresponding author

Correspondence to Matthew Motta.

Supplementary information

  1. Supplementary Information

    Supplementary Tables 1–6 and Figures 1 & 2

  2. Replication Syntax

    This DO file (syntax file for Stata 13) contains all syntax necessary to reproduce the tables and figures presented in both the letter and Supplementary Information. File paths have been removed, so the user must point to locations on their own drive. As a reminder, the data necessary to reproduce these findings can be downloaded for free (consult the Methods for a full citation and link)

About this article

Publication history






Further reading