Public knowledge and concern about polar-region warming

Abstract

In 2006 and 2010, before and after the International Polar Year, the General Social Survey asked cross-sections of the US public for their knowledge and opinions about polar regions. The opinion items sought respondents’ levels of concern about global warming in polar regions, and whether they favored opening Antarctica for development or reserving it for science. Polar knowledge scores show significant improvement from 2006 to 2010, while general science literacy scores and opinions remain largely unchanged. Regression of concern and Antarctic items on background characteristics, ideology, education and the two knowledge tests finds that ideology and knowledge have the most consistent effects. Conservative ideology negatively predicts all six concern items and supports for reserving the Antarctic. Polar knowledge exhibits a positive effect on most of the concern items and on support for reserving the Antarctic. General science knowledge has mainly positive effects on concern and Antarctic opinions as well, but its effects are moderated by ideology. These findings support two contrasting but not mutually exclusive views about the role of information: that more science information generally leads to greater concern about environmental changes, or greater support for science; but also that some informed but strongly ideological respondents acquire information selectively in ways that reinforce their existing beliefs.

Acknowledgements

The research described here was supported by a grant from Arctic Social Sciences in the Office of Polar Programs, US National Science Foundation (OPP-1136887).

Notes

1. Burstein (2006) distinguishes between issues with public importance where public opinion often affects policy and other issues with less public importance where interest groups are more likely to dominate. Although climate change does not show up as a top priority on most polls, it receives substantial media attention – particularly with respect to polar changes, and also extreme weather events. Moreover, the topic recurs often in US political discourse, such as statements by political candidates in the 2012 presidential campaign.

2. Some people have questioned the inclusion of Big Bang and evolution items on the GSS science literacy list, arguing that those two items conflate knowledge with beliefs (Bhattacharjee 2010). Their inclusion does not affect the general conclusions of this paper, however. The signs, significance and magnitudes of coefficients for these models in Table 3, including their interaction effects, turn out to be quite similar regardless of whether science literacy is represented by the original 0–11 scale or a truncated 0–9 version. Kahan et al. (2011b) construct a combined science literacy/numeracy scale (with different data) partially based on 8 of the 11 items in the upper part of Table 1 – without Big Bang, evolution or continental drift. For most of these the eight, the percent of correct answers by respondents from their pool appears substantially higher than either of the GSS samples in Table 1. This different knowledge measures, together with a different sampling approach, different dependent variable (risk assessment vs. degree of concern), and different control variables likely accounts for the divergence in our findings.

3. For comparisons of responses by the 309 panel respondents in 2006 and 2010, see Hamilton et al. (2012).

4. A secondary analysis, not shown, established that science literacy scores did not significantly change among college graduate or non-graduate subsets of the sample.

5. Ideology×polar knowledge interaction terms also were tested, but found to have no significant effects. They did raise problems with multicollinearity, and for both of these reasons are left out of the Table 3 models.

6. Several year interaction terms were tested to detect possible changes between 2006 and 2010 in the effects of other predictors. These interactions brought no improvement in fit and raised problems with multicollinearity, however, so they are left out of the final models.

7. See Hamilton (2009) for practical details on conditional effect plotting.

8. Versions of the models with ideology×science literacy interactions (Table 3) performed as well or slightly better than alternatives containing ideology×education interaction terms instead (not shown). Although we could not include both types of interactions in the same models due to multicollinearity, the ideology×science literacy versions here should be viewed not as a rejection of earlier ideology×education findings, but as an equally supported and more interpretable alternative.