Climate voting in the US Congress: the power of public concern

abstract

In the United States, few constituents know and understand climate policy, prioritize it as a political topic, or let their voting decisions depend on it. In these conditions, representatives would not be expected to pay heed to constituents’ climate concern in their voting decisions. Still, even after controlling for the presence of interest groups, campaign finance, and legislators’ party affiliation and ideology, there is a consistent link between public opinion and votes on cap-and-trade legislation in the House (and to a lesser degree in the Senate). The same is true when public opinion is simulated based on pre-vote district characteristics. Explanations for these findings are discussed.

Keywords:

• Climate change
• public opinion
• Congress
• representation
• responsiveness
• global warming

Acknowledgments

We are grateful to Michele Coscia, Chris Warshaw, Alex Copulsky, and three anonymous reviewers for critically reading the manuscript and suggesting important improvements. We would also like to thank Stephen Ansolabehere for addressing our methodological questions regarding the CCES. Finally, we are grateful to Matto Mildenberger for introducing us to the MRP technique.

Notes

1. Additional polls in 2004 and 2005 showed very similar results.

2. Although the salience of environmental issues has shifted over time in the US, other Winston Group polls show comparable results, peaking at 4% in April 2007 and reaching a low point of 1% in October 2009.

3. However, when a 2000 poll asked about the most important problem ‘25 years from now,’ 14% of respondents cited the environment, making it the top-rated problem (Guber 2003).

4. Tests showed that especially in the case of the House, random effects were not likely to be normally distributed. However, fixed effect sizes and their standard errors seem quite robust to such nonnormality’ (Maas and Hox 2004). In addition, for the House, there might be a concern about correlation between random state effects and the effect of public opinion. To alleviate these concerns, we reran models 1–4 for the House with a control variable for mean state opinion (cf. Bafumi and Gelman 2006). Conclusions were identical to those reported here.

5. By mixing climate belief, climate concern, and desire for climate action, this question is unfortunately not a clear indicator of constituency policy preferences. However, results from Howe et al. (2015) show that different aspects of (district-level) climate opinion are very highly correlated: Cronbach’s alpha among 14 diverse measures is 0.98. So, while the distinction between these aspects is conceptually important, it may not pose empirical problems.

6. NAICS codes: 211, 213, 2212, 324, and 2121.

7. NAICS codes: 221111, 221119, and 221113.

8. Three Congress members in our data set were independent. Because all of them were to some extent linked to the Democratic Party in the period of interest, we8 grouped them together with Democratic members.

9. Another possibility is that once we control for public concern about climate, NRDC membership comes to indicate the prioritization of other, perhaps more local and/or conservationist environmental issues. Yet another explanation might be that NRDC members actually found the proposal to be too restricted.

10. There is a moderately negative bivariate correlation (r = –0.32) between nuclear industry campaign donations and Senators’ climate votes. One possible explanation is that Senators with high shares of nuclear industry donations tend to come from rural states. In fact, after controlling for state urban–rural balance, benefiting industry donations only have a marginally significant connection with votes. Another possibility is that nuclear interest groups found that the cap-and-trade bills did not provide enough support for the nuclear industry.

11. When abstentions were treated as ‘no’ votes, public opinion ceased to be a significant predictor for the Senate in both models 3 and 4.

12. p-Values based on χ² likelihood ratio tests of model with and without public concern.

13. An important nuance here is that we tested only for broad characteristics of the legislator – not those of the constituents. Additional analyses showed that climate opinion in a state or district is highly correlated with general ideological leaning of the public in that district. The analyses in this study do not allow us to measure to what extent the connection between opinion and votes is due to the connection of both to constituent ideology.

14. This R² is an estimate of the upper bound on the expected reliability of our final ‘pre-vote’ public opinion measure: we will at most be able to replicate approximately 81% of the variance in ‘pre-vote’ public opinion.