How weather experiences strengthen climate opinions in Europe

Abstract

Previous research has shown that we believe more in the reality of climate change when we experience warmer-than-usual temperatures. This reflects a psychological process in which easily accessible information from personal weather experiences is used as a heuristic to form climate opinions. This paper replicates and extends upon a research design and results of Egan and Mullin to provide the first systematic European study of the Local Warming Effect. Based on data from 12 European countries, the analysis shows that when objective temperatures increase by two standard deviations (5 °C) relative to normal temperatures, climate opinions are strengthened by around 0.5–1.0 percentage points – comparable to the effect of a full step to the left on a 0–10 political ideology scale.

Keywords:

• Local warming effect
• global warming
• temperature
• personal experience
• public opinion

Acknowledgements

I am very grateful to Professor Kasper M. Hansen from the Department of Political Science at the University of Copenhagen for convincing me to publish this work and offering indispensable advice. I also want to give thanks to Patrick J. Egan and Megan Mullin for their inspiring research and to the anonymous reviewers for their very helpful comments and suggestions. Finally, I acknowledge the data providers in the ECA&D project.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The ESS8 data is available at https://www.europeansocialsurvey.org/data and weather data from the ECA&D Project is available at https://www.ecad.eu.

Notes

1 23 countries participated, but Belgium, Portugal and the Czech Republic had insufficient weather data.

2 The issue of measurement error is discussed further in Online Appendix C.2 where the alternative strategy of modelling national temperature variance without excluding countries a priori is also presented.

3 Egan and Mullin (2012) apply a more comprehensive control logic by using date fixed effects to rule out temporal omitted variable bias. This is not possible in the present study as the central variation is between days. In Online Appendix A.3, the risk of temporal omitted variable is assessed. The most plausible source of bias would be discrete events related to climate politics or climate science (e.g. a COP meeting or the presentation of a new climate-related political programme) that affect opinions in a significant number of countries while coinciding with unusual temperatures.

4 Age, gender, ideology (11-point left–right scale), party identification (5-point degree of identification with any party), level of education (seven steps from the ISCED standard), household income (10 intervals), religiosity (attendance in religious services) and belonging to an ethnic minority. All covariates are treated as continuous. Missing income and ideology values are imputed with medians (5).

5 The effect of a two SD shift is 0.53 in model 1, 0.62 in model 4 and 0.94 in model 5. See Online Appendix B.1 for other effect size representations. Ideology has 11 levels; see its effect in Online Appendix B.4.

6 A supplementary analysis shows similar results with an alternative two-dimensional party identification measure (see Online Appendix B.3).

7 See further details in Online Appendix A.5 including a discussion of the shortcomings of the self-reported news exposure measure (Prior 2009).

8 This analysis with added control for political interest is presented in Online Appendix C.7.

9 In contrast, this trend is flat for education level and party identification, cf. Online Appendix B.2.

10 An analysis in Online Appendix B.5 shows that the effect may be limited to beliefs only, rather than concrete policy preferences.

Additional information

Notes on contributors

Søren Damsbo-Svendsen

Søren Damsbo-Svendsen is a Master of Political Science, Department of Political Science, University of Copenhagen. His research interests include political behaviour and opinion formation, media effects, climate change and data science. [[email protected]]