Climatic Change

, Volume 140, Issue 2, pp 135–147

Climate change may speed democratic turnover



The electoral fate of incumbent politicians depends heavily upon voters’ well-being. Might climate change – by amplifying threats to human well-being – cause incumbent democratic politicians and parties to lose office more frequently? Here I conduct the first-ever investigation of the relationship between temperature, electoral returns, and future climate change. Using data from over 1.5 billion votes in over 4,800 electoral contests held in 19 countries between 1925 and 2011, coupled with meteorological data, I show that increases in annual temperatures above 21 °C (70 °F) markedly decrease officeholders’ vote share. I combine these empirical estimates with an ensemble of climate models to project the impact of climate change on the fate of future officeholders. Resulting forecasts indicate that by 2099 climate change may reduce average incumbent party vote share across all nations in the sample, with the most acute worsening occurring in poorer countries. If realized, these predictions indicate that climate change could amplify future rates of democratic turnover by causing incumbent parties and their politicians to lose office with increasing frequency.


Elections Democracy Political instability Climate change impacts 

1 Introduction

Reductions in voter well-being regularly cause democratic politicians to lose office. This is because voters consider their own well-being and the well-being of those around them when deciding how to cast their ballots (Fiorina 1981). When voters are doing well they more frequently vote for their incumbent politicians (Healy and Malhotra 2013). When voters are doing poorly, whether economically or psychologically, they vote for political challengers at higher rates (Lewis-Beck and Stegmaier 2000). Importantly, scholars have determined that climate change is likely to undermine future economic (Burke et al. 2015) and psychological (Doherty and Clayton 2011) well-being. Might climate change – by reducing citizens’ well-being – induce voters to cast out their incumbent politicians at increasing rates in the future?

That diminished voter well-being can produce electoral losses for incumbent politicians is one of the most extensively documented findings in political science (Lewis-Beck 2006). Most studies focus on the ways that economic outcomes can affect ballot choices, with the conclusion that reductions in macroeconomic performance often precede incumbent politicians’ electoral losses (Duch and Stevenson 2010; 2008; Erikson 1989; Fair 1978; Kramer 1971; Lewis-Beck and Stegmaier 2008). Tufte (1978) articulated this relationship as a basic principle of politics: “When you think of economics, think elections; when you think of elections, think economics” (Tufte 1978). Yet, alterations in well-being not directly tied to the formal economy can also shape voter behaviors. Harmful events such as hurricanes (Abney and Hill 1966; Malhotra and Kuo 2008), tornadoes (Healy and Malhotra 2010), floods (Arceneaux and Stein 2006; Barry 2007; Gasper and Reeves 2011), and droughts (Barnhart 1925; Walker Jr. H and Hansen 1946; Cole et al. 2012) have also shaped the outcome of historical electoral contests. Even more minor reductions in psychological well-being, such as the loss of a favored sports team, have been linked to fewer ballots cast for incumbent politicians (Healy et al. 2010).1

Climate change induced warming is likely to reduce future economic well-being (Nordhaus1991; Stern2006; Weitzman2009) in both rich (Burke et al. 2015; Deryugina 2014) and poor (Dell et al. 2012) countries, in part by reducing individual productivity (Graff Zivin and Neidell 2014), and is likely to amplify the incidence and severity of extreme weather events (Hansen et al. 2012; Min et al. 2011; Rahmstorf and Coumou 2011). Future warming may also undermine human psychological well-being through mechanisms directly tied to increases in temperature extremes, such as worsened emotional states (Klimstra et al. 2011; Connolly 2013). These projected impacts of global climate change include many of the exact weather and climate-induced stressors that have historically caused incumbent democratic politicians to lose votes. Thus, a changing climate may indeed induce citizens to cast out their incumbent politicians at increasing rates. Yet, while this hypothesis flows readily from over a century of literature, this study is the first to explore it.

Here I conduct a multi-national investigation of the relationship between historical temperatures and constituency-level electoral outcomes and link these findings to predictions of future climatic changes. I examine four questions. First, have exogenous increases in temperature harmed the historical vote share of incumbent democratic parties? Second, do the effects of hotter temperatures vary by level of economic development or by density of agriculture? Third, might climate change alter vote shares in the future? Finally, which countries may see the highest future increases in warming-induced democratic turnover?

2 Temperature and changes to incumbent vote share

To investigate if hotter temperatures have indeed reduced historical incumbent party vote share, I employ a dataset of national lower house electoral returns based on over 1.5 billion votes cast in over 4,800 constituency-level electoral contests held in 19 countries variously between 1925 and 2011 (Kollman et al. 2015). These data include countries from North America, South America, and Europe as well as one African nation (see SI: Map of Constituency Boundaries). I link these data to constituency spatial boundaries to map historical monthly meteorological conditions onto each electoral constituency (Mitchell and Jones 2005) (see Supporting Information (SI): Data Description). The theoretical relationship of interest is the total causal effect of increases in constituency-level average temperature in the twelve months prior to an election on changes in the vote share of major incumbent party politicians. I empirically model this relationship as:
$$ {\Delta} Y_{ijt} = f(temp_{ijt}) + precip_{ijt} + \alpha_{i} + \zeta_{m} + \gamma_{t} + \nu_{jt} + \epsilon_{ijt} $$

I control for precipitation ( precipit) as it is correlated with temperature but could independently cause changes in voter behaviors (Fraga and Hersh 2010) (though excluding precipitation does not notably alter parameter estimates, see SI: Main Effect). In this time-series cross-sectional model, i indexes electoral constituencies, j indexes countries, m indexes election months, and t indexes election years. ΔYit represents the change in vote share of the incumbent party ( YitYit−1), defined as the party that won the plurality of votes in that constituency in the prior election (Wilkin et al. 1997). Taking this first difference removes some potentially confounding secular factors from the data – like strength of incumbent party – that may evolve incrementally in each electoral constituency over time (Dell et al. 2012; Burke et al. 2015).

The main independent variable of interest, tempit, represents the average temperature over the twelve months prior to an election held in month m for constituency i in country j and year t (see SI: Temperature and Precipitation). The relationship of interest is represented by f(), which I implement empirically using indicator variables for each 5 °C annual temperature bin, allowing for flexible estimation of a non-linear relationship (Auffhammer et al. 2013; Graff Zivin and Neidell 2014) between temperature and alterations in incumbent party vote share (the functional form remains similar across the use of 2 °C or 1 °C temperature bins, see SI: Alternative Temperature Bins).

Unobserved geographic or temporal factors may influence electoral outcomes in a way that correlates with temperature. For example, voters may be better off on average in constituencies that have better legal institutions, in certain months of the year, or in years with better global economic performance. To ensure that these factors do not bias estimates of the effect of temperature on incumbent party vote share, I include in Eq. 1 three terms, αi, ζm, and γt, that represent constituency, electoral month, and calendar year of election indicator variables, respectively. These variables control for all constant unobserved characteristics for each constituency and for each election month and year (Wooldridge 2010). Further, there may be unobserved, country-specific factors that influence changes in political outcomes over time (Burke et al. 2015). In order to control for these potential confounds I include νjt in Eq. 1, representing country-specific year indicator variables (results are robust to the use of continent-specific year indicators instead, see SI: Time and Location Controls). The identifying assumption, consistent with the literature (Dell et al. 2014), is that annual temperature is as good as random after conditioning on these fixed effects. The estimated model coefficients on temperature terms can thus be interpreted as the causal effect of temperature on changes in incumbent vote share (Burke et al. 2015; Dell et al. 2014; Auffhammer et al. 2013; Hsiang et al. 2013).

I adjust for within-constituency and within-year correlation in 𝜖it by employing heteroskedasticity robust standard errors clustered on both constituency and year (Cameron et al. 2011) (the results are also robust to accounting for spatial and serial dependence (Bester et al. 2011; Hsiang 2010), see SI: Spatial and Serial Correlation). I exclude non-climatic control variables from Eq. 1 because of their potential to generate bias – a phenomenon known as a ‘bad control’ (Burke et al. 2015; Hsiang et al. 2013) – in the parameters of interest. Because of heterogeneous constituency sizes, I weight the regression in Eq. 1 by the number of votes cast in each constituency election (to observe the variance in constituency size and number of votes in each country, see SI: Map of Constituency Boundaries and SI: Table S1). Of note, unweighted regressions return similar estimates of the relationship (see SI: Unweighted Regressions with Varying Temperature Bins). Finally, I omit the 16 °C-21 °C (60-70 °F) temperature indicator variable when estimating Eq. 1. This range contains the temperatures associated with optimal well-being from a number of recent studies (Keller et al. 2005; Maddison and Rehdanz 2011; Albouy et al. 2016). I thus interpret the parameter estimates of f(tempit) as the change in incumbent party vote share associated with a particular temperature range relative to this baseline category.

The results of estimating Eq. 1 for the effects of temperature on changes to incumbent party vote share indicate that after controlling for time and location fixed factors and country-specific trends, increases in annual temperatures above 21 °C (70 °F) significantly reduce incumbents’ electoral performance (see Fig. 1, panel (b) and SI: Regression Tables for full estimation results). For example, annual temperatures in the range of 21 °C-26 °C reduce incumbent vote share by over nine percentage points relative to the 16 °C-21 °C baseline (coefficient: −9.024, p: 0.003, n: 4,880) while constituency annual temperatures above 26 °C reduce incumbent vote share by over sixteen percentage points (coefficient: −16.100, p < 0.001, n: 4,880) (of note, these results remain highly significant even after Bonferroni correction for each temperature bin included in the regression (Mundfrom et al. 2006), see SI: Bonferroni Correction).
Fig. 1

Changes to incumbent party vote share decline with increases in annual temperature. Panel (a) depicts the relationship between average temperature in the year prior to an election and changes in the constituency-level vote share of national lower house incumbent politicians from 1,256 constituencies across 19 countries between 1925 and 2011. Points represent the average change in incumbent vote share for each 5 °C annual temperature bin. The line represents a loess smoothing of the raw data. Panel (b) draws from the estimation of the fixed effects model in Eq. 1 and plots the predicted change in vote share associated with each 5 °C temperature bin. As annual temperature increases beyond 16-21 °C (60-70 °F), changes to incumbent vote share become markedly negative. Shaded error bounds represent 95 % confidence intervals

A 5 °C increase in temperature – the average increase predicted under the RCP8.5 ‘business as usual’ scenario for 2099 as compared to 2010 – that produced a reduction in incumbent vote share of over nine percentage points could be politically substantial. Examining the constituencies in the 16 °C-21 °C temperature range (approximately 11 % of historical constituencies) indicates that 31 % of historical elections in this range had parties that won by less than this nine point margin. In two party constituencies – where electoral swings are mechanically equal to twice the reduction in incumbent vote share – in this temperature range 41 % of historical elections would have been altered by a nine percentage point reduction in the winning party vote share (19 % of all constituencies in the data have only two parties while 39 % have either only two or only three parties, see SI: Frequency of Close Elections). Thus, the effects of hotter annual temperatures on changes in vote share are of a magnitude that is politically meaningful and likely would have resulted in alterations to the historical democratic process if applied to past electoral returns.

3 Income and agriculture

The above estimates represent the average effect of increasing temperatures on changes to incumbent party vote share across all constituencies in the sample. However, democratic constituencies may vary in their response to increasing temperatures. For example, politicians and voters in rich countries may be better able to respond and adapt to the social stressors associated with hotter temperatures, while politicians and voters in poor countries may lack the resources needed to smooth temperature shocks and thus experience more notable decreases in well-being (Dell et al. 2012; Burke et al. 2015). Moreover, not all voters in rich or poor countries are likely to be equally affected by the costs of exposure to hotter temperatures. For example, voters in agricultural areas may experience more direct and costly effects of hotter annual temperatures than do voters in areas less reliant on agriculture for their overall well-being (Burke et al. 2015; Morton 2007). This leads to the second question, do the effects of hotter temperatures vary by level of economic development or by density of agriculture?

To examine whether richer or poorer countries’ voters are more sensitive to hotter temperatures, I stratify the sample by median country-level incomes in 1980 (measured in per-capita purchasing power parity units) and estimate Eq. 1 for both rich and poor country subsamples (Dell et al. 2014; Burke et al. 2015). Figure 2, panel (a), shows that the effect of increases in annual temperatures greater than 21 °C on changes to incumbent party vote share in rich country constituencies is negative, though this effect is significant only at the p < 0.10 level (coefficient: −13.677, p: 0.084, n: 3,933). Panel (b) of Fig. 2 shows that the effect of annual temperatures between 21 °C and 26 °C on changes to incumbent party vote share in poor country constituencies is also negative (coefficient: −7.841, p: 0.054, n: 947). Further, temperatures in excess of 26 °C in poor countries produce markedly negative and highly statistically significant changes to incumbent party vote share (coefficient: −15.162, p: 0.004, n: 947). Thus both in richer and poorer countries I find evidence indicating declines in incumbent party vote share due to an increase in temperature above 21 °C, suggesting that higher incomes may not substantially mute the impact of warming on electoral outcomes (see SI: Rich and Poor). This is consistent with the observation that increasing temperatures reduce economic well-being in both rich and poor nations (Burke et al. 2015; Deryugina 2014; Dell et al. 2012).
Fig. 2

Hot temperatures produce negative changes in incumbent vote share in both rich and poor countries. Panel (a) plots the predicted changes in incumbent party vote share associated with estimating Eq. 1 on the sample of above-median income countries in the data and panel (b) plots this relationship for constituencies in countries falling below median income (Burke et al. 2015). Past 21 °C (70 °F), changes to incumbent vote share decline for both sets of countries, though rich country reductions are significant only at the p < 0.10 level. Shaded error bounds represent 95 % confidence intervals

Using data on remote-sensed crop-cover (Friedl et al. 2010) to split constituencies along the median of percent of croplands, I repeat the above procedure to examine whether agricultural constituencies demonstrate differential electoral responses to increasing temperatures as compared to non-agricultural constituencies (see SI: Agricultural and Non-Agricultural). Figure 3, panel (a), shows that the effect of annual temperatures greater than 26 °C on changes to incumbent party vote share in non-agricultural constituencies is markedly negative, though this effect is estimated with higher variance and fails to gain significance at standard thresholds (coefficient: −18.175, p: 0.130, n: 2,281). Panel (b) of Fig. 3 shows that this effect in agricultural constituencies is also negative and is statistically significant (coefficient: −14.847, p: 0.010, n: 2,271). Thus both agricultural and non-agricultural constituencies’ coefficient estimates suggest a decline in incumbent party vote share due to an increase in temperature above 21 °C. These findings are consistent with the observation that increasing temperatures reduce both agricultural and non-agricultural economic growth (Burke et al. 2015).
Fig. 3

Increases in temperature produce negative changes in incumbent vote share in both non-agricultural and agricultural constituencies. Panel (a) plots the predicted changes in incumbent party vote share associated with estimating Eq. 1 on the sample of constituencies with below-median percentage of remote-sensed agricultural croplands and panel (b) plots this relationship for constituencies with above-median percentages of crop cover. As temperatures increase across both, changes in incumbent vote share decline. Past 21 °C (70 °F), changes in incumbent vote share decline for both sets of constituencies, though the declines in non-agricultural constituencies fail to gain significance. Shaded error bounds represent 95 % confidence intervals

Combining these insights, I split the sample along rich and poor countries’ agricultural and non-agricultural constituencies and estimate Eq. 1 in each sub-sample (see SI: Income and Agriculture). Figure 4, panel (a), shows that the effect of increases in annual greater than 21 °C on changes to incumbent party vote share in rich country non-agricultural constituencies is negative, though this effect is significant only at the p < 0.10 level (coefficient: −20.604, p: 0.053, n: 2,006). Panel (d) of Fig. 4 shows that this effect in poor country agricultural constituencies is also negative and is highly statistically significant (coefficient: −11.760, p < 0.001, n: 344). The regression models suggest the decline in incumbent party vote share due to an increase in temperature above 21 °C may be driven primarily by non-agricultural constituencies in rich nations and by agricultural constituencies in poor nations, though the effects in the rich agricultural and poor non-agricultural sub-samples are estimated with relatively high standard errors. These results could implicate differential causal political mechanisms underlying the relationship between temperature and vote shares in rich versus poor nations and suggest an important area for future research.
Fig. 4

Rich country non-agricultural and poor country agricultural constituencies show largest electoral effects of temperatures. Panel (a) plots the predicted changes in incumbent party vote share associated with estimating Eq. 1 on the sample of rich country constituencies with below-median percentage of remote-sensed agricultural croplands panel (b) plots this relationship for rich country constituencies with above-median percentages of crop cover, panel (c) plots this relationship for poor country constituencies with below-median percentages of crop cover, and panel (d) plots this relationship for poor country constituencies with above-median percentages of crop cover. Vote share in rich non-agricultural areas displays a reduction in incumbent vote share in response to increases in annual temperature above 21 °C (70 °F), though this result is only significant at the p < 0.10 level. Poor country agricultural constituencies also exhibit alterations in vote share in response to positive shifts in temperatures. Shaded error bounds represent 95 % confidence intervals

4 Constituency forecast

The historical data indicate that past temperatures increases have likely altered historical electoral outcomes in meaningful ways. Further, climate change is likely to produce positive shifts in annual temperature distributions in the future (Seneviratne et al. 2014) (see Fig. 5, panel (a)). Positive shifts in annual temperatures may reduce incumbent party vote share in the future, increasing the rate at which incumbent democratic parties and their politicians lose office. These facts lead to the third question: might climate change alter constituency-level vote share in the future?
Fig. 5

Climate change may speed democratic turnover via reductions to incumbent party vote share. Panel (a) depicts the distributions of annual temperature calculated from 21 downscaled climate models for the constituencies in the sample in 2010, 2050, and 2099. Annual temperatures increase in both magnitude and variation by 2050 and 2099 as compared to 2010. Panel (b) depicts the constituency-level forecasts for the impact of climate change on alterations in incumbent vote share in the future. To incorporate downscaled climate model uncertainty, I calculate an estimated change for an ensemble of 21 climatic models for each of the 1,256 constituencies, producing 26,376 estimates for both 2050 and 2099. I take the constituency average of these estimates, plotting the change between 2010 and 2050 with green lines and the predicted change between 2050 and 2099 with purple lines. The black vertical lines indicate the 2.5th to 97.5th percentile range across the average constituency estimates. As can be seen, currently hotter constituencies may experience markedly more negative changes to incumbent party vote share

To examine this question, I calculate projected average annual temperatures for 2050 and 2099 from NASA Earth Exchange’s (NEX) bias-corrected, statistically downscaled temperature forecasts drawn from the 21 CMIP-5 ensemble models in this data run on the RCP8.5 ‘business as usual’ emissions scenario (see SI: Climate Model Data). I chose the CMIP-5 ensemble as it was used for the most recent IPCC report (Taylor et al. 2012) and chose the RCP8.5 scenario as it closely tracks the world’s current emission trend (Riahi et al. 2011). I couple these predicted temperatures with the historical estimate of the relationship between annual temperatures and changes in incumbent party vote share – employing a spline regression model that closely matches the results from Eq. 1 – to calculate a forecast of possible alterations in future vote share due to climate change for each constituency across each downscaled climate model. I employ linear temperature splines in order to better account for changes in temperature that fall within the temperature bins from Eq. 1 and to enable forecasting vote shares for temperatures outside of the historical sample. Further, I employ the coefficient on historical precipitation and again apply it linearly to future precipitation (see SI: Constituency-Level Forecast for details).

I define the constituency-level forecast of the predicted change in incumbent party vote share due to climate change by 2050 ( Vi2050) as:
$$ V_{i2050} = \overline{\hat{\Delta Y}_{ki2050} - \hat{\Delta Y}_{ki2010}} $$
and for the change from 2010 to 2099 ( Vi2099) as:
$$ V_{i2099} = \overline{\hat{\Delta Y}_{ki2099} - \hat{\Delta Y}_{ki2010}} $$

Where k indexes the 21 specific climate models and i indexes the constituencies. Further, \(\hat {\Delta Y}_{ki}\) represents the fitted values derived from the a spline fit of the downscaled climate model data using the functional form from the estimated parameters of Eq. 1 for 2050 and 2099 (see SI: Main Forecast Model). Of note, the results remain similar under the use of the fitted values from Eq. 1 directly (see SI: Alternative Forecast Model). Using a full ensemble of climate models allows for incorporating uncertainty regarding the underlying climatic forecasts into the change in incumbent vote share predictions (Auffhammer et al. 2013; Dell et al. 2014).

Figure 5 panel (b) plots the forecast results. Each of the 1,256 constituencies in the sample has a mean prediction across all of the 21 downscaled climate models. The first quartile predicted reduction in incumbent vote share by 2099 is -5.8 percentage points while the median reduction is -1.9. Constituencies with higher historical annual temperatures experience the largest predicted future declines in incumbent vote share while cooler constituencies may experience more mild declines to even slight increases in vote share. However, the predicted negative impacts of climate change are over thirteen times greater in magnitude than are the positive impacts (the maximum mean prediction by 2099 among sample constituencies is 0.95 percentage points while the minimum mean prediction is -12.43 percentage points).

5 Country forecast

Some nations are hotter than others on average. This fact, coupled with the observation that the effects of temperature on changes to incumbent vote share are non-linear, with most acute effects observed at higher temperatures, leads to the fourth question: which countries may see the highest future increases in warming-induced democratic turnover?

Figure 6 plots country-level forecast results for 2050 and 2099, respectively. Bars for each country represent the average prediction across all of the 21 climate models across each of the constituencies within that country (see SI: Country-Level Forecast). Countries that have higher spatial variation in annual temperatures – such as the United States and Argentina – have a higher range of underlying constituency forecasts. Importantly, countries with higher average historical temperatures – such as Zambia, Brazil, and Colombia – may experience the most significant future reductions in incumbent vote share.
Fig. 6

Climate change may increase the frequency of democratic turnover most in warmer, poorer nations. This figure depicts the country-level averages across the 26,376 constituency-level climate model forecasts for the impact of climate change on alterations to future incumbent vote share by 2050 and 2099. As can be seen, countries with constituencies that experience presently hotter annual temperatures – countries that include many of the poorest countries in the sample – are likely to experience the greatest climate-induced increase in democratic turnover. To incorporate both downscaled climate model uncertainty and intra-country variance, I present the 2.5th to 97.5th percentile range of the 21 climate models across each country’s set of constituencies via the black vertical lines. Countries with greater intra-country variance in historical annual temperatures, like the United States, have a larger range of future constituency-level predictions

6 Discussion

Voting is central to modern politics. It provides the primary means of democratic participation, shapes politicians’ incentives, and regulates the nature of policies. The available evidence indicates that climate change may alter voting patterns in the future, increasing incumbent electoral losses and potentially speeding rates of democratic turnover.

There are several considerations important to the interpretation of these results. First, while I have data from over a billion votes cast across more than a thousand constituencies, optimal data would also include countries not within the present sample. Of special import would be countries with high average annual temperatures, like those in Sub-Saharan Africa. The lack of available spatial data on such countries’ historical electoral boundaries limits the current sample. Second, because I spatially average temperature and precipitation values to the constituency-level, measurement error may exist between average climatic conditions and those that voters actually experienced, possibly attenuating the estimated magnitude of the effects (Hausman 2001). Third, these estimates are based exclusively on annual temperature and precipitation. Because climate change is likely to increase extreme weather events like flooding and heatwaves (Fischer and Knutti 2015), and because such events can also reduce incumbent vote share (Healy and Malhotra 2010), these results may underestimate the full impact of climate change on future democratic turnover. Fourth, the data I have available do not provide a clear understanding of the exact causal mechanisms that drive the observed effects. Future studies should attempt to discern whether these results are being driven primarily by economic factors or by psychological considerations. Finally, it is possible that voters may adapt to altered future climates with political behaviors not seen in the historical data.

Ultimately, turnover – when directly related to politician performance – is vital to well-functioning democracy (Przeworski 2000). However, the empirical results I present here indicate that democratic turnover might increase as a result of climatic events outside the control of individual politicians. This exogenously driven political turnover may shorten democratic time horizons, inducing parties and their politicians to focus on short-run policies at the expense of important longer-run strategies (Healy and Malhotra 2009). Such altered political time horizons may have a particularly deleterious impact on climate mitigation, as the long-run benefits of mitigation are unlikely to be observed from one election to the next. Moreover, the uncertainty induced by increasing rates of democratic turnover can directly upset macroeconomic outcomes (Fowler 2006; Vaaler et al. 2006). Even more starkly, turnover in nations with weak democratic institutions can upend political stability. If incumbents in weak democracies foresee a greater risk of losing office, they sometimes employ electoral fraud and pre-electoral violence to maintain power (Collier and Vicente 2012; Bratton 2008). If these methods fail, incumbents’ loss occasionally precipitates post-electoral violence that can in turn induce broader civil conflict (Straus 2011; Dercon and Gutiérrez-Romero 2012). These insights, when coupled with the empirical findings above, suggest climate change may alter the nature of democratic politics in costly ways in the future.


  1. 1.

    Numerous studies directly link perceptions and personal experience of climate change related events to changes in political attitudes and behaviors regarding climate change. One topic includes literature on the “local warming effect”, or the propensity of individuals to report greater belief in and political concern about climate change when they experience warmer temperatures (Egan and Mullin 2012; Lang 2014; Zaval et al. 2014). Other studies directly examine the underpinnings for political behaviors regarding climate change (Krosnick et al. 2006; Gifford 2011; Myers et al. 2012; Roser-Renouf et al. 2014; Brügger et al. 2015; Linden 2015; Obradovich and Guenther 2016). Here I do not focus on climate change related political attitudes or behaviors themselves but instead on the potential for climatic changes to alter broader political behaviors.



This work was supported by the National Science Foundation (Grant Nos. DGE0707423, 0903551, TG-SES130013, and 1424091). I thank D. Alex Hughes and the San Diego Supercomputer Center for their assistance and J. Burney, J. Fowler, C. Gibson, B. LeVeck, A. Lo, D. Victor, and members of the UCSD Human Nature Group and Comparative Politics Workshop for their helpful comments.

Supplementary material

10584_2016_1833_MOESM1_ESM.pdf (1.1 mb)
(PDF 1.05 MB)


  1. Abney F, Hill L (1966) Natural disasters as a political variable: the effect of a hurricane on an urban election. Am Polit Sci Rev 60(4):974–981CrossRefGoogle Scholar
  2. Albouy D, Graf W, Kellogg R, Wolff H (2016) Climate amenities, climate change, and american quality of life. J Assoc Environ Res Econ 3(1):205–246Google Scholar
  3. Arceneaux K, Stein RM (2006) Who is held responsible when disaster strikes? The attribution of responsibility for a natural disaster in an urban election. Journal of Urban Affairs 28(1):43–53CrossRefGoogle Scholar
  4. Auffhammer M, Hsiang SM, Schlenker W, Sobel A (2013) Using weather data and climate model output in economic analyses of climate change. Rev Environ Econ Policy. ret016Google Scholar
  5. Barnhart J (1925) Rainfall and the populist party in Nebraska. Am Polit Sci Rev 19:527–540CrossRefGoogle Scholar
  6. Barry JM (2007) Rising tide: The great Mississippi ood of 1927 and how it changed America. Simon and SchusterGoogle Scholar
  7. Bester CA, Conley TG, Hansen CB (2011) Inference with dependent data using cluster covariance estimators. J Econ 165(2):137–151CrossRefGoogle Scholar
  8. Bratton M (2008) Vote buying and violence in Nigerian election campaigns. Elect Stud 27(4):621–632CrossRefGoogle Scholar
  9. Brügger A, Dessai S, Devine-Wright P, Morton TA (2015) Psychological responses to the proximity of climate change. Nat Clim ChangGoogle Scholar
  10. Burke M, Hsiang S, Miguel E, et al (2015) Global non-linear effect of temperature on economic production. Nature 527(7577):235–235CrossRefGoogle Scholar
  11. Cameron AC, Gelbach JB, Miller DL (2011) Robust inference with multiway clustering. J Bus Econ Stat 29(2)Google Scholar
  12. Cole S, Healy A, Werker E (2012) Do voters demand responsive governments? Evidence from Indian disaster relief. J Dev Econ 97(2):167–181CrossRefGoogle Scholar
  13. Collier P, Vicente PC (2012) Violence, bribery, and fraud: The political economy of elections in Sub-Saharan Africa. Public Choice 153(1-2):117–147CrossRefGoogle Scholar
  14. Connolly M (2013) Some like it mild and not too wet: The influence of weather on subjective well-being. J Happiness Stud 14(2):457–473CrossRefGoogle Scholar
  15. Dell M, Jones B, Olken B (2012) Temperature shocks and economic growth: Evidence from the last half century. Am Econ J Macroecon 4(3):66–95CrossRefGoogle Scholar
  16. Dell M, Jones BF, Olken BA (2014) What do we learn from the weather? The new climate-economy literature. J Econ Lit 52(3):740–798CrossRefGoogle Scholar
  17. Dercon S, Gutiérrez-Romero R (2012) Triggers and characteristics of the 2007 Kenyan electoral violence. World Dev 40(4):731–744CrossRefGoogle Scholar
  18. Deryugina T (2014) Does the environment still matter? Daily temperature and income in the United States (Rapport technique). National Bureau of Economic ResearchGoogle Scholar
  19. Doherty TJ, Clayton S (2011) The psychological impacts of global climate change. Am Psychol 66(4):265CrossRefGoogle Scholar
  20. Duch RM, Stevenson R (2010) The global economy, competency, and the economic vote. J Polit 72(01):105–123CrossRefGoogle Scholar
  21. Duch RM, Stevenson RT (2008) The economic vote: How political and economic institutions condition election results. Cambridge University PressGoogle Scholar
  22. Egan PJ, Mullin M (2012) Turning personal experience into political attitudes: the effect of local weather on Americans perceptions about global warming. J Polit 1(1):1–14Google Scholar
  23. Erikson R (1989) Economic conditions and the presidential vote. Am Polit Sci Rev 83:567–573CrossRefGoogle Scholar
  24. Fair R (1978) The effect of economic events on votes for president. Rev Econ Stat 60(2):159–173CrossRefGoogle Scholar
  25. Fiorina MP (1981) Retrospective voting in American national elections. Yale University PressGoogle Scholar
  26. Fischer E, Knutti R (2015) Anthropogenic contribution to global occurrence of heavy-precipitation and high-temperature extremes. Nat Clim ChangGoogle Scholar
  27. Fowler JH (2006) Elections and markets: The effect of partisanship, policy risk, and electoral margins on the economy. J Polit 68(1):89–103CrossRefGoogle Scholar
  28. Fraga B, Hersh E (2010) Voting costs and voter turnout in competitive elections. Quarterly Journal of Political Science 5(4):339–356CrossRefGoogle Scholar
  29. Friedl MA, Sulla-Menashe D, Tan B, Schneider A, Ramankutty N, Sibley A, et al (2010) MODIS Collection 5 global land cover: Algorithm refinements and characterization of new datasets. Remote Sens Environ 114(1):168–182CrossRefGoogle Scholar
  30. Gasper J, Reeves A (2011) Make it rain? Retrospection and the attentive electorate in the context of natural disasters. Am J Polit Sci 55(2):340–355CrossRefGoogle Scholar
  31. Gifford R (2011) The dragons of inaction: Psychological barriers that limit climate change mitigation and adaptation. Am Psychol 66(4):290CrossRefGoogle Scholar
  32. Graff Zivin JS, Neidell M (2014) Temperature and the allocation of time: Implications for climate change. J Labor Econ 32(1):1–26CrossRefGoogle Scholar
  33. Hansen J, Sato M, Ruedy R (2012) Perception of climate change. Proc Natl Acad Sci 109(37):E2415—E2423CrossRefGoogle Scholar
  34. Hausman J (2001) Mismeasured variables in econometric analysis: Problems from the right and problems from the left. J Econ Perspect:57–67Google Scholar
  35. Healy A, Malhotra N (2009) Myopic voters and natural disaster policy. Am Polit Sci Rev 103(3):387–406CrossRefGoogle Scholar
  36. Healy A, Malhotra N (2010) Random events, economic losses, and retrospective voting: Implications for democratic competence. Quarterly Journal of Political Science 5(2):193–208CrossRefGoogle Scholar
  37. Healy A, Malhotra N (2013) Retrospective voting reconsidered. Annual Review of Political Science 16:285–306CrossRefGoogle Scholar
  38. Healy A, Malhotra N, Mo CH (2010) Irrelevant events affect voters’ evaluations of government performance. Proc Natl Acad Sci 107(29):12804–12809CrossRefGoogle Scholar
  39. Hsiang SM (2010) Temperatures and cyclones strongly associated with economic production in the Carribean and Central America. Proc Natl Acad Sci 107 (35):15367–15372CrossRefGoogle Scholar
  40. Hsiang SM, Burke M, Miguel E (2013) Quantifying the in uence of climate on human con ict. Science 341(6151):1235367CrossRefGoogle Scholar
  41. Keller MC, Fredrickson BL, Ybarra O, Côté S, Johnson K, Mikels J, et al (2005) A warm heart and a clear head the contingent effects of weather on mood and cognition. Psychol Sci 16(9):724–731CrossRefGoogle Scholar
  42. Klimstra TA, Frijns T, Keijsers L, Denissen JJ, Raaijmakers QA, Aken MAV, et al (2011) Come rain or come shine: Individual differences in how weather affects mood. Emotion 11(6):1495CrossRefGoogle Scholar
  43. Kollman K, Hicken A, Caramani D, Backer D (2015) Constituency-level elections archive (CLEA). Ann Arbor. University of Michigan Center for Political Studies, MIGoogle Scholar
  44. Kramer GH (1971) Short-term uctuations in US voting behavior: 1896-1964. Cowles Foundation for Research in Economics at Yale UniversityGoogle Scholar
  45. Krosnick JA, Holbrook AL, Lowe L, Visser PS (2006) The origins and consequences of democratic citizens’ policy agendas: A study of popular concern about global warming. Clim Chang 77(1-2):7–43CrossRefGoogle Scholar
  46. Lang C (2014) Do weather uctuations cause people to seek information about climate change Clim Chang 125(3-4):291–303CrossRefGoogle Scholar
  47. Lewis-Beck M (2006) Does economics still matter? Econometrics and the vote. J Polit 68(1):208–212CrossRefGoogle Scholar
  48. Lewis-Beck M, Stegmaier M (2000) Economic determinants of electoral outcomes. Pol Sci 3(1):183Google Scholar
  49. Lewis-Beck M, Stegmaier M (2008) The economic vote in transitional democracies. Journal of Elections Public Opinion and Parties 18(3):303–323CrossRefGoogle Scholar
  50. Linden SVD (2015) The social-psychological determinants of climate change risk perceptions: Towards a comprehensive model. J Environ Psychol 41:112–124CrossRefGoogle Scholar
  51. Maddison D, Rehdanz K (2011) The impact of climate on life satisfaction. Ecol Econ 70(12):2437–2445CrossRefGoogle Scholar
  52. Malhotra N, Kuo AG (2008) Attributing blame: The public’s response to Hurricane Katrina. J Polit 70(01):120–135CrossRefGoogle Scholar
  53. Min S, Zhang X, Zwiers F, Hegerl G (2011) Human contribution to more-intense precipitation extremes. Nature 470(7334):378–381CrossRefGoogle Scholar
  54. Mitchell TD, Jones PD (2005) An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int J Climatol 25 (6):693–712CrossRefGoogle Scholar
  55. Morton JF (2007) The impact of climate change on smallholder and subsistence agriculture. Proc Natl Acad Sci 104(50):19680–19685CrossRefGoogle Scholar
  56. Mundfrom DJ, Perrett JJ, Schaffer J, Piccone A, Roozeboom M (2006) Bonferroni adjustments in tests for regression coefficients. Multiple Linear Regression Viewpoints 32(1):1–6Google Scholar
  57. Myers TA, Maibach EW, Roser-Renouf C, Akerlof K, Leiserowitz AA (2012) The relationship between personal experience and belief in the reality of global warming. Nat Clim Chang 3(4):343–347CrossRefGoogle Scholar
  58. Nordhaus WD (1991) To slow or not to slow: The economics of the greenhouse effect. Econ J 101(407):920–937CrossRefGoogle Scholar
  59. Obradovich N, Guenther SM (2016) Collective responsibility amplifies mitigation behaviors. Clim Chang 137(1):307–319CrossRefGoogle Scholar
  60. Przeworski A (2000) Democracy and development: Political institutions and well-being in the world, vol 3. Cambridge University Press, pp 1950–1990Google Scholar
  61. Rahmstorf S, Coumou D (2011) Increase of extreme events in a warming world. Proc Natl Acad Sci 108(44):17905–17909CrossRefGoogle Scholar
  62. Riahi K, Rao S, Krey V, Cho C, Chirkov V, Fischer G, et al (2011) RCP 8.5 – a scenario of comparatively high greenhouse gas emissions. Clim Chang 109 (1-2):33–57CrossRefGoogle Scholar
  63. Roser-Renouf C, Maibach EW, Leiserowitz A, Zhao X (2014) The genesis of climate change activism: from key beliefs to political action. Clim Chang 125(2):163–178CrossRefGoogle Scholar
  64. Seneviratne SI, Donat MG, Mueller B, Alexander LV (2014) No pause in the increase of hot temperature extremes. Nat Clim Chang 4(3):161–163CrossRefGoogle Scholar
  65. Stern N (2006) Review on the economics of climate change. London HM TreasuryGoogle Scholar
  66. Straus S (2011) It’s sheer horror here: Patterns of violence during the first four months of Côte d’Ivoire’s post-electoral crisis. African Affairs, adr024Google Scholar
  67. Taylor KE, Stouffer RJ, Meehl GA (2012) An overview of CMIP5 and the experiment design. Bull Amer Meteor Soc 93(4):485–498CrossRefGoogle Scholar
  68. Tufte E (1978) Political control of the economy. Princeton University Press, PrincetonGoogle Scholar
  69. Vaaler PM, Schrage BN, Block SA (2006) Elections, opportunism, partisanship and sovereign ratings in developing countries. Rev Dev Econ 10(1):154–170CrossRefGoogle Scholar
  70. Walker Jr H, Hansen P (1946) Rural local government: Local government and rainfall: The problem of local government in the northern great plains. Am Polit Sci Rev 40:1113–1123CrossRefGoogle Scholar
  71. Weitzman ML (2009) On modeling and interpreting the economics of catastrophic climate change. Rev Econ Stat 91(1):1–19CrossRefGoogle Scholar
  72. Wilkin S, Haller B, Norpoth H (1997) From Argentina to Zambia: a world-wide test of economic voting. Elect Stud 16(3):301–316CrossRefGoogle Scholar
  73. Wooldridge JM (2010) Econometric analysis of cross section and panel data. MIT pressGoogle Scholar
  74. Zaval L, Keenan EA, Johnson EJ, Weber EU (2014) How warm days increase belief in global warming. Nat Clim Chang 4:143–147CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2016

Authors and Affiliations

  1. 1.Belfer Center for Science and International Affairs, Kennedy School of GovernmentHarvard UniversityCambridgeUSA
  2. 2.Media Lab, Massachusetts Institute of TechnologyCambridgeUSA

Personalised recommendations