Retrospectively sweaty? The effects of temperature changes on presidential approval

POL837 Issues in Comparative Politics Research Paper

Daniel Sánchez Pazmiño

Simon Fraser University

April 2024

Introduction

Ideally, voters can adequately assess the performance of elected officials to hold them accountable at the ballot box.
However, evaluations of political performance can be influenced by factors unrelated to politician performance (Achen & Bartels, 2017; A. J. Healy et al., 2010; A. Healy & Malhotra, 2010).
I investigate whether transient temperature changes affect presidential approval in Ecuador.
Results show that a 1°C increase in maximum temperature leads to a 1.1 to 2.0 percentage points decrease in the likelihood of presidential approval.

Theoretical framework

Schwarz & Clore (1983) and Bower (1981) show that, when induced negative moods, individuals are more likely to search for outside factors which justify their feelings.
A “higher than normal” temperature would then induce a negative mood, leading to citizens looking for reasons to justify their feelings of discontent.
Existing research has found significant effects of weather on behaviour (Barrington-Leigh & Behzadnejad, 2017; Kämpfer & Mutz, 2013; Keller et al., 2005; Lignier et al., 2023).

Hypothesis

Citizens will be more likely to disapprove of the president on days with higher temperatures.

Empirical strategy

Data

Global daily gridded temperature data from the U.S. National Oceanic and Atmospheric Administration Climate Prediction Center (2024).
Collects temperature of 50x50 km grids across the world mapped to Ecuadorian municipalities (cantons).

Average daily temperature for 2023, cantons map

Data: The AmericasBarometer Survey

Biennial public opinion survey conducted in Ecuador since 2004 (LAPOP, n.d.).
I use a pooled cross-section of individuals from 2008 to 2023.
Daily weather data is joined with the survey by interview date and canton of the survey respondent.

Identification strategy

I use fixed effects logistic regression to estimate presidential popularity functions:

\[ \text{approval}_{ijt} = \alpha + \tau_t + \theta_j + \beta \ \text{temp}_{jt} + \mathbb{X'}_{ijt} \ \gamma + u_{ijt} \]

\(\text{approval}_{ijt}\) is a binary indicator of approval by respondent \(i\) in canton \(j\) on day \(t\).
\(\tau_t\) and \(\theta_j\) are day and canton fixed effects, \(\text{temp}_{jt}\) is a daily temperature measure in canton \(j\) on day \(t\), and \(\mathbb{X'}_{ijt}\) is a vector of control variables for presidential approval.
Short-term temperature changes can be seen as random and thus unrelated to variables that affect presidential approval.
\(\hat{\beta}\) is consistent and measures the causal effect of temperature on presidential approval.
Clustered standard errors at the canton level.

Results

Baseline results

Four specifications with no controls other than fixed effects.
- Minimum temperature only, maximum temperature only, average temperature (mean of min and max), and both temperatures with daily precipitation.
Statistically significant and negative effect of maximum temperature.
Average marginal effect (AME) of -1.1%.

Coefficient plot of average partial effects of baseline specifications

Models with controls

The same four specifications with controls for respondent characteristics.
Statistically significant and negative effect of maximum temperature; AME of -2%.

Coefficient plot of average partial effects of specifications with controls

Conclusion

Important to consider the implications of weather for retrospective voting, especially for developing countries.
Survey respondents are 1.1 to 2.0% less likely to approve of the president after 1C increase in maximum temperature.
The evidence supports the theory of poor retrospective voters due to mood-induced attribution errors.
The CPC weather data could be subject to measurement error, which would downward bias my estimates.

References

Achen, C. H., & Bartels, L. M. (2017). Blind retrospection: Electoral responses to droughts, floods, and shark attacks. In Democracy for Realists: Why Elections Do Not Produce Responsive Government (REV - Revised). Princeton University Press. https://doi.org/10.2307/j.ctvc7770q

Barrington-Leigh, C. (2008). Weather as a transient influence on survey-reported satisfaction with life. MPRA Paper, 25736.

Barrington-Leigh, C., & Behzadnejad, F. (2017). The impact of daily weather conditions on life satisfaction evidence from Canadian cross-sectional and panel data. Journal of Economic Psychology, 59, 145–163. https://doi.org/10.1016/j.joep.2017.01.003

Bassi, A. (2019). Weather, risk, and voting: An experimental analysis of the effect of weather on vote choice. Journal of Experimental Political Science, 6(1), 17–32. https://doi.org/10.1017/XPS.2018.13

Berlemann, M., & Enkelmann, S. (2014). The economic determinants of U.S. Presidential approval: A survey. European Journal of Political Economy, 36, 41–54. https://doi.org/10.1016/j.ejpoleco.2014.06.005

Bower, G. H. (1981). Mood and memory. American Psychologist, 36(2), 129–148. https://doi.org/10.1037/0003-066X.36.2.129

Connolly, M. (2013). Some like it mild and not too wet: The influence of weather on subjective well-being. Journal of Happiness Studies, 14(2), 457–473. https://doi.org/10.1007/s10902-012-9338-2

Deller, C., & Michels, J. (2022). The effect of weather on subjective performance evaluation ({{SSRN Scholarly Paper}} 3780405). https://doi.org/10.2139/ssrn.3780405

Feddersen, J., Metcalfe, R., & Wooden, M. (2016). Subjective wellbeing: Why weather matters. Journal of the Royal Statistical Society Series A: Statistics in Society, 179(1), 203–228. https://doi.org/10.1111/rssa.12118

Healy, A. J., Malhotra, N., Mo, C. H., & Laitin, D. (2010). Irrelevant events affect voters’ evaluations of government performance. Proceedings of the National Academy of Sciences of the United States of America, 107(29), 12804–12809. https://www.jstor.org/stable/25708619

Healy, A., & Malhotra, N. (2010). Random events, economic losses, and retrospective voting: Implications for democratic competence. Quarterly Journal of Political Science, 5(2), 193–208. https://doi.org/10.1561/100.00009057

Kämpfer, S., & Mutz, M. (2013). On the sunny side of life: Sunshine effects on life satisfaction. Social Indicators Research, 110(2), 579–595. https://www.jstor.org/stable/24718723

Keller, M. C., Fredrickson, B. L., Ybarra, O., Côté, S., Johnson, K., Mikels, J., Conway, A., & Wager, T. (2005). A warm heart and a clear head: The contingent effects of weather on mood and cognition. Psychological Science, 16(9), 724–731. https://doi.org/10.1111/j.1467-9280.2005.01602.x

LAPOP. (n.d.). The AmericasBarometer by the Latin American Public Opinion Project (LAPOP) Ecuador 2004 - 2023 Merged File [Dataset with Codebook].

Li, M., Ferreira, S., & Smith, T. A. (2020). Temperature and self-reported mental health in the United States. PLOS ONE, 15(3), e0230316. https://doi.org/10.1371/journal.pone.0230316

Li, X., & Patel, P. C. (2021). Weather and high-stakes exam performance: Evidence from student-level administrative data in Brazil. Economics Letters, 199, 109698. https://doi.org/10.1016/j.econlet.2020.109698

Lignier, P., Jarvis, D., Grainger, D., & Chaiechi, T. (2023). Does the climate impact satisfaction with life? An Australian spatial study. Weather, Climate, and Society, 15(1), 159–175. https://doi.org/10.1175/WCAS-D-22-0063.1

Lucas, R. E., & Lawless, N. M. (2013). Does life seem better on a sunny day? Examining the association between daily weather conditions and life satisfaction judgments. Journal of Personality and Social Psychology, 104(5), 872–884. https://doi.org/10.1037/a0032124

Mullins, J. T., & White, C. (2019). Temperature and mental health: Evidence from the spectrum of mental health outcomes. Journal of Health Economics, 68, 102240. https://doi.org/10.1016/j.jhealeco.2019.102240

National Oceanic and Atmospheric Administration (NOAA) Physical Sciences Laboratory (PSL). (2024). CPC Global Unified Temperature [Datasets].

Quijano-Ruiz, A. (2023). Assessing the reliability of self-rated health: the effects of transient weather fluctuations on perceived health [{Working Paper}].

Schmiedeberg, C., & Schröder, J. (2014). Does weather really influence the measurement of life satisfaction? Social Indicators Research, 117(2), 387–399. https://www.jstor.org/stable/24720831

Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well-being: Informative and directive-effects of affective states. Journal of Personality and Social Psychology, 45(3), 513–523.

Appendices

Documented weather effects

Existing research has found significant effects of weather on self-reported well-being and life satisfaction (Barrington-Leigh, 2008; Barrington-Leigh & Behzadnejad, 2017; Connolly, 2013; Kämpfer & Mutz, 2013; Keller et al., 2005; Lignier et al., 2023).
Individuals are affected by unpleasant weather, which is highly dependent on the regional and seasonal context.
In some cases, effects are significant though small and not robust (Lucas & Lawless, 2013; Schmiedeberg & Schröder, 2014).
Bassi (2019) also proposes that bad weather negatively impacts individuals’ mood, finding that individuals are less likely to vote by risky candidates when faced with bad weather.
Feddersen et al. (2016) and Quijano-Ruiz (2023) find small negative effects of temperature changes on self-rated health; M. Li et al. (2020) and Mullins & White (2019) find significant impacts on mental health.
Deller & Michels (2022) find that rain impacts the way that managers evaluate subordinates and X. Li & Patel (2021) find that students are only marginally sensitive to higher temperatures during exams.

Descriptive Statistics

Descriptive statistics for the matched AB data and weather variables
		N	Percent	Missing (%)	Mean	Std. dev.	Min	Median	Max	Percent
Education	None	153	1.12	100						1.12
	Primary	3255	23.81	100						23.81
	Secondary	6112	44.71	100						44.71
	Superior	2465	18.03	100						18.03
Female	Male	5992	43.83	100						43.83
	Female	6065	44.37	100						44.37
Labour market status	Employed	7319	53.54	100						53.54
	Not in Labour Force	4986	36.47	100						36.47
	Unemployed	1271	9.30	100						9.30
Worse perception of personal economy	Better or Same	8938	65.38	100						65.38
	Worse	4553	33.31	100						33.31
Worse perception of country economy	Better or Same	7744	56.65	100						56.65
	Worse	5711	41.78	100						41.78
Perception of corruption	Not Corrupt	4667	34.14	100						34.14
	Corrupt	6230	45.57	100						45.57
Tolerance to bribes	Not Tolerant	10580	77.40	100						77.40
	Tolerant	2688	19.66	100						19.66
	Presidential approval	13554	100.00	1	0.48	0.50	0.00	0.00	1.00	100.00
	Daily minimum temperature (C)	13139	100.00	4	16.55	6.52	1.56	18.33	27.78	100.00
	Daily maximum temperature (C)	13139	100.00	4	24.76	4.80	10.62	25.68	34.34	100.00
	Daily average temperature (C)	13139	100.00	4	20.65	5.45	8.96	22.19	29.28	100.00
	Daily precipitation (mm)	13195	100.00	3	5.73	8.24	0.00	2.65	56.95	100.00
	Age (years)	13644	100.00	0	38.81	15.89	16.00	36.00	96.00	100.00
	Ideology score (0-10)	9222	100.00	33	5.35	2.46	1.00	5.00	10.00	100.00
	Political pride score	13384	100.00	2	4.06	1.77	1.00	4.00	7.00	100.00
	Trust in police score (0-7)	13589	100.00	1	3.97	1.79	1.00	4.00	7.00	100.00
	Trust in local government score (0-7)	13529	100.00	1	3.91	1.75	1.00	4.00	7.00	100.00

Note: ^^ Note: Descriptive statistics for variables used in the empirical analysis. For categorical variables, the percent of observations in the category out of the total sample is presented. For numerical (either ordinal or continuous) variables, the mean, standard deviation, minimum and maximum are presented. For both, the number of observations and the percentage of missing values.

Explained variable: presidential approval

The AmericasBarometer asks a slight variation of the common Gallup poll question (Berlemann & Enkelmann, 2014), with a 5-point scale, five being full disapproval.
Dichotomized the variable to a binary indicator of approval, grouping responses of 1 and 2 into approval and 0 otherwise.

Presidential Approval Rating from 2008-2023

Treatment: daily temperature

I observe daily maximum and minimum temperature for cantons in the interview dates of the AB.

Baseline logit coefficients

Logit coefficients for baseline specifications
	(1)	(2)	(3)	(4)
Min. temperature (°C)	0.018			0.029
	(0.028)			(0.027)
Max. temperature (°C)		-0.044**		-0.051***
		(0.019)		(0.018)
Avg. temperature (°C)			-0.023
			(0.035)
Precipitation (mm)				-0.004
				(0.004)
N	14118	14118	14118	14118
AIC	18302	18297	18302	18297
RMSE	0.465	0.465	0.465	0.465
Canton fixed effects	X	X	X	X
Interview date fixed effects	X	X	X	X

Note: ^^ Baseline models explaining presidential approval through daily weather variables and canton and interview date fixed effects. Standard errors shown in parentheses are clustered by canton.

Note: ^^ ***p < 0.01, **p < 0.05, * p < 0.1.

Baseline average marginal effects

Average marginal effects for baseline models
	(1)	(2)	(3)	(4)
Min. temperature (°C)	0.004			0.006
	(0.006)			(0.006)
Max. temperature (°C)		-0.010**		-0.011***
		(0.004)		(0.004)
Avg. temperature (°C)			-0.005
			(0.008)
Precipitation (mm)				-0.001
				(0.001)
N	14118	14118	14118	14118
AIC	18302	18297	18302	18297
RMSE	0.465	0.465	0.465	0.465

Note: ^^ Average partial effects for baseline models explaining presidential approval through daily weather variables and canton and interview date fixed effects. Standard errors shown in parentheses are clustered by canton.

Note: ^^ *** p < 0.01, ** p < 0.05, * p < 0.1.

Logit coefficients with controls

Logit coefficients for specifications with controls
	(1)	(2)	(3)	(4)
Min. temperature (°C)	0.003			0.003
	(0.059)			(0.060)
Max. temperature (°C)		-0.107**		-0.106**
		(0.048)		(0.048)
Avg. temperature (°C)			-0.117
			(0.075)
Precipitation (mm)				0.001
				(0.009)
Female	-0.242***	-0.246***	-0.248***	-0.246***
	(0.072)	(0.071)	(0.071)	(0.072)
Age	0.004	0.004	0.004	0.004
	(0.003)	(0.003)	(0.003)	(0.003)
White (ref. Mestizo)	-0.129	-0.118	-0.115	-0.119
	(0.154)	(0.153)	(0.154)	(0.152)
Indigenous	0.439	0.403	0.421	0.403
	(0.277)	(0.279)	(0.280)	(0.279)
Black	0.033	0.012	0.026	0.012
	(0.268)	(0.271)	(0.273)	(0.267)
Mulatto	-0.044	-0.024	-0.031	-0.024
	(0.308)	(0.311)	(0.310)	(0.308)
Other ethnicity	0.177	0.180	0.204	0.179
	(1.091)	(1.100)	(1.098)	(1.093)
Rural area	-0.075	-0.094	-0.086	-0.095
	(0.135)	(0.136)	(0.134)	(0.135)
Religious	0.096	0.088	0.099	0.088
	(0.161)	(0.162)	(0.162)	(0.161)
Married (ref. single)	0.011	0.010	0.015	0.010
	(0.095)	(0.094)	(0.094)	(0.094)
Divorced/Separated/Widowed	0.178	0.172	0.176	0.172
	(0.209)	(0.207)	(0.208)	(0.207)
Primary education (ref. No education)	0.419	0.397	0.409	0.397
	(0.828)	(0.836)	(0.831)	(0.835)
Secondary education	0.553	0.528	0.540	0.528
	(0.812)	(0.820)	(0.816)	(0.819)
Higher education	0.468	0.440	0.454	0.439
	(0.796)	(0.804)	(0.800)	(0.803)
Not in Labour Force	0.137	0.143	0.146	0.143
	(0.091)	(0.091)	(0.091)	(0.091)
Unemployed	-0.202	-0.202	-0.205	-0.201
	(0.219)	(0.222)	(0.222)	(0.221)
Perceived worse personal economy	-0.386***	-0.385***	-0.387***	-0.385***
	(0.102)	(0.104)	(0.103)	(0.103)
Perceived worse country economy	-0.601***	-0.597***	-0.599***	-0.597***
	(0.103)	(0.105)	(0.104)	(0.105)
Voted for incumbent	1.228***	1.226***	1.228***	1.225***
	(0.115)	(0.113)	(0.114)	(0.114)
Ideology score (0-10)	-0.062***	-0.062***	-0.062***	-0.062***
	(0.021)	(0.021)	(0.021)	(0.021)
Supports democracy	0.319***	0.317***	0.319***	0.317***
	(0.095)	(0.097)	(0.096)	(0.097)
Political pride score (0-7)	0.179***	0.179***	0.177***	0.179***
	(0.033)	(0.034)	(0.034)	(0.034)
External efficacy score (0-7)	0.182***	0.180***	0.180***	0.180***
	(0.024)	(0.024)	(0.024)	(0.025)
Internal efficacy score (0-7)	0.030	0.031	0.031	0.031
	(0.028)	(0.028)	(0.028)	(0.028)
Perceives corruption	0.185*	0.192**	0.193**	0.192**
	(0.095)	(0.094)	(0.095)	(0.096)
Tolerates bribes	-0.210	-0.215	-0.216	-0.215
	(0.130)	(0.131)	(0.131)	(0.131)
Trust in police score (0-7)	0.118***	0.120***	0.120***	0.120***
	(0.030)	(0.031)	(0.031)	(0.031)
Trust in local gov. (0-7)	0.034	0.033	0.035	0.033
	(0.046)	(0.045)	(0.046)	(0.045)
N	3553	3553	3553	3553
AIC	4330	4325	4328	4329
RMSE	0.430	0.430	0.430	0.430
Canton fixed effects	X	X	X	X
Interview date fixed effects	X	X	X	X

Note: ^^ Models explaining presidential approval through daily weather variables and controls. Standard errors shown in parentheses are clustered by canton.

Note: ^^ *** p < 0.01, ** p < 0.05, * p < 0.1.

Average marginal effects with controls

Average marginal effects for models with controls
	(1)	(2)	(3)	(4)
Min. temperature (°C)	0.001			0.001
	(0.011)			(0.011)
Max. temperature (°C)		-0.020**		-0.020**
		(0.010)		(0.010)
Avg. temperature (°C)			-0.022
			(0.015)
Precipitation (mm)				0.000
				(0.002)
Female	-0.045***	-0.045***	-0.046***	-0.045***
	(0.014)	(0.013)	(0.014)	(0.013)
Age	0.001	0.001	0.001	0.001
	(0.001)	(0.001)	(0.001)	(0.001)
White (ref. Mestizo)	-0.024	-0.022	-0.021	-0.022
	(0.028)	(0.028)	(0.028)	(0.028)
Indigenous	0.080	0.073	0.076	0.073
	(0.049)	(0.051)	(0.050)	(0.051)
Mulatto	-0.008	-0.004	-0.006	-0.004
	(0.057)	(0.057)	(0.057)	(0.057)
Black	0.006	0.002	0.005	0.002
	(0.049)	(0.050)	(0.050)	(0.049)
Other	0.032	0.033	0.037	0.033
	(0.199)	(0.200)	(0.199)	(0.198)
Married (ref. Single)	0.002	0.002	0.003	0.002
	(0.017)	(0.017)	(0.017)	(0.017)
Divorced/Separated/Widowed	0.033	0.032	0.032	0.032
	(0.038)	(0.038)	(0.038)	(0.038)
Rural area	-0.014	-0.017	-0.016	-0.017
	(0.025)	(0.025)	(0.025)	(0.025)
Primary education (ref. No education)	0.077	0.073	0.075	0.073
	(0.153)	(0.154)	(0.154)	(0.154)
Secondary education	0.102	0.097	0.100	0.097
	(0.150)	(0.151)	(0.150)	(0.151)
Higher education	0.086	0.081	0.084	0.081
	(0.147)	(0.148)	(0.148)	(0.148)
Not in Labour Force	0.025	0.026	0.027	0.026
	(0.017)	(0.016)	(0.017)	(0.017)
Unemployed	-0.037	-0.037	-0.038	-0.037
	(0.041)	(0.041)	(0.041)	(0.041)
Perceived worse personal economy	-0.072***	-0.072***	-0.073***	-0.072***
	(0.019)	(0.020)	(0.019)	(0.019)
Perceived worse country economy	-0.115***	-0.114***	-0.114***	-0.114***
	(0.020)	(0.019)	(0.020)	(0.020)
Voted for incumbent	0.241***	0.241***	0.241***	0.241***
	(0.022)	(0.021)	(0.022)	(0.022)
Ideology score (0-10)	-0.011***	-0.011***	-0.011***	-0.011***
	(0.004)	(0.004)	(0.004)	(0.004)
Internal efficacy score (0-7)	0.006	0.006	0.006	0.006
	(0.005)	(0.005)	(0.005)	(0.005)
External efficacy score (0-7)	0.034***	0.033***	0.033***	0.033***
	(0.004)	(0.004)	(0.005)	(0.005)
Supports democracy	0.060***	0.059***	0.059***	0.059***
	(0.017)	(0.017)	(0.017)	(0.017)
Political pride score (0-7)	0.033***	0.033***	0.033***	0.033***
	(0.006)	(0.006)	(0.006)	(0.006)
Perceives corruption	0.034*	0.036**	0.036**	0.036**
	(0.018)	(0.017)	(0.018)	(0.018)
Tolerates bribes	-0.039	-0.040	-0.040	-0.040
	(0.024)	(0.025)	(0.024)	(0.024)
Trust in police score (0-7)	0.022***	0.022***	0.022***	0.022***
	(0.006)	(0.006)	(0.006)	(0.006)
Trust in local gov. (0-7)	0.006	0.006	0.006	0.006
	(0.009)	(0.009)	(0.009)	(0.008)
N	3553	3553	3553	3553
AIC	4330	4325	4328	4329
RMSE	0.430	0.430	0.430	0.430

Note: ^^ Average partial effects for models explaining presidential approval through daily weather variables, canton and interview date fixed effects, and political behaviour controls. Standard errors shown in parentheses are clustered by canton.

Note: ^^ *** p < 0.01, ** p < 0.05, `* p < 0.1.