1. Introduction
The interest in evaluating society's progress and the effectiveness of public policies based on individual wellbeing are increasingly the focus of researchers and governments. Consequently, prevalent measures of happiness or subjective wellbeing have considered determinants of life satisfaction such as income, job conditions, health, environment, human relations (White, Reference White2007; Kahneman and Deaton, Reference Kahneman and Deaton2010; Ferrer-i-Carbonell, Reference Ferrer-i-Carbonell2013; Clark, Reference Clark2018; Atkinson et al., Reference Atkinson, Bagnall, Corcoran, South and Curtis2020), and more recently climate variables (Welsch, Reference Welsch2006; Maddison and Rehdanz, Reference Maddison and Rehdanz2011; Connolly, Reference Connolly2013; Noelke et al., Reference Noelke, McGovern, Corsi, Jimenez, Stern, Wing and Berkman2016). The relationship between climate and happiness is a recent contribution to our understanding of climate conditions and human wellbeing. This relationship is even more critical in the context of climate change where rapid changes in climate patterns and drastic economic and social consequences of these changes are expected. Heatwaves and extreme temperatures directly affect human health and mortality (Pal and Eltahir, Reference Pal and Eltahir2015; Schär, Reference Schär2016), from heat exhaustion and heat cramps to fatal heat strokes (Becker and Stewart, Reference Becker and Stewart2011). The study of happiness and its relationship with extreme climate variables, especially concerning temperature, precipitation and humidity, can provide insights to understand this topic. This paper presents evidence of the relationship between climate variables and people's happiness in Ecuador, a tropical country with various climate conditions.
Ecuador is a country with abundant natural resources and is one of the most geographically-fragmented countries globally, with a geographic fragmentation indexFootnote 1 of 0.85, compared to an average of 0.3 for OECD countries (Lora et al., Reference Lora, Gallup and Gaviria2003). This fragmentation stems from the existence of three natural regions that determine a remarkable variation of climate conditions across local jurisdictions, as follows: the coastal area on the west towards the Pacific Ocean; the highlands in the middle and along the Andean mountains; and the Amazon to the east of the country. To identify the relationship between climate variables and happiness we compare the levels of overall satisfaction (i.e., how satisfied a person is with her life) and health satisfaction (i.e., individual satisfaction with own health status) of individuals across municipalities in the country, while controlling for critical socio-economic characteristics of the population. Although most determinants of happiness can have the same effect on overall and health satisfaction levels, climate variables can affect these two domains of happiness differently. Specifically, higher temperature levels can be associated with higher levels of satisfaction that result from undertaking outdoor or social activities (Keller et al., Reference Keller, Fredrickson, Ybarra, Côté, Johnson, Mikels, Conway and Wager2005; Connolly, Reference Connolly2008). In contrast, health status, especially among vulnerable populations such as the elderly and children, can worsen as temperature increases, dehydration occurs (Becker and Stewart, Reference Becker and Stewart2011), and the presence of harmful pathogens increases (Vergara et al., Reference Vergara, Rios, Galindo, Gutman, Isbell, Suding and Samaniego2013).
We look at how climate variables are associated with overall and health happiness separately to determine whether these relationships are different. We provide evidence on the relationship between climate variables and happiness by taking advantage of the 2006–2008 national employment surveys, including extensive information about household characteristics and individual self-reported assessments regarding different domains of happiness. We match information about individuals with information about climate variables at the municipal level. We identify the relationship between climate conditions and happiness by taking advantage of the fact that people sharing the same culture and country characteristics live in municipalities with different climate conditions. We employ a repeated cross-section approach while controlling for marital status, employment status, family size, age, and a self-assessment of the relative income level.
Although climate variables directly affect wellbeing, the combined effect of high temperature and humidity levels determines the heat stress people are exposed to and the subsequent impacts on their health. In the context of climate change, greater exposure to high temperatures is expected worldwide (IPCC, Reference Stocker, Qin, Plattner, Tignor, Allen, Boschung, Nauels, Xia, Bex and Midgley2013), which is associated with dehydration, kidney diseases, a higher risk of suffering heat strokes, and higher mortality (García-Trabanino et al., Reference García-Trabanino, Jarquín, Wesseling, Johnson, González-Quiroz, Weiss, Glaser, Vindell, Stockfelt, Roncal, Harra and Barregard2015). Climate conditions in Ecuador imply that municipalities where temperature and humidity levels are high experience heat stress often. To study the relationship between heat stress and happiness, we calculate a heat index (NOAA, 2009; Anderson et al., Reference Anderson, Bell and Peng2013) to reflect the combined effect of average maximum temperature and humidity recorded at the municipal level. Heat stress is notably unsafe for human health when temperature and humidity are higher than 30°C and 80 per cent, respectively, which produces a thermal effect of extreme heat equivalent to temperatures above 35°C (Willet and Sherwood, Reference Willet and Sherwood2012). The share of individuals in the sample living in locations with heat above this risk level represents approximately one-third of the population.
The main findings show that the relationship between climate and happiness is persistent and significant across econometric specifications and after controlling for individual and group characteristics. We explore this relationship by adopting linear and quadratic forms. The relationships between temperature variables and happiness are inverted U-shaped. Overall happiness peaks at an average temperature of 22°C. While the relationship between average temperature and health happiness is an inverted-U peaking at a level of 19°C, the relationship between heat and health happiness is significant only along the decreasing segment of the relationship. However, when looking at regional differences within the country, the relationship between temperature and happiness is mostly positive in cold and dry places, while mostly negative in warm and humid ones. The results suggest that the effect of climate on happiness is large in magnitude in terms of marginal willingness to pay. The effects of socio-economic and personal characteristics on happiness levels are also robust to different econometric specifications.
The evidence on the relationship between climate conditions and individual outcomes is inconclusive. Some research shows that higher temperatures influence people's mood, increase collaborative behavior, and lower anxiety and skepticism (Cunningham, Reference Cunningham1979), whereas low temperatures enhance work productivity (Howarth and Hoffman, Reference Howarth and Hoffman1984). Low and high temperatures in cold places are associated with increased respiratory and cardiovascular illnesses (Braga et al., Reference Braga, Zanobetti and Schwartz2002). Although contradictory results exist, the evidence suggests that higher temperatures trigger aggressive behavior and reduce social affection (Eliasson et al., Reference Eliasson, Knez, Westerberg, Thorsson and Lindberg2007). Similarly, humidity can directly affect people's moods (Cunningham, Reference Cunningham1979). Humidity is inversely related to good mood, diminishes physical energy, and reduces the interest in social interaction (Sanders and Brizzolara, Reference Sanders and Brizzolara1982). Moreover, high humidity reduces concentration, work productivity and collaborative behavior, and increases sleepiness events (Howarth and Hoffman, Reference Howarth and Hoffman1984). More extended periods of sunshine correlate with optimistic choices and positive evaluations of life satisfaction, and with increases in stock financial returns (Hirshleifer and Shumway, Reference Hirshleifer and Shumway2003).
The studies on climate conditions and happiness have been mostly limited to generating evidence for developed countries by taking advantage of differences in climate conditions across seasons. Most evidence shows a positive relationship between climate variables and happiness. The cross-country evidence shows that precipitation and temperature patterns are strong predictors of happiness (Rehdanz and Maddison, Reference Rehdanz and Maddison2005), and that a negative relationship between pollution and wellbeing exists (Welsch, Reference Welsch2006). Connolly (Reference Connolly2013) finds that happiness in summer is higher when the temperature decreases and lower when it increases. Brereton et al. (Reference Brereton, Clinch and Ferreira2008) and Cuñado and de Gracia (Reference Cuñado and de Gracia2013) find similar results from inter-regional analyses in Ireland and Spain, respectively. The economic value of droughts has also been approximated through subjective wellbeing measures (Carroll et al., Reference Carroll, Frijters and Shields2009). However, Lucas and Lawless (Reference Lucas and Lawless2013), studying the relationship between daily climate conditions and subjective wellbeing in the US, find negligible impacts of climate on happiness. Our understanding of the effect of climate on happiness in tropical countries, where climate patterns are different from developed countries and where climate change impacts are expected to be severe, is still scarce. Additionally, unlike countries with four seasons, the effect of climate on happiness in Ecuador corresponds to its inhabitants' long-term adaptation to natural conditions since climatic patterns in the tropics, namely temperature and humidity, are stable throughout the year.
The contribution of the paper is three-fold. First, it identifies the relationship between climate conditions and happiness and addresses the endogeneity in the estimation procedure as follows: (a) climate patterns are exogenous to the individuals and the degree of mobility across municipalities in the country is relatively low where 80 per cent of the individuals in the survey reports to live in the same municipality where they were born. Therefore, the estimation model controls for whether the respondent lives where she was born. (b) Our approach controls for individual characteristics that are usually unobservable and can be important determinants of happiness. These characteristics include personal aspirations (i.e., the difference between the income level that the individual considers as high enough to live well and her current income level). (c) Provincial and year fixed effects and clustered standard errors by municipality and year control for hedonic price differentials across municipalities. Second, the climatic fragmentation in Ecuador makes this country an ideal case to study the effect of climate conditions on economic outcomes. Third, the evidence comes from a country in the tropical region of the world that is expected to suffer significant changes in climatic patterns and where the understanding of the effect of climate on economic outcomes is essential to anticipating changes in wellbeing.
The paper is organized as follows. Section 2 describes the data and the methodology employed in the analysis and section 3 presents different econometric estimation models and discusses the findings on climate conditions, socio-economic variables and happiness. Finally, section 4 concludes.
2. Data and methodology
We use information from the employment survey of Ecuador conducted by the National Institute of Statistics and Census (INEC), the institution responsible for generating the country's official statistics.Footnote 2 We specifically employ the survey waves for the years 2006–2008, which include questions about subjective wellbeing regarding different domains of life. Although INEC collects information about employment conditions quarterly, information on happiness was only collected in the surveys conducted in the last quarter. The domains include the household financial situation, job conditions, health, environment, attitudes towards the government, social life, education level, neighborhood and housing conditions, and overall life assessment. The employment survey also includes information about the social, demographic, and economic profile of the population, and employment status and characteristics of the working-age population. The survey is conducted at the household level, covering urban and rural areas in every municipality of the country.Footnote 3
The sample includes 54,541 households distributed as follows: 17,726 households (32.5 per cent of the sample) in 2006, 18,271 (33.5 per cent of the sample) in 2007, and 18,544 (34 per cent of the sample) in 2008. Descriptive statistics for the main characteristics of respondents are shown in table 1. Household heads are mostly males (77.55 per cent) and were employed (82.36 per cent) on the survey day. Almost 80 per cent of the interviewees described themselves as Hispanic, and nine out of ten reported speaking Spanish exclusively. More than half of the population lives in urban areas (56 per cent of the sample), and married people represent half of the sample, whereas those who have at least completed elementary school represent slightly more than half of the sample. Respondents' ethnic identification and marital status of respondents coincide with the official information generated by the latest population census.Footnote 4 Regarding how optimistic or pessimistic people are, almost 40 per cent of respondents are optimistic, 21 per cent pessimistic and 39 per cent neutral about their living conditions.
Descriptive statistics of the sample at the household level
Questions about the level of satisfaction with different wellbeing domains are of the following form: how satisfied are you with your work /financial situation /health status /environment /government…? The survey adopts a scale from 0 to 10, corresponding to different satisfaction levels, with ten being the highest. People are more satisfied with their marital status, social life and neighborhood characteristics, whereas the financial situation, the government, and education levels achieve the lowest levels of satisfaction. Work condition, health status, leisure, and the environment reach intermediate ranges of satisfaction. For this paper, the focus is on overall life satisfaction (i.e., overall happiness), which is assumed to capture the general self-assessment of happiness and satisfaction with one's own health status (i.e., health happiness). Table 2 shows additional information about individuals' characteristics and climate conditions.
Descriptive statistics of socio-economic and climatic variables
As a measure of climate conditions, we use the monthly average values of temperature, precipitation, and humidity for November and December when the surveys with happiness questions took place. We also employ average maximum temperature levels and a measure of heat stress to identify the effect of temperature extremes on subjective wellbeing. Precipitation, however, is the climate variable exhibiting more variation with high and low peaks in the wet (December to May) and dry seasons (June to November), respectively. Information on climate variables comes from the 2006–2008 annual reports by the National Institute of Meteorology and Hydrology (INHAMI) of EcuadorFootnote 5 and the World Bank's Climate Change Knowledge Portal (see table 2 for details on climate patterns in the country).Footnote 6 Local climate conditions exhibit a remarkable variation across the country's natural regions, as shown in figure 1 which illustrates average temperature and precipitation levels at the municipal level. The analysis includes information on temperature, precipitation and humidity levels for 218 municipalities of Ecuador (86 in the coastal region, 91 in the highlands and 41 in the Amazon). At the municipal level, the average temperature ranges from 10.4°C to 27.6°C, monthly average precipitation from 14.4 to 900 mm, and average humidity levels from 62 to 94 per cent. Municipalities experiencing high temperature and humidity are primarily in the coastal and Amazon regions, whereas municipalities in the highlands are usually colder and drier. Regarding extreme temperatures, the average high temperatures recorded in the country can exceed 30°C, whereas the heat index can go above 40°C.
Annual average temperature (top panel) and precipitation (bottom panel) in Ecuador, 2006–2008.
The theoretical relationship between climate conditions and individual happiness implies that climate is an external factor to the individual (i.e., a factor that cannot be altered by individual decisions) that is an element of her utility function and provides positive or negative effects on wellbeing (van Praag and Ferrer-i-Carbonell, Reference van Praag and Ferrer-i-Carbonell2004). This notion of utility refers to the experienced utility that results from hedonic experiences of realized outcomes instead of decision utility inferred from individual choices (Kahneman and Thaler, Reference Kahneman and Thaler2006). Since the substitutability of components of the utility function is assumed (Mas-Colell et al., Reference Mas-Colell, Whinston and Green1995), it is possible to identify the trade-offs between climate conditions and income at the individual level or the shadow price of climate (van Praag and Ferrer-i-Carbonell, Reference van Praag and Ferrer-i-Carbonell2004). However, individual or collective (i.e., institutional arrangements) decisions can ameliorate or attenuate the effects of climate conditions on happiness. This element implies that the empirical estimation of the effect of climate on happiness needs to acknowledge both the exogenous nature of climate variables and the individual and social characteristics of populations that can mediate this relationship.
The exogenous source of variation allowing identification of the causal effect of climate on happiness consists of differences in individuals' happiness with similar socio-economic characteristics and affected by similar policies and institutions, but living in municipalities with different climate conditions. The variation in climate patterns comes mostly from climate differences across municipalities since the variation within municipalities over time is minimal. In terms of econometric methodology, the omitted variables problem ideally requires panel data to eliminate the bias from unobserved individual characteristics (Dell et al., Reference Dell, Jones and Olken2014; Hsiang, Reference Hsiang2016). Although we employ cross-section observations for three years, we can control for the effect of material aspirations respondents have that may affect their subjective assessment of wellbeing. This variable is the difference between the level of income respondents consider the minimum level for a ‘good’ life and their actual income level. This difference constitutes the gap between the lifestyle individuals aspire to enjoy and the income level they currently earn (variable ‘income gap’ in table 2). With this variable, we aim to determine how far individuals are from reaching their material aspirations and how this affects individual wellbeing.
The methodological treatment of answers to subjective wellbeing questions falling into different qualitative or quantitative categories can adopt two approaches: (1) the response of an individual stating being somewhat satisfied (instead of not satisfied or unsatisfied) may reflect an order mentally structured by the interviewee. In this case, respondents' self-reported wellbeing is treated as an ordinal variable that prevents the researcher from making interpersonal comparisons. (2) If the researcher assumes that a numerical scale reflects a quantitative assessment of the respondent's wellbeing, answers can be treated as cardinal variables that allow interpersonal comparisons. Econometrically, ordered logit models are suitable for analyzing ordinal dependent variables, whereas a cardinal probit approach or group-wise regressions are appropriate to treat wellbeing responses as cardinal variables (van Praag and Ferrer-i-Carbonell, Reference van Praag and Ferrer-i-Carbonell2004; Dolan et al., Reference Dolan, Peasgood and White2008). Some researchers have suggested other econometric models, such as probit ordinary least squares (OLS) that treats the values of the dependent variable as the conditional expectation of the measure within a wellbeing scale defined by intervals. The empirical evidence shows that these three econometric approaches yield similar results (van Praag and Ferrer-i-Carbonell, Reference van Praag and Ferrer-i-Carbonell2004). This paper will employ the probit OLS approach that consists of the cardinalization of subjective wellbeing responses based on Maddala's approach (Reference Maddala1983).Footnote 7 The most common happiness model in the literature adopts an additive functional form (Dolan et al., Reference Dolan, Peasgood and White2008):
where SWBi = self-reported measure of individual wellbeing in logarithmic form; Xi = demographic, social and economic factors at the individual level (income gap, age, gender, employment status, family size, marital status, whether people live where they were born); Zk = climate variables: annual average temperature and maximum temperature, precipitation and humidity, and their variations (linear and square terms, and deviations from averages), at the municipal level; and ε i = individual differences captured in the error term.
The assumptions for modelling subjective wellbeing are the following: (1) causality runs from the explanatory to the dependent variable; (2) no correlation between unobserved factors and the explanatory variables exists (Dolan et al., Reference Dolan, Peasgood and White2008); (3) similar responses from two individuals that fall into the same category represent the same level of satisfaction, which implies that ordinal interpersonal comparability is allowed; and (4) the substitutability between explanatory variables is plausible (van Praag et al., Reference van Praag, Frijters and Ferrer-i-Carbonell2003). Moreover, the causal relationship between climate and happiness does not suffer from endogeneity problems since climate patterns are exogenous to individuals, and mobility within the country is low. We also control for critical factors such as personal aspirations (measured by the variable ‘income gap’), which can be viewed as reference points to self-assess individuals' wellbeing (Kahneman and Krueger, Reference Kahneman and Krueger2006; Rojas, Reference Rojas2007; Clark et al., Reference Clark, Frijters and Shields2008).
It is also possible that individuals seeking to reach higher levels of happiness move to locations where climate conditions are associated with higher wellbeing, which would bias the regression estimates. To control for location choices, we include a dummy variable indicating whether the person lives where she was born (i.e., variable ‘place’ = 1, if the person lives in a different municipality, and ‘place’ = 0 otherwise). Moreover, we control for provincial and municipal level factors (i.e., provincial fixed-effects) in different specifications of the regression model. Although institutional quality at the provincial and municipal levels can result from climate conditions (Acemoglu et al., Reference Acemoglu, Johnson and Robinson2001; Dell et al., Reference Dell, Jones and Olken2014; Hsiang, Reference Hsiang2016), we do this to control the effect of hedonic or price differentials across municipalities in the country. This approach helps control for differences in the cost of living due to climate conditions (e.g., food prices) or exposure to climate events (e.g., seasonal flooding). We cluster standard errors by municipality and year to consider shared characteristics of individuals living in the same jurisdiction, which can also vary over time.
3. Results
As determinants of happiness, we consider socio-economic variables such as gender, age and age squared, family size, employment and marital status, level of income individuals consider enough to live well (i.e., income gap), and whether they currently live where they were born. Although individuals may assess their overall level of happiness (i.e., overall happiness) by taking into account the elements of their lives that are most important to them (Dolan et al., Reference Dolan, Peasgood and White2008), assessments of happiness for specific domains of life can bring a better understanding of particular aspects of individual wellbeing. Concretely, when individuals assess happiness with their health status (i.e., health happiness), they are expected to evaluate how the environment and their physical condition determine their level of satisfaction. Unlike the effect of climate conditions on overall happiness, happiness with one's health status is assumed to consider the direct effect of climate on respondents' health. Places that experience high or low temperatures combined with high humidity constitute locations where people can suffer from health problems related to respiratory and cardiovascular illnesses and changes in people's aggressiveness and mood (Cunningham, Reference Cunningham1979; Braga et al., Reference Braga, Zanobetti and Schwartz2002; Eliasson et al., Reference Eliasson, Knez, Westerberg, Thorsson and Lindberg2007). To determine whether climate conditions affect overall and health happiness differently, the determinants of these two domains are analyzed separately.
The identification of a causal relationship between climate and socio-economic outcomes is always a methodological concern. Deviations from average levels in climate variables have been suggested to establish this effect because of their random nature (Dell et al., Reference Dell, Jones and Olken2014; Hsiang, Reference Hsiang2016). To check the robustness of the relationship between temperature and happiness, we consider the deviations in average temperature as the temperature variable in the regression model. We define temperature deviations as the difference between the average temperature in the study years and the average temperature during 1991–2008. Since deviations in average temperatures vary across municipalities and over time, we can control for factors at the municipal level other than climate conditions that affect happiness. For example, these factors may refer to differences in the prices of food or housing, and ultimately the cost of living across municipalities. However, one should be aware that municipal dummies partially capture the effect of climate conditions on happiness, as suggested in the literature (Acemoglu et al., Reference Acemoglu, Johnson and Robinson2001; Dell et al., Reference Dell, Jones and Olken2014; Hsiang, Reference Hsiang2016).
Results in table 3 provide evidence suggesting a causal relationship between temperature and happiness. Columns 1 and 2 present the results for overall and health happiness as the dependent variables. The relationship between temperature deviations and overall and health happiness shows a positive, statistically significant effect. Specifically, an increase of 1°C in the temperature deviation raises overall and health happiness by 0.013 and 0.018 points, respectively. The results also suggest that women are less happy than men, with an average difference of 4.7 per cent in overall happiness and 11.3 per cent in health happiness. Age is positively associated with overall happiness and negatively associated with health happiness, while its quadratic term harms both types of happiness. Being unemployed decreases overall happiness by 3 per cent and health happiness by almost 16.5 per cent. Family size is negatively associated with overall and health happiness, but this effect is only statistically significant for the former, whereas people living in places other than their place of birth are generally happier. The variable income gap has a negative coefficient showing that the gap between actual income and what individuals consider an adequate level decreases happiness. Like the effect of unemployment, the income gap decreases health happiness more than overall happiness (elasticity values of 0.011 and 0.007, respectively), suggesting that the financial anxiety captured by this variable taxes the health condition more heavily than overall happiness.
Effect of temperature deviations on happiness
Notes: Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001.
The magnitude of the effect of socio-economic characteristics and its statistical significance are consistent across different econometric specifications and with existing evidence from Ecuador (Pontarollo et al., Reference Pontarollo, Orellana and Segovia2020). Additional controls include marital status and province and year fixed effects. The effect of precipitation and humidity levels on happiness is not statistically significant, suggesting that people perceive temperature at a given time (i.e., whether the day is hot or cold) more easily than precipitation and humidity. Standard errors in the econometric specification are clustered by municipality and year to control for factors affecting individuals living in the same locality, which can vary yearly. The results in table 3 suggest that temperature levels are important determinants of individual happiness even after controlling for shared characteristics of households at the provincial and municipal levels, such as cost of living and institutional quality.
Since the literature identifies a non-linear relationship between climate and happiness (van Praag and Ferrer-i-Carbonell, Reference van Praag and Ferrer-i-Carbonell2004; Maddison and Rehdanz, Reference Maddison and Rehdanz2011; Noelke et al., Reference Noelke, McGovern, Corsi, Jimenez, Stern, Wing and Berkman2016), we first explore the linear relationship between climate variables in levels and happiness and then impose a quadratic relationship to identify the temperature at which happiness reaches its highest peak. The results of the main econometric specifications are presented in tables 4 and 5 and show the relationship between climate variables and happiness while controlling for socio-economic determinants of wellbeing. While table 4 considers the linear effect of climate variables, table 5 introduces the model results, including the quadratic term of these variables. We alternate the temperature variable to consider the effect of annual average temperature (columns 1 and 4), annual average maximum temperature (columns 2 and 5), and heat stress (columns 3 and 6) to capture not only the average effect but also the effect of high peaks of temperature on wellbeing. In both tables, columns 1–3 have models for overall happiness, whereas columns 4–6 include models for health happiness. Table 4 shows that overall happiness is positively related to average temperature and average high temperature, whereas health happiness is negatively related to heat. Increases of 1°C in average and average high temperature are associated with an increase of 1.2 per cent and 1.4 per cent in overall happiness, respectively. The negative effect of heat on overall happiness and the positive effects of average and average high temperature on health happiness are statistically insignificant. The relationship between precipitation and happiness is negative, although statistically insignificant. Humidity and happiness are positively related with a statistically significant effect on health happiness and when the average high temperature is the temperature variable (column 5 in table 4). An increase of one percentage point in humidity levels is associated with an increase of 0.5 per cent in health happiness.
Linear relationship between climate variables and overall happiness
Notes: Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001.
Linear and quadratic relationship between climate variables and health happiness
Notes: Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001.
The shape of the quadratic relationships between temperature and happiness (table 5) is an inverted-U. Overall happiness increases as average and average high temperature increase only up to some level, above which it starts decreasing. Overall happiness peaks at an average temperature of 22°C and an average high temperature of 26°C (columns 1 and 2 in table 5). The pattern of this relationship implies that two segments are defining how temperature affects overall happiness. The first segment corresponds to increases in temperature in cold places (i.e., low temperature levels) where people became happier as temperature raises. In our sample, this occurs when average temperature gradually increases from 10.4°C (i.e., the minimum average temperature) to 22°C (i.e., the temperature at which overall happiness is the highest), or when average maximum temperature gradually increases from 12.5°C (i.e., the lowest average maximum temperature) to 26°C (i.e., the maximum temperature at which overall happiness is the highest). The second segment occurs in places already warm (i.e., with an average temperature of 22°C or average maximum temperature of 26°C) where higher temperature levels decrease overall happiness. Regarding the distribution of municipality-year observations, 28,198 observations (51.7 per cent) and 27,343 observations (48.3 per cent) are below and above 22°C, the turning point maximizing happiness. In terms of number, 100 municipalities are below, and 118 above, this temperature level. The relationship between temperature and health happiness exhibits a similar pattern. Health happiness reaches its maximum when average and average maximum temperature are 19°C and 24°C, respectively. The effect of heat on overall happiness is not statistically significant (column 3), whereas its effect on health happiness is only significant along the decreasing segment of the relationship (column 6). Health happiness decreases in heat for levels higher than 25.5°C. The existence of a temperature level maximizing happiness is consistent with other findings that determine the existence of a comfort range for individual satisfaction (Epstein and Moran, Reference Epstein and Moran2006; Budd, Reference Budd2008).
Although not statistically significant, the association of precipitation and humidity with overall happiness is represented by a U shape, whereas the relationship between humidity and health happiness adopts an inverted-U shape. While high humidity levels can be associated with a reduction in physical energy, people's concentration, work productivity and social interaction (Sanders and Brizzolara, Reference Sanders and Brizzolara1982; Howarth and Hoffman, Reference Howarth and Hoffman1984), we do not find evidence that this variable affects happiness. The effects of socio-economic determinants on happiness levels in tables 4 and 5 remain similar in direction and magnitude to those discussed above.
The effect of climate on happiness can differentially affect people with specific socio-economic characteristics. To test the effect of climate on happiness among different groups in the country, we modify the main econometric specification to include the interaction between average maximum temperature and income and education levels. We consider family income and education level as variables that may attenuate the effect of temperature on overall and health happiness. We run regressions similar to those in table 5 but include the interaction between average high temperature and individuals' family income and education.Footnote 8 The interaction variables correspond to the temperature variable and dummies for income quartiles and education levels (i.e., individuals with no formal education, primary education, and secondary education). Table 6 shows the regression specifications with overall and health happiness as the dependent variables, and average high temperature in levels and in interaction with individual socio-economic characteristics as the main independent variables.
Regression models with interaction variables
Notes: Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Additional controls include age, age square, family size, being unemployed, migrated from a different municipality, marital status, province and year dummies.
The results in columns 1 and 2 of table 6 show how income attenuates the effect of temperature on happiness. Although the effect of temperature and happiness remains U-inverted-shaped, the interactions between temperature measures and income quartiles suggest that overall and health happiness levels associated with temperature are higher for individuals with higher income. Whereas higher-income individuals are the happiest (i.e., the difference in happiness levels between the fourth quartile and the other three is negative and statistically significant), middle-income individuals are happier than low-income ones (i.e., the difference in happiness levels between the third and the second and the first quartiles is negative and statistically significant).Footnote 9 The coefficient of the interaction variables implies that the slope of the relationship between temperature and income is smaller for lower-income levels across the income distribution. The effect of temperature on happiness is larger for the first two income quartiles than for the highest quartile. The negative effect is also significant for those in the third quartile, although the effect size is smaller than for lower-income individuals. Families with higher income levels may cope better with climate conditions, allowing income to attenuate the negative effect of temperature on happiness through access to devices and infrastructure, making living conditions more comfortable. Air conditioning systems, weather monitoring, and construction with improved climatic-isolation mechanisms are just a few examples of affordable devices for people with higher incomes. The results remain persistent when the dependent variable is the average temperature.
The results in columns 3 and 4 in table 6 suggest that education also attenuates the effect of temperature on happiness. As with income, individuals with more education have higher overall and health happiness associated with temperature. Again, significant differences in happiness related to the effect of temperature levels exist between individuals with higher education and those with lower levels of education, and between those with primary education and no education and between those with secondary and primary education. Similarly, the coefficients of the interaction variables show that the relationship between temperature and happiness for individuals with less education (i.e., individuals with no education and primary education only) has a smaller slope. Being able to use information about climate and weather conditions and anticipate their risks and effects on health and infrastructure may be associated with higher education levels, which may explain why individuals with more education are happier than those with less education when experiencing the same climate conditions.
Breaking down the data by regions illustrates the effect of weather on happiness in places where climate conditions are noticeably different. For instance, the coastal and the Amazon regions are warm and humid, whereas the highlands are colder and drier. Regression models in table 7, with average maximum temperature as the temperature variable and overall and health happiness as the dependent variables, show the results for the coastal and Amazon regions taken together and the highlands.Footnote 10 The results by region show that climate affects individual happiness differently depending on the prevalent climate conditions in each region. In the highlands, cold and dry, the effect of average high temperature on overall happiness is positive and exhibits an inverted-U shape on health happiness. Health happiness peaks when the average high temperature reaches 24°C. In the coastal and Amazon regions, warm and humid, the effect of temperature on overall and health happiness is statistically significant only along the decreasing segment of the inverted-U relationship. Unlike the highlands, where most temperature levels are within a comfortable range and therefore increase happiness, the results by regions reveal that temperature in places already warm and humid harms both overall and health happiness.Footnote 11 The effects of precipitation and humidity remain insignificant in model specifications by region. These findings are consistent with the concern that heatwaves constitute a challenge for human health and wellbeing, mainly in tropical areas of the world characterized by high temperature and humidity levels.
Regression models by natural regions
Notes: Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001.
To determine the magnitude of the effect of climate variables on happiness, we determine a measure of willingness to pay (MWTP) for climate conditions. To solve endogeneity between income and happiness (i.e., happy people earn more money and a higher income makes people happier), we regress income on a set of individual characteristics and dummy variables by sector of the economy and occupation. We then estimate the predicted income to instrument the income variable in the regressions of happiness on income. We next follow the methodology proposed by Levinson (Reference Levinson2012) and Frey et al. (Reference Frey, Luechinger and Stutzer2010) to find the income level necessary to compensate individuals for an increase in temperature.Footnote 12 The average MWTP for a 1°C change in average and average high temperature for overall happiness is $22.6 and $26.5, or 5.3 per cent and 6.2 per cent of average income, respectively. MWTP slightly decreases when we consider health happiness the measure of wellbeing, with values of $19.3 and $24.2, or 4.5 per cent and 5.7 per cent of average income, for a change of 1°C in average and average high temperature, respectively.
The material in the online appendix compares our estimates to those of other studies and expands the calculation of MWTP by income and education levels, offering a perspective of inequitable impacts of climate conditions on wellbeing across the income distribution. We also explore the potential channels through which climate conditions affect overall and health happiness.
4. Concluding remarks
The paper shows that the relationship between climate conditions and overall and health happiness exists, indicating the direction of the effects and their magnitudes. Ecuador is an ideal case study to identify the effect of climate conditions on socio-economic outcomes because of its geographical fragmentation and its consequent diversity of climate conditions across municipalities. The robustness of the results stems from the fact that different econometric specifications are adopted and that potential confounding effects are controlled for. Personal and socio-economic characteristics are important determinants of happiness levels, confirming the findings in the happiness literature. The effect of climate variables, particularly temperature, on happiness is strong and consistent across econometric specifications. We find an inverted U-shaped relationship between the temperature variable (i.e., average and maximum temperature) and overall and health happiness levels, whereas the relationship between heat and health happiness is only negative. Significant differences regarding the effect of climate on happiness exist across income groups and education levels, and across regions of the country with different prevalent climate conditions. These last results suggest that mitigation and adaptation to changing climate conditions in a warmer world are possible.
The effect of climate or geographical variables on happiness may reflect long-term and short-term effects. The former may include the effect of climate on institutional arrangements (Sokoloff and Engerman, Reference Sokoloff and Engerman2000; Acemoglu et al., Reference Acemoglu, Johnson and Robinson2001) and might result from human adaptation (Landes, Reference Landes1998). The latter might be a direct and immediate effect of temperature, humidity or rain on human wellbeing. Moreover, changes in people's mood can result from prevailing climate conditions and impact the self-assessment of wellbeing. If this is the case, climate may have an immediate effect on the short-term evaluation of happiness (Schwarz and Clore, Reference Schwarz and Clore1983). This paper's findings do not provide a conclusive means to distinguish between the two effects, although the former might be related to provincial level variables given by social conditions and institutional quality. Nevertheless, analyses with panel data and within-year climate variation may contribute to a better understanding of the channels through which the climate directly impacts people's subjective wellbeing.
Additional channels through which climate affects happiness can be identified as well. Geographic conditions can determine the type of crops cultivated and, therefore, the main components of the diet affecting individuals' health. Changing climate conditions may also affect consumption, making people consume more ice cream during hot days, or more alcohol, tobacco and coffee during rainy days; or consumers in a good mood may be willing to spend more money going shopping (Murray et al., Reference Murray, Di Muro, Finn and Leszczyc2010). These are just a few examples of the importance of further understanding the effect of climate on individual behavior and subjective reports of wellbeing.
More research on how satisfaction domains relate to climate conditions and to each other is needed. For instance, the analysis of the degree of substitutability between domains can shed light on what matters most to people in the tropics, and public policy could be targeted to those aspects related to the most critical domains. Some of the open questions reported in the literature remain unexplained in the case of Ecuador. How climate conditions affect happiness with work conditions, financial situation, the environment, politics, and the neighborhood where the person lives, require further research. Complementarily, how climate-related disasters, such as floods or droughts, affect different domains of subjective wellbeing may also contribute to assessing the damages of disasters in terms of wellbeing and designing policies to help mitigate climate change and adapt to the changing patterns of world climate.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1355770X21000267.
