EXTREME HEAT REDUCES INDIVIDUAL HAPPINESS

1. Introduction

With the onset of the Industrial Revolution, productivity has since significantly increased, and the economy has also since experienced rapidly development. However, the subsequent increase in carbon emissions in the process of economic growth has caused issues such as global climate change. An important manifestation of climate change is global warming, which is the increase in the frequency of extreme temperatures. The World Meteorological Organization released its “State of Climate in 2021: Extreme events and major impacts” which pointed out that the past seven years were the warmest on record, indicating the increase in the frequency and intensity of extreme weather experienced worldwide. The consequent food crisis triggered by this extreme weather has also affected 161 million people, which is a 19% increase from the 135 million affected in the previous year. As the world’s largest developing country, China’s situation is far from optimistic. According to the “National Climate Change Adaptation Strategy 2035” issued by China’s Ministry of Ecology and Environment in 2022, the global average temperature has increased by 0.15^∘C every decade since the mid-20th century. The temperature in China has also risen significantly, with the average temperature warming rate reaching 0.26^∘C every decade from 1951 to 2020 — which is higher than the global average during the same period. Climate change has thus subsequently attracted global attention due to its negative effects on economic growth, agricultural production, and individual health (Deschênes and Greenstone, 2007; Dell et al., 2012; Barreca et al., 2016; Li et al., 2020; Wang et al., 2020; Aragón et al., 2021; Chen et al., 2022, 2023a,b).

The socioeconomic consequences of the increase in the intensity and frequency of extreme weather conditions have also made policymakers concerned. Numerous studies have assessed the consequences of extreme weather (Dell et al., 2014; Heal, 2017), discussing the impact of extreme weather from different perspectives. These discussions mainly include three aspects: The first explores the relationship between climate change and economic growth and conducts empirical tests around the environmental Kuznets curve (Gill et al., 2018). The second discusses the agricultural sector that is sensitive to climate change and its impact on farmers’ income (Habtemariam et al., 2017). Further research development reveal the relationship between climate change and health, thus extending the study of climate change to the labor market (Zivin and Neidell, 2014; Jessoe et al., 2018). Although these studies analyze the effects of climate change from different perspectives, there is little discussion on the happiness of micro-subjects in these analyses. The fundamental purpose of policymakers’ increased mitigation of climate change is to improve the social happiness. Therefore, it is necessary to analyze these discussions from the happiness viewpoint, which is closely related to individual happiness. Temperature and precipitation are the two most important variables that reflect the weather. With the recent increasing frequency and intensity of extreme heat in China, we will discuss the relationship between extreme heat and happiness to analyze the impact of climate change on social welfare.

There are two primary rationales for selecting China as the focus of our research. First, China’s substantial population and expansive territory offer valuable opportunities to explore the heterogeneous effects of extreme weather. Geographically, China spans over about 50 degrees of latitude, stretching from approximately 3^∘$51^{\prime }$N in the south to 53^∘$33.5^{\prime }$N in the north, encompassing both tropical and temperate climates. Moreover, coastal and inland regions of China exhibit a notable disparity in precipitation patterns. The southeastern coastal areas receive abundant rainfall, while the northwestern inland areas experience limited precipitation. The substantial geographic variations significantly improve our ability to capture and understand the impacts of climate change. Second, China provides a wealth of representative data on happiness. We leveraged data from the Chinese General Social Survey (CGSS), employing a stratified sampling approach that encompassed approximately 100 representative counties throughout China.¹ This survey collected individual-level data on happiness, which proves highly advantageous for analyzing underlying mechanisms given its micro-level nature.

This study matches the happiness in CGSS with the number of extreme high temperature days in the county in the current year and analyzes the impact of extreme high temperature on happiness. In an empirical strategy, the extreme weather in districts and counties is mainly affected by climate change caused by the increase in global emissions. Therefore, we can approximate the number of days experiencing extreme heat as an exogenous variable. Baseline results show that extreme heat reduces individual happiness. Specifically, each reduced day of extreme heat results in a 0.005 unit increase in happiness. Avoiding all the extreme high temperatures during that year then leads to a 4.883% increase in happiness. Using different extreme high temperature measures and model settings still allows us to obtain relatively robust results. Further analysis revealed how environmental inequalities are caused by extreme weather. We also found that as income increased, the negative effects of extreme weather were significantly reduced. Homes with air conditioning, unsurprisingly, effectively relieve extreme heat. Further mechanistic analysis found that extreme heat affects individual happiness mainly by reducing income levels, health status, and social activities with friends. Combined with the heterogeneity analysis, extreme high temperature was found to have a more obvious impact on low-income groups which is further amplified by income and health. This then forms a circular cumulative effect, ultimately further amplifying environmental inequality.

The contributions of this study are as follows. First, it considers the impact of extreme weather on individual happiness from the perspective of individual happiness. On the one hand, the study fills the gaps made by previous scholars’ neglect of individual happiness from the perspective of the environmental Kuznets curve. On the other hand, it further explores the influencing factors of happiness and provides support for climate change mitigative policies. Second, our study further explores the unequal impact of extreme heat and its amplification effect, thus providing a new perspective for studying environmental inequality. Previous studies and analyses mainly discussed pollution inequality. Pollution distribution is closely related to local industries which results in many environmental migrants and avoids the further amplification of the effects of pollution (Qin and Zhu, 2018). While the impact of extreme weather is relatively exogenous, this study provides a new perspective for the study of environmental inequality. Third, our research provides a reference for the formulation of environmental policy. Our study finds the negative effects of extreme heat and highlights the unequal effects of climate change. Therefore, policymaking should also consider the inequalities caused by climate issues.

The rest of the arrangement is as follows. The following section discusses the literature related to happiness and extreme temperatures. Section 3 explains the source of the data. Section 4 discusses the study’s empirical strategy. Section 5 reports the results and discusses the associated heterogeneity and mechanisms. The final section provides the study’s conclusion.

2. Literature Review

In this section, we will begin by examining the definition of happiness and its influencing factors. Subsequently, we will analyze the adverse effects of extreme weather and their significant influence on shaping happiness. Lastly, we will offer a succinct overview of the current literature in this field.

Improving human happiness is the ultimate goal of economic growth (Oswald, 1997), which explains the increasing scholarly attention on the intersection of psychology and economics. Easterlin (1974) made a breakthrough in exploring the relationship between income and happiness in different countries by finding no connection between a society’s economic development and its average level of happiness, which is also known as the “Easterlin Paradox”. Using multiple rich datasets spanning decades, Addoum et al. (2020) reassessed this paradox and instead established a clear positive relationship between average happiness and GDP per capita across countries. On the one hand, there are many quantitative dimensions of happiness (Frey, 2018). These various dimensions are mainly reflected in the subjective sense of experiential happiness and have nothing to do with income, socioeconomic status, and health. On the other hand, there is also the idea of an objectively assessable happiness, which reflects the quality of life. This viewpoint is what current studies often use, employing participants’ answers to certain questions to evaluate happiness (Graham and Nikolova, 2015). For example, based on a random sample of 1 million Europeans and Americans from the 1970s to the 1990s, Tella et al. (2003) found that the happiness index was closely related to macroeconomic variables, concluding that recessions take a greater toll on people’s mental health compared to the decline in GDP and the rise of unemployment.

Numerous factors also influence happiness. Most scholars agree on the existence of a U-shaped relationship between age and happiness (Frijters and Beatton, 2012). Education level has also been found to be highly correlated with happiness (Nikolaev, 2018). Employed women and women with higher incomes also have higher levels of happiness (Arrosa and Gandelman, 2016). Zimmermann and Easterlin (2006) concluded that the cohabitation life formed by a marital relationship has a significant positive impact on happiness following data from German residents living together. Interestingly, religious belief is often associated with higher happiness (Ash, 2007).

The increase in the frequency of heat waves has also led to the growing concern on how extreme heat affects people. Existing literature has conducted in-depth research from two aspects: Physiological impact and psychological impact. For physiological effects, exposure to heat can cause physical discomfort and, in severe cases, heat stroke, acute symptoms, and even death (Deschenes, 2014; Kenney et al., 2014; Xu et al., 2016). Michelozzi et al. (2009) found that high temperatures increase the rate of hospitalizations for respiratory diseases, especially in the elderly. Psychological effects mainly include emotional distress (Mullins and White, 2019), aggression (Anderson, 1989), cognitive impairment, and mental illness (Zhang et al., 2023). The body’s adverse reactions, in turn, lead to the reduction of people’s social happiness. Li et al. (2016) found that exposure to high temperatures reduces people’s labor productivity and their cognitive levels. Using anger and attack counts from Australian Twitter posts, Stevens et al. (2021) found an increase in offensive articles as temperatures rose. Heyes and Saberian (2019) used immigration decision-making data from US judges and found that exposure to extreme heat affected the cognitive output of high-quality decision-makers. The experiment of Keller et al. (2005) found that hot weather caused participants to feel depressed, which coincides with findings regarding seasonal affective disorder.

People are likely to experience different emotions in their natural environment affecting people’s psychological conditions (MacKerron and Mourato, 2013). Studies on the impact of environmental change on happiness is gradually enriched (Zapata, 2022). Van Praag and Baarsma (2005) found that airport noise has a significant negative impact on people’s happiness, focusing on data from the Amsterdam Airport. Pollution is considered to be an important factor which affects people’s level of happiness (Li et al., 2014). Rehdanz and Maddison (2005) sample consisting of 67 countries found that self-reported happiness levels were affected by climate change through causal identification following changes in temperature and rainfall. Some studies have also focused on the effects of exposure to extreme heat on people’s happiness. Zapata (2022) used a cross-sectional analysis to determine the effects of temperature, precipitation, and humidity on happiness in Ecuador and showed that climatic conditions constitute an important determinant of people’s happiness.

In conclusion, happiness serves as a crucial indicator of human aspirations. However, the exploration of temperature extremes and their effect on happiness has received insufficient attention in previous studies. The substantial decline in individuals’ happiness has a profound impact on social welfare. Scholars have consistently emphasized the significance of examining the effects of climate change on social happiness, particularly in the context of China — a developing nation with a population of 1.4 billion. The effects of extreme environmental conditions on social happiness in such a populous country can be substantial and warrant comprehensive investigation. Moreover, researchers have primarily focused on the direct implications of the natural environment on mental health, rather than the unequal distribution of losses resulting from climate change. The impact of extreme environments on individuals can be mitigated through adaptive behaviors adopted by different socioeconomic groups. For instance, higher-income groups can make adaptive investments, such as purchasing vehicles for commuting or installing air conditioners to counter the effects of extreme heat. Consequently, a fundamental question arises: Does shielding oneself from extreme environmental exposures alleviate the adverse effects of extreme heat on mental health? These crucial issues necessitate further in-depth research, building upon existing literature. Therefore, our study aims to explore the influence of frequent high temperatures on individuals’ happiness by employing a China-specific questionnaire, considering variations in heat sensitivity among different groups, and investigating the implications of climate change. Finally, we conducted a preliminary analysis of the underlying mechanisms involved.

3. Data

Our data mainly explore two aspects: The happiness and individual characteristic variables at the individual level from the CGSS, and county-level weather variables from the National Meteorological Science Data Sharing Service Platform of China.

3.1. Happiness and individual-level variables

Happiness data and individual-level variables were mainly extracted from CGSS.² The CGSS is a database that is organized and implemented yearly by Renmin University of China since 2003. It conducts a continuous cross-sectional survey of more than 10,000 households in various Chinese provinces, municipalities, and autonomous regions. To ensure a representative sample, the CGSS employed a three-stage stratified random sampling approach. The survey designated Shanghai, Beijing, Guangzhou, Shenzhen, and Tianjin as mandatory strata, while other cities were considered sampling strata. Approximately 2000 participants were selected from the mandatory stratum, and around 10,000 were sampled from the sampling stratum. The sampling process consisted of selecting streets in the first stage, village committees and neighborhood committees in the second stage, and households in the third stage. The resulting urban-rural ratio and other indicators closely align with macro survey data for China, confirming the representativeness of the sample size.

The survey mainly includes social structure, quality of life, and the internal connection mechanism between both. Its section on social structure contains data on individual and family socioeconomic characteristics, social class, family, and social relations. Meanwhile, its quality-of-life section includes some characteristics at the health, population, psychological, socioeconomic, and political or community levels.

The section on individual mechanism mainly includes the individual’s psychological and cognitive conditions, while the interpersonal mechanism mainly includes interpersonal relationships. The organizational mechanism section includes three aspects: Family, community, and work unit, and the institutional mechanism includes three aspects. Multi-stage stratified surveys with probability proportional to the sampling size were mainly adopted as the study’s sampling method. A total of 100 districts and counties and five megacities were selected for the study’s sample. Because the sample size is large and the questionnaires are more tolerant after the year 2010, we thus used data from 2010 to 2015 for primary analysis.

In the questionnaire, happiness was measured using the following question: “Generally speaking, do you feel happy or not?” to which participants answered using a 5-point Likert scale ($1=\mathrm{\text{Very Bad}}$; $2=\mathrm{\text{Bad}}$; $3=\mathrm{\text{Fair}}$; $4=\mathrm{\text{Good}}$; $5=\mathrm{\text{Very Good}}$). There are many ways to measure happiness: Some studies that specifically measure happiness are analyzed by scale (Deaton and Stone, 2016). Although this method provides a more comprehensive and accurate happiness index, this kind of questionnaire is unsuitable for large-scale questions because it leads to a small number of samples and fewer variables of individual socio-economic characteristics. Another more common method is to instead use a comprehensive questionnaire. This method directly asks about the evaluation of happiness, but it often only has a few questions and lacks a more detailed classification. However, it obtains more socio-economic characteristics. Thus, more topics can be discussed. This study uses a comprehensive survey to analyze the data herein. It is critical to note happiness and utility are very similar concepts. Utility is measured in terms of ordinal utility and cardinal utility. The measure in this paper is very similar to the cardinal utility. It cannot simply be assumed that individuals with a happiness score of 2 are twice as happy as those with a happiness score of 1. Therefore, we need to be cautious in our interpretation.

Other control variables include gender, age, the square of age, health, nationality, religion, party, marriage, hukou and education. Many studies have shown that these variables affect happiness, which will be discussed in further detail in the section on empirical strategy. Therefore, in terms of data, we coded male as “1” and female as “0”. Age was coded a continuous variable, and because age is usually nonlinear effects, we further calculated its square term. Health was coded as a discrete variable with values from 1 to 5; a higher value meant a higher degree of health. Because the study includes various nationalities, those who were of Han descent were coded “1”, while those from other groups were coded as “0”. In terms of political characteristics, we coded party members as “1” and “0” for those who were otherwise. For marital status, “1” indicated single, while those with another civil status were coded as “0”. China has adopted a household registration system, which divides individuals into both rural hukou and urban hukou. Those who were part of the rural hukou were coded as “1” while those from the urban hukou as “0”. Finally, in terms of educational level, the value ranges from 1 to 6, and the higher the level of education, the greater the value.

3.2. Weather variables

Weather variables mainly contain six variables, namely temperature, sunshine duration, relative humidity, pressure, precipitation, and wind speed. In the original data, these data are included in the monitoring data of the site. There are mainly the following methods to obtain data at the county level. First, the direct average of observations in the county. The disadvantage of this method is that some smaller counties may not be assigned to monitoring points. Some of the larger counties have fewer monitoring points, and a few monitoring points to measure the climate of the whole county may lead to bias. Second, the observation point closest to the sampling point was used as the proxy value. This method may cause a large bias because of the distance. To avoid the bias caused by both abovementioned methods, the observed values of monitoring points were linearly interpolated into a grid according to administrative division and then converted into raster data. The average value of each county containing the grid value was then calculated. To avoid the problem of excessive smoothing, only five sites near each grid point were used for interpolation and weighted according to geographical distance. Finally, we obtained the daily weather variables at the county level.

After getting the daily weather variables at the county level, the problem to consider is how to calculate the extremely high temperature. Measuring extreme high temperature faces several difficulties: First, the use of fixed thresholds may overlook regional differences. Some studies thus directly use a temperature higher than a certain temperature as the critical value of extremely high temperature (Deschênes and Moretti, 2009). However, the use of the same threshold for different regions may lead to overlooking the perceptions of extreme high temperatures in different regions. The perception of extreme temperature varies from region to region. For example, the extreme high temperature threshold is lower in cold areas and higher in hot areas (Addoum et al., 2020).

Second, individuals will adopt a series of adaptive behaviors, leading to a significant difference between the observed temperature and the actual exposure temperature. A typical example includes reducing the effects of extreme high temperatures by buying air conditioners (Biardeau et al., 2020). We took the 90th percentile as the threshold according to the temperature distribution of each county in the past 20 years. The temperature of the day above the threshold was defined as “1”, and the temperature below the threshold was defined as “0”. We calculated the annual cumulative value as the exposure to extreme high temperatures in the county. Measuring extreme high temperature allows for the threshold of extreme temperature to be different in different regions. In addition, the heterogeneity analysis combined the number of household air conditioners in the survey data to analyze the impact of air conditioning. Additionally, because of the subjectivity of using 90th percentile as the threshold, we further used 95th percentile and 99th percentile as the threshold in the robustness checks.

3.3. Summary statistics

After obtaining and matching the data, a total of 51,245 observations were collected. Table 1 reports summary statistics of key variables. The extreme temperature with the 90th quantile as the threshold is the core variable we focus on. The mean of the sample is about 37.20, accounting for 10.19% of the year, which indicates that the heatwave in the year of the sample is slightly higher than average. In areas with high heatwave incidence, 62 days of extreme high temperature occurred during the sample observation period, accounting for about 16.99% of the entire year. Areas with fewer heatwaves experienced 15 days of extreme heat during the sample observation period, which was about 4.11% of the year.

Generally, the differences in the appearance of heatwaves are relatively large, which is helpful to identify the effects of heatwave on happiness. The average level of happiness was about 3.809, and most individuals had a higher level of happiness. The gender ratio was about 48.8% and was relatively balanced. The average age was 48.70 years old, the minimum age was 17 years old, and the maximum age was 102 years old. The sample was mainly concentrated in adults. The average level of health was 3.532, indicating a high overall health level. Those of Han descent accounted for 91.7% of the total sample. 87.4% did not subscribe to a religion. Those who were single accounted for about 9.7%, and the education level was at a roughly lower middle school degree. The average duration of sunshine was 5.26h, accounting for about 21.92% of an entire day, indicating that the samples have more rainy weather. The relative humidity was about 69.04%. Pressure was about 973.5hPa. The average daily precipitation that year was 2.821mm, and the average wind speed was about 2.079m/s.

4. Empirical Strategy

To estimate the effects of heatwave on happiness, a fixed effect model was used for analysis :

where ${\mathrm{\text{Happiness}}}_{it}$ represents the happiness of individual i in county j at t year. ${\mathrm{\text{Heatwave}}}_{jt}$ is the cumulative number of days of extreme high temperature in county j at t year. $X_{ijt}$ references control variables at the individual level, including gender, age, the square of age, health, nationality, religion, party, marriage, hukou, and education. ${\mathrm{\text{Weather}}}_{jt}$ denotes weather control at the county level, including sunshine duration, relative humidity, pressure, precipitation, and wind speed. The vector ${\theta }_j$ is county fixed effects to absorb the inherent attributes that do not change with time at the county level. The vector ${\rho }_t$ is year fixed effects to capture the effect on happiness at the national level. ${ϵ}_{it}$ is the error term. Referring to Abadie et al. (2023), our explanatory variable is county-level variation. Therefore, standard errors are clustered at the county level.

${\beta }_1$ is a parameter that is crucial in evaluating the effects of heatwaves on happiness. To obtain the consistent estimator of ${\beta }_1$, that is, $\mathrm{\text{cov}}({\mathrm{\text{Heatwave}}}_{jt},{ϵ}_{ijt})=0$, the following aspects were considered: First, heatwave is a relatively exogenous variable. The main cause of extreme weather is global warming, which is then caused by massive anthropogenic emissions of greenhouse gases (Stott, 2016). CO2 do not play a single role in a county, but the global climate risk is a community. In other words, greenhouse gas emissions from various regions will eventually have an effect on climate change on a global scale. This is because the carbon emissions generated by the socio-economic activities of a county are negligible when compared with global carbon emissions and because the impacts of individuals on global carbon emissions are negligible. Therefore, the occurrence of extreme weather can be regarded as a relatively exogenous shock. In addition, many literatures also regard short-term weather changes as an exogenous shock (Dell et al., 2014; Heal, 2017).

Second, we included many individual control variables that affect happiness. Frey and Stutzer (2002) summarized the main factors that affect happiness, namely gender, age, the square of age, health, nation, religion, party, marriage, hukou and education to the control variables. Income is also an important factor that affects happiness, but because the variable of income in our sample is too small, we thus considered the impact of income in the robustness checks. Many studies have pointed out the necessity of adding these control variables. For example, Toshkov (2022) illustrates the relationship between gender, income, and happiness. Frijters and Beatton (2012) analyze the nonlinear relationship between age and happiness. Ifcher et al. (2018) discussed and analyzed the relationship between health and happiness. Field Zhang et al. (2017) considered these important control variables directly in the econometric model.

Third, we further consider other weather variables that are easy to affect extreme high temperatures. Although extreme high temperatures are relatively exogenous, there may be a correlation between temperature and other weather variables. Most of China lies in the monsoon region, hence there is a strong correlation between precipitation and temperature (Wu et al., 2018). We added other weather control variables to avoid extreme high temperatures to be affected individual happiness, which is by affecting other weather variables. However, some studies have pointed out that precipitation as a control variable may affect the estimated result (Auffhammer et al., 2013). Therefore, we excluded precipitation from the robustness checks.

Fourth, we added the fixed effects of counties and years to alleviate the problems caused by omitted variables. The common two-way fixed effect model was used to absorb the factors that do not change over time at the county level along with the impact of shocks at the national level on happiness (He and Wang, 2017). Notably, our analysis does not include the socio-economic characteristics at the city level; this includes GDP, population density, and industrial structure. To dispel this concern, we further added the city by year fixed effects in the robustness checks to absorb city socio-economic characteristics which change with the year.

Finally, our standard error was subjective to some extent. As the core variable of concern is at the county level, we clustered the standards to the county level. This specification assumes that there is a certain correlation between the samples in the county but does not exist in different counties. This assumption may be too strict — which is why we further relaxed the setting of standard error in the robustness checks, considering more abundant specifications.

5. Results

5.1. Baseline results

Benchmark regression was performed according to the model settings seen in Sec. 4. Regression results are shown in Table 2. Column 1 of Table 2 is a simple-form regression of the core explanatory variables and the happiness index. There is a significant negative relationship between extreme heat and happiness, suggesting that frequent heat waves can lead to a decrease in happiness. Column 2 further adds individual- and county-level weather controls, and standard errors are clustered at the county level. Column 3 is the preferred specification for adding control variables and fixed effects. We controlled for city fixed effects and year fixed effects. Results showed that for each additional day of extreme heat, people’s happiness index decreased by an average of 0.005. Table 2 clearly shows that the average accumulated time of people’s exposure to high temperatures is 37.2 days per year, and the average happiness index of people is 3.809. Therefore, we believe that exposure to extremely high temperatures leads to an average decrease of 0.186 points in people’s happiness, thereby accounting for an average of 0.186 points. 4.883%³ of happiness level.

Regarding the control variables, we observed notable gender disparities in happiness, with males reporting lower levels of happiness compared to women. The association between happiness and age exhibited a U-shaped pattern, potentially attributable to reduced work-related stress among older individuals and children. Furthermore, noteworthy variations in happiness were identified across ethnicity, party, and marital status. Notably, an increase in educational attainment emerged as an effective means to enhance overall happiness levels.

5.2. Robustness checks

This section reports a series of robustness test results. Regarding the subjectivity of the setting, we analyzed the results from five aspects: Sensitivity analysis of heatwave setting, substitution of fixed effects and control, substitution of cluster standard errors, substitution of model, and more controls.

According to the definition of different heat wave duration, this study constructs two core explanatory variables to explore whether the heat wave setting is sensitive. Column 1 is the preferred specification, and the heat wave duration is defined as the temperature threshold of the upper 90% quantile in the historical panel (lasting 20 years). The number of high temperature days in the city exceeding this threshold in a year was calculated. The second and third columns set the thresholds to 95% and 99%, respectively. The regression coefficients of focus in Table 3 are all significantly negative and have little change, meaning that the impact of extremely high temperature on happiness will not have major changes due to the difference in our threshold setting for high temperature.

In Table 4, we added the yearly fixed effect of the city based on the preferred paradigm, which was presented in column 2. Because China is a country with a wet/monsoon climate, there is a high correlation between rainfall and temperature, which coincides with the points earlier raised. To address our concerns about multicollinearity, we eliminated the rainfall variable and carried out a robustness check in column 3. Additionally, rainfall conditions and income conditions directly affect people’s happiness (Jebb et al., 2018; Lin et al., 2020; Zapata, 2022). Therefore, we removed the rainfall variable in columns 3. Results show that our transformation of the preferred specification did not significantly change the results.

We performed a more robust test to ensure heterogeneity in the magnitude and significance of the influence coefficients based on alternative clustering standard errors. Results of which are shown in Table 5. Column 1 provides standard errors that are unclustered and found no meaningful changes in their size. The standard errors of columns 2 to 4 are clustered at the level of year, city, and year by country, respectively. These were clustering at what could be considered as treatment level (for example, the happiness index will be affected by the city where residents live or the development of cities in different years). In summary, the abovementioned analysis did not affect our conclusions.

Our dependent variable was an ordinal discrete variable ranging from 1 to 5, with order logit or order probit being more suitable for estimation. Columns 1 and 2 of Table A.1 report the estimation results for order logit and order probit, respectively. The results show that extreme heat significantly reduces happiness, which is consistent with our baseline regression results. Since income has many missing values, we did not include income as a control variable in the baseline regression. But income is a significant variable affecting happiness. Table A.2 further reports the results after including income as a control variable. Although some samples were lost, it did not change our conclusion.

Our dependent variable was an ordinal discrete variable ranging from 1 to 5, with order Logit or order probit being more suitable for estimation. Columns 1 and 2 of Table A.1 reported the estimation results of order Logit and order probit, respectively. The results show that extreme heat significantly reduces happiness, which is consistent with our basic regression results. Since there are many missing values in income, we did not include income as a control variable in the basic regression. But income is an important variable affecting happiness. Table A.2 further reports the results after including income as a control variable. Although some samples were lost, it did not change our conclusion.

5.3. Heterogeneity analysis

This section mainly focuses on two perspectives: The heterogeneity of the surrounding environment and the influence of the heterogeneity of personal characteristics. We first analyzed whether differences in rainfall and relative humidity led to differences in effects. We then explored whether residents’ personal income, household income, age, education, and air-conditioning usage had differential effects on the sensitivity to extreme heat.

Table 6 presents the subsample regressions of people’s exposure to different amounts of rainfall. The situation of rainfall directly affects people’s actions. First, it affects people’s propensity to go out, such as outdoor travel and outdoor fitness. Second, rainfall also reduces the adverse effects of extreme heat on people and alleviates heat exposure. Based on the distribution of rainfall, we performed a subsample regression of annual rainfall according to the quartile of the sample. Columns 1 to 4 represent the classification of annual rainfall levels from low to high, respectively. We found that the effect of heatwaves on people’s happiness decreases with increasing rainfall (the coefficient gradually decreases until it is not significant at the end). Residents in areas with high rainfall are less sensitive to heat and are thus less affected by high temperatures.

People’s perception of heat is determined by the combined action of temperature and humidity (Berglund, 1998; Tian et al., 2011). Table 7 classifies residents exposed to different humidity levels based on ambient relative humidity data, which replicates the division found in Table 6. Columns 1 to 4 in Table 7 represent residents living in areas with low relative humidity to those areas with high surrounding humidity. We found that the detrimental effects of extreme heat on people’s happiness were significant in subsamples with relative humidity not exceeding the upper quartile. Meanwhile, residents’ susceptibility to heat exposure at high relative humidity, at the 75th percentile and above, did not have a significant effect on people’s happiness.

In this subsection, we employed interactive items to analyze extreme heat heterogeneity effects on individual characteristics. Specifically, columns 1 and 2 of Table 8 employed the logarithm of family income and individual income to construct an interactive item with extreme heat. The results showed that the interactive item between income and extreme heat was positive and statistically significant at a 5% level. This was contrary to the extreme heat coefficient. Therefore, increasing income can mitigate the negative impacts of extreme heat. As individuals’ incomes rise, they have more resources to deal with extreme heat. Subsequently, we examined heterogeneity differences in age and educational level in columns 3 and 4, respectively. However, no obvious differences were observed. In column 5, we looked at the influence of family air conditioning numbers. Air conditioners can significantly improve temperatures. The results showed that the interaction coefficient between air conditioners and extreme heat was positive and significant. Thus, buying air conditioners can effectively reduce the negative impacts on happiness caused by extreme heat. In summary, individuals with higher incomes or access to air conditioning are better able to alleviate the negative effects of extreme heat. Moreover, higher incomes are often positively correlated with air conditioners. In addition, extreme heat further decreases income levels, which then amplify the negative impacts of extreme heat. Therefore, income inequalities will be exacerbated by extreme heat.

5.4. Mechanism analysis

This subsection explores several possible pathways where heat sensitivity affects people’s happiness. The level of income is affected by extreme temperatures, and the income of those who work in heat-sensitive industries is directly related to temperature. For example, the income of farmers is directly affected by the temperature of the year. Therefore, we first performed regression analysis on the effects of extreme temperature on people’s personal and household income levels. Results are shown in columns 1 and 2 in Table 9. We found that extremely high temperature has a significant effect on people’s income. The more days with extreme heat there were in a year, the lower the income level of people, leading to less investment in heat adaptability.

We also explored the potential impact mechanism from two dimensions: People’s health and social status. Many studies have investigated the effects of extremely high temperature on people’s health. Consequently, we believe that physical health will also affect people’s mental health. First, the discomfort experienced by the body directly increases people’s negative emotions. Second, physical health affects people’s behavior and indirectly affects people’s psychological state. This includes the reduction of leisure methods such as going out for fitness and traveling. Results of column 3 of Table 9 shows that heatwaves have adverse health effects, and that this intuitive effect can be a direct cause of mental health impacts. Column 4 explores whether people’s social activities are a potential mechanism of action. When people are exposed to extreme heat, they spend less time going out and reduce social interactions, which can be detrimental to our release of negative emotions. Groups living in hotter areas were found to have reduced their frequency of seeing friends, thus reducing the amount of time residents were exposed to heat.

6. Conclusion and Policy Implications

Obviously, extremely high temperature directly affects people’s health, and measuring people’s happiness is an important topic in welfare economics. Based on the CGSS database, we explored the impact of extreme high temperature on residents’ happiness given the effects of global warming due to climate change. We controlled for personal and environmental characteristics that may affect people’s level of happiness. After a series of tests, we found that people’s average annual exposure to high temperature accumulated time led to a 4.883% decrease in happiness, implying that the surrounding loss of people’s welfare from high temperature is realized by reducing the sense of happiness. A heterogeneity analysis of the different populations in terms of thermal sensitivity was then performed, which found that residents living in areas with high rainfall were less sensitive to heat exposure and thus, less affected in mental health than those living in areas with high rainfall. Differences in age and education level were also not the main reasons for the heterogeneity of heat sensitivity in different groups. Our results coincide with related theories of environmental adaptability. The use of air conditioners was also found to effectively alleviate the thermal sensitivity of ambient temperature, which is helpful in understanding and discovering the extent people’s adaptive behaviors alleviate outdoor temperatures. Finally, we analyzed the mechanism mainly from the perspectives of income, health, and social interaction. We found that the frequent occurrence of extremely high temperature led to a decline in people’s income status, physical health, and social frequency, thus reducing their happiness.

This study provides new supporting data and research perspectives in studying the impact of climate and environment on people’s happiness. In the current context of global warming, it is helpful to analyze and understand people’s adaptive behaviors and channels to the thermal environment, and ultimately quantify the degree of damage to people’s mental health and happiness caused by extreme heat. Our results also map the research on environmental inequality, where different groups are unfairly affected by the environment due to differences in income, geographic location, nature of work, and other factors. Data herein found that people living in areas with high rainfall and high incomes are less sensitive to heat than people in income areas, thus reflecting the inconsistency of said environmental effects.

Our research has the following policy implications. First, effective measures need to be taken to reduce greenhouse gas emissions, to mitigate the negative impacts of extreme weather caused by carbon emissions. Our research results have shown that extreme weather caused by greenhouse gas emissions has already had a negative impact on residents’ happiness, which can affect individuals through multiple channels. Therefore, governments need to take various measures to reduce carbon emissions. Additionally, while carbon emissions vary among countries, all face the risks of global climate change. Collaborative governance among countries is needed in the process of reducing carbon emissions. Second, relevant departments need to adopt related healthcare policies to reduce the health losses caused by extreme weather. Our research found that extreme weather mainly affects individuals’ happiness through health channels. Relevant departments can reduce the health losses caused by extreme weather through intermediate channels, especially by improving and expanding health insurance coverage. Finally, relevant departments can consider increasing subsidies for air conditioning and other equipment to expand air conditioning coverage. Our research found that households with air conditioning can significantly reduce the negative impacts of extreme temperatures. In addition, governments can also take more preventive measures and subsidies, such as providing heat subsidies for employees in some companies.

Acknowledgments

We thank the support provided by the Natural Science Foundation of Guangdong Province of China (Grant # 2021B1515020103).

Notes

¹ See http://cgss.ruc.edu.cn/xmwd/cysj.htm for details.

² See http://cgss.ruc.edu.cn/xmwd/cysj.htm for details.

³ 4.883%=37.2*0.005/3.809.