1. Introduction
The potential impact margins of climate change are wide and far-reaching (Deschênes 2014). The University College London (UCL)–Lancet Commission on Managing the Health Effects of Climate Change called climate change “the biggest global health threat of the 21st century” (Costello et al. 2009). Fossil fuels have led to substantive emissions of greenhouse gases since the industrial age, which result in a dramatic increase in average temperatures and more frequent extreme weather events (Zivin and Shrader 2016; Oppenheimer and Anttila-Hughes 2016) where the daily temperatures exceed the upper and lower limits of a certain range (Yan et al. 2002; Zhang et al. 2011). Extreme temperature events are expected to become more frequent and intense as climate changes continue, at least in certain regions (Chen et al. 2014). The damage of extreme heat and cold to human health has been recognized as one of the most important areas of concern and become a focus of global research (Deschênes 2014; Yang et al. 2019a).
Abnormal temperature is a threat to physical health. Recent studies have analyzed the impact of temperature on mortality and morbidity (Deschênes and Greenstone 2011; Allen and Sheridan 2015; Marsha et al. 2018; Yang et al. 2019a) and extreme temperatures lead to significant reductions in health through various channels such as inducing cardiovascular and respiratory diseases (Curriero et al. 2002; Zivin and Shrader 2016). Climate change also creates mental health challenges. A growing number of studies point out the link between mental health and environmental factors through biological mechanisms. Human emotions such as happiness are influenced by daily weather (Baylis 2020).
From the perspective of macro–health evaluation, we find consistent conclusions in the literature where mortality and morbidity are used to measure the health level (Curriero et al. 2002; IPCC 2007; Yang et al. 2019a). The body’s heat regulatory function enables us to cope with exposure to high and low temperatures. However, this adaptation increases the stress on many organs, and when the human body experiences excessive temperature, health will be impaired (Deschênes 2014). For physical health, heat exhaustion and heat stroke are the most serious illnesses caused directly by extreme heat. The body may overheat, leading to prostration, dizziness, and muscle cramps (Bouchama and Knochel 2002; Qu and Xiao 2019). In cold weather, the direct health impact of hypothermia is more likely to happen, and exposure to cold indirectly leads to frostbite, pneumonia, and influenza (Qu and Xiao 2019). When it comes to anomalous cold events, such as heavy snow or frost, people might fall, especially the elderly (Hajat et al. 2016). Exposure to cold and heat indirectly increases the risk of cardiovascular and respiratory disease. Studies have shown that every 1°C increase above 29°C leads to about 3% more adult hospitalizations due to respiratory disease; the incidence of cardiovascular disease also increases in many cases (Michelozzi et al. 2009; Lin et al. 2009). In addition, for infectious diseases such as malaria, enteric fever, and diarrhea, extreme heat also increases the risk (Chowdhury et al. 2018). The Lancet health report (Watts et al. 2015) shows the complex mechanisms of rising temperatures and changes in precipitation patterns alter the viable distribution of disease vectors such as mosquitoes carrying dengue or malaria, which increases the incidence of these infectious diseases. On the other hand, due to the role of certain neurotransmitters (such as biogenic amines) in both emotional and thermal regulation, mental health might also be affected adversely (Iversen 1982). There is also a link between hot temperatures and anxiety, aggression, and suicide risk based on previous studies (Maes et al. 1994; Anderson and Anderson 1998; Hansen et al. 2008).
Our research addresses three gaps in the existing literature. First, most studies have focused on the extreme heat effects (e.g., heat waves) on human health under global warming. Only a few studies consider the role of extreme cold events (Hajat et al. 2006; Yang et al. 2019a). The relationship between extreme cold temperatures and human health is not well explored. Second, previous studies focus on mortality and morbidity indicators and largely ignore the assessment of human daily health status, partly due to the lack of availability for credible and large-scale empirical data. In addition, as mental illness has become the major driver of the global burden of disease (Murray et al. 2015), many psychological mechanisms make an epidemiological linkage between environmental factors and mental health biologically plausible (Xue et al. 2019). More research is needed on the relationship between mental health and external environmental factors. Third, most studies are carried out in developed countries, and little research is conducted in developing countries with substantial regional differences. Developing countries have lower income levels, less access to health care, and more difficult living conditions. The response patterns to extreme temperature events in developing countries may differ significantly from developed countries. As a member of the developing countries, China has a huge population covering a wide region with substantial regional climate differences, which is suitable to investigate the complicated relationship between humans and the environment.
We use both physical and mental health indicators based on the Chinese Family Panel Studies (CFPS) national survey data from 2010 to 2016. The CFPS survey includes basic sociodemographic indicators such as age, gender, and income, allowing heterogeneous analyses across subpopulations. After matching the samples with the climate data from 839 meteorological stations in China spatially and temporally, we obtained 98 423 individual-level observations. We find that the relationship between temperature and overall health, physical health, and mental health all exhibited an inverted U-shaped curve, which suggests that poorer health conditions were more likely to occur on days with very high or very low temperatures. We also explored the impact of 10 extreme weather indices such as heating degree-days, warm nights, and cool days on residents’ health. In addition, we investigated the impact of extreme weather among different gender, age, and income categories. Our results reveal that women and low-income households are more likely to be influenced by extreme cold, while men, the elderly, and high-income households are more sensitive to extreme heat.
2. Data
a. Data sources
We use the most comprehensive individual health and climate data available in China based on a national-scale survey and the meteorological networks. The key dependent variable, residents’ health status, comes from the CFPS project hosted by the Institute of Social Science Survey at Peking University. The survey has a wide range of sociodemographic indicators and extensive geographical coverages (Yao 2021). The CFPS survey is conducted every two years since 2010 and collects information on individuals, families, and communities about economic activities, social attitudes, health, and other variables for more than 100 000 residents in 162 counties in 25 provincial regions. The survey also records residents’ geographical location and time of the interview, providing gateways to match health status and other time-varying information related to temperature. Since personal health status is a complicated measure influenced by the individual’s physiological state, living habits, and social environment, we include various indicators such as physical health, mental health, and a wide array of demographic variables in the years 2010, 2012, 2014, and 2016. Because of the complexity and wide coverage of the investigation, the length of the survey may last for more than a year. As a result, our final dataset also includes individual samples collected in 2011, 2013, and 2015.
Temperature and precipitation data are compiled from the China Meteorological Science Data Sharing Service Network, which contains daily data from 839 basic meteorological stations in China, including pressure, temperature, precipitation, relative humidity, and other indicators since January 1951. Stringent data quality controls in the network ensure high availability and accuracy rates. Studies have shown that PM2.5 contributed to respiratory and heart diseases (Song et al. 2017) because fine particles can cross the blood-brain barrier and damage the neurological system (Jia et al. 2018). We collected the PM2.5 data from the global surface PM2.5 datasets established by the Atmospheric Composition Analysis Group and use the original value in the regression model to control the influence of air pollution (https://sites.wustl.edu/acag/datasets/surface-pm2-5/).
Figure 1 details the data assembling and matching process. We extracted 143 358 CFPS observations collected between 2010 and 2017. To create a panel dataset, we only keep observations that are included in the 162 counties from the 2010 survey, resulting in 134 425 samples. We further match the weather data to the individual respondent and exclude observations that cannot be matched. We excluded 33 650 observations with missing individual demographic characteristics. We are able to identify 98 423 individual-level observations for our research.
Data assembly and matching flowchart.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Data assembly and matching flowchart.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Data assembly and matching flowchart.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
b. Variable descriptions
1) Health
We focus on the effects of temperature on residents’ overall health, physical health, and mental health. We use self-reported data in the CFPS questionnaire to measure the residents’ daily health levels. For physical health indicators, CFPS asks respondents to evaluate their health status at the time of the interview by asking “What do you think of your health?” In comparison with other health indicators, the self-assessment method of health is more comprehensive and may reflect the diagnosis of diseases and early onset diseases that may be missed by doctors. The self-assessment data also account for the personal feelings about physical conditions in daily life and the personal understanding of family history (Idler and Benyamini 1997; Jylhä et al. 2006), which has been successful in predicting mortality and disability rates with good reliability and validity (DeSalvo et al. 2005; Qi 2014; Halliday et al. 2021). To standardize the subjective evaluation, examples and references were provided during the interview to differentiate the varying cut points applied by respondents.
In the 2010 questionnaires, the physical health options are healthy, fair, relatively unhealthy, unhealthy, and very unhealthy. In the 2012, 2014, and 2016 questionnaires, the options are changed to extremely healthy, very healthy, relatively healthy, fair, and unhealthy. In our study, we assign health indicators to corresponding quantitative values each year to measure physical health. We use a score of 3 to represent “healthy,” which corresponds to the options of healthy in 2010, extremely healthy and very healthy options from the 2012 to 2016 questionnaires, a score of 2 to represent “fairly healthy,” which corresponds to the options of the fair in 2010, relatively healthy and fair options from the 2012 to 2016 questionnaires, and a score of 1 to represent “unhealthy,” which corresponds to the options of relatively unhealthy, unhealthy, and very unhealthy in 2010, unhealthy options from the 2012 to 2016 questionnaires.
Biological mechanisms for maintaining body temperature suggest that mental health may be affected by ambient temperature (Iversen 1982). Following Zhang et al. (2017), we chose the frequency of respondents’ depression during a given period to measure depression. The frequency of depression measures short-term hedonic unhappiness status and is more directly related to the immediate environment and emotional state in daily life when compared with the life satisfaction indicators (Stone and Mackie 2014). A more frequent occurrence of depression reflects lower mental health status. In the 2010 and 2014 questionnaires, the options to measure frequency are “almost every day,” “often,” “half the time,” “sometimes,” and “never.” In 2012 and 2016, the options are changed to almost none (less than a day), sometimes (1–2 days), often (3–4 days), and most of the time (5–7 days).
Similar to the classification of physical health, we assign different scores to individual mental health status. The questionnaires in 2010 and 2014 ask the frequency of depression in the most recent month and the questionnaires in 2012 and 2016 investigate the frequency in the most recent week. We divide the options according to the actual proportion of depressed days as below. We use 3 to represent “healthy,” which corresponds to the option of never in 2010 and 2014, and almost none (less than a day) option from the 2012 and 2016 questionnaires. We use 2 to represent “fairly healthy,” which corresponds to the options of often, half the time, sometimes, in 2010 and 2014, and sometimes (1 to 2 days) option from the 2012 and 2016 questionnaires. We use 1 to represent “unhealthy,” which corresponds to the option of almost every day in 2010 and 2014, and options often (3 to 4 days), and most of the time (5 to 7 days) from the 2012 and 2016 questionnaires.
The scores of physical health and mental health are combined to represent the overall health level of the respondents. Therefore, the scores are in the range of 2–6. A higher score indicates a healthier status. To test the robustness of the results, we used principal component analysis (PCA) to measure the overall health status. Our results are consistent across different categorizations of health status.
2) Temperature
We match individual responses and weather stations to ensure that temperature data in different regions can be accurately linked with individual response data. We use data from the weather stations that are nearest to the respondent’s county to analyze the effects of temperature on the health status.
The weather measurements include the maximum, minimum, and average temperatures on each day (Fig. 2). To incorporate potential lagged effects, we calculate the average of these three temperature indicators for the month prior to each interview (the monthly average of the daily average temperature, the monthly average of the daily maximum temperature, and the monthly average of the daily minimum temperature). Moreover, two categories of extreme cold and extreme hot indicators are constructed to explore the effect of extreme weather on human health based on a series of temperature indicators. The cold temperature category includes the cold temperature frequency, heating degree-days, cool nights (TN10p), cool days (TX10p), and cold-spell duration indicator (CSDI), and the hot temperature category includes the hot temperature frequency, cooling degree-days, warm nights (TN90p), warm days (TX90p), and warm-spell duration indicators (WSDI). Detailed definitions of each indicator are shown in Table 1.
Temperature distribution (2010–16).
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Temperature distribution (2010–16).
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Temperature distribution (2010–16).
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Temperature indicators. The 10th and 90th percentiles of daily minimum temperature from 2010 to 2016 are −10.6° and 23.0°C, respectively, and the 10th and 90th percentiles of daily maximum temperature from 2010 to 2016 are 1.2° and 32.3°C, respectively.
To further analyze the health effect of extreme heat and cold, we divide the temperatures by Fahrenheit degrees into nine segments according to a 12.22°C (or 10°F) interval in the range from −12.11°C (or 10°F) to 32.22°C (or 90°F). Each temperature critical point (−12.11°, −6.67°, …, 26.67°, 32.22°C) is designated as a balance point temperature. We choose the daily temperature frequency index to calculate the proportion scale in which the temperature was higher or lower than the balance point temperature. For example, for the balance point temperature in the heat environment (e.g., 90°F), we calculate the monthly frequency in which the daily average temperature exceeds 90°F in the month before being interviewed. For the balance point temperature in the cold environment (e.g., 10°F), we calculate the monthly frequency in which the daily average temperature is below 10°F. As the balance point temperatures become more extreme, the frequency of days in the corresponding range will become lower, which implies a more extreme temperature on either end of the spectrum.
In addition, we use the cooling degree-days (CDD) and heating degree-days (HDD) defined by European Environment Agency to project climate change and its impacts on environmental, social systems, and human health (European Environment Agency 2020). CDD is the sum of cumulative degrees above the balance point temperature over a period of time; the corresponding HDD reflects the cumulative degrees of the daily average temperature below the balance point temperature. These indicators have been frequently used to estimate energy consumption (Ahmed et al. 2012; Shen et al. 2017; Alola et al. 2019; Yang et al. 2019b).
We also consider six indicators produced by the World Meteorological Organization (WMO), such as warm nights, cool days, and warm-spell duration indicators to measure extreme temperature. The nights and days indicators are constructed based on the observation data with cutoff points, while the warm-spell duration indicator and cold-spell duration indicator consider spatial and temporal heterogeneity. As proxy indices of extreme temperature, these indicators were widely used in the literature (Alexander et al. 2006; New et al. 2006; Choi et al. 2009; Zhou and Ren 2010).
3) Other control variables
We control for sociodemographic and health behavior variables, including age, gender, marital status, rural–urban residence, income, educational level, smoking, drinking, and physical exercise in regression models (Xu and Xie 2017). We use dummy variables to represent gender, marriage, and rural–urban residential settlements, as well as personal behavioral patterns such as smoking and drinking behavior according to reported frequency. We also take into account the effects of average precipitation and average PM2.5 on health.
3. Empirical strategy
4. Results
a. Summary statistics
Table 2 reports the summary statistics of our sample. Based on the general climate data statistics, the average temperature of all the counties included in our sample is 23.71°C (standard deviation SD = 5.32°C), which varies from −25.08° to 31.37°C. We also calculate that the means of the daily maximum and minimum temperatures are 28.66°C (SD = 5.07°C) and 19.86°C (SD = 5.74°C), respectively. The average values of precipitation and PM2.5 in the region are 5.00 mm (SD = 3.68 mm) and 52.61 μg m−3 (SD = 26.32 μg m−3), respectively.
Statistical descriptions (for the entire sample, N = 98 423).
The mean value of the aggregated individual health index is 4.14 (SD = 1.10). When divided into physical health and mental health, the means of individual physical and mental health are 2.18 (SD = 0.70) and 1.96 (SD = 0.80), respectively. The subjects in the CFPS survey range from 16 to 116 years old. The average age of the respondents is around 46.90 yr (SD = 16.52 yr). Nearly 81.39% are married, and 51.21% of the population are women. Most of the residents do not have a college diploma (junior college and above accounted for 7.41%), the families earn an annual income per capita of CNY 12,064 (SD = 20,881, around USD 1,743), and 46.85% of them live in the urban areas.1 From the lifestyle index data, we find that, among all the respondents, about 29.21% and 15.58% of the population have frequent smoking and drinking habits, respectively
b. Overall regression results analysis
Table 3 presents the regression results for estimating the overall effect of temperature on human health. The effects of average temperature, maximum temperature, and minimum temperature on health were shown in sequence (Table 3, models 1–3). We find that the influence of temperature on health presents a clear inverted U-shaped curve across different model specifications, indicating that both hot and cold weather produce negative effects on human health, while a moderate temperature is best for health. This result is consistent with our prior expectations (e.g., Yang et al. 2019a). The coefficient of the square of the average temperature and the average temperature are −0.000 254 and 0.009 495 in Table 3, model 1, respectively. We also calculate the inflection point of the curve. When the average temperature is less than 18.69°C, a higher temperature leads to more health benefits. When the average temperature is higher than 18.69°C, the increase in temperature leads to negative effects on health (Table 3, model 1). We classify nine balance point temperatures into two ranges, cold and hot temperatures, based on the 18.69°C benchmark.
Nonlinear relationship between temperature and health (the dependent variable is the health score). Significance levels p < 0.01, p < 0.05, and p < 0.1 are indicated by three asterisks, two asterisks, and one asterisk, respectively; robust standard errors are in parentheses; N = 98 423.
The impact of temperature on physical health and mental health are investigated separately in Table 4. Our regression results are consistent with the results of overall health, and we observed an inverted U-shaped relationship between the temperature and both physical and mental health. Our results show that the when the physical health is used as the dependent variable, the R square is smaller relative to the specification when mental health is used as the dependent variable, suggesting the current control variables better explain the variations for mental health. (In the appendix, Tables A1 and A2 present the coefficient estimates for the full set of control variables based on the specifications used in Tables 3 and 4, respectively).
Nonlinear relationship between temperature and health by symptom. Significance levels p < 0.01, p < 0.05, and p < 0.1 are indicated by three asterisks, two asterisks, and one asterisk, respectively; robust standard errors are in parentheses; N = 98 423.
c. Robustness checks
Table 5 presents results from additional robustness tests. The total health measurement scale ranged between 2 and 6, representing a limited dependent variable. We use the Tobit model and PCA method for robust analysis. Results are shown in model 2–3 of Table 5. The results are consistent with regressions in Tables 3 and 4, which confirm the nonlinear relationship between temperature and health.
Nonlinear relationship between temperature and health: robustness checks (the dependent variable is the health score). Significance levels p < 0.01, p < 0.05, and p < 0.1 are indicated by three asterisks, two asterisks, and one asterisk, respectively; robust standard errors are in parentheses; the health score: PCA = (0.5395 × the physical health score) + (0.4605 × the mental health score); the average temperature of distance aggregation = [(shortest distance × the average temperature of shortest distance) + (second shortest distance × the average temperature of the second shortest distance)]/(shortest distance + second shortest distance).
We also analyze alternative methods of matching meteorological station data to test the robustness of our results. We detect a nonlinear, quadratic relationship using alternative matching methods (Table 5, models 4–6). In Table 5, model 1, we match each respondent to the nearest meteorological station to link meteorological data and the individual. To address potential measurement errors and attenuation effects, we defined a cutoff radius of 60 km and eliminated the sample individuals who did not meet the conditions (Table 5, model 4). Alternatively, we used the weather data from the second nearest weather station and a temperature–distance-weighted index to measure this regression relationship (Table 5, models 5 and 6). We find alternative distance calculations have little influence on our main conclusions.
d. Health effects of hot and cold temperature
In this section, we construct a series of temperature indicators and evaluate if extreme hot and cold weather events have a more pronounced negative effect on health. Our main results are presented in Tables 6–8 and Fig. 3. In Table 6, the temperature frequency index is calculated as the proportion of days with temperatures in the designated range. Results show that both hot and cold temperatures will contribute to negative health effects. The negative impact gradually deepened as the temperature rises or drops farther. When the temperature falls below −12.22° or −6.67°C threshold, and if the cold duration increases by 1 percentage point, the health score will decrease by 0.29% or 0.14%, respectively (models 1 and 2 in Table 6). When the threshold is 26.67°C, a 1 percentage increase in the duration only decreases the health score by 0.09% (model 8 in Table 6). When the temperature rises above the 32.22°C threshold, a 1 percentage increase in the duration will lead the health score to decrease by 0.14% (model 9 in Table 6). Based on the method proposed by Joyce et al. (1989), and Chen and Shi (2013), we use differential transformation to evaluate the money metric value of temperature by calculating the average marginal rate of substitution between daily temperature and household per capita income, also known as “willingness to pay” (WTP).2 According to the formula, when the temperature falls below −12.22° or −6.67°C, and rises above 26.67° or 32.22°C, the corresponding WTP are CNY 4,525.91 (or USD 654), CNY 2,146.68 (or USD 310), CNY 1,311.02 (or USD 189), and CNY 2,132.27 (or USD 308), respectively, if the temperature becomes less extreme by 1%. Figure 3 also shows that the negative effect estimates for cold temperatures are much larger than for hot temperatures, but less precisely estimated, which is consistent with results in Anderson and Bell (2009) and Yang et al. (2019a) where they find that the extreme cold temperatures have a larger impact on mortality than does extreme heat. Based on Fig. 2, the distribution of temperature is heavily right skewed, and we have fewer observations for cold temperature (e.g., below 0°C), which may lead to relatively large standard errors for cold temperature.
Effect of cold and hot temperature on health: cold temperature and hot temperature (the dependent variable is the health score). Significance levels p < 0.01, p < 0.05, and p < 0.1 are indicated by three asterisks, two asterisks, and one asterisk, respectively; robust standard errors are in parentheses; N = 98 423.
Effect of cold and hot temperature on health: HDDs and CDDs (the dependent variable is the health score). Significance levels p < 0.01, p < 0.05, and p < 0.1 are indicated by three asterisks, two asterisks, and one asterisk, respectively; robust standard errors are in parentheses; N = 98 423.
Effect of cold and hot temperature on health: extreme cold and extreme heat (the dependent variable is the health score). Significance levels p < 0.01, p < 0.05, and p < 0.1 are indicated by three asterisks, two asterisks, and one asterisk, respectively; robust standard errors are in parentheses; N = 98 423.
Effect of cold and hot temperature on health. The vertical axis represents the change in the health score. The vertical lines show the 95% confidence interval.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Effect of cold and hot temperature on health. The vertical axis represents the change in the health score. The vertical lines show the 95% confidence interval.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Effect of cold and hot temperature on health. The vertical axis represents the change in the health score. The vertical lines show the 95% confidence interval.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Previous studies have shown that temperature has a cumulative effect on health (Huang et al. 2015; Liang et al. 2007). We use HDD and CDD indicators based on the European Environment Agency (2020) to evaluate climate change and its corresponding impacts on environmental and social systems, as well as human health. Based on Table 7, when the temperature is lower than −12.22°C, a 1°C increase in the monthly accumulation of temperature, which means an increase of one unit degree in the HDD value will lower the health score by 0.06%, and the WTP index is CNY 963.67 (or USD 139; model 1 in Table 7). When the temperature is lower than −6.67°C, the degree of unit health affected is −0.03%, and the corresponding WTP index is CNY 461.61 (or USD 67; model 2 in Table 7). When the temperature threshold is higher than 26.67°C, the unit health effect is −0.006%, which implies CNY 240.56 willingness to pay (or USD 35; model 8 in Table 7). When the temperature is higher than 32.22°C, a one unit in the CDD value will decrease the health score by 0.016%, which implies CNY 92.47 willingness to pay (or USD 13.38; model 9 in Table 7). The results in HDD also suggest that both extreme low temperatures create greater adverse burdens on human health. Changes in extreme cold indicators are negatively correlated with the health status of residents.
Table 8 presents the regional adaptation patterns using the temperature threshold based on WMO extreme weather indicators. The regression results reveal that for each 1% increase in TN10p, TX10p, CSDI, TN90p, TX90p, and WSDI, the health score decreased by 0.231%, 0.132%, 0.229%, 0.076%, 0.031%, and 0.035%, and the WTP indices are CNY 3,586.80, CNY 2,064.90, CNY 3,566.91, CNY 1,164.64, CNY 478.00, and CNY 543.63, respectively (models 1–6 in Table 8). The absolute magnitude of the parameters related to cold weather is larger than their warm weather counterparts, indicating that the extremely cold weather has a greater impact on human health.
e. Regression results in different subgroups
Based on the hierarchical regression method, we further estimated the differences in temperature and health among subgroups at the sociodemographic level. When using the six WMO extreme temperature indicators for regression, we find that subgroups of sex, age, and low income show the same results as the general population, with greater health impacts in extreme cold environments, while only the high-income population experience little effects when measured against the cold indicators. Figure 4a reveals that men are more sensitive to high temperatures, and women are more vulnerable in cold temperatures. We further divide the population into youth and middle-elderly groups using 40 as the cutoff point, referring to Zhang et al. (2017), which suggests that a person’s health, immunity, and various aspects of metabolism decline after the age of 40 in general. Many studies have shown that the elderly are more vulnerable to extreme weather (Zeng et al. 2010; Kinay et al. 2019). Our results show that middle or old people are more sensitive to extreme heat (Fig. 4b). When the population is categorized by economic income,3 wealthy people are more susceptible to extreme heat. However, under an extreme cold environment, the low-income groups will be more adversely affected (Fig. 4c).
Effect of extreme cold and heat on health for different subgroups. The vertical axis represents the change in the health score. The vertical lines show the 95% confidence interval. The young group consists of people age 16–40, and the middle-old group consists of people over age 40. We calculate the average value (CNY 12,064) of the households’ per capita annual income and subsequently classify residents with higher-than-average incomes as the high-income group and those with lower-than-average incomes as the low-income group.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Effect of extreme cold and heat on health for different subgroups. The vertical axis represents the change in the health score. The vertical lines show the 95% confidence interval. The young group consists of people age 16–40, and the middle-old group consists of people over age 40. We calculate the average value (CNY 12,064) of the households’ per capita annual income and subsequently classify residents with higher-than-average incomes as the high-income group and those with lower-than-average incomes as the low-income group.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
Effect of extreme cold and heat on health for different subgroups. The vertical axis represents the change in the health score. The vertical lines show the 95% confidence interval. The young group consists of people age 16–40, and the middle-old group consists of people over age 40. We calculate the average value (CNY 12,064) of the households’ per capita annual income and subsequently classify residents with higher-than-average incomes as the high-income group and those with lower-than-average incomes as the low-income group.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
We further test the robustness of the results of the above subgroups by using temperature frequency indicators, and the HDD and CDD indicators. The results are consistent with the main conclusions and are displayed in Figs. 5 and 6).
As in Fig. 4, but for cold and hot temperature.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
As in Fig. 4, but for cold and hot temperature.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
As in Fig. 4, but for cold and hot temperature.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
As in Fig. 4, but for HDDs and CDDs.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
As in Fig. 4, but for HDDs and CDDs.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
As in Fig. 4, but for HDDs and CDDs.
Citation: Weather, Climate, and Society 13, 4; 10.1175/WCAS-D-21-0038.1
5. Discussion
Based on the observations of nearly 100 000 Chinese residents, our research is the first to study the relationship between temperature and physical and mental health based on self-reported data at an individual level. We also investigate the heterogeneous effects of sociodemographic factors, including gender, age, and income types, on the health effects of temperature.
We confirm the regression relationship between daily health level and the temperature is an inverted U curve under various model specifications. The inflection point of the average temperature on health was estimated at 18.69°C (or 62.62°F), an average indicator based on a large sample from China. The effect of temperature on health could vary across geography or climatic baselines (Deschênes 2014). As the temperature moves away from the inflection point, the health of residents will be more adversely affected.
Due to different latitudes and impacts of climate change, the population’s adaptability to temperature varies. Our results suggest that extreme cold days have a greater impact on overall health. Previous studies suggested that temporary warming can make people more sensitive to intense cold waves (Qu and Xiao 2019). As the global warming trend continues, people can initiate a more gradual increase in thermal adaptation, while adaptation to extreme cold weather may be more difficult.
Six extreme temperature indicators produced by WMO are used as the main temperature explanatory variables to determine the influence of health heterogeneity under different population characteristics. Existing studies have found females were more vulnerable to the impact of temperature variation than men (Stafoggia et al. 2006; Ishigami et al. 2008; Yu et al. 2010; Gao et al. 2019), while others have not found differences between genders (Basu and Ostro 2008; Stafoggia et al. 2008). Our study finds that men and women show different responses to cold and heat. Women exhibit a stronger response to cold, while men are more sensitive to heat. Lan et al. (2008) investigated the gender difference in thermal comfort in China and found that females are more sensitive to the cool environment according to the analysis of skin temperature and the relationship between bodily state and sensation.4 Previous conclusions indicate that the temperature-related health effects are more pronounced in the middle-aged and elderly (Anderson and Bell 2009; Hajat et al. 2006; Ishigami et al. 2008). We find that the middle and old age groups will be affected by extreme cold and heat, and the influence was more significant under extreme heat. For this group, preexisting disorders create difficulty in thermoregulation (Stafoggia et al. 2008). The elderly are less resistant to heat, which might be caused by poor aerobic fitness, differences in body composition, and chronic health conditions (Pandolf 1997). On the psychological level, Hayes and Poland (2018) summarized that while the mental health implications of climate change can affect all age groups, these impacts tend to be greatest among individuals who are the most vulnerable, such as seniors.
We also find significant differences between incomes. The health of low-income groups is more affected by cold weather conditions, while high-income groups are significantly more affected by extreme heat, though the impact of very low temperatures is not precisely estimated for the high-income group. Our results depart slightly from previous studies. Yu et al. (2020) suggested that people with higher socioeconomic status have more ability to resist external heat pressure. Yang et al. (2019a) concluded that extreme temperatures have a greater influence on the mortality rate in regions with low economic development, due to the lack of prevention and adaptation measures, such as air conditioning and medical facilities. In the case of extreme heat, our results show that the rich are slightly more affected than the poor. The vast majority of wealthy people live in urban areas in China, where there might be a more serious heat island effect (Hua et al. 2008), which brings a more severe and lasting negative effect to local people.
6. Conclusions
We establish the relationship between temperature and respondents’ physical and mental health conditions based on individual-level survey and climate data in China. Our study also uncovers the effects of extreme temperatures among subgroups of populations. Our analyses contribute to the growing empirical results with regard to the climate effects on gender (Yu et al. 2010). Our data do not include individuals younger than 16. Existing literature suggests that young children are particularly vulnerable to heat-related deaths, and extreme heat and cold will primarily burden children with infectious diseases (Xu et al. 2012). Climate-related impacts on children are important for further studies.
Our research results have important implications for climate change mitigation policies. We find significant effects of extreme temperatures on the daily health of respondents. For example, low-income groups are more vulnerable to extremely cold weather, and the government or agencies can actively help low-income groups install and improve heating equipment, increase the proportion of renewable and clean energy heating, and optimize operating subsidy policies.
Acknowledgments
The authors are grateful for financial support from the National Natural Science Foundation of China (NSFC) (71704009; 72074021), the Joint Development Program of Beijing Municipal Commission of Education, the Science and Technology Research Project of Hubei Provincial Department of Education (Q20201104), and the Fundamental Research Funds for the Central Universities (FRF-BR-20-04A). We acknowledge the China Family Panel Studies team for providing data and the training of using the dataset.
Data availability statement
Data generated for this project are openly available online (http://isss.pku.edu.cn/cfps/download/login).
APPENDIX
Full Set of Control Variables
Tables A1 and A2 present the coefficient estimates for the full set of control variables based on the specifications used in Tables 3 and 4, respectively.
Nonlinear relationship between temperature and health (the dependent variable is the health score). Significance levels p < 0.01, p < 0.05, and p < 0.1 are indicated by three asterisks, two asterisks, and one asterisk, respectively; robust standard errors are in parentheses; N = 98 423.
Nonlinear relationship between temperature and health: by symptom. Significance levels p < 0.01, p < 0.05, and p < 0.1 are indicated by three asterisks, two asterisks, and one asterisk, respectively; robust standard errors are in parentheses; N = 98 423.
REFERENCES
Ahmed, T., K. M. Muttaqi, and A. P. Agalgaonkar, 2012: Climate change impacts on electricity demand in the state of New South Wales, Australia. Appl. Energy, 98, 376–383, https://doi.org/10.1016/j.apenergy.2012.03.059.
Alexander, L. V., and Coauthors, 2006: Global observed changes in daily climate extremes of temperature and precipitation. J. Geophys. Res., 111, D05109, https://doi.org/10.1029/2005JD006290.
Allen, M. J., and S. C. Sheridan, 2015: Mortality risks during extreme temperature events (ETEs) using a distributed lag non-linear model. Int. J. Biometeor., 62, 57–67, https://doi.org/10.1007/s00484-015-1117-4.
Alola, A. A., S. S. Akadiri, A. C. Akadiri, and U. V. Alola, 2019: Cooling and heating degree days in the US: The role of macroeconomic variables and its impact on environmental sustainability. Sci. Total Environ., 695, 133832, https://doi.org/10.1016/j.scitotenv.2019.133832.
Anderson, B. G., and M. L. Bell, 2009: Weather-related mortality: How heat, cold, and heat waves affect mortality in the United States. Epidemiology, 20, 205–213, https://doi.org/10.1097/EDE.0b013e318190ee08.
Anderson, C. A., and K. B. Anderson, 1998: Temperature and aggression: Paradox, controversy, and a (fairly) clear picture. Human Aggression: Theories, Research and Implications for Social Policy, R. G. Geen and E. Donnerstein, Eds., Academic Press, 247–298, https://doi.org/10.1016/B978-012278805-5/50011-0.
Basu, R., and B. D. Ostro, 2008: A multicounty analysis identifying the populations vulnerable to mortality associated with high ambient temperature in California. Amer. J. Epidemiol., 168, 632–637, https://doi.org/10.1093/aje/kwn170.
Baylis, P., 2020: Temperature and temperament: Evidence from Twitter. J. Public Econ., 184, 104161, https://doi.org/10.1016/j.jpubeco.2020.104161.
Bouchama, A., and J. P. Knochel, 2002: Heat stroke. N. Engl. J. Med., 346, 1978–1988, https://doi.org/10.1056/NEJMra011089.
Chen, Y. W., and Y. P. Shi, 2013: Air quality pricing from the perspective of happiness economics: A study based on CFPS 2010 data (in Chinese). Econ. Sci., 35, 77–88.
Chen, Z. H., G. F. Yang, and H. B. Hu, 2014: Research progress on the effects of temperature on human health in the context of climate change (in Chinese). Chin. J. Publ. Health, 30, 1318–1321.
Choi, G., and Coauthors, 2009: Changes in means and extreme events of temperature and precipitation in the Asia-Pacific network region, 1955–2007. Int. J. Climatol., 29, 1906–1925, https://doi.org/10.1002/joc.1979.
Chowdhury, F. R., Q. S. U. Ibrahim, M. S. Bari, M. M. J. Alam, S. J. Dunachie, A. J. Rodriguez-Morales, and M. I. Patwary, 2018: The association between temperature, rainfall and humidity with common climate-sensitive infectious diseases in Bangladesh. PLOS ONE, 13, e0199579, https://doi.org/10.1371/journal.pone.0199579.
Costello, A., and Coauthors, 2009: Managing the health effects of climate change: Lancet and University College London Institute for Global Health Commission. Lancet, 373, 1693–1733, https://doi.org/10.1016/S0140-6736(09)60935-1.
Curriero, F. C., K. S. Heiner, J. M. Samet, S. L. Zeger, S. Lisa, and J. A. Patz, 2002: Temperature and mortality in 11 cities of the eastern United States. Amer. J. Epidemiol., 155, 80–87, https://doi.org/10.1093/aje/155.1.80.
DeSalvo, F. B., V. S. Fan, M. B. McDonell, and S. D. Fihn, 2005: Predicting mortality and healthcare utilization with a single question. Health Serv. Res., 40, 1234–1246, https://doi.org/10.1111/j.1475-6773.2005.00404.x.
Deschênes, O., 2014: Temperature, human health, and adaptation: A review of the empirical literature. Energy Econ., 46, 606–619, https://doi.org/10.1016/j.eneco.2013.10.013.
Deschênes, O., and M. Greenstone, 2011: Climate change, mortality, and adaptation: Evidence from annual fluctuations in weather in the US. Amer. Econ. J. Appl. Econ., 3, 152–185, https://doi.org/10.1257/app.3.4.152.
European Environment Agency, 2020: Heating and cooling degree days. Accessed 19 June 2020, https://www.eea.europa.eu/data-and-maps/indicators/heating-degree-days-2/assessment.
Gao, J., and Coauthors, 2019: The association between cold spells and admissions of ischemic stroke in Hefei, China: Modified by gender and age. Sci. Total Environ., 669, 140–147, https://doi.org/10.1016/j.scitotenv.2019.02.452.
Hajat, S., R. S. Kovats, and K. Lachowycz, 2006: Heat-related and cold-related deaths in England and Wales: Who is at risk? Occup. Environ. Med., 64, 93–100, https://doi.org/10.1136/oem.2006.029017.
Hajat, S., Z. Chalabi, P. Wilkinson, B. Erens, L. Jones, and N. Mays, 2016: Public health vulnerability to wintertime weather: Time-series regression and episode analyses of national mortality and morbidity databases to inform the Cold Weather Plan for England. Public Health, 137, 26–34, https://doi.org/10.1016/j.puhe.2015.12.015.
Halliday, T., B. Mazumder, and A. Wong, 2021: Intergenerational mobility in self-reported health status in the US. J. Public Econ., 193, 104307, https://doi.org/10.1016/j.jpubeco.2020.104307.
Hansen, A., P. Bi, M. Nitschke, P. Ryan, D. Pisaniello, and G. Tucker, 2008: The effect of heat waves on mental health in a temperate Australian city. Environ. Health Perspect., 116, 1369–1375, https://doi.org/10.1289/ehp.11339.
Hayes, K., and B. Poland, 2018: Addressing mental health in a changing climate: Incorporating mental health indicators into climate change and health vulnerability and adaption assessments. Int. J. Environ. Res. Public Health, 15, 1806, https://doi.org/10.3390/ijerph15091806.
Hua, L. J., Z. G. Ma, and W. D. Guo, 2008: The impact of urbanization on air temperature across China. Theor. Appl. Climatol., 93, 179–194, https://doi.org/10.1007/s00704-007-0339-8.
Huang, F., A. Zhao, R. J. Chen, H. D. Kan, and X. Y. Kuang, 2015: Ambient temperature and outpatient visits for acute exacerbation of chronic bronchitis in Shanghai: A time series analysis. Biomed. Environ. Sci., 28, 76–79, https://doi.org/10.3967/bes2015.008.
Idler, E. L., and Y. Benyamini, 1997: Self-rated health and mortality: A review of twenty-seven community studies. J. Health Soc. Behav., 38, 21–37, https://doi.org/10.2307/2955359.
IPCC, 2007: Climate Change 2007: Impacts, Adaptation, and Vulnerability. Cambridge University Press, 976 pp., https://www.ipcc.ch/pdf/assessment-report/ar4/wg2/ar4_wg2_full_report.pdf.
Ishigami, A., S. Hajat, R. S. Kovats, L. Bisanti, M. Rognoni, A. Russo, and A. Paldy, 2008: An ecological time-series study of heat-related mortality in three European cities. Environ. Health, 7, 5, https://doi.org/10.1186/1476-069X-7-5.
Iversen, L. L., 1982: Neurotransmitters and CNS disease: Introduction. Lancet, 320, 914–918, https://doi.org/10.1016/S0140-6736(82)90876-5.
Jia, Z., and Coauthors, 2018: Exposure to ambient air particles increases the risk of mental disorder: Findings from a natural experiment in Beijing. Int. J. Environ. Res. Public Health, 15, 160, https://doi.org/10.3390/ijerph15010160.
Jorgenson, A. K., 2009: Political-economic integration, industrial pollution and human health: A panel study of less-developed countries, 1980–2000. Int. Sociol., 24, 115–143, https://doi.org/10.1177/0268580908099156.
Joyce, T. J., M. Grossman, and F. Goldman, 1989: An assessment of the benefits of air pollution control: The case of infant health. J. Urban Econ., 25, 32–51, https://doi.org/10.1016/0094-1190(89)90042-9.
Jylhä, M., S. Volpato, and J. A. Guralnik, 2006: Self-rated health showed a graded association with frequently used biomarkers in a large population sample. J. Clin. Epidemiol., 59, 465–471, https://doi.org/10.1016/j.jclinepi.2005.12.004.
Kinay, P., A. P. Morse, E. V. Villanueva, K. Morrissey, and P. L. Staddon, 2019: Direct and indirect health impacts of climate change on the vulnerable elderly population in East China. Environ. Rev., 27, 295–303, https://doi.org/10.1139/er-2017-0095.
Lan, L., Z. Lian, W. Liu, and Y. Liu, 2008: Investigation of gender difference in thermal comfort for Chinese people. Eur. J. Appl. Physiol., 102, 471–480, https://doi.org/10.1007/s00421-007-0609-2.
Liang, W. M., W. P. Liu, S. Y. Chou, and H. W. Kou, 2007: Ambient temperature and emergency room admissions for acute coronary syndrome in Taiwan. Int. J. Biometeor., 52, 223–229, https://doi.org/10.1007/s00484-007-0116-5.
Lin, S., M. Luo, R. J. Walker, X. Liu, S. A. Hwang, and R. Chinery, 2009: Extreme high temperatures and hospital admissions for respiratory and cardiovascular diseases. Epidemiology, 20, 738–746, https://doi.org/10.1097/EDE.0b013e3181ad5522.
Maes, M., F. Demeyer, P. Thompson, D. Peeters, and P. Cosyns, 1994: Synchronized annual rhythms in violent suicide rate, ambient temperature and the light-dark span. Acta Psychiatr. Scand., 90, 391–396, https://doi.org/10.1111/j.1600-0447.1994.tb01612.x.
Marsha, A., S. R. Sain, M. J. Heaton, A. J. Monaghan, and O. V. Wilhelmi, 2018: Influences of climatic and population changes on heat-related mortality in Houston, Texas, USA. Climatic Change, 146, 471–485, https://doi.org/10.1007/s10584-016-1775-1.
Michelozzi, P., and Coauthors, 2009: High temperature and hospitalizations for cardiovascular and respiratory causes in 12 European cities. Amer. J. Respir. Crit. Care Med., 179, 383–389, https://doi.org/10.1164/rccm.200802-217OC.
Murray, C. J. L., and Coauthors, 2015: Global, regional, and national disability-adjusted life years (DALYs) for 306 diseases and injuries and healthy life expectancy (HALE) for 188 countries, 1990–2013: Quantifying the epidemiological transition. Lancet, 386, 2145–2191, https://doi.org/10.1016/S0140-6736(15)61340-X.
New, M., and Coauthors, 2006: Evidence of trends in daily climate extremes over southern and West Africa. J. Geophys. Res., 111, D14102, https://doi.org/10.1029/2005JD006289.
Oppenheimer, M., and J. M. Anttila-Hughes, 2016: The science of climate change. Future Child., 26, 11–30, https://doi.org/10.1353/foc.2016.0001.
Pandolf, K. B., 1997: Aging and human heat tolerance. Exp. Aging Res., 23, 69–105, https://doi.org/10.1080/03610739708254027.
Qi, Y. Q., 2014: Reliability and validity of self-rated general health (in Chinese). Chin. J. Sociol., 6, 196–215.
Qu, F., and Z. N. Xiao, 2019: Assessment of the impact of climate change on human health. Adv. Meteor. Sci. Technol., 9, 34–47.
Shen, X. J., B. H. Liu, and D. W. Zhou, 2017: Spatiotemporal changes in the length and heating degree days of the heating period in Northeast China. Meteor. Appl., 24, 135–141, https://doi.org/10.1002/met.1612.
Song, C. B., and Coauthors, 2017: Health burden attributable to ambient PM2.5 in China. Environ. Pollut., 223, 575–586, https://doi.org/10.1016/j.envpol.2017.01.060.
Stafoggia, M., and Coauthors, 2006: Vulnerability to heat-related mortality: A multicity, population-based, case-crossover analysis. Epidemiology, 17, 315–323, https://doi.org/10.1097/01.ede.0000208477.36665.34.
Stafoggia, M., and Coauthors, 2008: Factors affecting in-hospital heat-related mortality: A multi-city case-crossover analysis. J. Epidemiol. Community Health, 62, 209–215, https://doi.org/10.1136/jech.2007.060715.
Stone, A. A., and C. Mackie, Eds., 2014: Subjective Well-Being: Measuring Happiness, Suffering, and other Dimensions of Experience. National Academies Press, 190 pp., https://www.nap.edu/read/18548/chapter/2.
Watts, N., and Coauthors, 2015: Health and climate change: Policy responses to protect public health: The Lancet Commissions. Lancet, 386, 1861–1914, https://doi.org/10.1016/S0140-6736(15)60854-6.
Xu, H., and Y. Xie, 2017: Socioeconomic inequalities in health in China: A reassessment with data from the 2010–2012 China Family Panel Studies. Soc. Indic. Res., 132, 219–239, https://doi.org/10.1007/s11205-016-1244-2.
Xu, Z., R. A. Etzel, H. Su, C. Huang, Y. Guo, and S. Tong, 2012: Impact of ambient temperature on children’s health: A systematic review. Environ. Res., 117, 120–131, https://doi.org/10.1016/j.envres.2012.07.002.
Xue, T., T. Zhu, Y. Zheng, and Q. Zhang, 2019: Declines in mental health associated with air pollution and temperature variability in China. Nat. Commun., 10, 2165, https://doi.org/10.1038/s41467-019-10196-y.
Yan, Z., and Coauthors, 2002: Trends of extreme temperatures in Europe and China based on daily observations. Climatic Change, 53, 355–392, https://doi.org/10.1023/A:1014939413284.
Yang, Z. M., Q. Wang, and P. Liu, 2019a: Extreme temperature and mortality: Evidence from China. Int. J. Biometeor., 63, 29–50, https://doi.org/10.1007/s00484-018-1635-y.
Yang, Z. M., W. R. Li, and Z. M. Yan, 2019b: The relationship between temperature change and electric power demand—Evidence from China urban panel data from 2000 to 2014. J. Beijing Inst. Technol., 5, 44–55.
Yao, J., 2021: Housing wealth and labor participation decision: Evidence from CFPS data (in Chinese). Rev. Econ. Manage., 1, 77–88.
Yu, G. L., T. T. Chen, and F. Q. Zhao, 2020: The influence of air temperature and temperature variability on mental health (in Chinese). Adv. Psychol. Sci., 28, 1282–1292, https://doi.org/10.3724/SP.J.1042.2020.01282.
Yu, W., P. Vaneckova, K. Mengersen, X. Pan, and S. Tong, 2010: Is the association between temperature and mortality modified by age, gender and socio-economic status? Sci. Total Environ., 408, 3513–3518, https://doi.org/10.1016/j.scitotenv.2010.04.058.
Zeng, Y., D. Gu, J. Purser, H. Hoenig, and N. Christakis, 2010: Associations of environmental factors with elderly health and mortality in China. Amer. J. Public Health, 100, 298–305, https://doi.org/10.2105/AJPH.2008.154971.
Zhang, X., L. Alexander, G. C. Hegerl, P. Jones, A. K. Tank, T. C. Peterson, B. Trewin, and F. W. Zwiers, 2011: Indices for monitoring changes in extremes based on daily temperature and precipitation data. Wiley Interdiscip. Rev.: Climate Change, 2, 851–870, https://doi.org/10.1002/wcc.147.
Zhang, X., X. Zhang, and X. Chen, 2017: Happiness in the air: How does a dirty sky affect mental health and subjective well-being? J. Environ. Econ. Manage., 85, 81–94, https://doi.org/10.1016/j.jeem.2017.04.001.
Zhou, Y. Q., and G. Y. Ren, 2010: Variation characteristics of extreme temperature indices in mainland China during 1956–2008 (in Chinese). Climate Environ. Res., 15, 405–417.
Zivin, J. G., and J. Shrader, 2016: Temperature extremes, health, and human capital. Future Child., 26, 31–50, https://doi.org/10.1353/foc.2016.0002.
The definition of city is based on the CFPS data, including municipalities, prefecture-level cities, and county-level cities.
With respect to Eq. (2), we take the total derivative of health for a given level (i.e., let dH = 0), and we can calculate the WTP as the average marginal rate of substitution between temperature and per capita household income, ∂Yijt/∂ETji | (dHijt = 0) = −Yijt × β1/β2.
We calculate the average value (CNY 12,064) of the households’ per capita annual income and subsequently classify residents with higher-than-average incomes as the high-income group and those with lower-than-average incomes as the low-income group.
The reason for reference may be that men are more muscular and have a faster metabolism, leading to a higher body surface temperature. And there are also medical explanations saying that this is related to the structure of the male and female reproductive systems.
