In this study, we constructed a new database to investigate the complex relationship between hazy weather, heightened air pollution concerns due to elevated PM 2.5 levels, and residents’ well-being in China. Our study concludes that the results of the baseline regression showing that PM2.5 significantly increases the well-being of the population is at odds with common sense, which is attributed to potential endogeneity issues, including reverse causation and omitted variables. For this purpose, first, we used the instrumental variables method to exclude the endogeneity problem and obtained a causal relation between objective air pollution and well-being, that is, objective air pollution significantly and negatively affects residents’ well-being. Second, we also considered the effect of individual differences and verified the mechanistic pathways by which objective air pollution affects well-being through residents’ subjective air pollution. Specifically, we found that in the air quality evaluation process, “pessimistic” residents have a lower level of well-being, and “optimistic” residents have a higher level of well-being. Finally, we verified that the level of household income can weaken the negative impact of objective air pollution on residents’ well-being and has a positive moderating effect. By integrating macro and micro-level data through innovative technology, our research provides policy recommendations for effective pollution management and improved overall well-being in China.
Hinweise
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
Air pollution is the largest environmental health hazard in the world, affecting 7 million people annually (Murray, 2020). Air pollution has caused and worsened many epidemics, from asthma to cancer, lung illness, and cardiovascular disease. Exposure to PM2.5 increases the risk of viral infections in this COVID-19 timeframe as well as the chance of severe post-infection symptoms. Current quantitative research on the relationship between air pollution levels and health hazards indicates that even exposure to low levels of air pollution can negatively affect health, suggesting that there is no recommended limit for PM2.5. 2021 Chinese air pollution data show that 66% of Chinese cities have improved air quality and reduced PM2.5 compared with 2020 data. Coal remains one of the most significant sources of PM2.5 pollution in China, the world’s largest producer and consumer of coal (Pui et al., 2014). The PM2.5 levels in the north continue to increase due to seasonal dust storms brought on by desertification and other factors. These factors have led to no Chinese cities in the 2021 World Air Quality Report reaching the revised WHO PM2.5 guideline of 5μg/m3, and the air pollution situation in China shows little improvement.
Recently, many parts of China have suffered from widespread smoggy days, and air pollution has caused concern and panic in society (Dong et al., 2022). There is a growing awareness that economic growth is no longer the only goal of urban development and that increasing per capita income does not necessarily lead to increased national well-being (Clark et al., 2008). At the 20th National Congress of the Communist Party of China held in October 20221, it was mentioned that people’s sense of access, well-being, and security should be made more abundant, guaranteed, and sustainable, and new results should be achieved in terms of common prosperity. Therefore, improving well-being has become a long-term goal of China’s development and a fundamental path towards achieving common prosperity.
Anzeige
Factors influencing well-being can be divided into two main types: economic and noneconomic. Economic factors include income, employment, and consumption, whereas non-economic factors include sex, age, health, marital status, education level, social relationships, and intergenerational transmission. Among the many factors that affect well-being, the living environment is gradually being emphasized, and noise and well-being have been found to be significantly negatively affected by climate and natural calamities (Rehdanz & Maddison, 2008; Van Praag & Baarsma, 2005). The effect of severe air pollution on well-being has attracted the attention of many researchers. Using data from the General Social Survey, Levinson found (Levinson, 2012) that air pollution in the United States negatively affects residents’ well-being and that residents interviewed during dates of high air pollution have lower levels of well-being. A study by Sanduijav et al. on Ulaanbaatar, the capital of Mongolia (Sanduijav et al., 2021), identified a negative correlation between air pollution and well-being and found that a 10 g/m3 increase in PM10 concentration was related to a 0.0017-point decrease in the self-reported life satisfaction of an individual on an 11-point scale. Using information from social surveys and urban macro data, Yuan et al. (2018) investigated the relationship between air pollution, greenery, and residents’ well-being in China and stated that the reduced mobility brought on by haze and the elevated health risks brought on by air pollution may be responsible for the negative relationship between air pollution and people’s daily lives.
Although there is consensus among academics about the negative effects of air pollution on well-being (Ambrey et al., 2014; Li et al., 2018), existing studies do not distinguish between objective and subjective air pollution. Regarding objective air pollution (OAP), some scholars have focused on different pollutant concentrations (Lin et al., 2019; Xu et al., 2022), such as SO2, NO2, PM2.5, and PM10, or have used local composite air quality indices, such as the AQI (Saha et al., 2021), to measure OAP levels and thus analyze their effects on well-being. Subjective air pollution (SAP) is an individual’s subjective assessment of OAP (Song et al., 2020). OAP may not capture its effect on well-being, whereas SAP may directly affect well-being (Ma et al., 2020). Therefore, this study measures both OAP and SAP and examines the differences between them and the impact they have on residents’ well-being. This study also discusses the pathways through which OAP affects residents’ well-being through SAP.
Academics have investigated the connection between air pollution and citizens’ well-being, and some studies have used it to support the local existence of the Easterlin paradox. Easterlin’s paradox (Easterlin, 1974), also known as the “happiness-income puzzle” and the “happiness paradox,” is a puzzling and important question posed by R. Easterlin (Easterlin, 2016), a professor of economics at the University of Southern California, “Why doesn’t more wealth lead to greater happiness?” Research has been conducted to explain this phenomenon in terms of relative income and omitted variables (Ferrer-i-Carbonell, 2005; Gardes & Merrigan, 2008). The omitted variable explanation considers that many noneconomic factors other than income affect well-being, such as employment status, health status, and environmental quality (Fanning & O’Neill, 2019). Existing studies would join the discussion of income level as an important factor affecting well-being (Song et al., 2019), but do not study it in conjunction with the relationship between air pollution and well-being. Therefore, in addition to verifying the effect of income level on residents’ well-being, this study also discusses the role of household income level in the process by which air pollution affects residents’ well-being.
By comparing our study to earlier literature, the following three major contributions are made. First, the objectively extracted county-level air pollution data were matched with individual-level well-being data of residents, and regressions were conducted using county ventilation coefficients as instrumental variables. The combination of the micro-database and the county-level macro-data creates the unique database used in this paper, providing micro-evidence from China for the study of the relationship between the two. Second, existing studies have only focused on the effects of OAP or SAP on residents’ well-being; there is a lack of research on the relationship between the two. This study examines the connection between OAP and SAP, the impacts of both on inhabitants’ levels of well-being, and the variations in these effects. Third, existing research has concentrated on how income affects well-being, but there has not been much discussion of how income affects well-being as a result of air pollution. This study analyzed the moderating role of income level in the effect of air pollution on residents’ well-being, enriching research on the relationship between income, air pollution, and well-being.
Anzeige
This study extracted PM2.5 data at the county level in China using ArcGIS (ESRI, Redlands, CA, USA) and matched it with the CHIP household micro database to analyze the connection between OAP and the level of well-being among Chinese citizens. We also measured residents’ SAP ratings, measured the difference between them, analyzed the difference in subjective well-being levels between “pessimistic” and “optimistic” residents of air pollution, and verified the mechanism path by which OAP affects well-being through SAP. After fully considering possible endogeneity issues, regressions were conducted using county ventilation coefficients as instrumental variables while examining the moderating role of household income level in the process of OAP affecting residents’ well-being.
The remainder of this paper is organized as follows. Section 2 presents data sources, variable descriptions, and econometric models. Section 3 provides the theoretical analysis and hypotheses. Section 4 presents our empirical findings. Section 5 concludes the paper with a discussion of policy implications.
2 Theoretical Analysis
Well-being has become a focus of academic research. Well-being, an individual’s subjective and comprehensive evaluation of life conditions, involves several disciplines, including economics, sociology, and psychology. Maslow’s Hierarchy of Needs theory believes that well-being is a subjective feeling after people’s needs levels are satisfied and closely links well-being to people’s life needs (Maslow, 1943). Subsequently, American economist Galbraith further proposed the concept of Quality of Life (QoL) as a subjective evaluation of living conditions (Galbraith, 1958), covering one’s satisfaction with life situations, self-evaluation, and satisfaction with one’s social fulfillment. Since then, research on subjective quality of life has largely followed this concept and called it well-being (Veenhoven, 1994). Based on existing studies, this study defines well-being as an individual’s comprehensive evaluation of life conditions and quality of life; a subjective identification of individuals in their emotional experience; and subjective, relatively stable, and holistic in nature. In recent years, some scholars have tried to use social media big data to assess residents’ well-being indices (Mitchell et al., 2013), but the results are not representative because of the limited number of people using social media software. Therefore, this study uses a structure questionnaire, a practice used by most scholars, to measure resident well-being.
Air pollution is the destruction of air quality in a country or region and is considered a public good because of its non-exclusive and non-competitive characteristics. Existing studies have concluded that air pollution affects residents’ well-being through three main channels: health status, productivity level, and convenience of life. First, air pollution can impair residents’ psychological and physiological conditions (Li et al., 2019), thereby reducing their well-being. On the one hand, air pollution impacts people’s perception of the risk of disease, and there is concern about the increased prevalence of disease from long-term exposure to poorer-quality air (Hadley et al., 2018; Petrowski et al., 2021). However, air pollution can eventually have substantial effects on the health of the population; for example, it can lead to respiratory diseases, physiological dysfunction, and irritation of mucous membrane tissues, such as the eyes and nose (Song et al., 2019a, b; Xu et al., 2020). Second, damage to health due to air pollution can further affect the income level of residents and reduce their well-being by reducing labor supply time and productivity. In terms of labor supply time, increased air pollution concentrations increase the number of days without work (Hausman et al., 1984) and reduce the number of hours worked per week by the labor force (Hanna & Oliva, 2015). In terms of productivity, air pollution can directly reduce the productivity of the workforce working outside (Li et al., 2017) and can also affect the productivity of the indoor workforce by affecting the psychological and physiological health of the workforce. A decline in working hours and efficiency can lead to lower household income levels, thereby affecting residents’ well-being. Third, air pollution, as a type of spatial friction, can cause residents to exhibit avoidance behavior, reduce the scope and frequency of people’s activities outside, and reduce the possibility of social interactions (Davis et al., 2015). Air pollution can reduce consumption opportunities, such as eating out and going to the cinema (Noonan, 2014), as well as reduce the time spent on exercise and leisure, such as outdoor cycling and walking in the park (Xu et al., 2022). Air pollution also increases resident protective expenditures (Zhang & Mu, 2018), including the purchase of masks, air purifiers, and health insurance, leading to a crowding-out effect on other aspects of resident consumption.
This investigation focused on how residents’ well-being is affected by air pollution; however, its impact on well-being may not be captured based on objective air quality assessment alone (Li et al., 2014). One possible reason for this is that objective risk is often a measure that does not consider psychosocial conditions. For individuals, the transformation of their psychological conditions is derived from processing and transforming objectively measured risks through perceptions (Elias & Shiftan, 2012). Individuals with different backgrounds may perceive the same OAP level differently (Tian et al., 2022); for example, individuals with lung disease may perceive air quality more significantly and individuals with strong environmental knowledge may also perceive air quality more significantly, which affects their subjective well-being (Gu et al., 2015). Thus, although subjective perceptions are influenced by objective risks, there may be differences between the two depending on issues such as personal experience (Guo et al., 2021), imperfect information, or a lack of confidence in official sources of information (Ward-Caviness et al., 2020). Therefore, this study considers not only OAP but also residents’ subjective evaluations of objective pollution, measured by residents’ self-rated air pollution levels in the area in which they live, to test whether the mechanism path that OAP affects residents’ well-being by influencing SAP is valid. Meanwhile, the difference between OAP and SAP was measured to analyze the extent to which residents’ ‘overestimation’ or ‘underestimation’ of the level of air pollution affects their well-being.
Easterlin’s ‘Happiness Paradox’ argues that there is limited scope for economic growth to improve national happiness in the long-term. However, China is still in the ranks of developing countries; the population base is large, the per capita income level is still low, and the increase in the income level of individual residents in the short term can still increase their well-being (Ye & Zhang, 2020). Additionally, under the same OAP conditions, households with higher incomes are more likely to invest in costlier and more efficient air pollution protection gear (Sun et al., 2017), such as masks and air filters, and reduce the effects of air pollution on household members by spending more on medical care (Barwick et al., 2018) and receiving health insurance. Thus, we believe that the level of household income can increase the well-being of the residents and simultaneously weaken the negative effect of air pollution on well-being (Sun et al., 2022).
Based on this analysis, Fig. 1 illustrates the theoretical framework and working hypotheses.
Fig. 1
Theoretical Framework of air pollution affecting subjective well-being
×
H1
Air pollution negatively affects residents’ subjective well-being.
H2
OAP levels affect resident subjective well-being through SAP evaluation.
H3
Income level can mitigate the detrimental impact of air pollution on residents’ subjective well-being.
3 Methodology
3.1 Data and Variables
The main purpose of this study was to systematically examine the effects of OAP on residents’ subjective air quality evaluations and well-being levels as well as the potential moderating role of household income levels. A total of 15,551 samples were obtained nationwide by matching the CHIP micro-database, air pollution data, and macro data from the China County Yearbook, with the following sources and processing of each type of variable. The meanings and descriptive statistics of the selected variables are reported in Table 1. The samples in the CHIP database2 are derived from a nationwide survey conducted by the China Bureau of Statistics in 2019. Employing a systematic sampling approach, the CHIP project team strategically selected samples through the stratification of regions into East, Central, and West, encompassing a total of 15 provinces. This approach enhances the data’s representativeness to a significant extent. The database itself encompasses a broad spectrum of information, including details on household income, consumption patterns, and personal characteristics of residents. Notably, the database also discloses the names of the cities and counties where the surveyed samples are located. This information proves invaluable as it facilitates the matching of corresponding variables, such as air pollution levels, ventilation coefficients, and climate types, at both the city and county levels. This, in turn, offers crucial support for constructing the database employed in our study.
Table 1
Definition of variables and descriptive statistics
Variable type
Variable
Definition
Mean
SD
Min
Max
Dependent variable
Well-being
Very unhappy-Very happy = 1–5
3.955
0.770
1.000
5.000
Independent variable
OAP
PM2.5 value (μg/m3)
37.121
10.458
14.846
63.102
SAP
Not serious-Very serious = 1–5
2.188
0.998
1
5
IV variable
VC
Ventilation coefficient, log
7.074
0.451
5.803
8.144
Control variable
Age
Age
45.183
15.659
17
92
Gender
Female = 0, Male = 1
0.498
0.500
0
1
Married
Unmarried = 0, Married = 1
0.802
0.399
0
1
Health
Self-rated health status: Bad = 1, Poor = 2, General = 3, Good = 4, Very good = 5
1.874
0.898
1
5
Educ
Elementary school and below = 1; Junior high school = 2; High school = 3; College and above = 4
Have you engaged in new types of employment in 2018, such as take-out, opening an online store, or online taxi? Yes = 1; No = 0
0.004
0.067
0
1
Per capita household income
Per capita household income = total household income/ household size, log
10.148
0.706
5.638
13.868
Economic development level
County GDP per capital, log
10.797
0.758
8.903
12.958
Industry structure
County secondary industry as a proportion of GDP (%)
0.439
0.116
0.119
0.722
Degree of fiscal decentralization
County budget revenue as a share of fiscal expenditure (%)
0.427
0.280
0.037
1.218
Medical coverage level
Number of beds in medical and health institutions, log
8.042
0.750
5.220
9.328
Precipitation
Average annual precipitation in the county (mm), log
6.888
0.488
5.163
7.734
Temperature
Average annual temperature in the county (℃)
14.979
4.761
-0.142
23.383
3.1.1 Residents’ well-being and micro data Sources
Within the construct of subjective well-being (Steptoe et al., 2015), at least three different approaches capture a different aspect – life evaluation, hedonic well-being, and eudemonic well-being. Therefore, this study followed the practice of most studies (Chen et al., 2024; Huang et al., 2024) and adopted life satisfaction to measure residents’ subjective well-being. This study uses the China Household Income Survey data (CHIPS) as the research sample. Well-being was measured by the question, “Do you feel happy considering all aspects of life?” Respondents answered, “Very unhappy,” “Unhappy,” “General,” “Happy,” and “Very happy,” which were indicated by a rating of 1–5 respectively according to the level of well-being.
3.1.2 Air Pollution data Sources
This study uses PM2.5 throughout 2018 to reflect regional air pollution levels. In this study, the air pollution indicator used is PM2.5. Contrary to single chemical components, it is made up of tiny airborne particles with a diameter of less than or equal to 2.5 microns and is widely regarded as the most dangerous particulate matter to human health owing to its negative effects on health and widespread prevalence in the environment. PM2.5 data was taken from Global/Regional Estimates (V5.GL.02), a monthly global PM2.5 dataset made accessible to the public by Washington University in St. Louis’s Atmospheric Composition Analysis Group. In this study, PM2.5 data of China for each month in 2018 are selected. The average PM2.5 concentration for the whole year of 2018 on the scale of Chinese county-level administrative regions is obtained by overlaying with vector layers of Chinese county-level administrative region boundaries, and the above operation is realized based on ArcGIS.
3.1.3 Meteorological Indicators data Sources
Temperature and precipitation data from the ERA5 dataset, made available by the European Center for Medium-Range Weather Forecasts (ECMWF), were used as meteorological indicators in this study. ERA5 is the fifth generation of the ECMWF reanalysis for the previous four to seven decades of global climate and weather. The ERA-Interim reanalysis was replaced by ERA5. The dataset was stored in a grid shape with a spatial resolution of 0.25°×0.25° and spanned the entire planet. In this study, precipitation, and temperature data for each month of 2018 in China were selected, and relevant meteorological indicators for each month of 2018 and the entire year at the county-level administrative district scale in China were obtained by overlaying the vector layers with the county-level administrative district boundaries of China in the ArcGIS platform.
3.1.4 District and County Economic Statistic Sources
This study used economic development, industrial structure, degree of fiscal decentralization, and healthcare coverage as control variables at the district and county levels. According to the regions involved after matching CHIPS with air pollution data, corresponding statistical indicators were collected in the China County Statistical Yearbook and the China City Statistical Yearbook to control the effect of regional differences on the results.
3.2 Variable Description
3.2.1 Socio-demographic Characteristics of the Respondents
The baseline profile of the matched sample was as follows Table 2: 49.57% and 50.43% of the total sample were female and male, respectively, and 69.44% of the respondents were under the age of 50. The proportions of respondents in junior high school and university and those older accounted for 53.89% and 20.25%, respectively. The percentages of employers, employees, self-employed workers, and household helpers were 2.13%, 40.92%, 9.32%, and 1.08%, respectively, whereas 46.55% of the respondents were not employed in these categories. Respondents with per capita household income levels between 10,000 and 50,000 yuan accounted for 75.68% of the sample. In addition, we briefly described OAP and SAP. According to the air quality standard of the PM2.5 detection network, the air quality grade is excellent at 0 ∼ 35 µg/m³, and the PM2.5 concentration greater than 35 µg/m³ in the respondents’ area accounted for 51.84% of the total sample. Respondents who considered SAP serious accounted for 10.64% of the total sample, whereas 68.07% considered it not serious.
Table 2
Socio-demographic characteristics of the respondents
Freq.
Percent
Freq.
Percent
Age
OAP
0 ∼ 25
4396
28.16%
0 ∼ 25 µg/m³
1775
11.37%
25 ∼ 50
6444
41.28%
25 ∼ 35 µg/m³
5744
36.79%
50 ∼ 75
4397
28.16%
35 ∼ 45 µg/m³
4011
25.69%
75 ∼ 100
375
2.40%
45 ∼ 55 µg/m³
3492
22.37%
Gender
> 55 µg/m³
590
3.78%
Male
7873
50.43%
SAP
Female
7739
49.57%
Not serious
4110
26.33%
Education
Less serious
6517
41.74%
Primary school
4509
22.88%
General
3323
21.28%
Middle school
4842
31.01%
More serious
1257
8.05%
High school
3099
19.85%
Very serious
405
2.59%
Bachelor’s degree
3162
20.25
Per capita household income
Employment status
0 ∼ 10,000
1514
9.70%
Employer
332
2.13%
10,000 ∼ 25,000
6352
40.69%
Employee
6389
40.92%
25,000 ∼ 50,000
5463
34.99%
Self-employed worker
1455
9.32%
50,000 ∼ 75,000
1484
9.51%
Household helper
169
1.08%
75,000 ∼ 100,000
478
3.06%
none
7267
46.55%
100,000+
321
2.06%
3.2.2 Visual Analysis of air Pollution Situation
The spatial distribution of PM2.5 in China varies significantly, see in Fig. 2. High values of PM2.5, were mainly concentrated in the North China Plain, Fenwei Plain, and south-central Xinjiang. The North China Plain and the Fenwei Plain are the main population-gathering areas and urban concentration zones in northern China. The terrain is flat, which is prone to air pollution and not conducive to pollutant dispersion. The south-central part of Xinjiang is the main desert distribution area, and coupled with high wind speed, it is easy to produce sandy and dusty weather in this area, thus increasing the concentration of particulate matter. The low-value areas of PM2.5, were mainly located in the Qinghai-Tibet Plateau and the Northeast Daxinganling region. The Qinghai-Tibet Plateau is an important ecological protection zone with a sparse population and low industrial pollution. The northeastern Daxinganling region has a high degree of vegetation cover and is a natural ecological barrier that helps improve air quality.
Fig. 2
The average PM2.5 concentrations of China in 2018
×
3.3 Empirical Models
3.3.1 Benchmark Model
Referring to Levinson’s (Levinson, 2012) econometric model, we constructed the following basic model to estimate the effect of air pollution on population well-being.
In Eq. (1), \(\:{Happiness}_{ji}\) denotes the degree of well-being of the ith resident; \(\:{PM}_{2.5j}\) denotes the degree of air pollution in the jth district and county3; \(\:{X}_{ji}\) denotes the personal characteristics variable of the resident; \(\:{C}_{j}\) denotes the economic characteristics variable of the district and county; \(\:{M}_{j}\) denotes the climate variable of the district and county; \(\:{\beta\:}_{1}\) denotes the influence coefficient of air pollution on the well-being of the resident, \(\:{\beta\:}_{0}\)、\(\:{\beta\:}_{2}\)、\(\:{\beta\:}_{3}\)、\(\:{\beta\:}_{4}\) are all coefficients to be estimated; \(\:{\epsilon\:}_{i}\) is the residual term, which obeys normal distribution and has a variance of \(\:{\sigma\:}^{2}\), i.e., \(\varepsilon \sim N\left( {0,{\sigma ^2}} \right)\).
As well-being is an ordered variable, we used an ordered probit model for the estimation. We also applied ordinary least squares (OLS) regression. Although the well-being of residents is not a continuous variable, as MacKerron and Mourato (2009) point out, OLS can produce outcomes similar to those of the ordered probit model, and OLS outcomes are easier to comprehend (Yuan et al., 2018).
3.3.2 Endogeneity Issue
The endogeneity of the relationship between objective air pollution index and subjective well-being is discussed in terms of mutual causation and omitted variables. Local governments often develop their economies at the expense of air pollution, and the higher the economic level, the higher the happiness of local residents, while subjective well-being is also affected by other omitted variables, such as regional and resident heterogeneity, all of which can lead to potentially biased results from the OLS estimation. In order to exclude this endogeneity disturbance, the ventilation coefficient is selected as an instrumental variable to be estimated to address the potential endogeneity issue in this study.
The ventilation coefficient, developed as an instrumental variable in this study, is an indicator of air movement at various locations in China (Hering & Poncet, 2014). The relationship between wind speed and atmospheric boundary layer height is given by the ventilation coefficient: ventilation coefficient = wind speed × atmospheric boundary layer height (Jacobson, 2002). Higher wind speeds aid the dilution and diffusion of airborne contaminants because wind speed controls the horizontal direction of pollution diffusion. The atmospheric boundary layer’s height controls the height of pollutant diffusion in the atmosphere; the lower the height of the layer, the more severe the air pollution (Li et al., 2021; Xiang et al., 2019). The conventional box model for atmospheric pollution identifies the ventilation coefficient as the factor that determines the dispersion rate of air pollutants (Jacobson, 2002). The latitude and longitude grids of the European Center for Medium-Range Weather Forecasts were used to gather pertinent information, which was then processed using ArcGIS software (ESRI) to produce the entire dataset. The ventilation coefficient impacts the air pollution concentration in the opposite direction; the higher the ventilation coefficient, the less pollution is present in the area. Although the ventilation coefficient can be regarded as exogenous to local economic activity and does not directly affect inhabitant satisfaction, it is determined by exogenous weather and geographical variables (Sun et al., 2022). Consequently, the hypothesis that the ventilation coefficient, as an instrumental variable, is correlated with the independent variable and not correlated with the dependent variable is satisfied.
4 Empirical Results
4.1 Impact of PM2.5 on Residents’ well-being
In conducting the empirical analysis, the respective variables were first tested for multicollinearity considering the possible multicollinearity between multiple variables. The results showed that the maximum value of the variance inflation factor (VIF) was less than 5 and the mean value was 1.86, indicating that the degree of correlated covariance among the variables was within a reasonable range and that there was no serious covariance problem. Ordered probit and OLS estimations of Eq. (1) are shown in columns (1) and (2) of Table 3, respectively. The estimation results show that OAP has a significant positive effect on residents’ well-being. This result is inconsistent with common sense and assumptions. Combining existing studies with the actual situation of Chinese residents and society, we conclude that there are several possible reasons for this positive relationship. First, the OAP index is highly correlated with regional economic development status. In China, the largest developing country, the rapid development of regional economic levels is often at the expense of air quality (Chen et al., 2019; Sarkodie & Strezov, 2019), while the regional socioeconomic development level can directly affect the well-being index of residents (Wu & Liu, 2021). Second, the PM2.5 index was objectively extracted, whereas residents’ well-being was a subjective variable. Although we selected many control variables, there were still some omitted variables affecting residents’ well-being that were not observed for a comprehensive index like residents’ well-being. Third, residents differ in their sensitivity to air pollution (Nam et al., 2019; Tian et al., 2022), and the level of well-being of residents living in the same area varies considerably when air pollution is the same within the same area. Fourth, PM2.5, which is more easily accessible to the lungs through breathing, is gradually gaining attention due to its smaller particle diameter. However, there are other pollutants in the air, such as PM10, SO2, and NO2, which also cause a decrease in air quality, which affects residents’ subjective sense of well-being. We have discussed endogeneity issues such as the reverse causality and omitted variables mentioned above, as well as the heterogeneous situation of individual residents, in later sections.
Table 3
The impact of PM2.5 on residents’ well-being
(1)
(2)
(3)
(4)
(5)
Ordered Probit
OLS
2SLS
2SLS
2SLS
Ln PM2.5
0.263***
0.179***
-2.071***
-2.521***
(0.035)
(0.024)
(0.270)
(0.325)
Ln VC
-0.089***
-0.083***
(0.006)
(0.007)
Age
0.005***
0.003***
0.000*
0.003***
-0.001
(0.001)
(0.000)
(0.000)
(0.001)
(0.001)
Gender
0.001
0.000
0.003
0.005
0.001
(0.018)
(0.012)
(0.004)
(0.015)
(0.016)
Married
0.102***
0.071***
0.016**
0.109***
0.107***
(0.028)
(0.019)
(0.006)
(0.024)
(0.025)
Health
-0.305***
-0.203***
-0.020***
-0.244***
(0.012)
(0.008)
(0.002)
(0.011)
Education
0.004
0.002
-0.010***
-0.017**
0.013
(0.009)
(0.006)
(0.002)
(0.008)
(0.008)
Employment status
0.053***
0.035***
0.000
0.036***
(0.008)
(0.005)
(0.002)
(0.006)
Housing
0.180***
0.124***
0.011*
0.134***
(0.034)
(0.023)
(0.007)
(0.028)
New employment type
-0.081
-0.045
-0.013
-0.074
(0.132)
(0.088)
(0.032)
(0.107)
Per capita household income
0.206***
0.138***
-0.005*
0.131***
(0.015)
(0.010)
(0.003)
(0.012)
Ln GDP
-0.010
-0.006
-0.134***
-0.336***
-0.350***
(0.024)
(0.016)
(0.004)
(0.042)
(0.052)
Industry structure
0.312***
0.213***
0.102***
0.592***
0.610***
(0.094)
(0.062)
(0.024)
(0.095)
(0.102)
Degree of fiscal decentralization
-0.481***
-0.325***
0.519***
0.764***
0.962***
(0.064)
(0.043)
(0.012)
(0.134)
(0.166)
Medical coverage level
-0.038***
-0.024**
0.051***
0.095***
0.116***
(0.015)
(0.010)
(0.004)
(0.020)
(0.022)
Precipitation
0.034
0.027
-0.471***
-0.887***
-1.092***
(0.034)
(0.023)
(0.009)
(0.110)
(0.135)
Temp
-0.002
-0.002
0.036***
0.077***
0.094***
(0.003)
(0.002)
(0.001)
(0.010)
(0.012)
_cons
2.011***
7.754***
17.290***
21.163
(0.279)
(0.121)
(1.818)
(2.210)
N
15,551
15,551
15,551
15,551
15,612
First Stage F-Stat
562.05***
523.64
Notes: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively
Possible reasons for the positive relationship between OAP and residential well-being are given above. Potential endogeneity issues, such as reverse causality and omitted variables, greatly underestimate the negative effects of air pollution on residential well-being. Therefore, in this study, the ventilation coefficient was selected as the instrumental variable to be estimated to address the endogeneity issue. The ordered probit and OLS regression results in Table 3 are comparable, and the OLS results are more intuitive (MacKerron & Mourato, 2009).Referring to Yuan (Yuan et al., 2018), we also used the 2SLS method to estimate the effect of OAP on residents’ well-being.
Columns (3) and (4) of Table 3 report the regression results of the instrumental variables method. The results of the first-stage regression showed that the ventilation coefficient variable provided a reasonable explanation for the OAP variable at the 1% significance level; the F-value was 294.45, which was much larger than the critical value of 10, indicating that the ventilation coefficient passed the weak instrumental variable test. The results of the second stage show that the coefficient estimate of OAP is -2.099 and is significant at the 1% level and that OAP significantly reduces residents’ well-being. This result is consistent with the hypothesis proposed in this study and with the findings of an existing study (Ahumada & Iturra, 2021), indicating that an endogeneity problem exists between OAP and residents’ well-being. The results also demonstrate that the instrumental variables approach can effectively address the endogeneity problem and ultimately reflect the actual relationship between the two.
In order to ensure the robustness of the results of the instrumental variables approach, we deleted a portion of the control variables and reran the regressions because we consider that this portion of the control variables has the potential to be a “bad control”, and the results of the regressions are shown in column (5) of Table 3. After removing this control variable, the regression results are consistent with the previous results, which indicates the robustness of the regression results in this study.
4.2 OAP, SAP, and Residents’ well-being
Due to the individual variability of residents, the OAP situation faced by residents living in the same region is consistent. However, the different sensitivities of residents to air pollution and aspects other than air quality, such as knowledge level and employment, also vary greatly, resulting in different subjective evaluations of the same air quality, which affects the level of well-being of residents. Therefore, does the objective level of air pollution affect residents’ subjective evaluation of air quality, and thus their level of well-being? The following is a detailed discussion of this question.
By exploring whether OAP affects the level of well-being through residents’ subjective perception of air pollution, we measured individual differences in residents’ exposure to OAP and tested the validity of the influence mechanisms of OAP, SAP, and residents’ well-being. We tested the mechanism using Ordered Probit and OLS, and the results in Columns (1) and (2) of Table 4 show that OAP significantly and positively affects residents’ SAP; that is, the higher the PM2.5 concentration, the more serious residents’ subjective evaluation of air pollution. The results in columns (3) and (4) of Table 4 show that SAP significantly and negatively affects residents’ well-being levels, with an OLS regression coefficient of -0.125 and is significant at the 1% level. The regression results further suggest the validity of the mechanism by which OAP affects resident well-being through SAP.
Table 4
PM2.5 affects well-being through SAP
OLS
OP
OLS
OP
SAP
SAP
Well-being
Well-being
PM2.5
0.227***
0.259***
(0.031)
(0.034)
SAP
-0.126***
-0.191***
(0.006)
(0.010)
Control Variables
Yes
Yes
Yes
Yes
_cons
1.678***
3.594***
(0.356)
(0.228)
N
15,551
15,551
15,551
15,551
Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively
4.3 Moderating Effect of Income Level
The previous 2SLS regression results show that household per capita income level positively affects residents’ well-being, whereas PM2.5 concentration negatively affects residents’ well-being. So, can the level of per capita household income weaken the negative effect of OAP on residents’ well-being? To answer the above question, we categorize households into four groups according to their income levels from low to high and observe the effects of OAP on residents’ well-being at different income levels. The following regression Table 5 using 2SLS compares the differences in the effects of OAP on residents’ well-being at different income levels.
Table 5
The effect of PM2.5 on residents’ well-being at different income levels
(1)
(2)
(3)
(4)
(6)
Income_1
Income_2
Income_3
Income_4
FULL
PM2.5
6.978
-2.889***
-1.364***
-1.214***
-2.071***
(6.839)
(0.794)
(0.372)
(0.321)
(0.276)
Household income per capita
-0.014
-0.103
0.352***
1.871***
0.131***
(0.109)
(0.132)
(0.108)
(0.126)
(0.012)
Control variables
Yes
_cons
-40.376
26.911***
11.651***
9.646***
17.290***
(42.944)
(6.148)
(2.679)
(1.458)
(1.892)
N
3903
3903
3903
3842
15,551
First Stage F-stat
116.39***
164.62***
153.37***
162.74***
365.22***
Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively
This study divided the sample data into four groups according to the level of per capita income. A comparison of the regression results for each group shows that OAP is not significant for low-income households, the coefficients on OAP are significantly negative for households in other income levels, and the negative effect of OAP on well-being diminishes as the income level increases. A possible reason is that when household income is too low, the group is more concerned about the improvement of income level and satisfaction of basic living needs, as opposed to the improvement of environmental quality. As income levels increase, residents spend more of their household income to prevent possible harm caused by air pollution, thereby weakening the negative effect of OAP on well-being. On this basis, we also added the interaction term between income and objective air pollution for testing, and the interaction term was significantly negative4, further verifying the positive moderating effect of income. Therefore, we believe that the level of household income can alleviate the negative effect of objective air pollution on residents’ well-being.
4.4 Further Analysis
4.4.1 Differences in Levels of well-being between “pessimistic” and “optimistic” Residents about air Pollution
The previous section verified that OAP affects residents’ well-being by influencing SAP, and the next section further analyzes the relationship between OAP and SAP and its impact on well-being. According to the distribution of subjective and OAP in Fig. 3, when OAP was severe, approximately 35% of the population believed it to be less severe, and when it was not, approximately 32% believed it to be more severe. To what extent, then, the inconsistency between residents’ SAP assessment and OAP levels affects the level of well-being of residents is explored in detail below.
Fig. 3
Residents’ subjective differential evaluation of objective air pollution
×
To make the OAP comparable with the SAP, we also divided the objective pollution into 5 levels, and categorized the PM2.5 index from small to large into 5 groups, “very not serious”-“very serious”, assigning values 1–5. Assigning values from 1 to 5. We then constructed the indicator K to represent the degree to which SAP deviates from OAP, K = SAP - OAP. This part of the deviation is caused by the residents themselves, and the difference in the degree of deviation also represents the heterogeneity of individual residents. When K > 0, we believe that the SAP index of the residents is greater than the OAP index, which represents that the residents have a high valuation of pollution and belong to the “pessimistic” air pollution residents. When K < 0, we believe that the SAP index of the residents is smaller than the OAP index, which represents the low valuation of the residents on pollution, and belongs to the “optimistic” air pollution residents. It is emphasized that the “pessimistic” and “optimistic” residents are only classified for the purpose of air quality assessment, and do not represent the overall psychological status of the residents.
To understand whether there are differences in the levels of happiness between “pessimistic” and “optimistic” residents and the factors that contribute to the differences, the Oaxaca-Blinder counterfactual decomposition method will be used to analyze the results. The Oaxaca-Blinder decomposition method, proposed by Oaxaca (Oaxaca, 1973) and Blinder (Blinder, 1973), has the advantage of constructing a counterfactual group in a simple manner and decomposing the differences between groups into “explainable"” and “unexplained” parts. We consider “Pessimistic” residents as Pessimistic = 1; “Optimistic” residents as Optimistic = 1, and k = 0 as the control group, which means that this group of residents is able to assess the air pollution situation more rationally.
The results in Table 6 show that the mean value of well-being of the “pessimistic” air pollution residents is lower, 3.777, compared to the well-being of the residents at k = 0. The gap between the two is significant, with a coefficient of 0.176, and the portion of the gap that can be explained by the control variables is 0.024, which is 13.63% of the gap, and it can be seen that the results of control variables are more powerful in explaining this gap than that of health and regional temperature. It can be seen that the gap is more strongly explained by the health of the population and regional temperature. This result illustrates that there is a gap in the level of well-being between pessimistic and non-pessimistic residents, and that 13.63% of the gap in well-being is related to individual differences in residents, differences in the regions they live in, and differences in climate between the two groups, and that 86.36% of the gap is due to the overvaluation of air pollution by residents. Similarly, we also examined the effect of air pollution “optimistic” residents on the difference in well-being. The difference between the well-being levels of optimistic and non-optimistic residents is significant at 0.067, where 28.36% of the difference in well-being is due to factors such as physical health, type of employment, level of household income, and the natural state of the regional economy of the residents in both groups, and 71.64% of the difference is due to the under-valuing of air pollution by the residents.
Table 6
Differences in happiness levels between “Pessimistic” and “Optimistic” residents: results of Oaxaca-Blinder counterfactual decomposition
Type
Overall
Difference
Explained
Unexplained
Overall
K = 0
3.953*** (0.013)
Pessimistic = 1
3.777***
0.176***
0.024***
0.152**
(0.014)
(0.019)
(0.007)
(0.019)
Optimistic = 1
4.020***
-0.067***
-0.019**
-0.048***
(0.008)
(0.015)
(0.007)
(0.017)
Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively
4.4.2 Robustness Tests and Heterogeneity Analysis
To further test the robustness of the empirical analysis results, we conducted robustness tests using the following two methods. First, we replaced the econometric model and conducted regressions using ordered probit and OLS methods. The regression results and significance levels were consistent. Second, regressions were performed using monthly data of PM2.5 concentrations. We first ran the regression using PM2.5 data for each month, and the regression results for each month were consistent with the above results. Third, considering that residential heating in winter can significantly increase the level of air pollution, the sample data were divided into two groups for regression during the heating and non-heating seasons. The Chinese government has set a heating period from November 15 to March 15; however, the start and end of heating in each city is determined by the local government. Therefore, we set April to October as the non-heating season and November to March as the heating season. The monthly distribution of PM2.5 in Fig. 4, also shows that air pollution is more serious during the heating season. This study used monthly air pollution data to determine the within-group mean as the dependent variable for regression. The 2SLS regression results in Table 7, showed that the negative effect of PM2.5 on well-being in the heating season was greater than the effect of the non-heating season on well-being. Although China is already promoting clean heating projects in winter, livelihood measures still need scientific guidance and active deployment owing to large regional differences and unfavorable conditions such as large diurnal variations in temperature and humidity in winter.
Fig. 4
County-wide monthly PM2.5 distribution
×
Table 7
Differences in the effects of PM2.5 on well-being between heating and non-heating seasons
(1)
(2)
Well-being
Well-being
PM2.5 (No heating)
-1.718***
(0.205)
PM2.5 (Heating)
-2.387***
(0.339)
Control Variables
Yes
_cons
13.360***
20.746***
(1.213)
(2.479)
N
15,551
15,551
First Stage F-stat
735.46***
507.35***
Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively
5 Conclusion, Limitations and Policy Implications
This study combined PM2.5 concentration data from Global/Regional Estimates (V5.GL.02), a monthly global PM2.5 dataset published by the Atmospheric Composition Analysis Group of Washington University in St. Louis, the Chinese Household Income Survey dataset published by the China Institute for Income Distribution, and the China County Statistical Yearbook data. Three main research aspects were discussed. First, the magnitude and path of the OAP impact on residents’ well-being; Second, Differences in subjective well-being levels of different types of residents during air quality assessments; Third, in addition to the direct effect of income level on well-being, how does income contribute to the path through which air pollution affects well-being levels was also analyzed.
This study used ArcGIS software for data extraction and visualization analysis. After matching and regression analysis of data using Stata software (StataCorp LLC), the following conclusions were drawn. The OLS and ordered probit regression results showed a positive effect of PM2.5 on residents’ well-being, yielding results inconsistent with the hypothesis. The possible reasons for this are discussed above. (1) After accounting for possible endogeneity issues such as omitted variables and reverse causality, the regression results using county ventilation coefficients as instrumental variables showed that OAP has a significant negative effect on residents’ well-being, a finding consistent with the hypothesis and indicative of the validity of the instrumental variables. (2) When considering PM2.5, as a measure of a criterion that does not include psychosocial conditions, there may be a problem of not capturing its true impact on well-being. Therefore, we verified the existence of a mechanistic pathway by which OAP affects well-being through residents’ subjective air quality evaluations. Meanwhile, using the Oaxaca-Blinder counterfactual decomposition method, we conclude that air pollution “pessimistic” residents have lower levels of well-being and 21.02% of the gap in their well-being is due to themselves; similarly, air pollution “optimistic” residents have higher levels of well-being and 27.56% of the gap comes from their optimistic residents themselves. (3) Besides the direct effect of household income level on well-being, we also find that it has a positive moderating effect that mitigates the negative effect of OAP on residents’ well-being. (4) Finally, we conducted a robustness check using monthly county-level PM2.5 data and divided the sample into two groups, “heating season” and “non-heating season,” in the context of China’s heating policy. The results showed that the negative effect of air pollution on residents’ well-being was greater during the heating season than during the non-heating season.
These results indicate that to improve air quality and enhance residents’ well-being, relevant policy measures should consider socioeconomic, technological, and subjective factors. First, China’s urban development should insist on giving more prominence to ecology and security, not considering economic growth as the only goal, and should focus on enhancing people’s sense of access, well-being, and security. Second, it should aim to improve residents’ recognition of air pollution and their environmental awareness. Successful implementation of air pollution control depends on educating locals about ecological civilization. This promotes the locals’ awareness of social, environmental, and self-protection by spreading air pollution knowledge more widely, raising public awareness of the effects of air pollution, and enhancing the locals’ quality of living conditions and psychological safety. Third, it promotes technological innovation among various enterprises. To achieve a “win-win” scenario for environmental protection and economic growth, local governments should develop reasonable environmental regulation policies that allow businesses to advance simultaneously in production and pollution control technologies, boost enterprise productivity, and become more competitive in the market. However, as residents are more sensitive to air pollution, companies should actively develop products that can be used by residents to reduce the damage caused by air pollution to the human body and reduce the psychological and physiological impacts of air pollution on residents. Fourth, a clean winter heating project should be promoted. Clean heating methods should be chosen to realistically meet the needs of urban development. Different types of green heating models should be actively explored, and the technical bottlenecks of nuclear energy, wind energy, solar energy, biomass, and other clean energy heating should be overcome, allowing residents to have a green and warm winter. Meanwhile, local governments should actively introduce pricing policies for clean heating, reduce the burden of electricity costs for clean heating on residents, and increase the level of well-being while bringing a sense of security to residents.
To further advance our understanding of the connection between air quality and individual well-being, future studies must address the limitations not fully addressed in this study. This study explored the potential logical relationship between air pollution and residents’ well-being as thoroughly as possible. First, this study used county-level annual average PM2.5 concentrations matched to a microscopic database. A more detailed spatial analysis of air pollution affecting resident behavior could be further developed if detailed satellite images and timely microscopic survey data for specific time periods are available. Second, for the identification of pollutants, this paper only uses PM2.5 concentration as the OAP measurement. Although PM2.5 is more damaging to the human body and of greater concern to residents, other pollutants in the air can also have an impact on the daily lives of residents. If more types of pollutants (SO2, NO2, and PM10) can be included in the OAP measurement in the future and fitted into a comprehensive air pollution index by scientific methods or the impact of different pollutants on residents’ behavior be compared, it will help regions adopt different treatment methods for different pollutants. Third, we will continue to conduct questionnaire research to collect data, with a view to breaking down the proportion of well-being influences that come from consciousness, as well as the proportion of physiological factors that are independent of consciousness, based on existing research ideas. Finally, we also observed that with the COVID-19 pandemic, exposure to airborne PM2.5 increases the risk of infection and exacerbates the symptoms of those infected with the virus. In this situation, the management of air pollution becomes particularly urgent and important. Therefore, we need to continue exploring the combined effects of air pollution and COVID-19 on the behavior of the population.
Declarations
Conflict of interest
We would like to declare on behalf of my co-author that the work described is original research that has not been published previously and is not under consideration for publication elsewhere, in whole or in part. We confirmed that no conflict of interest exists in the submission of this manuscript and is approved by all authors for publication in your journal.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The hierarchy of administrative divisions in China includes provincial, prefectural, county and township administrative divisions. China’s " county " and " district " belong to the same administrative level, both belong to the county administrative district, of which " county " refers to the basic administrative unit, " district " refers to the administrative unit within a city.