Do CEOs Realize the Negative Impact of Air Pollution? Evidence from Voluntary CEO Turnovers

ABSTRACT

We examine whether air pollution surrounding corporate headquarters affects CEO turnover. We assume that CEOs tend to have higher quality of life requirements and are more sensitive to air pollution. Using the data of A-share listed companies in China from 2000 to 2019, we empirically find that compared to CEOs experiencing clean air, those exposed to air pollution are more likely to depart and relocate to areas with better air quality. Additionally, these results are more pronounced when CEOs come from relatively less polluted overseas regions and when firms are located in regions with a higher degree of economic development, and they operate in highly competitive industries. These findings expand upon the factors that influence CEO turnover and provide empirical evidence that regional air pollution will bring about the flow of human capital, or even the loss of talents.

KEYWORDS:

• Air pollution
• CEO turnover
• cross-regional flow

JEL:

• G30
• G34
• J16

1. Introduction

Hugo Barra, former vice president of Xiaomi’s international business, wrote in a Facebook post, “The last few years of living in such a singular environment have taken a huge toll on my life and started affecting my health,” citing pollution as the main reason for his decision to go. With the acceleration of industrialization and global economic growth, the ecological environment is deteriorating day by day. How to coordinate economic growth and sustainable development has become an important challenge for human beings. As the world’s largest developing economy, China has experienced rapid economic growth, but inevitably faces serious environmental pollution problems due to low-cost manufacturing and loose environmental policies. According to the Global Environmental Performance Index (2022),¹ China’s air quality is among the worst in the world (160/180). The World Health Organization report (2019)² also shows that air pollution has become the biggest threat to human health, and more than seven million people die each year from diseases linked to air pollution. From the perspective of corporate human capital, anecdotal evidence reveals that most executives pay more attention to their health and the quality of the environment where they live. Therefore, our study seeks to find out whether air pollution around corporate headquarters affects CEOs’ turnover choices.

In this article, we examine the effect of air pollution on CEO voluntary departures, which is representative of similar issues among highly skilled professionals in the capital market. We argue that air pollution has several detrimental effects on CEOs. First, CEOs tend to belong to a high-income group and are more likely to have higher quality-of-life requirements (Xue, Zhang, and Zhao 2019). They have a greater economic ability to choose places with better air quality to live (Chay and Greenstone 2005; Roback 1982). Second, as people of high social status, CEOs have more knowledge and a better understanding of the harmful effects of air pollution, and the costs of finding other jobs are lower for them (Arntz 2010). Furthermore, if CEOs leave a company because of air pollution, it is possible to trace their movement and determine whether they have moved to firms in less polluted areas. Thus, we predict that, relative to CEOs working in areas with high air quality, those experiencing air pollution are more likely to depart and migrate to areas with less pollution.

With respect to the consequences of air pollution in finance and accounting research, most studies have focused on the impact of air pollution on the performance and decision-making of financial market participants (Li et al. 2020; Dong et al. 2021; Li, Luo, and Soderstrom 2020). For instance, air pollution has been shown to adversely affect analysts’ forecast accuracy (Dehaan, Madsen, and Piotroski 2017), investors’ decision making (Hirshleifer and Shumway 2003; Huang, Xu, and Yu 2020), agent performance, and worker productivity (Chang et al. 2019; Zivin and Neidell 2012). However, little micro-level evidence has shown whether and how firms and their Chief Executive Officers (CEOs) react to air pollution.

CEO turnover is an important topic in corporate finance and governance. Substantial empirical literature has focused on the relationship between turnover and compensation when a firm performs poorly (Bhagat and Bolton 2008; Engel, Hayes, and Wang 2003; Ghosh and Wang 2019; Jenter and Kanaan 2015; Murphy and Zimmerman 1993). Although many studies have consistently found that CEO turnover is inversely related to firm performance, the causal relationship between CEO turnover and corporate performance remains controversial (Brickley 2003). As Jiang et al. (2017) point out, it is difficult to ascertain whether turnover is forced because underpayment based on firm performance affects subsequent voluntary CEO departures. Consequently, it is better to explain the causality of CEO turnover from the perspective of non-monetary rather than monetary factors. More specifically, using air quality surrounding the corporate headquarters as a proxy for non-monetary factors, we investigate the causal effect of non-monetary factors on CEO turnover. This setting allows us to examine whether CEO turnover is voluntary as the air quality of corporate headquarters is a clean event that deteriorates the quality of the CEO’s living environment. Hence, if firms located in more polluted areas are less able to retain CEOs, it is reasonable to expect a causal relationship between air pollution and voluntary CEO departures.

Specifically, we created a unique dataset of career paths of executives and combine it with a dataset on air pollution. We first identified whether a firm is located in a polluted area by using the air quality index (AQI) in its headquarters. It is generally agreed that AQI values above 100 indicate unhealthy air conditions (Li et al. 2021). Then, we collected data on the career path for all executives at Chinese listed firms from 2000 to 2019 from the China Corporate Figure Characteristics Series database (GTA_TMT) in China Stock Market and Accounting Research (CSMAR). From the data, we traced where the executives worked, when they departed, and to which firms they migrated. Consistent with the perspective that participants in financial markets are not immune to poor air quality, our baseline results show that a city’s AQI is positively correlated with CEO turnover. Our results indicate that control for other variables, for every one-unit increase in air quality index (AQI), the odds of CEO departure become 2.36 times greater. Moreover, in the sample of CEO departures, the regression coefficient of AQI to migration was significant at 1% level, indicating that CEOs were more likely to depart and migrate to areas with better air quality.

We conducted several robustness checks to strengthen our interpretation of the results. First, we verified that our findings are not driven by poor firm performance or CEOs approaching retirement age. To solve endogeneity problems, we followed Chen, Oliva, and Zhang (2017) and used the strength of thermal inversion in a region as an instrument for the level of air pollution. Finally, considering that the official air quality index may be subject to manipulation, we re-estimated our models after excluding the three intervals of air quality index at 95–105, 90–110 and 80–120. We consistently found a significant and positive relationship between air pollution and CEO turnover.

We then conducted several cross-sectional tests to add more evidence to our argument. In particular, if air pollution at the corporate headquarters increases the probability of CEO turnover, such an effect should vary cross-sectionally depending on the CEOs’ ability to find a new job and their sensitivity to air pollution. First, we considered national culture and classified CEOs as foreign or non-foreign. These results suggest that the effect of air pollution on CEO turnover is more pronounced when foreign CEOs come from countries or districts with better air quality, where people’s health concerns are more sensitive to air pollution. Second, the positive pollution-turnover relationship is stronger for firms located in a region with a higher degree of economic development, which reflects local network spillovers and faster learning of skilled individuals, making it easier for departing CEOs to obtain new jobs in areas with less air pollution. Third, we show that the link between pollution and turnover is stronger for firms operating in highly competitive industries as firms with many industrial peers typically face tougher competition for managerial talent. Overall, the cross-sectional tests corroborated our hypothesis.

This study contributes to the existing literature in at least two ways. First, it documents a novel fact about CEOs negative reactions to air pollution. Previous studies have shown that air pollution adversely affects analysts’ forecast accuracy, investors’ decision-making, agent performance, and worker productivity (Chang et al. 2019, Dehaan, Madsen, and Piotroski 2017; Hirshleifer and Shumway 2003; Huang, Xu, and Yu 2020; Zivin and Neidell 2012). More broadly, discuss a wide range of decisions made by capital market participants under the influence of air pollution. In this article, we study the CEOs’ reactions to air pollution at their corporate headquarters. We argue that air pollution surrounding corporate headquarters prompts incumbent CEOs to leave and migration to areas with better air quality.

Second, research on CEO turnover has focused on the relationship between turnover and CEO compensation, firm performance (Dai et al. 2020; Denis and Denis 1995; Jenter and Kanaan 2015; Weisbach 1995). From the perspective of voluntary turnover, we examine whether non-monetary and external factors influence CEO departure, and our findings enrich the literature related to the causes and consequences of CEO turnover. We document novel and robust evidence that non-monetary and external factor—air pollution surrounding corporate headquarters engenders voluntary CEO turnover. In addition, our research also found that CEOs not only leave their jobs because of local air pollution, but also migrate, resulting in a movement of human capital. Xue, Zhang, and Zhao (2019) examined the relationship between air pollution and migration and confirmed that air pollution induces people to choose relatively less polluted areas as their intended workplace. Our results suggest that highly skilled professionals such as CEOs can realize the negative impact of air pollution and voluntarily leave firms located in polluted air areas. In addition to extending the labor economics literature on the negative effects of air pollution on human capital migration (Chen et al. 2018; Xue, Zhang, and Zhao 2019), the results in this article also shed light on a better understanding of environmental protection and green development.

The remainder of this article is organized as follows: Section 2 develops the research hypotheses, and Section 3 summarizes the data and introduces the empirical design of our study; Section 4 presents the empirical results; finally, in Section 5 concludes the article.

2. Literature Review and Hypotheses Development

2.1. The Impact of Air Pollution on CEO Departure

The influence of air pollution on financial market participants has been well identified and discussed in the economics and finance literature. The theoretical link between air pollution and decision making in these studies is the negative mood induced by air pollution, which has been well established in related studies (Huang, Xu, and Yu 2020). Dong et al. (2021) document a negative relationship between air pollution during corporate site visits by investment analysts and subsequent earnings forecasts. Given that higher pollution leads to increased pessimism, they show that a city’s AQI on the date of a site visit is negatively correlated with the visiting analyst’s subsequent earnings forecast. In addition to fostering a negative mood, an emerging body of research has investigated how the negative effects of air pollution on physical health affect participants’ behavior in the financial market (Chang et al. 2019; Zivin and Neidell 2012). For example, using a unique panel dataset on the daily productivity of employees in a pear-packing facility in Northern California, Chang et al. (2019) found a statistically significant and negative impact of PM2.5 on the productivity of indoor workers.

Building on the above literature, we argue that firms and executives also react to air pollution and that the negative effects (bad mood and physical effects) of air pollution may cause CEO to depart and migrate to areas with less air pollution. Recent studies have provided evidence that people living in areas with poor air quality can adopt defensive behaviors, such as purchasing air purifiers (Ito and Zhang 2020), to mitigate exposure to air pollution. If these provisional defensive behaviors cannot ensure health or are too costly in the long run, people will ultimately seek to settle in areas with better air quality. As Xue, Zhang, and Zhao (2019) indicate, migration sorting is an alternative defensive behavior that prevents exposure to air pollution, especially for skilled labor; hence, we expect air pollution to have a pronounced effect on CEOs.

Executives are more likely to be in a high-income group. These individuals tend to have a higher quality of life and are more sensitive to air pollution. They also have greater economic ability to sort themselves into locations where better air quality is impounded by housing prices (Chay and Greenstone 2005). Moreover, CEOs have more knowledge and a deeper understanding of the harmful effects of air pollution and thus lower tolerance for poor air quality. They also have more information on job opportunities and face lower costs when searching for new jobs (Arntz 2010). In line with this view, a substantial body of economic literature has addressed the importance of the living environment in people’s career choices. For example, Levine et al. (2018) found that firms exposed to the opening of toxic plants are more likely to experience CEO turnover. In sum, based on the literature, we expect air pollution to have a more detrimental effect on CEO departure and hypothesize the following:

H1:

Air pollution surrounding corporate headquarters triggers executives to leave and migrate to areas with less pollution.

2.2. The Impact of Air Pollution on CEO Turnover—Cross-Sectional Variation

In addition to determining the overall effect of air pollution surrounding corporate headquarters on CEO turnover, we further investigate the characteristics of different firms and CEOs. To influence the CEO turnover of firms located in air-polluted areas, CEOs need to have (1) the ability to find a new job and (2) corporate executives’ sensitivity to the air quality surrounding the corporate headquarters. Below, we develop predictions based on variations in the CEOs’ ability to find a new job and their sensitivity to air pollution surrounding corporate headquarters.

One of the key premises of the argument for H1 is the sensitivity of CEOs to the air quality surrounding corporate headquarters. Foreign executives increasingly complain about pollution in China and its perceived impact on their health and that of their families. In recent years, several high-profile executives have left China, citing pollution as the main reason for their decision; therefore, foreign CEOs have a strong sensitivity to air pollution. Accordingly, the effect of air pollution surrounding corporate headquarters should be stronger for foreign executives. Our second hypothesis is as follows:

H2:

The effect of air pollution surrounding corporate headquarters on CEO turnover is stronger for foreign CEOs.

Another premise underlying H1 is CEOs’ ability to find a new job. With China’s economic development, air pollution has grown rapidly in recent decades. In some regions, such as Beijing, Guangdong, and Shanghai, the economies are well developed, but the air quality is significantly different (Chen et al. 2018). We differentiate companies based on the degree of economic development of the regions in which they are located. According to Francis et al. (2016), cities with higher economic development reflect local network spillovers and faster learning by skilled individuals, and departing CEOs find it easier to obtain new jobs in less polluted areas. As such, we expect the effect of air pollution on CEO turnover to be stronger for firms in areas of higher economic development. Thus, our third hypothesis is as follows:

H3:

The effect of air pollution surrounding corporate headquarters on CEO turnover is stronger for firms located in areas with a higher economic development.

According to previous research, if an affected firm faces greater competition in its labor market, it may pay a higher premium to its CEO (Dai et al. 2020; Liu 2014). In other words, CEOs in a tight labor market can expand their outside employment options and thus increase turnover probability (Liu 2014). Firms in competitive industries with many industrial peers and facing tougher competition for managerial talent tend to accentuate the relationship between air pollution and CEO turnover. Thus, our fourth hypothesis is as follows:

H4:

The effect of air pollution surrounding corporate headquarters on CEO turnover is stronger for firms in more competitive industries.

3. Empirical Approach

3.1. Sample Selection and Data Sources

3.1.1. Sample Selection

Our study is based on all non-financial companies publicly traded in the Chinese A-share market from 2000 to 2019. The initial sample was based on the following criteria: (1) specially treated (ST) firms were excluded because they often have abnormal financial status and are at risk of being delisted; (2) the CEO’s identity was known at both the beginning and end of a given fiscal year so that turnover events could be identified; and (3) the incumbent CEO had been in office for at least 1 year, so that firm performance in the previous year could be attributed to them.

The CSMAR provides information on the reasons stated for a turnover (if any): (1) change of job, (2) retirement, (3) contract expiration, (4) change in controlling shareholders, (5) resignation, (6) dismissal, (7) health, (8) personal reasons, (9) corporate governance reform, (10) legal disputes, (11) no reason given, and (12) completion of acting duties. Since we were interested in how air pollution surrounding corporate headquarters impacted CEO voluntary turnover, we excluded observations if the incumbent CEO had retired, had been dismissed, or if the firm had been acquired during the fiscal year. After ensuring the availability of traditional explanatory variables for CEO turnover, our final sample consisted of 11,365 firm—year observations for 2000–2019. During the 20-year sample period, 6,739 CEO turnover events were identified.

3.1.2. Data Sources

The data used in this study were obtained from multiple sources. For each city in China, we obtained daily information on air pollution (AQI) from the official website of the Ministry of Environmental Protection of China (MEPC). For a small fraction of city-day observations, AQI readings were unavailable via the MEPC. We were able to fill in some of the missing data from the Qingyue Open Environment Data Center website, which obtains pollution data directly from local governments.

We collected executive career path information from the China Corporate Figure Characteristics Series database (GTA_TMT) in China Stock Market and Accounting Research (CSMAR). The CSMAR provides information on the reasons stated for a turnover, however, it does not provide location information of the headquarters of all publicly listed firms’ CEOs. Then we manually searched the firms’ annual reports, company websites, and other online sources (e.g., Google and Baidu.com) to complement it. In other words, we tracked the employment status of 10,356 CEOs in the sample firms, discovering where executives worked, when they departed, and to which firms they migrated.

Financial and market data were obtained from the CSMAR database. We collected information on CEOs and firms, such as CEOs’ compensation, age, educational background, firm size, and performance. Industry classification was based on the China Securities Regulatory Commission 2012 Industry Classification Standard.

3.2. Empirical Model Specifications and Main Variables

Following the literature on turnover, we conducted a multiple regression analysis of the effect of air pollution surrounding corporate headquarters on CEO turnover. We used the following logit regression model:

(1) $\mathit{turnove}{r}_{\mathit{i},\mathit{t}+1}={\beta }_0+\mathit{\hspace{0.17em}}{\beta }_1\mathit{AQ}{I}_{\mathit{i},\mathit{t}}+\mathit{\hspace{0.17em}}\sum {\beta }_n\mathit{Control}{s}_{\mathit{i},\mathit{t}}+\mathit{\hspace{0.17em}}{\epsilon }_{\mathit{i},\mathit{t}}$(1)

where turnover_i,t + 1 is firm i’s CEO turnover in year t + 1, which is an indicator that equals one if a firm experiences CEO turnover during the next fiscal year (t + 1) and zero in all other fiscal years. Specifically, CEO turnover is measured by CEO departure and migration. For all turnover variables, we followed Eisfeldt and Kuhnen (2013) by recording a departure event when the executive was at the firm at the end of the current fiscal year but not at the end of the following fiscal year. We constructed two measures of CEO turnover: we defined executive departure as a dummy that equaled one if a firm experienced a CEO departure during the following fiscal year (t +1) and zero in all other fiscal years; executive migration is a dummy that equaled one if the CEO was leaving the region where the firm was domiciled and migrated to a less air-polluted area during the next fiscal year (t + 1) and zero in all other fiscal years. When more than one CEO was in charge of the company in the same financial year, we chose the person who had been the CEO for the longest period (e.g., if a change occurred in 2008 but the entrant CEO was only in charge from October to December, that is, for only 3 months, we considered the 2008 CEO to be the predecessor and registered a CEO change in 2009). Although this is an imperfect measure, to the best of our knowledge, no database has classified turnover.

AQI_i,t is either the AQI or an indicator for AQI categories. Following Li et al. (2021), air pollution was measured using the AQI published by the Chinese Ministry of Environmental Protection (MEP). The AQI in China ranges from 0 to 500. It is generally agreed that AQI values above 100 indicate unhealthy air conditions. As robustness checks, we also consider alternative air pollution measures that particulate matters less than 2.5 micrometers in diameter (i.e., PM2.5) which can deposit in people’s lungs and pose grave health risks. In robustness tests, we repeated our analysis using alternative air pollution measures, such as particulate matters less than 2.5 micrometers in diameter (i.e., PM2.5) which can deposit in people’s lungs and pose grave health risks.

Following the turnover literature (Focke, Maug, and Niessen-Ruenzi 2017; Jenter and Kanaan 2015; Liu 2014), we controlled CEO and neighborhood region characteristics in our model. In addition, we included Year and Industry indicators to control for potential differences in executive turnover packages across time and industries. (Please see the Appendix for variable definitions.) To mitigate the effect of outliers, we winsorized all continuous variables at the 1% level in both tails of distribution.

4. Empirical Results

4.1. Descriptive Statistics and Univariate Results

Table 1 presents the descriptive statistics of the variables used in the analysis. In our study, departing CEOs were 20% on average, and 24% of these departing CEOs went to less polluted areas. With respect to the work environment, about 17% firms were located in polluted air areas, and the mean (median) AQI was 0.082 (0.081) thousand. All other variables were consistent with prior research (Chen, Oliva, and Zhang 2017; Xue, Zhang, and Zhao 2019).

Table 1. Summary statistics.

variable	N	mean	sd	min	p50	max
Turnover proximity
departure	11,365	0.197	0.443	0.000	0.000	1.000
migration	2,231	0.238	0.479	0.000	0.000	1.000
Environmental variables
AQI	11,365	0.082	0.114	0.023	0.081	0.379
AQI_dummy	11,365	0.171	0.394	0.000	0.000	1.000
AQI_median	11,365	0.073	0.097	0.012	0.071	0.483
Firm characteristics
Capital_exp	11,365	18.165	2.137	12.112	18.269	23.313
EPS	11,365	0.286	0.446	−1.227	0.231	2.007
Firm_age	11,365	2.713	0.489	0.000	2.773	4.176
HHI	11,365	0.087	0.113	0.009	0.077	0.591
ROE	11,365	0.073	0.162	−0.796	0.072	0.511
ROIC	11,365	0.071	0.114	−0.418	0.066	0.414
stockret vol	11,365	0.042	0.065	0.011	0.037	0.242
Growth	11,365	0.224	0.565	−0.706	0.179	3.875
Duality	11,365	0.194	0.395	0.000	0.000	1.000
outratio	11,365	0.351	0.117	0.000	0.330	0.571
Boardsizes	11,365	8.864	1.949	1.000	9.000	19.000
cr1_10	11,365	0.592	0.157	0.227	0.606	0.907
CEO characteristics
gender	11,365	0.800	0.500	0.000	1.000	1.000
ceoage	11,365	3.884	0.147	3.500	3.887	4.170
tenure	11,365	4.144	3.065	0.000	3.000	26.000
CEO comp	11,365	8.924	3.427	0.000	7.629	14.863
neighbourhood
Density	11,365	6.417	0.722	3.741	6.510	7.820
house price	11,365	8.854	0.781	7.311	8.802	10.988
per GDP	11,365	11.023	0.724	9.163	11.137	12.155
citybig4	11,365	0.334	0.472	0.000	0.000	1.000

We also manually collected the departure and cross-province migration data of CEOs of listed companies in China. The data shows that most executives who departed from firms located in polluted air areas subsequently worked in less polluted areas. Furthermore, we conducted mean-variance tests to assess whether firms located in air-polluted areas and those located in less polluted areas are significantly different. The results show that the two subsamples are statistically different and that the average turnover probability is higher for firms located in areas with polluted air. (Please see Appendix A and B of supplemental material.)

4.2. Effect of Air Pollution on Executive Turnover

We estimate logit models to examine the impact of air pollution on CEO turnover probability. Jenter and Kanaan (2015) document that aggregate industry and market conditions also affect the likelihood of CEO turnover; therefore, year and industry fixed effects were included in the regressions. All standard errors were adjusted for firm-level clustering.

Table 2 presents the baseline regression results for the effect of air pollution on CEO turnover. In Columns (1) to (4), we use departure to represent CEO turnover. The coefficients of AQI are positive and significant at the 1% level. Furthermore, the results are robust to excluding or including firm-specific control variables, CEO-specific characteristics, and region-specific characteristics, and the estimated coefficients on the AQI variables change little when conditioning on firm, CEO, and regional traits. Specifically, the coefficient on AQI in Column (4) is 0.860 (t = 3.87). In terms of economic effects, for one unit increase in air pollution (AQI), the odds of CEO departure are 2.36 times greater.³ These results confirm that the likelihood of observed CEO turnover events is significantly higher for firms located in polluted air areas than for those located in regions with better air quality.

Table 2. Air pollution and CEO turnover: baseline specifications.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	departure	departure	departure	departure	migration	migration	migration	migration
AQI	0.799***	0.828***	0.780***	0.860***	8.165***	8.133***	8.117***	7.762***
	(5.93)	(5.78)	(4.44)	(3.87)	(19.22)	(17.73)	(14.32)	(10.41)
Firm characteristics
capital_exp		−0.003**	−0.006***	−0.007***		−0.012***	−0.003	−0.005
		(−2.09)	(−2.81)	(−2.58)		(−3.96)	(−0.60)	(−0.92)
EPS		−0.007	−0.001	0.000		0.039***	0.024	0.020
		(−0.84)	(−0.14)	(0.01)		(2.75)	(1.23)	(0.95)
firm _age		0.017*	0.011	0.019		−0.000	0.001	−0.014
		(1.93)	(0.94)	(1.30)		(−0.02)	(0.04)	(−0.43)
HHI		0.123	0.140	0.245**		−0.660***	−0.662***	−0.472***
		(1.62)	(1.45)	(2.22)		(−6.10)	(−4.03)	(−3.23)
ROE		−0.111***	−0.065	−0.057		−0.080	−0.104	−0.092
		(−3.07)	(−1.25)	(−0.94)		(−1.51)	(−1.53)	(−1.31)
ROIC		−0.224***	−0.248***	−0.283***		0.235**	0.228**	0.189
		(−3.76)	(−2.96)	(−2.87)		(2.49)	(2.00)	(1.49)
stockret_vol		0.293**	0.085	0.149		0.050	0.473	0.418
		(2.02)	(0.41)	(0.62)		(0.24)	(1.20)	(1.01)
Growth		0.007	0.015**	0.018**		0.007	−0.009	−0.007
		(1.38)	(2.18)	(2.20)		(0.91)	(−0.89)	(−0.64)
Duality		−0.011	0.242***	0.222***		0.030**	0.062*	0.043
		(−1.58)	(6.31)	(5.07)		(2.33)	(1.70)	(0.98)
outratio		−0.027	−0.067	0.050		−0.163*	−0.091	−0.142
		(−0.53)	(−0.95)	(0.58)		(−1.92)	(−0.69)	(−0.96)
Boardsizes		−0.003*	−0.008***	−0.008***		−0.008***	−0.003	−0.002
		(−1.95)	(−3.79)	(−3.24)		(−2.88)	(−0.61)	(−0.50)
cr1_10		−0.033	−0.056**	−0.036		−0.043	−0.043	−0.086
		(−1.63)	(−2.14)	(−1.15)		(−1.07)	(−0.80)	(−1.46)
CEO characteristics
gender			0.075***	0.080***			0.061**	0.055*
			(7.46)	(6.59)			(2.16)	(1.70)
ceoage			0.271***	0.281***			−0.308***	−0.266***
			(10.69)	(9.09)			(−5.39)	(−3.94)
CEO_comp			0.000	0.001			0.020***	0.020***
			(0.30)	(0.59)			(8.01)	(6.89)
Tenure			−0.003**	−0.003**			−0.006**	−0.006**
			(−2.21)	(−2.31)			(−2.55)	(−1.97)
Neighbourhood
Density				0.002				0.033*
				(0.22)				(1.89)
house_price				−0.034**				0.015
				(−2.28)				(0.50)
per_GDP				0.020				0.055**
				(1.49)				(1.98)
citybig4				−0.000				−0.114***
				(−0.03)				(−3.25)
Ind	YES	YES	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES	YES	YES
N	11365	11365	11365	11365	2231	2231	2231	2231
Pseudo R^2	0.028	0.028	0.043	0.049	0.305	0.313	0.362	0.374

Note: This table presents the regression estimations of the impact of air pollution on CEOs mobility. Figures in brackets are z-value. Statistical significance at the 1%, 5%, and 10% levels is denoted by ***, **, and *, respectively.

We also introduced a second proxy for CEOs mobility, migration, to measure whether the CEO was leaving the region where the firm located and migrating to areas with less air pollution. The results are presented in Columns (5) to (8) of Table 2. The coefficients of AQI are significantly positive at the 1% level, which is consistent with the main prediction. Specifically, the coefficient on AQI in Column (8) is 7.762 (t = 10.41). This result implies that firms headquartered in air polluted areas induced CEO to relocate to areas with better air quality.

4.3. Sensitivity Tests

4.3.1. The Effect of CEO Age and Corporate Performance on the Association Between Air Pollution and CEO Turnover

In this section, we investigate alternative channels through which air pollution may lead to CEO turnover. Relative to prior research, age could drive the positive association between air pollution and turnover probability because older CEOs nearing retirement are more concerned about the quality of life in the work environment (Liu 2014). To address this concern, we split the sample into two age subsamples and ran logit turnover regressions separately. Using the median age of CEOs, we partitioned the full sample into old and young subsamples and reran our basic regression separately for each subsample. In Table 3, columns (1)–(2) of Panel A report the results of departure, and columns (3)–(4) report migration. The results show that the coefficient on AQI is consistently significant and positive. These findings rule out the possibility that our results are driven by CEOs’ age.

Table 3. The effect of CEO age and corporate performance on the association between the air pollution and CEO turnover.

Panel A: The effect of CEO age
	departure		migration
	(1) young	(2) old	(3) young	(4) old
AQI	1.147***	0.535*	8.548***	7.358***
	(3.75)	(1.78)	(10.06)	(6.75)
Firm characteristics	YES	YES	YES	YES
CEO characteristics	YES	YES	YES	YES
Neighborhood characteristics	YES	YES	YES	YES
Ind	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	5691	5674	982	1249
Pseudo R^2	0.057	0.051	0.454	0.345
Prob>chi2	0.36		0.77
Panel B: The effect of corporate performance
	departure		migration
	(5) high	(6) low	(7) high	(8) low
AQI	1.019***	0.709**	8.849***	6.861***
	(3.16)	(2.46)	(7.77)	(7.76)
Firm characteristics	YES	YES	YES	YES
CEO characteristics	YES	YES	YES	YES
Neighborhood characteristics	YES	YES	YES	YES
Ind	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	4936	6429	817	1414
Pseudo R^2	0.057	0.045	0.441	0.350
Prob>chi2	0.59		0.68

Note: Figures in brackets are z-value. Statistical significance at the 1%, 5%, and 10% levels is denoted by ***, **, and *, respectively.

Because most studies have focused on firm performance as a determinant of CEO turnover, we further addressed whether our results may be driven by firm performance. We defined a firm as having a low performance if its return on assets (ROA) was less than the median average ROA across all firms. We classified the samples into high and low performance firms and examined the results for the two kinds of firms in Columns (5)–(8) of Panel B. The results show that, for both high and low performance, the coefficients of AQI are positively significant; thus, our results were not affected by firm performance.

4.3.2. Alternative Ways of Measuring Air Quality

Based on the baseline test of Columns (4) and (8) in Table 2, we added a fixed effect to the regression. After controlling for the firm fixed effect in Columns (1) and (5) of Table 4, the coefficients of AQI are still positively related to departure and migration.

Table 4. Higher order fixed effects and alternative proxies for air pollution.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	departure	departure	departure	departure	migration	migration	migration	migration
AQI	0.828***				7.113***
	(3.55)				(9.05)
AQI_dummy		0.080***				0.402***
		(5.25)				(15.24)
AQI_median			0.936***				8.706***
			(3.77)				(9.54)
AQI_low_day				−.034***				−0.145***
				(−6.56)				(−14.02)
Firm characteristics	YES	YES	YES	YES	YES	YES	YES	YES
CEO characteristics	YES	YES	YES	YES	YES	YES	YES	YES
Neighborhood characteristics	YES	YES	YES	YES	YES	YES	YES	YES
Firm FE	YES	NO	NO	NO	YES	NO	NO	NO
Ind	NO	YES	YES	YES	NO	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES	YES	YES
N	11365	11365	11365	11365	2231	2231	2231	2231
Pseudo R^2	0.048	0.052	0.049	0.052	0.371	0.383	0.372	0.353

Note: Figures in brackets are z-value. Statistical significance at the 1%, 5%, and 10% levels is denoted by ***, **, and *, respectively.

We also considered three alternative proxies for air pollution. In Columns (2) and (6), AQI is a dummy variable that equals 1 if the AQI is larger than 100. In Columns (3) and (7), we define AQI as the median daily AQI for a region. In Columns (4) and (8), we calculated our proxy for air quality as the proportion of days with excellent air quality (AQI ≤ 50) in a year. Table 4 presents the results based on the alternative proxies for industrial policy. In each case, our original inferences remain unchanged, suggesting that they are not simply artifacts of any particular choice of proxy for the air quality of corporate headquarters.

4.3.3. Instrumental Variable Analysis

We considered an instrumental variable strategy likely to minimize bias from both endogeneity and classical measurement errors. Specifically, we exploited a meteorological phenomenon: thermal inversion, a common meteorological phenomenon that leads to higher concentrations of pollution near the ground. The mechanism through which this occurs is as follows: under normal conditions, temperature decreases as altitude increases; as air moves from hot to cool regions, air pollution circulates vertically, decreasing air pollution concentrations near the ground. However, under certain meteorological circumstances, the temperature of a layer of air above ground could be higher than that at lower altitudes, leading to an inversion in the temperature/height gradient or thermal inversion. When this occurs, air pollution is trapped near the ground, leading to higher air pollution concentrations (see Chen, Oliva, and Zhang 2017). Because thermal inversions are meteorological phenomena and, after conditioning on weather variables, are unrelated to turnover except through pollution, they are a valid instrument.

We followed previous studies and calculated thermal inversions using the first and second layers (110 and 330 meters, respectively; Chen, Oliva, and Zhang 2017). Thermal inversion data were obtained from NASA. Specifically, within each 6-hour period, we calculated the temperature difference using the temperature in the second layer (320 meters) minus that in the first layer (110 meters). If the difference was negative, this was a normal condition, and the difference was truncated to zero.

Table 5 shows the results of the instrumental variable approach. Column (1) presents the first-stage estimates of the effect of thermal inversions on air pollution, conditional on year and industry fixed effects. The regression results of the first stage show that the F statistics are much greater than the empirical value of 10, indicating no weak instrumental variable problem. Furthermore, we found significant and robust effects of thermal inversions on air pollution, indicating that thermal inversion strength leads to increases in air pollution concentrations. The regression results of the second stage in Columns (2) and (3) show that with the increase in air pollution, the possibility of executive turnover is significantly increased, and the regression coefficient values of air pollution show little change compared to those in Table 2, which implies that the endogenous deviation is not serious in terms of the impact of air pollution on CEO turnover.

Table 5. Instrumental variable (IV) analysis.

	First-stage	Second-stage
	(1) AQI	(2) departure	(3) migration
Thermal_Inversion	0.036***
	(4.32)
AQI		0.832***	6.063***
		(5.16)	(2.87)
Firm characteristics	YES	YES	YES
CEO characteristics	YES	YES	YES
Neighborhood characteristics	YES	YES	YES
Ind	YES	YES	YES
Year	YES	YES	YES
N	7207	7207	1536
adj. R^2/Pseudo R^2	0.025	0.131	0.096
F-value	12.58***

Note: In Column (1), figures in brackets are t-value. In Column (2) and (3), figures in brackets are z-value. Statistical significance at the 1%, 5%, and 10% levels is denoted by ***, **, and *, respectively.

4.4. Further Cross-Sectional Analysis

We further attempt to identify factors that may impact the relationship between air pollution and CEO turnover. We identify three specific firm and managerial characteristics that may give rise to greater non-monetary factor influence: (1) managerial nationality; (2) regional economic development; and (3) industry competition.

4.4.1. CEOs from Overseas

Foreign CEOs are more sensitive to air pollution. On the one hand, they are more concerned about the harmful effects of air pollution on the human body; on the other hand, they have more opportunities to leave China. For example, several high-profile executives have left China in recent years, citing pollution as the primary reason for their decision. Accordingly, we expect the effect of air pollution at the corporate headquarters to be stronger for foreign executives. Column (1) and (2) of Table 6 present the results of the study. The coefficients of the interaction AQI × FC are significantly positive, suggesting that foreign CEOs have a strong incentive to leave and migrate to less polluted areas because of the air pollution at corporate headquarters.

Table 6. Heterogeneity of the effect.

	Foreign CEOs		Regional economic development		Industry competition
	(1)	(2)	(3)	(4)	(5)	(6)
	departure	migration	departure	migration	departure	migration
AQI	0.876***	7.741***	0.286	2.140*	1.271***	10.361***
	(3.80)	(10.35)	(1.23)	(1.76)	(4.62)	(14.57)
AQI_FC	6.053**	1.998*
	(2.52)	(1.86)
Foreign CEOs	−0.282	−0.148
	(−1.56)	(−0.74)
AQI_GDP			1.826*** (2.91)	4.579*** (2.92)
GDP			0.052 (0.49)	0.081 (0.89)
AQI_HHI					−1.055*** (−3.14)	−6.322*** (−6.61)
HHI					0.087*** (3.21)	0.343*** (4.87)
Firm characteristics	YES	YES	YES	YES	YES	YES
CEO characteristics	YES	YES	YES	YES	YES	YES
Neighborhood characteristics	YES	YES	YES	YES	YES	YES
Ind	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES
N	11365	2231	11365	2231	11365	2231
Pseudo R^2	0.048	0.374	0.056	0.424	0.050	0.422

Note: Figures in brackets are z-value. Statistical significance at the 1%, 5%, and 10% levels is denoted by ***, **, and *, respectively.

4.4.2. Regional Economic Development

A large body of literature has documented that a high level of human capital (e.g., labor with higher education and richer work experience) is associated with high regional income and productivity (Black and Lynch 1996). We differentiated companies according to their degree of economic development. According to Francis et al. (2016), cities with higher economic development reflect local network spillovers and faster learning of skilled individuals; thus, departing CEOs find it easier to obtain new jobs in areas with less air pollution. Therefore, we expect the effect of air pollution on CEO turnover to be stronger for firms located in areas with higher levels of economic development than for other firms.

To measure the extent of economic development, we used the province-level GDP per capita. Columns (3) and (4) of Table 6 present the results. The results show that the coefficients of AQI×GDP are positive and significant, suggesting that CEOs in cities with higher levels of economic development are more likely to leave.

4.4.3. Industry Competition (HHI)

Since a company’s position in the industry affects employee behavior, we examined how air quality affects the turnover of corporate executives under different industry concentrations. We used the revenue-based Herfindahl index for each two-digit Standard Industry Classification (SIC) code industry to capture industry competition dynamics; higher scores indicate greater concentration and less industry competition. Faleye, Reis, and Venkateswaran (2013) argue that outside employment opportunities increase employees’ bargaining power through increased competition for labor among employers and by giving credibility to employees’ voluntary turnover threats. Thus, as managers in competitive industry firms have more outside employment opportunities, we expect them to “vote with their feet” when air pollution is at play. In contrast, in the case of an affected firm with high bargaining power, it appears impossible that air pollution at the headquarters will lead to CEO turnover.

Columns (5) and (6) of Table 6 present the regression results. Beyond the basic control variables, we included an interaction variable between AQI and HHI, that is, AQI×HHI. The results show that the coefficients of the interaction are all significantly negative, indicating that CEOs in competitive industries tend to experience turnover.

5. Conclusion

This study examined the effect of air pollution on CEO turnover. Based on unique data on executives’ career paths, we identify CEOs’ career moves and merge them with a dataset on air pollution at the firms’ headquarters. This study presents robust evidence that air pollution increases the probability of CEO turnover. Specifically, we found that air pollution surrounding corporate headquarters drives CEOs to escape from their work locations and relocate to areas with air quality advantages. Our identification tests suggest that air pollution has a causal effect on CEO turnover. Moreover, we discussed heterogeneity in the pollution—turnover relationship and found that pollution affects turnover only when foreign CEOs come from countries with better air quality, when they are located in a region with a higher degree of economic development, and when firms face lower competition.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Supplementary data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/1540496X.2023.2293971

Additional information

Funding

We gratefully acknowledge the financial support provided by the National Natural Science Foundation of China [Grant Nos. 71872017, 72002116].

Notes

1. Please find the Environmental Performance Index on the website https://epi.yale.edu/.

2. Please find the World Health Organization Report on the website https://www.who.int/publications/i/item/9789241565707.

3. In logit model, for every one unit increase in X, the odds of Y = 1 becomes e^βtimes greater (other variables are constant and the coefficient is greater than 0).

Appendix

Variable definition

Table

Variable	Description (Compustat data items in parentheses)
turnover variables
departure	A dummy variable that equals one if a firm experiences a CEO depart during the next fiscal year (t + 1).
migration	An indicator variable, which equals one if the CEO moves to a new firm located less air pollution areas, and zero otherwise.
Air pollution Variables
AQI	The average daily Air Quality Index for a region. China’s Ministry of Environmental Protection (MEP) is responsible for measuring the level of air pollution in China and publishing the Air Quality Index.
AQI_dummy	Dummy variable equals one if AQI ≥ 100.
AQI_median	The median daily Air Quality Index for a region.
AQI_low_day	The proportion of days with AQI ≤ 50 in a year.
Firm characteristics
Capital_exp	Capital expenditures divided by total assets.
EPS	Earnings per share.
Firm_age	Log of the number of years since the establishment of the Firm.
HHI	The Herfindahl index of firm assets in the industry in year t.
ROE	Net income divided by total equity.
ROIC	Return on Invested Capital.
Stockret_vol	The standard deviation of a firm’s daily stock returns. This variable is set to missing if the number of trading days is less than 50 in a given year.
Growth	Sales growth in percentage in a year.
Duality	A dummy variable that equals one if the CEO and chairman are the same person in a year and zero otherwise.
Outratio	The proportion of independent directors on the board in a year.
Boardsizes	The number of directors.
Cr1_10	The number of shares held by the top ten shareholder divided by the total issued shares in a year
CEO characteristics
gender	The gender of the CEO, equal to one if the CEO is male.
ceoage	Log of the average age of firm CEO and chairman of the board of directors.
tenure	The number of years that an executive has served in the firm.
CEO comp	The natural logarithm of average compensation of the CEO
Neighbourhood
Density	Log (number of population) for a region in a year.
House price	Log of the city average house price
Per_GDP	Log (GDP per capita) for a region in a year.
City_big4	A dummy variable that equals to one if firms located in Beijing, Shanghai, Shenzhen and GuangZhou.