Can Professional Football Players Adapt to Air Pollution? Evidence From China

Abstract

In this paper, we study the impact of air pollution on Chinese professional football players’ performance. Our primary research question is whether the negative effects of air pollution can be mitigated by adaptation, and which cohort of players can have higher adaptability. We find that a higher pollution level during the game, relative to the adapted pollution level in players’ home cities, has a negative and significant impact on the players’ efforts and accuracy. The impact of non-adapted air pollution can be greatly offset by the home advantage, but not by personal attributes such as the higher ability.

Keywords

air pollution football player performance adaptation

Introduction

Air pollution is a major environmental problem that brings substantial social and economic costs. Besides the rich evidence of severe health consequences (Beatty & Shimshack, 2014; Deryugina et al., 2019; Graff Zivin & Neidell, 2018), a growing literature documents a negative pollution–productivity relationship (Chang et al., 2019; Graff Zivin & Neidell, 2012; He et al., 2019). In particular, recent studies focus on the impact of air pollution on professional sports players, which are widely perceived as the most skillful type of labor force. Related studies cover various types of sports, including marathons (Guo & Fu, 2019), tennis (Liu et al., 2019), chess (Klingen & Ommeren, 2020), football (Lichter et al., 2017), and baseball (Archsmith et al., 2018). All these studies bring consistent empirical evidence that professional players are also inevitably affected by air pollution, though they are believed to have higher-level skills relative to most other types of labor forces.

However, one question remains unsolved: can sports players with highly professional skills adapt to air pollution and thus suffer less productivity loss during the games? Or in other words, have the negative impacts on professional players observed in the existing literature been already mitigated by players’ great adaptability? Existing studies reveal that individuals can adapt to some unfavorable external conditions such as high temperature (Grothmann & Patt, 2005; Hoffmann & Sgrò, 2011). So far, there is no empirical evidence on the adaptability in terms of air pollution, though some laboratory evidence has proved that humans and other animals can mitigate their cardiac and respiratory effects after exposure to air pollutants for some time (Hackney et al., 1977; Hamade & Tankersley, 2009). Can professional players adapt to air pollution and thus be affected less in short-run performance? If so, are there any types of players with higher adaptability? The answers to these questions are essential to a thorough understanding of the pollution-productivity relationship, especially among high-skilled labor forces such as professional sports players.

In this study, we answer these questions based on empirical evidence from professional football players in China. The unique context of football games is advantageous in answering the above research questions for three reasons. First, the professional football league provides an ideal setting to investigate air-pollution adaptation. A typical professional football player participates in routine training in his home city but attends football games in many other cities for the away games. Thus, we can not only observe the actual air quality at a football game but also quantitatively measure the air-pollution level that a football player should have been adapted to in his home city. Then we can quantitatively measure the unfamiliarity degree of game-day air pollution for each player in each game, which enables us to identify the effects of adaptation. Second, since the professional football league takes the home and visiting system, we can test whether home advantages exist in terms of air-pollution adaptability, which can help shed some light on the underlying mechanism of such adaptability. Third, we have comprehensive measures of a player's performance during a game, which are objectively recorded by professionals. These measures allow us to quantitatively examine the impact of non-adapted air pollution on players’ performances. We also have rich information regarding the characteristics of the players so that we can investigate the heterogeneous effect of non-adapted air pollution by individual attributes, such as age, ability and nationality.

Our dataset comes from the Chinese Super League and covers all football games held from May 2015 to November 2017, for a total of 576 games on 210 game days in 16 cities. We link the player's performance in a game with the air-pollution level during the game, relative to the regular pollution level in his home city. We use such a pollution gap as the core explanatory variable of a player's performance indicators with the assumption that the “sensitivity” of a player's performance to air pollution may depend on the pollution level to which he has adapted. Moreover, the gap measure provides us with more identification variations because the gap is different for two teams during a game, even though the actual air-pollution level during the game is the same for both teams. We measure a player's performance using two indicators. One is the number of passes, which is related to the player's effort provision. The other is the success rate of passes, which measures the accuracy of the efforts. We adopt an ordinary least squares (OLS) model to study the impact of the pollution gap (i.e., the air-pollution level during the game relative to the average pollution level in a player's home city in the previous three months) on the performance indicators of the players. Our model controls for the home team dummy, individual fixed effects, city fixed effects, team fixed effects, season-by-round fixed effects, pair-of-match fixed effects, position-of-players fixed effects, and the dummy variable for games held at night.

We have three main findings. First, a higher air-pollution level during the game, relative to the adapted pollution level in players’ home city, has a negative and significant impact on the performance of football players on both the effort and accuracy measures. Specifically, a one-standard-deviation increase in the air-pollution gap reduces the number of passes per game by 2.0% of a standard deviation and lowers the pass success rate by 3.1% of a standard deviation. As a falsification test, we do not find such a relationship if we measure the pollution gap using the air-pollution level three days before or after the actual game. These results are also robust to different measurements of the air-pollution gap, and remain unchanged under different sets of fixed effects and alternative ways of standard error clustering.

Second, we provide some suggestive evidence on the potential underlying mechanism of air-pollution adaptation. On the one hand, the air-pollution adaptation is likely to be associated with players’ mental familiarity with the situation. We find that the negative effect of non-adapted pollution on players’ performance only exists for the visiting teams. To be specific, for visiting team players, the number of passes and the success rate would decrease by 7.1% and 7.8% of a standard deviation, relatively, if the air-pollution gap increases by one standard deviation. By sharp contrast, home team players are hardly affected by unadapted air pollution. The results suggest the home team is less sensitive to the deviation of air pollution from the adapted level than is the visiting team. The home advantage can be considered another form of adaptation, which cannot lead to a relatively lower pollution level but indicates higher familiarity with the overall environment (Pollard, 2002; Ven, 2011; Waters & Lovell, 2002). In other words, either being exposed to pollution for some time, or being mentally familiar with the environment as a home-team player, is helpful to mitigate the negative impact of actual air pollution on players’ performance. On the other hand, the negative impacts of unadapted air pollution are less likely to be driven by players’ physical conditions, such as the possibility that players get tired faster under unadapted pollution, since we observe no significant heterogeneity of the effect over players’ time on the pitch.

Third, we find the negative impact of unadapted air pollution widely exists among all types of players within the visiting teams, no matter their age, level of ability, or nationality. The only heterogeneity we observe is that younger players’ (i.e., younger than the median age) pass number is less sensitive to the pollution gap than the elder cohort, which is consistent with the declining physical conditions with age. However, the younger players’ accuracy performance is not better than the elder players under non-adapted air pollution. We also divide the players according to their abilities, with the scores from a professional football website as the indicator. The results suggest that higher-ability players are unable to mitigate the negative impacts of unadapted pollution, compared with the lower-ability ones. In addition, we also find that both foreign and domestic players suffer from the air-pollution gap, showing no evidence that domestic players have higher adaptation ability. These findings indicate that almost no personal characteristics, including higher professional skills, can help enhance the air-pollution adaptation and further weaken the negative pollution–productivity relationship.

Our paper directly speaks to the emerging literature studying the impact of air pollution on sports outcomes (Archsmith et al., 2018; Guo & Fu, 2019; Klingen & Ommeren, 2020; Lichter et al., 2017; Liu et al., 2019). A most related paper is Lichter et al. (2017), which studies the effects of air pollution on productivity in the context of professional German soccer players. Our study can be distinguished from the existing literature in two main aspects. First, we provide new insights into the adaptation to air pollution as a possible mitigating factor in the pollution-productivity relationship. Second, our paper documents an important fact that air pollution may amplify the home advantage in sports contests.

More broadly, our research also contributes to the literature on the impact of air pollution on productivity (Chang et al., 2019; Graff Zivin & Neidell, 2012; He et al., 2019). Our findings suggest strong heterogeneity in the impact of air pollution on productivity, which depends on the contemporaneous pollution shock relative to the average pollution level to which an individual has adapted. To the best of our knowledge, this study provides the first field empirical evidence on air-pollution adaptation. More importantly, the results show that the negative impacts of the pollution gap almost disappear among home team players, which implies that familiarity with the overall environment can help enhance people's adaptability to air pollution.

Our paper is also related to the literature on adaptation. The existing literature reveals that humans do have the capacity to adapt to severe external conditions, such as low air pressure (Hochachka, 1998; Moore, 2000), low temperature (Mäkinen, 2007), or the ongoing climate changes (Grothmann & Patt, 2005; Hoffmann & Sgrò, 2011). However, studies about air pollution adaptation are relatively rare. Our paper provides suggestive evidence about the psychological channels of adaptation, apart from potential physiological channels. Specifically, the home advantage in terms of adaptation indicates people who are more mentally familiar with the environment might be less affected by the pollution shock. Admittedly, our identification strategy only exploits the immediate effect of air pollution adaptation since we only observe the impacts of air pollution on players’ short-term performances during the game. We cannot test the long-run impact of air pollution adaptation, which may affect the physical status of the players through health channels and cause potential productivity loss in future games.

The paper proceeds as follows. Section “Data” introduces the data; section “Empirical Strategy” discusses the identification strategy; section “Empirical Results” presents the empirical results; and section “Conclusion” concludes.

Data

Football Games Data

In this study, we focus on the Chinese Football Association Super League (CSL), which was founded in 2004 and is the top football league in China. Each season, 16 teams/clubs participate in CSL, with each team facing each opponent twice (both at home and away). Thus, each CSL season includes 30 rounds and 240 games in total. The 16 teams come from different cities across a vast territory of mainland China. The essential system of CSL is similar to the systems of top football leagues in other major countries. A season typically starts in late February and ends in early November. The games are held regularly in the afternoon or at night on weekends. Prior to the beginning of the season, the league committee announces the schedule specifying the date, time, location, and teams for all games during the whole season, all of which are beyond the control of any team or player. The season schedule is fixed unless a force majeure happens; so far, no season schedule has changed due to air pollution. Moreover, CSL takes the home and visiting system, which means all the teams are trained in their home cities but have to take half of the games in other cities. The rules of games in CSL are in accordance with the Fédération Internationale de Football Association (FIFA). For example, each team can make up to three substitutions in a game. The technical statistics indicating the performance of players are also summarized according to FIFA.¹ A considerable portion of players in CSL are transferred from foreign football clubs, including several world-renowned football stars. One team can have at most five foreign players in each season.

The football games data are collected from CSL's official website.² The sample covers all games held between May 2015 and November 2017, including two full seasons (2016 and 2017) and the last 20 rounds of the 2015 season.³ Twenty teams were involved in these three seasons. During the sample period, 576 games were held on 210 game days in 16 cities.⁴ Appendix Figure A.1 shows the geographical distribution of home cities in our sample across China. For each game, we have detailed information on the home and visiting teams, the final scores of each team, the stadium, date and kick-off time, and the list of players on the pitch. In these 576 games, 688 players participated, and we have 15,840 player-game observations. For each player-game observation, we observe his team, minutes played, position, and various indicators of his performance during the game.

We adopt two player-game-level indicators to measure the performance of a player during a game. One indicator is the number of passes he makes in the game, which measures how hard he performs during the game. This indicator is not only an essential physical input for football players but also highly relevant to the success of the game, through retaining ball possession and creating scoring opportunities (Carling et al., 2005; Lichter et al., 2017). The other indicator is the success rate of passes, reflecting the effective output produced by the player in the game. This indicator reflects the effectiveness of his efforts during the game. The summary statistics of the key variables are reported in Table 1. On average, a player passes 28.45 times per game, with an average success rate of 73.74%.

Table 1.

Summary Statistics.

Variables	Definition	Observations	Mean	Std. Dev.	Min	Max
Gap	Gap between game-day PM_2.5 and 3-month adapted PM_2.5	13,748	−3.92	35.67	−102.34	322.40
Gap_Relative	Gap between game-day PM_2.5 and 3-month adapted PM_2.5, normalized by the adapted PM_2.5	13,748	−0.02	0.84	−0.98	8.38
Gap_12m	Gap between game-day PM_2.5 and 12-month adapted PM_2.5	13,993	−13.58	36.94	−89.74	320.16
Home	Whether the player is from the home team in this game	15,840	0.50	0.50	0	1
Pass	Number of all passes	15,840	28.45	17.95	0	125
SuccPass	Success rate of all passes (%)	15,840	73.74	17.39	0	100
personYoung	Whether the player is young (i.e., younger than 27 years old)	15,840	0.75	0.44	0	1
personGood	Whether the player is of high ability	15,840	0.24	0.42	0	1
personForeign	Whether the player is from foreign country	15,840	0.27	0.44	0	1
Halftime	Whether the player plays for less than the full time in this game	15,840	0.40	0.49	0	1

Note. PM_2.5 data are collected from the nearest national monitoring station for each stadium. Football-game data are collected from the official website of the Chinese Football Association Super League. The dataset covers 210 game days from May 2015 to November 2017, with 688 players playing for 18 teams in 576 games held in 16 cities.

For each player, we also collect information on his personal attributes, including age (personYoung) and nationality (personForeign), from CSL's official website. Besides, to quantitatively measure ability, we collect the players’ scores awarded by Tzuqiu, an authoritative website on football data in China.⁵ Specifically, for all 688 players in our dataset, Tzuqiu awards the scores for 176 players, ranking between 5.36 and 8.17. For these 176 players, we use the dummy variable of personGood to indicate players whose scores are beyond the median value.

Air-Pollution Data

We measure air pollution by the hourly data from 163 national monitoring stations in 16 cities between May 2014 and December 2017.⁶ The data report the hourly readings of six major pollutants, including PM_2.5, PM₁₀, SO₂, NO₂, CO, and O₃. Following recent empirical studies on the effect of air pollution in China (Guo et al., 2020; Han et al., 2014; Hao & Liu, 2016; Xie et al., 2016), we use PM_2.5 as the core indicator of air pollution in the empirical analysis. The data also report the hourly measures of weather, including temperature, wind speed, dew point, visible distance, cloud height, rain depth, and air pressure, which serve as the control variables in the following empirical analysis. Additionally, the geographic coordinates of all the monitoring stations are also available.

For each stadium in the football games data, we pick the nearest monitoring station as the data source for its air-pollution level. For a specific game, we calculate the average PM_2.5 of the kick-off hour and the previous hour, recorded as PM_2.5_Actual, as the pollution level that players experience during the game. We also calculate the average PM_2.5 in the home city of each football player during a time period before the game day as the pollution level to which the player is adapted (PM_2.5_Adapted). Specifically, we measure the daily pollution level by calculating the hourly average PM_2.5 from 8 a.m. to 8 p.m., and then calculate the average daily pollution level for the previous 3 months before each game. Thus, we can define the gap between adapted pollution level and actual pollution level during the game as the following equation:

G a p = P M_{2.5}_A c t u a l - P M_{2.5}_A d a p t e d .

(1)

We focus on the gap measure, instead of the actual air pollution during the game, for two reasons. First, the gap measure allows us to test whether the negative impact of air pollution can be mitigated by adaptation, which is the major research question of this study. Second, the gap measure provides us richer identification variation because two teams of the same game have different levels of the gap measure even with the same actual air-pollution level. According to the definition, the gap can be either positive or negative for a given player and game. Specifically, a positive gap implies the pollution level that the player experiences during the game is more severe than the average pollution level that he has adapted to in his home city, whereas a negative gap means the game-day pollution is less severe than his adapted level.

Figure 1 plots the distribution of the gap in a histogram. The summary statistics of the gap are also reported in Table 1. On average, the air pollution that the players experience during the game is slightly lower than their adapted levels. In our sample, the average gap is −3.92 μg/m³.

Figure 1.

Distribution of the Gap between actual PM_2.5 and adapted PM_2.5

Empirical Strategy

In this study, we investigate the impact of non-adapted air pollution on the performance measures of football players. The main hypothesis to be tested is that a higher pollution level during the game, relative to the adapted pollution level in the home city, has a negative impact on the performance of football players. The main regression equation is as follows:

O u t c o m e_{i, j, k, t} = β_{1} \times G a p_{i, j, k, t} + β_{2} \times H o m e_{i, j, k, t} + W e a t h e r_{j, k, t} + X_{i} + α_{j} + δ_{t} + C i t y_{j, k, t} + P a i r_{j, k} + P o s i t i o n_{i, t} + β_{3} \times N i g h t_{j, k, t} + ϵ_{i, j, k, t},

(2)

where Outcome_i,j,k,t refers to the performance indicator of player i in team j in the game with team k held on day t. As described in section “Football Games Data,” we measure a player's performance by the number of passes and the pass success rate. The core explanatory variable is the gap between PM_2.5 during the game and the average PM_2.5 in the player's home city before the game day, which enters the regression equation as a continuous variable. We would expect

β_{1}

to be negative if our hypothesis is correct.

We also control for the Home_i,j,k,t dummy, which equals one for the home team. We use this dummy to capture the average level of the home advantage in football games. The vector of $W e a t h e r_{j, k, t}$ controls for the weather on day t for team j (and its opponent team k), including temperature, wind speed, dew point, visible distance, cloud height, rain depth, and air pressure. In the main specification, we also control for player fixed effects ( $X_{i}$ ), team fixed effects ( $α_{j}$ ), season-by-round fixed effects ( $δ_{t}$ ), city fixed effects ( $C i t y_{j, k, t}$ ), pair-of-teams fixed effects ( $P a i r_{j, k}$ ,with the same value as $P a i r_{k, j}$ ), position-of-players fixed effects ( $P o s i t i o n_{i, t}$ ), and the dummy for games held at night instead of in the afternoon ( $N i g h t_{j, k, t}$ ). Therefore, the identification variation comes from the same player on the same team playing in different games with the same rival team, with different air-pollution gaps. Note we are less concerned about the possible endogeneity of the gap measure because the time and venue of a game are predetermined. At least during our sample period, CSL never changed the time or venue of a game in response to the anticipated air-pollution level. We adopt the robust standard error clustering at the game level.

Empirical Results

Baseline Results

We start our empirical analysis from the baseline results. Table 2 shows the impact of the non-adapted air pollution on the performance of football players following equation (2). The indicators include the number of passes (columns 1 and 3) and the success rate of passes (columns 2 and 4). In all the regressions, we control for the weather conditions and all the fixed effects as described in the previous section.

Table 2.

The Impact of PM2.5 Gap on Players: Baseline.

	(1)	(2)	(3)	(4)
Variables	Pass	SuccPass	Pass	SuccPass
Gap	−0.010**	−0.015**	−0.010**	−0.015**
	(0.004)	(0.006)	(0.004)	(0.006)
Home			1.929***	1.225***
			(0.306)	(0.241)
Observations	12,535	12,535	12,535	12,535
R ²	0.727	0.281	0.730	0.282

Note. This table explores the effect of the air-pollution gap on players’ performance, following equation (2). Columns 1 and 3 adopt the number of all passes as the outcome variable, while columns 2 and 4 adopt the success rate of passes as the outcome variable. Gap measures the difference between the PM_2.5 during the game and the average PM_2.5 in the player's home city before the game day. In all specifications, we control for weather conditions, city fixed effects, team fixed effects, player fixed effects, season-by-round fixed effects, pair-of-teams fixed effects, the position-of-players fixed effects, and the dummy for games held at night. In columns 3 and 4, we also control for the Home dummy for the home teams. Standard errors are clustered at the game level. * indicates significance at the .1 level; ** indicates significance at the .05 level; *** indicates significance at the .01 level.

In column 1, we observe that the coefficient of Gap is negative and significant at the 5% level. According to the coefficient, a one-standard-deviation increase in the PM_2.5 gap leads to 2.0% of one-standard-deviation decrease in the number of passes per game. Similarly, in column 2, a one-standard-deviation increase in the PM_2.5 gap would reduce the success rate of passes by 3.1% of a standard deviation, which is also significant at the 5% level.⁷

Considering the existence of home advantage in football games, we further control for the home team dummy (Home) in columns 3 and 4 and find its coefficient significantly positive at the 1% significance level in both specifications, indicating home-team players pass more and achieve higher accuracy rate than players from the visiting team. Specifically, a home-team player passes 1.9 times more than the visiting-team player, and the success rate of passes is 1.2 percentage points higher than a visiting-team player, all else being equal. Meanwhile, the coefficients for Gap remain highly consistent after we control for Home. As a whole, the above findings suggest a higher non-adapted level of air pollution has a negative impact on both effort and accuracy measures of football players, and the magnitude of the effect is larger for accuracy.

Robustness Checks

We test the robustness of the baseline results from several aspects. First, considering that the air-pollution gap variable constructed in this study is a relatively novel indicator, we also try three different measurements of unadapted air pollution to replicate Table 2. The results are presented in three panels in Table 3. In Panel A, we introduce the relative gap (Gap_Relative) by normalizing Gap with PM_2.5_Adapted. Thus, we obtain the relative gap measured by percentage rather than level. In Panel B, we extend the adaptation period from three months to 12 months. In both panels, the results are consistent with those in Table 2. Thus, the baseline results are not sensitive to the way we construct the air-pollution gap indicator.

Table 3.

The Impact of PM2.5 Gap on Players: Robustness Check.

	(1)	(2)	(3)	(4)
Variables	Pass	SuccPass	Pass	SuccPass
Panel A: Relative Gap
Gap_Relative	−0.859***	−0.642***	−0.669***	−0.523**
	(0.204)	(0.242)	(0.187)	(0.240)
Home			1.857***	1.170***
			(0.306)	(0.252)
Observations	12,535	12,535	12,535	12,535
R ²	0.728	0.281	0.730	0.282
Panel B: 12-Month Gap
Gap_12m	−0.008**	−0.011**	−0.008**	−0.011**
	(0.003)	(0.005)	(0.003)	(0.005)
Home			1.936***	1.278***
			(0.303)	(0.248)
Observations	12,738	12,738	12,738	12,738
R ²	0.726	0.280	0.729	0.281
Panel C: Decomposed Gap
PM_2.5_Actual	−0.008**	−0.013**	−0.008**	−0.013**
	(0.003)	(0.005)	(0.003)	(0.005)
PM_2.5_Adapted	0.050*	0.048**	0.047*	0.045*
	(0.026)	(0.023)	(0.025)	(0.023)
Home			1.918***	1.216***
			(0.309)	(0.244)
Observations	12,535	12,535	12,535	12,535
R ²	0.727	0.281	0.730	0.282

Note. This table presents the results of robustness checks, following Table 2. In Panel A, we adopt the relative form of the air-pollution gap. In Panel B, we adopt the air-pollution gap with a 12-month adaptation period. In Panel C, we decompose the gap into PM_2.5_Actual and PM_2.5_Adapted. PM_2.5_Actual measures the average PM_2.5 of the kick-off hour and the previous hour, indicating the pollution level that players experience during the game. PM_2.5_Adapted measures the average PM_2.5 in the home city of each football player during a time period before the game day, indicating the pollution level that players have adapted before the game starts. In all specifications, we control for weather conditions, city fixed effects, team fixed effects, player fixed effects, season-by-round fixed effects, pair-of-teams fixed effects, the position-of-players fixed effects, and the dummy for games held at night. We also control for the Home dummy for the home teams in columns 3 and 4 in all the panels. Standard errors are clustered at the game level. * indicates significance at the .1 level; ** indicates significance at the .05 level; *** indicates significance at the .01 level.

In Panel C of Table 3, we try decomposing Gap into PM_2.5_Actual and PM_2.5_Adapted based on equation (2). Both the actual pollution level and the adapted pollution level have significant impacts on the efforts and accuracy of football players, while the direction of impacts is opposite. As expected, the football players give better performance under lower actual pollution level and/or higher adapted pollution level, which further proves the reliability of our gap measure.

Second, we implement the placebo test by replacing Gap with the air-pollution gaps of the kick-off hour on the 3 days before the game day (− 3 d, − 2 d, − 1 d) and on the 3 days after the game day ( + 1 d, + 2 d, + 3 d), respectively, with other variables remaining unchanged. As shown in Figure 2, the coefficients are insignificant for all the falsification gap variables.

Figure 2.

Visualization of the falsification analysis

Third, we test the robustness of the results with various combinations of control variables. Appendix Tables A.1 and A.2 report the results on the number of passes and the pass success rate, respectively, using different sets of control variables. Columns 1–6 exclude weather controls, city fixed effects, team fixed effects, player fixed effects, season-by-round fixed effects, and pair-of-teams fixed effects, one at a time from the most complete specification, and reach similar results. In columns 7 and 8, we change the way the standard errors are calculated by adopting standard errors clustered at the player level and two-way clustered at the city-date level, respectively. We also exclude the subsample of goalkeepers in column 9. In all the tables, the results are very similar to those shown in Table 2.

Finally, we exclude the alternative explanation that players’ ability may be correlated with their adapted pollution level. If better clubs tend to locate in more polluted cities, players from these teams would have both higher adapted pollution levels and higher abilities, and the latter might help them mitigate the negative impact of air pollution. In this case, the baseline results we observe may result from the omitted variable of players’ ability. However, at least in our sample, the ability of players from more polluted cities is not significantly higher than those from cleaner cities. Specifically, we divide the sample into two groups based on the home city's average PM_2.5 level during the previous three months before the game (i.e., the adapted level). For players from the more polluted (i.e., with adapted air pollution over the median value) and cleaner cities, the average values of personGood are 0.223 and 0.217, respectively, whose difference is not statistically significant.

Discussions on the Potential Mechanism

The above results suggest that football players’ performance is negatively affected by non-adapted air pollution; in other words, the negative effect of air pollution can be mitigated if the player has adapted to pollution. We are then interested in the underlying mechanism of such air-pollution adaptation. Although a stricter medical investigation based on experimental tests is well beyond the scope of this paper, we provide suggestive empirical evidence from two aspects.

First, from the psychological aspect, the negative effect of air pollution may be mitigated by adaption because players are mentally familiar with the contest condition with air pollution. To test this possible explanation, we focus on the relationship between the home advantage and players’ adaptability to air pollution. Home advantages significantly exist in most professional sports games. In particular, the existing literature suggests that home advantages are mainly caused by players’ mental familiarity with the overall conditions, while traveling fatigue, crowd support, and referee bias play a relatively smaller role (Pollard, 2002; Ven, 2011; Waters & Lovell, 2002). Thus, if the air-pollution adaptation results from the psychological mechanism such as mental familiarity, we expect that home team players are less affected by unadapted pollution since they should be mentally familiar with the overall environment, even with the same air-pollution gap.

We adopt two types of specifications to explore such a relationship. We first introduce the interaction term between Gap and Home. As shown in the first two columns in Table 4, the coefficients of the interaction terms are significantly positive, while the coefficients of Gap remain significantly negative. Specifically, the home advantages can fully offset the negative impacts of Gap on the number of passes, as shown in column 1. For the success rate of passes, as shown in column 2, such an home advantage is relatively weaker (with only 10% significant level and smaller magnitude), but still positive. These results suggest visiting-team players are more sensitive to the deviation of air pollution from their adapted level, even though both types of teams are likely to experience an unadapted air-pollution condition.⁸

Table 4.

The Impact of PM_2.5 Gap on Players: Home Advantage.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	Pass	SuccPass	Pass	SuccPass	Pass	SuccPass
			Home teams		Visiting teams
Gap	−0.023***	−0.022***	0.010	0.005	−0.032***	−0.035**
	(0.006)	(0.006)	(0.010)	(0.007)	(0.009)	(0.007)
Gap * Home	0.026***	0.015*
	(0.009)	(0.009)
Home	2.045***	1.292***
	(0.292)	(0.269)
Observations	12,535	12,535	6,537	6,537	5,998	5,998
R ²	0.730	0.282	0.768	0.352	0.765	0.351

Note. This table explores the potential psychological mechanism by testing the effect of the home advantage on air-pollution adaptability. In the first two columns, we adopt the interaction term of Gap and Home. In the last four columns, we run regressions using subsamples of home teams and visiting teams, respectively. In all specifications, we control for weather conditions, city fixed effects, team fixed effects, player fixed effects, season-by-round fixed effects, pair-of-teams fixed effects, the position-of-players fixed effects, and the dummy for games held at night. Standard errors are clustered at the game level. * indicates significance at the 0.1 level; ** indicates significance at the 0.05 level; *** indicates significance at the 0.01 level.

We also adopt another specification with looser restrictions by dividing the entire sample into two subsamples of “home teams” and “visiting teams.” As shown in the last four columns in Table 2, we find the previous effect of the air-pollution gap only exists within the visiting team players for both outcome variables. Specifically, a one-standard-deviation increase in unadapted air pollution leads to 7.1% of a one-standard-deviation decrease of the number of passes, and a 7.8% of a one-standard-deviation decrease of the success rate, both significant at the 1% level. Such effects are much higher than the average effects shown in Table 2. In sharp contrast, the effect of unadapted air pollution disappears for the home-team group. Such results provide further evidence that home teams would take advantage of the negative effect of air pollution on the visiting team to amplify the home advantage, which is also consistent with the explanation that mental familiarity plays an important role in the air-pollution adaptation.

The second possible mechanism comes from the physiological aspect. For instance, players under higher unadapted pollution levels may become tired faster and stand still more, which will lead to fewer passes and lower pass accuracy. To address this possible explanation, we focus on players’ variation on minutes on the pitch in a game. According to the game rule, each team can make at most three player substitutions in a game, which implies that, for each team-game, there are at most six players who do not play for the full time. The average minutes played of these non-fulltime players are 45 min in our sample, which is only about half of the full game time. Thus, it is reasonable to argue that these non-fulltime players are less likely to be tired than their full-time teammates. If the air-pollution adaptation results from the physiological channel, we should observe that these non-fulltime players are less likely to be affected by the air-pollution gap.

We define the dummy for these non-fulltime players as Halftime. In Table 5, we introduce the interaction term between Gap and Halftime and replicate the regressions following equation (2) on subsamples of home teams (columns 1 and 3) and visiting teams (columns 2 and 4), respectively. All the coefficients for the interaction term are insignificant, while Gap is still significantly negative for visiting teams. These results indicate that the effects of the pollution gap show little heterogeneity between players under different levels of fatigue. In particular, all players of visiting teams are affected by unadapted air pollution, no matter whether they are non-fulltime players or, equivalently, whether they are tired. Therefore, these results provide suggestive evidence that the air-pollution adaptation is less likely to be driven by the physiological channel, although we have to leave more conclusive arguments to the medical experts.

Table 5.

The Impact of PM_2.5 Gap on Players: Time on the Pitch.

	(1)	(2)	(3)	(4)
Variables	Pass	SuccPass	Pass	SuccPass
	Home teams		Visting teams
Gap	0.013	0.007	−0.029***	−0.040***
	(0.010)	(0.008)	(0.010)	(0.009)
Gap * Halftime	−0.003	−0.002	0.005	0.019
	(0.003)	(0.015)	(0.010)	(0.015)
Halftime	−15.425***	−2.009***	−15.485***	−3.306***
	(0.710)	(0.448)	(0.567)	(0.608)
Observations	6,537	6,537	5,998	5,998
R ²	0.665	0.341	0.671	0.331

Note. This table explores the potential physiological mechanism by introducing the interaction term between Gap and the dummy for whether the player plays for less than the full time in the game, using subsamples of the home teams and visiting teams, respectively. In all specifications, we control for weather conditions, city fixed effects, team fixed effects, player fixed effects, season-by-round fixed effects, pair-of-teams fixed effects, position-of-players fixed effects, and the dummy for games held at night. Standard errors are clustered at the game level. * indicates significance at the .1 level; ** indicates significance at the .05 level; *** indicates significance at the .01 level.

Heterogeneity Analysis of Players’ Attributes

One of the most important findings from the above analysis is the strong home advantage among football players in terms of adaptation to air pollution. These results imply that some types of players can further mitigate the negative impacts of the pollution gap by their higher adaptability to air pollution. A naturally related question is whether some other types of players are also more likely to adapt to air pollution. For this purpose, we further test the potential heterogeneity between different types of players from three aspects of personal attributes, including age, ability, and nationality. The results are listed in three panels in Appendix Table A.3, respectively.

First, we conduct the heterogeneity analysis of players’ age. For a physical game such as football, age is highly related to players’ physical conditions. Thus, one may expect older players to be more sensitive to non-adapted air pollution since they are more likely to suffer from physical decline. On the other hand, older players typically have more experience, which may help them deal with unfavorable external conditions. For example, Guo and Fu (2019) find that older marathon runners are less affected by air pollution, and they believe their rich experience should play a key role. In Panel A in Appendix Table A.3, we introduce the interaction term between Gap and personYoung, which equals one if the player's age exceeds 27 (i.e., the median value of the sample) that season. The results suggest limited advantages enjoyed by younger players. For visiting teams, younger players are less affected by the pollution gap if we adopt the number of passes as the outcome variable. According to the results shown in column 3, younger players can significantly mitigate the negative effect of unadapted air pollution by about 22% (0.008÷0.0358), compared with their elder counterparts. However, such an advantage disappears for the success rate of passes. We do not observe significant variations with player’ age for home team players, either, for either of the two outcome variables.

Second, we investigate the heterogeneity associated with players’ professional abilities. One may expect that players with higher professional abilities are less affected by unadapted air pollution. For instance, Guo and Fu (2019) reveal that top marathon runners are less affected by air pollution. For this purpose, we divide these 176 players into high- and low-ability groups based on the median of players’ scores awarded by Tzuqiu (6.57). The other 512 players without Tzuqiu scores are excluded from both groups.⁹ The results are shown in Panel B in Appendix Table A.3. The interaction term between Gap and personGood, the latter variable indicating high-ability players, is insignificant in all the columns. These results suggest that both high- and low-ability players are sensitive to unadapted air pollution, and the magnitude of the impact shows no significant differences.

Finally, we exploit the heterogeneity analysis over players’ nationality. In our sample, there are 192 foreign players, accounting for 27.91% of all the players. While one might expect the effect of unadapted air pollution to be larger for foreign players, because they are less familiar with China than their domestic counterparts, the results in Panel C show no significant differences.

Overall, the negative impacts of unadapted air pollution on players’ performance show little heterogeneity over different players, though younger players could take very limited advantage of their better physical conditions to resist the losses of efforts. These results reveal an important fact that the air-pollution gap would impose broad and inevitable impacts on almost all players from visiting teams. In particular, higher professional skills or richer experience cannot further help mitigate the adverse effects of unadapted air pollution.

Conclusion

In this paper, we focus on the effect of adaptation on the pollution–productivity relationship in the context of professional football games in China. We introduce a measurement of pollution gap indicating the difference between actual PM_2.5 on the game day and adapted PM_2.5 as the core explanatory variable, and investigate its impact on the performance of football players from both the effort perspective (indicated by the pass number) and accuracy perspective (indicated by the success rate of passes). The empirical results suggest individual players do have the ability to adapt, and greater non-adapted air pollution has a negative impact on both the pass number and success rate of passes of football players. We also find that home team players are hardly affected by unadapted air pollution, which indicates that home advantage still exists in terms of adaptation abilities to air pollution. However, the negative effects of unadapted air pollution on visiting teams’ players are widespread and cannot be further mitigated by higher ability or richer experience.

Our findings call for the improvement of air quality from the context of sports. We find that air pollution leads to bad performance for almost all types of football players, including both less passing and lower accuracy, which decreases the quality of the game for the audience. Our findings also provide insights to sports players, coaches, and team managers on how to weaken the negative effect of air pollution in the short run, especially for the visiting teams. According to our analysis, to mitigate the potential negative impact of air pollution, a possible solution is to reduce the difficulty of adaptation; that is, to make players feel like “home-team players.” In particular, ensuring that players are familiar with the overall environment by providing mental support is valuable, given that the adaptation effect is likely to be driven by players’ mental status. Additionally, the existence of adaptability also proves the rationality of the match system of professional football leagues, which requires each team to face its opponent both at home and away. Using similar rules to enhance the fairness of competition in other settings is worth consideration.

It is noteworthy that, although air pollution adaptation brings an immediate effect by weakening the short-term pollution–productivity relationship, such an arguably positive consequence does not necessarily apply to the long run. While the existing medical literature does provide some evidence that animals can mitigate their cardiac and respiratory effects after exposure to air pollutants for some time, the opposite effect may also exist. As an extreme case, players who adapt to air pollution better in the short run might be more physically vulnerable under severe air pollution than those without such adaptation. For instance, if a player adapts better to severe air pollution due to his mental familiarity with the overall environment, he would make more efforts (e.g., pass more) during the game compared to those who exert fewer efforts under the pollution, which may make him suffer from a heavier respiratory burden. Such a situation may result in worse performance or even worse health status in the long run. In this case, the short-term adaptation may be achieved at a long-term cost. More analysis on the long-run air pollution adaptation is called for.

Footnotes

Acknowledgments

The authors greatly appreciate the insightful comments from the editor, two anonymous referees and Tie Zan. Wu and Zhang thank the National Natural Science Foundation of China (Project Nos: 91546113, 71874093) for financial support.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Natural Science Foundation of China (grant nos. 71874093, 91546113).

ORCID iD

Jing Wu

Notes

Author Biographies

Yu Qin, Associate professor, Department of Real Estate, National University of Singapore, Singapore.

Jing Wu, Associate professor, Hang Lung Center for Real Estate, Department of Construction Management, Tsinghua University, Beijing, China.

Rongjie Zhang, PhD student, Hang Lung Center for Real Estate, Department of Construction Management, Heshanheng Building, Tsinghua University, Beijing 100084, China. Email: zhangrj19@mails.tsinghua.edu.cn.

Appendix

Table A.3.

The Impact of PM2.5 Gap on Players: Heterogeneity Analysis.

	(1)	(2)	(3)	(4)
Variables	Pass	SuccPass	Pass	SuccPass
	Home teams		Visiting teams
Pane A: Heterogeneity over age
Gap	0.012	0.001	−0.036***	−0.039***
	(0.009)	(0.007)	(0.010)	(0.009)
Gap * person young	−0.004	0.008	0.008**	0.006
	(0.007)	(0.011)	(0.004)	(0.011)
Person young	−0.169	0.563	0.869*	−1.784**
	(0.841)	(0.654)	(0.490)	(0.736)
Observations	6,537	6,537	5,998	5,998
R ²	0.768	0.352	0.765	0.351
Pane B: Heterogeneity over ability
Gap	0.007	0.003	−0.031***	−0.035***
	(0.009)	(0.007)	(0.011)	(0.008)
Gap * person good	0.011	0.011	−0.003	0.001
	(0.012)	(0.011)	(0.008)	(0.016)
Observations	6,537	6,537	5,998	5,998
R ²	0.768	0.352	0.765	0.351
Pane C: Heterogeneity over nationality
Gap	0.009	0.006	−0.032***	−0.036***
	(0.010)	(0.008)	(0.009)	(0.008)
Gap * person foreign	0.004	−0.002	0.001	0.002
	(0.007)	(0.011)	(0.008)	(0.012)
Observations	6,537	6,537	5,998	5,998
R ²	0.768	0.352	0.765	0.351

Note. This table presents the results of heterogeneous analysis, using subsamples of home teams and visiting teams, respectively. In Panel A, we interact Gap with the dummy for whether the player is younger than the median age (i.e., 27 years old), personYoung. In Panel B, we interact Gap with the dummy for whether the player is with higher ability, personGood. In Panel C, we interact Gap with the dummy for whether the player is foreign, personForeign. In all specifications, we control for weather conditions, city fixed effects, team fixed effects, player fixed effects, season-by-round fixed effects, pair-of-teams fixed effects, position-of-players fixed effects, and the dummy for games held at night. Standard errors are clustered at the game level. * indicates significance at the .1 level; ** indicates significance at the .05 level; *** indicates significance at the .01 level.

References

Archsmith

Heyes

Saberian

(2018). Air quality and error quantity: Pollution and performance in a high-skilled, quality-focused occupation. Journal of the Association of Environmental and Resource Economists, 5(4), 827–863. https://doi.org/10.1086/698728

Beatty

Shimshack

(2014). Air pollution and children's Respiratory health: A cohort analysis. Journal of Environmental Economics and Management, 67(1), 39–57. https://doi.org/10.1016/j.jeem.2013.10.002

Carling

Williams

Reilly

(2005). Handbook of soccer match analysis. Routledge.

Chang

Graff Zivin

Gross

Neidell

(2019). The effect of pollution on worker productivity: Evidence from call center workers in China. American Economic Journal: Applied Economics, 11(1), 151–172. https://doi.org/10.1257/app.20160436

Deryugina

Heutel

Miller

Molitor

Reif

(2019). The mortality and medical costs of air pollution: Evidence from changes in wind direction. American Economic Review, 109(12), 4178–4219. https://doi.org/10.1257/aer.20180279

Graff Zivin

Neidell

(2012). The impact of pollution on worker productivity. American Economic Review, 102(7), 3652–3673. https://doi.org/10.1257/aer.102.7.3652

Graff Zivin

Neidell

(2018). Air pollution's Hidden impacts. Science (New York, NY), 359(6371), 39–40. https://doi.org/10.1126/science.aap7711

Grothmann

Patt

(2005). Adaptive capacity and human cognition: The process of individual adaptation to climate change. Global Environmental Change, 15(3), 199–213. https://doi.org/10.1016/j.gloenvcha.2005.01.002

Guo

Zhang

Fan

(2020). Does air pollution stimulate electric vehicle sales? Empirical evidence from twenty major cities in China. Journal of Cleaner Production, 249, 119372. https://doi.org/10.1016/j.jclepro.2019.119372

10.

Guo

(2019). Running with a mask? The effect of air pollution on marathon runners’ performance. Journal of Sports Economics, 20(7), 903–928. https://doi.org/10.1177/1527002518822701

11.

Hackney

Linn

Mohler

Collier

(1977). Adaptation to short-term respiratory effects of ozone in men exposed repeatedly. Journal of Applied Physiology, 43(1), 82–85. https://doi.org/10.1152/jappl.1977.43.1.82

12.

Hamade

Tankersley

(2009). Interstrain variation in cardiac and respiratory adaptation to repeated ozone and particulate matter exposures. American Journal of Physiology-Regulatory, Integrative and Comparative Physiology, 296(4), 1202–1215. https://doi.org/10.1152/ajpregu.90808.2008

13.

Han

Zhou

(2014). Impact of urbanization level on urban air quality: A case of fine particles (PM2.5) in Chinese cities. Environmental Pollution, 194, 163–170. https://doi.org/10.1016/j.envpol.2014.07.022

14.

Hao

Liu

(2016). The influential factors of urban PM2.5 concentrations in China: A spatial econometric analysis. Journal of Cleaner Production, 112, 1443–1453. https://doi.org/10.1016/j.jclepro.2015.05.005

15.

Liu

Salvo

(2019). Severe air pollution and labor productivity: Evidence from industrial towns in China. American Economic Journal: Applied Economics, 11(1), 173–201. https://doi.org/10.1257/app.20170286

16.

Hochachka

(1998). Mechanism and evolution of hypoxia-tolerance in humans. Journal of Experimental Biology, 201(8), 1243–1254. https://doi.org/10.1242/jeb.201.8.1243

17.

Hoffmann

Sgrò

(2011). Climate change and evolutionary adaptation. Nature, 470(7335), 479–485. https://doi.org/10.1038/nature09670

18.

Klingen

Ommeren

(2020). Risk attitude and air pollution: Evidence from chess. Tinbergen Institute Discussion Papers 2020-027/VIII, Tinbergen Institute. https://ssrn.com/abstract=3609237

19.

Lichter

Pestel

Sommer

(2017). Productivity effects of air pollution: Evidence from professional soccer. Labour Economics, 48, 54–66. https://doi.org/10.1016/j.labeco.2017.06.002

20.

Liu

Salvo

(2019). Willingness to fight on: Environmental valuation in dynamic competitions. Working paper.

21.

Mäkinen

(2007). Human cold exposure, adaptation, and performance in high latitude environments. American Journal of Human Biology, 19(2), 155–164. https://doi.org/10.1002/ajhb.20627

22.

Moore

(2000). Comparative human ventilatory adaptation to high altitude. Respiration Physiology, 121(2-3), 257–276. https://doi.org/10.1016/S0034-5687(00)00133-X

23.

Pollard

(2002). Evidence of a reduced home advantage when a team moves to a new stadium. Journal of Sports Sciences, 20(12), 969–973. https://doi.org/10.1080/026404102321011724

24.

Ven

(2011). Supporters are not necessary for the home advantage: Evidence from same-stadium derbies and games without an audience. Journal of Applied Social Psychology, 41(12), 2785–2792. https://doi.org/10.1111/j.1559-1816.2011.00865.x

25.

Waters

Lovell

(2002). An examination of the homefield advantage in a professional English soccer team from a psychological standpoint. Football Studies, 5(1), 46–59. https://www-mendeley-com-s.web.bisu.edu.cn/catalogue/b8ba0122-bd13-3725-a6a3-1cdbea9017d7/

26.

Xie

Dai

Dong

Hanaoka

Masui

(2016). Economic impacts from PM2.5 pollution-related health effects in China: A provincial-level analysis. Environmental Science & Technology, 50(9), 4836–4843. https://doi.org/10.1021/acs.est.5b05576