A large body of evidence causally links pollution with poor health and mortality (Chay and Greenstone, 2003; Currie and Neidell, 2005; Currie and Walker, 2011; Moretti and Neidell, 2011; Schlenker and Walker, 2015). The negative impacts of pollution on individual performance and economic outcomes have also been documented, especially in physically demanding occupations (Aragón et al., 2017; Ebenstein et al., 2016; Hanna and Oliva, 2015; Kim et al., 2017; Ostro, 1983). However, these studies implicitly assume that “no hidden information” exists; that is, the actual pollution levels match those perceived by people, which is not always true. Several recent studies have examined behaviors in response to environmental information, such as increasing environmental awareness (Tu et al., 2020) or through propaganda to change health-related consumption and the willingness to pay for safe water (Jacobsen, 2011; Jalan and Somanathan, 2008; Jessoe and Rapson, 2014; Madajewicz et al., 2007; Moretti and Neidell, 2011; Tu et al., 2020; Zivin and Neidell, 2009). Few studies have focused on the impact of pollution information salience, which represents a complementary nudge policy that enhances the information effect (Zizzo et al., 2016).

This study provides quantitative evidence on how people might respond to the disclosure of pollution information and whether people's responses would further affect their health status. In recent decades, China's energy-intensive and coal-fueled economic growth has led to degraded air quality (Zhang et al., 2012). Owing to rising levels of concern and attention from the public toward the impact of air pollution, the Chinese government began an initiative in 2013 to disclose major pollution information to the public. In this study, we exploit this natural experiment—China's pollution information disclosure (PID) program since 2013—to explore the causal effects of PID adoption on people's behavior (e.g., outdoor activities) and health status.

This study draws on a novel dataset from several official sources. First, we collected a city-level dataset on the timing of PID, which revealed that 76, 38, 74, and 177 cities joined the PID program in January 2013, October 2013, January 2014, and January 2015, respectively. We combined this dataset with the China Health and Retirement Longitudinal Survey (CHARLS), which contains in-depth individual-level information including demographic and socioeconomic characteristics, health measures, and people's choice of outdoor activities1. Additionally, we utilized the city-level statistics yearbook to control for the main city-level socioeconomic indicators. PM2.5 satellite data were used to measure air pollution. Combining these datasets allowed us to examine individual responses (e.g., indoor physical activities) to PID and the subsequent impact on people's health status. Another individual-level micro dataset used was the Chinese General Social Survey (CGSS), which allowed us to examine whether people's awareness of environmental issues plays the role of an important channel through which PID affects individual responses.

To examine the causal relationship between PID and individual choice of outdoor physical activities, we used a difference-in-differences (DID) strategy. The biggest challenge was to select appropriate control groups for comparison with the treatment group, ensuring that cities that adopted PID earlier were valid counterfactuals for those that adopted PID later. However, as the timing of PID adoption might not be randomly selected, we expected a bias in the estimation results. To address this issue, we (i) controlled for some key city-level determinants, (ii) conducted a placebo test by randomly assigning PID adoption, and (iii) used an event study to estimate year-wise changes in outcomes before and after PID.

This study yielded three main findings. First, the adoption of PID significantly reduced the probability of outdoor exercise for the middle-aged and elderly living in more polluted cities; specifically, after PID, one standard deviation increase in PM2.5 reduced the probability of attending outdoor activities by 2.234%.

Second, we provided evidence for one of the mechanisms by incorporating a heterogeneity analysis. We found that the impact of PID was more pronounced for relatively higher educated people. This heterogeneity is linked to the mechanisms that demonstrate that the improvement of individual environmental pollution after PID plays a pivotal role. The evidence suggests that those with higher education are more likely to be concerned about pollution information and pay more attention to PID.

Third, using a series of health status measures, we did not find evidence that PID adoption had a positive impact on individual health states. This finding suggests that reducing the possibility of exposure to outdoor pollution does not directly improve the health status of the middle-aged and elderly. However, we interpreted these results with caution as the health benefits may appear over a longer period.

This analysis contributes to three streams of literature. The first investigates the direct impact of air pollution on individual health status. Given the consensus on the negative impact of both outdoor and indoor air pollution (Peabody et al., 2005; Qiu et al., 2019), much of this literature has examined the types of health damage that is likely to occur, such as mortality (Evans and Smith, 2005), cardiovascular disease (Chi et al., 2016), mental issues (Gu et al., 2020), and cognitive ability impairment (Qiu et al., 2019). Despite the short-term impact identified in most studies, some studies indicate that a negative impact can persist in the long term (Kim et al., 2017). Another important line of research examines the role of socioeconomic status (Neidell, 2004; Hajat et al., 2015), suggesting that pollution is a potential mechanism by which socioeconomic status affects health. This literature highlights the importance of reviewing heterogeneity in socioeconomic status when examining the health consequences of air pollution.

Second, this study complements the literature on the relationship between pollution, media, and welfare. Earlier research has found evidence on the impact of pollution information releases on economic performance. For instance, Hamilton (1995) found that stockholders in firms reporting Toxics Release Inventory figures experienced negative returns upon the first release of the information, which can be treated as a change in information salience. Other studies have also suggested that media warnings about environmental pollution would change individual responses to pollution (Winters et al., 2003). Conversely, air pollution is likely to change people's judgments on social media (Zheng et al., 2019). Anderson et al. (2022) found that Korean air-quality alert systems (AQAS) have positive welfare impacts. Additionally, much of the existing evidence comes from the effects of pollution on economic outcomes such as productivity, migration, and relevant consumption. Our results provide an empirical estimation of avoidance behavior in response to the disclosure of pollution information; furthermore, this research extends to the examination of pollution and its information effects on the middle-aged and elderly.

The third literature stream specifically examines individual responses to PID. To the best of our knowledge, little empirical evidence exists on the socioeconomic impact of information disclosure on air pollution. Among the few exceptions, Barwick et al. (2019) examined the effect of PID on online searches, day-to-day shopping, and mortality costs. Shen and Sun (2022) examined the effect of PID on outdoor workers labor supply, while other related studies have focused on health behavior in response to air pollution notifications (Skov et al., 1991). While Barwick et al. (2019) examined outdoor activity changes through credit card swipes at the city level, our study provides direct evidence that the “same” person living in the exact “same” location reported engaging in fewer outdoor activities on high pollution days after PID. Furthermore, for the identical individual, we observed changes in awareness (concern for the environment) and health behavior, which allows us to further test mechanisms.

The remainder of this paper is organized as follows: Section 2 briefly introduces the background of the PID program in China. Section 3 describes the data, and Section 4 presents the identification strategy. Section 5 presents the main findings on the impact of PID on individual behavior, followed by a discussion on the health status impact in Section 6. Section 7 concludes the paper.