Air pollution and analyst information production

Highlights

• · We examine whether air pollution affects analyst information production by exploiting variation in air quality in China.

• ·Compared with other analysts, analysts exposed to air pollution are less likely to issue timely or accurate forecasts.

• Air pollution deters analysts’ issuance of bold forecasts, especially negatively bold forecasts.

• Investors assign lower valuation to forecast revisions issued by analysts exposed to air pollution.

• The evidence suggests that air pollution jeopardizes analysts’ ability to provide information to capital market.

Abstract

Recent studies investigate the impact of air pollution on labor productivity. We extend this literature by showing that air pollution negatively affects equity analyst information production. Analysts exposed to air pollution are less likely to issue timely forecasts or improve their forecast accuracy. Investigating the underlying mechanism, we find that analysts exposed to air pollution are less likely to provide bold (especially, negatively bold) forecasts. We also find evidence that market pricing is less sensitive to forecast revisions issued by analysts exposed to air pollution. Our results are robust to controlling for firm/analyst and time fixed effects, as well as additional specifications employing difference-in-differences designs and placebo tests.

Introduction

In the past five years, particulate matter (PM) pollution has risen 8%, with over 80% of urban residents worldwide currently exposed to dangerous levels of outdoor pollution (Vidal, 2016; Walsh, 2016).¹ Recent economic studies find that outdoor pollution results in a reduction of worker productivity (Chang et al., 2014; Chang et al., 2016; Dong et al., 2019; Zivin and Neidell, 2012). We extend this enquiry by investigating whether outdoor air pollution affects analyst information production for the capital market.

We examine two related issues. First, we investigate whether PM pollution affects the timing and accuracy improvement of analyst earnings forecast revisions in response to firm earnings announcements. Our evidence is consistent with forecasts being less timely and of lower accuracy during polluted periods. Second, we investigate whether this effect is related to analysts having greater difficulty in information production during polluted periods. Consistent with forecasts being less-informed, we find that when analysts are in polluted cities during an earnings announcement, they are less likely to issue bold (especially, negatively bold) forecasts. Forecast boldness reflects the extent to which analysts map new information into revised forecasts (Clement and Tse, 2005; Gleason and Lee, 2003; Hong et al., 2000; Keskek et al., 2014). Our results are thus consistent with analysts reducing the extent to which they incorporate new information into their forecasts when they are forecasting under polluted conditions. Overall, our study advances understanding of the effects of air pollution on the work productivity of highly skilled professionals, providing additional insights about the hidden costs associated with air pollution.

We address our research questions by examining analyst forecast revisions in response to firm earnings announcements in China. This setting is particularly appropriate for our study. In the past two decades China has experienced severe PM pollution, which varies by location and across time. This allows us to detect differences in analyst information production relative to changes in levels of PM pollution for the same firm/analyst. Focusing on forecast revisions is also an important feature of our study. Although analysts may make a strategic trade-off between the timeliness and accuracy of their forecasts (Gul and Lundholm, 1995; Guttman, 2010), they face greater pressure to issue forecasts directly following firm earnings announcements (Zhang, 2008). Given this pressure, analysts are unlikely to delay their forecast revisions to reduce the negative effects of air pollution. On the other hand, our use of a short-term window after earnings announcements to capture the effects of PM pollution mitigates the concern that confounding factors other than air pollution are responsible for the changes in analysts forecast revisions that we observe. Taken together, these features allow a strong test of our research questions. Finally, China's air pollution is representative of similar issues in other developing countries in the world (e.g., Brazil, India, Iran). For decades, China has been criticized for developing its economy at the expense of the environment (Christmann and Taylor, 2001). Results from this setting may shed light on indirect costs related to such a country-level strategy, with practical implications for many developing countries/regions worldwide that follow a similar development path.

A large number of studies suggest that air pollution worsens the performance of agents. This effect is due to its negative effects on health (Chang et al., 2014; Chang et al., 2016; Lavy et al., 2014; Suglia et al., 2007; Zivin and Neidell, 2012) and mood (Bakian et al., 2015; Calderon-Garciduenas et al., 2015; Lim et al., 2012; Power et al., 2015; Szyszkowicz, 2007).² Extant research finds that air pollution leads to biased decisions by investors (Dong et al., 2019; Heyes et al., 2016; Li et al., 2019). We extend this research to investigate whether air pollution impacts analyst forecasts, which play an important informational role in capital markets.

Our empirical findings lend support to our hypotheses. Compared to analysts forecasting during periods of low PM pollution, analysts exposed to the highest levels of pollution are 14% less likely to revise their earnings forecasts within a two-day short window of earnings announcements. In addition, for forecast revisions issued during the short-term window of earnings announcements, analysts facing the worst air pollution levels are 4% less likely to improve their forecast accuracy than are analysts experiencing good air quality. In terms of forecast boldness, analysts facing the highest levels of air pollution are 4.8% (4.9%) less likely to issue bold (negatively bold) forecasts compared with analysts experiencing good air quality. When benchmarked against sample means of our dependent variables, these negative impacts are economically significant. All these findings remain robust when we include a battery of control variables at both the firm and analyst levels, as well as firm/analyst/time fixed effects and for alternative model specifications.

To strengthen our identification and rule out the possibility that the observed findings arise from high ability analysts migrating to cities with low air pollution, we conduct difference-in-differences (hereafter diff-in-diff) designs following Li et al., 2019. To identify the treatment sample, we require that for an analyst-firm observation, there is a stark change (at least one standard deviation) in the air quality index (AQI) between two adjacent quarters. Use of a stark quarter-to-quarter change in AQI not only takes the advantage of exogenous variation in air quality but also avoids the possibility that high ability analysts migrate to clean cities in short-term. For each observation in the treatment sample, we choose an observation from the remaining sample that did not experience drastic changes in AQI as the control observation. The control observation must have forecasts issued by an analyst from a different city but following the same firm over the same period as the treatment observation. As a result, the difference between changes in the dependent variables for the treatment sample and changes for the control sample over the same period should reflect the effects of air pollution. Results from the diff-in-diff tests show that analysts exposed to air pollution (clean air) tend to issue less (more) timely, less (more) accurate and less (more) negatively bold forecasts compared with analysts without experiencing drastic air quality changes. Moreover, we conduct two placebo tests by randomizing PM pollution within each analyst-firm and by randomizing cities to each analyst-quarter. Both placebo tests help mitigate the concern that our findings are driven by some time-variant city-level omitted variables. Both tests corroborate our main findings.

To understand whether there are certain factors which could alleviate the negative effects of air pollution on analyst information production, we partition the full sample based on several firm or analyst level characteristics and re-estimate our baseline models. We find that factors related to both analyst supply of forecasts and investor demand for forecasts play a key role. From the supply side, the negative effects of air pollution are more pronounced in the subsample where analysts have a higher workload or where competition among analysts is weak. From the demand side, investors' high level of demand for information following annual earnings announcements attenuates the negative effects of the PM pollution.

We also investigate whether investors understand the reduction in forecast quality when analysts provide forecasts during periods with high levels of air pollution. We examine whether investor response to forecast revisions is moderated by the levels of air pollution faced by analysts. We compute the cumulative abnormal return (CAR) on days t and t + 1 relative to the forecast revision date. We then regress the CAR on an interaction term of forecast revision and air quality over the two days preceding the forecast revision date. We find that investors respond less to forecast revisions issued after polluted days, suggesting that the market understands the negative effects of air pollution on analyst information production.

We contribute to three strands of literature. First, we extend research investigating the impact of air pollution on worker productivity (Archsmith et al., 2016; Chang et al., 2014; Chang et al., 2016; Zivin and Neidell, 2012) and research investigating how air pollution affects decisions made by capital market participants (Dong et al., 2019; Heyes et al., 2016; Li et al., 2019). Previous research focuses on output of physical goods, non-financial services or investor trading behaviors. However, the study of how air pollution affects highly skilled professionals (e.g., analysts) is very limited. To the best of our knowledge, the only paper closely related to our study is Dong et al. (2019), who find that air pollution during analyst site visits induces more pessimistic forecasts (forecast bias). We extend Dong et al. (2019) by examining the impact of air pollution on additional dimensions of analyst information production. First, we focus on analyst forecast revisions, which proxy for analyst information production (Clement and Tse, 2005; Keskek et al., 2014), rather than on forecast bias. Second, whereas Dong et al. (2019) examine a window of 15 days, we employ a 2-day time period. Our short-term window mitigates concerns of confounding factors underlying our results.

Second, we contribute to the literature linking weather conditions to information production (more broadly, decision making) of market participants (Chen et al., 2017; DeHaan et al., 2017; Heyes et al., 2016; Li et al., 2019; Huang et al., 2018). Air pollution is fundamentally different from the more general notion of weather conditions (e.g., sunshine). Unlike bad weather, air pollution can be directly influenced by policy makers, implying that the costs of air pollution observed in our study might be controllable.³ Further, our results hold after controlling for local weather conditions. Indeed, we find that the effect of air pollution on analyst forecast timeliness is as large as that of weather conditions. Unlike poor weather, however, we find that air pollution significantly affects analyst forecast accuracy and the likelihood of analysts to issue bold forecasts (particularly negatively bold forecasts). This indicates that relative to poor weather, air pollution has large incremental impacts on analyst information production.

Finally, our study adds to the literature on determinants of analyst forecast timeliness and properties of analyst forecasts (Clement and Tse, 2005; Yezegel, 2015; Zhang, 2008). Our findings imply that the air pollution may negatively affect functioning of the capital market. Given that air pollution is widespread and a growing problem throughout the world, particularly in emerging countries/regions, our findings provide insights regarding an additional indirect cost of prioritizing economic development over environmental protection.

The reminder of the paper proceeds as follows. Section 2 briefly discusses the air pollution in China. Section 3 summarizes prior literature and proposes hypotheses. Section 4 describes the research design. Section 5 reports empirical findings and Section 6 concludes.