The Impact of Pollution on Worker Productivity.

As one of the primary factors of production, labor is an essential element in every nation’s economy. Investing in human capital is widely viewed as a key to sustaining increases in labor productivity and economic growth. While health is increasingly seen as an important part of human capital, environmental protection, which typically promotes health, has not been viewed through this lens. Indeed, such interventions are typically cast as a tax on producers and consumers, and thus a drag on the labor market and the economy in general. Given the large body of evidence that causally links pollution with poor health outcomes (e.g., Bell et al. 2004; Chay and Greenstone 2003; Currie and Neidell 2005; Dockery et al. 1993; Pope et al. 2002), it seems plausible that efforts to reduce pollution could in fact also be viewed as an investment in human capital, and thus a tool for promoting, rather than retarding, economic growth. The key to this assertion lies in the impacts of pollution on labor market outcomes. While a handful of studies have documented impacts of pollution on labor supply (Carson, Koundouri, and Nauges 2011; Graff Zivin and Neidell forthcoming; Hanna and Oliva 2011; Hausman, Ostro, and Wise 1984; Ostro 1983), their focus on the extensive margin, where behavioral responses are nonmarginal, only captures highvisibility labor market impacts. Pollution is also likely to have productivity impacts on the intensive margin, even in cases where labor supply remains unaffected. Since worker productivity is more difficult to monitor than labor supply, these more subtle impacts may be pervasive throughout the workplace, so that even small individual effects may translate into large welfare losses when aggregated across the economy. There is, however, no systematic evidence to date on the direct impact of pollution on worker productivity. This paper is the first to rigorously assess this environmental productivity effect. Estimation of this relationship is complicated for two reasons. One, although datasets frequently measure output per worker, these measures do not isolate worker