Estimating the Adverse Health Effects of particulates.
The estimates of the health costs of particulates come from a large empirical literature on tion and health outcomes. The typical approach taken in this literature is to relate adult mortality in a geographical area to the level of particulates in the air in that area. The results from this type of analysis are suspect, however, due to the key empirical problem high lighted in Chapter 3: the areas with more particulates may differ from areas with fewer particulates in many other ways, not just in the amount of particulates in the air. Imagine, for example, that researchers compared two areas, one with old plants that emit a lot of particulates, and one with newer plants that are much cleaner. If the researchers found higher mortality in the areas with the older dirty plants, they might attribute this to the effects of particulates on human health. Suppose, however, that older plants are also less safe places to work than newer plants. In this case, the higher mortality areas with older plants might be due to workplace acci dents, not pollution. It is difficult to observe valid treatment and control groups in a situation like this; you can't just com pare dirty areas to cleaner ones because so many other things could differ between them, imparting bias to the estimates. Chay and Greenstone (2003) addressed this problem using the regulatory changes induced by the CAA. This act applied differentially to different counties in the United States, based on whether they were above or below a man dated "attainment" of clean air levels. Counties with emis sions above a mandated threshold (nonattainment counties)