Measuring the health effects of air pollution: to what extent can we really say that people are dying from bad air?

Abstract

Estimation of the effects of environmental impacts is a major focus of current theoretical and policy research in environmental economics. Such estimates are used to set regulatory standards for pollution exposure; design appropriate environmental protection and damage mitigation strategies; guide the assessment of environmental impacts; and measure public willingness to pay for environmental amenities. It is a truism that the effectiveness of such strategies depends crucially on the quality of the estimates used to inform them. However, this paper argues that in respect to at least one area of the empirical literature—the estimation of the health impacts of air pollution using daily time series data—existing estimates are questionable and thus have limited relevance for environmental decision-making. By neglecting the issue of model uncertainty—or which models, among the myriad of possible models researchers should choose from to estimate health effects—most studies overstate confidence in their chosen model and underestimate the evidence from other models, thereby greatly enhancing the risk of obtaining uncertain and inaccurate results. This paper discusses the importance of model uncertainty for accurate estimation of the health effects of air pollution and demonstrates its implications in an exercise that models pollution-mortality impacts using a new and comprehensive data set for Toronto, Canada. The main empirical finding of the paper is that standard deviations for air pollution-mortality impacts become very large when model uncertainty is incorporated into the analysis. Indeed they become so large as to question the plausibility of previously measured links between air pollution and mortality. Although applied to the estimation of the effects of air pollution, the general message of this paper—that proper treatment of model uncertainty critically determines the accuracy of the resulting estimates—applies to many studies that seek to estimate environmental effects.