Gary Koopa, , Ross McKitrickb, , and Lise Tolea, c,
a Department of Economics, University of Strathclyde, Glasgow, UK
b Department of Economics, University of Guelph, Guelph, ON, Canada
c School of Business, University of Edinburgh, Edinburgh UK
Received 6 March 2008; revised 21 December 2009; accepted 8 January 2010. Available online 3 March 2010.
Many studies have reported a relationship between urban air pollution levels and respiratory health problems. However, there are notable variations in results, depending on modeling approach, covariate selection, period of analysis, etc. To help clarify these factors we compare and apply two estimation approaches: model selection and Bayesian model averaging, to a new data base on 11 Canadian cities spanning 1974–1994. During this interval pollution levels were typically much higher than the present. Our data allow us to compare monthly hospital admission rates for all lung diagnostic categories to ambient levels of five common air contaminants, while controlling for income, smoking and meteorological covariates. In the most general specifications we find the here-observed health effects of air pollution are very small and insignificant, with signs that are typically opposite to conventional expectations. Smoking effects are robust across specifications. Considering the fact that we are examining an interval of comparatively high air pollution levels, and the contrast between our results and those that have been published previously, we conclude that extra caution should be applied to results estimated on short and/or recent data panels, and to those that do not control for model uncertainty and socioeconomic covariates.Read more.