Our experience with these variables has been that they are rarely significant in determining concentrations. In addition to site-specific fixed effects we also employ city-specific weather variables to capture differences across cities in their natural cleansing abilities and in seasonal influences on emissions. While weather variables are unlikely to be strongly correlated with our economic variables their inclusion may help us obtain more accurate estimates. To capture seasonal influences on the demand for fuels and hence emissions of SO2 we include the average monthly temperature from each site. As well we have included the variation in precipitation at the site as well to proxy for the ability of precipitation to wash out concentrations. If precipitation is largely concentrated in one season then its ability to wash out concentrations over the year is reduced. Seasonal influences have been found to be important in similar studies (See for example WHO (1984)).
Common-to-World Determinants, Error Components and Excluded Variables
We assume our error term sijkt is composed of three elements. First, a common-to-world but time varying component 1t reflecting trends in the public’s awareness of environmental problems, in abatement technology, and in world prices. We capture these common-to-world components via a linear time trend. Second, we include time invariant site components 9 ijk to reflect unmeasured meteorological or topographical features of a site as well as any time invariant country-specific effects such as government or country type. And finally we include an idiosyncratic component vijkt reflecting both human and machine measurement error at the site. Most of these assumptions are not controversial, although the issue of government type deserves some discussion website.
In developing our model we allowed pollution policy to be flexible and responsive to changes in the economy. In contrast, we took the existing level of trade frictions b as exogenous. Since trade frictions undoubtedly contain a component reflecting trade policy we have in fact taken this part of policy as exogenous. This assumption may be problematic if pollution and trade policies are correlated because political economy considerations, income levels, and other factors jointly determine them. Consider for example government type.
Suppose our sample of countries was divided into two types: democracies and communist countries. Suppose democracies are both relatively open and fairly clean, while communist countries are relatively closed to trade and very dirty. As a result, if we ignore the correlation of trade and environmental policy induced by political systems, our trade intensity measure may be correlated with our equation’s error term. All else equal, open economies will appear cleaner because they are open rather than because they are democracies.