Methodology and variables

• Pages: 128

Page 128

Econometric work with data for several countries spanning a number of years faces some daunting challenges: countryand time-specific effects, endogeneity of explanatory variables, omission of relevant variables, and uncertainty about the effectiveness of the underlying statistical model, among other things. In this study, we used two methodologies.

The first is a traditional framework to reproduce some of the existing results in the literature and to show that the methodology is perhaps inadequate for such analysis. In particular, while correcting for a number of the econometric problems noted, this framework cannot address model uncertainty, which arises because of the lack of clear theoretical guidance on the choice of explanatory variables.

The second methodology-a check for robustness-corrected for this problem along with the others. In simple terms, it attempts to account for all possible combinations of explanatory variables in the statistical regressions.We used 18 potential explanatory variables, basing the robustness check inferences on the results of a very large number of regressions.

The 18 potential explanatory variables accounted for

* overall average income (which was kept in...