Weak Instruments in Growth Regressions

Abstract

This paper revisits four recent cross-country empirical studies on the effects of inequality on growth. All four studies report strongly significant negative effects, using the popular system generalized method of moments estimator that is frequently used in cross-country growth empirics. This paper shows that the internal instruments relied on by this estimator in these inequality-and-growth regressions are weak, and that weak instrument-consistent confidence sets for the effect of inequality on growth include a wide range of positive and negative values. This suggests that strong conclusions about the effect of inequality on growth— in either direction—cannot be drawn from these studies. This paper also systematically explores a wide range of alternative sets of internal instruments, and finds that problems of weak instruments are pervasive across these alternatives. More generally, the paper illustrates the importance of documenting instrument strength, basing inferences on procedures that are robust to weak instruments, and considering alternative instrument sets when using the system generalized method of moments estimator for cross-country growth empirics.