This article focuses on the empirical approach proposed by Hall and Jones (1999) to estimate the effect of what they call ‘social infrastructure’ on productivity across countries. We consider the issue of weak identification in their linear instrumental variables model.
The evidence obtained from partially robust estimators, such as the k-class and jackknife estimators, is interpreted on the basis of Monte Carlo studies. Our findings suggest that using certain k-class estimators allows exclusive reliance on the linguistic variables to instrument
for institutional quality despite their low correlation with the endogenous regressor in question.