Estimation With Many Instrumental Variables

Abstract

Using many valid instrumental variables has the potential to improve efficiency but makes the usual inference procedures inaccurate. We give corrected standard errors, an extension of Bekker to nonnormal disturbances, that adjust for many instruments. We find that this adjustment is useful in empirical work, simulations, and in the asymptotic theory. Use of the corrected standard errors in t-ratios leads to an asymptotic approximation order that is the same when the number of instrumental variables grows as when the number of instruments is fixed. We also give a version of the Kleibergen weak instrument statistic that is robust to many instruments.