The Weak Instruments Problem in Factor Models

Abstract

This study considers the instrumental variable estimation of factor models. Specifically, we investigate the weak instruments problem, which is not well investigated in the literature, in detail. We show that the signal-to-noise ratios, which are defined by the variance ratios of the common components to the errors, of the scaling variable and the instruments are the major factors that determine the strength of the instruments. By means of Monte Carlo simulation, we investigate the performance of the 2SLS estimator with several types and strengths of instruments, and we confirm that the simulation results are consistent with the theoretical implications.

Keywords: factor model; two-stage least squares; weak instruments