How Fama-MacBeth Can Go Wrong – And an Informative Solution

Abstract

We find that weak identification can lead to econometric problems with Fama-MacBeth regressions, including serious size distortions and biased point estimates. Two sources of weak identification are particularly important and have been little studied in the finance literature – small betas and collinearity in the beta matrix. We introduce a technique (RTP) to deal with weak identification and compare the new technique with Fama-MacBeth and other alternatives. RTP has correct size and very good power to reject misspecified models. RTP has two further useful properties: 1) it provides a warning of weak identification; 2) model rejections can be informative.