Discussion of Identification Robust Testing of Risk Premia in Finite Samples
讨论了Fama-MacBeth两步法在有限样本中可能存在的两个问题：因子与收益的低相关性（类似弱工具变量）以及时间序列观测数相对于资产数过少，并探讨了识别稳健的检验方法。
The Fama–MacBeth (FM) two-pass procedure is widely used in empirical Finance. Given a linear asset pricing model, the first step is time-series regressions of asset returns onto pricing factors to estimate the betas, which are the relevant risk measure in this setting. The second step is cross-sectional regressions of average returns onto the estimated betas to estimate the prices of risk, or risk premia as called in this paper. The errors in the cross-sectional regressions are interpreted as the pricing errors of the model and are known as alphas. There are two potential problems for this empirical method. There may be low correlation between returns and factors, as pointed out by Kan and Zhang (1999) (useless factors), which is similar to weak instruments in IV methods. A second potential problem is a low number of time-series observations compared with the number of assets (limited T versus large...