Working Paper: NBER ID: w12055
Authors: Jay Shanken; Guofu Zhou
Abstract: In this paper, we conduct a simulation analysis of the Fama and MacBeth (1973) two-pass procedure, as well as maximum likelihood (ML) and generalized method of moments estimators of cross-sectional expected return models. We also provide some new analytical results on computational issues, the relations between estimators, and asymptotic distributions under model misspecification. The GLS estimator is often much more precise than the usual OLS estimator, but it displays more bias as well. A "truncated" form of ML performs quite well overall in terms of bias and precision, but produces less reliable inferences than the OLS estimator.
Keywords: Asset Pricing; Fama-Macbeth Procedure; Maximum Likelihood; Generalized Method of Moments
JEL Codes: G12
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Fama-Macbeth estimator (C51) | biases (D91) |
| model misspecification (C52) | biases in Fama-Macbeth estimator (C51) |
| GLS estimator (C51) | more precise than OLS estimator (C20) |
| GLS estimator (C51) | more bias than OLS estimator (C51) |
| truncated ML estimator (C51) | performs well in terms of bias and precision (C52) |
| truncated ML estimator (C51) | less reliable inferences than OLS (C20) |
| model misspecification (C52) | affects asymptotic distribution of Fama-Macbeth estimator (C51) |
| pricing errors (D49) | systematically affect risk premium estimates (C51) |
| ML and GMM methods (C38) | yield more accurate estimates than OLS and WLS (C51) |
| ML estimator (C51) | virtually unbiased in larger samples (C46) |
| ML estimator (C51) | lower root mean square errors (RMSE) than OLS and WLS (C20) |