Bian and Dickey (1996) developed a robust Bayesian estimator for thevector of regression coefficients using a Cauchy-typeg-prior. This estimator is an adaptive weighted average of the least squaresestimator and the prior location, and is of great robustness withrespect to at-tailed sample distribution. In this paper, weintroduce the robust Bayesian estimator to the estimation of theCapital Asset Pricing Model (CAPM) in which the distribution of theerror component is well-known to be flat-tailed. To support ourproposal, we apply both the robust Bayesian estimator and the leastsquares estimator in the simulation of the CAPM and in the analysisof the CAPM for US annual and monthly stock returns. Our simulationresults show that the Bayesian estimator is robust and superior tothe least squares estimator when the CAPM is contaminated by largenormal and/or non-normal disturbances, especially by Cauchydisturbances. In our empirical study, we find that the robustBayesian estimate is uniformly more efficient than the least squaresestimate in terms of the relative efficiency of one-step aheadforecast mean square error, especially for small samples.
展开▼