| 简介： | We propose a new method to estimate the risk premium of observable factors in a linear asset pricing model, that is valid even when the observed factors are just a subset of the true factors that drive asset prices. If some of the factors of the true model cannot be observed, standard methods yield biased estimates for the risk prices of observed factors due to omitted variable bias. Our approach marries principal component analysis with two-pass cross-sectional regressions to extract the priced latent factors from a large panel of testing assets, and use them to infer the risk price of the observable factors. In addition to correcting for omitted factors, the methodology accounts for potential measurement errors in the observed factor, and detects when such a factor is spurious or even useless. The methodology exploits the power of large cross-sections, and we therefore apply it to a large panel of equity portfolios to estimate risk prices for several workhorse linear models. |