Penalized Regressions in Factor Pricing Models and Validity of Cross-validaiton

Working paper


This paper investigates the consequences and their solutions of using k-fold cross validation methodology to determine the regularization parameter (λ) in penalized regressions, such as ridge and LASSO regressions, when serial correlations exist within data. In particular, I focus on the application of the post-LASSO of Belloni, Chernozhukov,et al. (2013) and the double-selection LASSO method proposed by Belloni, Chernozhukov,and Hansen (2014) and Feng, Giglio, and Xiu (2020) to asset pricing models.