Title: Time-varying non-linear predictions of asset returns
Authors: Frank Rotiroti - The University of Texas at Austin (United States) [presenting]
Carlos Carvalho - The University of Texas at Austin (United States)
Jared Murray - University of Texas at Austin (United States)
Abstract: A Bayesian approach is presented to modeling time-dependent data based on an extension of the Bayesian Additive Regression Trees (BART) model. Like BART, our approach consists of a Bayesian sum-of-trees model that constrains each tree to be a weak learner by a regularization prior; however, rather than equip each terminal node with a single mean parameter, we introduce a series of mean parameters generated according to a first-order autoregressive process. With this approach, we are better able to capture the dynamics of time-dependent data, while also taking advantage of the ability due to the BART framework to model nonlinearities and interactions among the predictors, as we demonstrate through simulation studies as well as an application to asset pricing.