CMStatistics 2022: Start Registration
View Submission - CFE
Title: A simplified decomposition model for asset return predictability Authors:  Arsene Brou - Laval University (Canada) [presenting]
Richard Luger - Laval University (Canada)
Abstract: A previous modelling approach decomposes asset returns into their signs and absolute values, and obtains the joint distribution by specifying a multiplicative error model for the absolute values, a dynamic binary choice model for the signs, and a bivariate copula for their interaction. With this copula-based approach, each component is specified conditional on past information. We propose a simplified approach that recovers the dynamics of the joint distribution with a specification for the sign component that conditions on past information and the contemporaneous absolute value. In sharp contrast to the original decomposition model, the simplified approach avoids the need to specify a copula for joining a continuous margin with a discrete one. Simulation results demonstrate that the simplified approach does as well as copula-based specifications, and empirical findings show that a larger degree of stock return predictability is revealed by decomposition modelling than by traditional predictive regressions.