Maximizing equity market sector predictability in a Bayesian time-varying parameter model

A large body of evidence has emerged in recent studies confirming that macroeconomic factors play an important role in determining investor risk premia and the ultimate path of equity returns. This paper illustrates how widely tested financial and economic variables from these studies can be employed in a time varying dynamic sector allocation model for U.S. equities. The model developed here is evaluated using Bayesian parameter estimation and model selection criteria. We find that using the Kalman filter to estimate time varying sensitivities to predetermined risk factors results in significantly improved sector return predictability over static or rolling parameter specifications. A simple trading strategy developed here using Kalman filter predicted returns as input provides for potentially robust long run profit opportunities. JEL classification: G12; C11

• Type: Journal
• Year: 2008
• Where: CSDA
• Authors: Lorne D. Johnson, Georgios Sakoulis
• Keywords: Bayesian Parameter Estimation, CSDA 2008, Investor Risk Premia, Kalman Filter