Predicting Stock Price by Applying the Residual Income Model and Bayesian Statistics

Abstract

Over the past decade, the Residual Income Model (RIM) has often been applied as a framework for accounting-based equity valuation. However, empirical implementations of the RIM are problematic, yielding large forecast errors. In this paper, we seek to improve the implementation of the RIM, specifically by using Bayesian statistics to improve the inference mechanics. Bayesian statistics has advantage over the commonly used Maximum Likelihood method because Bayesian parameters are stochastic and therefore Bayesian models are more adaptive to dynamical changes in the data. Indeed, our empirical results show that Bayesian forecasts are more accurate than Maximum Likelihood forecasts.