Modeling and Predicting Stock Returns:The Rule of Parsimony

초록

This study examines a variety of time-series forecasting models to uncover an economically effective forecasting model of stock returns using both valuation ratios and macroeconomic variables. We use the Schwarz criterion as a barometer of model selection and conclude that from the economic perspective as opposed to the statistical perspective the parsimonious multiple regression model with a select set of explanatory variables, mainly book-to-market ratio, default risk premium, and Treasury bill rate, shows as good a capacity to predict stock returns as more complicated forecasting models such as E-GARCH or I-GARCH. An out-of-sample test confirms our conclusion.

키워드