Multiple-Predictor Regressions: Hypothesis Testing

We propose a new hypothesis-testing method for multipredictorregressions in small samples, where the dependent variable isregressed on lagged variables that are autoregressive. The newtest is based on the augmented regression method (Amihud andHurvich, 2004), which produces reduced-bias coefficients andis easy to implement. The method's usefulness is demonstratedby simulations and by testing a model where stock returns arepredicted by two variables, income-to-consumption and dividendyield.