Return Predictability under the Alternative
University of Notre Dame - Department of Finance
Timothy T. Simin
Pennsylvania State University
Daniel R. Smith
Queensland University of Technology - School of Economics and Finance; Simon Fraser University; Financial Research Network (FIRN)
February 6, 2013
We propose a new standard error estimator useful in tests of long-run return predictability. Our standard error exhibits substantial power gains with only minor size distortions relative to popular tests, does not require imposing the null hypothesis that returns are unpredictable, and does not require estimation of additional autocovariance terms. Deriving the covariance matrix without imposing the null hypothesis also produces two new terms in the spectral density matrix capturing the volatility of the shock to the regressor and the correlation between the shocks of the predictor variable equation and the prediction equation. Empirically, we show that failure to detect return predictability at longer horizons is partially due to the lower power of tests derived under the null hypothesis. For many predictors, giving the alternative a chance allows short-run predictability to survive in long-horizon regressions.
Number of Pages in PDF File: 47
Keywords: Predictability, overlapping observations, analytical standard errors, size, power
JEL Classification: G12, C52working papers series
Date posted: August 26, 2012 ; Last revised: February 10, 2013
© 2013 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo6 in 0.406 seconds