Momentum Strategies: Some Bootstrap Tests

Abstract

This study introduces a new estimation-based bootstrap simulation procedure to test whether different returns-generating models can explain the profitability of momentum strategies first documented in Jegadeesh and Titman (1993). We incorporate simple random walk and multifactor models and allow for autocorrelation, cross-correlation, conditional heteroscedasticity and predictability through conditioning information variables. We also evaluate alternative sampling procedures for the bootstrap simulations. None of the models, however, are able to generate simulated profits as large as the actual profits. We do find, however, that accounting for time-varying expected returns with market-wide and macroeconomic instrumental variables can explain 75 to 80 percent of the profits.