Stock Return Serial Dependence and Out-of-Sample Portfolio Performance

We study whether investors can exploit stock return serial dependence to improve out-of- sample portfolio performance. To do this, we first show that a vector-autoregressive (VAR) model estimated with ridge regression captures daily stock return serial dependence in a stable manner. Second, we characterize (analytically and empirically) expected returns of VAR-based arbitrage portfolios, and show that they compare favorably to those of existing arbitrage portfolios. Third, we evaluate the performance of VAR-based investment (positive-cost) portfolios. We show that, subject to a suitable norm constraint, these portfolios outperform the traditional (unconditional) portfolios for transaction costs below 10 basis points.