Model Selection with Transaction Costs

Abstract

Failing to account for transaction costs materially impacts inferences drawn when evaluating asset pricing models, biasing tests in favor of those employing high cost factors. Ignoring transaction costs, the Hou, Xue, and Zhang (2015) q-factor model and the Barillas and Shanken (2018) six-factor model models have high maximum squared Sharpe ratios and small alphas across 120 anomalies. They do not, however, come close to spanning the achievable mean-variance efficient frontier. Accounting for transaction costs, the Fama and French (2015, 2018) five-factor model has a significantly higher squared Sharpe ratio than either of these alternative models, while variations employing cash profitability perform better still. More generally, these results highlight the importance of incorporating real-world concerns into financial research.

Keywords: Factor Models, Trading Costs, Mispricing

JEL Classification: G11, G12, G14