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, Hou, Xue, and Zhang (2015, Review of Financial Studies, 28, 650–705) q-factor model and Barillas and Shanken (2018, The Journal of Finance, 73, 715–754) six-factor models have high maximum squared Sharpe ratios and small alphas across 205 anomalies. They do not, however, come close to spanning the achievable mean-variance efficient frontier. Accounting for transaction costs, the Fama and French (2015, Journal of Financial Economics, 116, 1–22; 2018, Journal of Financial Economics, 128, 234–252) five-factor model has a significantly higher squared Sharpe ratio than either of these alternative models, while variations employing cash profitability perform better still.

Scientific Portfolio AI- Generated Summary

This paper explores the impact of transaction costs on asset pricing models. The authors argue that failing to account for transaction costs can bias tests in favor of high-cost factors, and that the Fama and French five-factor model performs better when transaction costs are considered.

The paper begins by describing the problems that arise when comparing factor models based on their maximum squared Sharpe ratios without accounting for transaction costs. The authors note that factor models are often judged by how well they price test assets, with small abnormal returns relative to the model indicating success and large abnormal returns indicating that an investor trading a model’s factors can significantly expand their investment opportunity set by additionally trading the test asset. However, comparing factor models by how well they price test assets is problematic for several reasons, including the fact that transaction costs can significantly impact performance.

The authors then introduce the factor models they compare, including the Fama and French five-factor model, the q-factor model, and the six-factor model. They note that these models have been shown to perform well in previous studies, but that these studies have not accounted for transaction costs.

The authors then present their main results comparing models after accounting for transaction costs. They find that the Fama and French five-factor model performs better than the q-factor and six-factor models when transaction costs are considered. They also find that the performance of all three models is significantly worse when transaction costs are accounted for, indicating that transaction costs can have a significant impact on performance.

The authors then examine how transaction costs impact comparisons of models based on their performance explaining anomalies. They find that the Fama and French five-factor model performs better than the q-factor and six-factor models in this context as well.

Finally, the authors expand on their main comparisons by incorporating cost-mitigation techniques in factor construction. They find that these techniques can significantly improve performance, but that the Fama and French five-factor model still performs better than the q-factor and six-factor models.

Overall, the authors argue that proper performance evaluation requires explicitly accounting for transaction costs, and that statistical techniques that explicitly account for the fact that investors incur transaction costs regardless of whether they buy or sell are necessary. They conclude that the Fama and French five-factor model is the best-performing model when transaction costs are considered, and that cost-mitigation techniques can further improve performance.