Can active portfolio management outperform passive benchmarks on a risk-adjusted basis when accounting for transaction costs?

Abstract

Mean-variance optimization remains a cornerstone in modern portfolio theory, yet the literature shows that optimized portfolios struggle to consistently and significantly outperform simple benchmarks, such as the naïve 1/N allocation strategy, in risk-adjusted out-of-sample performance. This thesis examines whether incorporating bid-ask spread-based transaction costs directly into the mean-variance optimization problem can improve risk-adjusted performance sufficiently to outperform the naïve 1/N benchmark. Using U.S. equity data from 2002 to 2024, we study two mean-variance-based portfolio strategies: a standard factor-implied MVO and a Black-Litterman-inspired fusion variant. Portfolios are constructed using a rolling-window estimation and are evaluated out-of-sample using a broad set of performance metrics. We find that explicitly accounting for transaction costs and varying turnover penalization substantially affects portfolio composition and trading intensity. While the actively managed portfolios outperform the naïve strategy for certain subperiods of the sample, neither strategy consistently outperforms the benchmark over the full sample. The results further suggest that although transaction cost awareness improves trading efficiency, residual estimation error and sensitivity to market conditions continue to limit the out-of-sample performance of mean-variance optimized portfolios. Overall, the findings contribute to the literature on portfolio optimization by providing new evidence on the performance of transaction cost-aware mean-variance strategies, and by reinforcing the role of the naïve diversification as a robust benchmark.