Markowitz portfolios under transaction costs

Markowitz portfolio selection is a cornerstone in finance, in academia as well as in the industry. Most academic studies either ignore transaction costs or account for them in a way that is both unrealistic and suboptimal by (i) assuming transaction costs to be constant across stocks and (ii) ignoring them at the portfolio-selection state and simply paying them after the fact. Our paper proposes a method to fix both shortcomings. As we show, if transaction costs are accounted for (properly) at the portfolio-selection stage, net performance in terms of the Sharpe ratio often increases, in particular for high-turnover strategies.