Transaction-cost-aware Factors

The asset pricing literature identifies hundreds of economic characteristics that appear to be priced in expected returns. However, candidate factors are often oblivious to how investors would optimally trade such signals in the presence of trading fees. In this paper, I show that factors constructed without taking transaction costs into account have little hope of spanning investors' achievable efficient frontier, which reflects the cost of trading. I introduce transaction-cost-aware versions of leading factors in the literature, assuming that investors trade off the benefits of tracking characteristics closely against trading fees incurred in the process. Transaction-cost-aware factors dramatically outperform traditional ones after fees. Models that incorporate transaction-cost-aware factors increase net maximum squared Sharpe ratios by up to a factor of 3. I further suggest that discretionary choices in factor construction can invalidate asset pricing inference. Excessively aggressive rebalancing can fully offset the pricing ability of otherwise promising economic characteristics.