Momentum Without Crashes

We construct a momentum factor that identifies cross-sectional winners and losers based on a weighting scheme that incorporates all the price data, over the entire lookback period, as opposed to only the first and last price points of the window. The weighting scheme is derived from the fractional-difference filter- a statistical transformation that preserves memory in the data, and has an economic interpretation of coherently combining reversal and momentum patterns in the returns. Our extensive out-of-sample analysis shows that the new fractional momentum strategy not only achieves significantly higher (risk adjusted) returns, but also mitigates the notoriously large drawdowns of the classical momentum and short-term reversal strategies. The performance results are robust with respect to transaction costs and other real world frictions; excess returns are not explained by other asset pricing factors; and they are pervasive across different asset universes and foreign markets.