Why Mutual Fund Alpha Predictors Weaken Out of Sample

Summary

This review examines whether published indicators of future mutual-fund performance retain their predictive power outside the periods in which they were studied. It compares high-minus-low portfolio alpha spreads and cross-sectional regression slopes across 27 predictors, using U.S. equity funds and additional evidence from corporate bond funds. The analysis finds that typical out-of-sample spreads are much smaller than in-sample spreads, with much of the decline occurring after the original samples end.

The authors test data mining, investor learning, market arbitrage, and competition among funds as explanations. After controlling for measures such as short interest, stock turnover, hedge-fund assets, and fund-industry competition, arbitrage activity accounts for much of the decline; competition appears to play a smaller role. The evidence does not convincingly support data mining or learning as the main causes. The findings use historical data and published predictors, so they do not establish that every indicator will fail or that past relationships will hold in other markets. The review also reports that estimated alpha among the highest-predicted funds fell over time, underscoring the need to account for changing market conditions.

Key ideas

• The study measures predictor performance using both high-minus-low alpha spreads and cross-sectional regression slopes.
• Most examined fund alpha predictors weaken substantially outside their original samples.
• Greater market-wide arbitrage activity explains much of the observed decline in predictive power.
• Fund competition has some explanatory power, but its effect is less robust than arbitrage activity.
• The historical results suggest that alpha prediction and the performance of selected funds can change over time.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.