An important question facing mutual fund investors is whether funds with positive alpha, net of fees and costs, can be distinguished, ex-ante, from funds with negative alpha. Most individual investors must believe this to be true, because the vast majority of their assets are invested in actively managed funds.
Reaching this conclusion requires that investors believe there are variables (such as past performance or active share) associated with future alphas.
While there have been some papers that claim to have found the holy grail of performance predictors, the research shows that skill-based alphas tend to be erased by fund fees and diseconomies of scale arising from fund flows and arbitrage (once anomalies are discovered, they tend to quickly disappear as the markets become ever more efficient).
As my co-author Andrew Berkin and I explain in “The Incredible Shrinking Alpha,” there are major trends that have led to an increase in arbitrage activity over the last several decades, making stock markets more efficient and alpha generation more difficult).
Christopher Jones and Haitao Mo contribute to the literature with their November 2016 study, “Out-of-Sample Performance of Mutual Fund Predictors.” They began by identifying 20 different predictors from 17 papers. Unfortunately, four of those predictors were impossible to replicate, as they relied on proprietary or hand-collected data, leaving 16 for which the authors were able to analyze out-of-sample performance.
Their study covered the period January 1961 to January 2013 and more than 3,500 mutual funds. On average, the authors’ quintile sort produced an in-sample alpha spread of 1.96% a year with an average t-statistic of 2.28.
Following is a summary of their findings: