Momentum is the tendency for assets that have performed well (poorly) in the recent past to continue to perform well (poorly) in the future, at least for a short period of time. The momentum effect is one of the most pervasive asset pricing anomalies documented in the ?nancial literature: Stocks with highest returns over the past six to 12 months continue to deliver above-average returns in the subsequent period.
In 1997, Mark Carhart, in his study “On Persistence in Mutual Fund Performance,” was the first to use cross-sectional (or relative) momentum, together with the three Fama–French factors (market beta, size, and value), to explain mutual fund returns. Initial research on cross-sectional momentum was published by Narasimhan Jegadeesh and Sheridan Titman, authors of the 1993 study “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency.”
As my co-author, Andrew Berkin, and I show in “Your Complete Guide to Factor-Based Investing,” the evidence supporting the momentum factor (both cross-sectional and time-series, or absolute momentum) and premium is persistent across time, pervasive around the globe and across asset classes, robust to various definitions, and implementable. We also provide the well-documented behavioral explanations for the factor’s existence.
Source: Quantitative Momentum (2016). The chart shows the invested growth of intermediate term momentum (2-12 month lookback) from 1927 to 2015 using data from Ken French’s website. The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.
One of the critiques of cross-sectional momentum is that it has a dark side, being subject to crashes (though this is only true for long-short portfolios, not long-only portfolios), such as the one that occurred in 2009 when the Fama-French French momentum factor (UMD, or up minus down) returned -83 percent. (See for example, “Momentum Crashes,” by Kent Daniel and Toby Moskowitz.)
Crashes to cross-sectional momentum occur because the long leg of momentum has positive exposures to the styles that performed well in the recent past, while the short leg is exposed to those that underperformed. For example, in bear markets, high beta stocks tend to underperform while low beta stocks outperform. As a result, momentum can exhibit negative returns if the market experiences a sharp turn. Below is Table 2 from the Momentum Crashes paper, which highlights some of the horrible monthly returns associated with the long/short momentum factor.
Source: Momentum Crashes (2016). The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.
Dealing with Momentum Crashes
A way to reduce the risk of cross-sectional momentum crashes was presented in the 2016 published paper, “Idiosyncratic Momentum: U.S. and International Evidence,” by Denis B. Chaves (working paper version here). Chaves utilized a regression approach to remove the return component due to market beta and thus produced a new definition of momentum with reduced volatility. His results held over a sample of 21 countries as well as in the United States.
The chart below highlights the effectiveness of idiosyncratic momentum (IUMD) relative to traditional momentum (UMD).
Source: Chaves (2016). The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.
Interestingly, Chaves found that this new version of momentum also works in Japan, where using the traditional academic definition of momentum produces weak results (although Cliff Asness would argue the result isn’t weak).
Idiosyncratic momentum was first investigated by Roberto C. Gutierrez and Christo A. Prinsky in their study “Momentum, Reversal, and the Trading Behaviors of Institutions,” which was published in the February 2007 issue of the Journal of Financial Markets. They identified two types of momentum in stock returns — one due to returns relative to other stocks (the traditional cross-sectional measure) and one due to firm-specific abnormal returns (idiosyncratic momentum), where abnormal is determined by a stock’s idiosyncratic return variation. They found that despite similar performances over the first year, these momentum portfolios perform dramatically differently beyond year one — relative-return momentum reverses strongly (producing returns of -0.40 percent per month in months 13 through 60), while abnormal-return momentum continues for years (producing returns of 0.20 percent per month in months 13 through 60).
Gutierrez and Prinsky also found that institutional investors contribute to both types of momentum they buy relative-return winners and sell relative-return losers at rates far above average, but buy abnormal-return winners and sell abnormal-return losers as if they were any other stock. These results suggest that institutions chase relative returns, possibly resulting in an overreaction, but institutions ignore firm-specific abnormal returns, possibly resulting in an underreaction.
David Blitz, Matthias X. Hanauer, and Milan Vidojevic, contribute to the literature on momentum with their April 2017 paper “The Idiosyncratic Momentum Anomaly.” This study built on the 2011 paper by David Blitz, Joop Huij, and Martin Martens “Residual Momentum” which was published in the June 2011 issue of the Journal of Empirical Finance. Their data sample covers the period from 1926 through 2015. The following is a summary of their findings:
- Idiosyncratic momentum (iMOM) cannot be explained by any of the established asset pricing factors, such as market, size, value, operating pro?tability, and investment, even if the total return momentum factor is included.
- Idiosyncratic momentum subsumes total return momentum in some tests, while the converse is never the case.
- Overcon?dence, overreaction, and risk-based explanations that arguably explain conventional momentum cannot explain idiosyncratic momentum.
- The existence of idiosyncratic momentum pro?ts is consistent with the underreaction to news (slow diffusion of information) hypothesis.
- Idiosyncratic momentum forecasts high short and long-term excess returns, while conventional momentum forecasts high short-term and negative long-term excess returns.
- There’s a monotonically decreasing pattern in excess returns, Sharpe ratios, and factor adjusted returns (i.e. alphas) going from high (Decile 1) to low (Decile 10) idiosyncratic and total return momentum portfolios. The D1-D10 idiosyncratic momentum portfolio generates a monthly return of 1.39 percent, somewhat lower than that of total return momentum (1.54 percent), but with a substantially lower volatility. The Sharpe ratio of the idiosyncratic momentum strategy is 0.48 per month, almost double that of conventional momentum (0.25).
- The Fama-French ?ve-factor (beta, size, investment, profitability, and value) model is unable to explain the extreme decile return spreads of the two strategies, and the t-statistic is substantially higher for idiosyncratic than for conventional momentum despite the lower abnormal return.
- A portfolio that