What conditions foretell stock momentum strategy crashes? In their October 2014 paper entitled “Momentum Trading, Return Chasing, and Predictable Crashes”, Benjamin Chabot, Eric Ghysels and Ravi Jagannathan examine stock momentum strategy performance for both widely used historical U.S. data (starting in 1926 through 2012) and for a hand-collected sample of stocks listed on the London Stock Exchange during 1866 to 1907. They consider two methods of measuring momentum strategy returns. One is the gross return to the Fama-French momentum factor portfolio. The other is the gross return to a portfolio that is each month long (short) the value-weighted 30% of stocks with the highest (lowest) returns per the Fama-French momentum decile portfolios. Both methods define momentum conventionally as the return from 12 months ago to one month ago, with a skip-month before portfolio formation to avoid short-term reversal. They focus on conditions that precede momentum strategy crashes based on a model that considers three factors: (1) the risk-free rate; (2) past stock market return; and, (3) past momentum strategy return. Using the specified stock return data sets, they find that:
- The momentum strategy generates statistically significant gross three-factor (adjusting for market, size and book-to-market ratio factors) monthly alphas in both data sets.
- For the 1926-2012 U.S. sample, average gross monthly returns for the two momentum measurement methods are 0.69% and 0.55%, with gross annualized Sharpe ratios 0.50 and 0.37. Gross monthly three-factor alphas are 1.0% and 0.88%.
- For the 1866-1907 UK sample, average gross monthly returns for the two momentum measurement methods are 0.30% and 0.27%, with gross annualized Sharpe ratios 0.48 and 0.37. Gross monthly three-factor alphas are 0.52% and 0.49%.
- However, the momentum strategy exposes investors to crashes during both sample periods.
- For the 1926-2012 U.S. sample, the two momentum measurement methods have 63 and 68 declines of at least 5% (on average, one every 15 and 16 months). The average loss during these declines is 14% and 15%.
- For the 1866-1907 UK sample, declines of at least 5% occur every 25 and 28 months on average for the two momentum measurement methods.These declines tend to be slower and smaller than those in the later U.S. sample.
- Momentum strategy crashes are more likely when borrowing is easy and/or return chasing is likely, thereby attracting capital and extinguishing the anomaly. This general finding translates into conditions when:
- Momentum recently performed well (evidence found in both samples).
- Interest rates are low (evidence observed only in the 1867-1907 sample).
- Momentum recently outperforms the overall stock market (evidence observed only in the 1926-2012 sample). In the U.S. sample, the probability of a momentum crash increases dramatically when the prior-year strategy return is in the top tenth of annual strategy returns.
- The times when momentum crash risk is highest are exactly the times when a momentum fund manager has a strong recent performance and finds it easy to attract funds from return-chasing investors.
In summary, evidence suggests the conventional stock momentum strategy inherently produces large, infrequent crashes based on the changing availability of capital to momentum traders.
Said differently, investors employing a conventional stock price momentum strategy should become cautious when leverage is easy and/or the strategy has recently performed well.
Cautions regarding findings include:
- Reported momentum strategy performance is gross, not net. Incorporation of trading frictions and shorting costs (which vary widely over the two sample periods) would reduce performance and could affect findings. Monthly portfolio reformation based on momentum tends to drive relatively high portfolio turnover.
- Shorting of some past losers may not be consistently feasible due to high shorting costs/lack of counterparties.
- Funds designed to exploit stock momentum would involve additional frictions (fees) for investors.
- The study does not investigate any strategy adjustments to avoid crashes.
