Understanding momentum in stock returns

(MoneyWatch) In 1993, Eugene Fama and Ken French released a paper explaining that stock returns are primarily due to three factors: market risk, size risk and value risk. However, this Fama-French three-factor model doesn't account for all stock returns, missing additional factors such as momentum. A recent paper sheds more light on how the momentum factor affects stock prices.

Momentum strategies have been highly profitable. For the period January 1927 through April 2013, the average annual return has been about 8.3 percent per year. However, momentum has been subject to occasional large crashes, such as losing 76 percent from March through September 2009.

• How much of your bond fund is actually in stocks?
• Trust models over instincts
• Why do so many investors ignore the evidence?

Dong Lou and Christopher Polk, professors at the London School of Economics, approached the issue of momentum in a new and unique way. They argue that when momentum strategies are the most crowded by arbitrage capital, those strategies are the least profitable and long-horizon abnormal returns are the most negative. They measured the extent of momentum crowding by the past degree of abnormally high correlation among those stocks on which a momentum arbitrageur would speculate. Their premise was that "when arbitrageurs take long positions in winner stocks and short positions in loser stocks, such momentum trades can have simultaneous, temporary price impacts on all momentum stocks and thus cause return comovement among these stocks." They called this measure comomentum.

Their assumption was that when comomentum is relatively low (meaning momentum strategies aren't crowded), abnormal returns to a standard momentum strategy should be positive and not revert. Conversely, when comomentum is relatively high, momentum strategies are crowded and abnormal returns on a standard momentum strategy should be low. In addition, when the momentum trade gets crowded, it can be destabilizing, resulting in subsequent reversal of the initial momentum returns. The following is a summary of their findings.

• Stock comomentum strongly and positively forecasts stock returns in the following months.
• When comomentum is relatively high, the long-run buy-and-hold returns to a momentum strategy are negative, consistent with times of relatively high amounts of arbitrage capital pushing prices farther away from fundamentals.
• International data are consistent with the U.S. momentum-predictability findings. In every one of the 19 largest non-U.S. stock markets that were examined, comomentum is negatively associated with subsequent profits from a standard momentum trading strategy.
• Country comomentum measures tend to move together over time, with an average pairwise correlation of 0.47, indicating times when global arbitrage capital is generally high or low.
• The results are present only for stocks with high institutional ownership.
• A strategy that invests only in momentum in those countries with low arbitrage capital and hedges out exposure to global market, size and value factors earns 18 percent per year, with a very high level of statistical significance (t-statistic of 6).

The authors also examined momentum-timing strategies of mutual and hedge funds and found that the typical long-short equity hedge fund decreases their exposure to the momentum factor when comomentum is relatively high. However, the ability of hedge funds to time momentum is decreasing in the size of the fund's assets under management -- large funds are unable to time a momentum strategy as easily as small funds.

This paper was presented at the National Bureau of Economic Research in October. The following are some insights (from the leading academics) from the discussion that followed. I want to thank University of Chicago professor Tobias Moskowitz for sharing them with me.

The empirical findings aren't that strong. The t-stats are barely above 2 for the main effect and, perhaps more troubling, there's no consistent pattern. When ranking on their comomentum (COMOM) factor, there are significantly weaker momentum (MOM) returns only for the highest quintile. This indicates the results hinge on the extremes and only the extremes, which is worrisome.

There's no reliable relation between COMOM and MOM returns after 1993. This is strange, since it doesn't seem likely that there was much capital chasing MOM before then. This is pretty devastating to the story of arbitrage capital determining the success of momentum strategy returns. It appears that something else is likely going on or the result is spurious.

The most interesting part of the discussion was when an alternative explanation for the results was offered. The issue is complex, but basically it has to do with the fact that MOM crashes are only about the losers rebounding suddenly after prolonged market downturns. When COMOM was low, losers had low market betas and/or the market didn't suddenly rise, and the losers continued to do poorly, making shorting them profitable. However, when COMOM and market betas of losers were high and the market suddenly went into an upswing, shorting the losers created big losses (momentum crashes). In other words, it's not a crowding effect at work, but the fact that market betas change over time. When the COMOM measure is high, the risk of a momentum crash is high (because the market betas of the losers is also high). This alternative explanation makes more sense, given the pre-1993 results.

Finally, it's important to note that momentum crashes are about only the short side. Long-only momentum doesn't suffer from these episodes.

Image courtesy of taxbrackets.org