The Profitability of Momentum Strategies

Momentum strategies based on continuations in stock prices have attracted a wide following among money managers and investors. We evaluated the profitability of price momentum strategies based on past return and earnings momentum strategies based on standardized unexpected earnings and revisions of consensus forecasts. The strategies proved to be profitable for intermediate horizons. Chasing momentum can generate high turnover, however; hence, implementation of momentum strategies requires a focus on managing trading costs. Comparing the strategies yielded evidence that they reflect distinct phenomena and provided information about the sources of profits. The results indicate that the market is slow to incorporate the full impact of information in its valuations. Momentum strategies that are based on continuations in stock prices over intermediate horizons have attracted a wide following among money managers and investors. In evaluating the usefulness of momentum strategies, measuring their profitability and also understanding why they work are important. In the absence of a reasonable explanation for the profitability of momentum strategies, the pattern of profits observed in the past may be a statistical fluke, in which case, it may not recur.Using a sample that included all U.S. stocks from the NYSE, Amex, and Nasdaq for 1973–1993, we evaluated the profitability of strategies based on price momentum (past return) and on earnings momentum (standardized unexpected earnings and revisions in consensus forecasts).We found that strategies based on price momentum and earnings momentum yielded significant profits over a 6–12-month horizon. For example, sorting stocks by prior six-month return yielded spreads in returns of 15.4 percentage points (pps) over the subsequent year. Similarly, ranking stocks by past earnings momentum produced large differences in future returns. The spreads in the subsequent year were 7.5 pps for portfolios based on standardized unexpected earnings and 9.7 pps based on a moving average of past revisions in consensus estimates of earnings.A critical comparison of these strategies yielded evidence that they reflect distinct phenomena. None of the momentum variables subsumed any of the others; they each exploited different pieces of information. In fact, combinations of these momentum strategies yielded higher profits than any of the strategies individually.The results presented are fairly robust. The strategies continued to generate profits when applied to a sample of large stocks only, although the spreads were lower than for the full sample. Furthermore, when we extended our results to the 1994–98 period, our main results held.These results indicate that the market is slow to incorporate the full impact of information in its valuations. For instance, the market seems to be pleasantly surprised around future earnings announcements for past winners and vice versa for past losers. In fact, a substantial portion of the momentum effect is concentrated around subsequent earnings announcements. Generally, if the market is surprised by good or bad earnings news, then on average, the market continues to be surprised in the same direction, at least over the next two subsequent announcements. Another piece of evidence compatible with a sluggish response is that analysts continue to revise their earnings forecasts downward for past losers for up to 12 months after the ranking period.We considered holding periods of six months and one year for the momentum strategies. Because of the resulting high turnover and because of potentially high transaction costs for some of the small, less liquid stocks in the winner and loser portfolios, practical implementation of momentum strategies will require skillful management of trade execution.