08. Winners and Losers in Momentum Investing

The Formation Process of Winners and Losers in Momentum Investing

Abstract:
Previous studies have focused on which stocks are winners or losers but have paid little attention to the formation process of past returns. This paper develops a model showing that past returns and the formation process of past returns have a joint effect on future expected returns. The empirical evidence shows that the zero-investment portfolio, including stocks with specific patterns of historical prices, improves monthly momentum profit by 59%. Overall, the process of how one stock becomes a winner or loser can further distinguish the best and worst stocks in a group of winners or losers.

Notes

p. 3: Intermediate-term (3–12 months) momentum has been documented by Jegadeesh and Titman (1993, 2001, hereafter JT), while short-term (weekly) and long-term (3–5 years) reversals have been documented by Lehmann (1990) and Jegadeesh (1990) and by DeBondt and Thaler (1985), respectively. Various models and theories have been proposed to explain the coexistence of intermediate-term momentum and long-term reversal. However, most studies have focused primarily on which stocks are winners or losers; they have paid little attention to how those stocks become winners or losers. This paper develops a model to analyze whether the movement of historical prices is related to future expected returns.

p. 4: This paper captures the idea that past returns and the formation process of past returns have a joint effect on future expected returns. We argue that how one stock becomes a winner or loser—that is, the movement of historical prices—plays an important role in momentum investing. Using a polynomial quadratic model to approximate the nonlinear pattern of historical prices, the model shows that as long as two stocks share the same return over the past n-month, the future expected return of the stock whose historical prices are convex shaped is not lower than one whose historical prices are concave shaped. In other words, when there are two winner (or loser) stocks, the one with convex-shaped historical prices will possess higher future expected returns than the one with concave-shaped historical prices.

p. 4: To test the model empirically, we regress previous daily prices in the ranking period on an ordinal time variable and the square of the ordinal time variable for each stock. The coefficient of the square of the ordinal time variable is denoted as gamma.