Momentum, contrarian, and the January seasonality
Outline Data and Portfolio Construction Long-term Contrarian Marginal strategies Trading profits Intermediate- and Short-term Momentum
Data and portfolio construction Sample: all of the stocks listed on the NYSE and the AMEX in the monthly files of the CRSP. The sample period is from January 1926 to December 2009. Marginal strategy: Novy-Marx (2011) At the beginning of each month t, NYSE and AMEX stocks are assigned into 10 deciles by single-month returns in month t-x, where x ranges from 1 to 60. The marginal strategy involves buying winners and selling losers. Winner-loser portfolios are reconstructed at the beginning of each month t and held for 1 month.
Data and portfolio construction This paper implements marginal strategy separately in January and outside of January. Jegadeesh and Titman (1993), Grundy and Martin (2001) point out that momentum strategy consistently lose money in January. Conversely, De Bondt and Thaler (1985) show that contrarian profits are concentrated mainly in January. Apart form examining the marginal strategy (return autocorrelation), this paper examine four kinds of strategies and evaluate the trading profits: the overall- year returns, the January returns and the non-January returns for each strategy.
Data and portfolio construction Long-term contrarian strategy: stocks are ranked by compounded returns in months t-60 to t-13 to form 10 deciles at the beginning of each month t. Zero- investment loser-winner portfolios (P10-P1) are reconstructed at the beginning of each month, and held for that month. Short-term momentum strategy: stocks are ranked by compounded returns in months t-6 to t-2 to form 10 deciles at the beginning of each month t. Zero- investment winner-loser portfolios (P10-P1) are reconstructed at the beginning of each month, and held for that month.
Data and portfolio construction Intermediate-term momentum strategy: stocks are ranked by compounded returns in months t-12 to t-7 to form 10 deciles at the beginning of each month t. Zero-investment winner-loser portfolios (P10-P1) are reconstructed at the beginning of each month, and held for that month. Seasonality strategy: assigns stocks into 10 deciles by the returns in month t-12 at the beginning of each month t. Zero-investment winner-loser portfolios (P10-P1) are reconstructed at the beginning of each month, and held for that month.
Long-term contrarian The empirical results of marginal strategies Figure 1 displays the average returns of 48 marginal strategies (t-60~t-13) that buy winners and sell losers. The upper panel of Fig.1 shows the long-term negative autocorrelations at lag 13 until lag 59, excluding multiples of 12 lags. The results appear to be consistent with the findings of De Bondt and Thaler (1985). The middle panel suggests that long-term contrarian is overwhelmingly successful in January. The marginal strategies sort stocks by single-month returns outside of January of the previous 2-5 years, but invest in January, earn sizable and negative returns.
Long-term contrarian The empirical results of trading profits The bottom panel shows that long-term reversal disappear after controlling for the January effect. The non-January returns of the marginal strategies become unreliable and non-persistent. The empirical results of trading profits The profits of the contrarian strategy are decomposed into two components, relating to the January- and non- January investing. Table 1 indicates that a huge portion of contrarian profits materialize in January, which is consistent with the findings in Fig. 1.
Long-term contrarian Given these findings in Section 3.1 and 3.2, the question naturally arises how behavioral stories reconcile with the evidence of the nonexistence of long-term contrarian outside of January. Subperiod checks-robustness
Intermediate- and short-term momentum The empirical results of marginal strategies Figure 2 documents the average returns of 11 marginal strategies (t-12 to t-2) that buy winners and sell losers. The upper panel shows short- and intermediate-term positive return autocorrelation. This findings confirms that recent winners continue to outperform recent losers. The middle panel reports the strong and negative return autocorrelations between January returns and their own lagged returns except for lag 12.
Intermediate- and short-term momentum This finding confirms that the momentum strategy consistently lose money in January. Besides the January momentum loss is particularly pronounced for small firms. The bottom panel shows that non-January returns exhibit stronger return autocorrelations with regard to their own lags than is the case for overall-year returns shown in the upper panel. The strong January annual seasonality creates the illusion of the dominant role of intermediate term prior returns on momentum.
Intermediate- and short-term momentum The empirical results of trading profits For both the short-term momentum and intermediate- term momentum strategies, trading profits are decomposed into two parts relating to the January- and non-January returns. Table 3 reports average monthly returns for intermediate-term (t-12 to t-7), short-term (t-6 to t-2) and seasonality (t-12) strategies. Similar to Novy-Marx (2011), the intermediate –term momentum strategy outperforms its short-term counterpart (0.91% vs. 0.4%; 1.1% vs. 0.62%).
Intermediate- and short-term momentum The second and fifth column of Table 3 report that momentum strategy produces loss of 6.43% (short- term) and loss of 2.39% (intermediate-term) each January. Q: Whether the superior performance of intermediate-term momentum is driven purely by the behavior of January returns. The third and sixth column show the non-January profitability of momentum strategy. There is no significant difference in the non-January returns between the short- and intermediate-term momentum strategy.
Intermediate- and short-term momentum Thus far, the author identified the dominant role of intermediate-term prior returns on momentum in January. Q: Why does the intermediate-term momentum strategy suffer considerably fewer January losses? The January seasonality causes the apparent success of intermediate-term momentum. The January seasonality strategy generates a disproportionately large magnitude of returns relative to the remaining 11 months (3.88% vs. 0.61% and 2.72% vs. 0.64%).
Intermediate- and short-term momentum The January seasonality strategy generates a disproportionately large magnitude of returns relative to the remaining 11 months (3.88% vs. 0.61% and 2.72% vs. 0.64%). Table 4 shows the average market capitalization (size) of relative strength portfolios (P1~P10) for each strategy. In January, winners tend to be very extremely small firms, whereas losers tend to be small firms. The considerable profitability of January seasonality is caused by extremely small firms outperforming small firms in January.
Intermediate- and short-term momentum The short-term momentum strategy buy small firms and sell extremely small firms (the third column). The occurrence of small firms underperforming extremely small firms in January eventually leads to substantially losses. The intermediate-term momentum strategy also buy small firms and sell extremely small firms (the fifth column). However, the January seasonality results in the intermediate-term winner being smaller firms than the short-term winner (608 vs. 976), while the intermediate-term losers are larger firms than the short-term losers (359 vs. 233).
Intermediate- and short-term momentum Thus, the intermediate-term winners underperform the intermediate-term losers to a lesser extent than the short-term winners underperform the short-term losers. The intermediate-term momentum strategy experiences less January loss than short-term strategy. Subperiod checks-robustness