Wednesday, March 17, 2010

“Short-term” reversals

In a publicly available paper “Diversification Across Characteristics” Erik Hjalmarsson extends the 2009 work of Asness, Moskowitz, and Pedersen. He studies the performance of long-short strategies based on seven reversal/momentum and value characteristics: short-term reversals, medium-term momentum, long-term reversals, book-to-market value, cashflow-price ratio, earnings-price ratio, and size. He builds seven single-characteristic portfolios and then compares their performance to that of an equal-weighted portfolio of the single-characteristic ones. The equal-weighted diversified portfolio, he finds, “almost always deliver[s] substantially better Sharpe ratios than any of the single-characteristic portfolios or the two-characteristic momentum-value portfolio considered by” Asness, Moskowitz, and Pedersen.

The three technical characteristics Hjalmarsson uses are short-term reversals (the prior month’s [t-1] return), medium-term momentum (returns from month t-12 to t-2), and long-term reversals (returns from month t-60 to t-13). For the medium-term momentum portfolio returns are calculated by taking the difference between the returns on the top decile of the universe of stocks and the bottom decile; for the short-term and long-term reversal portfolios returns are calculated by taking the difference between the returns on the bottom decile and the top decile.

Although Hjalmarsson touts the risk-adjusted strength of the portfolio of portfolios, I want to look at the short-term reversal single-characteristic portfolio since it delivered the best non-risk-adjusted returns. Throughout the entire period of the study, July 1951 through December 2008, it had an annualized mean return of 23.267% whereas the market returned only 6.104%. By comparison, medium-term momentum returned 14.956%, long-term reversals 9.314%, book-to-market value 11.651%, cashflow-price ratio 9.766%, earnings-price ratio 8.839%, firm size 5.145%, and the equal-weighted portfolio 11.849%. Its Sharpe ratio for the entire period (1.446) was also substantially higher than that of any other single-characteristic portfolio though lower than that of the equal-weighted portfolio (1.666). When the entire study period was halved, the short-term reversal portfolio once again trumped all the other individual strategies both in mean return and in Sharpe ratio. During the period July 1951 through December 1979 its Sharpe ratio was even higher than that of the equal-weighted portfolio. And during the difficult period of January 1999 through December 2008 its mean return (13.634% as opposed to a negative market return of 2.149%) beat all others, though no longer nearly so decisively. Its Sharpe ratio, however, dropped from first place among the seven strategies to sixth place; only the medium-term momentum portfolio had a lower Sharpe ratio. From a high of 2.013 during the 1951-1979 period it fell to 0.499 during 1999-2008. By contrast, the Sharpe ratio of the equal-weighted portfolio, which had an overall value of 1.666 and values of 1.658 and 1.672 when the study period was halved, fell to only 1.316 during the January 199 through December 2008 period.

The findings of Hjalmarsson’s paper are important for those concerned with portfolio management as well as those engaged in active trading. He clearly documents that over time the “short-term” mean reversion strategy has outperformed the “medium-term” momentum strategy. Of course, when developing a trading system for an account without the capital to buy and sell one-fifth of all the stocks on the NYSE, AMEX, and NASDAQ much more work has to be done—for starters, selecting trading vehicles, determining optimal timeframes, and developing a switching mechanism to adjust to changing market conditions.

By the way, Hjalmarsson makes use of the monthly return data from Kenneth French’s (of Fama and French fame) Dartmouth website. His data library, downloadable as text files, is definitely worth a look.

No comments:

Post a Comment