Do assets other than individual stocks exhibit a short-term (1-month) reversal effect? In their February 2019 paper entitled “Short-Term Momentum (Almost) Everywhere”, Adam Zaremba, Andreas Karathanasopoulos and Huaigang Long investigate short-term return predictability within long run global samples spanning five asset classes: equity indexes, government bonds, treasury bills, commodity futures and currencies. Each month they sort assets by class or overall into fifths (quintiles) on prior-month return. For classes with at least 10 assets available, they then construct long-short hedge portfolios that are long (short) the equal-weighted quintile of assets with the highest (lowest) prior-month returns. Using monthly returns for 45 equity indexes, 54 government bonds, 52 government bills, 48 commodity futures and 62 currency exchange rates in U.S. dollars as available during 1800 through 2018, they find that:
- The number of assets available overall increases from 16 in January 1800 to over 250 in recent years. Within asset classes, tests start in January 1934 for equity indexes, October 1822 for government bonds, June 1848 for government bills, February 1961 for commodity futures and July 1818 for currencies.
- Contrary to stock-level research, all class-level series except government bonds exhibit short-term momentum. There is no evidence of short-term reversal. Specifically:
- For government bonds, the hedge portfolio generates average monthly gross return -0.01%.
- For other classes, hedge portfolio average monthly gross returns range from 0.68% for government bills to 2.37% for commodity futures, with corresponding annualized gross Sharpe ratios ranging from 0.58 to 1.14.
- The equally weighted combination of asset class hedge portfolios generates average monthly gross return 0.84%, with annualized gross Sharpe ratio 0.84.
- A hedge portfolio formed from all assets treated individually generates average monthly gross return 1.30%, with annualized gross Sharpe ratio 1.08.
- This short-term momentum effect mostly remains significant after controlling for asset class beta, idiosyncratic volatility, standard (intermediate-term) momentum, value, skewness and seasonality factors. However, exposure to standard momentum, value and seasonality factors capture short-term momentum for currencies.
- The short-term momentum is strongest among assets with high idiosyncratic volatility and during times of elevated return dispersion.
- Pairwise correlations of asset class short-term momentum returns are positive but mostly weak.
- Findings are generally robust across subperiods, market states, calendar months, alternative implementation techniques and alternative datasets. However, there is no evidence of short-run momentum pre-1880.
In summary, evidence from long series indicates that class-level assets mostly exhibit short-term (1-month) return momentum and not short-term reversal.
Cautions regarding findings include:
- Findings are gross, not net. Accounting for monthly portfolio reformation costs would reduce all returns. There may not have been convenient/liquid vehicles for holding these assets during much of respective sample periods, making these costs high.
- As noted in the paper, “shorting many positions in our asset universe was difficult or even impossible, and …we do not consider the many costs that investors would face, such as borrowing or storage costs.”
- As noted in the paper, findings may impound material survivorship bias via omission of assets that were important in the past but are no longer tracked in databases.
- As noted in the paper, early data may be of poor quality. To mitigate, the authors “discard all the monthly observations with returns of less than -98% or exceeding 500% [and] drop all the zero-return months.”
