Is optimal stock anomaly exploitation long-only or long-short? If not long-short, does shorting the market rather than individual stocks work as well as shorting individual stocks? In his November 2017 paper entitled “How Do Short Selling Costs and Restrictions Affect the Profitability of Stock Anomalies?”, Filip Bekjarovski explores effects of short selling costs and constraints on the viability of exploiting seven U.S. stock anomalies: size, value, profitability, investment, momentum, accruals and net issuance. He constructs all anomaly portfolios via market capitalization weighting of stocks sorted into tenths (deciles). He measures portfolio alphas relative to the market excess return (1-factor). He considers long-only (long the top decile), conventional long-short (long the top and short the bottom deciles) and hybrid long-short (long the two highest alpha deciles, tilted toward the highest, while short the market). Anomaly portfolio rebalancing is annual for all except momentum (monthly). He analyzes effects of shorting costs based on an April 2017 proprietary snapshot of institutional stock borrowing fees. He specifically estimates a shorting cost threshold above which investors should switch between long-short and hybrid long-short exploitation methods. Using the stock borrowing fee snapshot and data required to construct seven anomalies from a broad sample of U.S. common stocks during July 1963 through December 2016, he finds that:
- Long-only anomaly gross alphas are large and statistically significant. Equal allocations to long-only anomaly portfolios generates a monthly gross Sharpe ratio 32% higher than that of the market.
- Equal allocations to hybrid long-short anomaly portfolios generates a monthly gross Sharpe ratio 40% higher than that for equal allocations to long-only allocations.
- Ignoring shorting costs, short sides of conventional long-short anomaly portfolios generate 63% of long-short profitability. Equal allocations to conventional long-short anomaly portfolios, ignoring shorting costs, generates a monthly gross Sharpe ratio 24% higher than that for equal allocations to hybrid long-short portfolios.
- For NYSE decile break points, capitalization-weighted (equal-weighted) stock annual borrowing cost is 0.46% (4.16%). This large difference derives largely from the hardest-to-borrow 37.3% of stocks that account for just 2.9% of market capitalization.
- Costs associated with capitalization-weighted short sides of conventional long-short anomaly portfolios based on NYSE decile break points are therefore small compared to their profitability contributions:
- Stock borrowing costs are on average only 15.4% of capitalization-weighted anomaly short side alpha.
- Highest annual capitalization-weighted borrowing costs are for low-profit (1.16%) and low-momentum (1.10%) short sides, but these costs are small fractions of their 5.04% and 11.2% respective annual gross alphas.
- For the other five anomalies, annual capitalization-weighted short side borrowing costs are less than 0.65%.
- Moreover, portfolios restricted to stocks that are easy to short have large and significant short anomaly alphas.
- However, average annual anomaly shorting cost for equal weighting and full-sample decile break points is 9.74%. Long-short anomaly portfolios based on this approach have higher gross alphas, but only on long sides.
- Hybrid long-short as specified above is more profitable than conventional long-short when annual borrowing costs for individual stocks exceed about 3%.
In summary, evidence supports belief that long-short stock anomaly exploitation is feasible (gross of portfolio reformation frictions) for capitalization weighting of anomaly decile portfolios based on NYSE break points.
Cautions regarding findings include:
- Analyses are gross of portfolio rebalancing frictions, which would reduce profitability of all anomaly portfolios, affecting momentum most dramatically. Annual rebalancing for other anomalies mitigates this concern.
- The April 2017 snapshot of institutional stock borrowing costs may not be representative of shorting costs/constraints throughout the sample period.
- Individuals may bear materially higher stock borrowing costs than used in the study.
- The hybrid long-short approach is complex enough to suspect some data snooping bias in its specification.
See also:
