Do anomaly portfolios that are long (short) the tenth, or decile, of stocks with the highest (lowest) expected value-weighted returns based on some firm accounting variable or stock behavior really work on a net basis? In the May 2019 version of their paper entitled “Shorting Costs and Profitability of Long-Short Strategies”, Dongcheol Kim and Byeung Joo Lee examine profitability of such portfolios after adjusting for: (1) unavailability of stocks to borrow for shorting as indicated; and, (2) stock loan fees paid to share lenders. They consider 14 value-weighted anomalies based on: return on assets, return on equity (ROE), momentum, net operating assets, investment-to-asset ratio, abnormal capital investment, accruals, asset growth, net stock issuance, composite equity issues, O-score, failure probability, gross profit and post-earnings announcement drift. They do not consider trading frictions (broker fees, bid-ask spread, impact of trading) incurred due to periodic reformation of anomaly portfolios. Using monthly stock prices and returns, data to construct value-weighted long-short anomaly portfolios, and share loan availability and fee data from Markit for a broad sample of U.S. stocks priced at least $1 during January 2006 through December 2017, they find that:
- Average gross monthly return of the 14 long-short anomaly portfolios is 0.262%.
- Average monthly shorting costs across anomalies are 0.029% and 0.075% for the two sources listed above, respectively. The total monthly shorting cost of 0.104% is about 40% of average gross anomaly monthly return.
- Shorting costs represent about 25% of widely used 3-factor model (adjusting for market, size and book-to-market factors) alpha and 80% of a 4-factor model (adjusting for market, size, investment-to-asset ratio and ROE) alpha.
- Incorporation of trading frictions for periodic reformation of anomaly portfolios would further reduce profitability.
In summary, evidence from actual shorting costs/constraints undermines belief in net profitability of long-short stock anomaly strategies.
For related research, see:
- “Lendable Share Supply a Roadblock to Shorting Strategies?”
- “Constrained Shorting and Factor Investing”
- “Stock Anomaly Short Side Costs Manageable?”
Frequently Asked Questions
What is a long-short stock anomaly strategy?
A long-short stock anomaly strategy buys the decile of stocks with the highest expected returns and shorts the decile with the lowest, based on a firm accounting variable or stock behavior such as momentum, return on assets, accruals, or asset growth. The study examined 14 value-weighted anomalies, including net stock issuance, O-score, failure probability, gross profit, and post-earnings announcement drift.
Do long-short stock anomaly strategies still make money after shorting costs?
Long-short stock anomaly strategies earn substantially less after shorting costs, undermining belief in their net profitability. Across 14 anomalies from January 2006 to December 2017, the average gross monthly return of 0.262% was reduced by total shorting costs of 0.104%, about 40% of gross return, leaving roughly 0.158%.
How are shorting costs measured for stock anomaly portfolios?
Shorting costs are measured using Markit share loan data on two sources: unavailability of stocks to borrow for shorting and stock loan fees paid to share lenders. Across the 14 value-weighted anomaly portfolios, these averaged 0.029% and 0.075% per month respectively, using U.S. stocks priced at least $1 from January 2006 through December 2017.
How much of a stock anomaly's alpha do shorting costs eliminate in factor models like the 3-factor and 4-factor?
Shorting costs consume about 25% of the alpha from the widely used 3-factor model (market, size and book-to-market) and about 80% of the alpha from a 4-factor model (market, size, investment-to-asset ratio and ROE). This means anomaly alpha is far more vulnerable to shorting costs under the 4-factor model than under the 3-factor model.
What costs are left out when estimating the profitability of long-short anomaly strategies?
The study accounts only for shorting costs and does not include trading frictions such as broker fees, bid-ask spreads, and price impact from periodically reforming anomaly portfolios. Including these frictions would further reduce profitability, so the 0.158% average net monthly return likely overstates realistic results. The findings also cover only 14 value-weighted anomalies in U.S. stocks from 2006 to 2017.