The Good News in Short Interest: 9/10

Investment Potential Rating: 9/10 (1 worst, 10 best)

============

The Good News in Short Interest

Ekkehart Boehmer
University of Oregon - Charles H. Lundquist School of Business
Zsuzsa R. Huszar
NUS Business School - Finance Department; California Polytechnic State University
Bradford D. Jordan
University of Kentucky - Gatton College of Business and Economics

Journal of Financial Economics, forthcoming

http://ssrn.com/abstract=1405511

Abstract:
We study the information content in monthly short interest using NYSE-, AMEX-, and NASDAQ-listed stocks from 1988 to 2005. We show that stocks with relatively high short interest subsequently experience negative abnormal returns, but the effect can be transient and of debatable economic significance. In contrast, we find that relatively heavily traded stocks with low short interest experience both statistically and economically significant positive abnormal returns. These positive returns are often larger (in absolute value) than the negative returns observed for heavily shorted stocks. Because stocks with greater short interest are priced more accurately, our results suggest that short selling promotes market efficiency. However, we show that positive information associated with low short interest, which is publicly available, is only slowly incorporated into prices, which raises a broader market efficiency issue. Our results also cast doubt on existing theories of the impact of short sale constraints.

Data Source:
The data on short interest come from the exchanges themselves. Although the authors don’t specify, I assume they merely downloaded csv files from the exchanges’ websites, each of which provide this data for free. (There are also many services that aggregate these data.) Other relevant data used in the study come from CRSP. The period tested is 1988-2005.

Data Specification:
Not surprisingly, a lot of research has come out recently on the topic of short selling. Traditionally, theory has held that constraints on short selling prevent the impounding of negative information in to stock prices and, since there is no similar constraint to going long, more positive information is impounded in to prices than negative. Thus, constraints on short selling impede market efficiency (and regulators would be foolish to try to limit its use, but we can save that discussion for another day).

This paper is about the performance of heavily shorted stocks versus lightly or un-shorted stocks. Market pundits usually assume heavy short interest is a bearish indicator because it signals that many investors anticipate the price to decline. But there are other interpretations; One is that shorting is usually used for hedging or arbitrage and thus says nothing about the valuation of the stock. Another is that shorting implies future buying because all short sales will need to be covered at some point in the future, therefore it may in fact be a bullish indicator. On balance, though, the academic literature has shown support for the conventional wisdom and found underperformance of heavily shorted stocks.

This paper is one of the first to actually compare the difference between high-short interest stocks and low-short interest stocks. The results are somewhat surprising.

Using the short interest data of the prior month, the authors rank all stocks based on the short interest ratio, which is the ratio of total shares shorted divided by shares outstanding, then observe the performance of those stocks in the subsequent month. There are a lot of stocks with zero short interest, about 15% of them, so when the authors to group stocks based on short interest ratio they have a lot more stocks in the 1st percentile portfolio than any of the others.

Results:
The results may surprise you. Heavily shorted stocks do underperform, but the outperformance of lightly or un-shorted stocks is far more significant. Hmm.

Stocks in the 99th percentile portfolio (the highest short interest stocks) on average experience negative absolute returns of 0.1% per month. The stocks in the 1% portfolio (the lowest short interest stocks) on average experience positive absolute returns of 2.1% per month. This is an impressive result.

After controlling for various risk factors through multifactor regressions, the authors find that the 99% portfolio produces -1.2% alpha monthly, and the 1% portfolio produces +1.4% alpha monthly. This means that a theoretical long/short portfolio adjusted for risk could produce 2.6% per month, or 36% annually!

In the full sample that the authors test, there are 45 stocks in this 99th percentile group, and 232 in the 1st percentile, so it the investor should be able to diversify himself fairly well.

Investment Strategy:
Each month, rank all stocks based on short interest ratio and assign each a percentile rank. Buy stocks in the lowest percentile and short stocks in the highest percentile. Rebalance monthly.

OR

If you’re not the shorting type, just rank the stocks as above but simply go long the stocks in the lowest percentile and rebalance monthly. Even doing this provides pretty good alpha.

Implementation Issues and Remarks:
It is nice to see a Wall Street rule of thumb formalized by academic work. I think this is what we can see in this paper. However it is important to note that this paper does not account for some very obvious costs, most noticeably the margin costs associated with shorting, but also the mere transactions costs of rebalancing a large portfolio so frequently. Nevertheless, I do feel the paper provides convincing evidence that following short interest data is not a fruitless exercise.

Another risk to investors is that short selling will be outlawed entirely. I would guess there is a nonzero probability of this happening sometime in the foreseeable future, and if it does clearly this strategy is worthless.

There is also reason to believe that the results of this paper persist for periods well beyond one month. Figure 3 of the paper charts the cumulative returns in months t=0 through 5 for stocks sorted in month t-1. In other words, it looks like we could hold a portfolio for a period of up to six months and still generate some good returns, although the authors do not fully test such a strategy in this paper.

Investment Potential Rating: 9/10