Short Interest And Aggregate Stock Returns by SSRN
Saint Louis University – John Cook School of Business
Washington University in Saint Louis – Olin Business School
Guofu Zhou
Washington University in St. Louis – Olin School of Business
February 27, 2015
WFA – Center for Finance and Accounting Research Working Paper No. 14/002
Abstract:
We show that short interest is arguably the strongest known predictor of aggregate stock returns. It outperforms a host of popular return predictors both in sample and out of sample, with annual R-squared statistics of 13% and 11%, respectively. In addition, short interest can generate utility gains of over 300 basis points per annum for a mean-variance investor. A vector auto-regression decomposition shows that the economic source of short interest’s predictive power stems predominantly from a cash flow channel. Overall, our evidence indicates that short sellers are informed traders who anticipate future aggregate cash flows and associated market returns.
Short Interest And Aggregate Stock Returns – Introduction
The equity market risk premium impacts many fundamental areas of finance, from portfolio theory to capital budgeting. Accordingly, a voluminous literature attempts to predict changes in future aggregate excess stock returns.1 In this paper, we show that short interest, aggregated across securities, is arguably the strongest predictor of the equity risk premium identified to date. Short interest outperforms a host of popular return predictors from the literature in both in-sample and out-of-sample tests. Short interest also generates substantial utility gains and Sharpe ratios that exceed those provided by popular predictors. Moreover, we show that the ability of short interest to predict future market returns stems predominantly from a cash flow channel. Taken together, our results suggest that short sellers are informed traders who are able to anticipate changes in future
aggregate cash flows and associated changes in future market returns.
We begin by constructing a long monthly time series of aggregate short interest spanning the period 1973 to 2013. Each month, using data recently made available from Compustat, we calculate the log of the equal-weighted mean of short interest (as a percentage of shares outstanding) across all publicly listed stocks on U.S. exchanges. The resulting series constitutes a measure of total short selling in the economy. The short interest series, which is plotted in Panel A of Figure 1, displays a strong upward trend over our sample period. Much of the upward trend is likely due to the continued development of the equity lending market, which has made it easier to short sell over time, as well as the large increase in the number of hedge funds in existence, which has led to a sharp increase in the amount of capital devoted to short arbitrage. This upward trend obscures the true information content in aggregate short interest. Accordingly, we detrend the short interest series to capture the variation in short interest that is due to changes in the beliefs of short sellers, and not simply secular changes in equity lending conditions or the amount of capital devoted to short arbitrage. We standardize the detrended series to create a short interest index (SII, hereafter) that can be viewed as a measure of market pessimism based on short interest data.
If short interest does contain information about future market returns, we would expect higher values of SII to predict lower future returns. Indeed, in-sample tests show that a one-standard-deviation increase in SII corresponds to a six to seven percentage point decrease in future annualized excess returns. SII produces predictive regression R2 statistics of 1.34% at the monthly horizon and 12.67% at the annual horizon. We also compare the predictive power of SII to that of 14 popular predictor variables from Goyal and Welch (2008). SII substantially outperforms all of the popular predictors at quarterly, semi-annual, and annual horizons and performs similarly or better than all of the predictors at the monthly horizon. Furthermore, the predictive power of SII is robust to concerns about persistence based on the recently developed test of Kostakis, Magdalinos, and Stamatogiannis (2015).
Goyal andWelch (2008) show that, despite significant evidence of in-sample predictive ability, popular predictor variables fail to predict the equity risk premium based on out-of-sample tests.
Consequently, we also examine the out-of-sample predictive ability of SII.2 We find positive out-of-sample R2 statistics (Campbell and Thompson, 2008) of 1.94%, 6.33%, 10.95%, and 10.94% at horizons of one, three, six, and twelve months, respectively, which are statistically significant and larger than those for all of the popular predictor variables considered by Goyal and Welch (2008). Using encompassing tests, we also show that forecasts based on SII have superior information content relative to forecasts based on the popular predictors.
See full PDF below.
