short selling and common stock prices

CFA Institute

Short Selling and Common Stock PricesAuthor(s): J. Randall Woolridge and Amy DickinsonSource: Financial Analysts Journal, Vol. 50, No. 1 (Jan. - Feb., 1994), pp. 20-28Published by: CFA InstituteStable URL: http://www.jstor.org/stable/4479709 .

Accessed: 13/06/2014 19:08

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

CFA Institute is collaborating with JSTOR to digitize, preserve and extend access to Financial AnalystsJournal.

http://www.jstor.org

This content downloaded from 91.229.248.187 on Fri, 13 Jun 2014 19:08:09 PMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=cfa

http://www.jstor.org/stable/4479709?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


Short Selling and Conmmon Stock PnCes

J. Randall Woolridge and Amy Dickinson

An analysis of overall market data and of individual companies traded on the NYSE, Amex and OTC markets indicates a positive but statistically insignificant relation between changes in short position and stock prices (even after accounting for market movements and risk). Short sellers, on average, do not earn abnormal returns. Furthermore, the relation between short selling and stock price does not seem to be materially affected by whether the stock is traded on the NYSE or Amex or over the counter.

The results provide strong evidence refuting the popular notion that short sellers earn abnormal profits at the expense of less informed traders. Short sellers do not appear to drive prices down through their short sales. In fact, if anything, short sellers provide market liquidity by shorting into up markets and reducing short positions in down markets.

While short selling has been an accepted form of trading in markets for centuries, it has come under

increased scrutiny by regulators and the media in recent years as some market observers have blamed it for increased market volatility and declines in both the overall market and individual security prices. This article evaluates the short sales-stock price relationship, focus- ing on three issues related to short selling:

* Do short sales, in and of themselves, affect security prices?

* Is a high level of short interest a bullish or bearish indicator for stock prices?

* Do short sellers, on average, earn abnormal returns?

These propositions are tested in two phases. Ini- tially, to assess the overall short selling and security return relationship, we evaluate monthly short sale and stock return data for the NYSE, Amex and NASDAQ markets. Then we analyze monthly short sale data and stock returns for individual companies using a random sample of 100 NYSE, Amex and NASDAQ companies. The study covers the 1986-91 period.

Overall, the results provide strong refutation of the popular notion that short sellers earn abnormal profits at the expense of less informed investors by artificially driving down stock prices. We find, on an aggregate market level as well as the individual company level, a positive relation between monthly short-selling activity and stock returns. The positive relation between short selling and returns for individual companies holds when

J. Randall Woolridge is Professor of Finance and the Goldman, Sachs & Co. and Frank P. Smeal Endowed University Fellow at the Smeal College of Business Administration at Pennsylvania State University. Amy Dickinson is Assistant Professor of Finance at Florida Atlantic University in Boca Raton.

both short ratios and short interest are regressed against raw as well as market and risk-adjusted returns. While these findings do not imply a causal relation between short-position changes and stock returns, they do indicate that, on average and on a month-to-month basis, short sellers are selling as stock prices are going up and reducing short positions when stock prices are declin- ing. Other tests show that a high level of short interest is not necessarily a bullish or bearish indicator for stock prices and that short sellers do not, on average, earn abnormal stock returns.

SHORT SEWNG IN THE EQUITY MARKETS Figure A shows monthly total short interest (in millions of shares) on the NYSE, Amex and NASDAQ for the 1986-91 period. While short interest increased in all three markets over the period, the NYSE experienced the most volume and growth of short interest. Also evident on the NYSE is a pre-Gulf War increase in short interest and a subsequent decline; no similar change is evident on the NASDAQ or Amex.

It also appears that the volatility of short interest was greater on the NYSE than on the NASDAQ or Amex markets. Figure B shows percentage changes in monthly short interest for the NYSE, Amex and NASDAQ. Monthly percentage changes declined slightly.

Figure C provides monthly short interest ratios (SIRs) for the NYSE and NASDAQ for the 1986-91 period. SIRs in both markets increased, with the SIR for the NYSE averaging nearly twice the SIR for NASDAQ over the entire period and being somewhat more volatile. Overall, the 1986-91 data indicate that short interest and SIRs are greater and more volatile for the NYSE than for NASDAQ.

20 Financial Analysts Journal / January-February 1994



Figure A Effcinty Priced Commodities 1,000

800-

600-

400-

200 -

El NASDAQ El NYSE

* Amex

Figure B. Monftly Short-nterest Pernage Change, Nvmb 1986-October 1991

30

20

10

u0 A. A

U-10 - l

-20

-30

* NASDAQ [1 NYSE

[E1 Amex

Financial Analysts Journal / January-February 1994 2



Figure C. NYSE and NASDAQ Shorltntrest Ra,v 1986-Octoe 1991 6

5-

4-

2-

0

* NYSE N D OTC

Short-SdIing Obecdves Short selling may be used to accomplish several

objectives. Bearish speculators may use it to seek profits from price declines. Much of the critical attention given short selling by regulators and other governing bodies over the years has reflected concern that some investors have earned unfair speculative profits by causing security price declines with short selling. It must be emphasized, however, that most short selling is done by market professionals providing liquidity to the market. NYSE data indicate that exchange members account for 80% to 85% of all short sales. In providing much needed liquidity to the market, market makers are also actively par- ticipating in the price-determination process.

Other motivations for short selling include hedging, arbitrage and tax considerations. In fact, much of the growth in the volume of short selling in recent years has been attributed to the hedging and arbitrage associated with the program trading activities of market professionals.

Preks Reseach Seneca, Mayor, Biggs, Smith and Kerrigan have all

examined the relation between short interest levels and subsequent stock price movements, with conflicting results.' Fosback, Reilly and Whitford, and Bowlin and Rozeff analyzed the relation between the monthly short- selling behavior of specialists, as measured by the spe-

cialists' short sale ratios (SSRs), and subsequent stock returns over alternative time intervals.2 These studies did not find a consistent relation between SSR and stock returns.

McDonald and Baron investigated whether short traders earn excess returns on a risk-adjusted basis.3 They analyzed short interest and short return data for a random sample of 100 NYSE stocks over a five-year period ending in 1966. The data revealed a direct relation between short interest (as measured by the number of months with a reported short position) and risk (as measured by beta). Their results indicated that short sellers, on average, earned negative returns and were not able to generate returns in excess of a naive short- selling portfolio strategy.

A number of more recent studies have investigated the impact of short-selling restrictions on common stock prices. For the most part, these studies developed mod- els to evaluate under what conditions short selling rules affect security pricing. The results indicate that assumptions relating to investors' expectations have a significant impact on whether short-selling rules affect prices.

Figlewski argues that, for individual stocks, the observed short interest is a proxy for unfavorable information.4 If this is the case, returns on high-short-interest portfolios should be lower than those on low-short- interest portfolios, after adjusting for portfolio risk. He analyzes monthly returns and short positions as of the

22 Financial Analysts Joumal / January-February 1994



15th of each month for each stock included in the Standard & Poor's 500 over the period January 1973 through June 1979 and obtains results supporting his hypothesis.

Diamond and Verrecchia develop a model in which constraints on short sales affect the distribution of security prices, the absolute speed of adjustment of prices to private information, and the relative speed of adjustment to good versus bad news.5 They assume two conditions-(1) asset trading is either information or liquidity motivated and (2) short-sale restrictions affect market participants differently so that the pool of information-motivated short sellers is larger with short-sale restrictions than without. They find that unanticipated increases in short interest reveal a greater-than-expected level of short sales. As a consequence, short-sale restrictions imply that the larger the number of short sales, the lower the price subsequent to the announcement. In support of this hypothesis, Conrad reports that an increase in unexpected short interest for NYSE-listed stocks is negatively correlated with excess returns after the end of the month over which short interest is measured.6

Overall, prior research does not provide a clear indication of the relation between short selling and stock prices.

TesWaIe Pmp n We test three hypotheses relating to the impact of

short-sale restrictions on stock prices. * Proposition 1. Securities do not have sufficiently

close substitutes, so demand is less than elastic. Because short sales effectively increase the supply of shares in the marketplace, they create a downward pressure on prices.

* Proposition 2. To the extent that short positions must eventually be covered, a high short interest ratio constitutes a boost in potential demand, which has a bullish effect. In contrast, a rise in short interest may be interpreted as an indicator of pessimism, precipitating a downward adjustment in stock prices.

* Proposition 3. If informed short sellers depress prices through security sales, at the expense of less informed traders, they may earn excess profits on a risk-adjusted basis. Evidence in support of this hypothesis would violate the implications of informational efficiency of the securities market.

A corollary to Proposition 3 involves the regulation of short sales in the form of the uptick rule. Shorting on upticks is required for short sales on exchanges but not for securities traded over the counter. If the uptick rule protects exchange-listed securities from being driven down by bearish speculators, then one would expect to see a stronger positive relation between short selling and stock prices for exchange-traded stocks than for NASDAQ stocks.

DATA AND METHOD We test Proposition 1 by analyzing the relation between monthly changes in short interest positions and returns

on three markets (NYSE, Amex and NASDAQ) for a random sample of individual common stocks. We then examine closely the risk and return of short positions in common stocks using individual company data. The most significant question addressed here is whether or not short sellers, on average, earn abnormal returns (Proposition 3). Resolution of this issue provides indirect evidence on Proposition 2 concerning the short selling- share price relationship. Specifically, if short sellers earn abnormal returns, it can be presumed that high-short- interest positions lead to lower stock prices. In both stages, we evaluate the impact of being traded on an exchange versus over the counter.

We selected a random sample of 50 companies from the NYSE and Amex and 50 NASDAQ companies as of November 15, 1986. This date corresponds to the first publication of monthly short interest data for NASDAQ common stocks. The companies selected had to have published short sales data available. We obtained the short interest position and average daily trading volume of each stock, reported as of the 15th of each month, from the Wall Street Journal. The S&P Daily Stock Record was the source of monthly data for the 1986-91 period for the following items-stock prices, dividends and splits for individual stocks, and NYSE, Amex and NASDAQ index averages as of the 15th of each month.

Charges in Short PoslUon and Stock Returms We employ linear regression to test for a relation

between monthly returns and monthly changes in short position. First, we regressed the returns for the market composites against percentage changes in market short position to assess if differences exist between markets:

Rmt = ai + bi (SHORTmt) + eit (1)

where:

Rmt = the return on the market (NYSE, Amex, NASDAQ) for month t,

ai = the constant regression coefficient, bi = the regression slope coefficient,

SHORTmt = the percentage change in short interest for firm i in month t and

eit = the regression error term, assumed to be serially independent.

Second, we regressed monthly returns for the individual stocks against monthly percentage changes in individual short positions over the five years:

Rit = ai + bi (SHORTit) + eit (2)

where:

R,t = the return on firm i's stock for month t and SHORTit = the percentage change in short interest for

firm i in month t.

Third, we evaluated the relation between stock returns and changes in short position after accounting for overall market movements and individual stock risk. This is accomplished by using an alternative form of the

Financial Analysts Joumal / January-February 1994 23



market model in which the individual stock's monthly returns are regressed on (1) the corresponding monthly returns for the appropriate market (NYSE, Amex or NASDAQ), which accounts for market movements and individual firm systematic risk; (2) the logarithm of the total market value of the company, which accounts for unsystematic stock risk (the largest portion of which is the small-firm effect); and (3) percentage changes in the individual stock's short position. This regression is spec- ified as:

Rit = ai + b1 Rmt + b2 MVi + b3(SHORTit) + eit (3)

where:

b1 = beta, or systematic-risk regression slope coefficient,

b2 = the regression coefficient associated with firm market value,

b3 = the regression coefficient associated with changes in short position and

MVi = the logarithm of the total equity market value of firm i.

If the results are to support Proposition 1, the regression coefficient associated with the change in short position must be significantly less than zero. A significantly negative short position coefficient in any or all of the above regressions, however, is not sufficient to prove Proposition 1, because other considerations (most nota- bly, information) are not held constant. A zero or positive coefficient, especially in the case of the third regression, would be sufficient to reject Proposition 1, showing that short selling itself does not lead to lower stock prices.

Do Shot Selle Earn Abnronl Reurms? To examine the relation between short sales and

abnormal returns, our study updates and improves upon a procedure developed in McDonald and Baron.7 First, we computed measures of risk and relative short selling to test the hypothesis developed in Houthakker and Pyle that, all else equal, riskier stocks have higher and more variable short positions.8 We estimated the risk measure for each stock by computing the five-year beta using the market model. According to modern capital market theory and the market model, beta is the relevant measure of risk for a security, because it is the amount of risk a security adds to a diversified portfolio. Beta is estimated by regressing monthly security returns on an index of monthly market returns:

Rit = ai + Bi (Rmt) + eit (4)

where Bi is the regression coefficient, or beta, for stock i. The average monthly short interest ratio (SIR) is

used as the measure of relative short selling for each stock. To test for a relation between relative short selling and risk, we estimated the following regression:

SIRi = ai + bi (By) + ei, (5)

where:

SIRi = the mean monthly short interest ratio for stock i, computed as the number of shares shorted, as of the 15th of the month, divided by the average daily volume over the previous four weeks, and

bi = the regression slope coefficient.

A positive relation between risk and short selling is presumed if the regression slope coefficient (bi) is significantly greater than zero.

To test if stocks with more variable short positions tend to be riskier, we estimated the following regression:

V(SIRi) = ai + bi (Bi) + ei (6)

where V(SIR,) is the variance of monthly short interest ratios for stock i. A positive relation between the variability of short positions and risk is presumed if the regression slope coefficient (bi) is significantly greater than zero.

Second, we approximated average returns on short positions by assuming that the stock short positions reported at mid-month were all opened at the mid- month market price and covered at the following month's mid-month price. This procedure results in a measure of the return on short positions that is similar to but more precise than that used by McDonald and Baron.9 The monthly return on a short position in common stock can be defined as:

SR* = [(mt - RImt] - 1 (7)

where:

SRTt = the return for month t on a short position in the stock of firm i,

mt = the margin requirement for month t and t= the month t return for long positions in stock i,

defined as the mid-month price plus dividends divided by the next mid-month price, minus 1.0.

We adopt two assumptions that simplify the analysis without having a significant effect on the results. First, we assume that the stock is loaned flat, so that the lender of the securities receives the proceeds of the short sale and interest is neither paid to nor received from the short seller. McDonald and Baron note that, to the extent the cash in a margin account can be used to reduce the debit balance of existing loans (thereby reducing interest on margin loans), this procedure may omit a small component of the short position return.10 Second, while in practice virtually all short sales are made in margin accounts, we ignore the effect of margin requirements on short positions. Hence we effectively assume a margin requirement of 100% each month for all participants. This assumption affects only the interpre- tation of the results, as the estimated returns understate the actual returns by a scale factor if the true m, is some constant less than one over the same period. With mt equal to 1.0, Equation 7 becomes:

SRit = - Rit. (8)




Under these assumptions, the total gain or loss (Iit) on short positions on stock i over month t equals:

Iit = Vit x SRit (9)

where Vi, is the current market value of the reported short position of sit shares. The mean monthly gain or loss (Wi) on "timed" short positions in stock i is defined as the sum of monthly income figures, divided by the sum of short position values for stock i over the five-year period. This weighted average return over the 60-month period is:

60 60

Wi = ( vit x SRit)/( vit). (10) t=l t=1

The 50 NYSE/Amex and 50 NASDAQ firms in the study are categorized into deciles based on risk (five- year betas). To test the hypothesis that short sellers earn abnormal returns, we compared the mean monthly returns on the timed short positions (W) with mean monthly returns from a naive short selling strategy. The naive strategy presumes that an equal dollar amount of stock i is sold short each month. The mean return (Q1) on stock i from the naive short strategy is expressed as:

60

Qi =( E SRJ1)60. ( t=1

If the mean timed short position returns (W) are consistently larger than the mean returns from the naive strategy (Q), the proposition that short sellers earn abnormal returns will be supported. These results suggest superior timing and/or price manipulation by short sellers.

EMPIRICAL RESULTS Table 1 provides summary statistics for the regression of monthly market returns on changes in monthly market short interest over the 1986-91 period. As noted above, a necessary condition for these results to support Prop- osition 1 is that the estimated regression coefficient for SHORT (bi) be significantly less than zero. The bi's for NYSESHORT, OTCSHORT and ASESHORT are 0.106, 0.132 and 0.308, respectively, with only the Amex coefficient being significant. The positive coefficients suggest that aggregate short positions are increasing in up months and decreasing in down months. This is consistent with acceptable market-making behavior (shorting into increasing markets and reducing short positions in down markets) and is not consistent with the notion that short sellers, in the aggregate, are selling into down markets and thereby driving down stock prices.

Table 2 gives the summary statistics for the regression of the individual common stock returns and the individual changes in short positions for the 1986-91 period. As was the case with the overall market data, the regression coefficient for the SHORT variable in each case is positive, with values of 0.018, 0.003 and 0.009 for the NYSE, NASDAQ and Amex companies, respec-

Tble 1. Regressin of Montfhy Maket Index Reums on Monthly Changes In Showt Inteb

NYSE Model (NYSEComp = 0.00723 + 0.106 NYSESHORT)

Standard Predictor Coefficient Deviation t-Statistic

Constant 0.00723 0.00607 1.19 NYSE SHORT 0.10553 0.08353 1.26 s = 0.04620 R2 = 2.7% Adj. R2 = 1.0%

NASDAQ Model (NASDAQComp = 0.00733 + 0.132 OTCSHORT)


Constant 0.00733 0.00839 0.87 OTC SHORT 0.13230 0.12850 1.03 s = 0.06352 R2 = 1.8% Adj. R2 = 0.1%

Amex Model (ASEIndex = 0.00182 + 0.308 ASESHORT)


Constant 0.00182 0.00723 0.25 ASE SHORT 0.30750 0.14400 2.14 s = 0.05181 R2 = 7.4% Adj. R2 = 5.8%

Tab 2. Regression of Monthly Common Stxk Retunms on Monthly Chges in Short Intenest

NYSE Model (NYSEReturn = 0.00796 + 0.0182 SHORT)


Constant 0.00796 0.00271 2.94 SHORT 0.01818 0.00439 4.14 s = 0.1153 R2 = 0.9% Adj. R2 = 0.9%

NASDAQ Model (OTC Return = 0.0163 + 0.00267 SHORT)



Amex Model (ASE Return = 0.00692 + 0.00869 SHORT)






tively. These figures indicate very little difference between markets. But, as in the previous case, the R-squares indicate that the regression fit is not very good, which tells us that monthly changes in a company's short position have virtually no effect on monthly returns.

Table 3 gives the summary statistics for the regression of the monthly common stock return on the respec- tive monthly market return, the log of the firm's equity market value and the percentage change in the company's short interest position over the period. Because market movement and firm risk are accounted for in this regression, it provides a stronger test of the relation between change in short interest and stock returns. In each regression, the market composite return is highly significant, which increases the explanatory power of each regression (as indicated by the R-squares). The regression coefficient for the SHORT variable in each case is positive, with values of 0.017, 0.003 and 0.005 for

NYSE, NASDAQ and Amex companies, respectively. Only the t-statistic for the NYSE is significant. Other- wise, the results indicate that the SHORT variable has little explanatory power.

Several conclusions can be drawn from the initial empirical analysis. First, in overall market as well as individual company data, a negative relationship does not exist between monthly stock returns and monthly changes in short interest position. If anything, the three regressions indicate a weak positive relation between these two variables. A positive relationship indicates that short interest is increasing in positive return months and decreasing in negative return months. This result is inconsistent with the proposition that short sellers, in the aggregate, are selling into down markets and exac- erbating overall market as well as individual stock price declines.

The overall results are not overly sensitive to where the security is traded. It may therefore be concluded that the lack of an uptick rule in the over-the-counter market does not increase the sensitivity of these stocks to changes in their short interest positions. Overall, these findings indicate that, in all the markets, on a month-to- month basis, increases in short selling are not associated with lower stock prices.

Risk Retum and Short Poslons Table 4 shows the summary statistics for the regres-

sion of mean SIRS (Model 1) and Table 5 the statistics for

Table 3. Regressio of Monthly Common Stock Rehns on Ma ltIdex Reurs MNt~hly Chnges in Short Inbrest and Manlet Value

NYSE Model (NYSE Return = -0.0102 + 1.25 NYSE Comp + 0.0171

SHORT + 0.00683 log(MV))


Constant -0.10207 0.02233 -4.57 NYSE Comp 1.25465 0.05059 24.80 SHORT 0.01713 0.00379 4.52 log(MV) 0.00683 0.00152 4.49 s = 0.09948 R2 = 26.3% Adj. R2 = 26.2%

NASDAQ Model (OTC Return = -0.100 + 1.28 NASDAQ Comp + 0.00327



Constant -0.09997 0.02242 -4.46 NASDAQ Comp 1.28162 0.03965 32.32 SHORT 0.00327 0.00262 1.25 log(MV) 0.00786 0.00167 4.71 s = 0.1331 R2 = 27.4% Adj. R2 = 27.3%

Amex Model (ASE Return = -0.0944 + 1.25 ASE Comp + 0.00519



Constant -0.09441 0.03912 -2.41 ASE Comp 1.25423 0.09992 12.55 SHORT 0.00519 0.00444 1.17 log(MV) 0.00750 0.00317 2.37 s = 0.1545 R2 = 16.1% Adj. R2 = 15.8%

Table 4. Regiession of Mean Monthly Short Ratio on Beta

NYSE Model (SIR = 5.80 - 0.17 BETA)


Constant 5.799 1.923 3.02 BETA -0.172 1.378 -0.12 s = 4.557 R2 = 0.1% Adj. R2 = 0.0%

NASDAQ Model (SIR = -2.25 + 6.21 BETA)


Constant -2.247 3.185 -0.71 BETA 6.209 2.342 2.65 s = 7.3981 R2 = 12.8% Adj.R2 = 11.0%

Amex Model (SIR = 33.1 - 4.1 BETA)


Constant 33.150 26.910 1.23 BETA -4.100 20.520 -0.20 s = 38.77 R2 = 0.3% Adj. R2 = 0.0%




Table 5. Regression of Varian of Monthly Short Raboon Beta

NYSE Model (V(SIR) = 17.2 + 28.1 BETA)


Constant 17.25 38.76 0.44 BETA 28.05 27.78 1.01 s = 91.85 R2= 3.2% Adj. R2 = 0.1%

NASDAQ Model (V(SIR) = -80.6 + 95.1 BETA)


Constant -80.65 38.06 -2.12 BETA 95.12 27.99 3.40 s = 88.42 R2 = 19.4% Adj. R2 = 17.7%

Amex Model (V(SIR) = 1238 + 422 BETA)


Constant 1238.00 3761.00 0.33 BETA 422.00 2867.00 0.15 s = 5417 R2 = 0.1% Adj. R2 = 0.0%

the regression of variance of SIRi (Model 2) on beta. For Model 1 (Table 4), the regression coefficient for BETA is positive and significant for the NASDAQ stocks, but negative and insignificant for the NYSE and Amex stocks. The Model 2 results (Table 5) indicate a positive relation between the variability of the SIR and risk as measured by beta for all three markets, with a significant coefficient once again for the NASDAQ stocks (t = 3.4). Overall, the evidence on the association between the mean and variance of the monthly short interest ratio is mixed. While the relationship is positive and significant for NASDAQ stocks in both cases, it is not particularly strong or consistent for the NYSE and Amex stocks.

Table 6 presents the following data by risk decile- the average beta (b), the mean monthly returns on timed (W) and naive (Q) short positions, and the resulting mean monthly abnormal returns (W-Q). The results are provided in two groups-the 50 NYSE/Amex companies and the 50 NASDAQ companies. For all risk deciles for both groups, the mean monthly returns for both strate- gies were negative because of the increase in the overall market over the 1986-91 period.

There appears to be a weak positive risk-return relationship in that higher-beta stocks offered lower returns, especially on the timed portfolios. For the 50 NYSE/Amex firms, the mean Q is -1.12%, while the mean W is -2.00%. The 50 NASDAQ firms had a mean Q of -1.60% and a mean W of -2.63%. The mean abnormal returns for the two groups are -0.89% and

-1.03%, respectively. A standard t-test indicates that the difference between these two mean abnormal returns is not significantly different from zero.

Two conclusions can be drawn. First, the results in Table 6 do not support the hypothesis that, on a month- to-month basis, short sellers earn abnormal returns. The timed short positions underperform the naive short positions. Second, the lack of an uptick rule does not result in superior returns to short sellers in the NASDAQ market. The mean abnormal short return for the NASDAQ firms is not significantly different from the mean abnormal return for the NYSE/Amex companies.

CONCLUSONS The results indicate that, on a month-to-month basis, short selling does not lead to lower stock prices. While

Table 6. Mean Monthly Returns and Mean Abnomia Monthly Returns onl Short Positions by Decde

NYSE/Amex Firms

Mean Mean Mean Monthly Monthly Abnormal

Returns on Returns on Returns on Timed Short Naive Short Timed Short

Risk Average Positions Positions Positions Decile Beta (W)% (Q)% (W-Q)%

1 0.33 -1.01% -0.19% -0.82% 2 0.64 -2.61 -1.06 -1.55 3 0.98 -3.24 -3.14 -0.10 4 1.10 -1.29 -1.10 -0.18 5 1.22 -1.79 -1.22 -0.58 6 1.31 -1.32 -0.74 -0.58 7 1.37 -1.11 -0.36 -0.75 8 1.54 -1.96 -1.17 -0.79 9 1.82 -3.00 -1.75 -1.25

10 2.27 -2.69 -0.43 -2.26 Mean 1.26 -2.00 -1.12 -0.89

NASDAQ Firms

Mean Mean Mean Monthly Monthly Abnormal

Returns on Returns on Returns on Timed Short Naive Short Timed Short

Risk Average Positions Positions Positions Decile Beta (W)% (Q)% (W-Q)%

1 0.66 -1.92% -1.71% -0.20% 2 0.81 -1.97 -1.57 -0.40 3 1.00 -2.13 -1.38 -0.75 4 1.10 -3.33 -2.63 -0.71 5 1.16 -1.63 -0.57 -1.06 6 1.25 -3.36 -2.09 -1.27 7 1.41 -4.12 -1.80 -2.32 8 1.56 -1.87 -0.61 -1.26 9 1.71 -3.26 -2.17 -1.08

10 2.19 -2.72 -1.51 -1.21 Mean 1.28 -2.63 -1.60 -1.03




we do not directly evaluate whether a relatively high level of short interest is a bullish or bearish indicator for stock prices, we can conclude that short sellers do not earn abnormally high or low returns on a monthly basis. This suggests that, in general, high levels of short interest are not necessarily a bullish or bearish indicator. It also implies that short sellers, on average, do not possess superior investment timing skills and do not generate unfair profits by driving security prices down. Finally, the market where a security is traded-NYSE or Amex versus NASDAQ-is not a significant factor in explaining the short selling-stock price relationship. The existence of the tick-test rule on the exchanges does not,

therefore, significantly influence the association between stock prices and short selling.

Overall, our study refutes the popular notion that short sellers earn abnormal profits at the expense of less informed investors by artificially driving down stock prices through short selling. In fact, the positive relation between short interest changes and stock returns, after accounting for market and specific risk factors, indicates that short interest is increasing in positive return months and decreasing in negative return months. This result suggests that short sellers are providing liquidity to the market by shorting into increasing markets and reducing short positions in down markets.

FOOTNOTES

1. J. Seneca, "Short Interest: Bearish or Bullish?" Journal of Finance, March 1967; T. Mayor, "Short Trading Activity and The Price of Equities: Some Simulation and Regression Results," Journal of Financial and Quantitative Analysis, Sep- tember 1968; B. Biggs, "The Short Interest-A False Prov- erb," Financial Analysts Journal, July/August 1966; R. Smith, "Short Interest and Stock Market Prices," Financial Analysts Journal, November/December 1968; and T. Kerrigan, "Be- havior of The Short Interest Ratio," Financial Analysts Jour- nal, November/December 1974.

2. N. G. Fosback, "Stock Market Logic" (Institute for Eco- nomic Research, Fort Lauderdale, 1976); F. Reilly and D. T. Whitford, "A Test of the Specialists' Short Interest Ratio," Journal of Portfolio Management, No. 2, 1982; and L. Bowlin and M. Rozeff, "Do Specialists' Short Sales Predict Re- turns?" Journal of Portfolio Management, Spring 1987.

3. J. McDonald and D. Baron, "Risk and Return on Short

Positions on Common Stocks," Journal of Finance, March 1973.

4. S. Figlewski, "The Informational Effects of Restrictions of Short Sales: Some Empirical Evidence," Journal of Financial and Quantitative Analysis 18 (1981), 463-76.

5. D. Diamond and R. Verrecchia, "Constraints on Short- Selling and Asset Price Adjustments to Private Informa- tion," Journal of Financial Economics 18 (1987), 277-311.

6. J. Conrad, "The Price Effects of Short Sale Restrictions: Some Empirical Evidence" (Ph.D. dissertation, University of Chicago, 1986).

7. McDonald and Baron, "Risk and Return," op. cit. 8. H. Houthakker, "Can Speculators Forecast?" Review of

Economics and Statistics, May 1957, and D. Pyle, "On the Theory of Financial Intermediation," Journal of Finance, June 1971.

9. McDonald and Baron, "Risk and Return," op. cit. 10. Ibid.




short selling and common stock prices

Documents