What determines the level of short-selling activity?

Financial Management (Financial Management Association), Winter, 2007 by Hung Wan Kot

III. Preliminary Statistics

Figure 1 shows the general performance of the stock markets and the short-selling activities from 1988 to 2002. In general, the NYSE composite index steadily increases before 1997 and becomes more volatile after that. The annual growth rate of short interest is 21% in the NYSE. The Nasdaq market shows a different pattern. Its composite index steadily increases from June 1988 to December 1997, at which point it produces the Dot Com bubble. The Nasdaq composite index increased to a peak of 4696 in February 2000, after which the bubble burst. By December 2002 the index has dropped to 1335. The annual growth rate of short interest is 23.8% in the Nasdaq.

Table 1 reports the descriptive statistics of sample stocks in the NYSE and Nasdaq. Panel A reports the stock characteristics of the NYSE sample. Relative short interest increased sharply from a median of 0.58% in 1988-1997 to a median of 1.21% in 1998-2002. Panel B reports the stock characteristics of the Nasdaq sample. Relative short interest increased sharply from a median of 0.23% in 1988.6-1997 to a median of 0.7% in 1998-2002.

[FIGURE 1 OMITTED]

Table II reports the Spearman correlations among variables. Except for LNAGE, the sign of correlations between relative short interest and other variables are the same on both the NYSE and Nasdaq. In addition, the correlations among market capitalizations, illiquidity measure, and institutional holdings, are high.

IV. Empirical Results

I present the results of testing the four hypotheses and then investigate which variables contribute the most to explaining the level of short-selling activity. I also investigate the difference of short-sellers' activity in the non-bubble and bubble periods.

A. Main Regression Results

Because of high correlations among LNCAP, ILLIQ, LNIH_NO, and IH_RATIO, I use an instrumental variables approach in the regression analysis. First, I obtain R_IHRATIO, which are the residual values of IH_RATIO from Equation (2), and then run Equation (3) to investigate the contemporaneous relation between short interest and predetermined variables.

IH_[RATIO.sub.it] = [[beta].sub.o] [[beta].sub.1]LNIH_[NO.sub.it] [[beta].sub.2][LNCAP.sub.it] [[beta].sub.3][ILLIQ.sub.it] [[epsilon].sub.it]. (2)

[RSI.sub.it] = [[beta].sub.0] [[beta].sub.1][PR1Y.sub.it] [[beta].sub.2][BM.sub.it] [[beta].sub.3][LNAGE.sub.it] [[beta].sub.4][SIGMA.sub.it] [[beta].sub.5]D_[DEBT.sub.it] [[beta].sub.6]D_[OPT.sub.it] [[beta].sub.7]D_[MEG.sub.it] [[beta].sub.8]R_[IHRATIO.sub.it] [[beta].sub.9]D_[SP500.sub.it] [[epsilon].sub.it]. (3)

I standardize all non-dummy variables into mean zero and standard deviation one. I adjust the t-value by Rogers standard errors clustered by firm (Rogers, 1993; and Petersen, 2005).

I test all hypotheses except the Taxation Hypothesis and report the main results in Table III. The Trend Hypothesis claims that short-sellers trade on the trend of the stock prices movements. The coefficient of PR1Y is significant and negative in the Nasdaq and marginally significant in the NYSE. The evidence shows that short selling is related to the past short-term stock performance, especially on the Nasdaq. The negative sign of coefficient indicates one type of short selling is to trade the stocks based on the past trend, i.e., buy the stock if the past short-term performance is good and short or sell the stock if the past short-term performance is bad. The evidence in Table III is consistent with momentum strategies (Jegadeesh and Titman, 1993), i.e., momentum traders short loser stocks over the previous one-year period, and expect the stock prices to continue to decrease in the near future.