What determines the level of short-selling activity?

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

D_MEG is the dummy variable of the merger and acquisition activities (one for stock that announces an acquisition and remains one until the acquisition is completed or withdrawn). I predict that the relation between D_MEG and the short-interest level is positive.

To control for the difference of investors' opinions, I construct two variables. SIGMA is standard deviation of daily returns in the previous three months. LNAGE is the natural logarithm of firm age, which defined as the difference of current time and the first time the stock is included in the CRSP monthly file. I predict that the relation between SIGMA and RSI is positive, and that the relation between LNAGE and RSI is negative.

For controlling the borrowing costs, I construct three variables. LNCAP is the natural logarithm of market capitalization. LNCAP is a proxy for the borrowing costs of short selling, i.e., relatively large stocks have a lower borrowing rate and small stocks have a higher borrowing rate (D'Avolio, 2002; Geczy, Musto, and Reed, 2002). Market capitalization is also a proxy for information transparency, which has a negative effect on the level of short selling. If the stock has higher information transparency, then the current prices will better reflect the intrinsic value of the firm and are less likely to be overpriced and shorted. Therefore, the relation between short interest and LNCAP could be positive or negative. I use two proxies for the institutional ownership: LNIH_NO and IH_RATIO. I define LNIH_NO as a natural logarithm of one plus the number of institutions holding the stock, and IH_RATIO as the fraction of shares outstanding held by institutional investors. I predict that both LNIH_NO and IH_RATIO have positive relation to RSI.

The relation between liquidity and the level of short selling should be strongly positive. I use ILLIQ as a proxy to control the liquidity. I define ILLIQ as the average ratio of the daily absolute return to the dollar trading volume on that day (Amihud, 2002):

[ILLIQ.sub.it] = 1/[D.sub.im] [[D.sub.im].summation over (t=1)] |[R.sub.imd]|/ [VOLD.sub.imd]. (1)

[D.sub.im] is the number of days for which data is available for stock i in month m, [R.sub.imd] is the return on stock i on day d of month t and [VOLD.sub.imd] is the respective daily volume in dollars. I predict that the relation between ILLIQ and RSI is negative. (2)

The level of short selling may increase for two reasons: first, because the liquidity of the stocks will be improved if added into the S&P 500 Index and the transaction cost will decrease (Hegde and McDermott, 2003; and Becker-Blease and Paul, 2006). Second, because the arbitrage between the index futures markets and the underlying stocks facilitates short selling (Jegadeesh and Subrahmanyam, 1993). However, the level of short selling may decrease because of enhanced investor awareness (Elliott, Van Ness, Walker, and Warr, 2006). Analysts pay more attention to those stocks added into the S&P 500 Index and to those for which the information asymmetry is reduced (Denis, McConnell, Ovtchinnikov, and Yu, 2003). Therefore, the impact of add or delete to the S&P 500 Index on level of short selling is an empirical question that I wish to address. I include D_SP500 as a dummy variable (one for yes, zero for no) to investigate whether including a stock in S&P 500 Index affects its short-selling level.