Printer Friendly

Short selling and the weekend effect for NYSE securities.

Using short-sale transactions data, we examine the relation between short selling and the weekend effect. We do not find that short selling is more abundant on Monday than on Friday, even for stocks that have higher Friday returns. We find that short sellers execute marc short-sale volume during the middle of the week, and that the positive correlation between short selling and returns on Monday is greater, on average, than the correlation on the other days of the week. Our results are robust to subsamples of stocks with larger weekend effects and stocks that do not have listed options.

**********

Using short-sale transactions data, we examine the relation between short selling and the weekend effect. We test the Chen and Singal (2003) hypothesis that speculative short sellers face risks in holding their positions over the weekend; thus, they will close out their position on Friday and reopen their position on the following Monday, hence, adding to the weekend effect. Chen and Singal hypothesize that the behavior of speculative short sellers adds to the weekend effect because of added selling pressure on Monday, and that the closing of short positions on Friday and the reopening of these positions on the following Monday may partially explain the weekend effect.

We test the Chen and Singal (2003) hypothesis in a different time period using short-sale transactions data. For our sample time period, we do not find evidence that supports the Chen and Singal hypothesis. Rather, we find evidence suggesting that there is more short selling during the middle of the week. Further, we find that daily returns are positively correlated with daily short-selling activity, supporting the notion that short sellers are contrarian in contemporaneous returns (e.g., Diether, Lee, and Werner, 2009). We find that on average, the positive correlation between returns and short-selling activity is greater on Monday than on other days, which is contrary to the prediction of the Chen and Singal hypothesis.

We use publicly available short-sale transaction data for NYSE stocks and find evidence contradictory to the Chen and Singal (2003) hypothesis. We subdivide our sample into stocks that do, and stocks that do not, exhibit a positive weekend effect. We create subsamples of stocks based on the size of the difference between Friday's returns and the following Monday's returns. For the sample of stocks that show a positive weekend effect, we find that Monday's returns are significantly less than Friday's returns. However, we find that Friday's short-selling activity (the percentage of short-sale volume relative to trading volume) is significantly higher than Monday's short-selling activity.

We create a subsample of stocks that do not have tradable options, following the intuition of Chen and Singal (2003), who suggest that stocks with options are likely to have less speculative short selling. We find that for stocks that de net have tradable options. Monday's short-selling activity is significantly less than Friday's short-selling activity. This finding contradicts the notion that speculative short sellers will reopen positions on Monday after closing them on Friday.

Similar to other papers that test the Chen and Singal (2003) hypothesis, we use finer short-sale data that give researchers information about the short sale transaction. However, we note that as in other papers, we de net have information about the closing of short positions. Having this information is important in testing the Chen and Singal hypothesis, because the closing of the short positions on Friday adds to the buying pressure and increases prices relative to Monday, when the opening of short positions decreases prices.

After sorting stocks into deciles based on Monday's short-selling activity, we do not find evidence that the weekend effect increases across short-selling deciles, but that it increases across the bottom five deciles, although it is mixed across the top five deciles.

Consistent with Diether, Lee, and Werner (2009), we find that short sellers are contrarian in contemporaneous returns. That is. we find that daily short-selling activity and daily returns are positively correlated for each day of the week. We find that the correlation is greatest on Monday, which contradicts the notion that short-selling pressure decreases returns on Monday, thus causing a weekend effect.

The results of our analysis are robust to subsamples of stocks that exhibit larger weekend effects and stocks without tradable options. Because our sample time period differs from that of Chen and Singal (2003), it is possible that short-selling activity has increased, which may affect the behavior of short sellers over time. Therefore, we argue that the evidence we find regarding the Chen and Singal hypothesis is robust, specifically for our sample time period. To test whether the use of finer data is the driving force behind our contradictory results, we obtain monthly short-interest data and replicate portions of Chen and Singal. We find that using monthly short-interest data produce different results than those that we obtain when we use short-sale transactions data. This finding suggests that the use of finer data produces more reliable results.

The paper is organized as follows. Section I discusses prier studies. Section II describes the data. Section III presents the methods we use to formally lest the Chen and Singal (2003) hypothesis, as well as the empirical findings of the tests. Section IV concludes.

I. Prior Studies

A number of studies examine the weekend effect. Lakonishok and Maberly (1990) and Dyl and Maberly (1992) suggest that individual investors may be responsible for trading-pattern anomalies that lead to the weekend effect. Lakonishok and Maberly find evidence that individuals increase the selling of stocks relative to buying on Monday, thereby adding to the weekend effect. Millet (1988) also finds that lower Monday returns are caused by individual investors.

Although individual investors may partially explain the weekend effect, Rogalski (1984) finds that a portion of the weekend effect can be attributed to the nontrading weekend. He finds that most of the negative Friday close-to-Monday close returns occur during the nontrading weekend and are apparent by observing the negative Friday close-to-Monday open returns. Harris (1986), looking at open-to-close returns, finds that the weekend effect occurs in the first 45 minutes of Monday's trading. Other studies suggest that the weekend effect depends on the amount of information entering the market during nontrading hours.

Keim and Stambaugh (1984) show that the weekend effect varies across firm size, and that Friday's returns and the following Monday's returns are positively correlated. Abraham and Ikenberry (1994) also present evidence that suggests that Monday's return has a strong relation to Friday's return. When Friday's return is negative, Monday's return is negative nearly 80% of the time. In Wang, Li, and Erickson's (1997) investigation of the weekend effect, the authors find that negative Monday returns are significantly greater in the last two weeks of the month.

Diamond and Verrecchia (1987) posit that a significant portion of short sales are executed by investors with information on a certain security. The recent consensus is that short sellers are informed about the true value of stocks. Using intraday short-sale data from the Australian Stock Exchange, Aitken, Frino, McCorry, and Swan (1998) examine the price reaction to short sales and find that abnormal returns decrease almost immediately after a short sale. (We note that the Australian Stock Exchange makes short sales transparent to the market immediately on execution.) Their findings are consistent with Diamond and Verrecchia, who hypothesize that informed short interest leads to an adjustment in prices. Senchack and Starks (1993) also find evidence for the Diamond and Verrecchia hypothesis. Senchack and Starks find that unexpected increases in short interest yield significantly negative, although relatively small, abnormal returns. Desai, Ramesh, Thiagarajan, and Balachandran (2002) examine the relation between the level of short interest and returns. They find that the increases in the level of short interest significantly decrease returns for Nasdaq stocks. Further, they find that heavily shorted stocks experience negative returns. For a sample of favorably performing stocks, Kadiyala and Vetsuypens (2002) find that monthly short interest declines around positive signals, suggesting that short sales can measure investor sentiment.

Boehmer, Jones, and Zhang (2008), using short-sale transaction data, find evidence that supports the Diamond and Verrecchia (1987) hypothesis, as stocks that are lightly shorted outperform stocks that are heavily shorted. These authors find that a large portion of short volume is made up of institutional traders who are likely more informed than individual traders (Lo and MacKinlay, 1990; Chakravarty and McConnell, 1997; Binay, 2005). Christophe, Ferri, and Angel (2004) find a negative relation between short selling and future abnormal returns on Nasdaq. Further, they find that during the preannouncement periods, informed traders abnormally increase the level of short sales for stocks with negative earnings announcements.

Examining the behavior of short sellers, Diether, Lee, and Werner (2009) find that short sellers are contrarian, rather than momentum, traders. These authors also find that short sellers are contrarian in contemporaneous returns, suggesting that daily short selling and daily returns are positively correlated. Consistent with the Diamond and Verrecchia (1987) hypothesis, they find that short sellers are able to predict future negative returns, at times up to five days out, suggesting that short sellers are indeed informed.

Although theoretical and empirical research shows that short sellers are able to predict negative returns, the findings of a positive contemporaneous correlation between daily short selling and daily returns warrants further investigation of the Chen and Singal (2003) hypothesis. By using monthly short-interest data, Chen and Singal find evidence that the behavior of the speculative short sellers adds to the weekend effect. Using return data from July 1962 to December 1999, they find a significant weekend effect. In a Fama and MacBeth (1973) framework, they find that the correlation between monthly short interest and returns is significantly negative on Mondays. Chen and Singal also find that the weekend effect is less for stocks with tradable options because speculators are more likely to use options than short sales because of less risk, and interpret their results to be consistent with the hypothesis that speculative short sellers add to the weekend effect by closing out their positions on Friday and increasing selling pressure on Monday as they reopen their positions.

Christophe, Ferri, and Angel (2009) test the Chen and Singal (2003) hypothesis and find a weekend effect in mean returns for Nasdaq stocks. Using a proprietary data set, which allows them to distinguish between dealer and customer short sales, they find that customer short sales as a percentage of trading volume is higher on Monday than on Friday. Although the weekend effect is apparent in their data, they conclude that customer (speculative) short selling only slightly contributes to the weekend effect. Their results provide weak evidence in support of the Chen and Singal hypothesis.

An interesting aspect of the Christophe, Ferri, and Angel (2009) study is that they find that dealer short sales make up the largest percentage of executed short sales. They assume that dealer shorts are a primary result of market making, and not of individual speculative trading.

Gao, Kalcheva, and Ma (2006) also test the Chen and Singal (2003) hypothesis by looking at short selling on the Stock Exchange of Hong Kong. They find a significant weekend effect before 1993, which was a period when short selling was not allowed. In 1994, after short selling was introduced for some stocks, the weekend effect exists for subsets of both shortable and nonshortable stocks. The authors find that Monday returns are less for shortable stocks than for nonshortable stocks, although the difference is not significant. Their finding weakly supports the Chen and Singal hypothesis. Further, the authors report that the difference in the size of the weekend effect between the subset of stocks is not significant.

The current research that is testing the Chen and Singal (2003) hypothesis has yet to provide substantial evidence for or against the notion that speculative short sellers are adding to the weekend effect.

II. Data

Our sample period is from January 1, 2005 to December 31, 2005. We obtain short-sale data from the NYSE in response to the SHO regulation, trade data from Trade and Quote (TAQ), and the daily return, price, capitalization, and volatility data from Center for Research in Security Prices (CRSP). We limit our sample to NYSE-listed common stocks, and exclude stocks with prices less than $5 and stocks that do not trade every day of the sample time period. These limitations leave 2,151 stocks, from which we calculate the weekend effect following Chen and Singal (2003). We subtract Monday's return from the previous Friday's return. If trading does not occur on Monday, then we calculate the weekend effect as the Friday's return less the following Tuesday's return. To isolate the contribution of short selling in explaining the weekend effect, we distinguish between stocks that do, and those that do not, exhibit a positive weekend effect. The total number of stocks that (do not) exhibit a positive weekend effect is 1,502 (649).

We note an important limitation in using the short sale transactions data from Regulation SHO to test the Chen and Singal (2003) hypothesis. Chen and Singal argue that the added selling pressure from speculative short sellers on Monday will likely decrease Monday's returns. However, the closing of the short positions on the previous Friday will add buying pressure and increase Friday's returns, thus increasing the weekend effect. The SHO data do not contain information on the covering of short positions, but our transactions data make it possible for us to observe the short-selling activity for each day of the week, something that monthly short-interest data are unable to do.

Out variables are as follows. Our price is the CRSP daily ending price. Market capitalization is the daily ending capitalization reported by CRSP. Volume is the average daily volume. We measure volatility as the standard deviation of returns from day [t.sub.-10] to day t, where day t is the current trading day. Table I presents statistics of the stocks used in the analysis. The average stock in our sample has a price of $33.07 with a daily capitalization and daily volume of $5,892,049 and 783,512 shares, respectively. The volatility of the average stock in our sample is 0.0429.

We create several subsamples. The first subsample consists of the 649 stocks with a nonpositive weekend effect (Friday's return minus the following Monday's return is negative). The second subsample consists of the 1,502 stocks with a positive weekend effect, that is, Friday's return minus the following Monday's return is positive. We also create subsamples of stocks with the largest weekend effects. We use subsamples of the 1,000 stocks with the largest weekend effects, and the 500 stocks with the largest weekend effects. From the stocks that exhibit a positive weekend effect, we also create a subsample of stocks that do not have tradable options (602 stocks). (1)

In Table I, Panels B and C report the different statistics for the stocks with a nonpositive and positive weekend effect. Panels D and E report the statistics for the top 1,000 stocks and top 500 stocks with the largest weekend effects. Consistent with Keim and Stambaugh (1984), we find that in terms of market capitalization, size is less for stocks with the highest weekend effects. Further, we find that volume (volatility) is smaller (greater) for stocks with the largest weekend effect. Panel F reports the characteristics of the 602 stocks that do not have tradable options. We find that these stocks are generally smaller (low market cap) and have lower volume than stocks in the other subsamples.

Table II reports the statistics for CRSP returns, by day, with each stock equally weighted. Panel A also shows the number of days. We find that Monday's and Friday's returns are similar, in the aggregate. Panels B and C of Table II show the mean returns, by day, for the subsamples of stocks with a nonpositive and a positive weekend effect. In Panel B (Panel C), we find that the Monday returns are significantly larger (smaller) than are the Friday returns. This difference in size is a product of subdividing the sample into positive and nonpositive weekend effect subsamples. In Panels D and E, we report the average returns, by day, for stocks with the largest weekend effects. In Panel F, we also find that Friday's returns are significantly larger than Monday's returns for stocks that do not have tradable options, and that the difference is similar to that in Panel C. An interesting pattern is that the size of the weekend effect depends largely on the size of Friday's return as opposed to the size of Monday's returns. We find that Friday's returns are the largest for stocks with the largest weekend effects. This finding is consistent with studies on the seasonalities in weekday returns (Lakonishok and Smidt, 1988).

III. Results

Our first test of the Chen and Singal (2003) hypothesis is to determine if there is more short-selling activity on Monday. If speculative short sellers are closing out positions on Friday and reopening positions on the following Monday, then more short-selling activity should occur on Mondays than on Fridays. Table III presents statistics for the short-activity ratio, which we define as the daily short volume divided by the daily total trade volume. The overall average short-activity ratio is almost 20%. Although this value may appear high when compared to the monthly short-interest data, Boehmer, Jones, and Zhang (2008) find that nearly 13% of NYSE SuperDOT

(1) We also create a subsample of stocks that are in the Regulation SHO pilot, which we do not tabulate. Because these stocks experience regulatory changes during out sample time period, we test to see if the results of our tests hold in this subsample. We perform our entire analysis for this subsample of stocks and find that results are qualitatively similar. The results from the tests on SHO pilot stocks are available on request. volume is short volume. For Nasdaq-listed securities, Diether, Lee, and Werner (2009) find that short volume makes up about 32% of volume, while Nasdaq reports only 3% in monthly short interest. Further, they find that nearly 24% of volume on the NYSE is made up from short-sale volume.

In Table III, Panel A shows the statistics that describe short activity for the entire sample, equally weighted by stock, for each day of the week. We find that Monday's short-activity ratio is significantly less than Friday's. This finding is contrary to the initial prediction that short selling will be higher on Monday than on Friday due to speculators reopening positions on Monday.

Panels B and C report the results for the subsamples of stocks with nonpositive and positive weekend effects, and Panels D and E report the results for the subsamples of stocks with the largest weekend effects. If short selling is a partial explanation for the weekend effect, then across subsamples with the larger weekend effects, we expect to see Monday's short-activity ratio increase relative to Friday's short-activity ratio. Surprisingly, we find that the difference between the Friday and Monday short-activity ratios increases, rather than decreases, from Panels B to E. That is, we find more short-selling activity on Friday than on Monday, and that the difference increases as the weekend effect increases. However, we note that although we find that the difference between Friday's and Monday's short-selling activity is increasing from Panels C to E, which contradicts the expectation of the Chen and Singal (2003) hypothesis, the differences between panels are not statistically significant. The t-statistic that tests whether or not Panel D's difference is larger than that of Panel C is 1.13, while the t-statistic that tests the differences between Panels D and E is 0.34.

Panel F reports the results for stocks that do not have tradable options. Chen and Singal (2003) argue that stocks without tradable options should have more speculative short sellers because speculators are not able to substitute short sales with tradable options. First, we find that our short-activity ratio is smallest in this subsample. Figlewski and Webb (1993) and Danielsen and Sorescu (2001) find that stocks with tradable options are sold short more often than are stocks that do not have tradable options. They argue that short selling for stocks with listed options is greater than for stocks without listed options, because option market makers use short sales to hedge their positions. Our finding of relatively less short activity for stocks without tradable options is consistent with the findings of Figlewski and Webb and Danielsen and Sorescu.

We also see that Monday's short activity is significantly less than Friday's. We perform a t-test to determine if the difference in Panel F is significantly larger than the difference in Panel A. We find that the difference is significant at the 5% level and has a t-statistic of 2.17, showing that, in our "no option" subsample, which should consist of more speculative short sellers, Monday's short activity is significantly less than that of Friday, and the difference is greater than the difference for the entire sample. This finding contradicts the prediction of the Chen and Singal (2003) hypothesis.

We also test to see if the difference between Monday's and Friday's short activity is significantly greater for stocks that do not have tradable options (Panel F) and stocks with a positive weekend effect (Panel C). We find that the difference is statistically significant at the 10% level and has a t-statistic of 1.84.

The difference between Friday's and Monday's short-selling activity does not depend on whether the stocks show a positive or nonpositive weekend effect. We do not find a significant difference between Monday's short activity for stocks without a weekend effect and Monday's short-selling activity for stocks with a weekend effect (t-statistic is 1.13). The difference in Friday's and Monday's short activity between stocks that exhibit positive and nonpositive weekend effects is not significant (t-statistic is 1.42). We argue that daily shorting activity is independent of the size and magnitude of the weekend effect.

However, we note the difference between out time period and that of Chen and Singal (2003). It is possible that the results from our tests differ from that of Chen and Singal because of the different time periods used in the two analyses. For instance, if short activity has increased between the study periods and short sellers are informed, as other studies indicate, then the effect of speculative short sellers may be deflated by the presence of informed short sellers.

Because there may be other factors influencing the level of short activity, we run a pooled regression of a standardized measure of short volume on several stock characteristics and day of the week dummy variables.

[SASV.sub.i,t] = [[beta].sub.0] + [[beta].sub.1][return.sub.i,t] + [[beta].sub.2] ln([volume.sub.i,t]) + [[beta].sub.3] ln([volatility.sub.i,t]) + [[delta].sub.1][MON.sub.i,t] + [[delta].sub.2][TUES.sub.i,t] + [[delta].sub.3][THUR.sub.i,t] + [[delta].sub.4][FRI.sub.i,t] + [[epsilon].sub.i]. (1)

The dependent variable, [SASV.sub.i,t] is the standardized abnormal short volume for stock i on day t.

We note that we specify Equation (1) in several different ways. We use year-end values for market cap, trading volume, volatility, and prices, and find that the results are qualitatively similar. We use the daily changes in prices and daily changes in market capitalization and again find that the results are similar. To confirm the positive relation between short-selling activity and daily returns, we choose to report the results for Equation (1). In each case, the results for abnormal short volume for day-of-the-week dummy variables are the same.

Following Lakonishok and Vermaelen (1986) and Koski and Scruggs (1998),

we define the abnormal short volume as

[ASV.sub.i,t] = [SV.sub.i,t] - [bar.SV].sub.i]. (2)

Here, [SV.sub.i,t] is the average daily short volume for stock i on day t, [bar.SV].sub.i] is the mean for average daily short volume for each stock, and [ASV.sub.i,t] is the abnormal short volume.

We define SASV, the standardized abnormal short volume, as

[ASV.sub.i,t] = [ASV.sub.i,t]/[sigma]([SV.sub.i]). (3)

Here, [sigma]([SV.sub.i]) is the standard deviation of short volume for each stock. This standardization allows each dependent variable in the panel data sample to be distributed with a zero mean and a unit variance.

The explanatory variables include the daily return, return; the natural log of the daily volume, In(volume); and the natural log of the standard deviation of returns for day [t.sub.-10] to day t, ln(volatility), where day t is the current trading day. We also use dummy variables for the days of the week. We test each stock for fixed effects, and find differences across stocks. We perform both a one-way fixed effects (by stock) regression and a two-way fixed effects (by stock and date) regression, and find that the results are qualitatively similar. In Table IV we report the one-way fixed effects results for Equation (1). Our finding of a contemporaneous positive relation between returns and abnormal short volume support that of Diether, Lee, and Werner (2009). We also find that daily volume is positively related to abnormal short-sale volume. The results for the estimates of volatility are mixed.

Panel A shows that abnormal short volume is highest during the middle of the week. We note that Monday's and Friday's short volume is significantly less than Wednesday's short volume, and that Tuesday's and Thursday's short volume is less than Wednesday's short volume, although the difference is not significant in each panel. The estimate for the Monday dummy variable is the most negative, which again contradicts the hypothesis that short sellers close out their positions on Friday and reopen them on the following Monday. Rather, the estimate supports the notion that there are more short sellers during the middle of the week.

Panels B through F also show that short sellers execute the least short volume on Mondays, and generally execute more short volume during the middle of the week. In each of the panels, we find that Monday's and Friday's short-selling activity appears to be lowest when compared to the other days of the week. The intraweek pattern of standardized abnormal short volume appears to be robust for stocks that exhibit positive and nonpositive weekend effects, stocks with the largest weekend effects, and stocks that do not have tradable options.

A. Weekend Effect

We define the weekend effect similar to Chen and Singal (2003) as the last day of the week's return minus the following first day of the week's return. Table V presents the mean of the weekend effect, equally weighted by stock.

We sort stocks into deciles by Monday's short-selling activity ratios. If short selling contributes to the weekend effect, then as Monday's short selling increases, the weekend effect should also increase. We present the mean of the weekend effect for each decile. In Panel A, we find that the weekend effect is increasing in the bottom five deciles but becomes mixed in the top deciles. Panel B reports a negative weekend effect and no specific pattern across increasing short-selling deciles. Panel C reports the results for stocks with a positive weekend effect and shows that the weekend effect is significantly greater than zero, with a t-statistic (unreported) greater than 20. In the remaining panels, we find that the mean weekend effect generally increases in the bottom five deciles but the relation is mixed in the top five deciles. Since Panels D and E contain stocks with the largest weekend effect, we expect a stronger increasing relation between the weekend effect and Monday's short-activity ratio. We do not find substantial evidence that the calculated weekend effect is positively related to Monday's short-activity ratio. We also find that the relation between the weekend effect and Monday's short-activity ratio is weak for stocks that do not have listed options (Panel F).

To further investigate the Chen and Singal (2003) hypothesis, we estimate correlation coefficients between returns and short-selling activity, by day of the week. If the Chen and Singal hypothesis holds, short activity and returns should be negatively correlated on Mondays. Recent studies on the informativeness of short sellers suggest that short sellers are able to predict future negative returns (Senchack and Starks, 1993; Desai et al., 2002; Christophe, Ferri, and Angel, 2004; Boehmer, Jones, and Zhang, 2008). As mentioned earlier, Diether, Lee, and Werner (2009) find that short sellers are contrarian in contemporaneous returns and lagged returns. Contrarian behavior suggests that daily short activity and daily returns are positively correlated.

Table VI reports the correlation coefficients between short activity and returns, by day. Panel A presents the coefficients with the corresponding p-values for the entire sample. Consistent with Diether, Lee, and Werner (2009), we find that short selling and returns are positively correlated, which suggests that short sellers are contrarian in contemporaneous returns. Comparing the coefficients, we find that Monday's coefficient is larger than any other day of the week, suggesting that short selling on Mondays does not decrease returns, but rather, the relation is exactly the opposite and stronger than on any other day.

The Chen and Singal (2003) hypothesis predicts that Monday's contemporaneous correlation between short-selling activity and returns should become more negative as the weekend effect becomes more prevalent. We find that the correlation between Monday's returns and Monday's short-activity ratio increases from Panels B to E, which is contrary to the hypothesis. Finally, Panel F shows that the correlation coefficient is largest on Monday for stocks that do not have tradable options. Further, Monday's correlation coefficient in Panel F is larger than any other coefficient in the table, thus adding to earlier evidence that speculative short selling does not explain the weekend effect in our sample time period. We test whether or not the relation between short activity and returns is significantly greater on Monday than on the other days in the next subsection.

B. Regression Results

Chen and Singal (2003) perform Fama and MacBeth (1973) regressions for returns on five day-of-the-week dummy variables and five interaction dummy variables. In these regressions, they interact day dummy variables with a dummy variable that equals one if the relative monthly short interest is high.

Since we have access to transaction data, we estimate a pooled model similar to Chen and Singal (2003). We do not use the estimation technique of Fama and MacBeth (1973). Because we only use one year of short-sale data, we use panel data models to regress daily returns on day dummy variables and interaction variables, where we interact the day of the week with the amount of short activity on that day. We specify the pooled model as follows

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (4)

The dependent variable is the return for stock i on day t. The independent variables include day-of-the-week dummy variables in which [MON.sub.i,t] is equal to one if Monday, and zero otherwise. The other day-of-the-week dummy variables are similarly specified. Further, [MON.sub.i,t] x [SA.sub.i,t] is equal to the short-activity ratio on day t, if day t is a Monday, and zero otherwise. The other interaction variables are similarly defined. To obtain full rank of the coefficient matrix, we exclude the intercept.

Chen and Singal (2003) find a negative interaction estimate for [[delta].sub.6], thus supporting the hypothesis that speculative short sellers add to the weekend effect by reopening short positions on Monday. In their estimation they use monthly short-interest data and daily return data for the time period from July 1988 to December 1999. We estimate Equation (4) using daily short-sale data available in compliance with Regulation SHO (January 3, 2005), which provides a more robust test. If the Chen and Singal hypothesis holds, then the estimated coefficient, [[delta].sub.6], from Equation (4) will be significantly negative.

Table VII reports the results from estimating Equation (4). In Panel A we estimate Equation (4) for the entire sample of stocks and find that the estimate of [[delta].sub.6] is significantly positive, which is contrary to Chen and Singal's (2003) results. We perform a t-test to determine if the estimate of [[delta].sub.6] is greater than 1/4 [summation].sup.10.sub.j = 7] [[delta].sub.j] where j = (7, 8, 9, and 10). We find that the estimate of [[delta].sub.6] is greater than the average of the other interaction estimates with a t- statistic of 5.37 for the entire sample. The results of this test add to the comparison of the correlation coefficients in Table VI and show that on average, the correlation between returns and short-activity ratios is greatest on Mondays.

Panels B and C report the results from estimating Equation (4) for stocks with a nonpositive weekend effect and stocks with a positive weekend effect, respectively. In Panels B and C, we find that the estimated coefficient, [[delta].sub.6], is significantly positive and greater than the average of the estimates for [[delta].sub.7], [[delta].sub.8], [[delta].sub.9], and [[delta].sub.10] with a t-statistic of 3.04 (4.57). We perform a similar examination for Panels D through E We find that t-statistic that tests the difference between [[delta].sub.6] and the average of [[delta].sub.7], [[delta].sub.8], [[delta].sub.9], and [[delta].sub.10] is significant at the 5% level for the 1,000 stocks with the largest weekend effect (t-statistic = 2.45). We find that the difference between the estimate for [[delta].sub.6] and the average estimate of [[delta].sub.7], [[delta].sub.8], [[delta].sub.9], and [[delta].sub.10] is positive but not significant for the 500 stocks with the largest weekend effect (Panel E). In Panel E we find that a similar test yields a t-statistic of 5.34, suggesting that the contemporaneous relation between returns and short-selling activity is higher on Monday than on the other days of the week. The results from our regressions, which use finer short-sale data, do not support the Chen and Singal (2003) hypothesis that short selling partially causes the weekend effect.

To further test the Chen and Singal (2003) hypothesis, we obtain monthly short-interest data for some of the stocks in our sample and replicate a portion of their analyses to see if the differences in our findings are a result of using finer data. We are able to obtain monthly short interest for only 1,806 of the 2,151 stocks for total sample. We have monthly short-interest data for 552 of the 649 stocks with a nonpositive weekend effect, 1,254 of the 1,502 stocks with a positive weekend effect, 897 of the 1,000 stocks with the largest weekend effect, 471 of the 500 stocks with the largest weekend effect, and 381 of the 602 stocks that do not have tradable options.

As do Chen and Singal (2003), we sort the short-interest sample into size deciles according to market capitalization at the beginning of the month and then, within each size decile, we subdivide the observations into quartiles based on the relative short interest (RSI), which we define as the number of shares that are shorted, but uncovered, relative to the number of shares outstanding. Using the highest and the lowest quartiles, we estimate the following model:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (5)

Following Chen and Singal, we set dummy variable H equal to one if stock i is in the highest quartile on day t, and zero if stock i is in the lowest quartile on day t. We perform the regression for the entire sample and each of the subsamples as before.

Table VIII reports the results from estimating Equation (5). We find that the interaction variable MON x H is significant in only two of the six panels. We find that Tuesdays, which have negative returns in our sample, have a greater impact when interacting with the high-RSI quartile in Panels A, C, D, and E. Tuesday's interaction results in Panel B are significantly negative at the 10% level.

When we compare the results in Table VIII to the results in Table VII, we see that the contemporaneous contrarian behavior of short sellers--the positive relation between daily short volume and daily returns--on each day of the week does not show up when we use monthly short-interest data. In general, we do not find that using monthly short interest provides the same results as we find in Table VII. We note that although the results from performing this analysis lead us to believe that the findings of Chen and Singal (2003) may be biased by the use of monthly data, we are unable to conclude that the differences in our results and the results of Chen and Singal are entirely due to use of finer short-sale data.

Wang, Li, and Erickson (1997) find evidence that the weekend effect is greater during the last two weeks of the month. Therefore, as a measure of robustness, we calculate the weekend effect for the last two weeks of the month. For stocks that exhibit a positive weekend effect, we find an average weekend effect of 0.0033, which is significantly greater than zero at the 1% level, and comparatively larger than the weekend effect for the entire sample of stocks for the entire time period. Similar to previous analyses, we include day-of-the-week dummy variables and three-way interaction variables that interact the short-activity ratio with day-dummy variables and monthend dummy variables (M_E). We use the three-way interaction to control for the possibility that the Chen and Singal (2003) results are driven by the larger weekend effect in the last two weeks of the month. The pooled model is

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (6)

In Equation (6), we focus on the estimate of [[delta].sub.7]. If the Chen and Singal (2003) results are driven by the larger weekend effect during the last two weeks of the month, then the estimate of [[delta].sub.7] should be negative and significant. The results from estimating Equation (6) are reported in Table IX.

Ariel (1987) documents a monthly pattern in daily returns and suggests that monthly returns are explained by large positive daily returns in the 10 consecutive trading days beginning with the last day of the month. He finds that returns become negative during the end of the month. Consistent with Ariel's results, we find a negative estimate for 36 that suggests that returns in the last two weeks of the month are significantly less than returns in the first two weeks of the month. We find that the estimates of [[delta].sub.7] are significantly positive for the entire sample and each of the subsamples. Performing similar t-tests as before, we find that the estimate of [[delta].sub.7] is significantly larger than the average of the other interaction estimates ([[delta].sub.8], [[delta].sub.9], [[delta].sub.10], and [[delta].sub.11]) at the 1% level for the entire sample of stocks and each subsample. The t-statistics that test the difference between the estimate of [[delta].sub.7] and the average of the other interaction estimates are 9.79 in Column 1, 8.78 in Column 2, 8.51 in Column 3, 6.18 in Column 4, 3.6 in Column 5, and 3.96 in Column 6. Our evidence adds to the strong evidence against the Chen and Singal (2003) hypothesis for our sample time period.

IV. Conclusion

Chen and Singal (2003) hypothesize that speculative short sellers, because they face a greater risk if they hold short positions over the weekend, will close out their short positions on Friday, which increases buying pressure before the weekend. By reopening their positions on the following Monday, increased selling pressure will drive prices down, thus adding to the weekend effect.

We test the Chen and Singal (2003) hypothesis, using a different time period and finer short-sale data. In light of their hypothesis, we expect to see more short activity on Monday than on Friday. Using short-sale transaction data for a sample of NYSE-listed stocks, we find contradictory evidence. Although the short-sale transaction data do not contain information about the covering of short sales, we are able to test whether or not short selling differs between the days of the week. We find that the level of short-selling activity is lowest on Monday and significantly less than short-selling activity on Friday. We find that the difference between Monday's short-activity ratio and Friday's short-activity ratio increases across stocks with larger weekend effects, and is largest for stocks that do not have tradable options, which Chen and Singal argue have more speculative short sellers. We also find that short volume is highest during the middle of the week.

According to the Chen and Singal (2003) hypothesis, we also expect to see the weekend effect become larger as the level of Monday's short selling increases. We do not find evidence that supports this proposition. After sorting stocks into deciles according to Monday's short-activity ratios, we show that the weekend effect generally increases in the bottom deciles, but does not increase in the top deciles. The results are robust to stocks with larger weekend effects and stocks without listed options.

We estimate correlation coefficients between returns and short-activity ratios for each day of the week and find that returns are positively correlated with short-selling activity for each day. This finding is consistent with the notion that short sellers are contrarian in contemporaneous returns (Diether, Lee, and Werner, 2009). We find that in our sample, the correlation is greatest on Monday. We note especially that we find that the size of the positive correlation increases across stocks with larger weekend effects and is largest for stocks that do not have tradable options.

Our regression analysis generates some surprising results. When we regress returns on day-of-the-week dummy variables and interaction variables that interact day dummy variables with daily short-activity ratios, we find that returns are increasing in short-activity ratios on Monday. Further, we find that the increasing relation between Monday's returns and Monday's short-activity ratios is significantly larger than is the average relation between returns and short-activity ratios on other days.

Using monthly short-interest data, we replicate a portion of Chen and Singal's (2003) analysis for out sample time period. We find that using finer short-sale data help explain the role short sellers play in determining the weekend effect. However, we are unable to conclude that the difference in our findings and the findings of Chen and Singal are entirely due to the use of finer short-sale data. When we control for the results of Wang, Li, and Erickson (1997), we find that during the last two weeks of the months, the relation between returns and short activity is positive and significant on Mondays. Similarly, we find that Monday's relation is significantly greater than is the average relation between short activity and returns on other days. Using finer and more recent short-sale data, out results do not support the hypothesis of Chen and Singal that speculative short sellers add to the weekend effect.

We thank Andriy Shkilko. Richard Warr, and seminar participants at the University of Mississippi, Wilfrid Laurier University, and the 2007 Eastern Finance Association Meeting for helpful comments. We also thank an anonymous referee for insightful comments and suggestions. All errors are the responsibility of the authors.

References

Abraham, A. and D.L. Ikenberry, 1994, "The Individual Investor and the Weekend Effect," Journal of Financial and Quantitative Analysis 29, 263-277.

Aitken, M., A. Frino, M. McCorry, and P. Swan, 1998, "Short Sales Are Almost Instantaneously Bad News: Evidence from the Australian Stock Exchange," Journal of Finance 53, 2205-2223.

Ariel, R.A., 1987, "A Monthly Effect in Stock Returns," Journal of Financial Economics 18, 161-174.

Binay, M., 2005, "Performance Attribution of US Institutional Investors," Financial Management 34, 127152.

Boehmer, E., C.M. Jones, and X. Zhang, 2008, "Which Shorts Are Informed?" Journal of Finance 63, 491-527.

Chakravarty, S. and J.J. McConnell, 1997, "An Analysis of Prices, Bid/Ask Spreads and Bid/Ask Depths Surrounding Ivan Boesky's Illegal Trading in Carnation's Stock," Financial Management 26, 18-34.

Chen, H. and V. Singal, 2003, "Role of Speculative Short Sales in Price Formation: The Case of the Weekend Effect," Journal of Finance 58, 685-705.

Christophe, S., M. Ferri, and J. Angel, 2004, "Short Selling Prior to Earnings Announcements," Journal of Finance 59, 1845-1875.

Christophe, S., M. Ferri, and J. Angel, 2009, "Short Selling and the Weekend Effect in Stock Returns," Financial Review 44, 31-57.

Danielsen, B. and S.M. Sorescu, 2001, "Why Do Options Introductions Depress Stock Prices? An Empirical Study of Diminishing Short Sale Constraints," Journal of Financial and Quantitative Analysis 36, 451484.

Desai, H., K. Ramesh, S. Thiagarajan, and B. Balachandran, 2002, "An Investigation of the Information Role of Short Interest in the NASDAQ Market," Journal of Finance 52, 2263-2287.

Diamond, D. and R. Verrecchia, 1987, "Constraints on Short Selling and Asset Price Adjustment to Private Information," Journal of Financial Economics 18, 277-312.

Diether, K., K. Lee, and I. Werner, 2009, "Can Short Sellers Predict Returns? Daily Evidence," Review of Financial Studies 22, 575-607.

Dyl, E. and E. Maberly, 1992, "Odd-Lot Transactions around the Turn of the Year and the January Effect," Journal of Financial and Quantitative Analysis 27, 591-604.

Fama, E. and J. MacBeth, 1973, "Risk Return and Equilibrium: Empirical Tests," Journal of Political Economy 81, 607-636.

Figlewski, S. and G. Webb, 1993, "Options, Short Sales, and Market Completeness," Journal of Finance 48, 761-777.

Gao, P., I. Kalcheva, and T. Ma, 2006, "Short Sales and the Weekend Effect Evidence from Hong Kong," University of Arizona Working Paper.

Harris, L., 1986, "A Transaction Data Study of Weekly and Intradaily Patterns in Stock Returns," Journal of Financial Economics 16, 99-117.

Kadiyala, P. and M.R. Vetsuypens, 2002, "Are Stock Splits Credible Signals? Evidence from Short-Interest Data," Financial Management 31, 31-49.

Keim, D. and R. Stambaugh, 1984, "A Further Investigation of the Weekend Effect in Stock Returns," Journal of Finance 39, 819-835.

Koski, J.L. and J.T. Scruggs, 1998, "Who Trades around the Ex-Dividend Day? Evidence from NYSE Audit File Data," Financial Management 27, 58-72.

Lakonishok, J. and E. Maberly, 1990, "The Weekend Effect: Trading Patterns of Individual and Institutional Investors," Journal of Finance 45, 231-243.

Lakonishok, J. and S. Smidt, 1988, "Are Seasonal Anomalies Real? A Ninety-Year Perspective," Review of Financial Studies 1, 403-427.

Lakonishok, J. and T. Vermaelen, 1986, "Tax-Induced Trading around Ex-Dividend Days," Journal of Financial Economics 3, 287-319.

Lo, A.W. and A.C. MacKinlay, 1990, "When Are Contrarian Profits Due to Stock Market Overreaction?" Review of Financial Studies 3, 175-206.

Miller, E., 1988, "Why a Weekend Effect?" Journal of Portfolio Management 14, 42-48.

Rogalski, R., 1984, "New Findings Regarding the Day-of-the-Week Returns over Trading and Non-Trading Periods: A Note," Journal of Finance 39, 1603-1614.

Senchack, A. and L. Starks, 1993, "Short Sale Restrictions and Market Reaction to Short-Interest Announcements," Journal of Financial and Quantitative Analysis 28, 177-194.

Wang, K., Y. Li, and J. Erickson, 1997, "A New Look at the Monday Effect," Journal of Finance 52, 2171-2186.

Benjamin M. Blau, Bannie F. Van Ness, and Robert A. Van Ness *

* Benjamin M. Blau is an Assistant Professor at Utah State University in Logan, UT. Bonnie F. Van Ness and Robert A. Van Ness are Associate Professors at the University of Mississippi in University, MS.
Table I. Statistics for Our Sample of Stocks

This table presents statistics that describe the stocks in our
sample. Our sample comprises NYSE stocks that have an average price
more than $5 and are traded every day of the sample time period
(January 1, 2005 to December 31, 2005). The total number of stocks
in the sample is 2,151. We subdivide the sample by stocks that
exhibit a positive and negative weekend effect, which we define as
Friday's return minus the following Monday's return. We also create
subsamples for the 1,000 and the 500 stocks that have the largest
weekend effect. Further, from the stocks that show a positive
weekend effect, we create a subsample for nonoptionable stocks that
likely contains more speculative trades than optionable stocks. We
present statistics for price, market capitalization, average
volume, and volatility. Price is the closing price for each stock,
Mkt Cap is daily market capitalization, and Volume is the average
volume. We measure Volatility as the standard deviation of returns
for the time period ([t.sub.-10] to t) where t is the current
trading day. The table gives the equally weighted (by stock)
average daily statistics.

 Price Mkt Cap Volume Volatility

Panel A. General Statistics for Entire Sample (2,151 Stocks)

Mean 33.07 5,892,049 783,512 0.0429
SD 31.58 19,373,217 1,732,327 0.0176
Min 5.59 25,822 3,990 0.0063
Max 836.30 373,034,463 29,339,069 0.1094

Panel B. General Statistics for Stocks with a Nonpositive
Weekend Effect (649 Stocks)

Mean 33.81 7,417,187 944,635 0.0425
SD 40.60 22,967,830 1,869,979 0.0172
Min 5.74 28,488 3,990 0.0103
Max 836.30 370,078,836 19,961,272 0.1090

Panel C. General Statistics for Stocks with a Positive
Weekend Effect (1,502 Stocks)

Mean 31.91 5,233,050 713,893 0.0430
SD 26.24 17,562,027 1,665,176 0.0178
Min 5.59 25,822 4,150 0.0063
Max 629.62 373,034,463 29,339,069 0.1094

Panel D. General Statistics for the 1,000 Stocks with
the Largest Weekend Effect

Mean 32.22 4,377,169 669,189 0.0468
SD 26.67 15,220,521 1,433,724 0.0169
Min 5.85 30,928 5,604 0.0133
Max 626.62 373,034,463 20,032,298 0.1094

Panel E. General Statistics for the 500 Stocks with
the Largest Weekend Effect

Mean 30.69 3,007,276 637,378 0.0533
SD 17.08 5,951,462 1,186,936 0.0163
Min 5.95 30,928 6,015 0.0205
Max 107.74 80,369,853 13,568,481 0.1094

Panel F. General Statistics.for the 602 Stocks with
No Tradable Option

Mean 25.92 857,586 103,022 0.0368
SD 31.48 2,110,093 129,908 0.0174
Min 5.59 25,822 5,604 0.0063
Max 626.62 34,259,019 1,638,410 0.0914

Table II. Statistics for Returns

This table presents statistics for the daily returns for each day of
the week and the difference between Friday's mean return and the mean
return of each day. Panel A shows the statistics for returns for the
sample stocks, the number of days in the sample, and the difference
between Friday's mean return and the entire sample. Panels B-F report
the results from the subsamples. The t-statistics test for the
significance of the difference between Friday's mean return and the
mean return for each day.

 Monday Tuesday Wednesday

Panel A. Mean Returns for Entire Sample, by Day.

Mean 0.0013 -0.0006 0.0002
SD 0.0024 0.0022 0.0022
Min -0.0073 -0.0128 -0.011
Max 0.0174 0.0129 0.0148
Number of days 46 52 52
Difference -0.0000 0.0020 *** 0.0012***
t-statistic (-0.01) (28.34) (18.17)

Panel B. Mean Returns for Stocks with a Nonpositive
Weekend Effect, by Day

Mean 0.0012 -0.0006 -0.0000
SD 0.0023 0.0023 0.0022
Min -0.0073 -0.0128 -0.0080
Max 0.0099 0.0092 0.0148
Difference -0.0016 *** 0.0002 * -0.0003 ***
t-statistic (-12.99) (1.78) (-3.17)

Panel C Mean Returns for Stocks with a Positive
Weekend Effect, by Day

Mean 0.0014 -0.0007 0.0002
SD 0.0025 0.0022 0.0022
Min -0.0066 -0.0122 -0.0110
Max 0.0174 0.0129 0.0105
Difference 0.0006 *** 0.0027 *** 0.0018 ***
t-statistic (8.57) (34.20) (24.99)

Panel D. Mean Returns for the 1,000 Stocks with
the Largest Weekend Effect, by Day

Mean 0.0016 -0.0007 0.0003
SD 0.0026 0.0023 0.0023
Min -0.0054 -0.0122 -0.0096
Max 0.0174 0.0106 0.0098
Difference 0.0011 *** 0.0034 *** 0.0024 ***
t-statistic (11.32) (34.91) (25.86)

Panel E. Mean Returns for the 500 Stocks with the
Largest Weekend Effect, by Day

Mean 0.0019 -0.0008 0.0005
SD 0.0028 0.0026 0.0025
Min -0.0054 -0.0122 -0.0085
Max 0.0150 0.0106 0.0098
Difference 0.0017 *** 0.0044 *** 0.0031 ***
t-statistic (11.56) (28.75) (21.02)

Panel F. Mean Returns for the 602 Stocks with No
Tradable Option, by Day

Mean 0.0009 -0.0003 0.0000
SD 0.0024 0.0019 0.0018
Min -0.0041 -0.0098 -0.0096
Max 0.0174 0.0064 0.0105
Difference 0.0008 *** 0.0020 *** 0.0017 ***
t-statistic (7.01) (18.06) (16.59)

 Thursday Friday

Panel A. Mean Returns for Entire Sample, by Day.

Mean 0.0002 0.0013
SD 0.0020 0.0022
Min -0.011 -0.009
Max 0.0098 0.0114
Number of days 51 51
Difference 0.0012*** N/A
t-statistic (17.67) N/A

Panel B. Mean Returns for Stocks with a Nonpositive
Weekend Effect, by Day

Mean 0.0016 -0.0004
SD 0.0018 0.0017
Min -0.0054 -0.009
Max 0.0098 0.0058
Difference -0.0020 *** N/A
t-statistic (-26.65) N/A

Panel C Mean Returns for Stocks with a Positive
Weekend Effect, by Day

Mean -0.0005 0.0020
SD 0.0018 0.0020
Min -0.0110 -0.0045
Max 0.0064 0.0114
Difference 0.0025 *** N/A
t-statistic (40.40) N/A

Panel D. Mean Returns for the 1,000 Stocks with
the Largest Weekend Effect, by Day

Mean -0.0008 0.0027
SD 0.0019 0.0020
Min -0.011 -0.0028
Max 0.0064 0.0114
Difference 0.0035 *** N/A
t-statistic (48.11) N/A

Panel E. Mean Returns for the 500 Stocks with the
Largest Weekend Effect, by Day

Mean -0.0015 0.0036
SD 0.0021 0.0022
Min -0.0110 -0.0026
Max 0.0064 0.0114
Difference 0.0051 *** N/A
t-statistic (49.55) N/A

Panel F. Mean Returns for the 602 Stocks with No
Tradable Option, by Day

Mean -0.0005 0.0017
SD 0.0015 0.0019
Min -0.0068 -0.0045
Max 0.0034 0.0113
Difference 0.0022 *** N/A
t-statistic (23.79) N/A

*** Significant at the 0.01 level.

* Significant at the 0.10 level.

Table III. Statistics for Short Activity

This table presents statistics for daily short activity for each
day of the week and the difference between Friday's short activity
and the short activity of each day. We define short activity as the
daily short volume divided by the daily total trade volume. Panel A
shows the statistics for short activity for the sample stocks and
the difference between Friday's short activity and the entire
sample. Panels B-F report the results for the subsamples. The
t-statistics test for the significance of the difference between
Friday's short activity and the short activity for each day.

 Monday Tuesday Wednesday

Panel A. Short Activity for the Entire Sample, by Day

Mean 0.1954 0.1937 0.1941
SD 0.0743 0.0726 0.0723
Min 0.0304 0.0348 0.0344
Max 0.6148 0.6080 0.6157
Difference 0.0022 *** 0.0039 *** 0.0035 ***
t-statistic (5.34) (9.97) (8.90)

Panel B. Short Activity for Stocks with a Nonpositive
Weekend Effect, by Day

Mean 0.1914 0.1896 0.1906
SD 0.0714 0.0697 0.0706
Min 0.0304 0.0365 0.0425
Max 0.6041 0.6080 0.6157
Difference 0.0015 ** 0.0033 *** 0.0023 ***
t-statistic (2.13) (4.78) (3.18)

Panel C. Short Activity for Stocks with a Positive
Weekend Effect, by Day

Mean 0.1971 0.1954 0.1956
SD 0.0754 0.0737 0.0730
Min 0.0328 0.0348 0.0344
Max 0.6148 0.5882 0.6105
Difference 0.0025 *** 0.0042 *** 0.0040 ***
t-statistic (4.96) (8.76) (8.58)

Panel D. Short Activity for the 1,000 Stocks with
the Largest Weekend Effect, by Day

Mean 0.2047 0.2032 0.2031
SD 0.0713 0.0699 0.0690
Min 0.0343 0.0362 0.0422
Max 0.6148 0.5881 0.6105
Difference 0.0032 *** 0.0047 *** 0.0048 ***
t-statistic (5.19) (7.81) (8.43)

Panel E. Short Activity for the 500 Stocks with
the Largest Weekend Effect, by Day

Mean 0.2084 0.2067 0.2061
SD 0.0626 0.0607 0.0600
Min 0.0343 0.0486 0.0521
Max 0.5651 0.5230 0.4878
Difference 0.0035 *** 0.0052 *** 0.0057 ***
t-statistic (3.91) (5.90) (6.65)

Panel F. Short Activity for the 602 Stocks with
No Tradable Option, by Day

Mean 0.1757 0.1750 0.1742
SD 0.0911 0.0892 0.0876
Min 0.0328 0.0348 0.0344
Max 0.5651 0.5230 0.4878
Difference 0.0042 *** 0.0049 *** 0.0057 ***
t-statistic (4.55) (5.45) (6.82)

 Thursday Friday

Panel A. Short Activity for the Entire Sample, by Day

Mean 0.1962 0.1976
SD 0.0736 0.0728
Min 0.0283 0.0315
Max 0.6165 0.6169
Difference 0.0014 *** N/A
t-statistic (3.72) N/A

Panel B. Short Activity for Stocks with a Nonpositive
Weekend Effect, by Day

Mean 0.1958 0.1929
SD 0.0717 0.0713
Min 0.0318 0.0315
Max 0.6165 0.6169
Difference -0.0030 *** N/A
t-statistic (3.72) N/A

Panel C. Short Activity for Stocks with a Positive
Weekend Effect, by Day

Mean 0.1963 0.1996
SD 0.0744 0.0734
Min 0.0283 0.0355
Max 0.6073 0.6056
Difference 0.0033 *** N/A
t-statistic (7.04) N/A

Panel D. Short Activity for the 1,000 Stocks with
the Largest Weekend Effect, by Day

Mean 0.2032 0.2079
SD 0.0708 0.0695
Min 0.0283 0.0355
Max 0.6073 0.6056
Difference 0.0047 *** N/A
t-statistic (8.19) N/A

Panel E. Short Activity for the 500 Stocks with
the Largest Weekend Effect, by Day

Mean 0.2054 0.2119
SD 0.0631 0.0601
Min 0.0283 0.0355
Max 0.5754 0.4956
Difference 0.0065 *** N/A
t-statistic (7.54) N/A

Panel F. Short Activity for the 602 Stocks with
No Tradable Option, by Day

Mean 0.1758 0.1799
SD 0.0897 0.0893
Min 0.0290 0.0355
Max 05754 0.4977
Difference 0.0041 *** N/A
t-statistic (4.53) N/A

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

Table IV. Regression Results from Short-Sale Volume on Stock
Characteristics and Day of the Week

This table presents the results from estimating the following
model.

[SASV.sub.i,t] = [[beta].sub.0] + [[beta].sub.1] [return.sub.i,t] +
[[beta].sub.2] ln([volume.sub.i,t]) +
[[beta].sub.3][volatility.sub.i,t] + [[delta].sub.2] [TUES.sub.i,t]
+ [[delta].sub.3][THUR.sub.i,t] + [[delta].sub.4][FRI.sub.i,t]

The dependent variable, [SASV.sub.i,t] is the standardized abnormal
short volume for day t and stock i. We subtract the mean of the
short volume for stock i from the average daily short volume for
stock i and then divide the result by the standard deviation of the
short volume for stock i. Here, [Return.sub.i,t] is the daily
return of stock i on day t, ln([volume.sub.i,t]) is the natural log
of the average daily volume for stock i on day t, and
ln([volatility.sub.i,t]) is the natural log of the standard
deviation of the daily high price less the daily low price for
day-of-the-week t and stock i. We also specify the day-of-the-week
dummy variables for stock i. We report P-values in parentheses.

 Intercept return In(volume)

Panel A. Regression Results for the Entire Sample of Stocks

FE estimate 0.0267 0.2345 0.0939
p-value (0.2159) (<0.0001) (<0.0001)

Panel B. Regression Results for Stocks with a
Nonpositive Weekend Effect

FE estimate 0.3250 0.2512 0.0812
p-value (<0.0001) (<0.0001) (<0.0001)

Panel C. Regression Results for Stocks with a Positive
Weekend Effect

FE estimate -0.1152 0.2269 0.1004
p-value (<0.0001) (<0.0001) (<0.0001)

Panel D. Regression Results for the 1,000 Stocks
with the Largest Weekend Effect

FE estimate -0.0748 0.2279 0.1143
p-value (0.0204) (<.0001) (<0.0001)

Panel E. Regression Results for the 500 Stocks with
the Largest Weekend Effect

FE estimate 0.0194 0.2295 0.1216
p-value (0.6753) (<0.0001) (<0.0001)

Panel F. Regression Results for Stocks without
a Tradable Option

FE estimate -1.5186 0.1858 0.2054
p-value (<0.0001) (<0.0001) (<0.0001)

 In(volatility) MON TUES

Panel A. Regression Results for the Entire Sample of Stocks

FE estimate -0.0418 -0.0841 -0.0025
p-value (<0.0001) (<0.0001) (0.6803)

Panel B. Regression Results for Stocks with a
Nonpositive Weekend Effect

FE estimate -0.0243 -0.1053 -0.0178
p-value (0.0003) (<0.0001) (0.1077)

Panel C. Regression Results for Stocks with a Positive
Weekend Effect

FE estimate -0.0494 -0.0750 0.0042
p-value (<0.0001) (<0.0001) (0.5668)

Panel D. Regression Results for the 1,000 Stocks
with the Largest Weekend Effect

FE estimate 0.0246 -0.0667 0.0118
p-value (<0.0001) (<0.0001) (0.1848)

Panel E. Regression Results for the 500 Stocks with
the Largest Weekend Effect

FE estimate 0.0971 -0.0521 0.0135
p-value (<0.0001) (<0.0001) (0.2813)

Panel F. Regression Results for Stocks without
a Tradable Option

FE estimate -0.1064 -0.0268 0.0217
p-value (<0.0001) (0.0234) (0.0619)

 THUR FRI

Panel A. Regression Results for the Entire Sample of Stocks

FE estimate -0.0050 -0.0643
p-value (0.4129) (<0.0001)

Panel B. Regression Results for Stocks with a
Nonpositive Weekend Effect

FE estimate 0.0382 -0.0864
p-value (0.0004) (<0.0001)

Panel C. Regression Results for Stocks with a Positive
Weekend Effect

FE estimate -0.0256 -0.0551
p-value (0.0004) (<.0001)

Panel D. Regression Results for the 1,000 Stocks
with the Largest Weekend Effect

FE estimate -0.0304 -0.0427
p-value (0.0006) (<.001)

Panel E. Regression Results for the 500 Stocks with
the Largest Weekend Effect

FE estimate -0.0374 -0.0331
p-value (0.0028) (0.0057)

Panel F. Regression Results for Stocks without
a Tradable Option

FE estimate -0.0169 -0.0271
p-value (0.1485) (0.0169)

 [R.sup.2] Fixed
 Effects

Panel A. Regression Results for the Entire Sample of Stocks

FE estimate 0.0961 Yes
p-value

Panel B. Regression Results for Stocks with a
Nonpositive Weekend Effect

FE estimate 0.1068 Yes
p-value

Panel C. Regression Results for Stocks with a Positive
Weekend Effect

FE estimate 0.0923 Yes
p-value

Panel D. Regression Results for the 1,000 Stocks
with the Largest Weekend Effect

FE estimate 0.1063 Yes
p-value

Panel E. Regression Results for the 500 Stocks with
the Largest Weekend Effect

FE estimate 0.1203 Yes
p-value

Panel F. Regression Results for Stocks without
a Tradable Option

FE estimate 0.0950 Yes
p-value

Table V. Weekend Effect

This table presents the calculated value of the weekend effect. We
define the weekend effect as Friday's return less the following
Monday's return. Further, if there is no trading on Monday, the
weekend effect is Friday's return less the following Tuesday's
return. We sort the weekend effect by Monday's short activity
deciles. Panel A reports the weekend effect across Monday's short
activity deciles for the entire sample. Panels B-F report the
results for the subsamples.

 All SA Low [2] [3]

Panel A. Weekend Effect by Short-Activity
Deciles for the Entire Sample

Mean 0.0014 0.0008 0.0012 0.0013
SD 0.0031 0.0018 0.0028 0.0033
Min -0.0125 -0.0052 -0.0058 -0.0077
Max 0.0140 0.0103 0.0125 0.0135

Panel B. Weekend Effect by Short-Activity Deciles
for Stocks with a Nonpositive Weekend Effect

Mean -0.0019 -0.0011 -0.0014 -0.0018
SD 0.0018 0.0011 0.0012 0.0016
Min -0.0125 -0.0052 -0.0058 -0.0078
Max 0.0000 0.0000 0.0000 0.0000

Panel C. Weekend Effect by Short-Activity Deciles
for Stocks with a Positive Weekend Effect

Mean 0.0028 0.0016 0.0025 0.0030
SD 0.0024 0.0014 0.0024 0.0027
Min 0.0000 0.0000 0.0001 0.0001
Max 0.0140 0.0103 0.0125 0.0135

Panel D. Weekend Effect by Short-Activity Deciles
for the 1,000 Stocks with the Largest Weekend Effect

Mean 0.0039 0.0029 0.0042 0.0041
SD 0.0023 0.0018 0.0027 0.0021
Min 0.0015 0.0015 0.0015 0.0015
Max 0.0140 0.0104 0.0135 0.0125

Panel E. Weekend Effect by Short-Activity Deciles
for the 500 Stocks with the Largest Weekend Effect

Mean 0.0055 0.0057 0.0059 0.0054
SD 0.0022 0.0022 0.0026 0.0018
Min 0.0032 0.0034 0.0032 0.0032
Max 0.0140 0.0125 0.0135 0.0107

Panel F. Weekend Effect by Short-Activity Deciles
for the 602 Stocks with No Tradable Options

Mean 0.0026 0.0012 0.0016 0.0022
SD 0.0022 0.0009 0.0012 0.0019
Min 0.0000 0.0000 0.0001 0.0002
Max 0.0128 0.0038 0.0057 0.0104

 [4] [5] [6] [7]

Panel A. Weekend Effect by Short-Activity
Deciles for the Entire Sample

Mean 0.0015 0.0018 0.0012 0.0013
SD 0.0034 0.0035 0.0032 0.0035
Min -0.0074 -0.0091 -0.0083 -0.0125
Max 0.0140 0.0132 0.0120 0.0119

Panel B. Weekend Effect by Short-Activity Deciles
for Stocks with a Nonpositive Weekend Effect

Mean -0.0021 -0.0021 -0.0019 -0.0023
SD 0.0017 0.0019 0.0020 0.0023
Min -0.0074 -0.0091 -0.0083 -0.0125
Max -0.0001 0.0000 0.0000 -0.0001

Panel C. Weekend Effect by Short-Activity Deciles
for Stocks with a Positive Weekend Effect

Mean 0.0031 0.0034 0.0027 0.0033
SD 0.0025 0.0028 0.0023 0.0026
Min 0.0001 0.0000 0.0000 0.0000
Max 0.0140 0.0132 0.0120 0.0119

Panel D. Weekend Effect by Short-Activity Deciles
for the 1,000 Stocks with the Largest Weekend Effect

Mean 0.0045 0.0039 0.0041 0.0040
SD 0.0027 0.0022 0.0024 0.0024
Min 0.0015 0.0015 0.0015 0.0015
Max 0.0140 0.0132 0.0120 0.0119

Panel E. Weekend Effect by Short-Activity Deciles
for the 500 Stocks with the Largest Weekend Effect

Mean 0.0058 0.0053 0.0058 0.0056
SD 0.0026 0.0021 0.0024 0.0024
Min 0.0032 0.0032 0.0032 0.0032
Max 0.0140 0.0132 0.0120 0.0119

Panel F. Weekend Effect by Short-Activity Deciles
for the 602 Stocks with No Tradable Options

Mean 0.0028 0.0031 0.0034 0.0031
SD 0.0027 0.0027 0.0031 0.0025
Min 0.0001 0.0001 0.0001 0.0000
Max 0.0125 0.0128 0.0125 0.0104

 [8] [9] SA High

Panel A. Weekend Effect by Short-Activity
Deciles for the Entire Sample

Mean 0.0017 0.0019 0.0014
SD 0.0031 0.0033 0.0027
Min -0.0078 -0.0071 -0.0095
Max 0.0118 0.0130 0.0101

Panel B. Weekend Effect by Short-Activity Deciles
for Stocks with a Nonpositive Weekend Effect

Mean -0.0019 -0.0023 -0.0017
SD 0.0019 0.0016 0.0018
Min -0.0084 -0.0071 -0.0095
Max 0.0000 0.0000 0.0000

Panel C. Weekend Effect by Short-Activity Deciles
for Stocks with a Positive Weekend Effect

Mean 0.0030 0.0032 0.0026
SD 0.0024 0.0023 0.0019
Min 0.0000 0.0000 0.0000
Max 0.0130 0.0103 0.0101

Panel D. Weekend Effect by Short-Activity Deciles
for the 1,000 Stocks with the Largest Weekend Effect

Mean 0.0039 0.0040 0.0034
SD 0.0023 0.0022 0.0016
Min 0.0015 0.0015 0.0015
Max 0.0130 0.0102 0.0083

Panel E. Weekend Effect by Short-Activity Deciles
for the 500 Stocks with the Largest Weekend Effect

Mean 0.0054 0.0055 0.0048
SD 0.0024 0.0019 0.0015
Min 0.0032 0.0032 0.0033
Max 0.0130 0.0102 0.0101

Panel F. Weekend Effect by Short-Activity Deciles
for the 602 Stocks with No Tradable Options

Mean 0.0027 0.0028 0.0031
SD 0.0015 0.0020 0.0021
Min 0.0000 0.0000 0.0001
Max 0.0080 0.0089 0.0083

Table VI. Spearman Correlation Tests

This table reports the Spearman correlation coefficients for our
measure of short-selling activity and returns, by day of the week.
Here, Short Act. is the short-activity ratio, which we define as
the daily short volume divided by the daily total trade volume. We
report the correlation coefficients between the short-activity
ratio and returns for each day of the week. Panel A reports the
correlation coefficients, with the corresponding p-values, for the
entire sample. Panels B-F report the results for the subsamples.

 Monday Tuesday Wednesday

Panel A. Correlation Tests between Shorting Activity
and Returns for Entire Sample, by Day

Short Act._returns 0.14337 0.10900 0.12625
p-value (<0.0001) (<0.0001) (<0.0001)

Panel B. Correlation Tests between Shorting Activity
and Returns, for Stocks with a Nonpositive
Weekend Effect, by Day

Short Act._returns 0.13964 0.11424 0.11776
p-value (<0.0001) (<0.0001) (<0.0001)

Panel C. Correlation Tests between Shorting Activity
and Returns for Stocks with a Positive
Weekend Effect, by Day

Short Act._returns 0.14488 0.10689 0.12984
p-value (<0.0001) (<0.0001) (<0.0001)

Panel D. Correlation Tests between Shorting Activity
and Returns for the 1,000 Stocks with the Largest
Weekend Effect, by Day

Short Act._returns 0.14498 0.11593 0.13639
p-value (<0.0001) (<0.0001) (<0.0001)

Panel E. Correlation Tests between Shorting Activity
and Returns for the 500 Stocks with the Largest
Weekend Effect, by Day

Short Act._returns 0.14931 0.13288 0.15060
p-value (<0.0001) (<0.0001) (<0.0001)

Panel F. Correlation Tests between Shorting Activity
and Returns for the 602 Stocks with No
Tradable Option, by Day

Short Act. _returns 0.15029 0.10135 0.12430
p-value (<0.0001) (<0.0001) (<0.0001)

 Thursday Friday

Panel A. Correlation Tests between Shorting Activity
and Returns for Entire Sample, by Day

Short Act._returns 0.11601 0.11581
p-value (<0.0001) (<0.0001)

Panel B. Correlation Tests between Shorting Activity
and Returns, for Stocks with a Nonpositive
Weekend Effect, by Day

Short Act._returns 0.11651 0.10438
p-value (<0.0001) (<0.0001)

Panel C. Correlation Tests between Shorting Activity
and Returns for Stocks with a Positive
Weekend Effect, by Day

Short Act._returns 0.11594 0.11923
p-value (<0.0001) (<0.0001)

Panel D. Correlation Tests between Shorting Activity
and Returns for the 1,000 Stocks with the Largest
Weekend Effect, by Day

Short Act._returns 0.12405 0.11581
p-value (<0.0001) (<0.0001)

Panel E. Correlation Tests between Shorting Activity
and Returns for the 500 Stocks with the Largest
Weekend Effect, by Day

Short Act._returns 0.14275 0.11830
p-value (<0.0001) (<0.0001)

Panel F. Correlation Tests between Shorting Activity
and Returns for the 602 Stocks with No
Tradable Option, by Day

Short Act. _returns 0.10796 0.12052
p-value (<0.0001) (<0.0001)

Table VII. Regression Results

This table reports the results from a regression of returns on
day-of-the-week dummy variables and five interaction variables. We
interact the short-activity ratio for stock i on day t,
[SA.sub.i,t], with the day dummy variables. We use both one- and
two-way fixed effects (by stock and by day) and find the results
are similar, so we only report the one-way fixed effects results.
The model is specified below:

[Ret.sub.i,t] = [[delta].sub.1][MON.sub.i,t] +
[[delta].sub.2][TUES.sub.i,t] + [[delta].sub.3][WED.sub.i,t] +
[[delta].sub.4][THUR.sub.i,t] + [[delta].sub.5][FRI.sub.i,t] +
[[delta].sub.6][MON.sub.i,t] x [SA.sub.i,t] + [[delta].sub.7][TUES
+ [[delta].sub.10][FRI.sub.i,t] x [SA.sub.i,t] +
[[epsilon].sub.i,t].

Panel A reports the results from the fixed-effects regression for
the entire sample. Panels B-F report the results for the
subsamples. We report the t-statistics in parentheses.

 MON TUES WED

Panel A. Regression Results for the Entire Sample

Estimate -0.0020 *** -0.0033 *** -0.0030 ***
t-statistic (-21.27) (-37.61) (-34.35)

Panel B. Regression Results for Stocks with a Nonpositive
Weekend Effect

Estimate -0.0020 *** -0.0033 *** -0.0029 ***
t-statistic (-11.83) (-20.80) (-18.64)

Panel C. Regression Results for the Stocks with a Positive
Weekend Effect

Estimate -0.0020 *** -0.0033 *** -0.0030 ***
t-statistic (-17.68) (-31.40) (-28.87)

Panel D. Regression Results for the 1,000 Stocks with the
Largest Weekend Effect

Estimate -0.0022 *** -0.0039 *** -0.0036 ***
t-statistic (-14.38) (-27.37) (-24.97)

Panel E. Regression Results for the 500 Stocks with the
Largest Weekend Effect

Estimate -0.0029 *** -0.0053 *** -0.0047 ***
t-statistic (-11.23) (-21.82) (-19.29)

Panel F. Regression Results for the 602 Stocks with No
Tradable Option

Estimate -0.0015 *** -0.0019 *** -0.0021 ***
t-statistic (-11.53) (-15.38) (-17.03)

 THUR FRI MON x SA

Panel A. Regression Results for the Entire Sample

Estimate -0.0029 *** -0.0014 *** 0.0168 ***
t-statistic (-32.06) (-15.70) (43.03)

Panel B. Regression Results for Stocks with a Nonpositive
Weekend Effect

Estimate -0.0014 *** -0.0027 *** 0.0165 ***
t-statistic (-8.94) (-17.17) (23.03)

Panel C. Regression Results for the Stocks with a Positive
Weekend Effect

Estimate -0.0346 *** -0.0007 *** 0.0169 ***
t-statistic (-32.51) (-7.20) (36.37)

Panel D. Regression Results for the 1,000 Stocks with the
Largest Weekend Effect

Estimate -0.0044 *** -0.0004 *** 0.0183 ***
t-statistic (-30.87) (-2.72) (29.54)

Panel E. Regression Results for the 500 Stocks with the
Largest Weekend Effect

Estimate -0.0066 *** -0.0003 0.0228 ***
t-statistic (-27.28) (-1.19) (21.49)

Panel F. Regression Results for the 602 Stocks with No
Tradable Option

Estimate -0.0023 *** -0.0002 * 0.0136 ***
t-statistic (-18.51) (-1.67) (24.64)

 TUES x SA WED x SA THUR x SA

Panel A. Regression Results for the Entire Sample

Estimate 0.0137 *** 0.0163 *** 0.0152 ***
t-statistic (36.88) (43.96) (41.00)

Panel B. Regression Results for Stocks with a Nonpositive
Weekend Effect

Estimate 0.0144 *** 0.0153 *** 0.0154 ***
t-statistic (21.08) (22.64) (22.92)

Panel C. Regression Results for the Stocks with a Positive
Weekend Effect

Estimate 0.0134 *** 0.0166 *** 0.0151 ***
t-statistic (30.35) (37.73) (34.00)

Panel D. Regression Results for the 1,000 Stocks with the
Largest Weekend Effect

Estimate 0.0156 *** 0.0190 *** 0.0177 ***
t-statistic (26.57) (32.35) (29.98)

Panel E. Regression Results for the 500 Stocks with the
Largest Weekend Effect

Estimate 0.0215 *** 0.0251 *** 0.0248 ***
t-statistic (21.37) (24.91) (24.75)

Panel F. Regression Results for the 602 Stocks with No
Tradable Option

Estimate 0.0093 *** 0.0121 *** 0.0104 ***
t-statistic (17.65) (23.03) (19.76)

 FRI x SA [R.sup.2] Fixed
 Effects
Panel A. Regression Results for the Entire Sample

Estimate 0.0136 *** 0.0175 Yes
t-statistic (37.08)

Panel B. Regression Results for Stocks with a Nonpositive
Weekend Effect

Estimate 0.0122 *** 0.0170 Yes
t-statistic (18.23)

Panel C. Regression Results for the Stocks with a Positive
Weekend Effect

Estimate 0.0140 *** 0.0179 Yes
t-statistic (31.92)

Panel D. Regression Results for the 1,000 Stocks with the
Largest Weekend Effect

Estimate 0.0148 *** 0.0229 Yes
t-statistic (25.42)

Panel E. Regression Results for the 500 Stocks with the
Largest Weekend Effect

Estimate 0.0183 *** 0.0286 Yes
t-statistic (18.52)

Panel F. Regression Results for the 602 Stocks with No
Tradable Option

Estimate 0.0108 *** 0.0184 Yes
t-statistic (20.71)

*** Significant at the 0.01 level.

* Significant at the 0.10 level.

Table VIII. Regression Results of High and Low Relative
Short-Interest Quartiles

This table reports the results from replicating Chen and Singal
(2003). We divide the short-interest sample (1,806 stocks) into
size deciles at the beginning of the month and then subdivide
stocks into quartiles based on relative short interest. We regress
returns on day-of-the-week dummy variables and five interaction
dummy variables for the highest and lowest quartiles. We define H
as equal to unity if the stock i on day t is in the highest RSI
quartile. We interact the dummy variable H with the day dummy
variables following Chen and Singal.

[Ret.sub.i,t] = [[delta].sub.1][MON.sub.i,t] +
[[delta].sub.2][TUES.sub.i,t] + [[delta].sub.3][WED.sub.i,t] +
[[delta].sub.4][THUR.sub.i,t] + [[delta].sub.5][FRI.sub.i,t] +
[[delta].sub.6][MON.sub.i,t] x [H.sub.i,t] + [[delta].sub +
[[delta].sub.10][FRI.sub.i,t] x [H.sub.i,t] + [[epsilon].sub.i,t].
(9)

Panel A reports the results from the regression for the entire
short interest sample. Panels B-F report the results for the
subsamples. We report the t-statistics in parentheses.

 MON TUES WED

Panel A. Regression Results for the Entire Short-Interest
Sample (1,806 Stocks)

Estimate 0.0014 *** -0.0004 *** 0.0003 ***
t-statistic (10.68) (-3.42) (2.67)

Panel B. Regression Results for Stocks with a Nonpositive
Weekend Effect (552 Stocks)

Estimate 0.0011 * -0.0003 0.0002
t-statistic (4.49) (-1.33) (0.92)

Panel C. Regression Results for the Stocks with a Positive
Weekend Effect (1,254 Stocks)

Estimate 0.0015 * -0.0005 * 0.0004 *
t-statistic (9.81) (-3.22) (-4.28)

Panel D. Regression Results for the 1,000 Stocks with the
Largest Weekend Effect (897 Stocks)

Estimate 0.0016 * -0.0005 * 0.0003 **
t-statistic (8.70) (-2.93) (1.96)

Panel E. Regression Results for the 500 Stocks with the
Largest Weekend Effect (471 Stocks)

Estimate 0.0017 * -0.0007 ** 0.0008 *
t-statistic (6.13) (-2.51) (2.91)

Panel F. Regression Results for the 602 Stocks with
No Tradable Option (381 Stocks)

Estimate 0.0014 * -0.0005 ** 0.0003 *
t-statistic (6.66) (-2.36) (1.70)

 THUR FRI MON x H

Panel A. Regression Results for the Entire Short-Interest
Sample (1,806 Stocks)

Estimate 0.0001 0.0016 *** 0.0005 ***
t-statistic (0.038) (12.69) (2.92)

Panel B. Regression Results for Stocks with a Nonpositive
Weekend Effect (552 Stocks)

Estimate 0.0017 * -0.0002 0.0005
t-statistic (7.39) (-0.86) (1.65)

Panel C. Regression Results for the Stocks with a Positive
Weekend Effect (1,254 Stocks)

Estimate -0.0006 * 0.0023 * 0.0006 *
t-statistic -2.58 (15.63) (2.58)

Panel D. Regression Results for the 1,000 Stocks with the
Largest Weekend Effect (897 Stocks)

Estimate -0.0010 * 0.0029 * 0.0005 *
t-statistic (-5.40) (16.39) -1.69

Panel E. Regression Results for the 500 Stocks with the
Largest Weekend Effect (471 Stocks)

Estimate -0.0015 * 0.0037 * 0.0007
t-statistic (-5.78) (13.82) (1.61)

Panel F. Regression Results for the 602 Stocks with
No Tradable Option (381 Stocks)

Estimate -0.0008 * 0.0020 * 0.0005
t-statistic (-4.10) (10.25) (1.17)

 TUES x H WED x H THUR x H

Panel A. Regression Results for the Entire Short-Interest
Sample (1,806 Stocks)

Estimate -0.0007 *** -0.0000 0.0002
t-statistic (-3.92) (-0.12) (1.24)

Panel B. Regression Results for Stocks with a Nonpositive
Weekend Effect (552 Stocks)

Estimate -0.0006 * -0.0003 0.0000
t-statistic (-1.94) (-0.83) (0.14)

Panel C. Regression Results for the Stocks with a Positive
Weekend Effect (1,254 Stocks)

Estimate -0.0007 * 0.0001 0.0001
t-statistic (-3.52) (0.60) (0.31)

Panel D. Regression Results for the 1,000 Stocks with the
Largest Weekend Effect (897 Stocks)

Estimate -0.0005 ** 0.0002 -0.0000
t-statistic (-2.28) (0.92) (-0.14)

Panel E. Regression Results for the 500 Stocks with the
Largest Weekend Effect (471 Stocks)


Estimate -0.0008 ** -0.0001 -0.0003
t-statistic (-2.01) (-0.24) (-0.86)

Panel F. Regression Results for the 602 Stocks with
No Tradable Option (381 Stocks)

Estimate -0.0000 -0.0004 0.0002
t-statistic (-0.08) (-1.00) (0.56)

 F-Value
 FRI x H [R.sup.2] (p-value)

Panel A. Regression Results for the Entire Short-Interest
Sample (1,806 Stocks)

Estimate -0.0004 ** 0.0031 69.61
t-statistic (-2.04) (<.0001)

Panel B. Regression Results for Stocks with a Nonpositive
Weekend Effect (552 Stocks)

Estimate -0.0004 0.0031 23.12
t-statistic (-1.31) (<.0001)

Panel C. Regression Results for the Stocks with a Positive
Weekend Effect (1,254 Stocks)

Estimate -0.0001 0.0054 83.21
t-statistic (-0.31) (<.0001)

Panel D. Regression Results for the 1,000 Stocks with the
Largest Weekend Effect (897 Stocks)

Estimate -0.0001 0.0072 78.25
t-statistic (-0.54) (<.0001)

Panel E. Regression Results for the 500 Stocks with the
Largest Weekend Effect (471 Stocks)

Estimate -0.0001 0.0095 56.71
t-statistic (-0.29) (<.0001)

Panel F. Regression Results for the 602 Stocks with
No Tradable Option (381 Stocks)

Estimate 0.0000 0.0052 23.86
t-statistic (0.07) (<.0001)

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table IX. Regression Results

This table presents the results from a one-way fixed effects
regression of returns on day-of-the-week dummy variables, a dummy
variable that captures the returns at the end of the month, and
five three-way interaction variables. Here, M E is a dummy variable
equal to one if the day is in the last two weeks of the month, and
SA is the short-activity ratio for stock i on day t. We use a
three-way interaction to control for the possibility that the
month-end drives the results of the weekend effect following Wang,
Li, and Erickson (1997). The model is specified below:

[Ret.sub.i,t] [[delta].sub.1]MON + [[delta].sub.2]TUES +
[[delta].sub.3]WED + [[delta].sub.4]THUR + [[delta].sub.5]FRI +
[[delta].sub.6]M_E + [[delta].sub.7]MON x SA x M_E +
[[delta].sub.8]TUES x SA x M_E + [[delta].sub.9]WED x SA x M_E +
[[delta].sub.10]THUR x SA x M_E + [[delta].sub.11]FRI x SA x M_E +
[[epsilon].sub.i,t].

We perform a one-way (by stock) and a two-way (by stock and date)
fixed effects and find the results to be similar, so we only report
the one-way results. We perform the analysis for the entire sample
as well as the subsamples, respectively. We report the t-statistics
in parentheses.

 All Stocks Nonpositive Positive
 Weekend Weekend
 Effect Stocks Effect Stocks

MON 0.0007 *** 0.0008 *** 0.0007 ***
 (10.96) (6.32) (8.95)
TUES -0.0002 *** -0.0001 -0.0003 ***
 (-3.450) (-0.96) (-3.55)
WED -0.0008 *** -0.0009 *** -0.0008 ***
 (-12.54) (-7.81) (-9.92)
THUR -0.0001 0.0015 *** -0.0007 ***
 (-0.85) (12.48) (-9.29)
FRI 0.0015 *** 0.0001 0.0021 ***
 (22.63) (0.73) -26.76
M_E -0.0025 *** -0.0027 *** -0.0024 ***
 (-39.24) (-23.62) (-31.33)
MON x SA x M_E 0.0186 *** 0.0185 *** 0.0186 ***
 (44.19) (23.79) (37.20)
TUES x SA x M_E 0.0091 *** 0.0102 *** 0.0086 ***
 (22.68) (13.74) (18.08)
WED x SA x M_E 0.0224 *** 0.0232 *** 0.0221 ***
 (56.45) (31.99) (46.51)
THUR x SA x M_E 0.0148 *** 0.0151 *** 0.0147 ***
 (37.24) (20.98) (30.74)
FRI x SA x M_E 0.0113 *** 0.0099 *** 0.0117 ***
 (28.41) (13.58) (24.77)
[R.sup.2] 0.0122 0.0127 0.0142
Fixed effects Yes Yes Yes

 1,000 Stocks 500 Stocks Nonoptionable
 with Largest with Largest Stocks
 Weekend Effect Weekend Effect

MON 0.0009 *** 0.0010 *** 0.0005 ***
 (8.28) (6.05) (4.41)
TUES -0.0004 *** -0.0006 *** 0.0001
 (-3.37) (-3.71) (0.85)
WED -0.0009 *** -0.0009 *** -0.0008 ***
 (-8.28) (-11.84) (-7.31)
THUR -0.0012 *** -0.0019 *** -0.0007 ***
 (-11.27) (-11.84) (-6.52)
FRI 0.0028 *** 0.0038 *** 0.0016 ***
 (27.000 (22.83) (15.51)
M_E -0.0028 *** -0.0036 *** -0.0015 ***
 (-26.69) (-20.92) (-15.43)
MON x SA x M_E 0.0200 *** 0.0249 *** 0.0130 ***
 (30.62) (22.98) (20.21)
TUES x SA x M_E 0.0101 *** 0.0156 *** 0.0046 ***
 (16.28) (15.01) (7.48)
WED x SA x M_E 0.0244 *** 0.0307 *** 0.0169 ***
 (39.30) (29.68) (27.61)
THUR x SA x M_E 0.0166 *** 0.0220 *** 0.0106 ***
 (26.65) (21.22) (17.33)
FRI x SA x M_E 0.0126 *** 0.0155 *** 0.0096 ***
 (20.41) (15.16) (15.82)
[R.sup.2] 0.0167 0.0206 0.0127
Fixed effects Yes Yes Yes

*** Significant at the 0.01 level.
COPYRIGHT 2009 Financial Management Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2009 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:New York Stock Exchange
Author:Blau, Benjamin M.; Van Ness, Bonnie F.; Van Ness, Robert A.
Publication:Financial Management
Article Type:Report
Date:Sep 22, 2009
Words:14405
Previous Article:Local effects of foreign ownership in an emerging financial market: evidence from qualified foreign institutional investors in Taiwan.
Next Article:Derivatives use, information asymmetry, and MNC post-acquisition performance.
Topics:


Related Articles
Daiwa Securities America fined $400,000 for short-sale rule violation.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |