the impact of institutional investors on the monday seasonal - chan et al 04

8/8/2019 The Impact of Institutional Investors on the Monday Seasonal - Chan Et Al 04

1/20

Su Han Chan Department of Finance, CaliforniaState University-Fullerton

Wai-Kin Leung Faculty of Business Administration,Chinese University of Hong Kong

Ko Wang Department of Finance, CaliforniaState University-Fullerton

The Impact of InstitutionalInvestors on the MondaySeasonal *

I. Introduction

One of the most noticeable anomalies docu-

mented in the nance literature is the Mondayseasonal. This anomalous Monday return patternis observed not only in stock markets in theU. S. (see, for example, French (1980)) and other parts of the world (see, for example, Jaffe,Westereld, and Ma (1989)), but also in debt markets (see, for example, Flannery and Proto- papadakis (1988)).

Numerous explanations have been developed

to rationalize the puzzling observations that themean return on Monday is signicantly negativeand is lower than that on other weekdays (i.e., theMonday seasonal). One of the most plausible

( Journal of Business , 2004, vol. 77, no. 4) B 2004 by The University of Chicago. All rights reserved.0021-9398/2004/7704-0012$10.00

967

* The authors acknowledge helpful comments from Kalok Chan, John Erickson, Tsong Yue Lai, Laura Starks, John Wei,and seminar participants at the Hong Kong University of Scienceand Technology. The authors are indebted to an anonymousreferee, whose detailed comments have signicantly improvedthe paper. The usual disclaimer applies.

It is well documentedthat the mean Mondayreturn is signicantlynegative and is lower than the mean return onother weekdays. Usinginstitutional stock holdings informationduring the 19811998 period, we document that the Monday sea-sonal is stronger instocks with low institu-tional holdings and that the Monday return is

not signicantly differ-ent from the meanTuesday to Fridayreturns for stocks withhigh institutional hold-ings during the 1990 1998 period. Our study provides direct evi-dence to support the belief that the Mondayseasonal may be related

to the trading activitiesof less sophisticatedindividual investors.


2/20

explanations is the trading pattern of individual investors. 1 Accordingto Miller (1988) and Lakonishok and Maberly (1990), individualinvestors typically do not have time during the weekday trading hoursand therefore process information and make investment decisions onlyduring the weekend. In addition, since brokers who tend to issue more buy than sell order recommendations do not work on the weekend,individual investors are less likely to be affected by brokers buy order recommendations when they make their investment decisions duringthe weekend. Lakonishok and Maberly (1990) provide empirical evi-dence to support this view. They nd that when the market reopens onMonday individual investors tend to increase trading activity (espe-cially sell transactions) on that day. Because of the increased sellingactivity by individual investors on Monday relative to the rest of theweek, it is possible that an association exists between the trading be-haviour of individual investors and the Monday seasonal. However,there is yet to be direct evidence indicating that those stocks traded byindividual investors on Monday are actually the stocks that cause theaverage price decline in the stock market on Monday.

Kamara (1997) reports that the Monday seasonal declines over the19621993 period for stocks in the S&P 500 index, but not for stocks inthe index of NYSEs smallest capitalization decile. He believes that thedecline in trading costs and the increased role of institutions in the stock market can explain this phenomenon. However, because Kamarasobservations are based only on the S&P 500 index and the small-capindex (which represent only a small fraction of stocks traded on the NYSE), it is difcult to draw a concrete conclusion as to whether theaverage decline in the stock market on Monday can be attributed tothe lack of institutional ownership in certain stocks. 2

The above is especially true given Sias and Starkss (1995) nding.Using a sample of stocks traded on the NYSE during the 19771991 period, they nd essentially no evidence of differences in Mondayreturns between stocks with high institutional ownership and stockswith low institutional ownership. 3 Furthermore, conditional on the previous Fridays return and controlling for rm size, Sias and Starks

1. Other explanations include settlement period (see Lakonishok and Levi (1982)),measurement error or specialist-related explanation (see Keim and Stambaugh (1984)), in-stitutional factors (see Flannery and Protopapadakis (1988)), and correlation with Fridayreturn (see Damodaran (1989) and Abraham and Ikenberry (1994)).

2. However, analyzing S&P 500 future and spot returns and small-cap returns onMondays, Kamara (1997) makes a convincing argument that informed traders use less costlymarkets to exploit the Monday seasonal.

3. Kamara (1997) uses the ratio of block trading to odd-lot trading to proxy for institu-tional trading while Sias and Stark (1995) use the institutional ownership level of stockstraded on NYSE (reported by Vickers Stock Research Corporation) to conduct their analysis.This might explain the difference in their ndings.

968 Journal of Business


3/20

(1995) nd that the Monday seasonal is more pronounced in stockswith higher institutional ownership. It should be noted, however,that the sample period of Sias and Starkss (1995) study ends in 1991and therefore their result does not fully capture the change in theU.S. equity markets caused by the signicant increase in institutionalholdings in the stock market after 1990. (We document the increase inthe role of institutional investors later in Table 1 of this paper.)

In sum, the results from studies on the relationship between the levelof institutional investors and the magnitude of the Monday seasonalseem to be sample specic and inconsistent with each other. It should be noted that prior studies use small samples and/or rms limited to NYSE. However, it is well known that the Monday seasonal exists not only for stocks traded on the NYSE, but also in other stock markets.For example, Wang, Li and Erickson (1997) document that the Mondayseasonal is more severe in the Nasdaq market than in the NYSE duringthe 19621993 period. In this paper we also show that Nasdaq stocks,which have higher transaction costs and exhibit the strongest Mondayseasonal among the three stock markets in the early years, also haverelatively lower institutional holdings than NYSE and Amex stocksduring the same period. 4 Consequently, it seems necessary to conduct a more comprehensive analysis using rm-specic information on in-stitutional holdings for all stocks traded on the NYSE, Amex, and Nasdaq to conrm whether the trading patterns of individual investorscan explain the magnitude of the Monday seasonal.

This paper examines the relation between the Monday seasonal andthe percentage of institutional holdings using all stocks traded on the NYSE, Amex and Nasdaq over the 19811998 period. We report that the post-1990 period is associated with a signicant increase in insti-tutional holdings in the U.S. stock market and that the magnitude of theMonday seasonal diminished during this period. Although there is a possibility that the less negative mean return on Mondays for the 1990scould be the result of the bull market during this period, this is lessof a concern because we also document that stock portfolios with highinstitutional holdings exhibit a weaker Monday seasonal than stock portfolios with low institutional holdings. In addition, we nd that the mean Monday returns of stock portfolios with high institutionalholdings are mostly positive (and are not signicantly different fromthe mean returns on other weekdays) during the 19901998 period.Our result provides direct evidence that the Monday seasonal is relatedto the trading patterns of less sophisticated individual investors. Thisnding seems to be consistent with the growing body of literature

4. For a partial list of studies on the Nasdaq market, see Huang and Stoll (1996), Chan andLakonishok (1997), Keim and Madhavan (1997), Bessembinder (1999), Christie and Schultz(1999) and Barclay et al. (1999).

969The Impact of Institutional Investors on the Monday Seasonal


4/20

documenting that institutional investors trading strategies affect themovement of security prices. 5

Section II of this paper describes our data gathering procedure whilesection III reports the patterns of institutional holdings during the19811998 period. Section IV provides our empirical ndings. Thelast section contains our conclusions.

II. Data Gathering Procedure

To gather institutional holdings and stock returns data, we rst obtainfrom the Center for Research in Security Prices (CRSP) the perma-nent numbers, CUSIP numbers, and names of all rms traded on the NYSE, Amex and Nasdaq during the 19811998 period. We thenobtain information on the number of shares held by institutionalinvestors from the Spectrum 3:13 ( f ) Institutional Stock HoldingsSurvey during the same period. (This data source is available on theSpectrum tapes provided by Computer Directions Advisors). Finally,we combine data from the Spectrum tapes with data in the CRSP tapesto create the sample for this study. Only rms with both conrmedinstitutional information as well as stock return information are in-cluded in the nal sample. (Appendix A provides details of the dataselection and combination process.)

The number of rms in our sample monotonically increases over time. We have 4,304 rms and 8,569 rms with complete informationin 1981 and 1998, respectively. The percentage of institutional hold-ings is computed as the number of shares held by institutional in-vestors in the second quarter (June 30) of the year (obtained fromSpectrum) divided by the number of shares outstanding at the end of that quarter (obtained from CRSP).

III. Patterns of Institutional HoldingsDuring our sample period, institutional investors have become moredominant players in the stock market in the later years. Except for the19831984 and 19871988 periods, we nd that the percentage of institutional holdings is a monotonically increasing function of time.The mean percentage of institutional holdings of stocks traded in allthe three markets (NYSE, Amex, and Nasdaq) is 14.6% in 1981,21.9% in 1990, and 31.0% in 1998. On the other hand, the percentageof stocks with zero institutional holdings decreases from 11.6% in

5. For a partial list of those studies, see Lakonishok, Shleifer, and Vishny (1992), Wang,Chan, and Gau (1992), Chan and Lakonishok (1993 and 1995), Keim and Madhavan (1995),Grinblatt, Titman, and Wermers (1995), Badrinath, Kale, and Noe (1995), Wang, Erickson,Gau, and Chan (1995), Sias and Starks (1997), and Nofsinger and Sias (1999).



5/20

1981 to 3.7% in 1990, and to 2.4% in 1998. For NYSE rms, theaverage percentage of institutional holdings is 25.7% in 1981 and44.2% in 1998. The mean institutional holdings for Amex and Nasdaqrms are 9.1% and 8.3%, respectively in 1981 and 18.6% and 25.4%,respectively in 1998.

Table 1 presents summary statistics on the distribution of institu-tional holdings for NYSE, Amex and Nasdaq rms during the 1981 1989, 19901998, and 19811998 periods. Panel A of Table 1 reportsthat the percentage of institutional holdings for all rms in all the threestock markets in the 19901998 period (27.0%) is signicantly higher (t-statistics = 65.21) than that in the 19811989 period (18.5%). If we believe that institutional investors are better informed about theMonday seasonal than individual investors are, then our evidencesuggests that stock markets should become more efcient in the recent years.

Panels B, C, D, and E of Table 1 show that NYSE rms on averagehave a signicantly higher percentage of institutional holdings than both Amex rms and Nasdaq rms in the period of analysis. WhenAmex rms are compared to Nasdaq rms, Panel E of Table 1 showsthat Amex rms have a signicantly higher percentage of institu-tional holdings during the 19811989 period, but a signicantly lower percentage of institutional holdings during the 19901998 period.From 1981 to 1998, institutional holdings in Nasdaq stocks show asubstantial increase of over 206% (from 8.3% in 1981 to 25.4% in1998) while the increases for Amex and NYSE stocks are 104% (from9.1% in 1981 to 18.6% in 1998) and 72% (from 25.7% in 1981 to44.2% in 1998), respectively. The signicant change in the level of institutional ownership in the Nasdaq market suggests that an exam-ination of the relationship between institutional investors and price behavior in the Nasdaq market might yield interesting results.

IV. Empirical Findings

A. Patterns of Monday ReturnsTo examine whether the magnitude of the Monday seasonal changeswith the increase in institutional investors in the stock market, wecalculate the mean Monday return and the mean Monday minusTuesday to Friday return by period (19811989, 19901998 and19811998) and by stock market. Panels A, B, C, and D of Table 2report that, regardless of the stock market chosen, the mean Mondayreturn and the mean Monday minus Tuesday to Friday return are higher in the 19901998 sub-period than in the 19811989 sub-period.

While it is possible that the less negative mean Monday return in the19901998 period could be the result of the bull market of the1990s, this market condition should also affect the mean returns on



6/20

other weekdays during the same period as well. Since we also nd that the magnitude of the mean Monday minus Tuesday to Friday return issignicantly reduced in the 19901998 period, it is more likely that the weaker Monday seasonal during this period is associated with anincrease in institutional stock holdings in the markets.

We note that NYSE, which has the highest average percentage of institutional holdings among the three markets during each of the

TABLE 1 Institutional Holdings by Market and by Period for All NYSE, Amex,and Nasdaq Firms, 19811998

Period

198189 199098 198198

Panel A: Institutional Holdings for Firms in All Three Markets

Mean 0.185 0.270 0.234Median 0.116 0.201 0.161Maximum 0.999 1.000 1.000Ratio of rms with zero holdings 0.065 0.020 0.039 Number of observations 48865 65761 114626

Panel B: Institutional Holdings for NYSE Firms

Mean 0.324 0.394 0.366

Median 0.315 0.403 0.362Maximum 0.996 1.000 1.000Ratio of rms with zero holdings 0.002 0.003 0.003 Number of observations 13822 20575 34397

Panel C: Institutional Holdings for Amex Firms

Mean 0.139 0.176 0.158Median 0.086 0.112 0.098Maximum 0.938 1.000 1.000Ratio of rms with zero holdings 0.021 0.006 0.014 Number of observations 6620 7082 13702

Panel D: Institutional Holdings for Nasdaq FirmsMean 0.128 0.220 0.181Median 0.065 0.152 0.110Maximum 0.999 0.994 0.999Ratio of rms with zero holdings 0.106 0.032 0.064 Number of observations 28423 38104 66527

Panel E: T-Statistics for the Differences in Mean Institutional Holdings between Firms in Different Markets

NYSE vs. Amex 72.5** 75.8** 105.5** NYSE vs. Nasdaq 98.3** 81.1** 121.0**Amex vs. Nasdaq 5.3** 17.7** 13.6**

Note. The above summary statistics are based on data for the second quarter (June 30) of each year during the period. The ratio of shares held by institutional investors for each rm is computed as thetotal number of shares of the rm held by institutional investors at the end of the second quarter divided by the number of shares outstanding. The number of observations reported in the table includes bothrms with positive as well as zero institutional holdings.

*, ** Signicant at the 10% and 5% levels for a two-tailed test, respectively.



7/20

TABLE 2 Mean Monday and Mean Monday Minus Tuesday to FridayReturns (%) for All NYSE, Amex, and Nasdaq Stocks Partitioned byPeriod and Market, 19811998 (T-statistics in Parentheses)

Period

Difference inReturns between

198189 199098 198198199098 and

198189 Period

Panel A: For all NYSE, Amex, and Nasdaq Stocks Combined

Mean Monday Return 0.183 0.031 0.107 0.152(4.21**) (0.88) (3.82**) (2.74**)

Mean Monday minusTuesdayFriday Return 0.299 0.182 0.240 0.118

(6.51**) (4.86**) (8.09**) (2.01**)

Panel B: For NYSE StocksMean Monday Return 0.145 0.010 0.068 0.155

(2.65**) (0.29) (2.09**) (2.40**)Mean Monday minus

TuesdayFriday Return 0.265 0.059 0.163 0.206(4.62**) (1.63) (4.75**) (2.91**)

Panel C: For Amex Stocks


Mean Monday minusTuesdayFriday Return 0.310 0.184 0.247 0.126(5.99**) (5.62**) (8.05**) (2.13**)

Panel D: For Nasdaq Stocks


Mean Monday minusTuesdayFriday Return 0.317 0.248 0.283 0.069

(7.77**) (6.12**) (9.78**) (1.22)

Panel E: T-Statistics for the Differences in Returns between Firms in Different Markets

Mean Monday Return NYSE vs. Amex (1.90*) (3.38**) (3.72**) NYSE vs. Nasdaq (2.57**) (3.81**) (4.24**)Amex vs. Nasdaq (1.70*) (0.25) (1.41)

Mean Monday minusTuesdayFriday Return

NYSE vs. Amex (0.62) (2.66**) (1.93*) NYSE vs. Nasdaq (0.83) (3.57**) (2.83**)Amex vs. Nasdaq (0.16) (1.33) (0.96)

Note .The Monday minus Tuesday to Friday return refers to the difference (%) between theMonday return and the average Tuesday to Friday return of the same week.

*, ** Signicant at the 10% and 5% levels for a two-tailed test, respectively.



8/20

19811989, 19901998, and 19811998 periods, also has the highest (least negative) mean Monday return and mean Monday minus Tues-day to Friday return during the same period. It is also interesting tonote that during the 19811989 sub-period Nasdaq has the lowest per-centage of institutional holdings and also has the lowest mean Mon-day return and mean Monday minus Tuesday to Friday return amongthe three markets.

Panel E of Table 2 provides t-statistics on the differences in meanMonday returns and the differences in mean Monday minus Tuesdayto Friday returns between stocks in NYSE, Amex, and Nasdaq. Theresult indicates that the mean Monday returns for NYSE stocks aresignicantly higher (less negative) than that for Amex and Nasdaqstocks in all the three periods (19811989, 19901998, 19811998)examined. The mean Monday minus Tuesday to Friday return for NYSE stocks is also signicantly higher than that for Amex and Nasdaq stocks during the 19901998 sub-period. These results pro-vide additional evidence that the Monday seasonal may be related tothe institutional holdings level in the stock market.

B. Analyses Based on Institutional PortfoliosOur evidence so far indicates that the Monday seasonal is stronger inthe 19811989 sub-period than in the 19901998 sub-period. Thiscoincides with the evidence that the percentage of institutionalholdings is lower in the 19811989 period than in the 19901998 period. We also know that during the 19811998 period, the Mon-day seasonal is stronger in Amex and Nasdaq stocks than in NYSEstocks. This coincides with the evidence that the institutional hold-ings level of NYSE stocks is much higher than that of Amex and Nasdaq stocks. These two pieces of evidence indicate that theMonday seasonal and the level of institutional holdings may becorrelated. To provide direct evidence on this issue, this sectionanalyzes the impact of institutional holdings on Monday returns by partitioning stocks into portfolios sorted by the level of institutionalholdings.

For each year during the 19811998 period, we divide all NYSE,Amex and Nasdaq stocks in our sample into ten equal portfolios,according to the percentage of institutional holdings in each stock.Portfolio 1 is the portfolio with the lowest mean institutional holdingswhile portfolio 10 is the one with the highest mean institutional hold-ings. Table 3 reports the result. The institutional holdings level variessignicantly among deciles. In the 19811998 period, the mean in-stitutional holdings level is 0.2% for the lowest decile and 68.6% for the highest decile. We also observe that the institutional holdings levelis a monotonically increasing function of time for nearly all decilesduring the period examined.



9/20

TABLE 3 Institutional Holdings in Each Holdings Decile by Year and Specied Periods, 19811998

Year

Portfolio 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95

Lowest 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.001 0.001 0.001 0.001 0.002 0.003 0.003 0.003 0.Decile 2 0.003 0.002 0.005 0.005 0.007 0.011 0.012 0.011 0.011 0.010 0.011 0.014 0.018 0.018 0.022 0.Decile 3 0.013 0.012 0.020 0.020 0.027 0.036 0.037 0.033 0.034 0.033 0.035 0.043 0.052 0.052 0.060 0.Decile 4 0.031 0.031 0.046 0.046 0.057 0.071 0.072 0.067 0.070 0.068 0.074 0.086 0.101 0.101 0.112 0.Decile 5 0.057 0.060 0.081 0.082 0.097 0.115 0.117 0.112 0.118 0.119 0.125 0.142 0.161 0.163 0.176 0.Decile 6 0.093 0.097 0.130 0.128 0.150 0.174 0.171 0.164 0.179 0.184 0.191 0.214 0.232 0.236 0.250 0.Decile 7 0.150 0.156 0.193 0.188 0.216 0.243 0.239 0.229 0.251 0.260 0.274 0.300 0.321 0.328 0.345 0.Decile 8 0.233 0.237 0.277 0.269 0.302 0.326 0.329 0.324 0.348 0.358 0.381 0.405 0.426 0.441 0.457 0.Decile 9 0.350 0.347 0.393 0.383 0.413 0.438 0.446 0.445 0.469 0.489 0.511 0.536 0.554 0.571 0.590 0.Highest 0.537 0.537 0.570 0.571 0.592 0.615 0.624 0.632 0.658 0.671 0.696 0.718 0.736 0.745 0.763 0.

Note .This table reports the mean percentage of institutional holdings of ten portfolios of NYSE, Amex, and Nasdaq stocks sorted by tand in each specied period during the 19811998 period. The percentage of institutional holdings is dened as the number of shares held shares outstanding on June 30 of each year. Portfolio 1 (10) is the portfolio with the lowest (highest) mean institutional holdings.


10/20

Panel A of Table 4 reports the mean Monday returns of the ten portfolios sorted by the percentage of institutional holdings in eachyear and specied periods from 1981 to 1998. Two clear patterns can be observed from the table. First, the portfolio returns are, in general,more negative in the earlier years (when institutional holdings arelower) than in the later years (when institutional holdings are higher).Second, the portfolio returns are, in general, more negative in thelower decile portfolios (with lower institutional holdings) than in thehigher decile portfolios (with higher institutional holdings). Except for portfolios 1, 3 and 4, the portfolio return is a monotonically increasingfunction of the level of institutional holdings in the 19811998 period.For the portfolio with the highest institutional holdings (portfolio 10),the mean Monday return is signicantly negative in only one out of the 18 years examined. Not surprisingly, the mean Monday return(0.057%) of the portfolio with the highest institutional holdings inthe 19811998 period is not signicantly different from zero. Simi-larly, for the portfolio with the next highest institutional holdings(portfolio 9), only three out of the 18 years are with a signicantlynegative mean Monday return.

A comparison of the results in the 19811989 sub-period with that in the 19901998 sub-period yields additional insight. The stock returns of every portfolio in the 19811989 period are lower than thecorresponding portfolio in the 19901998 period. All the portfolioMonday returns are negative and signicant in the 19811989 period.During the 19811989 period, except for the portfolios 1 and 7, thereturn is a monotonically increasing function of the level of institu-tional holdings. In contrast, only the returns of two portfolios (port-folios 5 and 6) are negative and signicant in the 19901998 period.More interestingly, the two portfolios with the highest institutionalholdings (portfolios 9 and 10) exhibit positive mean Monday returnsduring the 19901998 period.

As a simple test to see if the portfolio Monday return is identicalacross deciles, we run regressions for each of the years and periodsusing the portfolio Monday return as the dependent variable and the percentage of institutional holdings in each holdings decile as the in-dependent variable. We predict the coefcient of this variable to besignicantly positive. Panel B of Table 4 reports the results. For the18 years examined, 16 out of the 18 coefcients for the institutionalholdings variable are positive. Out of these 16 coefcients, ten aresignicant. For the remaining two negative coefcients, only one issignicant. When examining the relationship between the meanMonday return and the mean institutional holdings during the periods19811989, 19901998 and 19811998, we nd that all the threecoefcients associated with the institutional holdings variable are positive and signicant. This regression result seems to indicate that



11/20

the mean Monday returns across institutional holdings deciles are not identical. 6

When the difference between the Monday return and the averageTuesday to Friday return is partitioned by institutional holdings level and by year, we nd an interesting pattern. Panel A of Table 5 reports that the differences become insignicantly different from zero in recent years(1990 onwards) for portfolios with higher institutional holdings. For the19901998 period, the difference between the Monday return and theaverage Tuesday to Friday return is not signicant for the two portfolios(portfolios 9 and 10) with the highest institutional holdings level. This at least shows that, in recent years, the Monday return is not different fromthe mean Tuesday to Friday return for stocks with more institutionalinvestors. Similar to Table 4, we also run regressions for each of the yearsand periods using the difference between the Monday return and theaverage Tuesday to Friday return as the dependent variable and the in-stitutional holdings in each decile as the independent variable. Again, we predict the coefcient of this variable to be signicantly positive.

Panel B of Table 5 reports the result. For the 18 years examined, 16of the coefcients associated with the institutional holdings variableare signicantly positive. For the three periods examined (19811989,19901998 and 19811998), all the three coefcients associated withthe institutional holdings variable are also positive (two are signi-cant). Our regression result suggests that the difference between theMonday return and the average Tuesday to Friday return is not iden-tical across holdings deciles. This further shows that the Monday sea-sonal is typically stronger in stocks with low institutional holdings.

A comparison of the result in Table 4 (based on Monday returns)with that in Table 5 (based on Monday minus Tuesday to Fridayreturns) yields an interesting observation. We observe in Table 4 that the effect of holdings is not very signicant on a year-to-year basis after 1990, while in Table 5 the effect of holdings is always signicantly positive. This is probably because, although the mean Monday returnsare not signicantly negative for most holdings deciles for the yearsduring the bull market of 1990s, the Monday returns (at least for thelower holdings deciles) are still signicantly lower than the returns onother weekdays during the period. This lower mean Monday minusTuesday to Friday return may be the reason why the institutionalholdings variable is more signicant in explaining the Monday minus

6. We also conduct t-tests to see if the mean Monday returns are equal among institutionalholdings deciles in the 19811998 period. When we divide all stocks into three institutionalholdings portfolios, the difference in the Monday returns between the high holdings port-folio and low holdings portfolio is 0.068% (t-statistic = 2.29). When we divide all stocksinto ve institutional holdings portfolios, the difference in the Monday returns between thehighest holdings portfolio and lowest holdings portfolio is 0.056% (t-statistic = 2.26). Theseresults support the conclusions derived from the regression analyses.



12/20

TABLE 4 Impact of Institutional Holdings on Mean Monday Returns of Portfolios of NYSE, Amex, and Nasdaq Stock

Year

81 82 83 84 85 86 87

Panel A: Equally-Weighted Monday Returns (%) of Portfolios of NYSE, Amex, and Nasdaq Stocks Sorted by the Level o(T-Statistics in Parentheses)

PortfolioLowest .249 .174 .123 .328 .146 .17 .424

(3.48**) (2.61**) (1.33) (5.17**) (3.13**) (3.38**) (2.22**) (Decile 2 .438 .3 .121 .339 .087 .169 .43

(3.99**) (2.73**) (.97) (4.80**) (1.56) (2.77**) (1.86*) Decile 3 .304 .207 .063 .267 .061 .194 .554

(2.72**) (1.97**) (.62) (4.26**) (1.14) (2.90**) (2.18**) Decile 4 .256 .107 .047 .277 .068 .217 .486

(2.14**) (.87) (.47) (3.96**) (1.05) (2.91**) (1.77*) Decile 5 .22 .146 .088 .271 .013 .184 .531

(2.08**) (1.22) (.83) (3.49**) (.20) (2.55**) (1.75*) Decile 6 .235 .116 .079 .215 .035 .238 .546

(2.16**) (1.01) (.73) (2.58**) (.56) (3.02**) (1.84*) Decile 7 .204 .097 .068 .212 .009 .197 .459

(1.89*) (.74) (.62) (2.53**) (.13) (2.56**) (1.57) Decile 8 .187 .167 .093 .175 .001 .212 .532

(1.54) (1.23) (.81) (2.02**) (.01) (2.25**) (1.65*) Decile 9 .185 .084 .068 .223 .004 .23 .547

(1.51) (.54) (.58) (2.32**) (.04) (2.25**) (1.50) Highest .119 .045 .064 .168 .016 .169 .6

(.96) (.28) (.57) (1.69*) (.17) (1.47) (1.45)

Panel B: Regression Estimates Obtained by Regressing the Mean Monday Return of Each Portfolio on the Institutional H(T-Statistics in Parentheses)

Intercept .292 .185 .090 .290 .078 .194 .472(10.20**) (7.42**) (7.49**) (20.78**) (4.86**) (18.24**) (21.98**) (2

Holdings .355 .277 .049 .253 .198 .018 .191 (4.18**) (3.94**) (1.57) (5.01**) (4.09**) (.37) (3.85**) (

Adjusted R 2 .49 .39 .03 .62 .56 .10 .39 Number of Observations 10 10 10 10 10 10 10


13/20


14/20


15/20


16/20

Tuesday to Friday return than in explaining the Monday return duringthe 19901998 period. 7

We also conduct two regressions to see if the effect of institutionalholdings is the same for the 19811989 and 19901998 sub-periods.In regression 1, we regress the 20 mean Monday returns (for the decile portfolios in the 19811989 period and the decile portfolios in the19901998 period) against three independent variables: dummy for 19901998, percentage of institutional holdings in each decile, and aninteraction variable (dummy for 19901998 * percentage holdings).Regression 2 estimates the same equation, with the exception that weuse the mean Monday minus Tuesday to Friday return as the dependent variable. We expect the coefcient of the institutional holdings variableto be signicantly positive. Table 6 reports the result.

As expected, the coefcients for the percentage of institutionalholdings are signicantly positive in both regressions. When theMonday return is used as the dependent variable, the interaction term isnot signicant indicating that the percentage of institutional holdingsdoes not have a different effect in the two sub-periods. However, when

Monday minus Tuesday to Friday return is used as the dependent vari-able, we nd that the effect of institutional holdings is stronger inthe 19901998 sub-period than in the 19811989 sub-period. (The

TABLE 6 Regression Analysis of the Effect of Institutional Holdings onMonday Returns and Monday minus Tuesday to Friday Returnsfor the 19811989 and 19901998 Sub-Periods

Dependent Variable Regression 1Monday Return

Regression 2

Monday minus Tuesdayto Friday Return

Intercept 0.208 0.313(25.35**) (30.03**)

Percentage Holdings 0.135 0.080(4.35**) (2.03*)

Dummy 9098 0.151 0.047(12.58**) (3.11**)

Percentage Holdings * Dummy 9098 0.041 0.225(1.03) (4.50**)

Adjusted R-square 0.95 0.92

Number of Observations 20 20 Note .Regression 1 uses the average Monday returns of the ten portfolios sorted by institutional

holdings during the 19811989 sub-period and the average Monday returns of the ten portfolios sorted by institutional holdings level during the 1990 1998 sub-period as the dependent variable. Regression 2uses the 20 average Monday minus Tuesday to Friday portfolio returns sorted by institutional holdingslevel during the 19811989 and the 19901998 sub-periods as the dependent variable. PercentageHoldings is the average percentage of institutional holdings of each portfolio sorted by holdings levelduring the two sub-periods. Dummy 9098 takes a value of 1 for the ten portfolio returns during the19901998 period. Percentage holdings * Dummy 9098 is an interaction variable.

*, ** Signicant at the 10% and 5% levels, respectively, for a two-tailed test (t-statistics).

7. We thank an anonymous referee for pointing out this observation.



17/20

coefcient of the interaction term in regression 2 is positive and sig-nicant.) This probably explains why the mean Monday minus Tuesdayto Friday returns of larger holdings deciles (deciles 7 to 10) are mostlynot signicantly different from zero for the years in the 19901998 sub- period, but are mostly negative and signicant during the years in the19811989 sub-period. However, the fact that the dummy for the 1990 1998 period is also signicantly positive in both regressions indicatesthat during the period there are forces other than institutional holdingsaffecting the magnitude of the Monday seasonal as well.

C. Additional TestsAbraham and Ikenberry (1994) and Sias and Starks (1995) nd that rm size may also inuence the magnitude of the Monday return. Inaddition, Wang, Li, and Erickson (1997) report that the mean Mondayreturn is more negative in the last two weeks of the month. Theliterature also indicates that the Monday seasonal is highly correlatedwith the previous Fridays return. To further test the impact of insti-tutional holdings on Monday returns while controlling for these fac-tors, we run two separate sets of regressions.

The rst set of regressions tests the impact of institutional holdingson the Monday return of individual stocks by year and by specied period over the 19811998 period while controlling for rm size andthe type of stock market. (The dependent variable is the Mondayreturn of individual stocks and the independent variables are the percentage of institutional holdings of the stock, the logarithm of the stocks market capitalization, and two dummies representing dif-ferent stock markets). The second set of regressions analyzes therelation between institutional holdings and the Monday returns of individual stocks for size-sorted portfolios over the 19811998 periodwhile accounting for the previous trading day return, the last twoMondays effect, and the type of stock market. Overall, our con-clusions regarding the effects of institutional holdings on the Mondayreturn do not seem to be affected when we control for these additionalvariables.

V. Conclusions

The Monday seasonal has been one of the most puzzling phenomenain the nance literature. Our paper reports several interesting ndingsrelated to this phenomenon. We rst document that the mean Mondayreturn is not signicantly negative if we consider only stock returndata from 1990 to 1998. There are two possible explanations for thisobservation. First, the bull market of the 1990s could have resultedin a less negative mean Monday return during the period. Second,



18/20

because more institutional investors participate in the stock market during the 19901998 period than in the pre-1990 period, the declinein the magnitude of the Monday return after 1990 may be linked to theincreased role of institutional investors in the equity markets.

We further explore the relationship between institutional investorsand the Monday seasonal, nding that the Monday seasonal is weaker for stock portfolios with high institutional holdings than those withlow institutional holdings. More importantly, we nd that the meanMonday returns of stock portfolios with high institutional holdings are positive during the 19901998 period and most of these returns arenot signicantly different from the corresponding mean Tuesday toFriday returns during the same period.

Our paper provides direct evidence that the well-known Mondayseasonal is typically stronger in stocks with low institutional holdings.Lakonishok and Maberly (1990)document that individual investors tendto increase trading activity (especially sell transactions) on Monday.This evidence indicates that the Monday seasonal could be related to thetrading pattern of individual investors. Our result supports Lakonishok and Maberlys (1990) ndings. Our nding also suggests that an active participation of institutional investors in a stock market may reduce themagnitude of the Monday seasonal. This is true because institutionalinvestors may be able to actively arbitrage the seasonal given that they should be more aware of the Monday seasonal and have lower trading costs than most individuals.

Appendix

Our data collection process begins with the Spectrum tapes. From these tapes, weobtain the following information: CUSIP, stock name, end-of-quarter shares out-standing, and shares held by institutional investors. However, because the Spec-

trum le reports the number of shares outstanding gure rounded to the nearest onemillion, some of the percentages exceed 100 percent when the percentage of sharesheld by institutional investors is calculated. To solve this problem, we use sharesoutstanding data from CRSP tapes. We rst identify rms that are listed by both theCRSP and Spectrum tapes and then use the 8-digit CUSIP number from the CRSPtapes as the key to merge information from the two tapes. This process results in ale that contains the permanent number (from CRSP), the number of shares held by each institutional investor (from Spectrum), the number of shares outstanding at the end of each quarter (from CRSP), and stock returns data (from CRSP) duringthe 19811998 period. For each rm the percentage of institutional holdings in

each quarter is computed as the number of shares held by institutional investorsdivided by the number of shares outstanding in that quarter.We use the data reported for 1984 as an example to discuss the data collection

process. In 1984, there are 5,266 rms in the Spectrum tapes that have informationon the number of shares held by institutional investors. Of these rms, 158 cannot be matched (using the 8-digit CUSIP number) with any rm in the CRSP tapes



19/20

during the same quarter. (The majority of these rms are foreign rms.) Since wedo not have return data for these 158 rms, we delete them from the sample. After eliminating these rms, the resulting number of Spectrum rms that can becompletely matched with CRSP rms in the same quarter using the 8-digit CUSIPnumber is 5,108. From these 5,108 rms we exclude 8 observations where thecomputed percentage of institutional share holdings exceeds 100 percent and an-other 50 observations with no CRSP shares outstanding data for the quarter. (In thelatter case, Spectrum reports the share holdings information on the rm at least aquarter earlier than when CRSP begins to report shares outstanding data for therm. This is likely in situations where rms go public using the best-efforts method.)After going through the above-mentioned procedures, the resulting number of usefulobservations is 5,050.

We then match rms listed in CRSP tapes with rms listed in Spectrum in the19811998 period. In the second quarter of 1984, we nd 6,335 observations inthe CRSP tapes with shares outstanding data along with stock return data duringthe quarter. Among these 6,335 rms, 381 cannot be matched (using the 8-digit CUSIP number) with rms listed in Spectrum during any quarter in the period19811998. In other words, during the 19811998 period none of the 13(f )institutions hold the stocks of these rms. Hence these rms are assigned a zero percentage of institutional holdings for the purpose of analysis. 5,954 rms listedin the CRSP tapes in the second quarter of 1984 can be matched using the 8-digit CUSIP with rms listed in the Spectrum in one of the quarters during the 1981 1998 period.

From these 5,954 rms we exclude 8 rms (similar to the 8 rms in the Spectrumtapes) that have a computed percentage of institutional holdings in excess of 100 percent. After we exclude these 8 rms we have 5,946 completely matched CRSPrms. Of these 5,946 rms, 5,050 are matched with Spectrum rms in the samequarter and 896 are matched with Spectrum rms in a quarter other than the current quarter. Since we are not certain about the institutional holdings information in thecurrent quarter for the 896 rms, we delete these rms from the sample. This process gives us a total of 5,431 observations. Of these, 5,050 observations have a positive percentage of institutional holdings and 381 observations have zero in-stitutional holdings (these are rms that cannot be matched with rms listed in

Spectrum during any quarter in the 19811998 period).

References

Abraham, A., and Ikenberry, D. L. 1994. The individual investor and the weekend effect. Journal of Financial and Quantitative Analysis 29: 263277.

Badrinath, S. G.; Kale, J.; and Noe, T. 1995. Of shepherds, sheep and the cross-autocorrelationin equity returns. Review of Financial Studies 2: 401430.

Barclay, M.; Christie, W.; Harris, J.; Kandel, E.; and Schultz, P. 1999. The effects of market reform on the trading costs and depths of Nasdaq stocks. Journal of Finance 54: 134.

Bessembinder, H. 1999. Trade execution costs on NASDAQ and NYSE: A post-reform

comparison. Journal of Financial and Quantitative Analysis 34: 387407.Chan, L. K. C., and Lakonishok, J. 1993. Institutional trades and intraday stock price behavior. Journal of Financial Economics 33: 173199.

Chan, L. K. C., and Lakonishok, J. 1995. The behavior of stock prices around institutionaltrades. Journal of Finance 50: 11471174.

Chan, L. K. C., and Lakonishok, J. 1997. Institutional equity trading costs: NYSE versus Nasdaq. Journal of Finance 52: 713735.



20/20

Christie, W., and Schultz, P. 1999. The initiation and withdrawal of odd-eighth quotesamong Nasdaq stocks: An empirical analysis. Journal of Financial Economics 52: 409 442.

Damodaran, A. 1989. The weekend effect in information releases: A study of earnings anddividend announcements. Review of Financial Studies 2: 607623.

Flannery, M. J., and Protopapadakis, A. A. 1988. From T-bills to common stocks: Inves-tigating the generality of intra-week return seasonality. Journal of Finance 33: 431450.

French, K. R. 1980. Stock returns and the weekend effect. Journal of Financial Economics8: 5570.

Grinblatt, M.; Titman, S.; and Wermers, R. 1995. Momentum investment strategies, port-folio performance and herding: A study of mutual fund behavior. American Economic Review 85: 10881105.

Huang, R., and Stoll, H. 1996. Dealer versus auction markets: a paired comparison of execution costs on Nasdaq and the NYSE. Journal of Financial Economics 41: 313357.

Jaffe, J. F.; Westereld, R.; and Ma, C. 1989. A twist on the Monday effect in stock prices:Evidence from the U.S. and foreign stock markets. Journal of Banking and Finance 13:641650.

Kamara, A., 1997. New evidence on the Monday seasonal in stock returns. Journal of Business 70: 6384.

Keim, D., and Madhavan, A. 1995. Anatomy of the trading process: Empirical evidence onthe behavior of institutional traders. Journal of Financial Economics 32: 371398.

Keim, D., and Madhavan, A. 1997. Transactions costs and investment style: An inter-exchange analysis of institutional equity trades. Journal of Financial Economics 46: 265 292.

Keim, D., and Stambaugh, R. 1984. A further investigation of the weekend effect in stock returns. Journal of Finance 30: 81984.

Lakonishok, J., and Maberly, E. 1990. The weekend effect: Trading patterns of individualand institutional investors. Journal of Finance 40: 231243.

Lakonishok, J., and Levi, M. 1982. Weekend effects on stock returns: A note. Journal of Finance 37: 883889.

Lakonishok, J.; Shleifer, A.; and Vishny, R. 1992. The impact of institutional trading onstock prices. Journal of Financial Economics 32: 2343.

Miller, E. 1988. Why a weekend effect? Journal of Portfolio Management 14: 4248. Nofsinger, J., and Sias, R. W. 1999. Herding and feedback trading by institutional and

individual investors. Journal of Finance 54: 22632295.Sias, R. W., and Starks, L. 1995. The day-of-the-week anomaly: The role of institutional

investors. Financial Analysts Journal 51: 5867.Sias, R. W., and Starks, L. 1997. Return autocorrelation and institutional investors. Journal

of Financial Economics 46: 103131.Wang, K.; Erickson, J.; Gau, G.; and Chan, S. H. 1995. Market microstructure and real estate

returns. Real Estate Economics 23: 85100.

Wang, K.; Chan, S. H.; and Gau, G. 1992. Initial public offerings of equity securities:Anomalous evidence using REITs. Journal of Financial Economics 31: 381 410.Wang, K.; Li, Y.; and Erickson, J. 1997. A new look at the Monday effect. Journal of

Finance 52: 21712186.


the impact of institutional investors on the monday seasonal - chan et al 04

Documents