Who Gains More by Trading – Institutions or Individuals?
Granit San∗
Tel-Aviv University
[email protected]
First Draft: September 2004
This Version: April 2006
∗
This paper is based on parts of my dissertation. Portions of this research were
implemented while I was a Visiting Doctoral Fellow at the Wharton School. I owe
special thanks to my advisor, Simon Benninga, for his dedicated support. I am especially
indebted to Roni Michaely for his guidance and numerous helpful comments and
suggestions. I would also like to thank Avner Kalay, Eugene Kandel, Shmuel Kandel,
Pete Kyle, Gideon Saar, and Jacob Sagi for helpful discussions and suggestions.
Comments from seminar participants at the Hebrew University, Tel-Aviv University, and
the Technion are gratefully acknowledged. This research was supported by a grant from
the Ministry of Science and Technology, Israel.
Who Gains More by Trading – Institutions or Individuals?
Abstract
We calculate four measures of institutional trading and find significant differences in the
return patterns of stocks with different proportions of trading activity by institutions and
individuals. In a two-year window around the trading, the lowest returns coincide with
intense institutional selling, whereas high returns correspond to intense institutional
buying. Our results demonstrate that individuals realize superior gains by selling. In the
late 1990s bubble, they also gain about 2% per month more than institutions by buying.
They suggest that a possible explanation for the inferior performance of institutions is
that institutions hold winners too long and mistime the momentum cycles. Though in
line with existing empirical evidence, they do not support the conventional wisdom that
individuals are the noise traders who lose money by trading.
I. Introduction
“Noise trading is trading on noise as if it were information… Most of the time, the noise
traders as a group will lose money by trading, while the information traders as a group
will make money… Because the actual return on a portfolio is a very noisy estimate of
expected return… there will always be a lot of ambiguity about who is an information
trader and who is a noise trader.” Black (1986, p. 531-532)
Which group of investors, institutions or individuals, are the noise traders who
lose money by trading? Financial academics and practitioners tend to view individual
investors as the noise traders. This paper demonstrates that this should not be held as a
self-evident truth, by presenting evidence that during the period 1986 through 2001
individuals gain more than institutions, thus they are not necessarily the noise traders.
We expose a significant difference in the return patters of stocks with different
proportions of trading activity by institutions and individuals. These differences indicate
that individuals are the ones who buy low and sell high, and that individuals as a group
do not lose money by trading, but realize superior gains.
Using data on institutional holdings and trading volume for all NYSE and
Nasdaq-NM stocks during the period 1986 through 2001, we calculate four measures of
institutional trading activity.
Our proxies measure the relative trading activity by
institutions and individuals at the single stock level.
We use these measures to
investigate the cross-sectional variations in stocks with different proportions of trading
activity by institutions and individuals. Since our measures includes the trading activity
of all public participants, a difference between stocks with various amount of trading by
institutions and individuals, which emerges through the return patterns, reflects the
systematic influence of each group on market prices.1
In this study, we focus on Black’s definition of noise traders. Accordingly, in
order to explore which group of investors, institutions or individuals, are the noise traders
(who lose money by trading) we use our measures to compare the performance of stocks
1
In that, we are different from most previous research, which focuses on a specific group of investors. For
example, Jensen (1968), Malkiel (1995), Grinblatt, Titman and Wermers (1995), Wermers (1999, 2000),
and many others, study mutual funds; Lakonishok, Shleifer and Vishny (1992a,b) analyze pension funds;
Del Guercio (1996) examines mutual funds and banks; and Odean (1998, 1999) investigates individual
investors with accounts in a particular discount brokerage. The extent to which the trading of each group of
investors affects market prices also depends on the trading of other market participants, yet this cannot be
manifested by these researches.
1
with different proportions of institutional and individual trading activity. We implement
the four measures of performance that are commonly used in the literature: raw return,
CAPM alpha, Fama-French alpha and four-factor alpha. Both, the return patterns and the
risk-adjusted returns, are investigated over the two years before and after the trading.
Figure 1 graphs the cumulative market-adjusted returns for the top and bottom
decile portfolios of institutional net trading. The bottom decile contains the stocks with
the most intense institutional net selling, while the top decile contains the stocks with the
most intense institutional net buying. Likewise, since individual trading is counter to
institutional trading, the bottom decile contains the stocks with the most intense
individual net buying, and the top decile contains the stocks with the most intense
individual net selling.2 The return patterns are established by event-times in the two-year
window around the trading. The figure summarizes the main results of the paper: the
lowest returns coincide with intense institutional net selling (and intense individual net
buying), whereas high returns correspond to intense institutional net buying (and intense
individual net selling).
At first glance, it would seem that our results are puzzling. However, a closer
scrutiny reveals that they are, in fact, coincide perfectly with existing empirical evidence.
First, consistent with the literature, Figure 1 shows that institutions are momentum
investors whereas individuals are contrarian traders. Furthermore, it reveals that one of
the implications of this evidence is that individuals time their trades better than
institutions and thereby might realize superior gains. Second, numerous studies find that
individuals exhibit the disposition effect (i.e. the tendency to sell winners too early).
Figure 1 confirms these findings, while exposing that even though individuals sell
winners too soon, they sell at a higher price than institutions.
Third, our results
corroborate the recent findings of Kaniel, Saar and Titman (2004) and Campbell,
Ramadorai and Vuolteenaho (2005). Investigating trading at the daily frequency and
short-term (up to month) performance, Kaniel, Saar and Titman (2004) find superior
performance of individuals, and Campbell, Ramadorai and Vuolteenaho (2005) find
inferior performance of institutions. We present similar evidence in the long run.
2
Section III contains the details of this presumption.
2
We provide a number of new results on the difference between stocks traded by
institutions and individuals, and, by implication, on the performance of institutions and
individuals and their role in the stock market.
First, our results indicate that institutions buy high and sell low, whereas
individuals are the ones who buy low and sell high. This is apparent when we examine
the return patterns over the four-year period around their trading (two-year before and
two-year after). Comparing the event-time market-adjusted returns of stocks heavily
traded by individuals and institutions, we find that the trading activity of institutions (or
their counterparts, individuals) signals a change in the return trends. For example, the
returns of stocks excessively bought by individuals decline in the two-year period
preceding individual purchases, while in the two years that follow the purchases the
returns increase. Moreover, when we investigate whether our trading measures convey
information that is not captured by the holding measure, we find that among the highmomentum stocks (winners) that institutions prefer to hold, institutions are net buyers in
stocks whose past returns are significantly higher than the past returns of the stocks in
which they are net sellers.
Second, our evidence suggests that if a certain group of investors is the noise
traders who lose money by trading, this group is not individuals but institutions. We find
that individuals time their exit from the market better than institutions and realize
superior gains by selling. Independent of the measure of performance we use, stocks
heavily sold by individuals, after been held by them over one quarter to two years, have
experienced significant abnormal excess past returns relative to stocks heavily sold by
institutions. For example, if both institutions and individuals sell a NYSE stock after
holding it for one year, the average return of stocks with excess institutional net selling
underperforms the average return of stocks with excess individual net selling by 15.7%
per year; and this is statistically significant, with a t-statistic of 17.63. Interestingly, the
inferior profits of institutions are salient despite the disposition effect of the individuals in
our sample. Our results indicate that although individuals do not realize the highest
(optimal) returns from their sales, they do sell at higher price than institutions. We also
find that in the pre-bubble period, neither institutions nor individuals realize superior
3
gains by buying. Stocks heavily bought by institutions realized about the same future
returns as stocks heavily bought by individuals.
Third, our subperiod analysis reveals distinct characterizations of the trading
activity of institutions and individuals in the late 1990s bubble. In the late 1990s bubble,
particularly in Nasdaq, not only do individuals gain more than institutions by selling but
they also gain more by buying. In this is period, Nasdaq-stocks excessively bought (and
sold) by institutions realize lower returns than Nasdaq-stocks excessively bought (and
sold) by individuals; and this holds over any investment horizon of up to two years.
Furthermore, when we adjust the returns to the four-factor risk, and include the
momentum risk, we find that the four-factor risk-adjusted returns of the portfolio of
intense institutional buying significantly underperform the portfolio of intense individual
buying. For example, in Nasdaq, the risk-adjusted return earned by the portfolio of
intense institutional buying, in the year that follow the purchases, is 28.3% lower than the
risk-adjusted return earned by the portfolio of intense individual buying.
Fourth, our findings suggest a possible explanation for the inferior performance of
institutions. In line with previous evidence, we find that institutions are momentum
traders who hold stocks with high past returns. However, our overall results indicate that
they tend to stick to momentum trading style and hold winners too long, and in doing so
they fail to time their trades to fully exploit the intermediate momentum effect. We find
that institutions tend to hold stocks which have high past returns (winners) not only over
the previous six-month and one-year but also over the previous two-year. Moreover, they
are net sellers in stocks whose past returns are lower than the past returns of the stocks in
which they are net buyers, and they are momentum traders with respect to the returns in
the previous six-month to two-year. This harms their performance, in particular in light
of Jegadeesh and Titman’s (1993) results on the intermediate momentum effect in stock
returns, suggesting that they might mistime the momentum cycle. Furthermore, the
differences between the raw returns and the four-factor risk-adjusted returns indicate that
not only are institutions not properly compensated for taking high momentum risk but
they also worsen their performance by taking this risk. A possible reason for this
evidence is window dressing (e.g., Lakonishok et al. (1991)) by institutions.
4
Our fifth finding is that our trading measures convey information that cannot be
captured by the holding measure. We find that stocks with high level of institutional
holdings have distinct characteristics from stocks with intense institutional trading. Large
stocks, with high beta and low book-to-market ratio have large institutional holdings;
while among the stocks with high institutional holdings, the stock with frequent
institutional trading are smaller, and have lower beta and higher book-to-market ratio
than the stocks with thin institutional trading.
Finally, our results indicate that the superior performance of individual is not
merely a compensation for high systematic risks. We find that though institutions and
individuals differ in their trading style, institutions do not consistently trade stocks with
risk-characteristics that provide lower returns.
The rest of the paper is organized as follows. Section II reviews related literature.
Section III presents the measures of institutional trading. We first describe the datasets
and methodology used to calculate them, and then demonstrate that it contains significant
information that cannot be captured by the holding measure. Section IV presents our
main results graphically, and section V investigates them further by a detailed empirical
analysis of the trading style, the past returns, and the future returns. We end this section
by a discussion of the complement implications of the empirical analysis. In Section VI
we look into the late 1990s bubble by a subperiod analysis. Section VII concludes.
II. Related Literature
Prior empirical works examine aspects of the relation between institutions and
individuals and stock returns.
The evidence on the preferences of institutions and
individuals is rather conclusive. On one hand, institutional studies find a positive relation
between past returns and net change in institutional holdings. For example, Grinblatt,
Titman and Wermers (1995), and Sias, Starks, and Titman (2001) document this relation
for one quarter; Wermers (1999), and Chen, Jegadeesh and Wermers (2000), document it
for one and two quarters; and Nofsinger and Sias (1999) for one year. On the other hand,
individual studies find that individuals tend to be contrarians. For example, Odean (1998,
1999) finds that U.S. individuals tend to hold on to their losers and sell their winners; and
Hvidkjaer (2005a) finds small-trade buying pressure for loser stocks. Individuals outside
5
the U.S. also tend to be contrarians.
For example, Grinblatt and Keloharju (2000)
document this for Finish individuals, Choe, Kho, and Stulz (1999) for individuals in
Korean, Shapira and Venezia (2001) for Israeli amateurs, Jackson (2003) for Australian
individuals, and Richards (2005) for individuals in six Asian markets. Our results, that
institutions are momentum traders and individuals are contrarians, are in line with these
works. We add to them by exploring that these results are not limited to few quarters but
persist over the two years prior to the trading; and by presenting their implications for the
realized profits of individuals and institutions and for their long-term performance.
The evidence on the performance of institutions is mixed. For example, Jensen
(1968), and Malkiel (1995) find that mutual funds underperform relevant market indices
over horizons of one year; and Lakonishok, Shleifer and Vishny (1992b) find that active
money managers of pension funds underperform the market index. In contrast, Wermers
(2000) finds that mutual funds hold stocks that outperform a broad market index by 1.3%
per year over one quarter, and Nofsinger and Sias (1999) find that following large
changes in institutional ownership stocks institutions purchase subsequently outperform
those they sell.
Gompers and Metrick (2001) find that the aggregate institutional
portfolio outperforms the aggregate individual portfolio by 0.67% per annum.
The
disparity in these findings corresponds to our results, which indicate that during the prebubble period the differences between the future raw returns of stocks with various levels
of institutions and individuals buying are insignificant, particularly in the quarter and
year subsequent to their purchases.
We add to these papers by comparing the
performance of institutions and individuals; and by suggesting that the tendency of
institutions to hold on to winners might drive the results, thus the performance measure
that should be applied is the four-factor alpha.
Carhart (1997) finds that some apparent persistent in the performance of mutual
funds is due to momentum in stock returns, hence it is not extended beyond a year. His
results indicate that active managers fail to outperform passive benchmark portfolio.
Similarly, Grinblatt, Titman and Wermers (1995), and Daniel et al. (1997) attribute much
of mutual funds outperformance to the high average returns of the stocks they hold, thus
to the momentum effect. Chen, Jegadeesh and Wermers (2000), like us, examine trades
of mutual funds rather than holdings. They show that funds tend to buy stocks that
6
outperform the stocks they sell, but only in the first year following the trades. They also
find that the persistence in performance is mostly due to the momentum effect in stock
returns. These findings are in line with our argument that the performance measure that
should be applied is the four-factor alpha. We add to them by presenting evidence that
institutions tend to stick to momentum trading style too long, thus they are not properly
compensated for taking high momentum risk; and by demonstrating that this was
particularly pronounced in the late 1990s bubble. Furthermore, unlike these papers,
which focus on mutual funds, we take a broader approach and investigate all institutions
and individuals.3 It is this extension that enables us to expose market effects.
The evidence on the performance of individuals in the U.S. is also mixed. Odean
(1999) investigates individual investors with U.S. discount brokerage accounts in the
period 1987 to 1993. He finds that individual buying portfolio underperform their selling
portfolio over the following two years. Hvidkjaer (2005b) studies small-trade volume to
infer retail trading, and finds that stocks with intense sell-initiated small-trade volume
outperform those with intense buy-initiated small-trade volume. Barber, Odean and Zhu
(2003) analyze trading records of individuals with discount broker accounts between
1991 and 1996, as well as larger sample of investors with retail broker accounts between
1997 and 1999. They find no convincing evidence that stocks heavily purchased by these
clients outperform those heavily sold by them. In fact, in their second sample, they find
outperformance (though not reliable). Possible reasons for the differences between these
findings and ours are that our group of individuals is broader and includes investors that
are not included in the particular group of individuals that these studies examine, and
their sample periods.4 Barber, Odean and Zhu’s (2003) results support this conjecture.
Several recent papers investigate the short-term dynamic relation between
institutions and individuals and stock returns. Kaniel, Saar and Titman (2005) use a
unique NYSE dataset for the period 2000 through 2002 to examine the relation between
daily individual investor trading and short horizons (up to a month) returns. They find
that individuals tend to be contrarian and that the stocks that individuals buy exhibit
3
It is worth noting that contrary to the extensive research on mutual funds, mutual funds constitute only a
fraction of all institutions. For example, in December 1986 their holdings represent only 5.7% of the value
of total institutional holdings; and in December 1995 their holdings represent only 22.2% of the value of
total institutional holdings (Gompers and Metrick (2001)).
4
We show that the significant outperformance of individuals is unique to the late 1990s bubble.
7
positive excess returns in the following month. Griffin, Harris and Topaloglu (2003a)
use the type of brokerage house to identify individual and institutional trading in Nasdaq100 stocks for May 2000 through February 2001. They find positive relation between
daily institutional trading and past daily and intra-daily returns, but no significant relation
to future daily returns.
Campbell, Ramadorai and Vuolteenaho (2005) apply a
sophisticated method to infer daily institutional trading from 13F filling and TAQ. They
find that institutions are momentum traders at the daily frequency, and that daily
institutional sales strongly predict positive return while institutional purchases only
weakly predict negative returns. Our study complements these studies by analyzing
longer horizons. We find similar results at the quarterly frequency. While these papers
focus on daily trades and their relatively short-term (days to a month) dynamics, we
investigate quarterly trades and their longer-term (a quarter to two-year) influences; thus,
we evaluate noise at different resolutions. In addition, our result expose that the relative
trading activity by institutions and individuals has different implications for the future
returns in the late 1990s bubble and in the period preceding it; suggesting that the validity
of their results to other periods should be considered with caution.
Our evidence is related to existing theories in two ways. First, some researchers
(e.g., Trueman (1988), Allen and Gorton (1993), and Dow and Gorton (1997)) recognize
that institutions might be the noise traders and provide various explanations of why they
engage in noise trading. Our findings support the premise of these models. Furthermore,
recent studies support our results and provide alternative explanations for them.
Dasgupta, Prat and Verardo (2005) model the conformist tendency of institutions. They
show that conformism could cause institutions to herd and present empirical evidence
that institutions lose from their trades in stocks that have been persistently bought (sold)
by them. Frazzini and Lamont (2005) evidence suggest that institutional flows could also
derive our results. They find that high mutual funds flows predict low future returns.
Second, many studies use the interaction between noise traders and information
traders in modeling the way in which information is incorporated into the markets and
affects prices. Despite the wide use of the distinction between trader types, the various
models are hard to verify since the identity, role and impact of the different traders are
largely unexplored empirically.
Our results have implication for these theories, in
8
specifying the identity of the traders in the models. We briefly give two examples. De
Long at al. (1990b) suggest that rational speculators might move ahead of noise traders,
in order to push up prices and trigger behavioral feedback traders to buy, so that they
profit from driving stock price movements. In this case, our results imply that institutions
are the noise traders, particularly in the late 1990s bubble. Alternatively, De Long at al.
(1990a), among others, recognize that noise traders can lead prices away from
fundamental values by creating noise trader risk which information traders cannot
arbitrage, and thereby affect prices and earn high returns. In this case, our evidence
might imply that individuals are the noise traders and they gain due to the noise they
create. Obviously, further empirical study is required in order to derive the validity of the
various models.
III. The Measures of Individual Investor Trading
A. Data
We calculate our institutional trading measures using two databases: institutional
holdings and trading volume. In this subsection, we describe each database, the way we
use it to calculate our measures, and our sample selection.
A.1 Institutional Holdings
A 1978 amendment to the Securities and Exchange Act of 1934 requires all
institutions with greater than $100 million of securities under discretionary management
to report their holdings to the SEC. Holdings are reported quarterly on the SEC’s form
13F, where all common stock positions greater than 10,000 shares or $200,000 must be
disclosed.5 13F filings were drawn from CDA/Spectrum Institutional Holdings database,
currently distributed by Thomson Financial.
We use the quarterly holdings from this database to calculate institutional
trading, defined as the total institutional dollar trading (i.e. buying and selling) in a
5
Other types of securities holdings (e.g. convertible bonds, stock options, preferred stock) are also required
to be disclosed and count toward the $100 million limit, but only common stocks are included in our study.
9
specific stock, i, during a given quarter, t.6 Letting h j ,i ,t denote institution j’s holdings of
stock i at the end of quarter t, and pi ,t denote the average daily price (taken from CRSP
daily stock file) of stock i in quarter t;7 we define
⎛
⎞
buying )i ,t = ⎜⎜ ∑ Max(h j ,i ,t − h j ,i ,t −1 ,0 )⎟⎟ ⋅ pi ,t
⎝ j
⎠
⎛
⎞
(institutional selling )i ,t = ⎜⎜ ∑ Max(h j ,i ,t −1 − h j ,i ,t ,0)⎟⎟ ⋅ pi ,t
⎝ j
⎠
(institutional
(institutional
trading )i ,t = (institutional buying )i ,t + (institutional selling )i ,t
(1)
(2)
(3)
To uniquely define each stock, we match each cusip code (which is used in the
Spectrum 13F file to identify a security) to its CRSP permanent number (permno). To
uniquely define each institution, we correct for reused institution’s identification numbers
by assigning the reused number a different institution number.8 Quarters without any
holding report are considered as having the same holding as in the previous quarter.
The use of 13F filings yields a good, though not perfect, proxy for institutional
trading. The main limitation of this proxy is that the quarterly snapshots of institutional
positions do not measure intraquarter roundtrip transactions. However, existing evidence
supports the assumption that such transactions are infrequent and should have a minor
impact on the results. First, institutions’ turnover is about 80 percent per year, which
corresponds to a holding period of fifteen months, well above a quarter.9 Second, until
1997 the so-called “short-short” rule of the IRS imposed tax penalties on funds that
derive more than 30 percent of their profits from holding periods of 91 or fewer days; this
ruling discourages funds from turning over stocks during short time periods. Third, both,
Lakonishok, Shleifer and Vishny (1992a), who analyze pension funds, and Wermers
6
We calculated trading in both dollars and number of shares (adjusted to distribution events). The results
established with the two measures are qualitatively identical. Hence, we report only the results based on
dollar volume and interchangeably use the two definitions.
7
Since institutions could have transacted any time during the quarter, we do not use the end-of-quarter
prices imputed in the 13F reports, but the average price for the quarter (calculated using CRSP daily
prices).
8
Spectrum reused institution’s identification numbers. A gap of more than one year in the reported
holdings for the same institution’s number typically reflects a different and unrelated institution.
9
80% per year is actually an upper estimate. For example, for mutual funds, which are among the most
active institutional traders, Carhart (1997) finds a turnover of 77.3% per year, Wermers (2000) documents
that the turnover is 59% per year, and Jin (2004) finds a minimum holding period of four months.
10
(1999), who examines mutual funds, indicate that roundtrip intraquarter trades are
infrequent and represent a small minority of all fund trades. Fourth, Griffin, Harris, and
Topaloglu (2003a), who use proprietary data to distinguish between individual and
institutional trading on Nasdaq, find a strong relation between their proxy for institutional
trading and the proxy that measure institutional trading by quarterly changes in
institutional ownership from Spectrum.
A.2 Trading Volume
We use volume traded and prices from CRSP daily file to calculate total trading,
defined as the total dollar trading (i.e. buying and selling) in a specific stock during a
given quarter. While using CRSP volume to calculate total trading, it is important to
notice that there are two different reporting conventions. In one, a transaction is reported
by each side of a trade, hence buy and sell are counted as two separate transactions. We
use this convention, along with the term trading (defined as the sum of buying and
selling) to indicate a transaction that is reported by the two parties involved in it. In the
other convention a transaction is reported only by one side of the trade (either the buyer
or the seller), and hence buy and sell are considered as one transaction. Volume traded,
reported on CRSP, use this convention. Therefore, to compute total buying (selling), we
first multiply the daily volume traded in a stock by its daily price, and then aggregate this
over the quarters’ trading days to get total buying (selling) in this stock during the
quarter. The sum of total buying and selling (i.e. twice CRSP volume) is total trading.
A.3 Sample Selection
We merge the three trading datasets (described above) to create our trading
sample. Our study covers all equity securities traded on the NYSE and Nasdaq with
available data from CRSP. The trading period is from the beginning of 1986 through the
end 2001.10 We only include ordinary common shares of firms incorporated in the
United States (share code 10 or 11). If a firm is delisted, we exclude the delisting quarter
(due to lack of holding data at the end of this quarter). Since we analyze the NYSE and
10
The trading period was set by data availability. When we implemented our empirical study, the data of
institutional holdings were available from the beginning of 1986, and the return data (from CRSP) were
available through the end of 2003. Since we are interested in the return patterns in the two years following
the trades our trading sample ends in 2001.
11
Nasdaq stocks separately, we only include quarters in which the stock is traded during the
whole quarter in the same exchange. To ensure that the results are not driven primarily
by small and illiquid stocks, we exclude Nasdaq Small Cap stocks and stocks whose
average market capitalization is lower than $100 million. We do not drop from our
sample observations with either no institutional trading, or institutional trading greater
than public trading (defined below). If a stock in CRSP is not traded by any institution
(insider) in a given quarter, we set institutional (insider) trading to zero. If institutional
buying (selling) in a given stock-quarter is greater than public buying (selling), we
restrict it to be equal to public buying (selling). Our final trading sample consists of
166,976 stock-quarter observations, with 7,245 stocks (2,858 traded in the NYSE and
4,387 in Nasdaq-NM) and 64 quarters.
B. Methodology
We use our proxies of institutional buying, selling and trading, defined in
equations (1), (2), and (3), to calculate four trading measures:
%(institutional buying ) =
institutional buying
total buying
(4)
%(institutional selling ) =
institutional selling
total selling
(5)
%(institutional trading ) =
institutional trading
total trading
(6)
%(institutional net trading ) =
%(institutio nal buying ) − %(institutio nal selling )
(7)
Our analysis combines the information in each of these measures, in order to
provide a comprehensive picture of stocks with different proportions of institutional and
individual trading activity. The net trading measure (equation 7) provides information on
the direction and magnitude of imbalances in institutional investor trading.
When
purchases among institutions exceed their sales, it is positive and institutions are the net
buyers (i.e. there is a positive trade imbalance where institutions are the net buyers).
When institutional selling exceeds their buying, it is negative and institutions are the net
sellers. Thus, the measure of institutional net trading quantifies the relation between net
12
buying and net selling, and enables to compare stocks with buying and selling pressures
by institutions.11 The buying, selling and trading measures (equations 4, 5 and 6) provide
information on the magnitude and intensity of buying, selling and trading by institutions;
thus on the impact of their trading activity itself.
In this paper, we are interested in the cross-sectional variations of different stocks.
Hence, the four trading measures are normalized so that they measure the percentage of
institutional trading. To ease the terminology, we henceforth use the terms “institutional
trading” and “percentage of institutional trading” interchangeably. Main issue that has to
be considered while calculating the percentage of institutional trading is that the trading
mechanisms are different for the NYSE and Nasdaq stocks. The NYSE is primarily an
auction market whereas Nasdaq is a dealer market.
Consequentially, total trading
reported for securities listed on the NYSE is not directly comparable to total trading
reported for securities listed on Nasdaq. To address this, throughout the study, we
separate between the two exchanges.
For simplicity, we refer to investors that file 13F as “institutions” and to all other
investors as “individuals”, even though this group also includes very small institutions
and short-term (intraquarter) institutional traders, since they make up only a tiny
percentage of the category. In addition, while the trades of individuals are likely to be the
counterparties to institutional trades, the complement of the percentage of institutional
trading is only an approximate estimate of the percentage of individual trading. Besides
individual trading, the complement of institutional trading includes insider trading, and
intermediary trading. Insider trading is smaller than institutional trading by an order of
magnitude, thus constitutes only a negligible fraction of the total trading.12 With the
exception of small, illiquid stocks (which are not included in our sample), the fraction of
intermediary trading is not substantially different for stocks that are traded in the same
11
In order to exclude the possibility that this measure is strongly dependent on the amount of trading (e.g.,
it is close to zero mainly due to either negligible trading or intense trading (with negligible difference
between buying and selling)), we also normalized the net trading measure by institutional trading (instead
of public investor trading). This has insignificant effect on the results (if anything, it reinforces them) since
the two measures are highly correlated (the correlation is about 0.9).
12
In previous version of this paper, we calculate insider trading using insider transactions information from
Forms 3, 4 and 5 fillings with the SEC, and incorporated it with institutional trading. This has no material
effect on the results.
13
exchange.13 Hence, the fraction of a stock traded by institutions (i.e. our institutional
trading measures) should be negatively correlated with the fraction of a stock traded by
individuals.
Although our method is not perfect, it does have a meaningful ability to detect
significant differences in the levels of trading activity by institutions and individuals, thus
provide information on the trading activity of all public participants. Since the extent to
which the trading of each group of investors affects market prices also depends on the
trading of other market participants, comparing the trading activity of the two groups
provides an opportunity to identify their systematic influences and measure the
performance of their underling stocks.
C. Are Our Trading Measures Different from the Holding Measure?
A prerequisite to an understanding of the dynamics of stock prices is an
understanding of the trading activity of the various market participants and their effects
on market prices. But, does trading activity by different investors convey information
that cannot be captured by their holdings? Could the return patterns of stocks with
intense institutional trading be different from those of stocks with high level of
institutional holdings?
Do stocks preferably held by institutions have distinct
characteristics from stocks frequently traded by them? Are the trading strategies of
institutions important only in the stocks they prefer to hold?
Since trading involves transaction costs and tax concerns, the decision to actively
trade a stock is likely to reflect stronger views, or at least different considerations, about
value than the decision to passively hold it. Since trading reflects realized gains and
losses whereas holdings reflect paper (or unrealized) gains and losses, any evidence of
stock selection ability would be more discernible in trading rather than holdings.
Furthermore, institutions face a variety of constraints on their investment decisions (e.g.
different regulation rules, prudence restrictions), which required them to hold stocks with
certain characteristics. Hence, it is possible that the stocks institutions heavily hold are
different from the stocks they frequently trade. All of the above lead us to hypothesize
13
The Appendix provides supporting evidence of this presumption for intermediary trading in each
exchange.
14
that trading conveys additional and important information that is not captured by
holdings.
To test whether our trading measures contain information that is not captured by
the holding measure, we perform conditional sorts. Each quarter, we conduct doublequintile sorts, sorting first by institutional holdings and then by our institutional trading
measures. We then calculate the time series average of the cross-sectional means of
stocks’ characteristics in the resulting 25 portfolios. Institutional holding measure is
calculated as in previous studies (e.g., Gompers and Metrick (2001), Sias, Starks and
Titman (2001), and Campbell, Ramadorai and Vuolteenaho (2005), among other). Using
13F filings, we compute institutional holdings in a stock-quarter, as the sum of shares
held by all institutions divided by the total shares outstanding.
In the current study, we focus on the cross-sectional variation in the return
patterns of stocks with different proportions of trading by institutions and individuals.
Academic research has shown that stocks with high past returns (“momentum”), small
stocks, stocks with high beta, and stocks with high book-to-market ratio, have higher
returns than stocks without those characteristics, and attribute this to risk differences. We
therefore investigate these characteristics for possible differences in the returns and the
risks associated with them. All variables are measured at the beginning of the trading
quarter. We compute the characteristics using the four trading measures. For simplicity
of presentation, we only report one trading measures for each characteristic, but the
results using all other trading measures were very similar.
The results, presented in Table I, conclusively confirm our hypothesis. Stocks
with high institutional holdings differ significantly from stocks with intense institutional
trading, and this is true with respect to each one of the above characteristics, as well as
for stocks traded both on the NYSE and on Nasdaq.
In Panel A of Table I, we report the average past returns (“momentum”) for
stocks in the double-quintiles of holding and net trading. We present past returns for
three time periods: six-month, one-year and two-year, calculated as the compounded
return over these time terms. The double-sort is by holdings and then by net trading. We
present the results for the net trading measure, since it enables to test explicitly whether
institutions realize their predicted momentum gains. The results are quite striking. Not
15
only do institutions prefer to hold stocks with high past returns in the previous six and
twelve months (previous studies), but also the stocks with high institutional holdings
have high returns in the preceding six months to two years. Moreover, they are net
buyers in stocks whose past returns, over all three time periods, are significantly higher
than the past returns of the stocks in which they are net sellers. For example, among the
stocks with the highest percentage of institutional holdings in the NYSE, institutions are
net buyers in stocks with an average past two-year return of 61.58% while the respective
return of the stocks where they are net sellers is 33.04%. In other words, institutions do
not seem to realize optimally their expected gains from holding winners. This establishes
our explanation that institutions might lose due to misting the intermediate momentum in
stock returns (Jegadeesh and Titman (1993)).
Panel B of Table I presents the average size decile of the stocks in each portfolio.
The size decile is the rank of the market capitalization of equity, based on NYSE size
decile cutoff, with the size rank of one being the smallest and the size rank of ten being
the largest. As apparent from the last row, and consistent with previous research (Daniel,
Grinblatt, Titman and Wermers (1997); Chen, Jegadeesh and Wermers (2000); and
Gompers and Metrick (2001)), institutions prefer to hold large capitalization stocks, and
in particular have an aversion to small stocks (Falkenstein (1996)).
However, our
evidence reveals that not only do institutions prefer large stocks, but also that the stocks
with high institutional holdings are larger than stocks with low institutional holdings.
Furthermore, among the stocks with high institutional holdings, the stocks with intense
institutional trading are smaller than the stocks with thin institutional trading.
For
example, among the stocks with the highest proportion of institutional holdings in the
NYSE, the average size decile of stocks with the highest institutional trading activity is
6.19 versus 7.38 for stocks with the lowest institutional trading activity.14 This indicates
that though the preferred holdings of institutions are large stocks, it is not necessarily the
case that this is also where their trading strategies are most important.
14
The results persist both in the NYSE and Nasdaq, and along the different holdings quintiles. The only
exception is very small stocks. According to our results, those stocks are rarely either held or traded by
institutions. However, this could be an artifact of our proxy. The threshold reporting levels required to fill
in the 13F form will import a small bias to both the holdings and trading proxies, which will be more
pronounced in the smaller stocks.
16
Panel C of Table I reports the average beta within each portfolio.
Beta is
estimated for each stock by the market model, using monthly returns in the 24-60 months
prior to the trading quarter.15 The results show a clear difference between institutions’
preference to hold more stocks with higher beta (beta increase with institutional
holdings), and their tendency to trade more often in the stocks with the lower beta
(controlling for holdings, beta decreases with trading). Panel D of Table I presents the
natural log of the book-to-market ratio of the stocks in each portfolio. The book-tomarket ratio is the book value for the calendar quarter, lagged by six-month, divided by
market capitalization at the beginning of the trading quarter. As can be seen, stocks with
high level of institutional holdings have lower book-to-market ratio than stocks with low
holdings, whereas, controlling for holdings, stocks with frequent institutional trading
have higher book-to-market ratio than stocks with low institutional trading.
In sum, in light of our results, it is clear that stocks with high level of institutional
holdings have distinct characteristics from stocks with intense institutional trading. This
demonstrates that our measures contain significant information, conveyed in investors’
trading activity, which cannot be captured by the holdings measure.
IV. Graphical Presentation of the Main Results
Figure 1 presents graphically the principal results of this paper. The figure graphs
event-time, cumulative market-adjusted returns for the top and bottom decile portfolios of
institutional net trading. It shows the return patterns to stocks before and after they are
intensively purchased and sold by institutions. At the end of each quarter, hereafter date
0, stocks are sorted into decile portfolios according to their percentage of institutional net
trading during that quarter. Cumulative buy-and-hold, market-adjusted returns for each
portfolio, p, for a period of τ trading months relative to date -24 (two-year before the end
of trading quarter), are calculated as:
R p ,τ =
1
N
τ
⎧ τ
(
)
(1 + RM ,t )⎫⎬
1
R
+
−
⎨∏
∑
∏
i ,t
i =1 ⎩t = −24
t = −24
⎭
N
15
(8)
We also test Scholes-Williams beta, estimated from the return data in the calendar year prior to the
trading quarter. Results are essentially identical to those reported herein.
17
Where Ri ,t is the CRSP monthly return for stock i on month t, RM ,t is month t return on the
CRSP value-weighted index including distributions, and τ = -18, -12, -6, -3, 0, 3, 6, 12,
18, and 24 months.16 Negative dates are τ months before the end of the trading quarter,
corresponding to the period over which returns are calculated to characterize investor
preferences. Positive dates are τ months after the portfolio formation date, the period
over which return performance is evaluated. The time-series average of the event-time
cumulative returns for each portfolio is its cumulative market-adjusted return in event
time. The returns are depicted for stocks in decile 1, which have the highest institutional
net selling (i.e. institutional net trading is the lowest and negative); and for stocks in
decile 10, which have the highest institutional net buying (i.e. institutional net trading is
the highest and positive). Since the complement of institutional trading is individual
trading, stocks in decile 1 have the highest individual net buying and stocks in decile 10
have the highest individual net selling. Panel A shows the results for stocks traded on the
NYSE and Panel B for stocks traded on Nasdaq.
The figure shows the return patterns from two years before the trading quarter
until two years after it; hereby includes the three investment horizons we examine in
detail in this study: one-quarter, one-year and two-year. We focus on these horizons for
the following reasons. One-quarter is the most frequently studied period in the various
institutional researches and is also close to Odean’s (1998) approximate median of
individual holding period for stocks. One-year is Benartzi and Thaler’s (1995) estimate
of the average investor’s investment horizon, and it is close to Carhart’s (1997) mean
holding period of mutual fund. Two-year is the average turnover of NYSE securities
during this period. We present the market-adjusted returns, not the raw returns, in order
to remove the effect that market timing might have on our results; particularly since part
of our research period was characterized by high returns and was highly volatile.
The most striking results in Figure 1 are the evident difference between the return
patterns of stocks with intense institutional purchases and sales, and the trend change that
is associated with the imbalances in their trading activity. Moreover, taking into account
the fact that intense institutional purchases coincide with intense individual sales (and
16
We checked the robustness of the results to calculations that are done for the CRSP equal-weighted or
exchange equal-weighted indices, and found that the change in indices has virtually no effect on the
market-adjusted returns of the portfolios relative to each other.
18
vise versa for institutional sales), it is apparent that if a certain group of traders better
time its exit and entry from a stock, this group is individuals. This highlights the
question: Are individuals indeed the noise traders who lose money by trading?
Figure 1 shows that institutions buy high and sell low. Furthermore, it reveals
that the trading activity of institutions signals a change in trend. In the stocks with the
highest percentage of institutional net selling, selling pressure by institutions is associated
with a reversal in the return patterns: in the two-year period preceding institutional sales
the return of stocks with the highest percentage of institutional net selling declines,
whereas following the sales the return increases. In stocks with the highest percentage of
institutional net buying the return rises steeply in the two-year period preceding
institutional purchases, while following the purchases the increase in return is mild.
The graphs also demonstrate that institutions are momentum investors and
individuals are contrarian traders. Moreover, the graphs show that not only do they
exhibit these trading styles with respect to the short-term (one-quarter), but also with
respect to the long-term (one- and two-year). Stocks with intense institutional net selling
have experienced at least two years of a decrease in return before being sold by
institutions, while stocks with intense institutional net buying have experienced an
increase in return over the same period. Contrary to the distinct difference in the returns
preceding institutional trading, which reflect superior timing of individuals relative to
institutions, the implications of the returns following institutional trading are not as
conclusive. While Nasdaq stocks with intense institutional net buying underperformance
(by 10.34% within two years) those with intense individuals net buying, the return
differences following institutional trades in the NYSE are insignificant.17
In Figure 1 we present the first-order results, namely institutional net buying and
net selling, and compare between institutional net purchases and sales. This obscures
possible impact differences of the trading activity by institutions and individuals. To
investigate further these differences, Figures 2 (3) present the second-order results for
institutional and individual buying (selling). It enables us to explicitly characterize and
compare stocks with intense institutional buying (selling) to those with intense individual
17
The overall increase in returns stems from our sample selection (see III.A.3 for details). The average
return of the stocks in our sample is higher than CRSP value-weighted return.
19
buying (selling). We plot Figures 2 and 3 using the same methodology and notations as
those we use to graph Figure 1. The only difference is that in Figure 2 (3) stocks are
sorted into decile portfolios according to their percentage of institutional buying (selling);
hence, stocks in decile 1 are the stocks with the lowest institutional buying (selling) or the
highest individual buying (selling), and stocks in decile 10 are the ones with the highest
institutional buying (selling).
Figures 2 and 3 reveal a dramatic difference between the return patterns of stocks
intensively traded by institutions and individuals, as well as superior trading timing for
individuals. Figure 2 shows that stocks with intense individual buying significantly
outperform, over the following two years, those with intense institutional buying (with
the exception of a slight underperformance in the first two quarters following the
purchases of NYSE stocks). This means that the gains individuals realize from the stocks
they purchase are higher than institutions’ gains (at least with respect to two out of our
three trading horizons).
For example, in Nasdaq, the return of stocks with intense
individual buying rises, on average, by 42.4% in the two years subsequent to being
purchased by individuals, whereas the average return of stocks with intense institutional
buying rises only by 8.15% in the same period. Moreover, individuals buy stocks
cheaper than institutions: institutions buy stocks after they rise, while individuals buy
them after they decline.18
When an investor sells a stock, the gains he realizes are determined by its past
returns; hence, for the selling decision the relevant returns are the past returns.
Accordingly, Figure 3 shows that not only do individuals gain more than institutions from
their sales, but they also enjoy gains while institutions suffer losses. For example, if both
individuals and institutions sold a NYSE stock after holding it for two years, on average,
individuals gained 12% while institutions lost 10%. Figure 3 also shows that although
the returns of stocks with intense individual selling significantly increase in the two years
prior to individuals sales, providing nice gains to their owners, the returns continue to rise
after individuals have sold them, suggesting that individuals could have benefited even
18
The only exception is individual buying in Nasdaq, where the decline in stocks’ return before individuals
bought them is less evident. This could reflect high dispersion in individuals’ opinions, or a high degree of
heterogeneity, in those stocks. It could stem, inter alia, from the formidable search problem involved in
choosing a stock to buy, when there are well over 10,000 securities available.
20
more if they had postponed their sales. These results confirm the disposition effect (i.e.
the tendency of investors to sell winners too early (Shefrin and Statman (1985));
documented by Odean (1998) for U.S. individuals, and by many others for U.S. and nonU.S. individuals) for the individuals in our sample.
It is worth mentioning briefly some additional implications of our results. First,
the institutions in our sample tend to buy recent good performers (Figure 2) and sell off
dismal performers (Figure 3). This is in line with the literature on window dressing (i.e.
institutions’ tendency to massage their positions just prior to reporting their holdings so
that they will look better in the eyes of investors, e.g. Lakonishok et al. (1991)). Second,
a comparison of Figure 1 with Figures 2 and 3 indicates that there are differences
between institutional net trading and institutional buying and selling. This reflects a
certain degree of heterogeneity (or herding) in both institutional and individual trading
activity. Third, our results suggest that if a certain group of investors is among the last
buyers to contribute to the rise of overvalued momentum securities and among the first to
suffer losses when these securities decline (De Long at al. (1990b)) these investors are
institutions, not individuals.
V. Empirical Analysis
In the previous section, we demonstrate graphically the implications of institutional
and individual trading activity for stock prices. Stocks with intense institutional selling
have experienced lower past returns than stocks with either intense institutional buying or
intense individual selling; and stocks with intense institutional buying realize lower
future returns than stocks with intense individual buying. This suggests that individuals
are the ones that time their purchases and sales better than institutions, and thereby gain
more by trading. In this section, we investigate these results further, by a detailed
quantitative analysis of these gains and their sources.
A. Trading Style
What is the source of the differences in the return patterns of stocks with different
rates of institutional trading activity? Do institutions differ from individuals in their
trading style? Do the stocks they buy and sell differ in their risk-characteristics? Are
21
individuals’ gains due to high systematic risk? The literature shows that cross-sectional
variation in returns can be explained by a compensation for risk, or by systematic risk
factors in returns. In line with this, in subsection III.C we study the risk-characteristics of
stocks institutions trade by a separate analysis of the four characteristics that are most
commonly associated with risk: beta, size, book-to-market ratio, and past returns. To
emphasize the robustness of our results, in this section we investigate whether investor
trading styles could explain the return differences, by analyzing the risk factor
sensitivities on different sets of factors: Sharpe-Lintner CAPM, Fama and French (1993)
three-factor model, and Carhart’s (1997) four-factor model.
At the end of each quarter, stocks are sorted into quintile portfolios according to
either their institutional trading (Table II) or their institutional net trading (Table III).
The returns on the quintile portfolios are calculated over the three months of institutional
trading measure formation, equally weighting the stocks within each quintile.19
In
addition, we calculate the returns on the “Q5-Q1” portfolio, constructed by going long
the top quintile and short the bottom quintile, to demonstrate clearly the difference
between the variables in the extreme quintiles. The three-month return series are linked
across quarters to form a monthly series of returns on each quintile portfolio. Using these
monthly returns, we estimate three model specifications of the time series regression:
R p ,t − R f ,t = α p + β p (RM ,t − R f ,t ) + s p SMBt + h p HMLt + m pUMDt + e p ,t
(9)
R p ,t is the monthly return for portfolio p. R f ,t is the monthly return on one-month T-bill.
RM ,t is the monthly return on a value-weighted market portfolio. SMBt is the monthly
return on a factor-mimicking portfolio for size (i.e. a zero-investment portfolio formed by
subtracting the return on a large firm portfolio from the return on a small firm portfolio).
HMLt is the monthly return on a factor-mimicking portfolio for book-to-market (a
portfolio of high book-to-market stocks less a portfolio of low book-to-market stocks).
And UMDt is the monthly return on a factor-mimicking portfolio for one-year past return
momentum (a zero-investment portfolio formed by subtracting the return on a portfolio of
low return stocks over the preceding year from the return on a portfolio of high return
19
We checked the robustness of the results to calculation of the returns over three and twelve months, both
following and preceding the trading quarter. The time period over which the returns are calculated has
virtually no effect on the risk factors sensitivities’ results.
22
stocks). The three model specifications we use are: the CAPM, that includes only the
first factor (market excess return) in equation (9); Fama and French (1993) three-factor
model, that extend the CAPM by including also size and book-to-market factors; and
Carhart’s (1997) four-factor model, that includes all four factors. The slopes are factor
loadings that have a clear interpretation as risk factor sensitivities for stocks.
In order to compare between the risk-characteristics of stocks traded by
institutions and individuals, stocks are sorted into quintile portfolios based on their
institutional trading measure.20 Q1 represents the portfolio with the lowest institutional
trading activity (highest individual trading), and Q5 represent the portfolio with the
highest institutional trading activity. Table II reports the risk factors sensitivities of two
model specifications of equation (9): CAPM beta, and Fama and French three-factor
loadings on beta, size, and book-to-market.21 Stocks with high institutional trading
activity have lower beta and higher book-to-market ratio than stocks with low
institutional trading. With respect to size, there is a difference between stocks traded on
the NYSE and on Nasdaq. On the NYSE, small stocks (high s) are traded more heavily
by both institutions and individuals, whereas in Nasdaq small stocks are traded more
heavily by individuals. Taking into account the size difference between NYSE and
Nasdaq stocks, it seems that both institutions and individuals trade in small stocks, but
institutions avoid trading in the smallest ones. In general, despite the differences in
institutional and individual trading preferences, the results do not indicate a coherent
trend in their attribute toward risk. For example, institutional trading activity decreases
with beta, which is associated with a decrease in risk, but at the same time it increase
with book-to-market (especially in Nasdaq), which is associated with an increase in risk.
In order to compare between the risk-characteristics of stocks purchased and sold
by institutions, we sort the stocks into quintile portfolios based on their institutional net
trading measure. Q1 represents the portfolio with the lowest, and negative, institutional
net trading, thus the portfolio of intense institutional net selling; and Q5 represents the
20
We also sort stocks into quintile portfolios based on their institutional buying and on their institutional
selling, and repeat the analysis for these portfolios. The results are qualitatively identical and therefore are
not reported.
21
We do not include the past return (momentum) factor here since, in the context of trading activity, this
factor contains meaningful information only if we separate between the buying and selling decision. We
elaborate on this issue in Table III.
23
portfolio with the highest, and positive, institutional net trading, thus the portfolio of
intense institutional net buying. Table III presents the risk factor sensitivities of Carhart’s
four-factor loadings on beta, size, book-to-market, and past return (momentum). Since
the past return factor might have different implications for the decision to buy or sell,
contrary to Table II, Table III adds this factor to the regression. The most salient result in
Table III is the difference in past return factor sensitivity between stocks in which
institutions are net buyers and stocks in which they are net sellers.
Stocks where
institutions are net buyers have, on average, a higher past return factor loading than
stocks in which institutions are net sellers. This also holds when we compare (not
reported here) the quintiles of institutional buying and institutional selling. Thus, in line
with the rest of this paper, particularly with the results presented previously, we clearly
see that the past returns of stocks heavily bought by institutions are significantly higher
than the past returns of stocks heavily sold by them. Moreover, this factor captures most
of the difference between the stocks institutions purchase and sell, as the differences with
respect to the other variables are negligible.
In sum, the results indicate that the differences in risk-characteristics could not be
the source of the return differences. While institutions differ from individuals in their
trading style, there is no consistent pattern in the risk attributes of the stocks they trade.
Moreover, if we take into account the four common risk factors, the sole risk factor that
significantly distinguishes stocks with excess institutional net buying from those with
excess institutional net selling is the one-year past return momentum.
B. Past Returns
Are the results that were presented graphically in the previous section significant
and robust? In the following two subsections, we address this question and investigate
these results further, through a detailed quantitative analysis of the returns to stocks
before and after they are sold and purchased by institutions and individuals. At the end
of each quarter, stocks are sorted into quintile portfolios according to one of our trading
measures. We label this quarter, hereafter the trading or portfolio formation quarter, as
“Qtr 0”, and measure all investment horizons relative to it. Thus, in the tables to follow,
Qtr -1 denotes one quarter before the trading quarter, Year 1 one year after it, -2Years the
24
two years before it, and so on. As explained in the previous section, we examine three
investment horizons: one-quarter, our short-term horizon, and one- and two-year, our
long-term horizons. In this subsection, we study these horizons for the periods preceding
the trading quarter, and in the next subsection, we analyze them for the periods following
it. Monthly returns on the quintile portfolios over each investment horizon are calculated
as the rebalanced mean monthly returns for the portfolio, equally weighting the stocks
within each quintile. The horizon months return series are linked across quarters, equally
weighting the portfolios’ overlapping returns, to form a monthly series of returns on each
quintile portfolio.22
Panel A of Table IV presents average monthly percentage returns on the quintile
portfolios of institutional net trading for the periods preceding the trading quarters. Q1
represents the portfolio with the lowest, and negative, institutional net trading, thus the
portfolio of intense institutional net selling; and Q5 represents the portfolio with the
highest, and positive, institutional net trading, thus the portfolio of intense institutional
net buying. Q5-Q1 represents the abnormal return difference between the quintile of
stocks that are most heavily bought (Q5) and the quintile of stocks most heavily sold
(Q1) by institutions. Hence, it provides a test of the null hypothesis that the difference in
the returns preceding intense institutional net buying and intense institutional net selling
is zero. The results reveal a dramatic and significant difference in the past returns to
stocks in which institutions are net buyers and those in which they are net sellers.
Moreover, the differences persist, and are statistically significant, both for stocks traded
on the NYSE and Nasdaq, as well as over each one of the investment horizons (though
they are more prominent for Nasdaq stocks and for the shorter horizons). Past returns are
decreasing with institutional net selling. Thus, stocks with increasing institutional net
selling have experienced lower past returns than stocks with increasing institutional net
selling (or, as presented in the table, increasing institutional net buying). For example, in
Nasdaq, institutions are intense net buyers (Q5) in stocks with an average monthly return
of 2.49% over the previous two years, while they are intense net sellers (Q1) in stocks
with a return of 0.98% over the same period. Hence, the returns difference between the
22
Similar to Jegadeesh and Titman (1993), to increase the power of our tests, we include portfolios with
overlapping investment period by revising the weights in the portfolio each month. This also allows us to
use simple t-statistics while testing the significance of the results.
25
stocks institutions buy and sell is 1.51% monthly, or 43.3% over the previous two years,
and this difference is statistically significant, with a t-statistic of 9.48. The evident
differences have two important implications.
First, the momentum trading style of
institutions hurts their performance, while the contrarian trading style of individuals
benefits them. Considering that institutional net buying is equivalent to individual net
selling, and that past returns reflect the realized gains from sales; our results indicate
clearly that the portfolio of intense institutional net selling significantly underperforms,
over the previous two years, the portfolio of intense individual net selling. This suggests
that institutions gain more than individuals by selling. Second, not only are institutions
momentum traders and individuals contrarian traders with respect to the short-term past
returns of one-quarter, but they also exhibit the same trading styles with respect to the
long-term past returns of two years.
Would the evidence hold if we adjust the returns to risk? To answer this question,
we use the traditional Jensen’s alpha (Jensen (1968)) to measure the risk-adjusted
performance of the quintile portfolios. Three versions of alpha, the intercept term in
equation (9), are calculated, by estimating equation (9) for three benchmark models. The
Sharpe-Lintner CAPM alpha is calculated with respect to the market benchmark, the
Fama-French alpha with respect to the market, size, and book-to-market benchmarks of
Fama and French (1993), and the four-factor alpha with respect to the market, size, bookto-market, and momentum benchmarks, following Carhart (1997). Panels B, C and D of
Table IV reports the alphas, in percentage per month, of the quintile portfolios of
institutional net trading for the periods preceding the trading quarters. The clear picture
that emerges from these three panels of the table is that the risk differences cannot
explain the difference between the past returns of stocks that bought by institutions and
those sold by them. For all three benchmark models, the trends and differences in riskadjusted returns (Panels B through D) are similar to those in raw returns (Panel A). In
particular, the similarity between the raw returns and the market-adjusted returns (CAPM
alpha) indicates that the performance differences are not solely due to market timing; and
the similarity between the raw returns and the four-factor adjusted returns indicates that
institutions are not properly compensated for buying stocks with high momentum risk.
26
Thus, the superior returns that individuals gain from their sales sustain even if we control
for systematic risks.
In order to compare explicitly between the returns that institutions and individuals
realize from their sales, Table V presents the four performance measures (raw returns and
alphas) of the quintile portfolios of institutional selling for the periods preceding the
trading quarters. Q1 represents the portfolio with the lowest institutional selling (highest
individual selling), and Q5 represents the portfolio with the highest institutional selling.
Q5-Q1 represents the portfolio that buys the institutional selling portfolio and sells the
individual selling portfolio, thus it provides a test of the null hypothesis that the
difference in past returns to stocks heavily sold by institutions and to those heavily sold
by individuals is zero. The results indicate that past returns decrease with institutional
selling, and that stocks with intense individual selling have experienced significantly
higher past returns than stocks with intense institutional selling. For example, in the year
that precedes the sales, Nasdaq stocks with intense individual selling experienced a
statistically significant excess return of 19% (1.46% monthly) relative to stocks with
intense institutional selling. Furthermore, as is evident by comparing the raw returns
(Panels A of Table V) to the performance measures that take explicit account of the
effects of risk on the return (Panels B through D of Table V), adjusting the returns for
systematic risks has minor effect on the results.
In sum, whether we compare institutional net selling to their net buying or
institutional selling to individual selling, and whether we take into account the systematic
risks or not; our findings show clearly that the portfolio of intense institutional selling
significantly underperform, over the previous two years, the portfolio of intense
individual selling.
Therefore, our past returns analysis suggests that not only are
institutions momentum traders whereas individuals contrarian traders in the short (onequarter) and long (one- and two-year) terms, but also that an important implication of the
difference between theirs trading styles is that the gains individuals realize from their
sales are higher than those of institutions, independent of risks.
27
C. Future Returns
In this subsection, we repeat the analysis of the previous subsection for the future
returns, using the same methodology and notations.
Table VI presents the four
performance measures, in percentage per month, of the quintile portfolios of institutional
net trading for the periods following the trading quarters. Panel A reports average
monthly returns, and Panels B through D report the performance measures that take into
account the effect of risk. In general, the results show minor and insignificant differences
between the returns (and risk-adjusted returns) to stocks with various levels of
institutional net trading. However, in Nasdaq, the future returns to stocks in which
institutions are net buyers underperform those of stocks in which they are net sellers in
the second year and over the two years following their trades.
Table VII compares explicitly between the future returns that institutions and
individuals realized from their purchases. The results are quite similar to those of Table
VI. In the NYSE, there are insignificant differences between the returns to stocks
following institutional and individual buying; and the minor outperformance of
institutions in the quarter that follow their purchases eliminates when adjusting for the
momentum in stock returns. In Nasdaq, the returns decrease with institutional buying,
and the four-factor risk-adjusted returns of the portfolio of intense institutional buying
significantly underperform the portfolio of intense individual buying over each one of the
trading horizons. A puzzling result, apparent both in tables VI and VII for Nasdaq-traded
stocks, is the significant excess future returns of the portfolio of individual buying
relative to the portfolio of institutional buying in the second year. This will be cleared
through the investigation of the following section.
D. Discussion
Our findings lead to a coherent picture, with important, though somewhat
surprising, implications. The relative trading activity of institutions and individuals
conveys relevant information for market prices in a two-year window around their
trading. Stocks heavily sold by individuals have experienced significant abnormal excess
past returns relative to stocks heavily sold by institutions; indicating that individuals time
their exit (sales) from the market better than institutions, and thereby realize higher gains
28
by selling. Stocks heavily bought by institutions realize about the same future returns as
stocks heavily bought by individuals; indicating that neither institutions nor individuals
gain more by buying. Put together, our results cast doubts on the standard view of
individuals as the noise traders who lose money by trading.
Moreover, the inferior performance of institutions relative to individuals is not
due to the lower systematic risks of their portfolios. First, we show that institutions do
not consistently trade stocks with risk-characteristics that provide lower returns, thus the
abnormal returns that individuals achieve cannot be explained solely by risk
compensation. Second, most of the results do not change when we analyze the different
measures of performance that take explicit account of the effects of risk on the return of
the portfolios. Therefore, our findings suggest that individuals’ outperformance is not
merely a compensation for risks but might reflect stock-picking ability.
Additional evidence that emerges from our analysis suggests that a possible
explanation for the previous result is that institutions mistime the intermediate
momentum in stock returns (Jegadeesh and Titman (1993)). First, in subsection III.C we
find that institutions tend to hold stocks which have high past returns (winners) not only
over the previous six-month and one-year but also over the previous two-year; and that
they are net buyers in stocks whose past returns are significantly higher than the past
returns of the stocks in which they are net sellers. Second, when we compare the trading
styles of institutions and individuals, we see that if we consider the four risk factors market, size, book-to-market, and momentum - the sole risk factor that significantly
distinguishes between their trading styles is the momentum risk, which is higher for
institutions than for individuals. Third, when we analyze the past returns we see that
institutions are momentum traders and individuals are contrarian traders with respect to
both the short-term past returns of one-quarter, and the long-terms past returns of one and
two years. Finally, when we adjust the returns to the momentum risk, it is apparent that
institutions are not properly compensated for taking high momentum risk. In particular,
in Nasdaq, adjusting for momentum risk results in a significant abnormal future return of
the portfolio of intense individual buying relative to the portfolio of intense institutional
buying. Overall, these findings imply that institutions tend to stick to momentum trading
style and hold winners too long. In doing so, they mistime the intermediate momentum
29
effect and this damages their performance. Alternatively, the reversal that is associated
with the relative trading activity of institutions and individuals might not be related to
Jegadeesh and Titman’s (1993) intermediate momentum effect, but to the long-term
reversals documented by De Bondt and Thaler (1985), and the irrational biases accounted
for them.
VI. The Late 1990s Bubble
On January 1995, the Nasdaq Composite Index opened at 751; on March 10, 2000,
it closed at 5,048. The unusual rise and fall in the prices of Nasdaq stocks in the late
1990s has led many academics and practitioners to describe this period as a stock price
bubble. Who gained more by trading in the late 1990s bubble, institutions or individuals?
Were their gains in the bubble different from their gains in the period that precede it?
Did they trade differently? In this section, we investigate these questions by a subperiod
analysis.
We separate our trading sample into two subperiods: 1995 through 2000, the late
1990s bubble period, and 1986 through 1994, the pre-bubble period.23,24
For each
subperiod, we replicate the analysis of the previous sections. We present here only the
results that are different from the results in our entire sample period, since these are the
results that could shed some light on the distinct characteristics of the bubble. Therefore,
results that are not presented here are similar (both in each subperiod and in the entire
sample period) to those reported in previous sections.
Table VIII repeats the analysis of Table VI, and presents subperiods’ results of the
returns following institutional net trading. Panel A reports the average monthly returns,
and Panel B reports the four-factor alphas. The most striking result in this table is the
clear difference between the four-factor alphas in the bubble and in the period preceding
it, a difference that is most pronounced in Nasdaq stocks. In the pre-bubble period, there
are insignificant differences between the returns, and the alphas, of stocks with various
levels of institutional net trading, and this holds both for stocks traded on the NYSE and
23
We do not investigate the post-bubble period due to data availability (see footnote 13).
Similar results were established when we include within the late 1990s bubble period trading that were
implemented in 2001. However, since 2001 could be considered as part of the post-bubble period, in order
to focus the discussion here on the bubble, we do not present these results here.
24
30
Nasdaq. However, this changes in the late 1990s bubble period, in particular where the
bubble was most salient, in Nasdaq. In the late 1990s bubble, Nasdaq-stocks with intense
individual net buying (institutional net selling) have significant excess four-factor riskadjusted returns relative to stocks with intense institutional net buying, over each one of
the trading horizons. Furthermore, adjusting the returns to the four systematic risk
factors has major effect on the short-term results, though minor effect in the long-run. In
the quarter and year subsequent to the trading quarter, there is a minor and insignificant
difference between the raw returns to stocks with various percentage of institutional net
trading; whereas after adjusting the returns to the systematic risks, particularly to the
momentum risk, stocks with intense institutional net buying underperform stocks with
intense individual net buying. In the long-term (the second year and over the two years
following the trades), the underperformance is evident both in the raw returns and alphas.
For example, in Nasdaq, the raw returns of stocks in which institutions are net buyers
underperform those of stocks in which individuals are net buyers (institutions are net
sellers), in the second year after the trades, by 11.6%; and this is statistically significant
(at the one-percent level) with a t-statistic of 3.04.
Table IX repeats the analysis of Table VII, and shows subperiods’ results for the
future returns that institutions and individuals realized from their purchases. Panel A
reports the average monthly returns, and Panel B reports the four-factor alphas. Similar
to Table VIII, the most salient result in this table is the difference between the returns in
the bubble and in the period preceding it, particularly for Nasdaq stocks. In the prebubble period, there are insignificant performance differences between stocks with
various levels of institutional buying, both for stocks traded on the NYSE and Nasdaq.
However, this is changed in the late 1990s bubble period, specifically in Nasdaq. In the
late 1990s bubble, Nasdaq-stocks heavily bought by institutions underperform those
heavily bought by individuals, and this holds over each trading horizon in the subsequent
two years. The underperformance is most prominent in the four-factor risk-adjusted
returns.
For example, in Nasdaq, the underperformance of the portfolio of intense
institutional buying relative to the portfolio of intense individual buying ranges from
2.05% to 2.31% per month (27.57% to 31.53% per year) over the different trading
horizons.
31
The findings of our subperiod analysis suggest some important implications for
the characterization of the trading activity of institutions and individuals in the late 1990s
bubble.
First, in the late 1990s bubble, not only do individuals gain more than
institutions by selling but they also gain more by buying. Contrary to the similar gains of
institutions and individuals from their purchases in our entire sample period, in the
bubble, the gains individuals realize from their purchases are higher than the gains
institutions realize from their purchases.
Thus, whereas in the pre-bubble period
individuals time their exit (sales) from the market better than institutions, in the bubble
they time both, their entry (purchases) and exit (sales) from the market better than
institutions. Intriguingly, any doubts, raised previously, about the conventional wisdom
of individuals as the noise traders who lose money by trading, become even more serious
in the late 1990s bubble.
Second, in line with the rest of our findings, the subperiod analysis also indicates
that institutions are not properly compensated for taking high momentum risk, thus their
tendency to stick to momentum trading style does not benefit them but hurts their
performance. However, the subperiod analysis reveals that this is most severe in the late
1990s bubble. In the bubble, when we adjust the future returns to the four-factor risks
(and thereby include the risk factor that should compensate institutions for their
momentum trading style), the risk-adjusted returns of stocks excessively bought by
institutions do not outperform those of stocks excessively bought by individuals, but
significantly underperform them by up to 2.31% per month. This supports our previous
argument, that a possible explanation for institutions’ underperformance is their
mistiming of the intermediate momentum effect in stock returns.
Furthermore, our
finding, that this is most distinct in the bubble, might imply that the underperformance of
stocks with excess institutional buying may be due to institutions’ mistiming of the
momentum cycle, i.e. buying at the top of it, and increasing the risk. It is possible that
institutions, which follow momentum strategies, are among the last buyers that contribute
to the rise of overvalued momentum stocks and are among the first to suffer losses when
trends reverse and these stocks decline. In this case, the noise traders, in models such as
De Long et al. (1990b), are not the individuals.
32
Overall, our findings of the inferior performance of institutions relative to
individuals, along with it being most pronounced in the late 1990s bubble, suggest that
the standard attributes of institutions (sophisticated-informed investors) and individuals
(noise-irrational traders) should not be held as a self-evidence truth. Furthermore, they
lead to many puzzling implications. For example, since we show that the superior
profits, realized by individuals, are not merely a compensation for high systematic risks;
this leaves a room for other possible explanations, such as stock-picking ability, trading
on different information, agency problems, institutions’ constraints, etc.
VII. Conclusion
Black (1986) defines noise traders as traders that as a group, most of the time, lose
money by trading. In this paper, we use this definition and test the performance of two
groups of investors, institutions and individuals.
We calculate four measures of
institutional trading, and expose that the relative trading activity by institutions and
individuals has important implications for market prices in a two-year window around
their trading. Taken as a whole, our evidence indicates that during the period 1986
through 2001 individuals gain more than institutions by trading, thus it should not be
taken for granted that individuals are the noise traders.
We find that when comparing the two groups of investors, institutions are the
ones who realize inferior profits. They buy high and sell low, hence realize inferior
gains. Stocks heavily sold by individuals earn significant excess past returns relative to
stocks heavily sold by institutions. In the late 1990s bubble, particularly in Nasdaq, the
portfolio of intense institutional buying underperforms the portfolio of intense individual
buying with respect to future returns. This implies that individuals time their entry and
exit from the market better than institutions and gain more. Overall, our findings cast
doubts on the common assumption that individuals are the noise traders who lose money
by trading while institutions are the information traders who make money; and call for
further investigation of possible sources of our results (e.g. agency costs, institutions’
constraints, fund flows).
Interestingly, our evidence suggests a possible explanation for the inferior
performance of institutions. First, we find that institutions tend to hold stocks that are
33
winners over the previous two-year, but, among the high-momentum stocks they hold
they are net sellers in stocks whose past returns are lower than the past returns of the
stocks in which they are net buyers. Second, we find that institutions are momentum
traders with respect to the past two-year returns. Third, when we adjust the returns to the
four-factor risks, it is apparent that institutions are not properly compensated for taking
high momentum risk but lose from it, particularly in the late 1990s bubble. These
findings imply that institutions might gain less than individuals due to holding winners
too long. In so doing, they mistime the intermediate momentum effect (Jegadeesh and
Titman (1993)), and increase their risk without exploit its profit. This has important
implications for both practitioners and academics. For instance, it seems that some
practitioners do in fact recognize this: “Before I look at a stock, I take a look at the (SEC)
filings to see who the major shareholders are. If you see a large amount of momentum
money in there, you have to accept that there’s a high risk…”25
Our results also indicate that the superior performance of individuals is not merely
a compensation for high systematic risks. Individual outperformance could arise from
better timing the momentum cycle, but also from numerous other reasons. For example,
institutions might underperform due to their structural constrains; individuals and
institutions might possess different information or interpret it differently; individuals
might have superior stock selection skills or forecasting ability. What are the reasons that
explain the differences in return patterns is an open question.
What are the implications of our evidence to existing theories? They support
models of noise trading by institutions, in particular when a bubble exists (e.g. Allen and
Gorton (1993)).
They could be explained differently by models of the interaction
between noise traders and informed traders. For example, if noise traders mistime the
momentum cycle (e.g. De Long at al. (1990b)) institutions are the noise traders, and this
corresponds to our pronounced results in the bubble; however, if noise traders gain due to
the noise they create (e.g. De Long at al. (1990a)) individuals are the noise trader.
Clearly, further research is required in order to apply existing theories to our findings.
25
Quoted from money manager in an article by Greg, 1997, Red flag on wall street: Momentum investors,
Wall Street Journal, February 24.
34
We suggest two avenues for future research. First, in this study we focus on the
performance of stocks with different proportions of institutions and individuals trading
activity. In order to explore the dynamics that derives our results we plan to investigate
additional characteristics of stocks that differ in the relative trading activity by
institutions and individuals.
Second, our results reveal that the trading volume by
institutions and individuals contains important information for market prices. Why is
trading volume by institutions and individuals able to predict the magnitude and
persistence of future movements is an open question for further research.
We end with some reservations.
Although the evidence suggests inferior
performance of stocks with more institutional trading, for a variety of reasons investors
should interpret them with caution. First, individuals may have reasons, other than
institutions’ performance or stock-picking ability, to invest through institutions: saving
time and effort, piece of mind, and others. Second, our evidence applies on average, only
for stocks, and to the market as a whole. It does not suggest that the portfolios of
institutions perform worse than the portfolios of individuals. Third, our results might be
unique to our study period.
Appendix. Intermediary Trading
Main issue that has to be considered while estimating the fraction of intermediary
trading is that the trading mechanisms are different for the NYSE and Nasdaq. Thus, we
discuss intermediary trading separately for each exchange.
A1. NYSE
The NYSE reports its member trading in the NYSE Fact book. The report
includes monthly and annual data of the aggregate total member purchases and sales. In
addition to the total member trades, it also specifies separately specialists’ purchases and
sales, as well as non-specialists members’ purchases and sales that originated on the floor
and off the floor. According to these data, about half of the member purchases and sales
are by specialists, about half of them are by non-specialist off-floor members, while nonspecialist on-floor trades constitute only negligible fraction of the total member purchases
and sales.
35
Hasbrouck and Sofianos (1993) use data from November 1988 through August
1990 and find that the specialists’ participation rate is higher for stocks in the lowest
trade freqency decile. Madhavan and Sofianos (1998) use data from July 1993 and find
that specialist trading varies across stocks and is inversely related to trading frequency.
Sofianos and Werner (2000) use data of common stocks for January and February 1997
and find that specialists’ participation rate decreases with trading volume, however the
only distinct difference is in the lowest trading volume decile; the differences between
the other deciles are not substantial. The results of this paper are most relevant for us
since, like us, it includes only common stocks.26 Based on this evidence we should use
different estimates for intermediary trading in stocks with different trading volume,
specifically for the most illiquid stocks.
However, as was emphasized above, besides specialists’ trading, non-specialist
off-floor members trading contributes equally to the NYSE intermediary trading volume.
Several papers suggest that the trades of non-specialists off-floor members decrease with
trading volume. First, Hasbrouck, Sofianos and Sosebee (1993), and Madhavan and
Cheng (1997) provide evidence that non-specialists member trades that originate off the
floor tend to be very large block trades, and that the percentage of upstairs-facilitated
block volume increases with block size. Sofianos and Werner (2000) find that the
average block size increases with trading volume and that this pattern is most apparent
for upstairs-facilitated block trades. Madhavan and Sofianos (1998) find that specialist
trading is inversely related to trading frequency and proxies for off-exchange
competition, and that specialists participate less in stocks for which block trading volume
is significant and trade size in larger, because larger trades are more likely to be directed
to upstairs brokers. Madhavan and Cheng (1997) remark that small capitalization stocks
have only a few upstairs transactions in a month. Taking all of these together, it is
apparent that in the less liquid and smaller capitalization stocks there are fewer block
trades, specifically larger-block trades that are originated off the floor.
Second,
Madhavan and Cheng (1997) report that floor brokers play an important role in the
execution of upstairs-facilitated trades, and Sofianos and Werner (2000) find that when
26
Madhavan and Sofianos (1993) include in their sample, not only common stocks but also preferred stock
and closed-end fund. For these issues they find lower specialist participation, evidence that might bias their
findings relative to a sample of common stocks.
36
trading volume decreases, floor broker participation decreases, specifically for the
upstairs-facilitated block trades.
Overall, the evidence shows that low trading volume enhances the need for the
specialist to bridge gaps in natural liquidity, but also reduces the trading by non-specialist
off-floor members. Therefore, the differences in the fraction of intermediary trading
between NYSE stocks should not be substantial. Moreover, the only distinct difference
in specialists’ trading is in the lowest trading decile. Most of the stocks in the illiquid
decile are small capitalization stocks, which are excluded from our sample (see A.3).
Thus, the differences in the percentages intermediary trading across stocks should not be
significant, particularly for the stocks in our sample.
A2. NASDAQ
Nasdaq is largely a dealer market, where market makers are on the opposite side
of most trades. Ellis, Michaely and O'Hara (2002) analyze the mechanisms of market
making of dealers for 313 stocks in the first seven months after their IPO. Although their
sample might seem specific, their evidence shows that the data of stable market makers’
participation is similar for different stocks. First, they find that markets for newly issued
stocks adjust rapidly to a permanent equilibrium and are stable after twenty trading days.
Second, they show that the mean trading volume in months two through seven for their
sample firms is identical to the average turnover of Nasdaq stocks; and that despite the
differences in sample stocks, trading and dealer participations is consistent across stocks.
Griffin, Harris and Topaloglu (2003a,b) use proprietary data and qualitative analysis to
identify the parties involved in each trade, and classify both sides of all trades as
originating from an individual, an institution or a market maker. Their findings indicate
that the percentages of the different investors’ type participation are stable. Put together,
despite the variation in the stocks, time periods, and methodologies, the differences in the
resulting percentages of intermediary trading are not significant.27 Evidently, market
makers constitute a significant percentage of total trading for Nasdaq stocks, and only a
small fraction of the trading volume is executed directly by brokers.
27
Moreover, since we are interested in the cross-sectional variations of different stocks, throughout the
analysis, we sort the stocks into portfolios at the end of each quarter. Therefore, the time series differences
in intermediary trading do not affect our results.
37
References
Allen, Franklin and Gary Gorton, 1993, Churning Bubbles, Review of Economic Studies
60, 813-836.
Benartzi, Schlomo and Richard H. Thaler, 1995, Myopic loss aversion and the equity
premium puzzle, Quarterly Journal of Economics 110, 73-92.
Black, Fischer, 1986, Noise, Journal of Finance 41, 529-543.
Barber, Brad M., Terrance Odean and Ning Zhu, 2003, Systematic noise, Working paper,
University of California, Davis.
Campbell, John Y., Tarun Ramadorai and Tuomo O. Vuolteenaho, 2005, Caught on type:
Institutional order flow and stock returns, Working paper, Harvard University.
Carhart, Mark M., 1997, On persistence in mutual fund performance, Journal of Finance
52, 57-82.
Chen, Hsiu-Lang, Narasimhan Jegadeesh and Russ Wermers, 2000, The value of active
mutual fund management: An examination of the stockholdings and trades of fund
managers, Journal of Financial and Quantitative Analysis 35, 343-368.
Choe, Hyuk, Bong-Chan Kho and Rene M. Stulz, 1999, Do foreign investors destabilize
stock markets? The Korean experience in 1997, Journal of Financial Economics 54, 227264.
Daniel, Kent, Mark Grinblatt, Sheridan Titman and Russ Wermers, 1997, Measuring
mutual fund performance with characteristic-based benchmarks, Journal of Finance 52,
1035-1058.
Dasgupta, Amil, Andrea Prat and Michela Verardo, 2005, The price of conformism,
Working paper, London School of Economics.
De Bondt, Werner F. M. and Richard Thaler, 1985, Does the stock market overreact?
Journal of Finance 40, 793-805.
De Long, J. Bradford, Andrie Shleifer, Lawrence H. Summers and Robert J. Waldmann,
1990a, Noise trader risk in financial markets, Journal of Political Economy 98, 703-738.
De Long, J. Bradford, Andrie Shleifer, Lawrence H. Summers and Robert J. Waldmann,
1990b, Positive feedback investment strategies and destabilizing rational speculation,
Journal of Finance 45, 379–395.
Del Guercio, Diane, 1996, The distorting effect of prudent-man laws on institutional
equity investments, Journal of Financial Economics 40, 31-62.
38
Dow, James and Gary Gorton, 1997, Noise trading, delegated portfolio management, and
economic welfare, Journal of Political Economy 105, 1024-1050.
Ellis, Katrina, Roni Michaely and Maureen O'Hara, 2002, The making of a dealer market:
From entry to equilibrium in the trading of Nasdaq stocks, Journal of Finance 57, 22892316.
Falkenstein, Eric G., 1996, Preferences for stock characteristics as revealed by mutual
fund portfolio holdings, Journal of Finance 51, 111-135.
Fama, Eugene F. and Kenneth R. French, 1993, Common risk factors in the returns of
stocks and bonds, Journal of Financial Economics 33, 3-56.
Frazzini, Andrea, and Owen A. Lamont, 2005, Dumb money: Mutual fund flows and the
cross-section of stock returns, Working paper, Yale University.
Gompers, Paul A. and Andrew Metrick, 2001, Institutional investors and equity prices,
Quarterly Journal of Economics 116, 229-259.
Griffin, John M., Jeffrey H. Harris and Selim Topaloglu, 2003a, The dynamics of
institutional and individual trading, Journal of Finance 58, 2285-2320.
Griffin, John M., Jeffrey H. Harris and Selim Topaloglu, 2003b, Investor behavior over
the rise and fall of Nasdaq, Working paper, Yale University.
Grinblatt, Mark and Matti Keloharju, 2000, The investment behavior and performance of
various investor types: A study of Finland’s unique data set, Journal of Financial
Economics 55, 43-67.
Grinblatt, Mark, Sheridan Titman and Russ Wermers, 1995, Momentum investment
strategies, portfolio performance, and herding: A study of mutual fund behavior,
American Economic Review 85, 1088-1105.
Hasbrouck, Joel, George Sofianos and Deborah Sosebee, 1993, New York Stock
Exchange systems and trading procedures, NYSE working paper #93-01.
Hvidkjaer, Soeren, 2005a, A trade-based analysis of momentum, forthcoming Review of
Financial Studies.
Hvidkjaer, Soeren, 2005b, Small trades and the cross-section of stock returns, Working
paper, University of Maryland.
Jackson, Andrew, 2003, The aggregate behavior of individual investors, Working paper,
London Business School.
39
Jegadeesh, Narasimhan and Sheridan Titman, 1993, Returns to buying winners and
selling losers: Implication for stock market efficiency, Journal of Finance 48, 65-91.
Jensen, Michael C., 1968, The performance of mutual funds in the period 1945-1964,
Journal of Finance 23, 389-416.
Jin, Li, 2004, How does investor short-termism affect fund manager short-termism,
Working paper, Harvard Business School.
Kaniel, Ron, Gideon Saar and Sheridan Titman, 2004, Individual investor trading and
stock returns, Working paper, New York University.
Lakonishok, Josef, Andrei Shleifer, Richard Thaler, and Robert W. Vishny, 1991,
Window dressing by pension fund managers, American Economic Review 81, 227-231.
Lakonishok, Josef, Andrei Shleifer, and Robert W. Vishny, 1992a, The impact of
institutional trading on stock prices, Journal of Financial Economics 32, 23-43.
Lakonishok, Josef, Andrei Shleifer, Robert W. Vishny, 1992b, The structure and
performance of the money management industry, Brookings Papers on Economic
Activity: Microeconomics 1992, 339-391.
Madhavan, Ananth and Minder Cheng, 1997, In search of liquidity: Block trades in the
upstairs and downstairs markets, Review of Financial Studies 10, 175-203.
Madhavan, Ananth and George Sofianos, 1998, An empirical analysis of NYSE specialist
trading, Journal of Financial Economics 48, 189-210.
Malkiel, Burton, 1995, Returns from investing in mutual funds 1971 to 1991, Journal of
Finance 50, 549-572.
New York Stock Exchange, 1995-2000, New York Stock Exchange Fact Book (New York
Stock Exchange, New York).
Nofsinger, John R. and Richard W. Sias, 1999, Herding and feedback trading by
institutional and individual investors, Journal of Finance 54, 2263-2295.
Odean, Terrance, 1998, Are investors reluctant to realize their losses?, Journal of
Finance 53, 1775-1798.
Odean, Terrance, 1999, Do Investors Trade Too Much?, American Economic Review 89,
1279-1298.
Richards, Anthony J., 2005, Big fish in small ponds: The trading behavior of foreign
investors in Asian emerging equity markets, Journal of Financial and Quantitative
Analysis 40, 1-27.
40
Shapira, Zur, and Itzhak Venezia, 2001, Patterns of behavior of professionally managed
and independent investors, Journal of Banking and Finance 40, 777-790.
Shefrin, Hersh, and Meir Statman, 1985, The disposition to sell winners too early and
ride losers too long: Theory and evidence, Journal of Finance 40, 777-790.
Sias, R., L. Starks, and S. Titman, 2001, The price impact of institutional trading,
Working paper, Washington State University.
Sofianos, George and Ingrid M. Werner, 2000, The trades of NYSE floor brokers,
Journal of Financial Markets 3, 139-176.
Trueman, Brett, 1988, A theory of noise trading in securities markets, Journal of Finance
43, 83-95.
Wermers, Russ, 1999, Mutual fund herding and the impact on stock prices, Journal of
Finance 54, 581-622.
Wermers, Russ, 2000, Mutual fund performance: An empirical decompositions into
stock-picking talent, style, transaction costs, and expenses, Journal of Finance 55, 581622.
41
Figure 1
Returns Patterns around Intense Institutional Net Trading
This figure presents event time cumulative market-adjusted returns for the top and bottom decile portfolios
of institutional net trading. At the end of each quarter (date 0), stocks are sorted into decile portfolios
according to their percentage of institutional net trading in this quarter. The figure plots the cumulative
market-adjusted returns for stocks in decile 1, with the highest institutional net selling, and for stocks in
decile 10, with the highest institutional net buying. Cumulative market-adjusted return is the time-series
average of cross-sectional means of the cumulative, buy-and-hold, market-adjusted return (in excess of
CRSP value-weighted index), in event time. The event times are τ = -18, -12, -6, -3, 0, 3, 6, 12, 18, and 24
months. The trading sample period is January 1986 through December 2001 (hence the returns are from
1984 to 2003). Panel A shows the results for stocks traded on the NYSE and Panel B for stocks traded on
Nasdaq.
Panel A. NYSE
18
8
-2
-12
-22
-24
-18
-12
-6
0
6
12
18
24
Event Tim e (m onths)
intense institutionl net buying
intense institutionl net selling
Panel B. Nasdaq
80
Cumulative Market-Adjusted
Returns (%)
Cumulative Market-Adjusted
Returns (%)
28
60
40
20
0
-20
-24
-18
-12
-6
0
6
12
18
24
Event Tim e (m onths)
intense institutionl net buying
42
intense institutionl net selling
Figure 2
Returns Patterns around Intense Purchases by Institutions and Individuals
This figure presents event time cumulative market-adjusted returns for the top and bottom decile portfolios
of institutional buying. At the end of each quarter (date 0), stocks are sorted into decile portfolios
according to their percentage of institutional buying in this quarter. The figure plots the cumulative
market-adjusted returns for stocks in decile 1, with the lowest institutional buying (highest individual
buying), and for stocks in decile 10, with the highest institutional buying. Cumulative market-adjusted
return is the time-series average of cross-sectional means of the cumulative, buy-and-hold, market-adjusted
return (in excess of CRSP value-weighted index), in event time. The event times are τ = -18, -12, -6, -3, 0,
3, 6, 12, 18, and 24 months. The trading sample period is January 1986 through December 2001 (hence the
returns are from 1984 to 2003). Panel A shows the results for stocks traded on the NYSE and Panel B for
stocks traded on Nasdaq.
Panel A. NYSE
12
7
2
-3
-8
-13
-18
-23
-24
-18
-12
-6
0
6
12
18
24
Event Tim e (m onths)
intense institutional buying
intense individual buying
Panel B. Nasdaq
60
Cumulative Market-Adjusted
Returns (%)
Cumulative Market-Adjusted
Returns (%)
17
50
40
30
20
10
0
-24
-18
-12
-6
0
6
12
18
Event Tim e (m onths)
intense institutional buying
43
intense individual buying
24
Figure 3
Returns Patterns around Intense Sales by Institutions and Individuals
This figure presents event time cumulative market-adjusted returns for the top and bottom decile portfolios
of institutional selling. At the end of each quarter (date 0), stocks are sorted into decile portfolios
according to the percentage of institutional selling in this quarter. The figure plots the cumulative marketadjusted returns for stocks in decile 1, with the lowest institutional selling (highest individual selling), and
for stocks in decile 10, with the highest institutional selling. Cumulative market-adjusted return is the timeseries average of cross-sectional means of the cumulative, buy-and-hold, market-adjusted return (in excess
of CRSP value-weighted index), in event time. The event times are τ = -18, -12, -6, -3, 0, 3, 6, 12, 18, and
24 months. The trading sample period is January 1986 through December 2001 (hence the returns are from
1984 to 2003). Panel A shows the results for stocks traded on the NYSE and Panel B for stocks traded on
Nasdaq.
Panel A. NYSE
19
15
11
7
3
-1
-5
-9
-13
-24
-18
-12
-6
0
6
12
18
24
Event Tim e (m onths)
intense institutional selling
intense individual selling
Panel B. Nasdaq
Cumulative Market-Adjusted
Returns (%)
Cumulative Market-Adjusted
Returns (%)
23
93
73
53
33
13
-7
-24
-18
-12
-6
0
6
12
18
Event Tim e (m onths)
intense institutional selling
44
intense individual selling
24
Table I
Stocks Characteristics, Comparing our Trading Measures and the Holding Measure
At the end of each quarter, stocks are sorted into quintiles portfolios according to institutional holdings at
the beginning of the quarter, and then sorted within the quintiles according to institutional net trading
(Panel A) or institutional trading (Panels B, C, and D). Stocks characteristics are averaged within each of
the 25 portfolios for this quarter, and then the time series average over all quarters is calculated. All
characteristics are measured at the beginning of the quarter. Panel A reports six-month, one-year and twoyear past returns, calculated as the compounded return over these time periods. Panel B reports the average
size decile, where size decile is the rank of the market capitalization of equity, based on NYSE size decile
cutoff. Panel C reports the average beta, where beta is estimated for each stock by the market model, using
monthly returns in the 24-60 months prior to the trading quarter. Panel D reports the average natural log of
the book-to-market ratio, computed as the book value of the stock for the calendar quarter, lagged by sixmonth, divided by its market capitalization at the beginning of the quarter. The sample’s trading period is
January 1986 through December 2001, and the results are shown separately for stocks traded on the NYSE
and Nasdaq.
Quintiles of
Institutional
Net Trading
Quintiles of Institutional Holdings
low
2
3
4
high
Quintiles of Institutional Holdings
low
2
3
4
high
Panel A. Past six-month, one-year, and two-year Returns (Momentum)
NYSE
Nasdaq
Q1 (net sellers)
-1.51
0.93
13.59
1.56
5.27
21.37
4.20
9.92
26.90
4.27
10.27
27.03
6.87
14.71
33.04
-0.05
6.54
31.72
-2.90
0.91
24.13
-0.10
5.32
26.78
1.14
8.30
32.84
5.35
17.43
48.95
Q2
1.62
4.89
19.54
4.22
11.81
31.17
5.42
13.68
35.71
7.26
16.11
38.97
9.72
22.06
46.52
4.37
14.08
38.62
3.86
14.13
47.26
4.98
16.48
48.06
8.52
26.39
67.39
13.49
37.81
97.40
Q3
6.74
11.78
24.29
6.75
14.76
35.07
7.38
16.45
38.32
8.59
18.82
43.46
10.62
23.72
49.34
15.22
26.87
57.19
13.49
29.28
61.79
13.81
29.30
69.34
15.32
34.42
78.83
18.41
47.61
115.81
Q4
8.23
14.89
30.57
8.99
16.42
34.11
9.08
17.35
36.79
8.72
18.45
41.61
10.47
22.49
48.95
14.96
27.33
62.67
14.83
31.61
65.28
16.28
32.26
61.28
19.18
42.09
79.36
20.87
47.74
104.24
Q5 (net buyers)
10.70
19.04
34.51
12.03
22.41
42.48
12.80
25.11
49.52
12.43
24.41
50.55
14.37
30.57
61.58
14.63
30.11
59.71
17.80
29.85
59.38
21.38
38.99
70.39
24.48
48.22
92.04
29.19
62.44
126.75
All
5.04
10.02
24.12
6.67
14.02
32.64
7.75
16.43
37.30
8.24
17.56
40.22
10.40
22.66
47.69
9.59
20.50
49.43
9.08
20.55
50.82
11.07
23.88
54.41
13.65
31.54
69.34
17.41
42.35
98.06
45
Quintiles of
Institutional
Trading
Quintiles of Institutional Holdings
low
2
3
4
high
Quintiles of Institutional Holdings
low
2
3
4
high
Panel B. Size Decile
NYSE
Nasdaq
Q1 (low)
3.74
5.29
6.57
7.20
7.38
2.54
2.97
3.60
4.28
5.54
Q2
4.76
6.29
6.95
7.41
7.37
2.85
3.22
3.64
4.17
5.06
Q3
5.21
6.28
6.82
7.23
7.14
3.17
3.46
3.87
4.21
4.86
Q4
5.18
5.81
6.33
6.79
6.80
3.40
3.70
4.05
4.17
4.63
Q5 (high)
5.02
5.26
5.73
6.06
6.19
3.59
3.72
3.89
4.01
4.35
All
4.78
5.79
6.48
6.94
6.98
3.11
3.42
3.81
4.17
4.89
Panel C. Beta
NYSE
Nasdaq
Q1 (low)
1.02
1.09
1.21
1.22
1.26
1.30
1.51
1.59
1.60
1.72
Q2
0.86
0.94
1.07
1.10
1.19
1.15
1.31
1.37
1.42
1.53
Q3
0.80
0.91
1.03
1.07
1.15
0.89
1.06
1.18
1.27
1.35
Q4
0.83
0.92
1.00
1.05
1.10
0.73
0.94
1.02
1.12
1.22
Q5 (high)
0.82
0.88
0.95
1.01
1.05
0.66
0.84
0.89
1.00
1.08
All
0.87
0.95
1.05
1.09
1.15
0.95
1.14
1.21
1.28
1.38
Panel D. ln(B/M)
NYSE
Nasdaq
Q1 (low)
-0.52
-0.58
-0.69
-0.74
-0.78
-1.45
-1.05
-1.07
-1.13
-1.35
Q2
-0.58
-0.62
-0.76
-0.79
-0.74
-1.09
-0.95
-0.94
-1.00
-1.19
Q3
-0.62
-0.66
-0.75
-0.77
-0.75
-0.92
-0.86
-0.83
-0.93
-1.04
Q4
-0.65
-0.68
-0.74
-0.77
-0.73
-0.82
-0.84
-0.84
-0.84
-0.93
Q5 (high)
-0.62
-0.64
-0.69
-0.69
-0.71
-0.75
-0.77
-0.77
-0.76
-0.89
All
-0.60
-0.63
-0.72
-0.75
-0.74
-1.02
-0.90
-0.89
-0.93
-1.08
46
Table II
Comparing between the Risk-Characteristics of Stocks Traded
by Institutions and Individuals
At the end of each quarter, stocks are sorted into quintiles portfolios according to their
institutional trading measure. Stocks in quintile 1 (Q1) are the stocks with the lowest institutional
trading (highest individual trading), stocks in quintile 5 (Q5) are the ones with the highest
institutional trading, etc. The returns on the quintile portfolios are calculated over the three
months of institutional trading proxy formation, equal weighting the stocks within each quintile.
The three-month return series are linked across quarters to form a monthly series of returns on
each quintile portfolio, which are used to estimate the time series regressions of Sharpe-Lintner
CAPM model, and Fama and French (1993) three-factor model. The table reports the risk factors
sensitivities of CAPM on beta, and of Fama and French three-factor model on beta, size (s), and
book-to-market (h). The factors sensitivities are reported for each quintile portfolio as well as for
the Q5-Q1 portfolio, constructed by going long quintile 5 and short quintile 1. The results are
shown separately for stocks traded on the NYSE and Nasdaq. The sample period is January 1986
through December 2001. We report the statistical significance of the results (*, ** denote
significant at the 5-percent and 1-percent levels, respectively) only for the Q5-Q1 portfolio.
Quintiles of
Institutional
Trading
CAPM
β
FF Factors Sensitivities
CAPM
s
β
β
h
FF Factors Sensitivities
s
β
NYSE
h
Nasdaq
Q1 (low)
1.05
1.18
0.56
0.55
1.69
1.22
1.28
-0.73
Q2
0.94
1.14
0.29
0.62
1.52
1.21
1.00
-0.41
Q3
0.89
1.10
0.23
0.63
1.33
1.14
0.82
-0.15
Q4
0.90
1.11
0.29
0.63
1.14
1.10
0.73
0.16
Q5 (high)
0.88
1.07
0.37
0.63
0.94
0.98
0.61
0.31
-0.17**
-0.11
0.07
-0.75
Q5-Q1
*
-0.19
**
47
**
-0.24
**
-0.67
**
1.04
**
Table III
Comparing between the Risk-Characteristics of Stocks in which
Institutions are Net Sellers and Net Buyers
At the end of each quarter, stocks are sorted into quintiles portfolios according to their
institutional net trading measure. Stocks in quintile 1 (Q1) are the stocks with the highest
institutional net selling, and stocks in quintile 5 (Q5) are the ones with the highest institutional net
buying. The returns on the quintile portfolios are calculated over the three months of institutional
net trading proxy formation, equal weighting the stocks within each quintile. The three-month
return series are linked across quarters to form a monthly series of returns on each quintile
portfolio, which are used to estimate the time series regressions of Carhart’s (1997) four-factor
model. The table reports the risk factors sensitivities of the market (β), size (s), book-to-market
(h) and one-year past return momentum. The factors sensitivities are reported for each quintile
portfolio as well as for the Q5-Q1 portfolio, constructed by going long quintile 5 and short
quintile 1. The results are shown separately for stocks traded on the NYSE and Nasdaq. The
sample period is January 1986 through December 2001. We report the statistical significance of
the results (** denote significant at the 1-percent level) only for the Q5-Q1 portfolio.
Quintiles of
Institutional
Net Trading
Four-Factor Sensitivities
β
s
h
Four-Factor Sensitivities
m
β
NYSE
s
h
m
Nasdaq
Q1 (net sellers)
1.09
0.42
0.61
-0.23
1.07
0.79
-0.03
-0.43
Q2
1.14
0.29
0.54
-0.22
1.16
1.03
-0.43
-0.41
Q3
1.13
0.29
0.53
-0.18
1.17
0.95
-0.45
-0.31
Q4
1.10
0.33
0.56
-0.13
1.12
0.95
-0.21
-0.09
Q5 (net buyers)
1.07
0.47
0.60
-0.06
1.02
0.82
-0.01
-0.01
Q5-Q1
-0.02
0.05
-0.00
-0.06
0.04
0.02
0.43**
0.17**
48
Table IV
Returns Preceding Institutional Net Buying and Net Selling
This table reports four performance measures, in percentage per month, of institutional net trading quintile
portfolios, for the periods preceding the trading quarters. At the end of each quarter, stocks are sorted into
quintile portfolios according to their institutional net trading measure. Stocks in quintile 1 (Q1) are the
stocks with the highest institutional net selling, stocks in quintile 5 (Q5) are the ones with the highest
institutional net buying (individual net selling), and the portfolio Q5-Q1 is constructed by going long Q5
and short Q1. The returns on the quintile portfolios are calculated over one quarter (Qtr -1), one year (Year
-1), year two (Year -2) and two years (-2Years) periods preceding the trading quarter, equal weighting the
stocks within each quintile. The monthly returns series on each quintile portfolio are used to calculate the
average monthly returns (raw returns), as well as the risk-adjusted returns, estimated from the time series
regressions of the CAPM model, Fama and French (1993) three-factor model and Carhart’s (1997) fourfactor model. The table reports the raw returns (Panel A), and alphas in each model, as well as their tstatistics (in parentheses). The CAPM alpha (Panel B) is defined with respect to the market benchmark, the
Fama-French alpha (Panel C) with respect to the market, size, and book-to-market benchmarks, and the
four-factor alpha (Panel D) with respect to Fama and French and momentum benchmarks. The sample’s
trading period is January 1986 through December 2001 (hence the returns are from 1984 to 2001), and the
results are shown separately for stocks traded on the NYSE and Nasdaq.
Quintiles of
Institutional
Net Trading
Time Period
Qtr -1
Year -1
Year -2
Time Period
-2Years
Qtr -1
Year -1
Year -2
-2Years
Panel A. Raw Returns
NYSE
Q1 (net sellers)
0.31
0.64
Nasdaq
1.16
0.83
-0.32
0.50
1.89
0.98
Q2
0.87
1.09
1.43
1.20
0.89
1.47
2.62
1.79
Q3
1.28
1.34
1.45
1.34
2.21
2.34
2.66
2.25
Q4
1.49
1.44
1.41
1.38
2.89
2.59
2.31
2.31
Q5 (net buyers)
2.18
1.86
1.42
1.61
3.52
2.97
2.19
2.49
1.87
1.22
0.26
0.78
3.85
2.46
0.31
1.51
(16.78)
(17.63)
(4.65)
(13.42)
(14.29)
(14.07)
(2.22)
(9.48)
Q5-Q1
Panel B. CAPM Alphas
NYSE
Q1 (net sellers)
Q2
Q3
Q4
Q5 (net buyers)
Q5-Q1
Nasdaq
-0.78
-0.46
-0.15
-0.24
-1.59
-0.82
0.37
-0.30
(-4.07)
(-2.65)
(-1.04)
(-1.53)
(-5.25)
(-2.94)
(1.50)
(-1.11)
-0.22
-0.02
0.10
0.12
-0.56
0.01
0.98
0.40
(-1.33)
(-0.11)
(0.75)
(0.85)
(-1.62)
(0.02)
(3.32)
(1.30)
0.19
0.23
0.11
0.26
0.74
0.87
1.03
0.86
(1.25
(1.59)
(0.89)
(1.94)
(2.22)
(2.77)
(3.46)
(2.88)
0.38
0.31
0.06
0.28
1.52
1.22
0.79
1.02
(2.36)
(2.00)
(0.43)
(1.95)
(5.29)
(4.58)
(3.22)
(4.05)
1.09
0.75
0.11
0.55
2.26
1.73
0.84
1.34
(6.48)
(4.80)
(0.74)
(3.61)
(8.49)
(7.81)
(4.57)
(6.66)
1.88
1.21
0.26
0.79
3.85
2.55
0.47
1.64
(16.64)
(17.34)
(4.61)
(13.46)
(14.10)
(14.79)
(3.66)
(11.02)
49
Quintiles of
Institutional
Net Trading
Time Period
Qtr -1
Year -1
Year -2
Time Period
-2Years
Qtr -1
Year -1
Year -2
-2Years
Panel C. Fama-French Alphas
NYSE
Q1 (net sellers)
Q2
Q3
Q4
Q5 (net buyers)
Q5-Q1
Nasdaq
-1.07
-0.72
-0.30
-0.50
-1.40
-0.58
0.67
0.01
(-7.61)
(-5.64)
(-2.48)
(-4.12)
(-6.01)
(-3.14)
(6.77)
(0.08)
-0.49
-0.29
-0.05
-0.15
-0.16
0.40
1.39
0.86
(-4.01)
(-2.59)
(-0.45)
(-1.36)
(-0.83)
(2.53)
(11.08)
(5.74)
-0.06
-0.03
-0.02
0.01
1.15
1.27
1.45
1.31
(-0.53)
(-0.28)
(-0.20)
(0.11)
(6.85)
(8.69)
(11.99)
(9.78)
0.13
0.05
-0.07
0.03
1.82
1.50
1.08
1.32
(1.06)
(0.46)
(-0.61)
(0.29)
(13.07)
(12.85)
(10.70)
(11.75)
0.86
0.52
-0.03
0.30
2.53
1.91
0.97
1.49
(7.13)
(4.47)
(-0.24)
(2.64)
(18.64)
(18.32)
(10.20)
(15.21)
1.93
1.24
0.27
0.81
3.93
2.49
0.30
1.48
(16.96)
(17.51)
(4.70)
(13.41)
(14.17)
(14.22)
(3.00)
(10.42)
Panel D. Four-Factor Alphas
NYSE
Q1 (net sellers)
Q2
Q3
Q4
Q5 (net buyers)
Q5-Q1
Nasdaq
-0.78
-0.48
-0.01
-0.26
-0.88
-0.19
0.71
0.35
(-6.47)
(-4.21)
(-0.07)
(-2.40)
(-4.60)
(-1.21)
(6.87)
(2.40)
-0.26
-0.09
0.24
0.05
0.25
0.69
1.45
1.12
(-2.41)
(-0.90)
(2.55)
(0.47)
(1.47)
(4.78)
(11.02)
(8.00)
0.10
0.12
0.26
0.17
1.43
1.49
1.48
1.52
(0.97)
(1.24)
(2.81)
(1.67)
(9.38)
(10.85)
(11.66)
(11.84)
0.31
0.27
0.24
0.24
1.95
1.65
1.15
1.47
(2.73)
(2.49)
(2.57)
(2.33)
(13.97)
(14.69)
(10.85)
(13.52)
0.97
0.67
0.25
0.47
2.45
1.95
1.05
1.57
(7.94)
(6.01)
(2.37)
(4.24)
(17.63)
(18.12)
(10.60)
(15.72)
1.75
1.15
0.26
0.73
3.33
2.14
0.33
1.21
(16.39)
(16.75)
(4.27)
(12.37)
(14.38)
(14.11)
(3.17)
(9.39)
50
Table V
Returns Preceding Institutional and Individual Sales
This table reports four performance measures, in percentage per month, of institutional selling quintile
portfolios, for the periods preceding the trading quarters. At the end of each quarter, stocks are sorted into
quintile portfolios according to their institutional selling measure. Stocks in quintile 1 (Q1) are the stocks
with the lowest institutional selling (highest individual selling), stocks in quintile 5 (Q5) are the ones with
the highest institutional selling, and the portfolio Q5-Q1 is constructed by going long Q5 and short Q1.
The returns on the quintile portfolios are calculated over one quarter (Qtr -1), one year (Year -1), year two
(Year -2) and two years (-2Years) periods preceding the trading quarter, equal weighting the stocks within
each quintile. The monthly returns series on each quintile portfolio are used to calculate the average
monthly returns (raw returns), as well as the risk-adjusted returns, estimated from the time series
regressions of the CAPM model, Fama and French (1993) three-factor model and Carhart’s (1997) fourfactor model. The table reports the raw returns (Panel A), and alphas in each model, as well as their tstatistics (in parentheses). The CAPM alpha (Panel B) is defined with respect to the market benchmark, the
Fama-French alpha (Panel C) with respect to the market, size, and book-to-market benchmarks, and the
four-factor alpha (Panel D) with respect to Fama and French and momentum benchmarks. The sample’s
trading period is January 1986 through December 2001 (hence the returns are from 1984 to 2001), and the
results are shown separately for stocks traded on the NYSE and Nasdaq.
Quintiles of
Institutional
Selling
Time Period
Qtr -1
Year -1
Year -2
Time Period
-2Years
Qtr -1
Year -1
Year -2
-2Years
Panel A. Raw Returns
NYSE
Nasdaq
Q1 (low)
1.68
1.41
1.32
1.30
3.17
2.57
2.45
2.28
Q2
1.27
1.38
1.50
1.38
2.38
2.45
2.75
2.34
Q3
1.15
1.29
1.48
1.32
1.67
2.12
2.54
2.14
Q4
1.10
1.23
1.40
1.26
1.10
1.64
2.30
1.82
Q5 (high)
0.91
1.04
1.18
1.06
0.84
1.11
1.71
1.29
-0.77
-0.37
-0.14
-0.24
-2.33
-1.46
-0.74
-0.99
(-4.72)
(-2.67)
(-1.19)
(-1.87)
(-5.33)
(-4.03)
(-2.68)
(-3.07)
Q5-Q1
Panel B. CAPM Alphas
NYSE
Q1 (low)
Q2
Q3
Q4
Q5 (high)
Q5-Q1
Nasdaq
0.53
0.25
-0.05
0.18
1.66
1.10
0.90
0.91
(2.80)
(1.44)
(-0.33)
(1.13)
(3.50)
(2.55)
(2.46)
(2.27)
0.16
0.25
0.16
0.29
0.88
0.95
1.09
0.93
(0.97)
(1.65)
(1.26)
(2.10)
(2.51)
(2.86)
(3.48)
(2.94)
0.07
0.19
0.15
0.26
0.27
0.69
0.93
0.78
(0.44)
(1.22)
(1.12)
(1.78)
(0.93)
(2.52)
(3.63)
(3.03)
0.03
0.14
0.07
0.20
-0.19
0.32
0.79
0.56
(0.14)
(0.84)
(0.50)
(1.32)
(-0.84)
(1.48)
(3.84)
(2.72)
-0.13
-0.02
-0.12
0.02
-0.27
-0.05
0.36
0.17
(-0.74)
(-0.14)
(-0.78)
(0.14)
(-1.36)
(-0.27)
(2.07)
(0.92)
-0.66
-0.27
-0.07
-0.16
-1.93
-1.15
-0.54
-0.75
(-4.20)
(-2.04)
(-0.57)
(-1.31)
(-4.83)
(-3.40)
(-1.98)
(-2.46)
51
Quintiles of
Institutional
Selling
Time Period
Qtr -1
Year -1
Year -2
Time Period
-2Years
Qtr -1
Year -1
Year -2
-2Years
Panel C. Fama-French Alphas
NYSE
Q1 (low)
Q2
Q3
Q4
Q5 (high)
Q5-Q1
Nasdaq
0.35
0.05
-0.16
-0.02
2.28
1.65
1.37
1.50
(2.70)
(0.44)
(-1.36)
(-0.13)
(8.41)
(6.49)
(7.54)
(6.34)
-0.11
-0.01
0.02
0.03
1.36
1.40
1.53
1.43
(-0.93)
(-0.12)
(0.16)
(0.32)
(8.35)
(9.57)
(11.66)
(10.12)
-0.21
-0.09
-0.01
-0.02
0.55
1.00
1.27
1.13
(-1.64)
(-0.79)
(-0.06)
(-0.20)
(3.88)
(8.64)
(11.95)
(10.42)
-0.26
-0.14
-0.08
-0.07
-0.02
0.50
1.02
0.77
(-1.94)
(-1.13)
(-0.64)
(-0.58)
(-0.16)
(4.92)
(11.38)
(8.23)
-0.40
-0.29
-0.25
-0.24
-0.26
-0.03
0.47
0.23
(-2.97)
(-2.22)
(-1.88)
(-1.79)
(-1.82)
(-0.22)
(5.33)
(2.07)
-0.75
-0.34
-0.09
-0.22
-2.53
-1.68
-0.91
-1.27
(-5.05)
(-2.74)
(-0.73)
(-1.84)
(-8.98)
(-6.78)
(-4.56)
(-5.66)
Panel D. Four-Factor Alphas
NYSE
Q1 (low)
Q2
Q3
Q4
Q5 (high)
Q5-Q1
Nasdaq
0.58
0.31
0.15
0.25
2.60
1.99
1.38
1.82
(5.02)
(3.08)
(1.53)
(2.49)
(9.81)
(8.13)
(7.19)
(7.95)
0.07
0.17
0.29
0.21
1.60
1.60
1.56
7.95
(0.63)
(1.63)
(3.07)
(2.12)
(10.40)
(11.37)
(11.28)
(11.67)
-0.04
0.07
0.27
0.14
0.76
1.16
1.35
1.29
(-0.34)
(0.66)
(2.74)
(1.33)
(5.61)
(10.35)
(12.30)
(12.27)
-0.06
0.02
0.21
0.10
0.18
0.64
1.10
0.92
(-0.50)
(0.20)
(1.89)
(0.80)
(1.63)
(6.72)
(11.93)
(10.34)
-0.22
-0.10
0.05
-0.04
0.04
0.22
0.55
0.45
(-1.66)
(-0.79)
(0.43)
(-0.33)
(0.37)
(2.15)
(6.16)
(4.55)
-0.80
-0.41
-0.10
-0.29
-2.56
-1.77
-0.83
-1.37
(-5.20)
(-3.15)
(-0.82)
(-2.33)
(-8.73)
(-6.94)
(-3.96)
(-5.95)
52
Table VI
Returns Following Institutional Net Buying and Net Selling
This table reports four performance measures, in percentage per month, of institutional net trading quintile
portfolios, for the periods following the trading quarters. At the end of each quarter, stocks are sorted into
quintile portfolios according to their institutional net trading measure. Stocks in quintile 1 (Q1) are the
stocks with the highest institutional net selling (individual net buying), stocks in quintile 5 (Q5) are the
ones with the highest institutional net buying, and the portfolio Q5-Q1 is constructed by going long Q5 and
short Q1. The returns on the quintile portfolios are calculated over one quarter (Qtr 1), one year (Year 1),
year two (Year 2) and two years (2Years) periods following the trading quarter, equal weighting the stocks
within each quintile. The monthly returns series on each quintile portfolio are used to calculate the average
monthly returns (raw returns), as well as the risk-adjusted returns, estimated from the time series
regressions of the CAPM model, Fama and French (1993) three-factor model and Carhart’s (1997) fourfactor model. The table reports the raw returns (Panel A), and alphas in each model, as well as their tstatistics (in parentheses). The CAPM alpha (Panel B) is defined with respect to the market benchmark, the
Fama-French alpha (Panel C) with respect to the market, size, and book-to-market benchmarks, and the
four-factor alpha (Panel D) with respect to Fama and French and momentum benchmarks. The sample’s
trading period is January 1986 through December 2001 (hence the returns are from 1986 to 2003), and the
results are shown separately for stocks traded on the NYSE and Nasdaq.
Quintiles of
Institutional
Net Trading
Time Period
Qtr 1
Year 1
Year 2
Time Period
2Years
Qtr 1
Year 1
Year 2
2Years
Panel A. Raw Returns
NYSE
Nasdaq
Q1 (net sellers)
1.30
1.08
1.16
1.22
1.58
1.36
1.89
1.79
Q2
1.14
1.02
1.12
1.16
1.39
1.35
1.80
1.74
Q3
1.17
1.03
1.14
1.16
1.84
1.52
1.65
1.71
Q4
1.19
1.09
1.16
1.20
1.82
1.57
1.60
1.71
Q5 (net buyers)
1.36
1.14
1.17
1.24
1.79
1.43
1.34
1.51
0.06
0.06
0.01
0.02
0.20
0.07
-0.55
-0.28
(0.47)
(0.52)
(0.13)
(0.20)
(0.78)
(0.38)
(-3.50)
(-2.11)
Q5-Q1
Panel B. CAPM Alphas
NYSE
Q1 (net sellers)
Q2
Q3
Q4
Q5 (net buyers)
Q5-Q1
Nasdaq
0.30
0.24
0.29
0.30
0.42
0.39
0.88
0.72
(1.39)
(1.14)
(1.44)
(1.56)
(1.24)
(1.36)
(3.14)
(2.64)
0.13
0.17
0.22
0.22
0.07
0.25
0.66
0.52
(0.77)
(1.02)
(1.34)
(1.39)
(0.20)
(0.73)
(1.98)
(1.61)
0.15
0.16
0.25
0.21
0.51
0.39
0.47
0.46
(0.95)
(1.05)
(1.57)
(1.43)
(1.51)
(1.18)
(1.49)
(1.49)
0.18
0.22
0.28
0.26
0.56
0.50
0.48
0.52
(1.07)
(1.42)
(1.66)
(1.67)
(1.98)
(1.87)
(1.83)
(2.05)
0.36
0.28
0.28
0.30
0.61
0.44
0.30
0.41
(2.24)
(1.69)
(1.65)
(1.84)
(2.56)
(1.94)
(1.33)
(1.91)
0.06
0.04
-0.01
-0.00
0.19
0.05
-0.58
-0.31
(0.47)
(0.41)
(-0.08)
(-0.01)
(0.72)
(0.27)
(-3.66)
(-2.32)
53
Quintiles of
Institutional
Net Trading
Time Period
Qtr 1
Year 1
Year 2
Time Period
2Years
Qtr 1
Year 1
Year 2
2Years
Panel C. Fama-French Alphas
NYSE
Q1 (net sellers)
Q2
Q3
Q4
Q5 (net buyers)
Q5-Q1
Nasdaq
-0.04
-0.08
-0.13
-0.09
0.47
0.38
0.71
0.60
(-0.26)
(-0.57)
(-1.11)
(-0.72)
(1.76)
(1.81)
(4.46)
(3.52)
-0.16
-0.11
-0.12
-0.10
0.36
0.44
0.66
0.59
(-1.24)
(-0.97)
(-1.13)
(-1.00)
(1.65)
(1.99)
(3.21)
(2.93)
-0.13
-0.10
-0.08
-0.10
0.82
0.66
0.55
0.60
(-1.18)
(-1.00)
(-0.79)
(-0.97)
(5.01)
(3.57)
(2.75)
(3.37)
-0.11
-0.04
-0.06
-0.05
0.80
0.67
0.49
0.59
(-0.93)
(-0.35)
(-0.56)
(-0.52)
(7.65)
(5.20)
(3.32)
(4.53)
0.09
0.02
-0.05
-0.01
0.75
0.54
0.22
0.40
(0.82)
(0.19)
(-0.47)
(-0.12)
(6.21)
(4.85)
(1.53)
(3.38)
0.13
0.10
0.08
0.07
0.28
0.16
-0.49
-0.20
(1.00)
(0.97)
(0.92)
(0.92)
(1.03)
(0.87)
(-3.28)
(-1.59)
Panel D. Four-Factor Alphas
NYSE
Q1 (net sellers)
Q2
Q3
Q4
Q5 (net buyers)
Q5-Q1
Nasdaq
0.27
0.17
0.04
0.10
1.10
0.91
0.96
0.96
(2.09)
(1.39)
(0.35)
(0.97)
(5.83)
(6.28)
(6.77)
(7.37)
0.09
0.10
0.06
0.08
0.83
0.97
1.04
1.01
(0.87)
(1.08)
(0.59)
(0.87)
(4.91)
(5.90)
(6.08)
(6.43)
0.02
0.07
0.09
0.06
1.08
1.05
1.00
0.99
(0.23)
(0.72)
(0.96)
(0.68)
(7.33)
(7.01)
(6.90)
(7.20)
0.00
0.10
0.09
0.07
0.90
0.93
0.82
0.86
(0.04)
(0.92)
(0.83)
(0.71)
(8.75)
(8.58)
(7.69)
(8.60)
0.14
0.11
0.09
0.09
0.77
0.68
0.53
0.62
(1.21)
(0.92)
(0.80)
(0.86)
(6.21)
(6.36)
(4.98)
(6.33)
-0.13
-0.07
0.05
-0.01
-0.33
-0.23
-0.42
-0.34
(-1.24)
(-0.69)
(0.54)
(-0.17)
(-1.66)
(-1.61)
(-2.80)
(-2.81)
54
Table VII
Returns Following Institutional and Individual Purchases
This table reports four performance measures, in percentage per month, of institutional buying quintile
portfolios, for the periods following the trading quarters. At the end of each quarter, stocks are sorted into
quintile portfolios according to their institutional buying measure. Stocks in quintile 1 (Q1) are the stocks
with the lowest institutional buying (highest individual buying), stocks in quintile 5 (Q5) are the ones with
the highest institutional buying, and the portfolio Q5-Q1 is constructed by going long Q5 and short Q1.
The returns on the quintile portfolios are calculated over one quarter (Qtr 1), one year (Year 1), year two
(Year 2) and two years (2Years) periods following the trading quarter, equal weighting the stocks within
each quintile. The monthly returns series on each quintile portfolio are used to calculate the average
monthly returns, as well as the risk-adjusted returns, estimated from the time series regressions of the
CAPM model, Fama and French (1993) three-factor model and Carhart’s (1997) four-factor model. The
table reports the raw returns (Panel A), and alphas in each model, as well as their t-statistics (in
parentheses). The CAPM alpha (Panel B) is defined with respect to the market benchmark, the FamaFrench alpha (Panel C) with respect to the market, size, and book-to-market benchmarks, and the fourfactor alpha (Panel D) with respect to Fama and French and momentum benchmarks. The sample’s trading
period is January 1986 through December 2001 (hence the returns are from 1986 to 2003), and the results
are shown separately for stocks traded on the NYSE and Nasdaq.
Quintiles of
Institutional
Buying
Time Period
Qtr 1
Year 1
Year 2
Time Period
2Years
Qtr 1
Year 1
Year 2
2Years
Panel A. Raw Returns
NYSE
Nasdaq
Q1 (low)
1.06
0.98
1.18
1.19
1.70
1.67
2.20
2.16
Q2
1.24
1.07
1.14
1.19
1.75
1.53
1.71
1.75
Q3
1.26
1.08
1.10
1.16
1.82
1.45
1.56
1.63
Q4
1.22
1.09
1.14
1.20
1.55
1.35
1.46
1.52
Q5 (high)
1.38
1.13
1.19
1.24
1.63
1.26
1.35
1.42
0.32
0.15
0.01
0.05
-0.07
-0.41
-0.85
-0.74
(1.45)
(0.80)
(0.09)
(0.30)
(-0.15)
(-0.90)
(-2.01)
(-1.78)
Q5-Q1
Panel B. CAPM Alphas
NYSE
Q1 (low)
Q2
Q3
Q4
Q5 (high)
Q5-Q1
Nasdaq
0.00
0.09
0.25
0.20
0.30
0.50
0.97
0.85
(0.01)
(0.39)
(1.21)
(1.01)
(0.55)
(0.97)
(1.99)
(1.78)
0.23
0.22
0.26
0.25
0.39
0.39
0.52
0.49
(1.52)
(1.44)
(1.64)
(1.75)
(1.10)
(1.13)
(1.62)
(1.52)
0.26
0.23
0.22
0.23
0.56
0.38
0.46
0.45
(1.53)
(1.38)
(1.32)
(1.47)
(2.10)
(1.52)
(1.81)
(1.87)
0.23
0.24
0.25
0.27
0.39
0.37
0.44
0.43
(1.21)
(1.33)
(1.41)
(1.55)
(1.74)
(1.73)
(2.03)
(2.10)
0.39
0.29
0.32
0.31
0.54
0.34
0.39
0.40
(2.24)
(1.62)
(1.78)
(1.84)
(2.84)
(1.77)
(1.90)
(2.16)
0.38
0.20
0.07
0.11
0.24
-0.16
-0.58
-0.45
(1.76)
(1.13)
(0.46)
(0.70)
(0.51)
(-0.38)
(-1.46)
(-1.16)
55
Quintiles of
Institutional
Buying
Time Period
Qtr 1
Year 1
Year 2
Time Period
2Years
Qtr 1
Year 1
Year 2
2Years
Panel C. Fama-French Alphas
NYSE
Q1 (low)
Q2
Q3
Q4
Q5 (high)
Q5-Q1
Nasdaq
-0.25
-0.16
-0.10
-0.12
0.78
0.88
1.14
1.09
(-1.30)
(-0.98)
(-0.69)
(-0.81)
(2.03)
(2.48)
(3.54)
(3.36)
-0.05
-0.04
-0.08
-0.05
0.75
0.66
0.60
0.63
(-0.44)
(-0.40)
(-0.78)
(-0.54)
(3.95)
(3.13)
(2.95)
(3.23)
-0.05
-0.06
-0.12
-0.09
0.75
0.51
0.41
0.47
(-0.42)
(-0.54)
(-1.13)
(-0.88)
(5.89)
(3.56)
(2.80)
(3.41)
-0.10
-0.06
-0.12
-0.08
0.45
0.39
0.30
0.36
(-0.75)
(-0.50)
(-0.98)
(-0.67)
(3.81)
(3.38)
(2.38)
(3.10)
0.10
0.01
-0.03
-0.01
0.52
0.28
0.19
0.26
(0.78)
(0.08)
(-0.25)
(-0.12)
(4.41)
(2.74)
(1.59)
(2.44)
0.35
0.17
0.07
0.10
-0.26
-0.60
-0.94
-0.83
(1.63)
(0.98)
(0.45)
(0.66)
(-0.67)
(-1.71)
(-3.07)
(-2.68)
Panel D. Four-Factor Alphas
NYSE
Q1 (low)
Q2
Q3
Q4
Q5 (high)
Q5-Q1
Nasdaq
0.18
0.21
0.18
0.18
1.47
1.59
1.67
1.66
(1.24)
(1.68)
(1.57)
(1.63)
(4.55)
(5.29)
(5.95)
(5.98)
0.10
0.13
0.09
0.10
1.10
1.13
1.03
1.06
(0.97)
(1.39)
(1.00)
(1.18)
(6.95)
(7.11)
(6.76)
(7.19)
0.08
0.07
0.01
0.02
0.95
0.83
0.74
0.77
(0.63)
(0.60)
(0.08)
(0.23)
(8.37)
(7.57)
(6.68)
(7.40)
0.01
0.04
0.01
0.02
0.59
0.60
0.55
0.58
(0.07)
(0.30)
(0.05)
(0.21)
(5.18)
(6.15)
(5.53)
(6.18)
0.17
0.11
0.08
0.08
0.59
0.41
0.43
0.44
(1.32)
(0.85)
(0.68)
(0.69)
(4.98)
(4.23)
(4.34)
(4.69)
-0.01
-0.10
-0.10
-0.10
-0.88
-1.18
-1.24
-1.22
(-0.04)
(-0.65)
(-0.66)
(-0.69)
(-2.58)
(-3.66)
(-4.14)
(-4.20)
56
Table VIII
Returns Following Institutional Net Buying and Net Selling: Subperiod Analysis
This table reports subperiod results of two performance measures, in percentage per month, for institutional
net trading quintile portfolios in the periods following the trading quarters. At the end of each quarter,
stocks are sorted into quintile portfolios according to their institutional net trading measure. Stocks in
quintile 1 (Q1) are the stocks with the highest institutional net selling (individual net buying), stocks in
quintile 5 (Q5) are the ones with the highest institutional net buying, and the portfolio Q5-Q1 is constructed
by going long Q5 and short Q1. The returns on the quintiles portfolios are calculated over one quarter (Qtr
1), one year (Year 1), year two (Year 2) and two years (2Years) periods following the trading quarter, equal
weighting the stocks within each quintile. The monthly returns series on each quintile portfolio are used to
calculate the average monthly returns (raw returns), as well as the four-factor risk-adjusted returns,
estimated from the time series regression of the Carhart’s (1997) four-factor model with respect to the
market, size, book-to-market and momentum benchmarks. The table reports the raw returns (Panel A), and
the four-factor alphas (Panel B), as well as their t-statistics (in parentheses). This table reports the returns
within two subperiods: the late 1990s bubble period, where the trading period is 1995-2000, and the prebubble period, where the trading period is 1986-1994. The results are shown separately for stocks traded
on the NYSE and Nasdaq.
Panel A. Raw Returns
Quintiles of
Institutional
Net Trading
Time Period
Qtr 1
Year 1
Year 2
Time Period
2Years
Qtr 1
Year 1
Year 2
2Years
The Late 1990s Bubble Period: 1995-2000
NYSE
Nasdaq
Q1 (net sellers)
1.37
1.36
1.05
1.18
1.66
2.04
1.85
1.85
Q2
1.30
1.32
0.92
1.09
1.77
2.18
1.94
1.94
Q3
1.24
1.30
0.94
1.06
2.25
2.31
1.52
1.78
Q4
1.26
1.27
0.84
1.02
1.79
2.08
1.35
1.57
Q5 (net buyers)
1.45
1.41
0.89
1.11
1.95
1.96
0.93
1.34
Q5-Q1
0.08
0.06
-0.16
-0.07
0.29
-0.08
-0.92
-0.50
(0.31)
(0.30)
(-1.13)
(-0.46)
(0.51)
(-0.21)
(-3.04)
(-1.90)
The Pre-Bubble Period: 1986-1994
NYSE
Nasdaq
Q1 (net sellers)
1.19
1.14
1.03
1.17
1.35
1.35
1.41
1.41
Q2
0.98
1.10
1.06
1.15
1.17
1.39
1.26
1.37
Q3
1.07
1.12
1.09
1.15
1.56
1.54
1.36
1.47
Q4
1.07
1.18
1.20
1.25
1.79
1.72
1.40
1.57
Q5 (net buyers)
1.24
1.23
1.18
1.26
1.64
1.60
1.26
1.46
Q5-Q1
0.06
0.09
0.15
0.09
0.29
0.26
-0.15
0.04
(0.35)
(0.69)
(1.23)
(0.88)
(1.50)
(2.04)
(-0.99)
(0.37)
57
Panel B. Four-Factor Alphas
Quintiles of
Institutional
Net Trading
Time Period
Qtr 1
Year 1
Year 2
Time Period
2Years
Qtr 1
Year 1
Year 2
2Years
The Late 1990s Bubble Period: 1995-2000
NYSE
Q1 (net sellers)
Q2
Q3
Q4
Q5 (net buyers)
Q5-Q1
Nasdaq
0.31
0.18
0.27
0.27
1.57
1.45
1.35
1.39
(1.69)
(1.03)
(1.46)
(1.59)
(4.02)
(4.80)
(5.38)
(6.04)
0.12
0.11
0.20
0.21
1.39
1.56
1.74
1.64
(0.64)
(0.67)
(1.22)
(1.40)
(3.76)
(4.45)
(5.48)
(5.71)
-0.15
0.05
0.23
0.16
1.57
1.45
1.58
1.53
(-0.88)
(0.31)
(1.42)
(1.16)
(5.14)
(5.34)
(5.74)
(6.20)
-0.17
-0.03
0.11
0.09
0.74
1.09
1.15
1.12
(-0.85)
(-0.20)
(0.67)
(0.60)
(3.37)
(5.51)
(5.39)
(5.75)
-0.00
0.08
0.13
0.15
0.73
0.89
0.62
0.78
(-0.03)
(0.42)
(0.79)
(0.98)
(2.98)
(4.08)
(3.00)
(4.01)
-0.32
-0.11
-0.14
-0.12
-0.84
-0.56
-0.74
-0.61
(-1.71)
(-0.63)
(-0.98)
(-0.86)
(-2.18)
(-1.77)
(-2.74)
(-2.42)
The Pre-Bubble Period: 1986-1994
NYSE
Q1 (net sellers)
Q2
Q3
Q4
Q5 (net buyers)
Q5-Q1
Nasdaq
0.40
0.21
0.02
0.13
0.60
0.48
0.48
0.44
(3.19)
(1.92)
(0.15)
(1.46)
(4.21)
(4.63)
(3.90)
(4.37)
0.08
0.08
0.03
0.06
0.47
0.56
0.41
0.45
(1.07)
(1.51)
(0.43)
(1.09)
(3.45)
(4.39)
(2.84)
(3.65)
0.12
0.07
0.03
0.03
0.76
0.66
0.46
0.51
(1.57)
(1.17)
(0.64)
(0.56)
(4.95)
(5.29)
(3.90)
(4.54)
0.11
0.13
0.18
0.14
0.98
0.81
0.49
0.60
(1.55)
(2.02)
(2.60)
(2.41)
(8.58)
(7.78)
(5.35)
(6.77)
0.23
0.16
0.18
0.16
0.76
0.64
0.40
0.48
(2.69)
(2.24)
(2.33)
(2.34)
(5.66)
(6.45)
(3.93)
(5.24)
-0.16
-0.05
0.16
0.04
0.16
0.16
-0.08
0.04
(-1.17)
(-0.42)
(1.44)
(0.42)
(0.92)
(1.48)
(-0.59)
(0.35)
58
Table IX
Returns Following Institutional and Individual Purchases: Subperiod Analysis
This table reports subperiod results of two performance measures, in percentage per month, for institutional
buying quintile portfolios in the periods following the trading quarters. At the end of each quarter, stocks
are sorted into quintile portfolios according to their institutional buying measure. Stocks in quintile 1 (Q1)
are the stocks with the lowest institutional buying (highest individual buying), stocks in quintile 5 (Q5) are
the ones with the highest institutional buying, and the portfolio Q5-Q1 is constructed by going long Q5 and
short Q1. The returns on the quintiles portfolios are calculated over one quarter (Qtr 1), one year (Year 1),
year two (Year 2) and two years (2Years) periods following the trading quarter, equal weighting the stocks
within each quintile. The monthly returns series on each quintile portfolio are used to calculate the average
monthly returns (raw returns), as well as the four-factor risk-adjusted returns, estimated from the time
series regression of the Carhart’s (1997) four-factor model with respect to the market, size, book-to-market
and momentum benchmarks. The table reports the raw returns (Panel A), and the four-factor alphas (Panel
B), as well as their t-statistics (in parentheses). This table reports the returns within two subperiods: the
late 1990s bubble period, where the trading period is 1995-2000, and the pre-bubble period, where the
trading period is 1986-1994. The results are shown separately for stocks traded on the NYSE and Nasdaq.
Panel A. Raw Returns
Quintiles of
Institutional
Buying
Time Period
Qtr 1
Year 1
Year 2
Time Period
2Years
Qtr 1
Year 1
Year 2
2Years
The Late 1990s Bubble Period: 1995-2000
NYSE
Nasdaq
Q1 (low)
1.31
1.37
0.99
1.14
2.46
2.88
2.32
2.48
Q2
1.37
1.30
0.90
1.05
1.99
2.18
1.60
1.75
Q3
1.21
1.27
0.88
1.04
1.71
2.00
1.31
1.50
Q4
1.31
1.35
0.86
1.06
1.57
1.81
1.28
1.46
Q5 (high)
1.41
1.37
1.00
1.16
1.75
1.75
1.03
1.29
Q5-Q1
0.10
0.00
0.02
0.02
-0.71
-1.13
-1.29
-1.20
(0.24)
(0.00)
(0.06)
(0.09)
(-0.61)
(-1.14)
(-1.36)
(-1.41)
The Pre-Bubble Period: 1986-1994
NYSE
Nasdaq
Q1 (low)
0.87
1.00
1.02
1.09
1.33
1.62
1.49
1.58
Q2
1.13
1.19
1.10
1.21
1.53
1.66
1.38
1.54
Q3
1.23
1.21
1.12
1.23
1.79
1.55
1.34
1.46
Q4
1.06
1.16
1.13
1.22
1.43
1.44
1.25
1.38
Q5 (high)
1.24
1.20
1.18
1.24
1.43
1.37
1.24
1.35
0.37
0.20
0.15
0.16
0.10
-0.25
-0.25
-0.24
(1.43)
(1.04)
(0.79)
(0.91)
(0.33)
(-0.95)
(-0.85)
(-0.92)
Q5-Q1
59
Panel B. Four-Factor Alphas
Quintiles of
Institutional
Buying
Time Period
Qtr 1
Year 1
Year 2
Time Period
2Years
Qtr 1
Year 1
Year 2
2Years
The Late 1990s Bubble Period: 1995-2000
NYSE
Q1 (low)
Q2
Q3
Q4
Q5 (high)
Q5-Q1
Nasdaq
0.52
0.37
0.43
0.44
2.86
2.68
2.57
2.60
(2.13)
(1.60)
(2.15)
(2.23)
(4.08)
(4.29)
(4.85)
(5.09)
0.02
0.07
0.16
0.15
1.48
1.44
1.63
1.55
(0.12)
(0.45)
(1.02)
(1.08)
(4.79)
(4.71)
(5.55)
(5.78)
-0.24
-0.07
0.11
0.08
0.75
1.10
0.99
1.01
(-1.22)
(-0.41)
(0.66)
(0.49)
(3.53)
(5.57)
(4.84)
(5.59)
-0.16
-0.02
0.05
0.06
0.41
0.72
0.81
0.80
(-0.75)
(-0.10)
(0.30)
(0.36)
(1.83)
(3.85)
(4.09)
(4.56)
-0.02
0.04
0.20
0.17
0.55
0.57
0.51
0.56
(-0.12)
(0.23)
(1.10)
(1.03)
(2.48)
(3.03)
(3.01)
(3.53)
-0.54
-0.33
-0.24
-0.26
-2.31
-2.10
-2.06
-2.05
(-1.82)
(-1.19)
(-1.14)
(-1.24)
(-3.25)
(-3.18)
(-3.71)
(-3.82)
The Pre-Bubble Period: 1986-1994
NYSE
Q1 (low)
Q2
Q3
Q4
Q5 (high)
Q5-Q1
Nasdaq
0.10
0.07
0.02
0.05
0.67
0.82
0.65
0.68
(0.56)
(0.49)
(0.16)
(0.39)
(2.79)
(3.93)
(2.83)
(3.29)
0.21
0.14
0.09
0.10
0.78
0.83
0.52
0.62
(2.88)
(2.46)
(1.61)
(1.91)
(4.70)
(6.23)
(4.10)
(5.13)
0.26
0.15
0.08
0.11
1.03
0.63
0.46
0.49
(3.55)
(2.38)
(1.27)
(1.79)
(7.73)
(5.91)
(4.65)
(5.07)
0.10
0.12
0.08
0.11
0.59
0.51
0.32
0.39
(1.08)
(1.54)
(1.03)
(1.43)
(5.40)
(5.52)
(3.49)
(4.41)
0.27
0.16
0.16
0.15
0.50
0.37
0.30
0.32
(2.84)
(1.98)
(1.78)
(1.90)
(4.31)
(3.89)
(3.02)
(3.53)
0.17
0.10
0.14
0.11
-0.17
-0.46
-0.35
-0.36
(0.77)
(0.56)
(0.81)
(0.69)
(-0.63)
(-2.02)
(-1.31)
(-1.59)
60