Directory UMM :Data Elmu:jurnal:J-a:Journal Of Business Research:Vol50.Issue3.2000:
The Pattern of Intraday Portfolio
Management Decisions: A Case Study of
Intraday Security Return Patterns
Stanley B. Block
TEXAS CHRISTIAN UNIVERSITY
Dan W. French
NEW MEXICO STATE UNIVERSITY
Edwin D. Maberly
UNIVERSITY OF CANTERBURY
This article examines causes of observed stock trading patterns that show
high hourly returns and trading volume during early and late trading
hours. Using time-stamped data from an institutional investor, we document high levels of portfolio managers’ early-morning and late-afternoon
decisions to trade that can result in the volume pattern and relatively
higher proportions of buy decisions that could contribute to the return
pattern. J BUSN RES 2000. 50.321–326. 2000 Elsevier Science Inc.
All rights reserved.
I
nvestors who trade stocks are competing with professional
traders and institutional investment managers, and to be
competitive, investors must be aware of the influences
that other traders have on market returns. Researchers have
identified a number of regularities in common stock returns,
and this article addresses the well-documented pattern of
stock returns during the average trading day that shows a
repeated occurrence of higher-than-expected returns at the open
and close of the market and lower returns during the middle
of the day. The hypotheses of the article is that the timing of
institutional investors’ decisions stimulates this pattern because managers tend to make the majority of their transaction
decisions toward the end of the day. These late-day decisions
affect end-of-day market returns and volume, and the beginning-of-day returns and volume are high because of the accumulation of orders from managers’ decisions made after the
market close the previous day plus decisions made before the
opening of trading each day.
Address correspondence to Dan W. French, Department of Finance, MSC
3FIN, New Mexico State University, Las Cruces, NM 88003-8001; Tel.: (505)
646-3201; fax: (505) 646-2820; E-mail: [email protected]
Journal of Business Research 50, 321–326 (2000)
2000 Elsevier Science Inc. All rights reserved.
655 Avenue of the Americas, New York, NY 10010
Possible Causes of Intraday
Return Patterns
The literature refers to the observed return regularity as the
“U-shaped pattern of intraday returns.” It has been identified
in a number of studies including work by Harris (1986, 1989),
Jain and Joh (1988), Wood, McInish, and Ord (1985), McInish
and Wood (1990), and Smirlock and Starks (1986). A similar
pattern (high early and late in the day, low during the middle
hours) applies to trading volume (see, for example, Foster
and Viswanathan, 1989; Jain and Joh, 1988; McInish and
Wood, 1991; and Wood et al., 1985), return variance (see,
for example, Admati and Pfleiderer, 1989a; Lockwood and
Linn, 1990; and Wei, 1992), and bid-ask spreads (see, for
example, Brock and Kleidon, 1992; McInish, and Wood,
1992; and Stoll and Whaley, 1990). The values for these
variables during market opening and closing times are distinctly larger than the midday values.
Some of the attempts to explain these non-stationarities
have relied on market microstructure models that allow patterns to develop based solely on the internal structure and
operation of the market and in the absence of any external
influences. Admati and Pfleiderer (1989b) develop an information-based model to explain trade clustering that can occur
at any time during the trading day, and they discuss the role
of the open and the close as special clustering points. They
contend that market makers have adverse selection problems
with informed traders that they attempt to alleviate through
their interaction with nondiscretionary liquidity traders, discretionary liquidity traders, and informed traders, so they offer
inducements for liquidity traders to transact at certain times
during the day.
ISSN 0148-2963/00/$–see front matter
PII S0148-2963(99)00023-5
322
J Busn Res
2000:50:321–326
Brock and Kleidon (1992) maintain that market evidence
is at odds with the Admati and Pfleiderer (1989b) informationbased model. The Admati-Pfleiderer model partially relies on
lower transaction costs to explain higher volume and returns
at the opening and close of the day. However, Brock and
Kleidon (1992), extending the inventory-based model of Garman (1976), show that spreads (transactions costs) should be
higher at the open and the close. They present evidence of
observed transaction costs as measured by the bid-ask spread
often being higher at these times.
In addition to explanations based on market microstructure, the U-shaped trading pattern may be attributable to
variables that are exogenous to the market structure as noted
in Brock and Kleidon (1992). Primary exogenous considerations are variations in supply and demand from investors and
the discontinuous nature of market trading. For example,
Brock and Kleidon argue that much of the trading at the open
and close is due to the fact that investors cannot trade as
easily when major markets are closed. The incentive to transact
at the opening and close may be related to the desire of
portfolio managers to bring their portfolios to a particular
target risk, a factor that is unrelated to the market microstructure. Another exogenous factor could be that market activity
is responding to the release of information. Berry and Howe
(1994) find distinct intraday patterns in the release of public
information but conclude that they are only moderately related
to trading volume and not significantly related to price variability. Other exogenous considerations relate to covering
short positions overnight (see Miller, 1989) and the desire of
portfolio managers of index funds to trade at the end of the
day to reduce tracking errors. Mutual fund managers, judged
by net asset value at the end of the day, also have some
incentive to correlate their trading patterns with their benchmarks. Furthermore, brokers are often given orders which
must be filled by the end of the day or canceled.
Whether the causes of intraday return patterns lie with
external or internal (market microstructure) considerations,
the fact remains that a predictive pattern is in place that
challenges market efficiency, although it is doubtful that a
trading strategy could exploit the patterns to the extent of
producing superior after transactions-costs returns. Studies of
the return phenomenon have typically dealt with transactions
data from major exchanges such as the New York Stock Exchange, the American Stock Exchange, and the Toronto Stock
Exchange. Since the data represent realized transactions, the
observed U-shaped pattern of returns and volume includes
both the external influence of investors and portfolio managers
who order the execution of trades for portfolio rebalancing
reasons and the internal influence of market makers and traders who attempt to minimize transaction costs and make trading profits. A shortcoming of the use of aggregated market
transactions data is that the external influence from portfolio
managers cannot be measured separately from the effects of
traders and market makers.
S. B. Block et al.
Gerety and Mulherin (1992) attribute the U-shaped pattern
in trading volume to short-term investors. These short-term
investors, such as market makers and day traders, have relatively little ability to bear risk overnight and desire to exchange
their positions with investors having a greater ability to bear
risk overnight. Gerety and Mulherin (1992) reason that if
investors are transferring risk of holding positions while the
market is closed, then end-of-day volume should be directly
related to overnight return variance. Using closing and opening-hour trading volume on the NYSE from 1933 to 1988,
they find evidence to support their position and conclude that
short-term traders are the cause of the U-shaped pattern in
volume.
Gerety and Mulherin (1992) mention anecdotal discussion
that attributes the U-shaped pattern to institutional investors.
However, they observe a downward trend in last-hour volume
over the last several decades and a concurrent increase in
block trading (to represent institutional trades) and suggest
that institutional traders do not cause the U-shaped pattern.
However, the simple observation of a negative correlation
between these two variables is not proof of a causal relationship. It is quite possible that both institutional investors and
short-term traders are parties to the causes of the U-shaped
pattern, especially if institutional investors represent the other
side of the trades with short-term traders. It is our hypothesis
in this article that a major part of the demand for opening
and closing trades originates with institutional and other longterm investors, and short-term traders are willing to fill that
demand.
While these previous studies have proposed and examined
various causes of intraday trading patterns, none has been
able to explain more than a small portion of the pattern. It
is the intent of this article to further contribute to the evidence
in this area by using a case study to illustrate the importance
of the timing of transaction decisions by institutional portfolio
managers and other long-term investors. These decisions
quickly become orders that impact the market and contribute
to the formation of the daily U-shaped patterns observed in
this article. As these intraday decisions are transmitted to
security traders, they are translated into intraday volume,
variance, and price patterns.
Transactions data typically used in market studies identifies
only the time of the transaction, not the time of the decision
to transact. To overcome this problem, our sample tracks
the times at which portfolio managers made their decisions.
Conversations with the portfolio managers led to the conclusion that the portfolio managers were, for the most part,
indifferent to the problems and issues related to the traders
(endogenous factors). Their concerns were related to rebalancing their portfolios. The extent to which they weight their
actions toward buying activities at market opening and market
closing times in the absence of concerns about execution
costs would tend to indicate the importance of the exogenous
supply-demand variable in the pattern of intraday returns.
Intraday Portfolio Management Decisions
J Busn Res
2000:50:321–326
323
Table 1. Portfolio Manager Transaction Decisions by Trading Period, 1988 (Percent of Total in Parentheses)
Number of Transaction Decisions
Trading period
(Eastern-time)
Before 10 a.m.
10 a.m. to 11 a.m.
11 a.m. to 12 noon
12 noon to 1 p.m.
1 p.m. to 2 p.m.
2 p.m. to 3 p.m.
3 p.m. to 4 p.m.
Total
Buys
500
168
235
212
124
133
302
1674
Volume, Millions of Dollars
Sells
(29.9)
(10.0)
(14.0)
(12.7)
(7.4)
(7.9)
(18.0)
(100.0)
440
207
346
341
118
163
258
1873
Any clustering of either buying or selling decisions at market
open and close would affect volume at those times.
Database and Methodology
The Trust Division of NationsBank Dallas provided a sample
of 3,547 equity transactions completed by the bank’s security
traders during 1988. During that period, the bank had approximately 25 portfolio managers. Their operation is such that
as soon as the portfolio managers reach a decision to place
an order for a security transaction, they communicate to the
trading desk by delivering a written ticket directing the trader
to initiate the purchase or sale of the security. The first task
of the trader is to stamp the time of receipt of the written
order. Then, the trader enters into the market and begins to
“work” the order, searching for the best execution possible
and trying to achieve execution of the order within a reasonably short time. Most orders are completed within 30 minutes
of receipt by the trading desk.
Thus, the time stamp on the order approximates the time
that the portfolio manager reached a transaction decision. It
also approximates the time that the trader began revealing
the order in the market. Thus, the important distinction of
this data set is that observations represent times of decisions
to trade, not necessarily transaction times. If demand and
supply from portfolio managers affect the pattern of intraday
returns, then we would expect to see the U-shaped pattern
occur in a metric that measures relative demand. In addition,
to the extent that the trading desk completes trades relatively
soon after receipt, the volume of portfolio manager decisions
should also exhibit the U-shaped pattern.
Each observation in the sample contains the following information: (1) the time stamp, (2) whether the order was a
buy or a sell, (3) the number of shares, (4) the stock symbol,
the trading location (New York Stock Exchange, American
Stock Exchange, or over the counter), and (5) the execution
price.
Results
We divided the trading day into the same seven trading periods
as did Gerety and Mulherin (1992). The first period is from
Buys
(23.5)
(11.1)
(18.5)
(18.2)
(6.3)
(8.7)
(13.8)
(100.0)
100.9
27.0
35.7
26.9
29.3
24.7
64.6
309.0
Sells
(32.7)
(8.7)
(11.6)
(8.7)
(9.5)
(8.0)
(20.9)
(100.0)
93.4
42.3
69.0
62.5
29.3
26.4
57.6
380.4
(24.6)
(11.1)
(18.1)
(16.4)
(7.7)
(6.9)
(15.1)
(100.0)
the market open (9:30 a.m. Eastern time) to 10:00 a.m. Each
of the other periods begins on the hour and lasts for one
hour, so that there are a total of seven trading periods during
the day.
Orders time-stamped before the open of trading are included in the first time period. Even though these orders do
not represent decisions made during the first trading period,
this period is obviously the time during which these orders
flow into the market. Similarly, the first trading period contains all of the orders stamped after the close of trading from
the previous day.
Descriptive Statistics
Table 1 presents the 3,547 buy and sell decisions by portfolio
managers broken down by trading period and classified into
buys and sells measured by number of transactions and the
dollar volume of the transaction. The percentage of each as
a portion of the total is in parentheses. A scan of the table
reveals that most of the decisions, as measured by both number
of decisions and dollar value, occurred during the opening
and closing periods of the day.
The 3,547 trading decisions generated transactions on 742
different stocks traded on the New York Stock Exchange
(NYSE), American Stock Exchange (AMEX), and the overthe-counter markets (OTC). Table 2 shows these decisions
broken down by trading location of the stock. A scan of this
table indicates that the tendency of portfolio managers’ trading
decisions to cluster at the beginning and end of the day occurs
for stocks traded in all locations.
While Tables 1 and 2 indicate that there is a definite clustering of trading decisions during the first and last trading periods
using aggregated data for the year, tests are necessary to show
the statistical significance of the U-shaped pattern. To accomplish statistical tests of the buy and sell decisions, the aggregated data in Tables 1 and 2 are not sufficient. To be useful
for a statistical test, buy decisions for each period on each
day should be compared to sell decisions during that same
period. This would give us an idea of the “relative demand”
during that period.
324
J Busn Res
2000:50:321–326
S. B. Block et al.
Table 2. Portfolio Manager Transaction Decisions by Trading Location of Stock, 1988 (Percent of Total in Parentheses)
Number of Transaction Decisions
Trading
Period
Before 10 a.m.
10 a.m. to 11 a.m.
11 a.m. to 12 noon
12 noon to 1 p.m.
1 p.m. to 2 p.m.
2 p.m. to 3 p.m.
3 p.m. to 4 p.m.
Total
NYSE
678
278
387
357
189
216
352
2457
(27.6)
(11.3)
(15.8)
(14.5)
(7.7)
(8.8)
(14.3)
(100.0)
AMEX
27
3
3
0
2
7
27
69
(39.1)
(4.3)
(4.3)
(0.0)
(2.9)
(10.1)
(39.1)
(100.0)
Volume, Millions of Dollars
OTC
235
94
191
196
51
73
181
1021
NYSE
(23.0)
(9.2)
(18.7)
(19.2)
(5.0)
(7.1)
(17.7)
(100.0)
158.4
57.6
94.5
75.1
53.4
42.9
92.4
574.4
AMEX
(27.6)
(10.0)
(16.4)
(13.1)
(9.3)
(7.5)
(16.1)
(100.0)
6.4
0.3
0.3
0.0
0.1
0.2
3.7
11.0
(58.1)
(2.9)
(2.9)
(0.0)
(1.3)
(1.5)
(33.4)
(100.0)
OTC
29.5
11.4
9.8
14.3
5.1
8.0
26.0
104.1
(28.4)
(10.9)
(9.4)
(13.8)
(4.9)
(7.6)
(25.0)
(100.0)
NYSE 5 New York Stock Exchange; AMEX 5 American Stock Exchange; OTC 5 Over-the-counter markets.
Statistical Tests
To verify that a U-shaped pattern in returns existed during
the year of the sample, we computed the returns on the Dow
Jones Industrial Average (DJIA) for each trading period and
averaged them for the year 1988. Table 3 presents these mean
DJIA returns. The hourly returns are high during the opening
and closing hours and lower during the middle trading hours.
To test for the existence of a pattern for the variables
in this study, we use the chi-square approximation of the
nonparametric Kruskal-Wallis test. This test identifies whether
grouped data might have come from a single population. For
the DJIA returns, the test statistic of 14.39 allows us to reject
the null hypothesis that the returns observed over the different
trading hours originated from the same population at the a 5
0.05 level of significance. This existence of a significant pattern
verifies that market returns during 1988 exhibited the familiar
U shape reported in other studies.
To measure the relative demand behavior of portfolio managers in the sample, we computed a “buy decision ratio” for
each of the seven trading periods for each day. The buy
decision ratio represents the dollar value of the buy decisions
Table 3. Daily Mean Dow Jones Industrial Average Returns and
Buy-Decision Ratios by Intraday Time Period, 1988a
Time Period
(Eastern time)
Before 10 a.m.
10 a.m. to 11 a.m.
11 a.m. to 12 noon
12 noon to 1 p.m.
1 p.m. to 2 p.m.
2 p.m. to 3 p.m.
3 p.m. to 4 p.m.
Kruskal-Wallis test
Dow Jones Industrial
Average Return
Buy-Decision Ratio
(1)
(2)
0.018
20.031
20.029
0.004
0.012
0.024
0.026
14.39*
0.136
0.044
0.045
0.044
0.048
0.028
0.105
140.26**
*Significant at the a 5 0.05 level.
** Significant at the a 5 0.01 level.
a
The buy-decision ratio represents buy decisions during that time period as a proportion
of total buy and sell decisions for that day. The before 10 buy-decision ratio includes
all decisions made after the close of trading on the previous day and before 10 a.m.
Eastern time. The Kruskal-Wallis test statistic is for the null hypothesis that the buydecision ratio in all time periods originated from the same population.
made during each period of the trading day as a portion of
total dollar value of all transactions decisions (buy and sell
decisions) during that day. Its definition is as follows:
buy decision ratio (period i)
total buy decisions in period i
5
total buy and sell decisions for the day
We computed this ratio for each of the seven trading periods
during which the NYSE is open.
Figure 1 shows a definite U-shaped pattern in both the
mean buy-decision ratio and the DJIA returns, and Table 3
presents the numerical values. The buy-decision ratio begins
the trading day with a value of 0.136 in the first time period,
falls to a low of 0.028 during the 2:00 to 3:00 hour, and rises
to close at 0.105. The Kruskal-Wallis test statistic for the buydecision ratios is 140.26, indicating a significant pattern exists
at the a 5 0.01 level.
These results suggest that factors exogenous to the market
trading mechanism may contribute to the pattern of intraday
returns. In particular, the relative frequency of buy and sell
decisions of the institutional investor follows the same
Table 4. Mean Volume of Portfolio Managers Decisions to Trade
by Intraday Time Period, 1988a
Time Period
(Eastern time)
Before 10 a.m.
10 a.m. to 11 a.m.
11 a.m. to 12 noon
12 noon to 1 p.m.
1 p.m. to 2 p.m.
2 p.m. to 3 p.m.
3 p.m. to 4 p.m.
Kruskal-Wallis test
Mean Volume
Share-Volume Ratio Dollar-Volume Ratio
0.278
0.109
0.138
0.157
0.071
0.078
0.169
188.88a
0.268
0.115
0.140
0.152
0.078
0.073
0.175
183.32a
* Significant at the a 5 0.01 level.
a
The mean share-volume ratio is the total share volume during that time period as a
proportion of total share volume for that day. The dollar-volume ratio is the dollar
volume of decisions made during that time period as a proportion of the total dollar
volume of all transactions decisions for that day. The before 10 values includes all
decisions made after the close of trading on the previous day and before 10 a.m. Eastern
time. The Kruskal-Wallis test statistic is for the null hypothesis that the volume ratio
in all time periods originated from the same population.
Intraday Portfolio Management Decisions
J Busn Res
2000:50:321–326
325
Figure 1. Daily mean Dow Jones Industrial Average returns and buy-decision ratios by intraday time period, 1988.
U-shaped pattern as do returns. However, it is also important
that when higher returns occur on the DJIA during a particular
trading period on a given day, they are simultaneously associated with a high level of buy decisions. To test for an association between concurrent returns and buy decisions, we estimated the correlation between the buy-decision ratio and
the return on the DJIA using the buy-decision ratio and its
corresponding DJIA return for each trading period of each
day in the sample. Because the data (particularly the buydecision ratios) are not normally distributed, we computed
the nonparametric Spearman rank correlation coefficient. Its
value was 0.159, which is significantly different from zero at
the a 5 0.05 level, allowing us to conclude that there was a
significant relationship between the level of buy decisions and
DJIA returns during this sample.
In addition to examining returns, we checked for the existence of the U-shaped pattern in the volume of decisions. To
measure volume during each trading period, we computed a
“volume ratio” as the total decisions to transact (buy and sell)
during the trading period as a portion of the total decisions
to transact for the day.
We computed total transactions ratios using both the number of shares (the “share-volume ratio”) and the total dollar
value (the dollar-volume ratio”). Table 4 presents these ratios.
As is evident from the table, both of the volume measures are
highest during the market open period and the market closing
hour, and they are at their lowest during mid-day hours. Note
that on average over one-quarter of all transactions decisions
entered into the market during the first trading period of the
day. The Kruskal-Wallis test indicates a significant pattern in
both ratios at the a 5 0.01 level.
Summary and Conclusion
This article investigates institutional investor buy and sell
decisions as a cause of the intraday patterns of common stock
returns and trading volume. A sample containing buy and
326
J Busn Res
2000:50:321–326
sell decisions of an institutional investor and the time at which
they were made shows that buy decisions tend to be concentrated during the market opening and closing hours, the same
time when stocks tend to show higher-than-average returns.
There was a statistically significant relationship between the
relative number of buy decisions and the intraday DJIA returns. In addition, the volume of portfolio manager decisions
(buy and sell) tends to be concentrated during the opening
and closing hours, coinciding with results from prior studies
that report a higher volume of transactions during opening
and closing hours. These results are consistent with the notion
that the actions of portfolio managers contribute to the intraday pattern of volume.
Although the behavior of other institutional investors may
not follow the same patterns as the institution (NationsBank)
in this case study, it is reasonable that the association between
timing of buy and sell decisions of our institution and market
returns and trading volume is more than coincidence. Portfolio
managers probably follow a systematic pattern in the tendency
of their transaction decisions, particularly their buy decisions,
during the time before the market opens, the opening period,
the closing period, and the time after the market closes.
While Gerety and Mulherin (1992) mention that institutional investors could influence daily trading patterns, they
attribute end-of-day volume to short-term traders. Our evidence in this article shows that institutional investors provide
formidable demand at the end of the day. While short-term
traders may be on the other sides of those transactions, it is
clear that the fundamental demand arises from the institutional
investors. This supports the contention of Brock and Kleidon
(1992) that forces exogenous to market microstructure provide demand or supply imbalances that could lead to observed
daily trading patterns.
The authors thank Mason Gerety, Larry Lockwood, and Ken Martin for their
comments and suggestions. All opinions and any errors that may have crept
into the article are the responsibility of the authors.
References
Admati, Anat R., and Pfleiderer, Paul: A Theory of Intraday Patterns:
Volume and Price Variability. Review of Financial Studies 1 (1989a):
41–66.
Admati, Anat R., and Pfleiderer, Paul: Divide and Conquer: A Theory
S. B. Block et al.
of Intraday and Day-of-the-Week Mean Effects. Review of Financial
Studies 2 (1989b): 189–223.
Berry, Thomas D., and Howe, Keith M.: Public Information Arrival.
Journal of Finance 49 (September 1994): 1331–1346.
Brock, William A., and Kleidon, Allan W.: Periodic Market Closure
and Trading Volume. Journal of Economic Dynamics and Control,
16 (July/October 1992): 451–489.
Foster, F. Douglas, and Viswanathan, S.: A Theory of the Interday
Variations in Volume, Variance, and Trading Costs in Securities
Markets. Review of Financial Studies 4 (1989): 593–624.
Garman, Mark: Market Microstructure. Journal of Financial Economics
3 (1976): 257–275.
Gerety, Mason S., and Mulherin, J. Harold: Trading Halts and Market
Activity: An Analysis of Volume at the Open and the Close. Journal
of Finance 47 (December 1992): 1765–1784.
Harris, Lawrence: A Transaction Data Study of Weekly and Intradaily
Patterns in Stock Returns. Journal of Financial Economics 16 (May
1986): 99–117.
Harris, Lawrence: A Day-end Transaction Price Anomaly, Journal of
Financial and Quantitative Analysis 24 (March 1989): 29–45.
Jain, Prem C., and Joh, Gun-Ho: The Dependence Between Hourly
Prices and Trading Volume. Journal of Financial and Quantitative
Analysis 23 (September 1988): 269–283.
Lockwood, Larry J., and Linn, Scott C.: An Examination of Stock
Market Return Volatility During Overnight and Intraday Periods
(1964–1989). Journal of Finance 45 (June 1990): 591–601.
McInish, Thomas H., and Wood, Robert A.: An Analysis of Transactions Data for the Toronto Stock Exchange: Return Patterns and
End-of-the-day Effect. Journal of Banking and Finance 14 (1990):
441–458.
McInish, Thomas H., and Wood, Robert A.: Hourly Returns, Volume,
Trade Size, and Number of Trades. Journal of Financial Research
14 (Winter 1991): 303–315.
McInish, Thomas H., and Wood, Robert A.: An Analysis of Intraday
Patterns in Bid/ask Spreads for NYSE Stocks. Journal of Finance
47 (June 1992): 753–764.
Miller, E. M.: Explaining Intraday and Overnight Price Behavior.
Journal of Portfolio Management (Summer 1989): 10–16.
Smirlock, Michael, and Starks, Laura: Day-of-the-week and Intraday
Effects in Stock Returns. Journal of Financial Economics 17 (September 1986): 197–210.
Stoll, Hans R., and Whaley, Robert E.: Stock Price Structure and
Volatility. Review of Financial Studies 3 (1990): 37–71.
Wei, Pei-Hwang: Intraday Variations in Trading Activity, Price Variability, and the Bid-Ask Spread. Journal of Financial Research 15
(Fall 1992): 265–276.
Wood, Robert A., McInish, Thomas H., and Ord, Keith: An Investigation of Transaction Data for NYSE Stocks. Journal of Finance 40
(July 1985): 723–741.
Management Decisions: A Case Study of
Intraday Security Return Patterns
Stanley B. Block
TEXAS CHRISTIAN UNIVERSITY
Dan W. French
NEW MEXICO STATE UNIVERSITY
Edwin D. Maberly
UNIVERSITY OF CANTERBURY
This article examines causes of observed stock trading patterns that show
high hourly returns and trading volume during early and late trading
hours. Using time-stamped data from an institutional investor, we document high levels of portfolio managers’ early-morning and late-afternoon
decisions to trade that can result in the volume pattern and relatively
higher proportions of buy decisions that could contribute to the return
pattern. J BUSN RES 2000. 50.321–326. 2000 Elsevier Science Inc.
All rights reserved.
I
nvestors who trade stocks are competing with professional
traders and institutional investment managers, and to be
competitive, investors must be aware of the influences
that other traders have on market returns. Researchers have
identified a number of regularities in common stock returns,
and this article addresses the well-documented pattern of
stock returns during the average trading day that shows a
repeated occurrence of higher-than-expected returns at the open
and close of the market and lower returns during the middle
of the day. The hypotheses of the article is that the timing of
institutional investors’ decisions stimulates this pattern because managers tend to make the majority of their transaction
decisions toward the end of the day. These late-day decisions
affect end-of-day market returns and volume, and the beginning-of-day returns and volume are high because of the accumulation of orders from managers’ decisions made after the
market close the previous day plus decisions made before the
opening of trading each day.
Address correspondence to Dan W. French, Department of Finance, MSC
3FIN, New Mexico State University, Las Cruces, NM 88003-8001; Tel.: (505)
646-3201; fax: (505) 646-2820; E-mail: [email protected]
Journal of Business Research 50, 321–326 (2000)
2000 Elsevier Science Inc. All rights reserved.
655 Avenue of the Americas, New York, NY 10010
Possible Causes of Intraday
Return Patterns
The literature refers to the observed return regularity as the
“U-shaped pattern of intraday returns.” It has been identified
in a number of studies including work by Harris (1986, 1989),
Jain and Joh (1988), Wood, McInish, and Ord (1985), McInish
and Wood (1990), and Smirlock and Starks (1986). A similar
pattern (high early and late in the day, low during the middle
hours) applies to trading volume (see, for example, Foster
and Viswanathan, 1989; Jain and Joh, 1988; McInish and
Wood, 1991; and Wood et al., 1985), return variance (see,
for example, Admati and Pfleiderer, 1989a; Lockwood and
Linn, 1990; and Wei, 1992), and bid-ask spreads (see, for
example, Brock and Kleidon, 1992; McInish, and Wood,
1992; and Stoll and Whaley, 1990). The values for these
variables during market opening and closing times are distinctly larger than the midday values.
Some of the attempts to explain these non-stationarities
have relied on market microstructure models that allow patterns to develop based solely on the internal structure and
operation of the market and in the absence of any external
influences. Admati and Pfleiderer (1989b) develop an information-based model to explain trade clustering that can occur
at any time during the trading day, and they discuss the role
of the open and the close as special clustering points. They
contend that market makers have adverse selection problems
with informed traders that they attempt to alleviate through
their interaction with nondiscretionary liquidity traders, discretionary liquidity traders, and informed traders, so they offer
inducements for liquidity traders to transact at certain times
during the day.
ISSN 0148-2963/00/$–see front matter
PII S0148-2963(99)00023-5
322
J Busn Res
2000:50:321–326
Brock and Kleidon (1992) maintain that market evidence
is at odds with the Admati and Pfleiderer (1989b) informationbased model. The Admati-Pfleiderer model partially relies on
lower transaction costs to explain higher volume and returns
at the opening and close of the day. However, Brock and
Kleidon (1992), extending the inventory-based model of Garman (1976), show that spreads (transactions costs) should be
higher at the open and the close. They present evidence of
observed transaction costs as measured by the bid-ask spread
often being higher at these times.
In addition to explanations based on market microstructure, the U-shaped trading pattern may be attributable to
variables that are exogenous to the market structure as noted
in Brock and Kleidon (1992). Primary exogenous considerations are variations in supply and demand from investors and
the discontinuous nature of market trading. For example,
Brock and Kleidon argue that much of the trading at the open
and close is due to the fact that investors cannot trade as
easily when major markets are closed. The incentive to transact
at the opening and close may be related to the desire of
portfolio managers to bring their portfolios to a particular
target risk, a factor that is unrelated to the market microstructure. Another exogenous factor could be that market activity
is responding to the release of information. Berry and Howe
(1994) find distinct intraday patterns in the release of public
information but conclude that they are only moderately related
to trading volume and not significantly related to price variability. Other exogenous considerations relate to covering
short positions overnight (see Miller, 1989) and the desire of
portfolio managers of index funds to trade at the end of the
day to reduce tracking errors. Mutual fund managers, judged
by net asset value at the end of the day, also have some
incentive to correlate their trading patterns with their benchmarks. Furthermore, brokers are often given orders which
must be filled by the end of the day or canceled.
Whether the causes of intraday return patterns lie with
external or internal (market microstructure) considerations,
the fact remains that a predictive pattern is in place that
challenges market efficiency, although it is doubtful that a
trading strategy could exploit the patterns to the extent of
producing superior after transactions-costs returns. Studies of
the return phenomenon have typically dealt with transactions
data from major exchanges such as the New York Stock Exchange, the American Stock Exchange, and the Toronto Stock
Exchange. Since the data represent realized transactions, the
observed U-shaped pattern of returns and volume includes
both the external influence of investors and portfolio managers
who order the execution of trades for portfolio rebalancing
reasons and the internal influence of market makers and traders who attempt to minimize transaction costs and make trading profits. A shortcoming of the use of aggregated market
transactions data is that the external influence from portfolio
managers cannot be measured separately from the effects of
traders and market makers.
S. B. Block et al.
Gerety and Mulherin (1992) attribute the U-shaped pattern
in trading volume to short-term investors. These short-term
investors, such as market makers and day traders, have relatively little ability to bear risk overnight and desire to exchange
their positions with investors having a greater ability to bear
risk overnight. Gerety and Mulherin (1992) reason that if
investors are transferring risk of holding positions while the
market is closed, then end-of-day volume should be directly
related to overnight return variance. Using closing and opening-hour trading volume on the NYSE from 1933 to 1988,
they find evidence to support their position and conclude that
short-term traders are the cause of the U-shaped pattern in
volume.
Gerety and Mulherin (1992) mention anecdotal discussion
that attributes the U-shaped pattern to institutional investors.
However, they observe a downward trend in last-hour volume
over the last several decades and a concurrent increase in
block trading (to represent institutional trades) and suggest
that institutional traders do not cause the U-shaped pattern.
However, the simple observation of a negative correlation
between these two variables is not proof of a causal relationship. It is quite possible that both institutional investors and
short-term traders are parties to the causes of the U-shaped
pattern, especially if institutional investors represent the other
side of the trades with short-term traders. It is our hypothesis
in this article that a major part of the demand for opening
and closing trades originates with institutional and other longterm investors, and short-term traders are willing to fill that
demand.
While these previous studies have proposed and examined
various causes of intraday trading patterns, none has been
able to explain more than a small portion of the pattern. It
is the intent of this article to further contribute to the evidence
in this area by using a case study to illustrate the importance
of the timing of transaction decisions by institutional portfolio
managers and other long-term investors. These decisions
quickly become orders that impact the market and contribute
to the formation of the daily U-shaped patterns observed in
this article. As these intraday decisions are transmitted to
security traders, they are translated into intraday volume,
variance, and price patterns.
Transactions data typically used in market studies identifies
only the time of the transaction, not the time of the decision
to transact. To overcome this problem, our sample tracks
the times at which portfolio managers made their decisions.
Conversations with the portfolio managers led to the conclusion that the portfolio managers were, for the most part,
indifferent to the problems and issues related to the traders
(endogenous factors). Their concerns were related to rebalancing their portfolios. The extent to which they weight their
actions toward buying activities at market opening and market
closing times in the absence of concerns about execution
costs would tend to indicate the importance of the exogenous
supply-demand variable in the pattern of intraday returns.
Intraday Portfolio Management Decisions
J Busn Res
2000:50:321–326
323
Table 1. Portfolio Manager Transaction Decisions by Trading Period, 1988 (Percent of Total in Parentheses)
Number of Transaction Decisions
Trading period
(Eastern-time)
Before 10 a.m.
10 a.m. to 11 a.m.
11 a.m. to 12 noon
12 noon to 1 p.m.
1 p.m. to 2 p.m.
2 p.m. to 3 p.m.
3 p.m. to 4 p.m.
Total
Buys
500
168
235
212
124
133
302
1674
Volume, Millions of Dollars
Sells
(29.9)
(10.0)
(14.0)
(12.7)
(7.4)
(7.9)
(18.0)
(100.0)
440
207
346
341
118
163
258
1873
Any clustering of either buying or selling decisions at market
open and close would affect volume at those times.
Database and Methodology
The Trust Division of NationsBank Dallas provided a sample
of 3,547 equity transactions completed by the bank’s security
traders during 1988. During that period, the bank had approximately 25 portfolio managers. Their operation is such that
as soon as the portfolio managers reach a decision to place
an order for a security transaction, they communicate to the
trading desk by delivering a written ticket directing the trader
to initiate the purchase or sale of the security. The first task
of the trader is to stamp the time of receipt of the written
order. Then, the trader enters into the market and begins to
“work” the order, searching for the best execution possible
and trying to achieve execution of the order within a reasonably short time. Most orders are completed within 30 minutes
of receipt by the trading desk.
Thus, the time stamp on the order approximates the time
that the portfolio manager reached a transaction decision. It
also approximates the time that the trader began revealing
the order in the market. Thus, the important distinction of
this data set is that observations represent times of decisions
to trade, not necessarily transaction times. If demand and
supply from portfolio managers affect the pattern of intraday
returns, then we would expect to see the U-shaped pattern
occur in a metric that measures relative demand. In addition,
to the extent that the trading desk completes trades relatively
soon after receipt, the volume of portfolio manager decisions
should also exhibit the U-shaped pattern.
Each observation in the sample contains the following information: (1) the time stamp, (2) whether the order was a
buy or a sell, (3) the number of shares, (4) the stock symbol,
the trading location (New York Stock Exchange, American
Stock Exchange, or over the counter), and (5) the execution
price.
Results
We divided the trading day into the same seven trading periods
as did Gerety and Mulherin (1992). The first period is from
Buys
(23.5)
(11.1)
(18.5)
(18.2)
(6.3)
(8.7)
(13.8)
(100.0)
100.9
27.0
35.7
26.9
29.3
24.7
64.6
309.0
Sells
(32.7)
(8.7)
(11.6)
(8.7)
(9.5)
(8.0)
(20.9)
(100.0)
93.4
42.3
69.0
62.5
29.3
26.4
57.6
380.4
(24.6)
(11.1)
(18.1)
(16.4)
(7.7)
(6.9)
(15.1)
(100.0)
the market open (9:30 a.m. Eastern time) to 10:00 a.m. Each
of the other periods begins on the hour and lasts for one
hour, so that there are a total of seven trading periods during
the day.
Orders time-stamped before the open of trading are included in the first time period. Even though these orders do
not represent decisions made during the first trading period,
this period is obviously the time during which these orders
flow into the market. Similarly, the first trading period contains all of the orders stamped after the close of trading from
the previous day.
Descriptive Statistics
Table 1 presents the 3,547 buy and sell decisions by portfolio
managers broken down by trading period and classified into
buys and sells measured by number of transactions and the
dollar volume of the transaction. The percentage of each as
a portion of the total is in parentheses. A scan of the table
reveals that most of the decisions, as measured by both number
of decisions and dollar value, occurred during the opening
and closing periods of the day.
The 3,547 trading decisions generated transactions on 742
different stocks traded on the New York Stock Exchange
(NYSE), American Stock Exchange (AMEX), and the overthe-counter markets (OTC). Table 2 shows these decisions
broken down by trading location of the stock. A scan of this
table indicates that the tendency of portfolio managers’ trading
decisions to cluster at the beginning and end of the day occurs
for stocks traded in all locations.
While Tables 1 and 2 indicate that there is a definite clustering of trading decisions during the first and last trading periods
using aggregated data for the year, tests are necessary to show
the statistical significance of the U-shaped pattern. To accomplish statistical tests of the buy and sell decisions, the aggregated data in Tables 1 and 2 are not sufficient. To be useful
for a statistical test, buy decisions for each period on each
day should be compared to sell decisions during that same
period. This would give us an idea of the “relative demand”
during that period.
324
J Busn Res
2000:50:321–326
S. B. Block et al.
Table 2. Portfolio Manager Transaction Decisions by Trading Location of Stock, 1988 (Percent of Total in Parentheses)
Number of Transaction Decisions
Trading
Period
Before 10 a.m.
10 a.m. to 11 a.m.
11 a.m. to 12 noon
12 noon to 1 p.m.
1 p.m. to 2 p.m.
2 p.m. to 3 p.m.
3 p.m. to 4 p.m.
Total
NYSE
678
278
387
357
189
216
352
2457
(27.6)
(11.3)
(15.8)
(14.5)
(7.7)
(8.8)
(14.3)
(100.0)
AMEX
27
3
3
0
2
7
27
69
(39.1)
(4.3)
(4.3)
(0.0)
(2.9)
(10.1)
(39.1)
(100.0)
Volume, Millions of Dollars
OTC
235
94
191
196
51
73
181
1021
NYSE
(23.0)
(9.2)
(18.7)
(19.2)
(5.0)
(7.1)
(17.7)
(100.0)
158.4
57.6
94.5
75.1
53.4
42.9
92.4
574.4
AMEX
(27.6)
(10.0)
(16.4)
(13.1)
(9.3)
(7.5)
(16.1)
(100.0)
6.4
0.3
0.3
0.0
0.1
0.2
3.7
11.0
(58.1)
(2.9)
(2.9)
(0.0)
(1.3)
(1.5)
(33.4)
(100.0)
OTC
29.5
11.4
9.8
14.3
5.1
8.0
26.0
104.1
(28.4)
(10.9)
(9.4)
(13.8)
(4.9)
(7.6)
(25.0)
(100.0)
NYSE 5 New York Stock Exchange; AMEX 5 American Stock Exchange; OTC 5 Over-the-counter markets.
Statistical Tests
To verify that a U-shaped pattern in returns existed during
the year of the sample, we computed the returns on the Dow
Jones Industrial Average (DJIA) for each trading period and
averaged them for the year 1988. Table 3 presents these mean
DJIA returns. The hourly returns are high during the opening
and closing hours and lower during the middle trading hours.
To test for the existence of a pattern for the variables
in this study, we use the chi-square approximation of the
nonparametric Kruskal-Wallis test. This test identifies whether
grouped data might have come from a single population. For
the DJIA returns, the test statistic of 14.39 allows us to reject
the null hypothesis that the returns observed over the different
trading hours originated from the same population at the a 5
0.05 level of significance. This existence of a significant pattern
verifies that market returns during 1988 exhibited the familiar
U shape reported in other studies.
To measure the relative demand behavior of portfolio managers in the sample, we computed a “buy decision ratio” for
each of the seven trading periods for each day. The buy
decision ratio represents the dollar value of the buy decisions
Table 3. Daily Mean Dow Jones Industrial Average Returns and
Buy-Decision Ratios by Intraday Time Period, 1988a
Time Period
(Eastern time)
Before 10 a.m.
10 a.m. to 11 a.m.
11 a.m. to 12 noon
12 noon to 1 p.m.
1 p.m. to 2 p.m.
2 p.m. to 3 p.m.
3 p.m. to 4 p.m.
Kruskal-Wallis test
Dow Jones Industrial
Average Return
Buy-Decision Ratio
(1)
(2)
0.018
20.031
20.029
0.004
0.012
0.024
0.026
14.39*
0.136
0.044
0.045
0.044
0.048
0.028
0.105
140.26**
*Significant at the a 5 0.05 level.
** Significant at the a 5 0.01 level.
a
The buy-decision ratio represents buy decisions during that time period as a proportion
of total buy and sell decisions for that day. The before 10 buy-decision ratio includes
all decisions made after the close of trading on the previous day and before 10 a.m.
Eastern time. The Kruskal-Wallis test statistic is for the null hypothesis that the buydecision ratio in all time periods originated from the same population.
made during each period of the trading day as a portion of
total dollar value of all transactions decisions (buy and sell
decisions) during that day. Its definition is as follows:
buy decision ratio (period i)
total buy decisions in period i
5
total buy and sell decisions for the day
We computed this ratio for each of the seven trading periods
during which the NYSE is open.
Figure 1 shows a definite U-shaped pattern in both the
mean buy-decision ratio and the DJIA returns, and Table 3
presents the numerical values. The buy-decision ratio begins
the trading day with a value of 0.136 in the first time period,
falls to a low of 0.028 during the 2:00 to 3:00 hour, and rises
to close at 0.105. The Kruskal-Wallis test statistic for the buydecision ratios is 140.26, indicating a significant pattern exists
at the a 5 0.01 level.
These results suggest that factors exogenous to the market
trading mechanism may contribute to the pattern of intraday
returns. In particular, the relative frequency of buy and sell
decisions of the institutional investor follows the same
Table 4. Mean Volume of Portfolio Managers Decisions to Trade
by Intraday Time Period, 1988a
Time Period
(Eastern time)
Before 10 a.m.
10 a.m. to 11 a.m.
11 a.m. to 12 noon
12 noon to 1 p.m.
1 p.m. to 2 p.m.
2 p.m. to 3 p.m.
3 p.m. to 4 p.m.
Kruskal-Wallis test
Mean Volume
Share-Volume Ratio Dollar-Volume Ratio
0.278
0.109
0.138
0.157
0.071
0.078
0.169
188.88a
0.268
0.115
0.140
0.152
0.078
0.073
0.175
183.32a
* Significant at the a 5 0.01 level.
a
The mean share-volume ratio is the total share volume during that time period as a
proportion of total share volume for that day. The dollar-volume ratio is the dollar
volume of decisions made during that time period as a proportion of the total dollar
volume of all transactions decisions for that day. The before 10 values includes all
decisions made after the close of trading on the previous day and before 10 a.m. Eastern
time. The Kruskal-Wallis test statistic is for the null hypothesis that the volume ratio
in all time periods originated from the same population.
Intraday Portfolio Management Decisions
J Busn Res
2000:50:321–326
325
Figure 1. Daily mean Dow Jones Industrial Average returns and buy-decision ratios by intraday time period, 1988.
U-shaped pattern as do returns. However, it is also important
that when higher returns occur on the DJIA during a particular
trading period on a given day, they are simultaneously associated with a high level of buy decisions. To test for an association between concurrent returns and buy decisions, we estimated the correlation between the buy-decision ratio and
the return on the DJIA using the buy-decision ratio and its
corresponding DJIA return for each trading period of each
day in the sample. Because the data (particularly the buydecision ratios) are not normally distributed, we computed
the nonparametric Spearman rank correlation coefficient. Its
value was 0.159, which is significantly different from zero at
the a 5 0.05 level, allowing us to conclude that there was a
significant relationship between the level of buy decisions and
DJIA returns during this sample.
In addition to examining returns, we checked for the existence of the U-shaped pattern in the volume of decisions. To
measure volume during each trading period, we computed a
“volume ratio” as the total decisions to transact (buy and sell)
during the trading period as a portion of the total decisions
to transact for the day.
We computed total transactions ratios using both the number of shares (the “share-volume ratio”) and the total dollar
value (the dollar-volume ratio”). Table 4 presents these ratios.
As is evident from the table, both of the volume measures are
highest during the market open period and the market closing
hour, and they are at their lowest during mid-day hours. Note
that on average over one-quarter of all transactions decisions
entered into the market during the first trading period of the
day. The Kruskal-Wallis test indicates a significant pattern in
both ratios at the a 5 0.01 level.
Summary and Conclusion
This article investigates institutional investor buy and sell
decisions as a cause of the intraday patterns of common stock
returns and trading volume. A sample containing buy and
326
J Busn Res
2000:50:321–326
sell decisions of an institutional investor and the time at which
they were made shows that buy decisions tend to be concentrated during the market opening and closing hours, the same
time when stocks tend to show higher-than-average returns.
There was a statistically significant relationship between the
relative number of buy decisions and the intraday DJIA returns. In addition, the volume of portfolio manager decisions
(buy and sell) tends to be concentrated during the opening
and closing hours, coinciding with results from prior studies
that report a higher volume of transactions during opening
and closing hours. These results are consistent with the notion
that the actions of portfolio managers contribute to the intraday pattern of volume.
Although the behavior of other institutional investors may
not follow the same patterns as the institution (NationsBank)
in this case study, it is reasonable that the association between
timing of buy and sell decisions of our institution and market
returns and trading volume is more than coincidence. Portfolio
managers probably follow a systematic pattern in the tendency
of their transaction decisions, particularly their buy decisions,
during the time before the market opens, the opening period,
the closing period, and the time after the market closes.
While Gerety and Mulherin (1992) mention that institutional investors could influence daily trading patterns, they
attribute end-of-day volume to short-term traders. Our evidence in this article shows that institutional investors provide
formidable demand at the end of the day. While short-term
traders may be on the other sides of those transactions, it is
clear that the fundamental demand arises from the institutional
investors. This supports the contention of Brock and Kleidon
(1992) that forces exogenous to market microstructure provide demand or supply imbalances that could lead to observed
daily trading patterns.
The authors thank Mason Gerety, Larry Lockwood, and Ken Martin for their
comments and suggestions. All opinions and any errors that may have crept
into the article are the responsibility of the authors.
References
Admati, Anat R., and Pfleiderer, Paul: A Theory of Intraday Patterns:
Volume and Price Variability. Review of Financial Studies 1 (1989a):
41–66.
Admati, Anat R., and Pfleiderer, Paul: Divide and Conquer: A Theory
S. B. Block et al.
of Intraday and Day-of-the-Week Mean Effects. Review of Financial
Studies 2 (1989b): 189–223.
Berry, Thomas D., and Howe, Keith M.: Public Information Arrival.
Journal of Finance 49 (September 1994): 1331–1346.
Brock, William A., and Kleidon, Allan W.: Periodic Market Closure
and Trading Volume. Journal of Economic Dynamics and Control,
16 (July/October 1992): 451–489.
Foster, F. Douglas, and Viswanathan, S.: A Theory of the Interday
Variations in Volume, Variance, and Trading Costs in Securities
Markets. Review of Financial Studies 4 (1989): 593–624.
Garman, Mark: Market Microstructure. Journal of Financial Economics
3 (1976): 257–275.
Gerety, Mason S., and Mulherin, J. Harold: Trading Halts and Market
Activity: An Analysis of Volume at the Open and the Close. Journal
of Finance 47 (December 1992): 1765–1784.
Harris, Lawrence: A Transaction Data Study of Weekly and Intradaily
Patterns in Stock Returns. Journal of Financial Economics 16 (May
1986): 99–117.
Harris, Lawrence: A Day-end Transaction Price Anomaly, Journal of
Financial and Quantitative Analysis 24 (March 1989): 29–45.
Jain, Prem C., and Joh, Gun-Ho: The Dependence Between Hourly
Prices and Trading Volume. Journal of Financial and Quantitative
Analysis 23 (September 1988): 269–283.
Lockwood, Larry J., and Linn, Scott C.: An Examination of Stock
Market Return Volatility During Overnight and Intraday Periods
(1964–1989). Journal of Finance 45 (June 1990): 591–601.
McInish, Thomas H., and Wood, Robert A.: An Analysis of Transactions Data for the Toronto Stock Exchange: Return Patterns and
End-of-the-day Effect. Journal of Banking and Finance 14 (1990):
441–458.
McInish, Thomas H., and Wood, Robert A.: Hourly Returns, Volume,
Trade Size, and Number of Trades. Journal of Financial Research
14 (Winter 1991): 303–315.
McInish, Thomas H., and Wood, Robert A.: An Analysis of Intraday
Patterns in Bid/ask Spreads for NYSE Stocks. Journal of Finance
47 (June 1992): 753–764.
Miller, E. M.: Explaining Intraday and Overnight Price Behavior.
Journal of Portfolio Management (Summer 1989): 10–16.
Smirlock, Michael, and Starks, Laura: Day-of-the-week and Intraday
Effects in Stock Returns. Journal of Financial Economics 17 (September 1986): 197–210.
Stoll, Hans R., and Whaley, Robert E.: Stock Price Structure and
Volatility. Review of Financial Studies 3 (1990): 37–71.
Wei, Pei-Hwang: Intraday Variations in Trading Activity, Price Variability, and the Bid-Ask Spread. Journal of Financial Research 15
(Fall 1992): 265–276.
Wood, Robert A., McInish, Thomas H., and Ord, Keith: An Investigation of Transaction Data for NYSE Stocks. Journal of Finance 40
(July 1985): 723–741.