Full text of Research Papers (Federal Reserve Bank of New York) : Price Formation and Liquidity in the U.S. Treasuries Market : Evidence from Intraday Patterns around Announcements, Research Paper 9633

View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

PRICE FORMATION AND LIQUIDITY IN THE U.S. TREASURIES MARKET:
EVIDENCE FROM INTRADAY PATTERNS AROUND ANNOUNCEMENTS
Michael J. Fleming and Eli M. Remolona

Federal Reserve Bank of New York
Research Paper No. 9633
October 1996

This paper is being circulated for purposes of discussion and comment only.
The contents should be regarded as preliminary and not for citation or quotation without
permission of the author. The views expressed are those of the author and do not necessarily
reflect those of the Federal Reserve Bank of New York or the Federal Reserve System.
Single copies are available on request to:

Public Information Department
Federal Reserve Bank of New York
New York, NY 10045

Draft, October 1996

Price Formation and Liquidity
in the U.S. Treasuries Market:
Evidence from Intraday Patterns
Around Announcements
MICHAEL

J. FLEMING AND ELIM. REMOLONA

Federal Reserve Bank of New York
·Capital Markets Function
New York, NY 10045
Tel.: (212) 720-6372 (Fleming)
Tel.: (212) 720-6328 (Remolona)
E-mail: michael.Jleming@frbny.sprint.com
E-mail: eli.remolona@Jrbny.sprint.com
Abstract

We find striking intraday adjustment patterns for price volatility,
trading volume, and bid-ask spreads in the U.S. Treasuries market
around the time of macroeconomic announcements. The patterns
suggest certain hypotheses about price formation and liquidity
provision in multiple-dealer markets. These hypotheses assign new
importance to public information, heterogeneous views, sluggish
price discovery, traditional inventory-control behavior by market
makers, and liquidity traders who react with a lag to price changes.
Keywords: Price formation, liquidity, intraday patterns, announcements, Treasury

securities market
JEL classification code: G14
• We thank Peter Antunovich, Richard Cantor, Young Ho Eom, Kose John, Yasuaki Hashiro,
Frank Keane, Patrick Lynch, Richard Lyons, Tony Rodrigues, Asani Sarkar, and Michael Strauss
for helpful discussions. We thank Debbie Gruenstein and Christina Hubbard for research
assistance. We are grateful to Gov PX, Inc. for the data. Views expressed are the authors' and do
not necessarily reflect the views of the Federal Reserve Bank of New _York or the Federal
Reserve System.

Draft, October 1996

Price Formation and Liquidity
in the U.S. Treasuries Market:
Evidence from Intraday Patterns
Around Announcements
MICHAEL J. FLEMING AND ELIM. REMOLONA

Federal Reserve Bank of New York

Introduction
We identify striking intraday patterns in the behavior of price volatility,
trading volume, and bid-ask spreads in the U.S. Treasury securities market
around the time of macroeconomic announcements. Using newly availa],le
interdealer broker data, we find that the bid-ask spread widens dramatically to
anticipate an announcement, commencing a brief period of illiquidity. There is
then a notable lack of trading volume at the time of the most volatile prices,
which occur in the first two minutes after a major announcement. Trading
volume surges only after an appreciable Jag. Liquidity returns in the form of a
narrowing spread as soon as trading volume surges. High levels of price
volatility and trading volume then persist for over an hour, with volume
persisting somewhat longer. Such patterns suggest hypotheses about the price
formation process and the provision of liquidity in a market where dealers
compete as both informed investors and market makers.
Three strands of the empirical literature on how markets process
information have together Jed to the conclusion that private information is the
dominant source of price volatility in stock markets. First, the "mixture of

distributions hypothesis" of Clark (1973), Epps and Epps (1976), and Tauchen
and Pitts (1983) suggests that prices and volume jointly depend on the arrival of
information, public or private. Second, French and Roll (1986) show that stock
return volatilities are higher when the exchanges are open than when they are
closed. By looking at Wednesdays when the exchanges happen ed to be closed,
they attribute the pattern not to the release of public information during normal
business hours but to the effect of private information conveyed through trading.
Third, Berry and Howe (1994) develop an intraday measure of public
information flow based on Reuter's news releases and find no significant
relationship between their measure and stock price volatility.
The provision of liquidity in financial markets is the responsibility of
market makers. In the market microstructure literature, recent models of market
making based on asymmetric information have challenged the more traditional
models based on inventory control. The inventory-control models (e.g.,
Demsetz, 1968; Tinic and West, 1972; and Ho and Stoll, 1983) emphasize the risks
to market makers of high price volatility and low trading volume, resulting in a
bid-ask spread that widens with volatility and narrows with volume. The
asymmetric-information models (e.g., Glosten and Milgrom, 1985; and Kyle,
1985), however, suggest that if some investors had superior information, the risk
to market makers and the spreads they quote would depend on when the
informed and uninformed trade. Admati and Pfleiderer (1988), for example,
propose that discretionary liquidity traders would cluster their trading activity
around certain times of the day and that the resulting liquidity would attract the
informed investors who would in turn bring price volatility. Hence high
volume, high volatility, and narrow spreads would coincide. The evidence from
the stock market tends to favor the asymmetric-information view (e.g.,
Madhavan and Smidt, 1991; and Hasbrouck, 1991), while evidence from the
foreign exchange market supports both views (Lyons, 1995).

In this paper, we study the price formation process and the provision of
liquidity in the interdealer cash market for U.S. Treasury securities. The market
is dominated by primary dealers who compete both as informed investors and
market makers. We focus on the intraday price and trading patterns around the
time of major macroeconomic announcements, examining the patterns at both
one-minute and five-minute intervals. These scheduled announcements are
useful because they identify precise times at which information is released.
Ederington and Lee (1993) and others rely on such announcements to explain
intraday price volatility in various futures and currency markets.' Until now,
however, the lack of data on trading volumes and bid-ask spreads has inhibited
the discussion about the way markets process information and provide liquidity.

We propose several hypotheses to explain the stylized facts. First, the
volatility spike immediately after an announcement implies a dominant role for
public information, for which price adjustment requires no trading activity.
Second, the ensuing surge in volume arises from wide disagreement among
investors regarding the price adjustment. Third, the persistence of high volatility
along with high volume represents a price-volume relationship that arises from
sluggishness in the formation of a consensus price. Fourth, the brief period of
illiquidity around announcements demonstrates the importance of inventorycontrol factors in market making. Finally, the persistence of high volume beyond
the period of high volatility indicates the presence of liquidity traders who react
with a lag to price changes.

The paper is organized as follows: In Section I, we characterize the
market described by our data. In Section II, we describe the data used in the .
' Crain and Lee ( 1995), Andersen and Bollerslev ( 1996), and Locke ( 1996) examine
announcement effects in precious metals markets and other futures and currency markets.

analysis. In Section III, we examine the basic intraday patterns and distinguish
the patterns that can be attributed to the macroeconomic announcements. In
Section IV, we identify the announcements that have the most impact on the
market. In Section V, we document five stylized facts about the timing and
persistence of the announcements' effects on volatility, volume, and the spread.
In Section VI, we discuss hypotheses about the price formation process and
liquidity provision to explain the stylized facts.

The lnterdealer Treasury Securities Market'
Trading in U.S. Treasury securities takes place primarily in a multiple-

dealer over-the-counter market rather than an organized exchange.' While there
are 1,700 brokers and dealers trading in the secondary market, the majority of
trading volume is accounted for by the 37 primary dealers.' Primary dealers are
those eligible to trade directly with the Federal Reserve Bank of New York.'
These dealers are expected to bid competitively at the Treasury auctions,
participate in the Fed's open market operations, and provide market intelligence
to the Fed. Until recently, primary dealers were also required to transact
significant volume with customers and thereby maintain a liquid secondary

Additional sources on the U.S. Treasuries market are Bollenbacher (1988), Department of the
Treasury (1992), Fleming (1996), Madigan and Stehm (1994), Stigum (1990), and U.S. General
Accounting Office (1986).
3

While Treasuries are traded on the New York Stock Exchange and the American Stock
Exchange, exchange trading is negligible compared to that of the over-the-counter market.

Department of Treasury ( 1992).

The ranks of primary dealers consist of diversified securities firms, money center banks, and
specialized securities firms, and include foreign-owned as well as domestically owned firms. A
list of the primary dealers is provided in Appendix A.

market for Treasury securities.' While trading with customers is no longer
required, primary dealers remain the predominant market makers in U.S.
Treasury securities. It is evident that the dealers compete as both market makers
and informed investors.
As market makers, the primary dealers take positions and starid ready to
buy and sell securities for their own account at their quoted bid and ask prices.
The positions taken, in fact, tend to be highly leveraged, as they are typically
financed with borrowings in the overnight repo market. The dealers trade 22 to
23 hours per day during the five day trading week, although 95% of trading
activity occurs during New York trading hours.7 Primary dealers trade an
average of $125 billion a day in the U.S. Treasuries cash market.' Just over half of
this volume is with customers and just under half is transacted with other
primary dealers. Roughly 90% of the interdealer volume occurs through
interdealer brokers.
The interdealer brokers are at the core of the secondary market for U.S.
Treasury securities.' The brokers provide primary dealers with electronic
screens that post the best bids and offers phoned in by the dealers. Dealers
execute trades by phone through the brokers, who then also post the resulting
trade price and size electronically. For the most part, the brokers act only as

Customers refer to entities that are not primary dealers or brokers, and in practice include nonprimary dealers, other financial institutions (e.g., banks, insurance companies, pension funds, and
mutual funds), non-financial institutions, and individuals.

Madigan and Stehm (1994) and Fleming (1996) describe the round-the-clock market. The
volume statistics in this paragraph are from Fleming (1996).

· In contrast, trading volume on the New York Stock Exchange averages $9.7 billion a day
(NYSE Fact Bookfor the Year 1994).

The six major interdealer brokers are: Cantor Fitzgerald Inc., Garban Ltd., Hilliard Farber &
Co. Inc., Liberty Brokerage Inc., RMJ Securities Corp., and Tullett & Tokyo Securities Inc.

agent and serve only primary dealers." The interdealer broker market is
extremely liquid with minimum trade sizes of $1 million ($5 million for bills), bidask spreads of less than two hundredths of a point for the most active 5-year
note, and modest brokerage fees."

An important feature of the interdealer broker market is the anonymity of

trading. The brokers are "blind" in the sense that they do not reveal to the
counterparties to a transaction the other's name." Since the market is limited to
dealers who are perceived to have been vetted by the Fed, there is little concern
that a counterparty might renege on a transaction. 13 The anonymity of trading
indicates the importance of individual primary dealers' private information or
views. Different dealers are known to have different analytical strengths,
investment strategies, and customers, and the same trade by one dealer, if
identified, would not convey the same information if undertaken by another
dealer.

Cantor Fitzgerald Inc. is the most notable exception on both points.

Typical brokerage fees (paid by the transaction initiator) are as follows: $12.50 per $1 million
on 3-month bills (¼ of a hundredth of a point), $25 per$ I million on 6-month and I-year bills (½
and¼ of a hundredth of a point respectively), and $39.06 per $1 million on notes and bonds (1/8
of a 32nd of a point). These are the fees reported by Stigum ( 1990), and recent communication
with market participants suggests that fees today are similar. The fees are negotiable, however,
and can vary with volume.·
12

Stigum ( 1990) explains that clearing trades through a clearing bank allows brokers not to
reveal names on trades done through them (p. 436).
13

Officially, the Fed states, "The designation [of primary dealer] is not an endorsement, is not
conferred under regulatory authority, and does not entail official supervision by the Federal
Reserve." (Federal Reserve Bank of New York, I 988).

II.

Data
Our data cover one year of tick-by-tick trading activity among the

primary dealers in the interdealer broker market. The source of the data is
GovPX, Inc., a joint venture set up by the primary dealers and interdealer
brokers in 1991 to improve the public's access to U.S. Treasury security prices.
GovPX consolidates and posts real-time quote and transactions data from 5 of
the 6 major interdealer brokers, accounting for two-thirds of the interdealer
broker market." Posted data include the best bids and offers, trade prices and
sizes, and the aggregate volume of trading for all Treasury bills, notes, and
bonds. GovPX data are distributed electronically to the public through several
on-line vendors.

The sample period is August 23, 1993 to August i9, 1994, a year in which
the Federal Reserve raised its target fed funds rate five times. After excluding 10
days when the market was closed, we have a sample of 250 trading days. We
focus our analysis on the on-the-run 5-year Treasury note. On-the-run securities
(also called active or current) are the most recently issued securities of a given
maturity and account for the majority of interdealer trading volume." Among
the on-the-run issues, the 5-year note is the most actively traded security among
the brokers reporting to GovPX (Fleming, 1996). During our sample period,
GovPX posted a daily average of 2,167 bid-ask quotations and 659 trades for the
on-the-run 5-year note. Appendix B describes the cleaning and processing of the
data in some detail.
14

Cantor Fitzgerald Inc. is not included. Cantor specializes in longer-tenn securities and in
particular the 30-year bond.
15

Fleming (1996) finds that 64% of interdealer trading is in on-the-run issues, 25% is in off-therun issues, and 12% is in when-issued securities. Off.-the-run securities are issued securities that
are no longer active, while when-issued securities are securities that have been announced for
auction but not yet issued.

We also collected data on the date and time of 19 different regularly
scheduled macroeconomic announcements. These include 18 monthly
announce ments that regularly appear in "The Week Ahead" section of Business

Week as well as one weekly announcement. 16 The federal governme nt's
"Schedule of Release Dates" and the Wall Street Journal were used in addition to
Business Week to determine announce ment dates and times. As seen in Table I, 11
of our announce ments are released at 8:30 AM eastern time (ET), one at 9:15 AM,
six at 10:00 AM and one at 2:00 PM." Seventeen of the announce ments come
from governme nt agencies and two from the private sector. Our period of
analysis encompasses 268 of these announcements on 173 separate days.

III.

Intraday Patterns
In this section, we describe the basic intraday patterns of price volatility,

trading volume, and bid-ask spreads and distinguish the patterns that can be
attributed .to the scheduled macroeconomic announcements. We examine these
patterns specifically for the on-the-run 5-year Treasury note for successive fiveminute intervals from 7:30 AM to 5:00 PM (ET). As mentioned before, these
trading hours account for nearly 95 percent of interdealer trading, and all of the
announce ments we examine occur during these hours. To isolate the role of the

One of these "monthly" announcements is gross domestic product (GDP). While GDP is a
quarterly statistic, advance, preliminary, and final estimates are released in .successive months.
Our list of announcements is the same as Ederington and Lee's (1993) except for the addition of
· initial jobless claims and consumer confidence, and the omission of installment credit. We
exclude installment credit since it does not occur at a fixed time.
17

Included in the 8:30 AM. count is personal income, which was released at 10:00 A.M. for the
first four announcements.

scheduled announcements, we compare the intraday patterns on announcement
days with the patterns on nonannouncement days. An announcement day is one
in which one of our 19 scheduled announcements was made, and a
nonannouncement day one in which none of the announcements was made.

To examine intraday price volatility we take the change in log prices for
each five-minute interval. Here and throughout the paper, prices are defined as
the midpoints between bid and ask quotes, and the price changes are calculated
with the last set of quotes posted during an interval." We then calculate the
standard deviation of these log price changes for each interval across our 250
trading days. We measure trading volume as the total face value of securities
traded during each interval and then take the mean across days. We measure the
bid-ask spread as the mean difference between bid and ask quotes divided by the
mean midpoint between bid and ask quotes for each interval and then take the
mean across trading days. In the figures and in our discussion we refer to the
different time intervals by the interval starting times, e.g., 8:30 AM for the 8:308:35 AM interval.

Price volatility
The most distinctive feature of the average intraday pattern of price
volatility is a spike in the 8:30 AM interval. As shown in Figure lA, volatility
rises sharply at that time and remains relatively flat the rest of the day, except for
minor spikes at 10:00 AM and 1:45 PM. Ederington and Lee (1993) find similar
spikes in the T-bond, Eurodollar, and deutsche mark futures markets, and they
attribute the pattern to the effects of a number of scheduled macroeconomic
announcements released at those times of the day." Indeed, as shown in Figure
18

We also have transactions prices but using the bid-ask midpoints allows us to avoid
complications associated with the "bid-ask bounce" and provides us with more observations.
19

Crain and Lee ( 1995), Andersen and Bollerslev ( 1996), and Locke ( 1996) find the same
effects in precious metals and other interest rate and currency mark~ts.

lB, volatility on announcement days exhibits a pronounced spike at 8:30 AM and
a less pronounced one at 10:00 AM. In our sample, 11 different scheduled
announcements are released at 8:30 AM and six at 10:00 AM. By comparison,
volatility stays relatively flat through nonannouncement trading days.

Trading volume
Our data provide new evidence on the intraday patterns of trading
volume in the U.S. Treasuries market. In this market, average trading volume
also surges around 8:30 AM but its subsequent decline is more gradual than the
decline in volatility. As shown in Figure 2A, volume is rising from the start of
New York trading at 7:30 AM. Trading volume then peaks occurs at 8:35 AM or
five minutes after the volatility peak. Except for a minor surge after 10:00 AM,
volume declines only gradually until 3:00 PM, when it then falls sharply. The
great volume surge around 8:30 AM and the minor surge around 10:00 AM can
be attributed to the announcements. As shown in Figure 2B, the surge at 8:35
AM is much more pronounced for announcement days, and the smaller surge
around 10:00 AM is clearly evident only for announcement days. For much of
the rest of day, volume on announcement days remains higher than volume on
nonannouncement days. Announcement and nonannouncement day patterns are
similar, however, in that both show a rise in volume before 8:30 AM and a sharp
drop after 3:00 PM. 20

The bid-ask spread
Our data also provide new evidence on the intraday patterns of liquidity
in the U.S. Treasuries market as measured by the bid-ask spread. In this market,
the average bid-ask ·spread displays a rough reverse J-shaped pattern interrupted

The early morning volume surge and the sharp drop in the afternoon may be related to the
opening of the futures markets in Chicago at 8:20 AM (ET) and the closing of those markets at
3:00 PM (ET).

by a sharp rise around 8:30 AM and a smaller rise around 10:00 AM. 21 As shown
in Figure 3A, the spread is widest at the start of New York trading indicating an
initial reluctance by dealers to provide liquidity. The spread then narrows
rapidly from 7:30 AM, then widens sharply for a brief period around 8:30 AM. It
then narrows again, and widens briefly around 10:00 AM. The spread then rises
gradually until the early afternoon, tapers off slightly and then rises sharply just
before trading ends. The momentary loss of liquidity around 8:30 AM and 10:00
AM, as indicated by the wide spreads, can be attributed to the announcements.
As shown in Figure 3B, the reverse J-shaped pattern is evident on both
announcement and nonannouncement days, whHe the interruptions of wide
spreads are evident only for announcement days.

IV.

The Most Important Announcements
Earlier announcement studies
Several studies, such as Roley and Troll (1983), Urich and Wachtel (1984),

Hardouvelis (1988), and Dwyer and Hafer (1989), use daily return data to
examine the impact of monthly economic announcements. 22 These studies
typically estimate the effects of the surprise components of the announcements
on the level of interest rates. While the producer price index (PPI) is generally
found to have an impact in the early studies, most monthly announcements are

Mclnish and Wood (1992) observe such a reverse J-shaped pattern for stocks traded on the
New York Stock Exchange.
22

There is also an extensive literature examining the impact of the weekly money supply
announcements on interest rates. See Hardouvelis (1984) for references to the earlier literature.
More recent papers include Thornton ( 1989) and Strongin and Tarhan ( 1990).

found to have no significant effects." More recently, however, Cook and Korn
(1991) demonstra te the importance of the employme nt report on daily returns
and document its increasing significance in recent years.
The availability of intraday price data has greatly increased the power of
tests that examine announce ment effects. Ederington and Lee (1993) examine the
impact of monthly economic announcements on the Treasury bond futures,
Eurodollar futures, and deutsche mark futures. They find employme nt, PPI, the
consumer price index (CPI), and durable goods orders in that order as having the
largest price volatility effects for interest rate futures, with generally similar
findings for deutsche mark futures markets. Extensions of this work have
considered spot market responses (Crain and Lee, 1995), market expectations
regarding the announce ments (Becker, Finnerty, and Kopecky, 1996), and other
determina nts of price volatility (Andersen and Bollerslev, 1996). Locke (1996)
finds a surge in trading activity as well as price volatility around announce ments
in interest rate futures, foreign exchange, and precious metals markets.

Measuring the impact of announcements
We now examine the effects of the various announce ments on price
volatility, trading volume, and the bid-ask spread in the U.S. Treasuries market.
Following Ederington and Lee (1993), we run cross-sectional regressions of
volatility, volume, and spread on announcement dummy variables for given
five-minute intervals. Specifically, we define dummy variables D1m where D1m =

The PPI is found to have an impact on interest rates in Urich and Wachtel (1984), Hardouvelis
( 1988), and Dwyer and Hafer ( 1989), but not Roley and Troll ( 1983). Roley and Troll ( 1983) do
find evidence in favor of the industrial production index impacting rates, but Hardouvelis ( 1988)
and Dwyer and Hafer ( 1989) do not. Hardouvelis ( 1988) finds significant effects for the
consumer price index, the trade balance, and the unemployment rate, but none of these other
studies that look at these variables find them significant. Hardouvelis (1988) also uncovers
evidence that durable goods orders, personal income, and retail sales impact interest rates.

1 if announcement k is made on day n and Dkn = 0 otherwise." The regression
equation is then

Y!, = a~, +

r:.,

alD.,, + e~ where the superscript j indicates

whether the dependent variable is volatility, volume, or spread, the subscript t
indicates the time interval, and K is the number of different announcements
included in the regression. The intercept a~, measures the average value of the
variable in time interval t in the absence of announcements and the coefficient

af, measures the average impact of announcement k.

We measure price volatility

here as the absolute value of the change in log prices. As before, trading volume
is defineli
aj,
thectotal face value of securities traded during an interval and the
.. . ..
- ..
bid-ask spread is defined as the mean bid-ask quote difference divided by the
mean midpoint of the bid and ask quotes for an interval.

For each dependent variable we conduct the regression analysis for four
different time intervals corresponding to the four times of day that the 19 reports
are released. In the case of price volatility and the bid-ask spread these
regressions are run for the five-minute intervals immediately following the
announcements (e.g., 8:30-8:35 AM for the 8:30 AM announcements). In the case
of trading volume these regressions are run for the succeeding five-minute
interval (e.g., 8:35-8:40 AM for the 8:30 AM announcements). The patterns
displayed in Charts lB, 2B, and 3B suggest that these are the intervals in which
the announcements have their most pronounced effects on the corresponding
variables.

The number of announcements K included in each set of regressions
varies. Specifically, we include dummy variables for those announcements

In our sample, two announcements -- construction.spending and the NAPM survey -- were
released on the same day in JO of 12 instances. In those JO cases, we set the dummy variable
equal to one; and for the days with only one of these announcements, we set it equal to one-half.

released at or before the corresponding interval time. The 9:15 AM set of
regressions, for example, includes dummy variables for both the 8:30 AM and
9:15 AM announcements, but not the 10:00 AM or 2:00 PM announc ements."
This procedure controls for the effects that earlier announcements might have on
subsequ ent volatility, volume, or spreads that day. To save space and focus on
the importa nt findings we do not present the control-variable results.
Our regression results in Table II therefore measure the immediate impact
of the various announcements on price volatility, trading volume, and the bidask spread. As shown in the table, nine of the 19 announcements show
significant effects on price volatility at the 1% level and none at the 5% level. Six
of the announcements show significant effects on trading volume at the 1 % level
and five at the 5% level. Ten show significant effects on the bid-ask spread at the
1% level and four at the 5% level.

The most important announcements
To compare announcements released at different times of the day, we
calculate the ratio of the coefficient

a;, to the intercept a~,.

The announcements

with the greatest effects on price volatility, which are also significant at the 1%
level, in the order of their importance are: employment, PPI, construction
spending and the NAPM survey, CPI, gross domestic product (GDP), retail sales,
industrial producti on and capacity utilization, consumer confidence, and new
single-family home sales. For trading volume, the most importa nt
announcements are: employment, construction spendin g and the NAPM survey,
PPI, new single-family home sales, retail sales, and industrial producti on and
capacity utilization in that order. For the bid-ask spread, the most importa nt
25

Personal income was released at I 0:00 AM for the first four announcements in our sample and
at 8:30 AM thereafter. When it is an 8:30 announcement we include it in the regressions for
intervals at and after 8:30. When it is a 10:00 announcement we include it only for intervals at
and after 10:00.

announcements are: employment, CPI, new single-family home sales, PPI, GDP,
durable goods, industrial production and capacity utilization, construction
spending and the NAPM survey, consumer confidence, and initial jobless claims
in that order.

One of our striking findings is the high correlation across variables in
which announcements seem to matter." For price volatility, trading volume, and
the bid-ask spread the 8:30 AM employment report is far and away the most
important announcement." In the first five minutes upon the report's release,
volatility rises 13-fold on average and the spread widens to three times its usual
value. Then in the next five-minute interval, volume surges to three-and-a-half
times its normal amount. While the rank ordering of announcements varies
somewhat after employment, there is nonetheless a strong relationship across
variables in which announcements matter. The PPI, for example, is ranked
second for impact on price volatility, third for trading volume, and fourth for the
.bid-ask spread.

This section has identified the immediate impact of the various
announcements on volatility, volume, and the spread. We next tum to the
adjustment of the market to announcements over somewhat longer time periods.
To conduct clearer tests, we focus on the major 8:30 AM announcements, which
we define as those that show significant effects on at least two of the variables at
the 1% level. By this criterion, the major 8:30 AM announcements are
employment, PPI, CPI, GDP, and retail sales.
26

The correlation between the 19 price volatility and trading volume coefficients in Table II is
0.89, significant at the ..01 level. Corresponding figures for the price volatility and bid-ask
spread coefficients and the trading volume and bid-ask spread coefficients are 0.80 and 0.91,.
respectively (both significant at the .01 level).
27

As noted by Ederington and Lee ( 1993), the employment report is the first government report
released concerning economic activity in a given month.

Five Stylized Facts
We use the major 8:30 AM announcements as conditioning variables to

clarify the stylized facts about the intraday patterns of price volatility, trading
volume, and bid-ask spreads. The precise time for the release of the
announcements provides an especially useful reference point for analyzing the
timing and persistence of various effects. In the analysis, we look at both oneminute and five-minute intervals. Table III compares days with any of the major
8:30 AM announcements identified in Section IV (employment, PPI, CPI, GDP, or
retail sales) with nonannouncement days for each one-minute interval from 8:25
to 8:37 AM. Table IV compares the same days for each.five-minute interval from
8:15 to 8:45 AM and for every third five-minute interval from 9:00 to 10:20 AM.
Figures 4A to 4C compare volatility, volume, and spread between announcement
and nonannouncement days for the whole trading day, indicating the instances
when the differences are significant. These tables and figures serve to document
five stylized facts:

Stylized fact 1: Volatility spikes without volume
The most volatile prices occur with a notable lack of trading volume in the
first few minutes after a major 8:30 AM announcement. Panel A of Table III
reports the standard deviation of changes in log prices over one-minute intervals
on announcement days, the standard deviation on nonannouncement days, and
the ratio of the standard deviations. Panel B reports mean trading volume on
announcement days, the mean on nonannouncement days, and the difference
between the two. As shown in Panel A, price volatility starts to rise a minute
before an announcement. 28 Volatility then spikes up in the next two minutes,

Ederington and Lee ( 1995) also note a rise in volatility in anticipation of an announcement.

reaching a peak of over 12 times the volatility for the same interval on
nonannouncement days. As shown in Panel B, during these two minutes,
trading volume tends to be less than the normal volume on nonannouncement
days, although the difference is not statistically significant. This lack of trading
volume during the time of the most volatile prices is striking, because to our
knowledge it has not been documented in the literature, even by researchers who
have analyzed the effects of public announcements." We find that not only is it
possible for prices to move with little trading volume but the most dramatic
price movements actually take place without a significant rise in trading volume.

Stylized fact 2: Volume surges with a lag
Trading volume does rise after an announcement but only with an
appreciable lag. As shown in Panel B of Table III, it takes two minutes for
volume to start to pick up after a major announcement, by which time price
volatility has already come down to a third of its level at the peak. Volume then
surges to its highest levels between four and seven minutes after an
announcement, when it averages about four times its average for
nonannouncement days. During this volume surge, volatility remains high but
at a level that is only a fifth of its peak level.

Stylized fact 3: Both volatility and volume persist at high levels and decline
together
Price volatility and trading volume both decline gradually while
remaining high for over an hour after a major 8:30 AM announcement. Panels A
and B of Table IV compare price volatility and trading volume on announcement
and nonannouncement days by five-minute intervals. As shown in the table,
volatility begins a process of decline after the spike but tends to remain
Locke (1996), for example, examines 15-minute intervals and finds that increased volume
accompanies the heightened volatility after announcements.
29

significantly higher than average for an hour after the announcement. In other
words, conditional volatility during the period is serially correlated. Beyond this
period, we observe significantly higher volatility for several more intervals but
these tend not to be consecutive intervals. Figure 4A shows the ratio of the
volatility on major 8:30 AM announcement days to the volatility on
nonannouncement days by five-minute intervals, indicating the ratios that are
significant at the 1% and 5% levels. Price volatility is significantly higher on
announcement days for every interval from 8:30 AM to 9:25 AM. Several
intervals of significantly higher volatility occur until early in the afternoon, but
after 11:30 AM such intervals are few and far between.

As shown in Table IV, trading volume on announcement days starts to
decline by 8:40 AM but remains significantly higher than on nonannouncement
days. Figure 4B shows the difference between trading volume on major 8:30 AM
announcement days and volume on nonannouncement days by five-minute
intervals. Trading volume is significantly higher on announcement days for
every interval from 8:30 AM to 10:10 AM. Intervals of significantly higher
volume are then fairly common until 1:00 PM. The fact that both volume and
volatility remain high for an extended period of time and decline together
suggests that their behavior is part of a process set off by the announcements.

Stylized fact 4: The market suffers a momentary liquidity loss at the time of
announcements
The market suffers some liquidity loss during a brief period around
announcements. An inverse measure of liquidity, the bid-ask spread widens
dramatically with a major announcement but narrows as soon as volume surges.
Panel C of Table III reports the mean bid-ask spread over one-minute intervals
on announcement days, the mean on nonannouncement days, and the difference
between the two. As shown in the panel, the spread starts to widen three

minutes before a major announcement while price volatility is still not
significantly higher than average. The most illiquid time coincides with the
volatility spike in the first minute after the announcement. During this time, the
spread is nearly six times wider than its average on nonannouncement days.,. In
the next two minutes, liquidity returns as the spread begins to narrow. Three
minutes after the announcement, the spread is no longer significantly different
from the nonannouncement day average. During this time, price volatility
remains high but trading volume has started to surge. Figure 4C shows the
difference between the mean bid-ask spread on major 8:30 AM announcement
days and the mean spread on nonannouncement days by five-minute intervals.
The figure shows that the spread is wider than the nonannouncement day
average for several more intervals after the announcement.

Stylized fact 5: High volume persists longer than high volatility
The announcement effects on trading volume appear to last longer than
the effects on price volatility. Table IV and Figure 4A suggest that significantly
higher volatility persists for at least an hour and possibly three hours after an
announcement. Such effects appear somewhat similar to those Ederington and
Lee (1993) found in their data, in which they detect significant effects on
volatility for 40 minutes and slight effects for several hours. We find the effects
on trading volume to be even more persistent. Table IV and Figure 4B suggest
that significantly higher volume persists certainly for an hour-and-a-half and
perhaps for four-and-a-half hours. Not only do we find the largest price
movements when there is little trading volume, we also find active trading when
there is little price movement. In the stock market, Beaver (1968), Morse (1981),
and Gallant, Rossi, and Tauchen (1992) also observe that high volume persists
after high price volatility has ended.
30

In contrast, Krinsky and Lee ( 1996) find an insignificant effect of earnings announcements on
bid-ask spreads for NYSE and AMEX stocks.

The stylized facts in one episode
The stylized facts we described are discernible even when we look at a
single episode. The largest volatility shock in our sample, took place just after
8:30 AM on August 5, 1994 when the July employment report was released.
Figure 5 shows the spread between the mean bid and ask quotes, the mean
transactions price, and trading volume by one-minute intervals from 8:00 AM to
10:00 AM on that day. As shown in the figure, in the half hour before the
announcement, the price of the 5-year note was relatively stable, the bid-ask
spread narrow, and trading volume low. The bid-ask spread then started to
widen a minute before the announcement. Upon the report's release, the price of
the 5-year note fell about 50 hundredth s of a point within three minutes of the
announcement, with trading still relatively thin. The spread, which was at its
widest in the first two minutes after the announcement, narrowed quickly in the
third minute as trading volume started to pick up. Volume then surged as the
price continued to fluctuate. For the next hour-and-a-half, both price volatility
and trading volume remained substantially higher than on nonannou ncement
days.

VI.

Some Hypotheses

We propose five specific hypotheses to explain the stylized facts. These
hypotheses assign important roles to public information, heterogeneity of
investors' views, sluggishness in price formation, inventory control behavior by
market makers, and liquidity traders who react with a lag to price changes.
These roles have not been emphasized in the recent empirical literature, but we
argue that they offer a consistent explanation of the price formation process and
the provision of liquidity in the U.S. Treasuries market.

Hypothesis 1: Public information plays a dominant role in bond price adjustment
The fact that the most volatile prices occur with relatively little trading
volume suggests a dominant role for public information. French and Roll (1986,
p. 9) distinguish between public information, "which affects prices before anyone
can trade on it," and private information, which "only affects prices through
trading." This dichotomy neatly identifies public information as the factor
driving the large price movements we find in the first few minutes after an
announcement. This result is reassuring, because the announcements should
contain largely public information. During these moments, the bid-ask spread is
at its widest, and this may explain the reluctance to trade.'1 During this period,
there is also no evidence of a price-volume relationship. Table V reports the
contemporaneous correlations between price volatility and trading volume
across days for one-minute intervals between 8:25 and 8:49AM. There is a
notable absence of a significant positive correlation between volatility and
volume in the first two minutes after an announcement, in contrast to the
significantly positive correlations in most of the other intervals.

Since volatility, volume, and spread are endogenous, it is also partly true that the spread is
wide because of low volume and high volatility.

The initial price adjustmen t after an announce ment is a measure of the
amount of public information contained in the announcem ent. We find that
price changes in the first five minutes after a major 8:30 AM announce ment
explain nearly half of price changes between 7:30 AM and 5:00 PM on those
days." The regressions reported in Table II measure this initial price adjustmen t
for the different announcements. The estimates suggest that the employme nt
report contains the most public information, followed by PPI, constructi on
spending and the NAPM survey, CPI, and GDP.
The dominant role we assign to public information in the bond market

,,. st.w4sJn sharfcont rast to its humhle role hf the stock market. French and Roll's
.
'.
.
ingenious study of stock return volatility around Wednesda ys when the New
York and American stock exchanges were closed to remedy a processing backlog
suggests that the high volatility during exchange trading hours is explained by
private rather than public information. Stoll and Whaley (1990) also conclude
that private information is the dominant factor in stock return volatility. Berry
and Howe (1994) develop an intraday measure of public informatio n flow based
on Reuter's news releases and find no significant relationship between their
measure and stock price volatility." In the bond market's case, Ederingto n and
Lee (1993) did point to the announcem ents as a major source of price volatility,
and it would seem obvious that the announce ments contain largely public
information. However, they lacked the data to show that much of the volatility
took place without a rise in trading volume.

The regression of the daily change in log prices (i.e., from 7:30 AM to 5:00 PM) on the fiveminute change results in an adjusted R-squared of 0.470, The adjusted R~squared is a slightly
higher 0.506 if absolute price changes are used instead of signed price changes.
33

The study by Jones, Kaul, and Lipson (1994) is an exception in this literature. They find that
the adverse-selection component of bid-ask spreads (a component they associate with private
information) on NASDAQ-NMS stocks is small and conclude that public information drives
short-term volatility. ·

Hypothesis 2: The surge in volume arises from wide disagreement about the price
It is remarkable that trading volume surges to its highest levels only after
the initial price adjustment following an announcement. Our hypothesis is that
· the trading activity arises from an initially wide disagreement among investors
regarding the price adjustment. The hypothesis harks back to Beaver (1968, p.
69) who attributed the price changes in the stock market in weeks of annual
earnings announcements to "changes in the expectations of the market as a
whole" and the rise in trading volume to "a lack of consensus regarding the
price." When an announcement is released, almost everyone in the market
would understand at least some of the price implications -- the direction of
prices, for example -- and prices should move sharply to reflect this common
understanding even without any trading. The precise magnitude of the
appropriate price change, however, would be a matter of analysis or
interpretation, which would differ among investors. Given some heterogeneity
of views conditional on the announcement, some investors would see the initial
price adjustment as an overreaction, others as an underreaction. These investors
would then take positions on the strength of their views and thus produce the
volume surge.

In the recent price formation literature, a number of formal models show
that some form of heterogeneity of views among investors can generate
speculative trading activity. To account specifically for surges in volume
following public announcements, Kim and Verrecchia (1991) assume investor
idiosyncrasies, Foster and Viswanathan (1993) represent beliefs with elliptically
contoured distributions, Harris and Raviv (1993) rely on differences in opinion,
and Kandel and Pearson (1995) specify agents with different likelihood functions.
Note, however, that the phenomenon we would like to explain is not just the fact
that volume surges after a public announcement but that it surges only after the

initial price adjustment that follows the announcem ent. These models take no
account of the bid-ask spread, and a lack of liquidity at the time of the
announce ment may help explain the apparent trading pause.
:<.

Hypothesis 3: The price-volume relationship reflects a sluggishness in the
formation of a consensus price

.'i"'

' along with high and
The persistenc e of high and declining price volatility
·"

declining trading volume suggests a slow process of price adjustm~n t to
',!'l

reconcile the heterogen eous views of investors. The hypothesi s assumes that the
consensus price that ultimately prevails correspon ds to a weighted average of
t~

the various investors' views. Investors partially reveal their views by trading
_·

when the market price deviates significantly from their forecasts of the consensus

~·

price. In making their forecasts, individual investors ra,tionally put some weight
:) \

on their own views and some weight on the market price at the time. The
0'

deviations between the market price and individual forecasts would tend to be
widest after the initial price adjustmen t following the announce ment, and
speculativ e trades would be correspon dingly large.

·1-~

p,
It is an important part of the hypothesi s that not all investors trade to fully
reveal their views at once so that a consensus price is not reached quickly. The
difficulty of establishin g a consensus price would arise from noise in prices and
::-

trades that serve to obscure traders' views. Such noise may come from strategic
12;
trading by dealers or from the use of customers ' liquidity trades to disguise
0

speculativ e trades. Thus the market price adjusts sluggishly over time to reflect
'rl1.

the views of more investors more closely at the same time that investors revise
tf-

their forecasts to put more weight on the market price and Jess on their own
::.;

initial views. Deviation s between price and forecast slowly narrow and
·,p .

speculativ e trades decrease. In this way, we would observe high and declining
1('

volume with high and declining volatility until the convergence to a consensus
price.

He and Wang's (1995) model of differentially informed investors is one
that gives rise to a slow convergence to a consensus price along with persistent
trading volume." In their model, investors trade for several rounds after they
receive information. Hence, volume and volatility persist even when the arrival
of information does not. Such persistence is possible because prices are noisy
and not fully revealing of all the traders' information. He and Wang, however,
base these results on differential private information. When they analyze the
effects of public announcements, they find that investors take speculative
positions before the announcements and unwind those positions immediately
afterwards. We find little evidence of such effects in the U.S. Treasuries market.
The investors in this market seem more willing to bet on the consensus price
conditional on an announcement than on the announcement itself. Our stylized
facts are more consistent with He and Wang's private information story, with
such information interpreted as private views induced by a public
announcement.
We have more difficulty interpreting our results in terms of the mixtureof-distributions hypothesis advanced by Clark (1973), Epps and Epps (1976),
Tauchen and Pitts (1983), and Lamoureux and Lastrapes (1990). The hypothesis
explains serial correlation in volatility and the price-volume relationship in terms
of the arrival of new information. The explanation is consistent with the idea
that private information is conveyed to prices through trading. Serial correlation

Foster and Viswanathan's (1996) model of straiegic trading with agents forecasting one
another's forecasts contains elements consistent with our hypothesis. Strategic behavior is one
way to prevent prices and trade flows from fully revealing investors' private forecasts. However,
they focus on the correlation structures of information signals rather than on price-volume
relationships.

in volume would require a sequential flow of new information. 35 While such
information flow may be possible, it is hard to conceive of the special kind of
new information that would produce not just the persistence of high volatility
and high volume but the consistent decline of volatility and volume over time
following public announcements.

Hypothesis 4: Inventory risk control drives liquidity
The brief period of illiquidity around a major announcement is consistent
with the behavior of competing dealers who are simply trying to control the risks
in their inventories. In this market, dealers face risk arising from the uncertainty
about returns on their inventories and from the uncertainty about transaction
flows. The effort to control risk would cause them to quote an extraordinarily
wide spread at the time of major announcements in anticipation of the high price
volatility. Once the announcement is out, they would begin to quote a much
narrower spread in spite of continued high price volatility because the ensuing
surge in trading volume greatly reduces the transaction flow uncertainty.
In the microstructure literature, recent models of market making based on
asymmetric information have challenged the more traditional models based on
inventory control. The inventory-control models of Demsetz (1968), Tinic and
West (1972), Amihud and Mendelson (1980), and Ho and Stoll (1983) emphasize
the increased risk to market makers of high price volatility and low trading
volume. Market makers' efforts to control this risk result in a positive
relationship between the bid-ask spread and volatility or other measures of price

" Bollerslev, Chou, and Kroner ( 1992) call for an explanation of "serial correlation in
· conditional second moments" as a property of speculative prices. The mixture-of-distributions
hypothesis with serially correlated information flows is a possible explanation. Jones, Lamont,
and Lumsdaine (1996) find some support for this explanation at daily frequencies. Our
hypothesis of a slow convergence to a consensus price after the arrival of information offers an
alternative explanation for intraday serial correlation.

risk and a negative relationship between the spread and volume. The
asymmetric-information models of Glosten and Milgrom (1985), Kyle (1985),
Admati and Pfleiderer (1988), and Easley and O'Hara (1992), however,
emphasize the risk to market makers of dealing with investors with superior
information. As a result, market makers would set the spread based on when the
informed and uninformed trade. Admati and Pfleiderer, for example, propose
that discretionary liquidity traders would cluster their trading activity around
certain times of the day and that the resulting liquidity would attract the
informed investors who would in tum bring price volatility. Hence high
volume, high volatility, and narrow spreads would coincide.
The evidence from the U.S. Treasuries market lends support to the
traditional inventory control models of the bid-ask spread. To see this, we
regress the spread on concurrent and anticipated next-period price volatility and
trading volume. For anticipated volatility and volume, we use the predicted
values from the regressions reported in Table II, which use announcement
dummy variables as explanatory variables. We run the regression on a panel of
five-minute intervals across our full sample of 250 trading days. The
estimated regression is
st = 1.494 + 1.029 Pt+ 0.756 Et (Pt+i) - 0.027 vt - 0.026 Et (vt+i)

where st is the bid-ask spread, Pt is price volatility, vt is trading volume, and
Et (Pt+i) and Et (vt+i) are anticipated volatility and volume for the next five-

minute interval.,. Each coefficient is significantly different from zero at the 1%
level and the adjusted R-squared statistic is 0.107. The slope coefficients have
the expected signs: the spread widens with both concurrent and anticipated

The bid-ask spread is measured in hundredths of a percent, price volatility is the change in log
prices times a thousand, and trading volume is measured in tens of millions of U.S. dollars.
·

price volatility and narrow s with both concurrent and anticipated trading
volume.

The performance of the regression across time intervals suggests that the
way the competing dealers set the spread around announcement times is similar
to the way they set it at other times of the day. Such consistency would not be
so appare nt if asymmetric information was important. Figure 6 compares the
regression's prediction of the bid-ask spread to the actual spread. The actual
difference is the difference between the mean spread on days with major 8:30 AM
announcements and the mean spread on nonannouncement days. The predicted
difference is the difference between the mean predicted spread on announcement
days and the mean spread on nonannouncement days. The figure shows that
our model performs consistently well in explaining the average spread before a
major announcement, around the announcement, and after the announcement.
To explain the spread' s dramatic behavior around major announcements,
Figure 6 also presents two other predicted spread measures. Predicted difference

with usual trading volume is the difference between the mean predicted spread on
announcement days assuming trading volume did not increase and the mean
spread on nonannouncement days. This counterfactual measure shows that the
bid-ask spread would likely have stayed high on major announcement days had
trading volume not also increased with price volatility. Predicted difference with
usual price volatility is the difference between the mean predicted spread on
announcement days assuming price volatility did not increase and the mean
spread on nonannouncement days. This counterfactual measure shows that the
bid-ask spread would have been even lower on announcement days had the high
trading volume that accompanies major announcements not also been
accompanied by high price volatility. The rapid decline of the bid-ask spread in

spite of continuing high price volatility seems to result from the offsetting effects
of high trading volume.

Hypothesis 5: Liquidity traders react with a lag to price changes
To explain why high volumes persist for a longer period than high
volatilities, we propose that certain liquidity traders react to price changes. In
existing microstructure models, liquidity traders do not need to react to price
changes; they trade for purely exogenous reasons. In the bond market,
however, there are good reasons for a reaction. Investment strategies that
involve duration targets and dynamic hedging strategies for swaps, options, and
mortgage-backed securities require a reaction to price changes. In theory the
reaction should be continuous, but in practice transactions costs lead to an
optimal reaction lag. Investors following these strategies are liquidity traders in
the sense that they do not speculate on prices and their trades would not cause
price volatility. Having an optimal reaction lag limits their discretion over when
they may trade during the day. Some liquidity traders may begin to react as
soon as the bid-ask spread narrows with the surge in volume after an
announcement, but with continued volatility some of them would still be
reacting later in the day. Some liquidity traders may react with lags of half an
hour or longer, generating high volumes even after volatility has returned to
normal levels.
Our data provide evidence that price volatility leads to trading volume.
Table VI reports Granger causality tests of volume and volatility based on six
lags of five-minute intervals for our full sample of trading days. 37 The tests
confirm a persistence in volatility controlling for past volume and a persistence
cif volume controlling for past volatility. More interestingly, the tests also show
37

Volatility is the absolute change in log prices times a thousand and volume is measured in
tens of millions of U.S. dollars.

that volatility causes subsequent volume and to a lesser extent that volume
causes subsequent volatility. The former is consistent with our hypothesis of
liquidity traders reacting with a lag to price changes.'" Similar evidence of a
persistence of high volume beyond the period of high volatility has been found
in the stock market by Beaver (1968), Morse (1981), Jain and Joh (1988), and
Gallant, Rossi, and Tauchen (1992).3' Such evidence in the stock market may also
be explained by liquidity traders who react to price changes.

VII.

Conclusion

The recent availability of data on the interdealer U.S. Treasuries market
allows us to examine in a new light how secondary markets form prices and
provide liquidity. The market we study is an extremely liquid one where many
dealers compete as both market makers and informed investors. We focus the
examination on intraday patterns around the times of macroeconomic
announcements, because they identify precise times for the arrival of information
in the market. In the body of the paper, we describe our most striking results in
the form of five stylized facts and we discuss various hypotheses to explain them
separately. In this section, we recapitulate by pulling the facts and hypotheses
together.

An alternative hypothesis is that the longer persistence of volume reflects the unwinding of
speculative positions later in the day. In theory dealers should be able to execute such unwinding
without causing price volatility and in practice such execution would be made easier by the
presence of liquidity traders.
39

Karpoff (1987, p. 123) finds such evidence puzzling, paticularly because it "casts doubt on the
interpretation of volume statistics as measures of 'information content' in event studies."

Our analysis suggests that the announcements set off an extended price
formation process. The sharpest price adjustments take place in the first few
moments of the process, when prices adjust to reflect public rather than private
information. We infer that the information is indeed public because we see no
rise in trading activity. From then on, the price formation process continues but
with a new shape, in which further price adjustment is accompanied by a surge
in trading volume. We believe this continuation of the process is driven by an
initially wide disagreement among investors about what the new information
means for prices. We do not think a sequential arrival of private information can
easily explain why the process should follow the release of public information
and involve a steady decline in volatility and volume. To us, the process looks
like a process of convergence toward a consensus price.

The persistence of high volatility and high volume suggests a rather
sluggish convergence process. Even so liquid a market seems to have great
difficulty resolving disagreement among investors. He and Wang's (1995) model
shows that such a sluggish process is at least a theoretical possibility. The source
of the difficulty is not that views are held stubbornly, since the consensus price
that ultimately prevails should reflect aH the views. The difficulty arises from
noise in prices and trades that serves to obscure investors' views from one
another. The noise may come from strategic trading by dealers or from the use
of customers' liquidity trades to disguise speculative trades.

The moments in which prices adjust sharply to public information are also
among the few times the market suffers from some illiquidity. These are the
moments of the widest bid-ask spreads away from the start and end of the
trading day. The dealers hesitate to provide liquidity simply because of the
uncertainty about the price changes immediately following the announcement
and not because of the risk of trading with better informed investors. The

asymmetric information assumption is inconsistent with the notion that the price
changes during these moments arise from public information. Indeed, the
dealers soon become willing to provide liquidity at the very time when a surge in
volume and high volatility might suggest trading by better informed investors.
Instead of widening, the bid-ask spread narrows. We believe the spread narrows
because the volume arises from speculative trading by investors who disagree on
the price adjustment as well as from liquidity trading by investors who react
with a lag to the price changes.

References

Admati, A. and P. Pfleiderer, 1988, A theory of intraday patterns: Volume and
price variability, Review of Financial Studies 1, 3-40.
Amihud, Y. and H. Mendelson, 1980, Dealership market: Market making with
inventory, Journal of Financial Economics 8, 31-53.
Andersen, T.G. and T. Bollerslev, 1996, OM-dollar volatility: Intraday activity
patterns, macroeconomic announcements, and longer run dependencies,
NBER Universities Research Conference (May).
Beaver, W.H., 1968, The information content of annual earnings announcements,
Empirical Research in Accounting: Selected Studies, supplement to The
Journal of Accounting Research, 67-92.
Becker, KG., J.E. Finnerty, and K.J. Kopecky, 1996, Macroeconomic news and the
efficiency of the international bond markets, Journal of Futures Markets 16,
131-145.
Berry, T.D. and K.M. Howe, 1994, Public information arrival, Journal of Finance
44, 1331-46.
Bollenbacher, G.M., 1988, The Professional's Guide to the U.S. Government Securities
Markets: Treasuries, Agencies, Mortgage-Backed Instruments. New York:
New York Institute of Finance.
Bollerslev, T., R.Y. Chou, and K.F. Kroner, 1992, ARCH modeling in finance: A
review of the theory and empirical evidence," Journal of Econometrics 52, 559.
Clark, P.K., 1973, A subordinated stochastic process with finite variance for
speculative prices, Econometrica 41, 135-155.
Cook, T. and S. Korn, 1991, The reaction of interest rates to the employment
report: The role of policy anticipations, Economic Review, Federal Reserve
Bank of Richmond, 3-12.
Crain, S.J. and J.H. Lee, 1995, Intraday volatility in interest rate and foreign
exchange spot and futures markets, Journal of Futures Markets, forthcoming.
Demsetz, H., 1968, The cost of transacting, Quarterly Journal of Economics 82, 3353.

Department of the Treasury, Securities and Exchange Commission, and Board of
Governors of the Federal Reserve System, 1992, Joint report of the
government securities market, January.
Dwyer, G. and H.W. Hafer, 1989, Interest rates and economic announcements,
Review, Federal Reserve Bank of St. Louis, 34-46.
Easley, D. and M. O'Hara, 1992, Time and the process of security price
adjustment, Journal of Finance 47, 577-605.
Ederington, L.H. and J.H. Lee, 1993, How markets process information: News
releases and volatility, Journal of Finance 48, 1161-91.
Ederington, L.H. and J.H. Lee, 1995, The short-run dynamics of the price
adjustment to new information, Journal of Financial and Quantitative
Analysis 30, 117-134.
Epps, T.W. and M.L. Epps, 1976, The stochastic dependence of security price
changes and transactions volumes: Implications for the mixture-ofdistributions hypothesis, Econometrica 44, 305-321.
Fleming, M., 1996, The round-the-clock market for U.S. Treasury securities,
Federal Reserve Bank of New York.
Federal Reserve Bank of New York, 1988, Primary dealers: Criteria and
procedures applied to firms interested in becoming and remaining
primary dealers, Dealer Surveillance Function (November).
Foster, F.D. and S. Viswanathan, 1993, The effect of public information and
competition on trading volume and price volatility, Review of Financial
Studies 6, 23-56.
Foster, F.D. and S. Viswanathan, 1996, Strategic trading when agents forecast
the forecasts of others, Journal of Finance 51, 1437-78.
French, K. and R. Roll, 1986, Stock return variances: The arrival of information
and the reaction of traders, Journal of Financial Economics 17, 5-26.
Gallant, A.R., P.E. Rossi, and G.E. Tauchen, 1992, Stock prices and volume,
Review of Financial Studies 5, 199-242.
Glosten, L.R. and P.R. Milgrom, 1985, Bid-ask and transaction prices in a
specialist market with heterogeneously informed traders, 14, 71-100.

Hardouvelis, G.A., 1984, Market perceptions of Federal Reserve policy and the
weekly monetary announcements, Journal of Monetary Economics 14, 225240.
Hardouvelis, G.A., 1988, Economic news, exchange rates, and interest rates,
Journal of International Money and Finance 7, 23-35.
Harris, M. and A. Raviv, 1993, Differences of opinion make a horse race, Review
of Financial Studies 6, 473-506.
Hasbrouck, J., 1991, Measuring the information content of stock trades, Journal of
Finance 46, 179-208.
He, H. and J. Wang, 1995, Differential information and dynamic behavior of
stock trading volume, Review of Financial Studies 8, 919-72.
Ho, T. and H. Stoll, 1983, The dynamics of dealer markets under competition,
Journal of Finance 38, 1053-74.
Jain, P.C. and G.-H. Joh, 1988, The dependence between hourly prices and
trading volume, Journal of Financial and Quantitative Analysis 23, 269-283.
Jones, C.M., G. Kaul, and M.L. Lipson, 1994, Transactions, volume, and
volatility, Review of Financial Studies 7, 631-51.
Jones, C.M., 0. Lamont, and R. Lumsdaine, 1996, Public information and the
persistence of bond market volatility, NBER Working Paper 5446
(January).
Kandel, E. and N.D. Pearson, 1995, Differential interpretation of public signals
and trade in speculative markets, Journal of Political Economy 103, 831-872.
Karpoff, J., 1987, The relation between price changes and trading volume: A
survey, Journal of Financial and Quantitative Analysis 22, 109-26.
Kim, 0. and RE. Verrecchia, 1991, Market reaction to anticipated
announcements, Journal of Financial Economics 30, 273-309.
Krinsky, I. andl Lee, 1996, Earnings announcements and the components of the
bid-ask spread, Journal of Finance 51, 1523-36.
Kyle, A., 1985, Continuous auctions and insider trading, Econometrica 53, 131535.

Lamoureux, C.G. and W.D. Lastrapes, 1990, Heteroskedasticity in stock return
data: Volume vs. GARCH effects, Journal of Finance 45, 487-98.
Locke, P., 1996, The locals' spin of the national news, Commodity Futures
Trading Commission (June).
Lyons, R.K., 1995, Tests of microstructural hypotheses in the foreign exchange
market, Journal of Financial Economics 39, 321-351.
Madhavan, A. and S. Smidt, 1991, A Bayesian model of intraday specialist
pricing, Journal of Financial Economics 30, 99-134.
Madigan, B. and J. Stehm, 1994, An overview of the secondary market for U.S.
Treasury securities in London and Tokyo," Board of Governors of the
Federal Reserve System, Finance and Economics Discussion Series 94-17
(July).
Mc!nish, T.H. and R.L. Wood, 1992, An analysis of intraday patterns in bid/ask
spreads for NYSE stocks, Journal of Finance 47, 753-764.
Morse, D., 1981, Price and trading volume reaction surrounding earnings
announcements: A closer examination, Journal of Accounting Research 19,
374-383.
Roley, V.V. and R. Troll, 1983, The impact of new economic information on the
volatility of short-term interest rates, Economic Review, Federal Reserve
Bank of Kansas City (February), 3-15.
Stigum, Marcia, 1990, The Money Market. Revised Edition. Homewood, Illinois:
Dow-Jones Irwin.
Stoll, H.R. and R.E. Whaley, 1990, Stock market structure and volatility,
Review of Financial Studies 3, 37-75.
Stronglin, S. and V. Tarhan, 1990, Money supply announcements and the
market's perception of Federal Reserve policy, Journal of Money, Credit, and
Banking 22, 135- 153.
Tauchen, G.E. and M. Pitts, 1983, The price variability-volume relationship on
speculative markets, Econometrica 51, 485-505.
Thornton, D., 1989, The effect of unanticipated money on.the money and foreign
exchange markets, Journal of International Money and Finance 8, 573- 587.

Tinic, S.M. and R.R. West, 1972, Competition and the pricing of dealer services
in the over-the-counter market, Journal of Financial and Quantitative
Analysis 8, 1707-1727.
U.S. General Accounting Office, 1986, U.S. Treasury securities: The market's
structure, risks, and regulation, GAO/GGD 86-80BR (August).
Urich, T. and P. Wachtel, 1984, The effects of inflation and money supply
announcements on interest rates, Journal of Finance 39, 1177-88.

Appendix A: Primary Government Securities Dealers
(as of October 1996)

BA Securities, Inc.
Barclays de Zoete Wedd Securities Inc.
Bear, Stearns & Co., Inc
BT Securities Corporation
Chase Securities Inc.
CIBC Wood Gundy Securities Corp.
Citicorp Securities, Inc.
CS First Boston Corporation
Daiwa Securities America Inc.
Dean Witter Reynolds Inc.
Deutsche Morgan Grenfell/C.J. Lawrence Inc.
Dillon, Read & Co. Inc.
Donaldson, Lufkin & Jenrette Securities Corporation
Eastbridge Capital Inc.
First Chicago Capital Markets, Inc.
Fuji Securities Inc.
Goldman, Sachs & Co.
Greenwich Capital Markets, Inc.
HSBC Securities, Inc.
Aubrey G. Lanston & Co., Inc.
Lehman Brothers Inc.
Merrill Lynch Government Securities Inc.
J.P. Morgan Securities, Inc.
Morgan Stanley & Co. Incorporated
NationsBanc Capital Markets, Inc.
Nesbitt Burns Securities Inc.
The Nikko Securities Co. International, Inc.
Nomura Securities International, Inc.
Paine Webber Incorporated
Prudential Securities Incorporated
Salomon Brothers Inc.
Sanwa Securities (USA) Co., L.P.
Smith Barney Inc.
SBC Capital Markets Inc.
UBS Securities LLC
Yamaichi International (America), Inc.
Zions First National Bank

Appendix B: Data Cleaning and Processing
GovPX distributes its information through on-line vendors by sending out a
digital ticker feed. Daily backup copies of the feed are used in this study. The data
provides a precise history of the tick-by-tick trading information sent to GovPX
subscribers. Any posting errors made by the interdealer brokers that are not filtered out
by GovPX are of course included in the backup files. Additionally, since the purpose of
the digital feed is to refresh vendors' screens, the data must be processed before it can be
effectively analyzed.

Trades
When a trade occurs two pieces of information are typically transmitted by
GovPX. First, during the "work-up" stage when traders are jumping into a transaction,
the news is posted that a bid is being "hit" or that an offer is being lifted (a "take"), as are
price and volume information. Seconds later the total volume of the trade or trades is
posted. Transactions through the same interdealer broker at the same price and
virtually the same time are thus counted as a single trade. Occasionally GovPX posts
several lines of data per trade, and sometimes there is only a single line.
The volume data is processed to ensure that each trade is counted once and only
once. This is done by retaining the price and volume data from the first line reporting
an increase in total daily volume for that security. This is typically the second line of
data reported for any given trade since the first line usually reports only that a trade is
occurring (with no trade size or increase in daily volume reported) . In our year of
intraday data there are a few instances where total daily volume shows a trade-to-trade

decline for the on-the-run 5-year note. These errors are corrected based on that day's
history of trades and the total daily volume reported for that security.'° In our year of
data we find 164,822 trades for the on-the-run 5-year note, or an average of 659 per day.
40

For example, at 3:32 PM eastern time (ET) on September 28, 1993 GovPX reports an
unprecedented trade of $810 million. Seventy minutes later GovPX reports a decline in
accumulated volume of $807 million for that same security. The size of the trade, its subsequent

Quotations
We put bid-ask quotes through a multi-step screening procedure:
•

Bid price movements of at least 3/4 of a point that are followed by a
similarly sized movement in the opposite direction (within 25%) are
dropped. This screen is imposed as several instances were found where a
security was erroneously reported to have moved a full point followed by a
full point correction within minutes. This eliminates an average of 0.3 quotes
per day.

•

As offers (or asks) in the dataset are quoted off of the bids, large positive
spreads are indistinguishable from small negative spreads. Spreads
calculated to be greater than 0.9 point (and less than 1.0 point) are likely to be
either errors or negative spreads that existed only momentarily when quotes
arrived from two different brokers. These quotes (22 per day) are dropped.

•

One-sided quotes (a bid or an offer but not both) are occasionally posted by
dealers. This study makes no use of these bids (22 per day) or offers (18 per
day).

•

Finally, spreads with bid-ask midpoints more than 10 standard deviations
from the daily bid-ask midpoint mean or daily price mean are dropped as are
spreads more than 10 standard deviations from the daily spread mean. This
screens out 1.5 quotes per day.

As spreads posted by the interdealer brokers do not include the brokerage fee
charged to the transaction initiator, zero spreads are common and can persist for
lengthy periods. Quotes calculated to be zero are therefore kept in the dataset. Our
dataset retains 541,745 quotes from the sample period for the on-the-run 5-year note, or
an average of 2,167 per day.

Times
GovPX's digital ticker feed contains a minute-by-minute time stamp. The
stamps typically appear 60 seconds apart within a day, but the exact timing of the stamp
varies across the days. One day may have time stamps of 8:28:40, 8:29:40, 8:30:40, etc.
while another day has time stamps of 8:28:53, 8:29:53, 8:30:53, etc. In our analysis each

reversion (as measured by accumulated volume), and the daily total volume statistic (reported in
a separate file) all suggest that the original trade was an error. We therefore scale down the
original trade size by the size of the reversion ($807 million).

interval is assumed to start at the beginning of the new minute that starts in that
interval. Our 8:30-35 interval therefore refers to the five minute period starting between
8:29 and 8:30 and ending five minutes later between 8:34 and 8:35, while our 8:30-31
interval refers to the one-minute period starting between 8:29 and 8:30 and ending
between 8:30 and 8:31.
Missing time stamps on a few of our sample days indicate times when we are
missing data. The most serious case of missing data in our sample is December 10, 1993
when our daily file ends at 12:10 PM (ET) instead of 6:00 PM. Most of our periods of
missing data occur outside of New York trading hours (7:30 AM - 5:00 PM), however,
and none appear within even a few hours of one our announcements.

Time Intervals
For tractability purposes our data is consolidated and analyzed at the fiveminute or one-minute interval. While the on-the-run 5-year note is the most actively
traded Treasury security, we are nevertheless missing some bid-ask spread and price
volatility observations." Specifically, the bid-ask spread is defined for 98.2% of the fiveminute trading intervals in our sample." As our definition of price volatility is based on
the change in price between successive intervals, this variable is defined for a slightly
lower 96.9% of five-minute trading intervals. Not surprisingly, most of our missing
observations are either early in the morning (7:30 - 8:00 AM) or late in the afternoon
(4:00 PM - 5:00 PM) when trading activity is light." Looking at the one-minute interval,
we have bid-ask spread (price volatility) observations for 94.1% (89.2%) of the trading
intervals on which we focus our one-minute analysis (8:25 - 8:37 AM).

There are no missing observations for trading volume as it is defined to be zero when there are
no transactions in an interval.

We count trading intervals as the number of five-minute intervals between 7:30 AM (En and
the market close for our 250 day sample. The market close is defined as the time of the last bidask quotation (for'the on-the-run 5-year note) for days when the market closed early and 5:00 PM
otherwise.
43

In the one hour period (8:20 AM - 9:20 AM) around the important 8:30 AM announcements,
we have bid-ask spread (price volatility) observations for 99.97% (99.77%) of the trading
intervals.

Holidays
We exclude 10 days from our sample when the Treasuries market was closed.
The market was closed on nine of these days in recognition of a holiday and on one day
(April 27, 1994) for the funeral of President Nixon." We retain one day in our sample,
April 1, 1994 (Good Friday), when the Public Securities Association (an industry trade
group) recommended that the bond market stay closed. The release of the March
employment report that day resulted in significant morning trading volume. We also·
retain several days in our sample when the market closed early (e.g., 2:00 PM), typically
before a holiday.

Gov PX recorded at least some overseas trading on five of these days and on two of the days
some light trading occurred during morning trading ho.urs in New York.

Table I

Macroeconomic Announcements
Announcement time, title, and reporting entitities for eighteen monthly macroeconomic announcements and one weekly (initial jobless claims) macroeconomic
announcement. All times are eastern time (ET).
Full Title

Short Title

Time

Rel!!!rtinll Enti~

8:30A.M.

Conswner Price Index (CPI)

Consumer Price Index

Bureau of Labor Statistics

8:30A.M.

Durable Goods Orders

Advance Report on Durable Goods Manufacturers'
Shipments and Orders

Bureau of the Census

8:30A.M.

Employment

The Employment Situation

Bureau of Labor Statistics

8:30A.M.

Gross Domestic Product (GDP)

Gross Domestic Product

Bureau of Economic Analysis

8:30AM.

Housing Starts

Housing Starts and Building Permits

Bureau of the Census

8:30A.M.

Initial Jobless Claims

Bureau of Labor Statistics

8:30A.M.

Leading Indicators

Composite Indexes of Leading, Coincident, and
Lagging Indicators

Bureau of Economic Analysis

8:30A.M.

Merchandise Trade

Report of U.S. Merchandise Trade

Bureau of the Census

8:30A.M.'

Personal Income

Personal Income and Outlays

Bureau of Economic Analysis

8:30AM.

Producer Price Index (PPI)

Producer Price Indexes

Bureau of Labor Statistics

8:30A.M.

Retail Sales

Advance Retail Sales

Bureau of the Census

Industrial Production and Capacity Utilization

Federal Reserve Board

Business Inventories

Manufacturing and Trade: Inventories and Sales

Bureau of the Census

10:00A.M.

Conswner Confidence

Consumer Confidence Index

Conference Board

10:00A.M.

Construction Spending

Value of New Construction Put in Place

Bureau of the Census

Factory Inventories

Manufacturers' Shipments, Inventories, and Orders

Bureau of the Census

10:00A.M.

NAPMSurvey

National Association of Purchasing Management
Index

National Association of
Purchasing Management

10:00A.M.

New Single-Family Home Sales

New One-Family Houses Sold and For Sale

Bureau of the Census

2:00P.M.

Federal Budget

Treasury Statement (The Monthly "Budj!et")

Department of the Treas!!!!

9:15A.M.
10:00A.M.

10:00A.M.

• Personal income was reported at 10:00 A.M. for the first four announcements in the period of analysis and at 8:30 A.M. thereafter.

Table II

The Impact of Announcements on Price Volatility, Trading Volume, and Bid-Ask Spreads for the Five-Year Treasury Note
Regressions of price volatility, trading volume, and bid-ask spread on announcement dummy variables for the five-year treasury note.• Results are presented for four
five-minute intervals (Panels A, B, C, and D) corresponding to the four sets of announcement times from Table I. For price volatility and the bid-ask spread the fiveminute period immediately following the announcements is presented (e.g., 8:30-8:35 for the 8:30 A.M. announcements), while for trading volume the succeeding
five-minute period is presented (e.g., 8:35-8:40 for the 8:30 A.M. announcements). The Panel B, C, and D regressions were run including dummy variables for
announcements that are reported earlier in the day, although these coefficients are excluded from the table. One and two asterisks indicate significance at the .05 and
.01 levels, respectively. The period of analysis is August 23, 1993 - August 19, 1994.
Price Volatility
Regression
Coefficient

P-Value

Trading Volume
Regression
Coefficient

Bid-Ask Spread

P-Value

Regression
Coefficient

P-Value

Panel A: 8:30 A.M. Announcements
Intercept

0.223**

0.001

7.717**

0.001

1.623**

0.001

Consumer Price Index

0.783**

0.001

4.443*

0.045

1.178**

0.001

Durable Goods Order

0.073

0.732

0.601

0.781

0.737**

0.006

Employment

2.625**

0.001

18.159**

0.001

3.189**

0.001

Gross Domestic Product

0.732**

0.001

4.210*

0.043

0.851**

0.001

Housing Starts

0.116

0.573

3.489

0.093

0.481

0.060

Initial Jobless Claims

0J43

0.194

2.492*

0.025

0.386**

0.005

Leading Indicators

-0.320

0.126

-0.148

0.944

0.213

0.413

Merchandise Trade

0.217

0.298

1.945

0.355

0.663*

0.011

Personal Income

0.230

0.356

-1.273

0.612

0.737*

0.018

Producer Price Index

1.401 **

0.001

7.875**

0.001

0.976**

0.001

Retail Sales

0.633**

0.006

6.705**

0.004

0.601*

0.035

AdjustedR2

0.480

0.298

0.462

Table II - Continued

Panel B: 9: 15 A.M. Announcements
Intercept

0.152**

0.001

6.038**

0.001

1.397**

0.001

Industrial Production and Capacity IJtilization

0.358**

0.001

4.823**

0.005

0.632**

0.001

Adjusted R2

0.167

0.215

0.107

Panel C: 10:00 A.M. Announcements
0.163**

0.001

6.012**

0.001

1.302**

0.001

0.075

0.480

3.570

0.063

0.414*

0.020

Construction Spending-NAPM Survey•

0.652**

0.001

8.440**

0.001

0.559**

0.001

Consumer Confidence

0.353**

0.001

3.900*

0.023

0.416**

0.009

0.061

0.537

4.092*

0.022

0.151

0.357

New Single-Family Home Sales

0.307**

0.004

6.049**

0.002

0.821**

0.001

Personal Income

-0.195

0.238

1.948

0.512

-0.292

0.288

Intercept

Business Inventories

Factory Inventories

Adjusted R2

0.181

0.186

0.202

Panel D: 2:00 P.M. Announcements
Intercept
Federal Budget
Adjusted R2

0.189**

0.001

4.629**

0.001

1.526**

0.001

0.006

0.893

-1.100

0.415

0.037

0.833

-0.020

0.019

-0.013

a Price volatility is defined as the absolute value of the log price change times lo'. Trading volume is reported in tens of millions of U.S. dollars. The bid-ask spread equals the actual
bid-ask spread times 10'.
b Construction spending and NAPM Survey are combined into a single dununy variable since they were released at the same time ten of twelve times in the sample. The dununy variable
is set equal to one for days when both reports were released, ½ for days when only one report was released, and zero otherwise.

Table ill

Persistence of Price Volatility, Trading Volume, and Bid-Ask Spread by One-Minute Intervals
One-minute log price change standard deviation, trading volume mean, and bid-ask spread mean are reported and compared for announcement (major 8:30 A.M.) and
nonannouncement days for the five-year treasury note.' All one-minute intervals between 8:25 and 8:37 A.M. are reported. One and two asterisks denote significance
at the .05 and .01 levels, respectively. The period of analysis is August 23, 1993 - August 19, 1994.
8:25-8:26

8:26-8:27

8:27-8:28

8:28-8:29

8:29-8:30

8:30-8:31

8:31-8:32

8:32-8:33

8:33-8:34 8:34-8:35

8:35-8:36. 8:36-8:37

Panel A: Price Volatility
Announcement day"

0.114

0.131

0.116

0.131

0.292

1.486

1.336

0.509

0.558

0.450

0.292

0.295

Nonannouncement day'

0.109

0.116

0.090

0.110

0.103

0.118

0.127

0.088

0.103

0.119

0.102

Standard Deviation Ratio

1.041

1.122

1.279

1.190

2.841**

5.800**

5.425**

3.796**

2.867**

2.893**

F-ratio p-value•

0.914

0.715

0.296

0.389

0.008

0.001

0.002

0.001

12.557** 10.556**
0.001

Panel B: Trading Volume
Announcement day

1.256

1.156

1.067

1.019

0.919

1.112

1.074

2.981

3.372

4.698

3.940

5.026

Nonannouncement day

1.340

1.183

1.609

1.370

1.066

1.359

1.584

1.320

1.459

1.159

1.247

1.379

Difference in Means

-0.084

-0.027

-0.542

-0.351

-0.148

-0.248

-0.509

1.662**

1.913**

3.538**

2.693**

3.647**

I-statistic p-value'

0.809

0.926

0.073

0.193

0.497

0.546

0.071

0.001

Panel C: Bid-Ask Spread
Announcement day

1.998

1.903

2.095

2.210

2.691

7.343

5.469

2.945

2.410

1.654

1.757

1.731

Nonannouncement day

1.627

1.590

1.633

1.490

1.453

1.489

1.463

1.322

1.501

1.563

1.549

1.578

Difference in Means

0.371

0.313

0.462*

0.720**

1.239**

5.854**

4.006**

1.623**

0.909

0.091

0.208

0.152

I-statistic p-value

0.087

0.096

0.016

0.001

0.055

0.573

0.209

0.283

• The reported Jog price change standard deviation is the actual standard deviation times Io'. Trading volume is reported in tens of millions of U.S. dollars. The reported bid-ask spread

is the actual spread limes IO'.
• Announcement days are defined as days with at least one of the following announcements: Consumer Price Index, Employmen~ Gross Domestic Product, Producer Price Index, Retail
Sales. These are the 8:30 AM. announcements significant at the .01 level in at least two models of Table II. Excluded are days in which there are any of our 9:15 A.M. or 10:00 A.M.
announcements.
t Nonannouncement days are defined as days in which none of our eighteen morning announcements occur.
d P-value from Brown-Forsythe-modified Levene F-statistic comparing variances for announcement and nonannouncement days.
c P-value from t-statistic comparing means for announcement and nonannouncement days assuming unequal variances.

Table IV

Persistence of Price Volatility, Trading Volume, and Bid-Ask Spread by Five-Minute Intervals
Five-minute log price change standard deviation, trading volume mean, and bid-ask spread mean are reported and compared for announcement (major 8:30 A.M.)
and nonannouncement days for the five-year treasury note.' All five-minute intervals between 8: 15 and 8:45 A.M. are reported as weU as intervals from every fifteen
minutes between 9:00 and 10:30 A.M.. One and two asterisks denote significance at the .05 and .Ol levels, respectively. The period of analysis is August 23, 1993 August 19, 1994.
8:15-8:20

8:20-8:25

8:25-8:30

8:30-8:35

8:35-8:40 8:40-8:45

9:00-9:05

9:15-9:20

9:30-9:35

9:45-9:50 10:00-10:05 10:15-10:20

Panel A: Price Volatility
Announcement day"

0.221

0.215

0.385

2.387

0.922

0.506

0.408

0.337

0.227

0.273

0.358

0.234

Nonannouncement day"

0.209

0.267

0.219

0.212

0.239

0.219

0.179

0.197

0.179

0.178

0.197

0.249

Std. Deviation Ratio

1.055

0.804

1.754

11.284**

3.865**

2.312**

2.276**

1.710**

1.265*

1.534**

1.822**

0.941

F-ratio p-value'

0.967

0.237

0.060

0.001

0.038

0.003

0.009

0.353

Panel B: Trading Volume
Announcement day

4.174

5.493

5.416

13.237

19.405

15.767

13.309

10.402

9.353

8.658

8.705

8.047

Nonannouncement day

3.224

5.702

6.567

6.881

7.063

6.817

5.064

5.187

5.791

5.559

5.917

6.013

Difference in Means

0.950

-0.209

-1.151

6.356**

8.245**

5.215**

3.563**

3.099**

2.787**

2.034*

I-statistic p-value'

0.176

0.757

0.098

0.001

0.005

0.003

0.035

12.342** 8.950 ••
0.001

0.001

Panel C: Bid-Ask Spread
Announcement day

1.773

1.832

2.160

3.313

1.706

1.487

1.614

1.479

1.442

1.343

1.573

1.456

Nonannouncement day

1.754

1.636

1.550

1.478

1.470

1.404

1.471

1.424

1.376

1.300

1.237

1.335

Difference in Means

0.019

0.196

0.610**

1.836**

0.236*

0.083

0.144

0.055

0.066

0.043

0.336**

0.121

I-statistic p-value

0.902

0.133

0.001

0.031

0.394

0.151

0.472

0.458

0.588

0.001

0.177

• The reported log price change standard deviation is the actual standard deviation times Io'. Trading volume is reported in tens of millions of U.S. dollars. The reported bid-ask spread

is the actual spread times to'.
• Announcement days are defined as days with at least one of the following announcemeots: Consumer Price Index, Employmeo~ Gross Domestic Product, Producer Price Index, Retail
Sales. These are the 8:30 A.M. announcemeots significant at the .01 level in at least two models ofTable II. Excluded are days in which there are any of our 9: 15 A.M. or 10:00 A.M.
announcements.
c Nonannouncement days are defined as days in which none of our eighteen morning announcements occur.
d P-value from Brown-Forsythe-modified Levene F-statistic comparing variances for announcement and nonannouncement days.
e P-vallle from t-statistic comparing means for announcement and nonannouncement days assuming unequal variances.

Table V

Correlations of Price Volatility and Trading Volume
Correlations of price volatility (absolute value of log price change) and trading volume for the five-year treasury note. All one-minute intervals between 8:25 and 8:49
A.M. are reported. One and two asterisks denote significance at the .05 and .01 levels, respectively. The period of analysis is August 23, 1993 - August 19, 1994.

Correlation
p-value

8:25-8:26

8:26-8:27

8:27-8:28

8:28-8:29

8:29-8:30

8:30-8:31

8:31-8:32

8:32-8:33

8:33-8:34

8:34-8:35

8:35-8:36

8:36-8:37

0.223**

0.189**

0.169*

0.195**

0.229**

0.049

-0.072

0.348**

0.298**

0.117

0.261**

0.169*

0.001

0.006

0.013

0.004

0.001

0.474

0.282

0.001

0.079

0.001

0.011

8:37-8:38

8:38-8:39

8:39-8:40

8:40-8:41

8:41-8:42

8:42-8:43

8:43-8:44

8:44-8:45

8:45-8:46

8:46-8:47

8:47-8:48

8:48-8:49

0.221**

0.234**

0.283**

0.107

0.156*

0.274**

0.251**

0.270**

0.344**

0.116

0.195**

0.139*

0.001

0.106

0.019

0.001

0.083

0.004

0.037

Table VI

Granger Causality Tests of Trading Volume and Price Volatility
Results from Granger causality tests of trading volume and price volatility for the five-year treasury note. Tests are conducted for New York trading hours (7:30 A.M.•
5:00 P.M.) using six lags of both price volatility and trading volume. Variables are measured in five-minute intervals with price volatility defined as the absolute value
of the log price change times 1o' and trading volume measured in tens of millions of U.S. dollars. One and two asterisks denote significance at the .05 and .OJ levels,
respectively. The period of analysis is August 23, 1993 - August 19, 1994.

Question

Dependent variable

Lagged variables

Sum of coefficients

F-statistic

p-value

Is price volatility persistent?

price volatility

0.472

306.219**

0.001

Is trading volume persistent?

trading volume

0.619

852.323**

0.001

Does price volatility cause trading volume?

trading volume

price volatility

6.501

183.752**

0.001

Does trading volume cause price volatility?

price volatility

trading volume

0.003

18.578••

0.001

Figure 1A
lntraday Price Volatility for the Five-Year Treasury Note
Standard deviation of log price changes by five-minute interval from August 23, 1993 - August 19, 1994. The standard deviation
equals the actual standard deviation times 1000 and times shown are interval start times eastern time (ET).

1~-------------------------------~
0.8

~0.6
~
0

>(I)
0

0.4

0.2

L-J,..L.U..U...L.U..u..J..1..U...U.U..U...L.U.....................u..u.............................u..u...........L.U..u..J..1..U..U...........L.U..u..J..I..U...U.U..U...L.U.....................U..U.U..U...L.U..u.J..l..U..U---'

8:00

9:00

10:00

11:00

12:00
1:00
Time-of-Day

2:00

3:00

4:00

5:00

Figure 1B
lntraday Price Volatility on Announcement and Nonannouncement Days
Standard deviation of log price changes for the five-year treasury note for days with at least one of the nineteen announcements
listed in Table I and days with none of these announcements. The standard deviation equals the actual standard deviation times
1000, the period of analysis is August 23, 1993 -August 19, 1994, and times shown are interval start times (ET).

1.2 ~ - - - , - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,

1
/

Announcement Days

0.8
~
~

~ 0.6
~
·c:

a..

Nonannouncement Days

0.4

JI
•:~ •'•.
...
-~.--- : A .. ~.

0.2

• •

711,. ....;;;;_

0 I

I 11

8:00

I I I ,/ 1' I I I

9:00

-·

I I I I t

I 'I

I It

10:00

I I I I

I ' II I

11:00

ti I I

1I I I

I 'I

-A

11=
:

1 JI I I

12:00
1:00
Time-of-Day

I If

I "

2:00

I 'I

t 11

I 'I

3:00

I 11

I " / II I l '

4:00

1/ I If

I I

5:00

Figure 2A

lntraday Trading Volume for the Five-Vear Treasury Note
Mean interdealer trading volume by five-minute interval from August 23, 1993 - August 19, 1994. Trading volume is reported in
tens of millions of U.S. dollars and times shown are interval start times (ET).

12 ~ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,

8
(I)

§ 6
~
4

0 .........J...J..LJ...U....U..U.........................................................................................................................................................................................................................................................
4:00
9:00
11:00
12:00
1:00
2:00
3:00
10:00
5:00
8:00
Time-of-Day

Figure 2B
lntraday Trading Volume on Announcement and Nonannouncement Days
Mean interdealer trading volume for the five-year treasury note for days with at least one of the nineteen announcements listed in
Table I and days with none of these announcements. Trading volume is reported in tens of millions of U.S. dollars, the period of
analysis is August 23, 1993 -August 19, 1994, and times shown are interval start times (ET).

Announcement Days

't••\ ~·

vv .
., :.. ....... ,~ ... . ♦ -; \.. :. .
:.

E
::,

\I i
•

• '

.• : ,: . '

• • • • .. ..
•

:-. '

.",/'-;

....

·-.~

• • • •y·

\.• \:
:;

: •

'....

'"'•

Nonannouncement Days

.....

•...: 'i\
♦

.......- - ~.......- - ~.......--...._.................................................u.u........u.w..u.....u.u........u.w..u...1..u.u.1.1.L..u..u.~
8:00
9:00
10:00
11:00
12:00
1:00
2:00
3:00
4:00
5:00

Time-of-Day

Figure 38

lntraday Bid-Ask Spread on Announcement and Nonannouncement Days
Mean interdealer bid-ask spread for the five-year treasury note for days with at least one of our nineteen announcements listed
in Table I and days with none of these announcements. The spread is measured in hundredths of one percent, the period of
analysis is August 23, 1993 - August 19, 1994, and times shown are interval start times (ET).

2.6 . . . - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,
2.4

2.2

...•. .•....•.
••

•••
••
••
•

:•

[ 1.8

en
1.6

Announcement Days

••
••
t.,

-•'•'••

••
••
••
••
••
••
••
••
•

Nonannouncement Days
••

........

.•. .•.•. - . ..
\:. ""/1..:'
r·.:~• •..,, •
.
.'...
.
..• .. ... ....

•

1.4

1.2

1 .................................8:00

•

t •

•

..•!•.
••

f~ .· ·,
.'/...,

•

........................._ .......................__......................._.............................................................................................................................
9:00
10:00
11:00
12:00
3:00
4:00
1:00
2:00
5:00
Time-of-Day

Figure 3A
lntraday Bid-Ask Spread for the Five-Year Treasury Note
Mean interdealer bid-ask spread by five-minute interval from August 23, 1993 -August 19, 1994. The spread is measured in
hundredths of one percent and times shown are interval start times (ET).

2.6 , - - - - - - - - - - - - - - - - , - - - - - - - - - - - - - - - - - - - - - - ,

2.4

2.2

"O
<G

I!!
~

1.8

1.6

1.4

1.2

L...LJ..U....U...U.J..U..U...U..U.U.J..U..U...U..U...U..U...U...U..J..U..U...U..U...U.J..U..U...U..U...U..U....U...U.J..U..U...U..U...U..U...U...U..U...U..U...U...U..J..U..U...U..U.U.J..U..U...U.~

8:00.

9:00

10:00

11:00

12:00
1:00
Time-of-Day

2:00

3:00

4:00

5:00

Figure 4A
Persistence of Price Volatility After Major 8:30 AM Announcements
Ratio of the standard deviation of log price. changes on major 8:30 AM announcement days to nonannouncement days for the
five-year treasury note. Ratios significant at the five percent level or better are indicated by diamonds. The period of analysis is
August 23, 1993 -August 19, 1994 and times shown are interval start times (ET).

12 . - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,

T
!

10 t-

llJi

u
<a
!I

"')

h
N
~ '[

I-

~ 'S

•"!
I"1

5% significance
◊
1% significance

t·1

1ii 6

1,,I

•

!l'}I

1,
'''
i ' ~

I-

I\ ♦
2

1, ♦. :··. .&

I-

J\
t

:'!Ir

I ~\-~'"""=·
... '•""

t"'"x:-i " i '

v V
,

8:00

'vC

, I \ ' ' ' I I,

9:00

''
II I

;

10:00

•

\7-.""'
·

~ t\,·.~..
5,

•·

g_f+,. -· ~n.

·v·AA~-

11:00

1:00
12:00
Time-of-Day

I," " I

2:00

,1 ..... 1 .....

3:00

,l .. ,,,J,,,,,

I
I

~j,,~,,,~
.. ""'cl\.,,
v'

C,,,,

4:00

5:00

Figure 4B
Persistence of Trading Volume After Major 8:30 AM Announcements
Difference in mean interdealer trading volume between major 8:30 AM announcement days and nonannouncement days for the
five-year treasury note. Differences significant at the five percent level or better are indicated by diamonds. The volume
difference is reported in tens of millions of U.S. dollars, the period of analysis is August 23, 1993-August 19, 1994, and times
shown are interval start times (ET).
14 . . - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

*''

·11

:'f

1+
I \n
' ♦I
I

8
Q)

:,I ll

••

5% significance

t \1

,
'!';_if

,....
½~
V

I /,,i'"'\r:l
'•

♦

- ;

t
~N~♦'¾k..t,
·

+,),
.•
, , ,'
i • •

. ';/
V_,i

♦

~•

l<U~
vi/5 ,

~i 1'1 1
"?I .,
,,

·,

◊
1% significance

•·

'. l.

••I

!f • !
,•

.
.••
\/

••

,,
(I
'
V
,
,,
•
v
>
W

-2

f [I I

t ! ! f I I

8:00

f t I I \

I I I I I

9:00

t I I I I I I t I I I

10:00

I I I I I

I 1I

I I I I I I I t l I I I t I I I I I I I I I I I l I I I I 1I I I I I I I I t I ! I I I I 1I I I I I I I I I I I I I / f I I ( I I I I I I

11:00

12:00
1:00
Time-of-Day

2:00

3:00

4:00

5:00

Figure 4C
Persistence of Bid-Ask Spread After Major 8:30 AM Announcements
Difference in mean interdealer bid-ask spread between major 8:30 AM announcement days and nonannouncement days for the
five-year treasury note. Differences significant at the five percent level or better are indicated by diamonds. The spread
difference is measured in hundredths of one percent, the period of analysis is August 23, 1993 -August 19, 1994, and times
shown are interval start times (ET).

2.--------,,-------------------------------,
5% significance 1% significance
◊
♦

T
I

1.5

!·

I
!

·,'!

·;

,lI :...'"~ , ♦

u, 0.5

· <j?

l ·t,.
l

.n ~

.•.

~1 t··/ i
1'L, 1v;
~

'.'.''°'

~
";/v•

l. 'I

r· \\,J \

-0.5
8:00

9:00

10:00

11:00

12:00
1:00
Time-of-Day

2:00

3:00

4:00

5:00

•

Figure 5
Market Response to August 5, 1994 Employment Report
Mean interdealer bids, asks, and transaction prices, and interdealer trading volume for the five-year treasury note by one-minute
interval between 8:00 AM and 10:00 AM on August 5, 1994. Trading volume is reported in tens of millions of U.S. dollars and
times shown are interval start times (ET).

1"00.6

rt~~~J

100.4
Ask

(I)

.g 100.2
Q.

Trade

Bid

i J.,..
.. 'T... F+.

...

100

41°.£.T

1 '
-tD"
lv.u......+

99.8
16
(1)

..2

r.,.

..;:!!'t'

++ .P.--+oJ.1• •+.._

~+1"
JfT

~~
-+r u-1•1::.,,1•~

...,_......,.&.

r----:---;----------------

12
8
4

o ~l ,1,,, ,..,,I
8:00

8:15

8:30

8:45

9:00
Time-of-Day

9:15

9:30

9:45

10:00

Figure 6
Bid-Ask Spread After Major 8:30 AM Announcements
Actual and predicted differences in mean interdealer bid-ask spread between major 8:30 AM announcement days and
nonannouncement days for the five-year treasury note. The spread is measured in hundredths of one percent, the period of
analysis is August 23, 1993 - August 19, 1994, and times shown are interval start times (ET).

:~•:

...: ·:
1.5

1
«I

. :
. :
. :

/ Predicted Difference
\

0.5

••
••
•

· ·-~

.._.-c, ••

~
0

•

0
··········
o·' ........_•••········•-•~ ~
• ~

••••

•••••• ,-,.

Actual Difference

'::....,"'--~••"' •

\!T'.

•• -·· ••

•••••

••• • -•

•
......
·-..
..... --

•,....

••
A,•r-i - ~Jiiilrt.r....
~-Q.
•,

..•

•

••

'·-....:

-0.5
8:00

•
.··~... ...:

·•.-·

8:30

••

9:00

""""'- Predicted Difference with Usual Price Volatility

9:30

10:00
Time-of-Day

10:30

11:00

11:30

12:00

FEDERAL RESERVE BANK OF NEW YORK
RESEARCH PAPERS
1996
•
The following papers were written by economists at the Federal Reserve Bank of
New York either alone or in collaboration with outside economists. Single copies of up to six
papers are available upon request from the Public Information Department, Federal
Reserve Bank of New York, 33 Liberty Street, New York, NY 10045-0001 (212) 720-6134.

9601. Bartolini, Leonardo, and Gordon M. Bodnar. "Are Exchange Rates Excessively Volatile?
And What Does 'Excessively Volatile' Mean, Anyway?" January 1996.
9602. Lopez, Jose A. "Exchange Rate Cointegration Across Central Bank Regime Shifts."
January 1996.
9603. Wenninger, John, and Daniel Orlow. "Consumer Payments Over Open Computer
Networks." March 1996.
9604. Groshen, Erica L. "American Employer Salary Surveys and Labor Economics Research:
Issues and Contributions." March 1996.
9605. Uctum, Merih. "European Integration and Asymmetry in the EMS." April 1996.
9606. de Kock, Gabriel S. P., and Tanya E. Ghaleb. "Has the Cost of Fighting Inflation Fallen?"
April 1996.
9607. de Kock, Gabriel S. P., and Tania Nadal-Vicens. "Capacity Utilization-Inflation Linkages:
A Cross-Country Analysis." April 1996.
9608. Cantor, Richard, and Frank Packer. "Determinants and Impacts of Sovereign Credit
Ratings." April 1996.
9609. Estrella, Arturo, and Frederic S. Mishkin. "Predicting U.S. Recessions: Financial
Variables as Leading Indicators." May 1996.
9610. Antzoulatos, Angelos A. "Capital Flows and Current Account Deficits in the 1990s: Why
Did Latin American and East Asian Countries Respond Differently?" May 1996.

9611. Locke, Peter R., Asani Sarkar, and Lifan Wu. "Did the Good Guys Lose? Heterogeneous
Traders and Regulatory Restrictions on Dual Trading." May 1996.
9612. Locke, Peter R., and Asani Sarkar. "Volatility and Liquidity in Futures Markets.•
May 1996.
9613. Gong, Frank F., and Eli M. Remolona. "Two Factors Along the Yield Curve." June 1996.
9614. Harris, Ethan S., and Clara Vega. "What Do Chain Store Sales Tell Us About Consumer
Spending?" June 1996.
9615. Uctum, Merih, and Michael Wickens. "Debt and Deficit Ceilings, and Sustainability of
Fiscal Policies: An Intertemporal Analysis." June 1996.
9616. Uctum, Merih, and Michael Aglietta. "Europe and the Maastricht Challenge." June 1996.
9617. Laster, David, Paul Bennett, and In Sun Geoum. "Rational Bias in Macroeconomic
Forecasts." July 1996.
9618. Mahoney, James M., Chamu Sundaramurthy, and Joseph T. Mahoney. "The Effects of
Corporate Antitakeover Provisions on Long-Term Investment: Empirical Evidence."
July 1996.
9619. Gong, Frank F., and Eli M. Remolona. "A Three-Factor.Econometric Model of the U.S.
Term Structure." July 1996.
9620. Nolle, Daniel E., and Rama Seth. "Do Banks Follow Their Customers Abroad?"
July 1996.
9621. McCarthy, Jonathan, and Charles Steindel. "The Relative Importance of National and
Regional Factors in the New York Metropolitan Economy." July 1996.
9622. Peristiani, S., P. Bennett, G. Monsen, R. Peach, and J. Raiff. "Effects of Household
Creditworthiness on Mortgage Refinancings." August 1996.
9623. Peristiani, Stavros. "Do Mergers lmprove the X-Efficiency and Scale Efficiency of U.S.
Banks? Evidence from the 1980s." August 1996.
9624. Ludvigson, Sydney. "Consumption and Credit: A Model of Time-Varying Liquidity
Constraints." August 1996.

9625. Ludvigson, Sydney. "The Channel of Monetary Transmission to Demand: Evidence from
the Market for Automobile Credit." August 1996.
9626. Sobol, Dorothy M. "Central and Eastern Europe: Financial Markets and Private Capital
Flows." August 1996.
9627. Evans, Joan, and James M. Mahoney. "The Effects of Daily Price Limits on Cotton Futures
and Options Trading." August 1996.
9628. Molyneux, Philip, and Rama Seth. "Foreign Banks, Profits and Commercial Credit
Extension in the United States." August 1996.
9629. Cantor, Richard, and Robert Driskill. "Can a Fiscal Contraction Strengthen a Currency?
Some Doubts About Conventional Mundell-Fleming Results." August 1996.
9630. Jayaratne, Jith, and Philip E. Strahan. "Entry Restrictions, Industry Evolution and Dynamic
Efficiency: Evidence from Commercial Banking." August 1996.
9631. Dziwura, Joseph R., and Eric M. Green. "Interest Rate Expectations and the Shape of the
Yield Curve." September 1996.
9632. Brewer, Elijah III, and Marc R. Saidenberg. "Franchise Value, Ownership Structure, and
Risk at Savings Institutions." September 1996.

To obtain more information about the Bank's RESEARCH PAPERS series and other
publications, visit the Federal Reserve Bank of New York's site on the World Wide Web
(http://www.ny.frb.org). From the research publications page, you can view abstracts of
RESEARCH PAPERS and STAFF REPORTS and order the full-length, hard-copy versions of
them electronically. Interested readers can also view and download articles from CURRENT
ISSUES IN ECONOMICS AND FINANCE and the ECONOMIC POLICY REVIEW.

Full text of Research Papers (Federal Reserve Bank of New York) : Price Formation and Liquidity in the U.S. Treasuries Market : Evidence from Intraday Patterns around Announcements, Research Paper 9633

FRASER