price discovery in the treasury futures marketmbrandt/papers/published/cti.pdfwiley periodicals,...

PRICE DISCOVERY

IN THE TREASURY

FUTURES MARKET

MICHAEL W. BRANDT*KENNETH A. KAVAJECZSHANE E. UNDERWOOD

The paper conducts a regression analysis utilizing both futures and cashmarket prices and net orderflow to determine where price discovery takesplace as well as the forces at play that influence the location. Specifically,given the strong theoretical linkage between the U.S. Treasury cash andfutures markets, they compare how orderflow contributes to price discov-ery and analyze how and when information flows from one market to theother. How a number of environmental variables (trader type, financingrates, and liquidity) impact the information flows between these two mar-kets is also considered. Their findings provide new evidence on the extentto which price discovery happens away from a primary market. © 2007Wiley Periodicals, Inc. Jrl Fut Mark 27: 1021–1051, 2007

For helpful comments, we thank seminar participants at Babson College and the Texas JuniorFinance Symposium as well as Michael Goldstein, Jim Hodder, Elizabeth Odders-White, NataliaPiqueira, and Mark Ready. In addition, we thank Brian Huber of Alpha Logic Technologies for pro-viding and explaining their data. All remaining errors are our own.

*Correspondence author, Fuqua School of Business, Duke University, One Towerview Drive,Durham, NC 27708. E-mail: [email protected]

Received November 2006; Accepted February 2007

� Michael W. Brandt is an Associate Professor of Finance at the Fuqua School ofBusiness of Duke University and is also affiliated with the NBER.

� Kenneth A. Kavajecz is an Associate Professor of Finance at the School of Business ofthe University of Wisconsin-Madison.

� Shane E. Underwood is an Assistant Professor of Management at the Jesse H. JonesGraduate School of Management of Rice University.

The Journal of Futures Markets, Vol. 27, No. 11, 1021–1051 (2007)© 2007 Wiley Periodicals, Inc.Published online in Wiley Interscience (www.interscience.wiley.com). DOI: 10:1002/fut.20275

1022 Brandt, Kavajecz, and Underwood

Journal of Futures Markets DOI: 10.1002/fut

INTRODUCTION

There are many ways that market participants can obtain exposure to themovement in riskless interest rates. One obvious way is through the cashmarket for U.S. Treasury securities. There has been a substantial litera-ture that investigates the inner workings of this market, includingFleming and Remolona (1997, 1999), Brandt and Kavajecz (2004), andGreen (2004). Another important avenue market participants use to gainexposure to riskless fixed income securities is through the tradingof futures on U.S. Treasury securities. The sheer size of the Treasuryfutures market foreshadows its importance; for example, the 10-yearTreasury futures contract traded on the Chicago Board of Trade (CBOT)has in excess of 100 billion dollars in notional volume traded every day.Financial theory suggests that these markets should be connectedthrough the diligent and repeated application of arbitrage strategies aswell as trades used to satisfy hedging demands. Yet despite the impor-tance of the Treasury futures market, there have been only a handful ofstudies dealing with the potential linkages between the U.S. Treasuryfutures and cash markets (Simpson & Ireland, 1985).

The goal of this study is to better understand the movements of theterm structure of interest rates through trading of both Treasury cashsecurities and Treasury futures contracts. Specifically, the followingquestions are of interest. First, are the mechanisms through which pricediscovery takes place the same in the U.S. Treasury futures and cashmarkets, i.e., does orderflow have the same impact in both markets?Second, does the primary location for price discovery switch between thecash and futures market? If so, when does that occur, and what environ-mental forces influence the transition?

This study fits in the intersection of three segments of the finance lit-erature and thereby provides a synthesis of many of the ideas therein.First, there is mounting evidence that aggregate orderflow is the mecha-nism through which information is impounded into prices in numerousmarkets. Examples of work in this area include Glosten and Milgrom(1985) and Kyle (1985) in the equity markets, and more recently Evansand Lyons (2002a, 2002b) and Rosenberg and Traub (2006) in the for-eign exchange market and Fleming and Remolona (1997, 1999) andBrandt and Kavajecz (2004) in the fixed income market. Second, there isa growing literature that investigates the connection between separate,but related markets. Examples include Simpson and Ireland (1985),Easley, O’Hara, and Srinivas (1998), Anderson, Bollerslev, Diebold, andVega (2004) and Underwood (2006). This study stands at the intersection

Price Discovery 1023


of these two strands of the literature, in that whether the relationbetween orderflow and prices in the Treasury futures market is similar to,or different from, the analogous relation in the Treasury cash market isinvestigated.

Lastly, conditional on orderflow being the primary source for pricediscovery, it is natural to question what forces have an impact on theextent and direction of price discovery across markets. For example,some researchers have considered how orderflow from different tradertypes impacts the extent of price discovery, e.g., Keim and Madhavan(1995, 1997) and He (2005) consider institutional and retail orderflow,whereas Manaster and Mann (1996) compare the general public andexchange members. Other researchers have investigated the impact ofliquidity on price discovery (see, for example, Brandt & Kavajecz, 2004).In the context of this analysis the impact of a number of environmentalvariables—including trader type, financing rates, and liquidity—on pricediscovery between the two markets is considered.

This article shares this intersection of the finance literature withtwo recent articles. Although these studies are similar to this investiga-tion, the focus of each of these studies is distinct. Menkveld, Sarkar, andvan der Wel (2006) focus on the informativeness of customer order flowin Treasury bond futures following macroeconomic announcements.Although they provide evidence on the time-varying importance of cus-tomer order flow in the futures market, they do not consider the interac-tion of futures and cash orderflow as is done here. Mizrach and Neely(2006) focus on the relative contributions of the Treasury cash andfutures markets to intraday price discovery unconditionally. They showthat the 5-year cash note and the 30-year futures contract contribute themost to price discovery in the Treasury market. Although the uncondi-tional impact of orderflow on price discovery in the two markets is alsoinvestigated here, the differing roles played by the two markets dependingon Repo financing rates and cash market liquidity is examined as well.

The analysis begins by comparing the direct impact that net order-flow has in the two markets. The results on the change in open interestreveals that traders in the futures market typically take a position on theslope of the yield curve, e.g., buy the 5-year contract and sell the 30-yearcontract. When futures net orderflow is disaggregated by trader type,trades originating from locals (i.e., market makers occupying the futurespits) were found to have almost no impact on daily returns, whereastrades originating from exchange members and retail customers con-tribute substantially to price discovery. That said, although exchange



members and retail trades have a permanent impact on prices, theirimpacts are in opposing directions. This study’s results show that retailtrades move prices in the direction of their trade (buy trades push pricesup and sell trades push prices down), whereas prices move opposite thedirection of exchange member trades. Given that much of the trading byexchange members is done to satisfy hedging demands, their trades areconsistent with muted, and even opposing movements, in futures prices.Consistent with our intuition, net orderflow is also important to price dis-covery in the cash market. Of particular importance toward price move-ments in all maturities are the on-the-run trades of the 2- and 5-yearnotes. In contrast, trading in the 10- and 30-year securities are far lessinformative.

The paper conducts a regression analysis utilizing both futures andcash market prices and net orderflow to determine where price discoverytakes place as well as the forces at play that influence the location.The results show that net orderflow in each market has a significantimpact on the prices in both the Treasury cash and futures markets. Theresults using own market orderflow are shown to be robust to condition-ing on orderflow variables from the cross market. Specifically, the 2- and5-year cash market net orderflow is significantly positively related to bothcash and futures price movements. The positive sign and strictly increas-ing size of the coefficients with respect to maturity suggests that cashpurchases of the 2- and 5-year note shift up the price level along theentire curve and in addition, steepen the slope. In contrast, hedging-related trades conducted by futures exchange members on the long endof the curve (5-, 10-, and 30-year contracts) have a significant negativeimpact and are strictly decreasing with respect to maturity. Thus, thesetrades shift down prices along the entire curve and also flatten the slope.Retail futures trades have a positive and significant impact on both cashand futures prices. The price reaction to retail futures trades is to shiftup and steepen prices along the yield curve similar to the reaction toshort maturity (2- and 5-year) cash orderflow; however, the magnitude ofthe reaction is much more muted for retail orderflow. In addition to trad-er type, financing rates and the level of liquidity in a market are found toimpact the extent and direction of price discovery. The results found hereshow that Repo specialness causes price discovery to occur relatively morein the futures market because transacting in the cash market is exceed-ingly expensive. Moreover, illiquidity in the cash market is associatedwith relatively more price discovery occurring in the cash marketas informed traders are likely transacting in that market. Interestingly,the instrument of central importance to the above results is the 5-year



maturity in both the futures and cash market, largely to the exclusion ofthe other maturities.

PRELIMINARIES

Data Sources

The data used in this analysis are daily time-series of price and orderflowfor both the futures and cash market over the period 1995 through 2000.1

The futures data are obtained from two different sources. The futurestransaction price and volume data are obtained from the Institute forFinancial Markets (Washington, DC). Daily closing prices and daily vol-ume information were used to construct a time-series of activity in eachof the 2-, 5-, 10-, and 30-year Treasury futures contracts. The futuresorderflow data are obtained from the Chicago Board of Trade’s LiquidityData Bank (hereafter LDB). These data provide net order flow (buy vol-ume less sell volume) differentiated by trader type for all CBOT futurescontracts at the daily frequency (beginning in 1995) as well as intraday(beginning in January 2003). Specifically, the LDB data assigns trades toone of four customer trade indicator (CTI) categories. The CBOT pro-vides the following descriptions of the four categories: Category 1 con-sists of trades executed by individual CBOT members trading for theirown accounts, Category 2 is made up of trades that are executed byCBOT clearing member firms trading for their house accounts, Category 3comprises trades executed by CBOT members filling orders for otherCBOT members, and Category 4 consists of trades filled for the public orany other type of “outside” customers. Daigler and Wiley (1999) providea nice summary of the different categories and the likely type of tradersthat are included in each. They suggest that Category 1 is largely popu-lated by market makers, “scalpers,” or “locals,” whose usual function isto provide liquidity and earn a small profit from the bid-ask spread.Category 2 appears to represent proprietary trading by clearing memberfirms, which may be conducted either for speculative or hedging purpos-es. The fact that Category 3 represents trades done on behalf of otherCBOT members suggests that hedging is a likely motive. For example, aCategory 3 trade might be a trade executed for a CBOT options traderwishing to hedge a position. Manaster and Mann (1996) and Fergusonand Mann (1999) also provide evidence on Categories 2 and 3 trading

1We end our sample in December 2000 because certain variables that allow us to calculate cashmarket order flow are missing from the GovPX dataset starting in early 2001.



for a variety of futures contracts that trade on the Chicago MercantileExchange. In general, they find that the bulk of Categories 2 and 3 volumeis related to hedging as many trades are part of a delta-hedging strategyfor options positions. Given the results of Manaster and Mann (1996)and in an effort to simplify this analysis, Categories 2 and 3 orderflowwere combined into one category. Lastly, Category 4 is typically referredto as the general public or retail orderflow, although in some cases smallinstitutional traders may fall into this category as well.

The Treasury cash data are obtained from GovPX (www.govpx.com),which consolidates quotes and transaction data from five of the six majorinterdealer Treasury securities brokers in the secondary U.S. Treasurycash market. Specifically, intraday U.S. Treasury security quotes andtransactions were used for all Treasury issues. The GovPX dataset con-tains security identifier information, including the CUSIP (Committeeon Uniform Security Identification Procedures, Standard & Poor’s, NewYork, NY), coupon, maturity date, as well as an indicator of whether thesecurity is trading when-issued, on-the-run, or off-the-run. The quotedata contain the best bid and ask prices, associated yields, and respectivebid and ask depths, all time-stamped to the nearest second. The transac-tion data include the time, initiator (i.e., signed trades), price and quan-tity. These data allow the calculation of the price, net orderflow, and liq-uidity measures for the Treasury cash market.

The basic futures and cash time-series data was supplemented withinformation regarding other pertinent aspects of the Treasury market, suchas the timing of Treasury auctions, macroeconomic announcements andfinancing (Repo) rates. For macroeconomic announcements, a listing ofthe date and time of release, the market’s expectation, and the announcedstatistic is obtained from Money Market Services (MMS). Data onannouncements of the latest consumer price index (CPI), producer priceindex (PPI), housing starts, civilian unemployment, nonfarm payroll, retailsales, industrial production, consumer confidence, the NAPM (nowInstitute for Supply Management, Tempe, AZ) reports, as well as on allFOMC (Federal Open Market Committee, Federal Reserve Board,Washington, DC) meetings are used. Financing rates are also obtainedfrom GovPX. Specifically, daily information on the overnight general collat-eral and on-the-run Repo rates for each of the specific maturities is used.

Aggregation, Timing, and Variable Definitions

Treasury futures end-of-day prices and daily net orderflow are takenfrom the nearby contract until the first delivery day, at which point the



series switches over to the second nearby contract. The trading activityin the delivery month is avoided because there may be some settlement-induced illiquidity as in Johnston, Kracaw, and McConnell (1991).Moreover, because the concentration of trading begins to naturally shiftto the next nearby contract, trading in the delivery month becomes lessand less representative of the overall market.

The Treasury cash securities are partitioned, and data aggregated, asin Brandt and Kavajecz (2004). Specifically, each Treasury security is cat-egorized along two dimensions common in the literature, remainingtime-to-maturity and seasonedness; seasonedness refers to how recently asecurity was auctioned. Securities are categorized into one of two groups:on-the-run and off-the-run. This partition will be important later whenfuture delivery options are discussed. The daily Treasury cash price andorderflow series are constructed by aggregating data within nonoverlap-ping periods. Daily orderflow figures are computed using executed ordersbefore 2:00 P.M. (Central time) on a given day and prices are averaged inthe final hours of trading (from 2:00 P.M. to the close). This samplingtechnique is adopted because trading in the Treasury cash market tends tobe concentrated in the morning and also to avoid any nonsynchronousmeasurement of orderflow and prices. The 2:00 P.M. Central time is cho-sen to line up with the close of trading in the futures market.

Any analysis of the connection between an exchange-traded futureand its underlying asset hinges upon the delivery options set out by theexchange. The fact that traders have the ability to deliver any one of awell-defined set of securities to satisfy a futures contract suggests thatthey will naturally avail themselves of the cheapest-to-deliver securitywithin the set of feasible and obtainable securities. Consequently, the setof securities that can be delivered against each futures contract each dayis identified. Within that feasible set of deliverables, i.e., eligible securitiesthat have the proper issuing maturity and remaining time-to-maturity, thesecurities are ordered by their implicit cost measured by the implied reporate (IRR). Identifying the cheapest-to-deliver security involves a numberof calculations. First, the current futures price is multiplied by a conver-sion factor, which converts the futures price to the comparable price atwhich that security would trade if it yielded 6% to first call.2 Multiplyingby the conversion factor puts all securities on a level playing field so thatthey may be compared. Second, the interest that has accrued since thelast interest payment is calculated. This represents a reduction in the cost

2Note that prior to March 2000, the conversion factor was set to convert the price of the security toone yielding 8% to first call.



to delivering the security as the receiver of the bond must pay the delivererfor the interest that has accrued before taking ownership. Lastly, thepercentage return is calculated on a 360-day basis until the first deliveryday, i.e., the implied Repo rate, for each security in the feasible set ofdeliverables.3 Because there are potentially many securities within the setof deliverables, for practicality, the two cheapest-to-deliver issues istracked for each contract for each day. These two cheapest-to-deliversecurities form the estimate of the price of the underlying security. As thefutures contract is rolled over to the next nearby contract at the first deliv-ery day, so too is the cheapest-to-deliver security rolled over at that time.

There are a number of facts about the cheapest-to-deliver securitiesthat are worth highlighting given their relevance to the forthcominganalysis. First, the cheapest-to-deliver securities are rarely the on-the-run issues, in fact, they are always in the off-the-run category during thesample.4 Second, despite the fact that these issues are cheapest-to-deliver,being off-the-run securities implies that these issues will be infrequentlytraded.

Summary Statistics

To appreciate the richness of the data used in this analysis, a number ofsummary tables is presented. Table I displays the average daily futurereturns, open interest and volume for each of the four futures contractsby year as well as the corresponding Treasury cash returns. Futures trad-ing is concentrated in the long and middle part of the curve as seenthrough the elevated figures for open interest and volume in the 10- and30-year contracts and the much lower figures for the 2-year contract.5

No particular year stands out as an anomaly with both futures and cashreturns averaging close to zero and open interest and volume trendingupwards over the sample period.

Table II presents the change in open interest and the futures netorderflow broken out by CTI category for each year in the sample.6 The

3For a more detailed explanation of the cheapest-to-deliver calculation see Burghardt, Belton, Lane,and Papa (1994).4This is partly because the on-the-run security typically trades at a premium relative to similar matu-rity securities. In addition, typical interest rate environments tend to disadvantage new issues interms of being cheapest to deliver. Burghardt et al. (1994) note that the regime of high, but falling,interest rates in the mid-1980s did lead to a role for the on-the-run securities as cheapest-to-deliver.5It is worth noting, however, that the 2-year contract trades based on an underlying security witha face value of $200,000, whereas the 5-, 10-, and 30-year contracts trade on a $100,000 basis.6Net orderflow broken out by category records both sides of a trade; thus the sum of the fourcategories must necessarily be 0.



results generally reflect those shown in Table I: Trading becomes muchmore active the longer the maturity of the futures contract. In addition toreviewing the category averages, it is interesting to consider the correla-tion of activity across trader types. In untabulated results, the correlationbetween categories within a futures contract is relatively low, with thenotable exception of Category 4, which is strongly negatively correlatedwith all other categories. The strong negative correlation with Category 1is to be expected, given traders in that category are mainly providing liq-uidity via market making. However, Categories 2 and 3 net orderflow arealso significantly negatively correlated with Category 4 orderflow, whichsuggests that these traders often take the opposite side of trades executedby Category 4 or retail traders. Moreover, very weak correlations arefound across contracts within a category. Lastly, the change in open inter-est appears to be uncorrelated with any of the specific net orderflow cate-gories, yet is highly positively correlated with the changes in open interestin other futures contracts.

TABLE I

Summary Statistics for U.S. Treasury Futures Contracts

1995 1996 1997 1998 1999 2000 Total

2 YearFutures return 0.0192 �0.0034 0.0042 0.0054 �0.0083 0.0078 0.0000Open interest 23.25 17.04 30.23 38.38 34.95 47.31 31.85Volume 3.46 2.78 4.10 5.71 4.30 6.89 4.54Cash return 0.0013 0.0075 0.0049 0.0025 �0.0064 0.0055 0.0000

5 YearFutures return 0.0398 �0.0116 0.0100 0.0138 �0.0196 0.0222 0.0001Open interest 164.76 149.29 203.84 274.13 267.85 362.73 237.09Volume 75.96 67.27 79.36 105.00 97.73 161.03 97.72Cash return 0.0103 0.0016 �0.0023 �0.0001 �0.0313 0.0181 �0.0000

10 YearFutures return 0.0575 �0.0169 0.0170 0.0220 �0.0307 0.0357 0.0001Open interest 235.09 256.25 318.31 435.77 502.13 527.72 379.23Volume 127.13 123.12 137.82 191.26 198.10 348.02 187.60Cash return 0.0587 �0.0095 0.0022 0.0060 �0.0501 0.0599 0.0001

30 YearFutures return 0.0842 �0.0292 0.0314 0.0298 �0.0492 0.0572 0.0002Open interest 341.66 362.92 486.24 680.41 564.11 417.76 475.39Volume 580.15 564.72 692.05 775.13 610.68 503.33 620.85Cash return 0.0820 �0.0379 0.0308 0.0142 �0.0700 0.0571 0.0001

Note. This table provides the mean of daily values for the futures return, open interest and volume figures for the 2-, 5-,10-, and 30-year Chicago Board of Trade Treasury futures contracts as well as the Treasury on-the-run cash return over theperiod 1995 through 2000. The futures and cash returns are quoted in percent and the open interest and volume figuresare quoted in 1000s of contracts for the nearby future.

TA

BL

E I

I

Fut

ures

Con

trac

ts N

et O

rder

flow

1995

1996

1997

1998

1999

2000

Tota

l

2 Ye

arC

ateg

ory

1�

13.8

4�

2.59

�17

.75

16.0

1�

18.1

3�

4.38

�6.

782�

3�

0.75

15.0

65.

6727

.58

30.3

8�

32.8

67.

514

14.5

9�

12.4

712

.08

�43

.60

�12

.25

37.2

4�

0.73

�in

ope

n in

t.�

103.

19�

61.1

7�

10.4

3�

271.

83�

290.

49�

149.

10�

147.

71C

ash

on-t

he-r

un27

2.82

342.

0526

9.95

273.

7326

8.32

14.1

023

8.90

Cas

h of

f-th

e-ru

n15

0.54

94.0

394

.06

76.4

360

.93

41.1

286

.24

5 Ye

arC

ateg

ory

118

.37

119.

70�

46.8

018

0.60

�25

6.16

�42

4.73

�68

.28

2�3

0.85

63.3

5�

103.

98�

81.9

837

2.12

�37

0.92

�19

.96

4�

19.2

2�

183.

0515

0.79

�98

.82

�11

5.96

795.

6588

.21

�in

ope

n in

t.82

.62

�39

8.85

26.4

8�

1502

.14

�16

62.4

0�

2397

.75

�97

6.70

Cas

h on

-the

-run

413.

8334

4.53

362.

9423

4.95

83.2

129

.54

245.

78C

ash

off-

the-

run

68.7

918

.97

13.8

741

.88

22.2

114

.15

29.9

910

Yea

rC

ateg

ory

118

.64

17.4

081

.33

11.9

0�

27.0

618

9.44

48.6

12�

3�

203.

40�

70.9

5�

152.

0328

.25

769.

95�

797.

31�

70.8

44

184.

7653

.56

70.7

0�

40.1

5�

742.

9260

7.87

22.2

2�

in o

pen

int.

�62

.27

�40

3.92

�10

3.57

�11

54.4

8�

2996

.62

�41

16.9

2�

1475

.29

Cas

h on

-the

-run

162.

2822

7.57

168.

4812

3.95

56.3

913

.69

125.

57C

ash

off-

the-

run

5.98

0.92

2.34

0.27

0.34

0.34

1.47

30 Y

ear

Cat

egor

y 1

�18

0.21

�18

5.40

�12

6.77

241.

73�

158.

69�

253.

01�

110.

622�

3�

111.

21�

133.

02�

158.

991.

9028

2.06

�29

7.11

�69

.40

429

1.42

318.

4228

5.76

�24

3.62

�12

3.40

550.

1218

0.01

�in

ope

n in

t.�

745.

47�

1376

.29

�99

9.59

�56

24.0

5�

4412

.46

�36

34.1

7�

2799

.79

Cas

h on

-the

-run

10.1

738

.38

4.32

21.9

57.

646.

7915

.53

Cas

h of

f-th

e-ru

n0.

293.

18�

0.61

�0.

05�

0.10

0.00

0.46

Not

e.T

his

tabl

e pr

ovid

es t

he n

earb

y co

ntra

ct n

et o

rder

flow

figu

res

for

the

cust

omer

tra

de in

dica

tor

(CT

I) c

ateg

orie

s, c

hang

e in

ope

n in

tere

st a

s w

ell a

s ne

t or

derfl

ow f

or t

he c

ash

on-

and

off-

the-

run

secu

ritie

s.

Cat

egor

y 1

cons

ists

of t

rade

s ex

ecut

ed b

y in

divi

dual

Chi

cago

Boa

rd o

f Tra

de (

CB

OT

) m

embe

rs tr

adin

g fo

r th

eir

own

acco

unts

, Cat

egor

y 2

is m

ade

upof

trad

es th

at a

re e

xecu

ted

by C

BO

Tcl

earin

g m

embe

r fir

ms

trad

ing

for

thei

r ho

use

acco

unts

, Cat

egor

y 3

com

pris

es tr

ades

exe

cute

d by

CB

OT

mem

bers

filli

ng o

rder

s fo

r ot

her

CB

OT

mem

bers

, and

Cat

egor

y 4

cons

ists

of t

rade

s fil

led

for

the

publ

ic o

r an

y ot

her

type

of o

utsi

de c

usto

mer

s. F

igur

es a

re a

vera

ges

of d

aily

val

ues

over

the

resp

ectiv

e ye

ar in

con

trac

t uni

tsfo

r fu

ture

s an

dm

illio

ns o

f dol

lars

for

the

cash

ser

ies.

Not

e th

at th

e C

TI fi

gure

s ar

e do

uble

cou

nted

in th

at e

ach

trad

e is

rec

orde

d fo

r bo

th th

e bu

yer

and

the

selle

r.



Table II also presents average net orderflow statistics for the cashTreasury market. The average net orderflow figures for the cash marketdisplay the usual pattern of a heavy concentration of trade in the on-the-run security with marked reductions in net orderflow in the off-the-runsecurities. Of note is the fact that GovPX net orderflow falls off substan-tially in 2000. Although this is potentially a result of a shift in tradingvenues, it also may result from the decreased issuance of Treasuries dueto budget surpluses at that time.

In other untabulated results, the distribution of trading activity acrossthe CTI categories is considered. The majority of futures trading is donethrough Categories 1 and 4, with the combined trading of Categories2 and 3 accounting for, at most, 25% of traded volume. This result is con-sistent across contracts. Despite the prominence of Categories 1 and 4,their relative importance varies across the contracts. For the 2-year con-tract Categories 1 and 4 account for roughly equal proportions of aver-age daily volume whereas for the 5-, 10-, and 30-year contracts,Category 1 makes up a much larger fraction of trading volume than doesCategory 4. Of all the contracts, the 5-year contact seems to exhibit themost trading in Categories 2 and 3. One possible explanation is that if amajority of fixed income portfolios have an average duration close to5 years, the 5-year futures contract becomes the natural hedge instru-ment. The earlier description of the likely participants encompassingCategories 2 and 3 trades is consistent with this conjecture.

CASH AND FUTURES ORDERFLOW

In this section, the ability of orderflow to explain price movements ineach of the respective markets is explored more directly. The daily returnin the futures (cash) market is regressed on net orderflow variablesfrom the different contracts (maturities). Owing to a strong seasonal“saw tooth” pattern within each of the net orderflow and open interestseries, those series are detrended by regressing each on a constant and atime-to-delivery variable.7 The variables used in the orderflow regres-sions represent the residuals from this detrending procedure. Lastly,after controlling for macroeconomic announcements, the impact oforderflow along the futures curve is estimated and is compared to theimpact of orderflow in the cash market.

7In particular, net orderflow, volume, and open interest are highest immediately after the rollover tothe next nearby contract and then slowly taper off until the first delivery day for the current contract.



Table III shows how changes in open interest impact futures prices.Recall that an increase in open interest simply means that additionalfutures contracts were created. Interestingly, the impact of changes inopen interest in the center of the curve (5- and 10-year contracts) isopposite the impact at the curve’s endpoints (2- and 30-year contacts).Moreover, the significance and uniform sign of the impact across returnsof different maturities for the simple creation of a new contract suggestthat market participants are likely to be executing strategies using multi-ple contract maturities. For example, a slope trade such as buying (sell-ing) a 5-year contract and selling (buying) a 30-year contract would be astrategy consistent with these results.

Table IV presents the results based on the current and lagged netorderflow of the various CTI categories. The results are consistent withthe description of each category given above. Category 1 trades appear tohave little impact on the direction of prices as coefficients are typicallyinsignificantly different from zero. This is consistent with the behavior oflocal market makers whose strategy is to temporarily supply liquidity inreturn for payment of the bid-ask spread. Anecdotally, these market makersdo not take a position in the security, opting instead to “go home flat.”8

TABLE III

Response of Futures Returns to Changes in Open Interest

Change in open interest

Futuresreturn Intercept 2 Year 5 Year 10 Year 30 Year R2

2 Year 0.0306 0.0742 �0.2000 �0.0425 0.2093 0.0230(1.1080) (1.9431) (�3.8973) (�0.9966) (4.6562)

5 Year 0.0799 0.1101 �0.4773 �0.1573 0.5646 0.0292(1.2903) (1.2851) (�4.1433) (�1.6429) (5.5951)

10 Year 0.1304 0.1392 �0.6442 �0.3339 0.9103 0.0339(1.4608) (1.1266) (�3.8802) (�2.4201) (6.2600)

30 Year 0.1914 0.1472 �0.8479 �0.5627 1.5938 0.0372(1.3787) (0.7659) (�3.2830) (�2.6221) (7.0456)

Note. This table presents the response of futures prices to changes in open interest separated by contract maturity. Eachchange in open interest series represents the residuals of a regression of the change in open interest regressed on a con-stant and a time-to-deliver variable. This adjustment accounts for the strong seasonal present in each open interest series.Coefficients are multiplied by 1,000 and t-statistics are shown in parentheses. Bold values represent significance at the5% level.

8Manaster and Mann (1996) examine the behavior of Category 1 traders in a number of contractstrading on the Chicago Mercantile Exchange. They find that although Category 1 traders do takepositions intraday, most of their positions are closed out by the end of the day.

TA

BL

E I

V

Res

pons

e of

Fut

ures

Ret

urns

to

Cur

rent

and

Lag

ged

Cus

tom

er T

rade

Ind

icat

or (

CT

I) C

ateg

ory

Ord

erfl

ow

Ord

erflo

w

2 Ye

ar5

Year

10 Y

ear

30 Y

ear

Fut

ures

retu

rnIn

terc

ept

Cur

rent

Lag

ged

Cur

rent

Lag

ged

Cur

rent

Lag

ged

Cur

rent

Lag

ged

R2

Cat

egor

y 1

2 Ye

ar0.

0416

0.05

050.

0095

�0.

0059

0.01

81�

0.05

320.

0406

�0.

0627

�0.

0307

0.00

64(1

.475

7)(1

.621

6)(0

.306

9)(�

0.14

94)

(0.4

586)

(�1.

3752

)(1

.048

1)(�

1.63

88)

(�0.

7871

)

5 Ye

ar0.

0883

0.29

000.

0542

�0.

0282

0.03

37�

0.11

560.

1019

�0.

1528

�0.

0751

0.01

76(1

.418

0)(4

.217

5)(0

.790

9)(�

0.32

32)

(0.3

875)

(�1.

3528

)(1

.189

7)(�

1.80

87)

(�0.

8703

)

10 Y

ear

0.14

400.

4559

0.06

890.

0607

0.05

74�

0.10

520.

1277

�0.

2693

�0.

1264

0.02

03(1

.604

4)(4

.598

3)(0

.697

5)(0

.482

3)(0

.456

8)(�

0.85

39)

(1.0

345)

(�2.

2112

)(�

1.01

63)

30 Y

ear

0.21

500.

8076

0.04

29�

0.10

930.

0070

0.13

260.

1312

�0.

4683

�0.

2355

0.02

55(1

.542

0)(5

.242

8)(0

.279

7)(�

0.55

87)

(0.0

358)

(0.6

925)

(0.6

839)

(�2.

4747

)(�

1.21

81)

Cat

egor

y 2

�3

2 Ye

ar0.

0510

�0.

0271

�0.

0173

�0.

2236

0.04

08�

0.21

86�

0.02

89�

0.28

91�

0.01

970.

1655

(1.9

736)

(�0.

8295

)(�

0.52

54)

(�5.

5496

)(1

.017

5)(�

6.75

79)

(�0.

9169

)(�

9.28

84)

(�0.

6346

)

5 Ye

ar0.

1135

�0.

0641

�0.

0510

�0.

6251

0.10

57�

0.52

36�

0.06

84�

0.79

67�

0.07

440.

2389

(2.0

880)

(�0.

9311

)(�

0.73

55)

(�7.

3688

)(1

.253

6)(�

7.68

87)

(�1.

0300

)(�

12.1

586)

(�1.

1399

)

10 Y

ear

0.17

00�

0.10

97�

0.08

72�

0.66

450.

1092

�0.

8967

�0.

0948

�1.

2727

�0.

1560

0.26

44(2

.199

6)(�

1.11

98)

(�0.

8852

)(�

5.50

82)

(0.9

109)

(�9.

2584

)(�

1.00

43)

(�13

.656

3)(�

1.67

98)

30 Y

ear

0.24

67�

0.14

79�

0.06

54�

0.81

060.

0961

�1.

2777

�0.

0430

�2.

5766

�0.

1985

0.32

37(2

.132

0)(�

0.10

97)

(�0.

4432

)(�

4.48

88)

(0.5

351)

(�8.

8131

)(�

0.30

41)

(�18

.469

6)(�

1.42

79) (C

onti

nued

)

TA

BL

E I

V (

Co

nti

nu

ed

)

Res

pons

e of

Fut

ures

Ret

urns

to

Cur

rent

and

Lag

ged

Cus

tom

er T

rade

Ind

icat

or (

CT

I) C

ateg

ory

Ord

erfl

ow

Ord

erflo

w

2 Ye

ar5

Year

10 Y

ear

30 Y

ear

Fut

ures

retu

rnIn

terc

ept

Cur

rent

Lag

ged

Cur

rent

Lag

ged

Cur

rent

Lag

ged

Cur

rent

Lag

ged

R2

Cat

egor

y 4

2 Ye

ar0.

0528

0.00

25�

0.00

150.

1407

�0.

0192

0.24

700.

0156

0.29

540.

0343

0.15

23(2

.048

5)(0

.083

6)(�

0.04

94)

(3.5

664)

(�0.

4881

)(8

.334

0)(0

.535

6)(9

.965

5)(1

.156

9)

5 Ye

ar0.

1165

�0.

0836

�0.

0027

0.42

60�

0.04

480.

6112

0.04

170.

8027

0.09

960.

2214

(2.1

309)

(�1.

3193

)(�

0.04

27)

(5.0

924)

(�0.

5361

)(9

.727

8)(0

.676

7)(1

2.77

51)

(1.5

829)

10 Y

ear

0.17

26�

0.12

260.

0355

0.41

91�

0.06

340.

9409

0.07

081.

3125

0.20

560.

2494

(2.2

170)

(�1.

3595

)(0

.397

8)(3

.517

8)(�

0.53

32)

(10.

5155

)(0

.807

4)(1

4.66

77)

(2.2

946)

30 Y

ear

0.25

07�

0.23

340.

0409

0.62

95�

0.03

751.

2551

0.07

192.

6073

0.25

720.

3131

(2.1

550)

(�1.

7326

)(0

.307

2)(3

.537

2)(�

0.21

07)

(9.3

896)

(0.5

489)

(19.

5037

)(1

.921

1)

Not

e.T

his

tabl

e pr

esen

ts t

he r

espo

nse

of f

utur

es p

rices

to

CT

I or

derfl

ow s

epar

ated

by

cont

ract

mat

urity

. E

ach

orde

rflow

ser

ies

repr

esen

ts t

he r

esid

uals

of

a re

gres

sion

of

orde

rflow

regr

esse

d on

a c

onst

ant

and

a tim

e-to

-del

iver

y va

riabl

e.

Thi

s ad

just

men

t ac

coun

ts f

or t

he s

tron

g se

ason

al p

rese

nt in

eac

h or

derfl

ow s

erie

s.

Coe

ffici

ents

are

mul

tiplie

d by

1,0

00 a

ndt-

stat

istic

s ar

e sh

own

in p

aren

thes

es.

Bol

d va

lues

rep

rese

nt s

igni

fican

ce a

t the

5%

leve

l.



In stark contrast stand Categories 2 and 3; for all but the 2-year contractthe coefficients are significantly negative. This suggests that whenCategories 2 and 3 traders buy, the futures prices actually fall. At firstglance these results seem odd; however, they are consistent with thesetrades being part of a hedging strategy. Category 4, or retail trades, have apositive and significant impact on futures prices, especially for the 5-, 10-,and 30-year contracts. Furthermore, the fact that the current impact oforderflow is significant and permanent whereas the lagged impact isinsignificantly different from zero suggests that Categories 2, 3, and 4orderflow is not driven by inventory considerations; rather all three cate-gories contribute meaningfully to price discovery in the futures market.9

The analogous results for the cash market are shown in Table V. Thefirst panel, labeled “On-the-run,” shows results for the most recentlyissued securities of each maturity. Net orderflow again plays a significantrole in determining the movements in prices. The coefficients for the2- and 5-year note are positive, significant, and strictly increasing withrespect to maturity, for all the different maturities. In contrast, the10- and 30-year contracts are far less informative. The weaker results forthe 10- and 30-year maturities may owe, in part, to duration issues thatare present when dealing with prices rather than yields. In addition, thescant coverage within GovPX for the 30-year Treasury is also likely a con-tributing factor here. Like the results for the futures market, movementsinduced by orderflow are permanent, as seen by the absence of a reversal inthe price moves. Thus, the cash results are consistent with work byBrandt and Kavajecz (2004), which shows that movements in the cashyield curve are permanent even absent news announcements and aretherefore indicative of price discovery rather than inventory considera-tions. The second panel further supports results documented in Brandtand Kavajecz (2004), namely that orderflow in on-the-run Treasuriesplays a significant role in price discovery for off-the-run Treasuries. Inthese regressions, the dependent variable is the return on off-the-runTreasuries, whereas the independent variables are the orderflows for theon-the-run securities. Note that, once again, the 2-year on-the-run order-flow is positive and significant in all regressions. In this case, however, the5-year orderflow is no longer significant.

The study results thus far suggest that locals in the pits (Category 1),in general, do not move prices, at least at the daily frequency, but insteadprovide short-term liquidity to the market. In contrast, movements

9Regressions were also done with up to five (daily) lags of orderflow, and consistent with the resultsof Table VIII, additional lags displayed little explanatory power over the current net orderflow.

TA

BL

E V

Res

pons

e of

Tre

asur

y C

ash

Ret

urns

to

Cur

rent

and

Lag

ged

Ord

erfl

ow

Ord

erflo

w

2 Ye

ar5

Year

10 Y

ear

30 Y

ear

Fut

ures

retu

rnIn

terc

ept

Cur

rent

Lag

ged

Cur

rent

Lag

ged

Cur

rent

Lag

ged

Cur

rent

Lag

ged

R2

On-

the-

Run

2 Ye

ar0.

0425

0.14

050.

0194

0.25

16�

0.00

930.

0050

0.00

790.

0455

0.03

670.

1317

(1.5

702)

(5.0

161)

(0.6

842)

(8.8

948)

(�0.

3303

)(0

.179

9)(0

.280

7)(1

.573

0)(1

.312

1)

5 Ye

ar0.

0800

0.29

670.

0452

0.50

660.

0029

0.04

060.

0174

0.07

910.

0656

0.13

10(1

.431

5)(5

.123

1)(0

.770

1)(8

.662

7)(0

.050

1)(0

.707

0)(0

.298

2)(1

.321

8)(1

.132

9)

10 Y

ear

0.16

520.

4973

0.08

491.

0608

0.05

100.

0788

�0.

0077

0.09

190.

0795

0.15

07(1

.643

8)(4

.776

5)(0

.805

0)(1

0.09

06)

(0.4

872)

(0.7

632)

(�0.

0738

)(0

.854

5)(0

.763

7)

30 Y

ear

0.25

710.

7075

0.16

711.

3753

0.30

070.

2532

�

0.03

21

0.03

630.

0814

0.12

19(1

.632

1)(4

.335

2)(1

.010

4)(8

.346

0)(1

.834

0)(1

.564

2)(�

0.19

56)

(0.2

155)

(0.4

990)

Off

-the

-Run

2 Ye

ar�

0.11

030.

1465

�0.

0171

0.02

080.

0433

0.10

18�

0.05

640.

0584

0.06

120.

0809

(�1.

6816

)(2

.669

5)(�

0.28

76)

(0.3

649)

(0.7

719)

(1.8

055)

(�1.

0879

)(2

.040

5)(1

.435

0)

5 Ye

ar�

0.22

380.

3327

�0.

0319

0.09

260.

0911

0.21

26�

0.13

000.

1324

0.07

840.

0925

(�1.

6706

)(2

.969

0)(�

0.26

23)

(0.7

962)

(0.7

958)

(1.8

476)

(�1.

2286

)(2

.266

7)(0

.900

1)

10 Y

ear

�0.

2213

0.41

170.

3767

0.03

380.

0527

0.42

98�

0.15

680.

1682

0.16

920.

0780

(�0.

9111

)(2

.026

8)(1

.710

1)(0

.160

1)(0

.254

0)(2

.060

1)(�

0.81

76)

(1.5

888)

(1.0

711)

30 Y

ear

�0.

9347

0.85

850.

3574

0.43

690.

0486

0.68

82�

0.65

65�

0.02

520.

3087

0.09

19(�

2.30

50)

(2.5

306)

(0.9

717)

(1.2

413)

(0

.140

2)(1

.975

6)(�

2.04

99)

(�0.

1425

)(1

.170

5)

Not

e.T

his

tabl

e pr

esen

ts th

e re

spon

se o

f Tre

asur

y ca

sh r

etur

ns to

net

ord

erflo

w in

the

cash

mar

ket s

epar

ated

by

cont

ract

mat

urity

. C

oeffi

cien

ts a

re m

ultip

lied

by 1

,000

and

t-st

atis

tics

are

show

n in

par

enth

eses

. B

old

valu

es r

epre

sent

sig

nific

ance

at t

he 5

% le

vel.



induced by Categories 2, 3, and 4 trades have a permanent impacton price suggesting that these trades all contribute to price discovery.Retail traders (Category 4) take positions such that prices move in thedirection of their trades; thus, as a group they are the driving forcebehind movements in futures prices. Hedgers in Categories 2 and 3attenuate movements in futures prices to the extent that they tradeopportunistically against retail traders. As in past research, net orderflowis found to be the mechanism through which price discovery is accom-plished in the Treasury cash market, with the 2- and 5-year net orderflowbeing particularly influential.

INTERACTION BETWEEN THE FUTURES AND CASH MARKET

In this section, the joint interaction is investigated between the futuresmarket and the cash market. Although arbitrage dictates that the two mar-kets must be linked, it is an empirical question in which market price dis-covery originates and how it is disseminated. Specifically, the return in thecash market is regressed on contemporaneous cash and futures orderflowand vice versa. The cash market return regressions are presented inTable VI. Although typically a market’s own orderflow is more influentialthan cross-market orderflow in explaining a market’s own return; that isnot the case here. The results suggest that net orderflow in the futuresmarket must contribute different information to price discovery in thecash market because the unconditional impact of cash market net order-flow is both quantitatively and qualitatively similar to the impact condi-tional on the cross market orderflow. Notice that like Table V, orderflow inthe 2- and 5-year note are still significantly positively related to cash pricemovements. However, the futures orderflow is also significantly related tocash price movements for CTI categories 2, 3, and 4. Categories 2 and 3orderflow is negatively related to cash prices throughout the curve, where-as Category 4 orderflow is positively related to cash prices. As is the casewith the cash orderflow, the effect increases in magnitude as one goestoward the long end of the curve. The gain in explanatory power fromadding futures orderflow to the cash return regressions is also notable; theadjusted R-squares nearly double with the addition of either Categories2 and 3 or Category 4 orderflow. These results clearly indicate that tradingactivity in the futures market contributes significantly to price discovery inthe cash market for U.S. Treasury securities.

The results in Table VII concerning the futures returns are alsoquite interesting with regard to the role of cross-market orderflow.

TA

BL

E V

I

Impa

ct o

f F

utur

es a

nd C

ash

Net

Ord

erfl

ow o

n Tr

easu

ry C

ash

Ret

urns

Ord

erflo

w

Fut

ures

net

ord

erflo

w b

y C

TI

cate

gory

Cas

h ne

t or

derfl

ow b

y m

atur

ity

Cas

hre

turn

Inte

rcep

t2

Year

5 Ye

ar10

Yea

r30

Yea

r2

Year

5 Ye

ar10

Yea

r30

Yea

rR

2

Cat

egor

y 1

2 Ye

ar0.

0312

�0.

0046

0.02

64�

0.07

37�

0.04

300.

1514

0.22

940.

0056

0.05

870.

1346

(1.2

257)

(�0.

1627

)(0

.736

9)(�

2.14

42)

(�1.

2841

)(5

.761

4)(8

.677

5)(0

.217

8)(2

.262

3)

5 Ye

ar0.

0541

0.08

840.

0799

�0.

1300

�0.

1225

0.31

190.

4655

0.03

850.

1023

0.13

59(1

.030

3)(1

.522

9)(1

.082

0)(�

1.83

52)

(�1.

7730

)(5

.756

6)(8

.536

8)(0

.730

3)(1

.911

2)

10 Y

ear

0.12

110.

2285

0.14

83�

0.08

69�

0.30

080.

5267

0.95

990.

0634

0.13

680.

1508

(1.2

795)

(2.1

824)

(1.1

126)

(�0.

6795

)(�

2.41

23)

(5.3

875)

(9.7

549)

(0.6

664)

(1.4

158)

30 Y

ear

0.22

300.

2803

�0.

0890

0.

1613

�0.

3104

0.73

741.

2358

0.17

740.

1277

0.10

94(1

.481

1)(1

.683

3)(�

0.42

02)

(0.7

932)

(�1.

5655

)(4

.743

2)(7

.897

7)(1

.172

8)(0

.830

9)

Cat

egor

y 2

�3

2 Ye

ar0.

0361

�0.

0477

�0.

1649

�0.

1113

�0.

1872

0.14

540.

1702

�0.

0067

0.04

580.

2120

(1.5

004)

(�1.

5326

)(�

4.04

77)

(�3.

6565

)(�

6.30

67)

(5.8

022)

(6.5

621)

(�0.

2760

)(1

.855

6)

5 Ye

ar0.

0675

�0.

1071

�0.

4419

�0.

2548

�0.

4593

0.29

940.

3233

0.01

200.

0792

0.24

70(1

.389

2)(�

1.70

38)

(�5.

3711

)(�

4.14

49)

(�7.

6640

)(5

.917

4)(6

.171

6)(0

.246

5)(1

.588

9)

10 Y

ear

0.14

13�

0.21

29�

0.69

55�

0.58

35�

0.95

130.

4884

0.67

660.

0234

0.09

010.

2797

(1.6

282)

(�1.

8949

)(�

4.73

01)

(�5.

3121

)(�

8.88

22)

(5.4

015)

(7.2

275)

(0.2

685)

(1.0

117)

30 Y

ear

0.23

59�

0.14

93�

0.83

05�

0.92

49�

1.43

880.

6926

0.81

890.

1397

0.06

300.

2247

(1.6

801)

(�0.

8216

)(�

3.49

14)

(�5.

2046

)(�

8.30

47)

(4.7

357)

(5.4

079)

(0.9

901)

(0.4

372)

(Con

tinu

ed)

TA

BL

E V

I (C

on

tin

ue

d)

Impa

ct o

f F

utur

es a

nd C

ash

Net

Ord

erfl

ow o

n Tr

easu

ry C

ash

Ret

urns

Ord

erflo

w

Fut

ures

net

ord

erflo

w b

y C

TI

cate

gory

Cas

h ne

t or

derfl

ow b

y m

atur

ity

Cas

hre

turn

Inte

rcep

t2

Year

5 Ye

ar10

Yea

r30

Yea

r2

Year

5 Ye

ar10

Yea

r30

Yea

rR

2

Cat

egor

y 4

2 Ye

ar0.

0376

0.04

700.

0776

0.14

860.

1859

0.15

220.

1769

�0.

0078

0.04

270.

2106

(1.5

634)

(1.6

355)

(2.0

373)

(5.3

417)

(6.5

211)

(6.0

710)

(6.8

793)

(�0.

3238

)(1

.730

8)

5 Ye

ar0.

0701

0.04

940.

2153

0.33

740.

4658

0.31

040.

3369

0.01

370.

0691

0.23

73(1

.436

8)(0

.845

8)(2

.785

8)(5

.976

4)(8

.055

5)(6

.103

1)(6

.455

5)(0

.279

0)(1

.378

6)

10 Y

ear

0.14

390.

0541

0.35

390.

6502

0.99

090.

5055

0.69

720.

0274

0.07

570.

2695

(1.6

453)

(0.5

180)

(2.5

554)

(6.4

292)

(9.5

652)

(5.5

488)

(7.4

575)

(0.3

122)

(0.8

431)

30 Y

ear

0.23

850.

0690

0.68

100.

7930

1.44

850.

7178

0.86

200.

1398

0.04

070.

2127

(1.6

856)

(0.4

078)

(3.0

397)

(4.8

466)

(8.6

429)

(4.8

701)

(5.6

990)

(0.9

856)

(0.2

806)

Not

e.T

his

tabl

e pr

esen

ts th

e im

pact

of f

utur

es a

nd c

ash

net o

rder

flow

on

Trea

sury

Cas

h re

turn

s se

para

ted

by c

ontr

act m

atur

ity.

Coe

ffici

ents

are

mul

tiplie

d by

1,0

00 a

nd t-

stat

istic

s ar

esh

own

in p

aren

thes

es.

Bol

d va

lues

rep

rese

nt s

igni

fican

ce a

t the

5%

leve

l. C

TI �

cust

omer

trad

e in

dica

tor.

TA

BL

E V

II

Impa

ct o

f F

utur

es a

nd C

ash

Net

Ord

erfl

ow o

n Tr

easu

ry F

utur

es R

etur

ns

Ord

erflo

w

Fut

ures

net

ord

erflo

w b

y C

TI

cate

gory

Cas

h ne

t or

derfl

ow b

y m

atur

ity

Cas

hre

turn

Inte

rcep

t2

Year

5 Ye

ar10

Yea

r30

Yea

r2

Year

5 Ye

ar10

Yea

r30

Yea

rR

2

Cat

egor

y 1

2 Ye

ar0.

0624

�0.

0024

�0.

0039

�0.

0631

�0.

0514

0.16

210.

2398

�0.

0022

0.09

100.

1215

(2.2

511)

(�0.

0778

)(�

0.09

86)

(�1.

6822

)(�

1.41

79)

(5.6

664)

(8.2

503)

(�0.

0788

)(3

.170

1)

5 Ye

ar0.

1297

0.15

450.

0214

�0.

1362

�0.

1206

0.38

340.

5830

0.04

070.

1213

0.14

85(2

.180

8)(2

.357

3)(0

.254

2)(�

1.69

33)

(�1.

5508

)(6

.247

8)(9

.350

5)(0

.675

4)(1

.970

7)

10 Y

ear

0.21

940.

2601

0.13

23�

0.13

42�

0.22

480.

5333

0.85

570.

0810

0.14

180.

1541

(2.5

921)

(2.7

880)

(1.1

061)

(�1.

1722

)(�

2.03

03)

(6.1

054)

(9.6

402)

(0.9

450)

(1.6

173)

30 Y

ear

0.33

250.

5099

0.04

780.

0350

�0.

4095

0.80

651.

4208

0.12

800.

1691

0.16

36(2

.516

7)(3

.501

5)(0

.256

0)(0

.196

1)(�

2.36

96)

(5.9

158)

(10.

2565

)(0

.956

7)(1

.235

9)

Cat

egor

y 2

�3

2 Ye

ar0.

0644

�0.

0614

�0.

1968

�0.

1542

�0.

2062

0.15

290.

1608

�0.

0124

0.08

640.

2097

(2.4

535)

(�1.

8492

)(�

4.52

09)

(�4.

6219

)(�

6.37

23)

(5.6

052)

(5.6

194)

(�0.

4670

)(3

.160

0)

5 Ye

ar0.

1386

�0.

1333

�0.

5751

�0.

3787

�0.

6011

0.36

640.

3669

0.01

850.

1060

0.29

48(2

.587

0)(�

1.96

57)

(�6.

4737

)(�

5.56

10)

(�9.

0996

)(6

.581

1)(6

.282

6)(0

.341

8)(1

.900

3)

10 Y

ear

0.22

85�

0.19

85�

0.65

42�

0.71

67�

0.95

190.

4986

0.53

330.

0504

0.10

770.

3233

(3.0

415)

(�2.

0884

)(�

5.25

36)

(�7.

5079

)(�

10.2

812)

(6.3

903)

(6.5

146)

(0.6

625)

(1.3

772)

30 Y

ear

0.34

83�

0.18

82�

0.80

94�

1.01

93�

2.11

620.

6933

0.85

700.

0815

0.11

100.

3829

(3.0

835)

(�1.

3171

)(�

4.32

31)

(�7.

1014

)(�

15.2

010)

(5.9

094)

(6.9

624)

(0.7

126)

(0.9

442) (C

onti

nued

)

TA

BL

E V

II (

Co

nti

nu

ed

)

Impa

ct o

f F

utur

es a

nd C

ash

Net

Ord

erfl

ow o

n Tr

easu

ry F

utur

es R

etur

ns

Ord

erflo

w

Fut

ures

net

ord

erflo

w b

y C

TI

cate

gory

Cas

h ne

t or

derfl

ow b

y m

atur

ity

Cas

hre

turn

Inte

rcep

t2

Year

5 Ye

ar10

Yea

r30

Yea

r2

Year

5 Ye

ar10

Yea

r30

Yea

rR

2

Cat

egor

y 4

2 Ye

ar0.

0616

0.05

200.

1196

0.18

490.

2080

0.16

420.

1691

�0.

0097

0.08

160.

2106

(2.3

718)

(1.7

117)

(2.9

324)

(6.0

407)

(6.7

873)

(6.0

894)

(6.0

328)

(�0.

3690

)(3

.017

6)

5 Ye

ar0.

1313

0.02

370.

3338

0.48

040.

5956

0.39

090.

3927

0.02

450.

0959

0.28

62(2

.449

5)(0

.378

2)(3

.964

3)(7

.602

1)(9

.413

5)(7

.018

1)(6

.784

1)(0

.452

1)(1

.716

9)

10 Y

ear

0.21

990.

0251

0.31

470.

7952

0.97

750.

5258

0.57

030.

0605

0.09

810.

3148

(2.9

192)

(0.2

841)

(2.6

594)

(8.9

555)

(10.

9952

)(6

.718

6)(7

.012

0)(0

.794

3)(1

.249

4)

30 Y

ear

0.33

73�

0.07

510.

4363

1.05

812.

1352

0.73

530.

9007

0.09

050.

1024

0.37

69(2

.975

2)(�

0.56

62)

(2.4

501)

(7.9

182)

(15.

9600

)(6

.243

6)(7

.359

2)(0

.789

8)(0

.867

0)

Not

e.T

his

tabl

e pr

esen

ts th

e im

pact

of f

utur

es a

nd c

ash

net o

rder

flow

on

Trea

sury

Fut

ures

ret

urns

sep

arat

ed b

y co

ntra

ct m

atur

ity. C

oeffi

cien

ts a

re m

ultip

lied

by 1

,000

and

t-st

atis

tics

are

show

n in

par

enth

eses

. Bol

d va

lues

rep

rese

nt s

igni

fican

ce a

t the

5%

leve

l. C

TI �

cust

omer

trad

e in

dica

tor.



Trading by Categories 2 and 3, and Category 4 groups still explains asubstantial amount of the variation in futures returns. However, as in thecase of the cash returns, the cross-market orderflow variables alsoexplain a great deal of the variation in returns for each of the futurescontracts. Here the orderflow in the 2- and 5-year notes is significantlypositively related to futures returns. The contribution to the R-squaredvalues is again notable, especially in the case of the Category 1 regres-sions. But even in the Categories 2 and 3 and Category 4 regressions,adding cash orderflow to the futures return regressions increases theR-squared by around 6–7%. In short, the two sets of regressions shownin Tables VI and VII indicate that both the futures and cash markets forTreasury securities contribute substantially to price discovery in the mar-ket for riskless securities. The results suggest that futures may be moreinfluential than cash in terms of cross-market effects, particularly forCategories 2, 3, and 4 orderflow.

As a final piece of evidence regarding the cross-market nature ofprice discovery in the Treasury market, further results are provided foroff-the-run Treasuries. Recall that Table V showed that on-the-runorderflow explains a large part of the variation in off-the-run returns.These results are consistent with Brandt and Kavajecz (2004), and showthat price discovery in the off-the-run portion of the curve largely occursin on-the-run securities. Given the arbitrage relation that must holdbetween the cheapest-to-deliver security and the futures price, and giventhat the cheapest security is typically off-the-run, which market con-tributes more to off-the-run price discovery: the cash market or thefutures market? Table VIII shows the results of adding futures marketorderflow to the regressions run in the second panel of Table V. Theresults show that again, futures orderflow plays a substantial role. All ofthe CTI categories come in significant (including Category 1), and theiraddition increases R-squared values sharply. In the most extreme case ofthe 5-year returns, the R-squared increases from 9–35% with the addi-tion of Category 4 orderflow. These results indicate that for off-the-runTreasuries, where the cheapest-to-deliver security typically resides, thefutures market contributes more to the price discovery process than trad-ing in on-the-run cash securities.

FORCES INFLUENCING THECASH/FUTURES RELATION

In the previous section, the case for multidirectional price discoverybetween the Treasury futures and cash markets was established. In this

TA

BL

E V

III

Impa

ct o

f F

utur

es a

nd O

n-th

e-R

un C

ash

Net

Ord

erfl

ow o

n O

ff-t

he-R

un T

reas

ury

Cas

h R

etur

ns

Ord

erflo

w

Fut

ures

net

ord

erflo

w b

y C

TI

cate

gory

Cas

h ne

t or

derfl

ow b

y m

atur

ity

Cas

hre

turn

Inte

rcep

t2

Year

5 Ye

ar10

Yea

r30

Yea

r2

Year

5 Ye

ar10

Yea

r30

Yea

rR

2

Cat

egor

y 1

2 Ye

ar�

0.08

42�

0.08

810.

2227

�0.

2328

0.04

640.

1500

0.26

19�

0.02

090.

0798

0.27

17(�

1.53

10)

(�1.

2910

)(2

.577

0)(�

2.40

83)

(0.5

453)

(2.7

680)

(5.0

990)

(�0.

4282

)(1

.485

3)

5 Ye

ar�

0.18

04�

0.15

080.

4425

�0.

4349

0.03

520.

3524

0.60

89�

0.07

340.

1763

0.31

18(�

1.61

59)

(�1.

0891

)(2

.522

3)(�

2.21

62)

(0.2

039)

(3.2

034)

(5.8

41)

(�0.

7403

)(1

.617

4)

10 Y

ear

�0.

2542

�0.

0651

1.10

98�

1.56

250.

2695

0.25

131.

0203

�0.

3846

0.03

260.

2233

(�1.

0782

)(�

0.22

25)

(2.9

953)

(�3.

7698

)(0

.738

5)(1

.081

8)(4

.633

9)(�

1.83

57)

(0.1

414)

30 Y

ear

�0.

9554

�0.

2106

0.72

24�

1.64

84�

0.84

781.

0851

1.38

97�

0.44

110.

238

0.18

07(�

2.38

43)

(�0.

4235

)(1

.146

9)(�

2.33

97)

(�1.

3666

)(2

.747

7)(3

.713

3)(�

1.23

87)

(0.6

081)

Cat

egor

y 2

�3

2 Ye

ar�

0.03

82�

0.04

�0.

2489

�0.

1736

�0.

0911

0.12

520.

2103

�0.

0446

0.07

910.

2725

(�0.

6859

)(�

0.40

36)

(�2.

2268

)(�

2.08

32)

(�1.

4005

)(2

.271

2)(3

.944

5)(�

0.90

38)

(1.4

759)

5 Ye

ar�

0.07

72�

0.09

21�

0.57

16�

0.32

62�

0.27

140.

2951

0.48

73�

0.12

30.

1774

0.33

13(�

0.69

28)

(�0.

464)

(�2.

5546

)(�

1.95

58)

(�2.

0854

)(2

.674

3)(4

.566

3)(�

1.24

39)

(1.6

542)

10 Y

ear

�0.

0372

�0.

0174

�1.

1282

�0.

5929

�0.

3872

0.21

690.

7887

�0.

5267

0.06

090.

1788

(�0.

1517

)(�

0.03

98)

(�2.

2883

)(�

1.61

31)

(�1.

3502

)(0

.892

1)(3

.354

3)(�

2.41

78)

(0.2

579)

30 Y

ear

�0.

6432

�0.

2204

�1.

4216

�0.

709

�1.

0277

0.86

831.

0294

�0.

5464

0.14

570.

1856

(�1.

5903

)(�

0.30

58)

(�1.

7495

)(�

1.17

05)

(�2.

1744

)(2

.167

2)(2

.656

6)(�

1.52

21)

(0.3

742)

(Con

tinu

ed)

TA

BL

E V

III

(Co

nti

nu

ed

)

Impa

ct o

f F

utur

es a

nd O

n-th

e-R

un C

ash

Net

Ord

erfl

ow o

n O

ff-t

he-R

un T

reas

ury

Cas

h R

etur

ns

Ord

erflo

w

Fut

ures

net

ord

erflo

w b

y C

TI

cate

gory

Cas

h ne

t or

derfl

ow b

y m

atur

ity

Cas

hre

turn

Inte

rcep

t2

Year

5 Ye

ar10

Yea

r30

Yea

r2

Year

5 Ye

ar10

Yea

r30

Yea

rR

2

Cat

egor

y 4

2 Ye

ar�

0.04

380.

1946

0.11

010.

3336

0.11

210.

1164

0.21

54�

0.03

650.

0862

0.30

36(�

0.80

4)(2

.304

3)(0

.941

4)(4

.248

6)(1

.756

8)(2

.177

5)(4

.122

7)(�

0.76

41)

(1.6

417)

5 Ye

ar�

0.08

790.

3845

0.29

660.

6701

0.31

970.

2814

0.49

53�

0.10

970.

1945

0.35

65(�

0.80

43)

(2.2

682)

(1.2

638)

(4.2

513)

(2.4

965)

(2.6

222)

(4.7

232)

(�1.

1451

)(1

.844

4)

10Ye

ar�

0.08

130.

409

0.28

341.

4523

0.42

580.

1484

0.81

78�

0.48

80.

0415

0.21

01(�

0.33

78)

(1.0

953)

(0.5

483)

(4.1

828)

(1.5

095)

(0.6

279)

(3.5

403)

(�2.

3134

)(0

.178

8)

30 Y

ear

�0.

6125

0.93

051.

3802

2.01

241.

2984

0.78

060.

9756

�0.

5564

0.20

10.

2271

(�1.

5544

)(1

.522

1)(1

.631

1)(3

.540

3)(2

.811

6)(2

.017

)(2

.579

8)(�

1.61

1)(0

.528

5)

Not

e.T

his

tabl

e pr

esen

ts th

e im

pact

of f

utur

es a

nd c

ash

net o

rder

flow

on

Trea

sury

Cas

h re

turn

s se

para

ted

by c

ontr

act m

atur

ity. C

oeffi

cien

ts a

re m

ultip

lied

by 1

,000

and

t-st

atis

tics

are

show

n in

par

enth

eses

. B

old

valu

es r

epre

sent

sig

nific

ance

at t

he 5

% le

vel.

CT

I �cu

stom

er tr

ade

indi

cato

r.



section, we take a more comprehensive approach by investigating twooutside forces that may influence the extent and/or direction of thecash/future relation. The costs of financing the purchase of fixed incomesecurities as well as the liquidity environment in the cash market are con-jectured to have a substantial impact on Treasury yield curve movements.

The rationale for considering these outside forces stems from thefollowing background. First, due to Brandt and Kavajecz (2004), as wellas results presented above, the on-the-run orderflow drives changes(price discovery) in the cash market. Second, has as shown above (andhas been shown elsewhere), price discovery occurs in both the underly-ing and derivatives markets. Third, prices in the fixed income futuresmarket apply directly to the cheapest-to-deliver issues, which are almostalways off-the-run issues. Given these observations, it stands to reasonthat there is no direct mechanism (rather there is an indirect link due to thecheapest-to-deliver typically being off-the-run securities) by whichthe futures market is the primary location for price discovery and canmove the cash curve. Said differently, even though price discovery occursin the futures market, it has a direct impact on the off-the-run (CTDissues) securities, which has a muted impact on the current (on-the-run)curve. Thus, what alters the relative attractiveness of transactingbetween the futures and cash market?

We hypothesize that the cost of financing (Repo rates), as well asthe liquidity of the Treasury cash market are key determinants of whereprice discovery takes place. If a trader is deciding between obtainingexposure to interest rate movements either through the cash market orthe futures market, what is paramount is the relative cost of the requiredexposure in the two markets.

With regard to financing rates, a trader wishing to buy an on-the-runissue that is on special (i.e., the applicable repurchase agreement rate islower than the analogous rate for general collateral), will pay relativelymore than he would for an issue that is not on special because the issueon special is in high demand (or short supply). However, if the on-the-runissue is not on special, that same trader would be able to pay a lowerprice for the issue, albeit the applicable Repo rate would necessarily behigher. In addition, it may be more difficult to establish a short positionin the cash market when the security is on special, as it would be moredifficult to locate a security to deliver. Under this scenario, we hypothe-size that specialness in a Treasury issue will induce price discovery totake place in the futures market, whereas more price discovery will takeplace in the cash market during periods when the issue can only be“Repoed” at the general collateral rate.



Finally, it is well understood that liquidity in a market, defined bythe bid-ask spread and quoted depth, is a prime determinant of transac-tion costs. Financial theory suggests that liquidity is a function of thelevel of adverse selection risk within a market. In particular, when thereis a high level of informed trading, spreads widen and depths fall, there-by reducing liquidity. Trading venues that have high levels of informedtrading contribute relatively more to price discovery. Thus, we hypothe-size that precisely when the Treasury cash market is less liquid is wheninformed traders are transacting in the cash market, which, in turn,leads to relatively more price discovery occurring in the cash market.

These hypotheses are tested by rerunning the orderflow regressionsconditioning on measures for each of the variables described above.Specifically, a dummy variable is created for each maturity/contract thatidentifies periods of specialness (defined as a specialness being in theupper quintile of its distribution for the entire sample) and illiquidity(defined as the bid-ask spread being in the upper quintile of its distribu-tion over the sample for that maturity). These dummy variables are mul-tiplied by the net orderflow variables to form interaction terms, whichallow us to measure the conditional impact of specialness and illiquidityon the price discovery relation.

Our hypothesis concerning financing rates suggests that specialnesswill shift price discovery to the futures market. Under this hypothesis,the impact of cash net orderflow would be muted and futures net order-flow would be accentuated during periods of specialness. Thus, interac-tion terms with cash net orderflow should be negative (muting the strongpositive impact) and interaction terms with futures net orderflow posi-tive for Category 4.10 Tables IX and X present the results; for brevity onlythe interaction terms are shown, although all the unconditional variables(not shown) maintain their sign and significance from the previousanalysis. Panel A shows results for the impact of futures orderflow oncash returns. Consistent with our hypothesis, the coefficients of the5- and 30-year interaction terms take on the appropriate sign and aresignificant in three of the four cases, whereas the interaction terms for2- and 10-year futures contracts are insignificantly different from zero.In addition, panel B shows that orderflow in the on-the-run 5-year notebecomes relatively less important in explaining futures returns when the5-year note is on special. In general, the results confirm our hypothesisthat the futures market will become relatively more important during

10For brevity, we only display results for Category 4 orderflow. As expected, results for Categories2 and 3 are similar in nature, but in the opposite direction.



TABLE IX

Cross-Market Orderflow Impact When Cash Treasury Is on Special

Panel A Futures net orderflow interaction variable by maturity

Cash return 2 Year 5 Year 10 Year 30 Year

Category 4

2 Year �0.0010 0.2086 0.0350 0.1164(�0.0136) (2.1375) (0.4692) (1.6940)

5 Year �0.0107 0.4072 0.1247 0.3001(�0.0675) (2.0601) (0.8255) (2.1596)

10 Year �0.0794 0.3037 0.3282 0.5289(�0.2778) (0.8548) (1.2106) (2.1201)

30 Year �0.2427 �0.9717 0.2154 0.8500(�0.5252) (�1.6930) (0.4911) (2.1070)

Panel BCash net orderflow interaction variable by maturity

Futures return 2 Year 5 Year 10 Year 30 Year

Category 4

2 Year 0.0587 �0.1656 �0.0070 �0.0871(0.8277) (�2.4711) (�0.1100) (�1.2482)

5 Year 0.1151 �0.2762 �0.0081 �0.0374(0.7929) (�2.0107) (�0.0618) (�0.2620)

10 Year 0.1682 �0.3337 �0.0063 0.0276(0.8290) (�1.7378) (�0.0344) (0.1380)

30 Year �0.1461 �0.3724 �0.0753 0.2469(�0.4795) (�1.2904) (�0.2747) (0.8232)

Note. This table presents the impact of futures net orderflow on cash Treasury returns when the given on-the-run Treasuryis trading on special in the repo market. Futures net orderflow is interacted with a dummy that is equal to 1 when the givenmaturity Treasury is trading in the upper quintile of specialness over the entire sample. In panel A, each estimate shownrepresents the interaction term added to the regressions shown in Table VII for CTI4 orderflow. In panel B, each estimateshown represents the interaction term added to the regressions shown in Table VI for CTI4 orderflow. Only the coefficientestimates and t-statistics (shown in parentheses) for the interaction variables are shown. Bold values represent signifi-cance at the 5% level.

periods where specialness in the on-the-run Treasury is high, with the5- and 30-year maturities most affected.

Our hypothesis concerning liquidity implies that illiquidity in thecash market is a manifestation of the presence of informed traders, thusa primary location for price discovery. Under our hypothesis, cash market



interaction terms should reinforce the positive impact that cash marketorderflow has on Treasury prices. The results (shown in Table X) aregenerally consistent with our hypothesis, with the exception of tradingin the 30-year futures contract. Panel A shows that Category 4 orderflow inthe 10-year futures contract becomes less important in explaining cashreturns when the 10-year cash Treasury is illiquid. Orderflow in the

TABLE X

Cross-Market Orderflow Impact When Cash Treasury Is Illiquid

Panel AFutures net orderflow interaction variable by maturity

Cash return 2 Year 5 Year 10 Year 30 Year

Category 4

2 Year �0.0499 �0.1408 �0.1033 0.1649(�0.6568) (�1.4947) (�1.5971) (2.3392)

5 Year �0.2494 �0.1420 �0.1914 0.4583(�1.6236) (�0.7437) (�1.4614) (3.2183)

10 Year �0.4711 �0.4206 �0.5021 0.4308(�1.7087) (�1.2277) (�2.1382) (1.6795)

30 Year �0.4091 �1.0675 �1.9090 �0.3448(�0.9168) (�1.9289) (�5.0758) (�0.8306)

Panel B Cash net orderflow interaction variable by maturity

Futures return 2 Year 5 Year 10 Year 30 Year

2 Year 0.0017 0.2678 �0.0836 0.1571(0.0210) (2.7616) (�0.7976) (1.6523)

5 Year 0.1985 0.6529 �0.1515 0.3169(1.1850) (3.2927) (�0.7060) (1.6271)

10 Year 0.3575 1.2263 �0.0208 0.3948(1.5277) (4.4436) (�0.0695) (1.4504)

30 Year 0.7556 1.9897 0.2976 0.5054(2.1518) (4.8073) (0.6607) (1.2358)

Note. This table presents the impact of futures net orderflow on cash Treasury returns when the given on-the-run Treasuryis less liquid in the cash market. Futures net orderflow is interacted with a dummy that is equal to 1 when the given maturityTreasury exhibits a bid-ask spread in the upper quintile of its distribution over the entire sample. In panel A, each estimateshown represents the interaction term added to the regressions shown in Table VII for CTI4 orderflow. In panel B, each esti-mate shown represents the interaction term added to the regressions shown in Table VI for CTI4 orderflow. Only the coef-ficient estimates and t-statistics (shown in parentheses) for the interaction variables are shown. Bold values representsignificance at the 5% level.



30-year contract, however, actually becomes more useful in terms of itsability to explain 2- and 5-year cash returns when the 30-year Treasury isrelatively illiquid. Panel B shows that trading activity in both the 2- and5-year notes takes on increased importance in explaining futures returnswhen the cash market is relatively illiquid. This is also consistent withour hypothesis concerning contributions to price discovery when thecash market is illiquid.

In general, the results in this section suggest that where price dis-covery takes place depends on two important environmental variables:financing (Repo) rates and liquidity. Moreover, the impact on prices oforderflow in the 5-year maturity appears to be a focal point for theseeffects as it is especially influenced by these factors.

CONCLUSION

The goal of this study is to understand the forces that shape move-ments in the Treasury yield curve. Orderflow within the cash marketfor U.S. Treasuries has been shown to be an important conduit throughwhich price discovery takes place; nevertheless, the trading of fixedincome futures is another important facet of this market that needs tobe understood if a complete picture of yield curve movements is to bepainted.

How prices respond to information flows in the fixed income futuresmarket were analyzed. As in the cash market, the results reinforce thenotion that orderflow drives movements in the futures prices. In particular,the type of trader executing the trade has an impact, with retail tradersmoving futures prices significantly in the direction of their trades,exchange members moving futures prices significantly against theirtrades likely owing to hedging demands, and local market mak-ers who supply short-term liquidity, having little long term impact onfutures prices. In the Treasury cash market, trading in the 2- and 5-yearnotes play significant roles in price discovery as they have significanteffects on prices along the entire yield curve.

When the joint interaction of orderflow between the cash andfutures markets is considered, both markets contribute to price discoveryas net orderflow in each market has explanatory power in explainingprice changes in both the Treasury cash and futures market contempora-neously. As such, the results suggest that futures and cash net orderflowcontain information that is different from the information contained intrading activity in the other market. Specifically, net orderflow in the2- and 5-year Treasury cash market has significant explanatory power



over price movements in both markets, as do futures net orderflow fromboth exchange members and retail traders. On the one hand, net order-flow originating on the short end of the cash market (2- and 5-year matu-rities) and retail trades on the long end of the futures market pushup prices and steepen the entire yield curve. On the other hand, netorderflow originating from exchange members in the futures marketdepresses prices and flattens the entire yield curve. In addition, tradingin the futures market has a substantial impact on price discovery withinoff-the-run Treasury securities.

The above analysis reveals the unconditional relation between pricediscovery in the two markets. The direction and extent of price discoveryis influenced by environmental variables such as Repo financing ratesand liquidity. Consistent with our hypothesis, specialness causes pricediscovery to occur relatively more in the futures market because trans-acting in the cash market when the issue is on special is exceedinglyexpensive. Moreover, illiquidity, measured by wide bid-ask spreads, in thecash market is associated with relatively more price discovery occurring inthe cash market as asymmetric information is high and therefore,informed traders are likely transacting in that market. Interestingly, theinstrument of central importance to the above results is the 5-year matu-rity in both the futures and cash market, largely to the exclusion of theother maturities.

In conclusion, this analysis highlights the importance of consider-ing the relation between separate, but related markets, and not simplyanalyzing the behavior of a market in isolation. As market participantsare always comparing the costs and benefits of transacting in relatedmarkets, so too should our models of the fixed income market. Failure toaccount for cross-market effects, at best, paints an incomplete picture,and at worst may lead to inaccurate conclusions.

BIBLIOGRAPHY

Anderson, T.G., Bollerslev, T., Diebold, F., & Vega, C. (2004). Real-time pricediscovery in stock, bond and foreign exchange markets (working paper).Evanston/Chicago: IL: Northwestern University.

Brandt, M., & Kavajecz, K. (2004). Price discovery in the U.S. Treasury market:The impact of orderflow and liquidity on the yield curve. Journal ofFinance, 59, 2623–2654.

Burghardt, G., Belton, T., Lane, M., & Papa, J. (1994). The Treasury BondBasis. New York: McGraw-Hill.

Daigler, R., & Wiley, M. (1999). The impact of trader type on the futuresvolatility-volume relation. Journal of Finance, 54, 2297–2316.



Easley, D., O’Hara, M., & Srinivas, P.S. (1998). Option volume and stock prices:Evidence on where informed traders trade. Journal of Finance, 53, 431–465.

Evans, M., & Lyons, R. (2002a). Orderflow and exchange rate dynamics.Journal of Political Economy, 110, 170–180.

Evans, M., & Lyons, R. (2002b). Informational integration and FX trading.Journal of International Money and Finance, 21, 807–831.

Ferguson, M., & Mann, S. (1999). The impact of off-exchange customer tradeson prices: evidence from the futures markets (working paper). Fort Worth,TX: Texas Christian University.

Fleming, M., & Remolona, E. (1997). What moves the bond market? FederalReserve Bank of New York Economic Policy Review, 3, 31–51.

Fleming, M., & Remolona, E. (1999). Price formation and liquidity in the U.S.Treasuries market: The response to public information. Journal of Finance,54, 1901–1915.

Glosten, L., & Milgrom, P. (1985). Bid, ask and transaction prices in a special-ist market with heterogeneously informed traders. Journal of FinancialEconomics, 14, 71–100.

Green, C. (2004). Economic news and the impact of trading on bond prices.Journal of Finance, 59, 1201–1234.

He, C. (2005). The influence of institutional trading (working paper). Madison,WI: University of Wisconsin-Madison.

Johnston, E., Kracaw, W., & McConnell, J. (1991). Day-of-the-week effects infinancial futures: An analysis of GNMA, T-bond, T-note and T-bill con-tracts. Journal of Financial and Quantitative Analysis, 26, 23–44.

Keim, D., & Madhavan, A. (1995). Anatomy of the trading process: Empiricalevidence on the behavior of institutional investors. Journal of FinancialEconomics, 37, 371–398.

Keim, D., & Madhavan, A. (1997). Transaction costs and investment style: Aninter-exchange analysis of institutional equity trades. Journal of FinancialEconomics, 46, 265–292.

Kyle, A. (1985). Continuous auctions and insider trading. Econometrica, 53,1315–1335.

Manaster, S., & Mann, S. (1996). Life in the pits: Competitive market makingand inventory control. Review of Financial Studies, 9, 953—975.

Menkveld, A., Sarkar, A., & van der Wel, M. (2006). The informativeness of cus-tomer order flow following macroeconomic announcements: Evidencefrom the Treasury futures markets (working paper). New York, NY: FederalReserve Bank of New York.

Mizrach, B., & Neely, C. (2006). Price discovery in the U.S. Treasury market(working paper). St. Louis, MO: Federal Reserve Bank of St. Louis.

Rosenberg, J., & Traub, L. (2006). Price discovery in the foreign currencyfutures and spot market (working paper). New York, NY: Federal ReserveBank of New York.

Simpson, W.G., & Ireland, T. (1985). The impact of financial futures on thecash market for Treasury bills. Journal of Financial and QuantitativeAnalysis, 20, 371–379.

Underwood, S. (2006). The cross-market information content of stock and bondorderflow (working paper). Houston, TX: Rice University.

price discovery in the treasury futures marketmbrandt/papers/published/cti.pdfwiley periodicals,...

Documents