i i

Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 1 le-tex

The Microstructure of the US Treasury Market 1

On-the-run On-the-run refers to the most recently auc- 45

The Microstructure tioned Treasury security of a particular maturity. After 46
of the US Treasury Market the next auction, the security goes o-the-run. 47

1 BRUCE MIZRACH1 , CHRISTOPHER J. N EELY2 Price discovery The process by which prices adapt to new 48

2
1 Department of Economics, Rutgers University, information. 49

3 New Brunswick, USA Primary dealers Primary dealers are large brokerage 50

4
2 Research Department, Federal Reserve Bank rms and investment banks that are permitted to trade 51

5 of St. Louis, St. Louis, USA directly with the Federal Reserve in exchange for mak- 52

ing markets in Treasuries. They provide the majority 53

of liquidity in the Treasury market, participate in Trea- 54

6 Article Outline sury auctions, and provide information to assist the 55

7 Glossary Fed in implementing open market operations. 56

8 Denition of the Subject Secondary market After the initial auction of Treasury 57

9 Introduction instruments, trading in on-the-run and o-the-run se- 58

10 Types of Treasury Issues curities makes up the secondary Treasury market. 59

11 Treasury Market Participants When issued When-issued bonds are those Treasuries 60

12 Stages of the Treasury Bond Market whose auctions have been announced but have not yet 61

13 The Treasury Futures Market settled. 62

14 Seasonality and Announcement Eects

15 Discontinuities in the US Treasury Market
16 Order Flow in the US Treasury Market
Definition of the Subject 63
17 Modeling the Limit Order Book
18 Price Discovery This article discusses the microstructure of the US Trea- 64

19 Future Directions sury securities market. 65

20 References US Treasury securities are default risk free debt instru- 66

ments issued by the US government. These securities play 67

an important, even unique, role in international nancial 68

21 Glossary markets because of their safety, liquidity, and low transac- 69

22 Algorithmic trading Algorithmic trading is the practice tions costs. Treasury instruments are often the preferred 70

23 of automatically transacting based on a quantitative safe haven during nancial crises, a process often referred 71

24

25

26
model.
Broker A broker is a rm that matches buyers and sellers
in nancial transactions. An interdealer broker (IDB)
to as a ight to quality.
According to the US Treasury, there was more than
$9 trillion in US government debt outstanding as of Au-
72

73

74

o f
27

28

29
is an intermediary providing trading services to hedge
funds, institutions, and other dealers. IDBs handle the
majority of Treasury securities transactions in the sec-
gust 31, 2007. Of this quantity, the public holds more than
$5 trillion and $4.5 trillion is tradable on nancial mar-
kets. Foreigners hold approximately $2.4 trillion of the
r o 75

76

77

30

31
ondary market.
Coupons Owners of Treasury notes and bonds receive
marketable supply, with Japan and China together holding
more than $1 trillion. According to the Securities Indus-
P 78

79

d
32 periodic payments called coupons. They are xed by try and Financial Markets Association (SIFMA), average 80

33 the Treasury at auction and are typically paid semi-an- daily trading volume in the US Treasury market in 2007 81

34

35
nually.
Depth Depth is the quantity the dealer is willing to sell at
was $524.7 billion.

t
Microstructure is the study of the institutional details

3 e 1 82

83

c 07
the bid or oer. of markets and trading behavior. Microstructural analysis
-
36 84

37 Electronic communications networks (ECN) The Secu- takes three ideas seriously that are often overlooked: the 85

38

39

40
rities and Exchange Commission denes electronic
communications networks (ECNs) as electronic trad-
ing systems that automatically match buy and sell or-
ders at specied prices.
e
institutional features of the trading process inuence how

rr 00 8 -
private information is impounded into prices; agents are
heterogeneous; and information is asymmetric. Empirical
microstructure research studies topics such as the causes
86

87

88

o
41 89

42 Market microstructure Market microstructure is a eld

2
and eects of market structure, how market structure in- 90

c
43 of economics that studies the price formation process uences price discovery, how trading and order ow reveal 91

44 and trading procedures in security markets. private information, how quickly public information is 92

n
Please note that the pagination is not nal; in the print version an entry will in general not start on a new page.

U
i i
Editors or typesetters annotations (will be removed before the nal TEX run)
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 2 le-tex

2 The Microstructure of the US Treasury Market

93 impounded into prices, the volatility-volume relation, and also pay semi-annual coupons, and make up 12:5% of the 135

94 the determinants of transactions costs (i. e., the compo- US debt. 136

95 nents of bid-ask spreads). The relatively recent availability The price of both notes and bonds are quoted as a per- 137

96 of tick-by-tick nancial data and limit order book data, as centage of their face value in thirty-seconds of a point. 138

97 well as the computer resources to manipulate them, have A quoted price of 98-08 means that the quoted price 139

98 been a great boon to nancial market microstructure re- of the note (or bond) is (98 C 8/32 D) $98.25 for each 140

99 search. $100 of face value. The cash price of bonds and notes 141

is equal to the quoted price plus accrued interest since 142

100 Introduction the last coupon payment, calculated with an actual/actual 143

day count convention. Quoted prices are sometimes called 144

101 We begin by describing the types of Treasury issues and
clean prices, while cash prices are said to be dirty. 145
102 the major Treasury market participants, including the
The US Treasury also issues 5-, 10-, and 20-year Trea- 146
103 Federal Reserve, primary dealers and the major electronic
sury InationProtected Securities (TIPS), whose pay- 147
104 brokers. We then outline the stages of the Treasury mar-
o is linked to changes in the US Consumer Price Index 148
105 ket, from auction announcements to the secondary mar-
(CPI). These make up about 10:2% of the total value of 149
106 ket. Next, we examine several closely related areas of the
Treasuries outstanding. The principal value of TIPS is ad- 150
107 literature: Seasonality in the Treasury market and the re-
justed daily and the semi-annual coupon payments and 151
108 actions of the Treasury market to macro and monetary an-
principal payment are then based on the adjusted principal 152
109 nouncements; discontinuities in Treasury prices; and the
amount. Economists extract ination forecasts by compar- 153
110 eect of order ow in Treasury markets. We then discuss
ing the TIPS yields to those on similar nominal instru- 154
111 modeling and other academic questions about the Trea-
ments. The Federal Reserve Bank of Saint Louis provides 155
112 sury market.
TIPS spreads through its publication, Monetary Trends. 156

There is also an active market in STRIPS (Separate 157

113 Types of Treasury Issues Trading of Registered Interest and Principal of Securi- 158

114 As of October 2007, the US Treasury issued four types of ties) which are popularly known as zero coupon bonds. 159

115 debt instruments. The shortest-maturity instruments are These instruments are created by the Treasury through an 160

116 known as Treasury bills. 22:6% of the marketable US debt accounting system which separates coupon interest pay- 161

117 is in bills, securities with maturities of 1 year or less. Bills ments and principal. Finally, the US Treasury also issues 162

118 are sold at a discount and redeemed at their face value at savings bonds, low denomination securities for retail in- 163

119 maturity. They do not pay any coupons prior to maturity vestors. 164

120

121

122
and currently have maturities up to 26 weeks. Treasury bill
prices are usually quoted in discount rate terms, which
are calculated with an actual/360 day count convention, Treasury Market Participants 165

o f
123
T-bill discount rate D [face value bill price]

(360/number of days until maturity) :
The Federal Reserve in the Treasury Market
The Federal Reserve Bank of New York, under the guid-
r o 166

167

124

125
Thus, a bill with a face value of $100,000, a cash price of
$97,500 and 90 days to maturity will have a discount rate of P
ance of the Federal Open Market Committee (FOMC), is
a uniquely important player in the Treasury market. The
168

169

d
FOMC meets approximately every six weeks to review eco- 170
126 10% D [100  97:5]
(360/90) in a newspaper. Treasury
nomic conditions and determine a target for the federal 171
bill yields are often quoted as bond equivalent yields,

e 1
127
funds rate, the rate at which US banks borrow/lend reserve 172

t
128 which are dened as,
balances from/to each other. The manager of the Open 173


face value  bill price
3
c 07
Market Desk (a.k.a., the Desk) at the Federal Reserve
-
174
T-bill yield D
129 bill price Bank of New York is responsible for ensuring that the 175

130

(365/number of days until maturity) :

Treasury instruments with intermediate maturities (2-,

5- and 10-year) are known as Treasury notes. Notes pay
e
average federal funds transaction is close to the target by

rr 00 8 -
buying and selling Treasury instruments (primarily short-
term). In practice, the Desk accomplishes this in two ways.
First the Desk buys sucient Treasuries to satisfy most
176

177

178

o
131 179

132 semi-annual coupons, and make up 54:7% of the debt.

2
but not all the markets demand for deposits at the Fed. 180

c
133 In February 2006, the US Treasury also resumed issu- Secondly, the Desk buys Treasuries via repurchase (repos) 181

134 ing 30-year instruments, known as Treasury bonds. Bonds agreements (overnight and for terms of several days) to 182

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 3 le-tex

The Microstructure of the US Treasury Market 3

183 achieve a desired repo rate that inuences the federal funds sis that are helpful in the formulation and implementation 232

184 rate and other short-term interest rates through arbitrage. of monetary policy. The Federal Reserve does not regulate 233

185 To determine day-to-day actions, every morning, sta primary dealers, but does subject them to capital require- 234

186 at both the Division of Monetary Aairs of the Board of ments. The Federal Reserve can withdraw a rms primary 235

187 Governors of the Federal Reserve System and the Desk dealer designation if it fails to participate in auctions or 236

188 forecast that days demand for reserve balances. The Desk open market operations or if its capital reserves fall below 237

189 sta also consults market participants to get their views on desired levels. 238

190 nancial conditions. The relevant Desk and Board stas The daily average trading volume in US Government 239

191 then exchange views in a 9 am conference call. Finally, securities of all the primary dealers was approximately 240

192 the relevant Desk sta, the Board sta, and at least one $550 billion during 2005. 241

193 of the voting Reserve Bank Presidents then confer dur-

194 ing a second conference call at about 9:20 am. The Desk Interdealer Brokers 242
195 sta summarizes market conditions, projects actions for
196 the day and asks the voting Reserve Bank President(s) for Prior to 2000, voice-assisted brokers dominated secon- 243

197 comments. Open market operations commence shortly af- dary market trading in Treasuries. Except for CantorFitz- 244

198 ter the conclusion of this call. gerald, all these brokers reported their trading activity to 245

199 When the Desk buys Treasuries, it increases available GovPX, a consortium. In the face of demands by the Secu- 246

200 liquidity (reserves) in debt markets and tends to lower in- rities and Exchange Commission and bond market deal- 247

201 terest rates. Selling Treasuries has the opposite eect, low- ers for greater transparency, ve IDBs formed GovPX as 248

202 ering reserves and raising interest rates. If the intention a joint venture in 1991. In March 1999, CantorFitzgerald 249

203 is to make a permanent change in reserves, then outright opened up its internal electronic trading platform, eSpeed, 250

204 purchases or sales are undertaken. In contrast, if the Desk to clients. The eSpeed system quickly grabbed a dominant 251

205 anticipates that only temporary changes in reserves are market share, and Cantor Fitzgerald spun o eSpeed as 252

206 necessary, it uses repos (for purchases) or reverse repos a public company in December 1999. In 2000, a competing 253

207 (for sales). Bernanke [52] notes that actual open market electronic brokerage, BrokerTec, joined the market. As in 254

208 sales of debt instruments are rare; it is more common for foreign exchange and equity markets, most interdealer and 255

209 the Federal Reserve to allow such securities to expire with- institutional trading in Treasuries quickly migrated from 256

210 out replacing them. Both open market sales and allowing voice networks to these electronic communications net- 257

211 the Feds securities to expire have the same balance sheet works (ECNs), which have dominated trading in Treasury 258

212 eects: The Fed holds fewer bonds and more cash, while instruments since 2001. Mizrach and Neely [4] describe 259

213

214

215
the public will hold more bonds and less cash.
The Federal Reserve provides several valuable refer-
ences on its operating procedures. The Annual Report
the transition from voice assisted trading, largely through
the primary dealers, to electronic trading in the Treasury
market.
260

261

262

o f
o
216 of the Markets Group of the Federal Reserve Bank of As of November 2007, the two dominant ECNs are 263

r
217 New York describes open market operations and current eSpeed and BrokerTec. London-based ICAP, PLC, owns 264

218 procedures (Federal Reserve Bank of New York, Markets BrokerTec while eSpeed merged in the summer of 2007 265

P
219 Group [61]). Meulendyke [63] provides a comprehensive with BGC, another London based interdealer brokerage. 266

220 view of Federal Reserve monetary policy operations with eSpeed and ICAP compete for both on- and o-the-run 267

liquidity. Hilliard Farber and TullettPrebon hold the 268

d
221 a historical perspective. Akhtar [64] explains how mon-
222 etary policy is decided and how such policies aect the largest brokerage share outside of the dominant two plat- 269

e 1
223 economy. Finally, Harvey and Huang [40] gives some his- forms. 270

224 torical perspective on operating procedures in the 1980s.

Stages of the Treasury Bond Market
t
c 07 - 3
The sale of Treasuries undergoes four distinct phases:
271

rr e200
225 Primary Dealers 272

when issued, primary, on-the-run and o-the-run. Each

8-
273
226 Among the most important private sector players in the of these stages has a distinct market structure. 274
227 Treasury markets are the 21 primary dealers. The Federal
Reserve Bank of New York explains that primary deal-

o
228

229 ers must participate meaningfully in both the Feds open The Primary Market 275

c
230 market operations and Treasury auctions and . . . provide In the primary market, the US Treasury sells debt to the 276

231 the Feds trading desk with market information and analy- public via auction. The US Treasury usually publishes 277

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 4 le-tex

4 The Microstructure of the US Treasury Market

278 a calendar of upcoming tentative auction dates on the Fabozzi and Fleming [6] estimate that 6% of total inter- 326

279 rst Wednesday of February, May, August, and Novem- dealer trading is in the when-issued market. Just prior to 327

280 ber and bids may be submitted up to 30 days in advance auctions though, these markets become substantially more 328

281 of the auction. In practice, however, the Treasury only an- active. In the bill market, when-issued trading volume ex- 329

282 nounces rm auction information several days in advance ceeds the volume for the bills from the previous auction. 330

283 and most bids are submitted at that time. Since August 8,
284 2002, the Treasury has made auction announcements (for On-the-Run Upon completion of the auction, the most 331

285 all new securities) at 11:00 am Eastern Time (ET). 13- and recently issued bill, note or bond becomes on-the-run and 332

286 26-week bills are auctioned weekly; 2- and 5-year notes the previous on-the-run issue goes o-the-run. Overall 333

287 are auctioned monthly; 10-year notes are auctioned eight Treasury trading volume is concentrated in a small num- 334

288 times a year. 30-year bonds, which were reintroduced on ber of on-the-run issues. Trading in these benchmark on- 335

289 February 9, 2006 after a ve year hiatus, are auctioned four the-run issues, which Fabozzi and Fleming [6] say con- 336

290 times a year. stitutes approximately 70% of total trading volume, has 337

291 The US Treasury has used a single price auction ex- migrated almost completely to the electronic networks. 338

292 clusively since November 1998. Garbade and Ingber [5] Mizrach and Neely [4] estimate a 61% market share for the 339

293 discuss the transition from multiple price auctions to the BrokerTec platform and a 39% share for eSpeed in 2005, 340

294 current format single price auctions. All securities are al- which is consistent with industry estimates. 341

295 located to bidders at the price that, in the aggregate, will

296 result in the sale of the entire issue. This mitigates the O-the-Run With more than 200 o-the-run issues 342

297 risk of a buyers curse the highest bidder paying more trading in October 2007 44 bills, 116 notes, and 343

298 than other auction participants. To prevent a single large 45 bonds most o-the-run volume takes place in voice 344

299 buyer from manipulating the auction, the Treasury re- and electronic interdealer networks. Barclay, Hendershott 345

300 stricts anyone from buying more than 35% of any single and Kotz [1] document the fall in ECN market share when 346

301 issue. Bids may be submitted up to thirty days prior to issues go o the run. They also report that transaction 347

302 the auction, and large institutions make use of the Trea- volume falls by more than 90%, on average, once a bond 348

303 sury Automated Auction Processing System (TAAPS). Re- goes o-the-run. The ECN market share falls from 75:2% 349

304 tail investors can participate through the Treasury Direct to 9:9% for the 2-year notes, from 83:5% to 8:5% for the 5- 350

305 program. The Treasury allocates a portion of nearly every year notes, and from 84:5% to 8:9% for the 10-year notes. 351

306 auction to small investors at the same price as the large in- Several IDBs handle most o-the-run securities trading. 352

307 stitutions. These are called non-competitive bids, and they

308

309

310
are quantity only orders that are lled at the market clear-
ing price.
Primary dealers dominate the auction process. In 2003,
On- versus O-the-Run Liquidity and Prices O-the-
run securities trade at a higher yield (lower price) than on-
the-run securities of similar maturity. Many researchers
353

354

355

o f
ro
311 they submitted 86% of auction bids, totalling more than have attempted to explain the yield dierential with rel- 356

312 $6 trillion. They were awarded $2.4 trillion, or 78% of the ative liquidity. Vayanos and Weill [7] utilize a search the- 357

313 total auction supply. oretic model that is motivated by the fact that bonds may 358

314 The Secondary Market

Hanke, and Nath [36] compare on-the-run and o-the-
P
be dicult to locate once they go o-the-run. Goldreich, 359

360

d
run Treasuries and show that the liquidity premium de- 361

315 The secondary market is composed of the when-issued, pends primarily on the amount of remaining future liquid- 362

316 on-the-run and o-the-run issues.

t e 1
ity, which is highly predictable. The study exploits the fact
that the liquidity of a Treasury is predictable. Due [18]

3
363

364

c 07
When-Issued Even prior to the primary auction, there argues that legal or institutional restrictions on supplying
-
317 365

318 is an active forward market in Treasury securities (apart collateral induces special repo rates that are much less 366

319

320

321
from TIPS) that are about to be issued. Trading in the
when-issued security market typically begins several days
prior to an auction and continues until settlement of auc-
tion purchases. Nyborg and Sundaresan [3] document
e
than market riskless interest rates. The price of the under-

rr 00
-
lying instrument is increased by the present value of the
8
savings in borrowing costs.
367

368

369

o
322

323 that when-issued trading provides important information

2
Supply Variation and Prices Although it is generally 370

c
324 about auction prices prior to the auction and also permits accepted that the on-the-run premium is due to greater 371

325 market participants to reduce the risk they take in bidding. liquidity, the theoretical relation between the supply of 372

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 5 le-tex

The Microstructure of the US Treasury Market 5

The Microstructure of the US Treasury Market, Table 1

Contract Details from the CBOT Treasury Market

Contract Quote Pricing example Deliverable asset characteristics

conven-
tion
2-year 1/32 and 95
060 D 95 C 6/32 US Treasury notes with a face value  $200,000 and
quarters 95
062 D 95 C 6:25/32 original maturity  5 years and 3 months and
of 32nds 95
065 D 95 C 6:5/32 remaining maturity  1 year and 9 months from the first day of the delivery month and
95
067 D 95 C 6:75/32 and remaining maturity  than 2 years from the last day of the delivery month.
5-year 1/32 90
170 D 90 C 17/32 US Treasury notes with a face value  $100,000 and
and 90
175 D 90 C 17:5/32 original maturity  5 years and 3 months and
halves remaining maturity  4 year and 2 months from the first day of the delivery month
of 32nds
10-year 1/32 90
170 D 90 C 17/32 US Treasury notes with a face value  $100,000 and
and 90
175 D 90 C 17:5/32 remaining maturity  10 years
halves remaining maturity  6 year and 6 months from the first day of the delivery month
of 32nds
30-year 1/32nds 85
12 D 85 C 12/32 US Treasury bonds with a face value  $100,000 and
if callable: Not callable for at least 15 years from the first day of the delivery month;
if not callable: Remaining maturity  15 years from the first day of the delivery month.

373 a given bond issue and prices is not clear. Do issue A variety of Treasury instruments meet the criteria to 402

374 sizes produce lower yields (higher prices) through their be deliverable issues. Table 1 describes the pricing conven- 403

375 liquidity eects or whether downward-sloping demand tions and the characteristics of the assets that may be de- 404

376 for individual securities would produce higher prices livered to satisfy the contracts. The CBOT denes con- 405

377 (lower yields) for larger issues? Empirically, the evidence version factors that adjust the quoted futures prices for 406

378 is mixed. Simon [65,66], Due [18], Seligman [64] and the asset that is actually delivered. Despite these conver- 407

379 Fleming [29] nd that the larger issues lead to lower prices sion factors, one issue will be the cheapest to deliver. 408

380 (higher yields), while Amihud and Mendelson [2], Ka- Cash prices at delivery depend on both the conversion fac- 409

381 mara [51], Warga [69], and Elton and Green [23] nd the tor for a particular bond and the interest accrued on that 410

opposite: The liquidity eect predominates, resulting in bond since the last coupon payment.

f
382 411

383 higher prices (lower yields) for larger issues. There might Although agents frequently use the futures markets for 412

o
384 be a nonlinear relationship. Liquidity may increase prices hedging or taking positions on future price movements, 413

385 up to a certain point, but then nite demand for any in- only a modest amount of microstructure research has fo- 414

386

387
dividual security reduces the attractiveness of additional
supply.
cused on futures markets. Brandt, Kavajecz, and Under-
wood [27] show that futures and spot market order ow
are useful in predicting daily returns in each market and
r o 415

416

417

388 The Treasury Futures Market

that the type of trader inuences the eect of order ow.
Mizrach and Neely [12] show that futures markets con-
P 418

419

d
389 Spot markets are not the only markets for US Treasuries. tribute a substantial amount of price discovery to US Trea- 420

390 The Chicago Board of Trade (CBOT) has active futures sury markets. Campbell and Hendry [51] compare price 421

391

392
markets for 2-, 5-, 10- and 30-year US Treasuries. Table 1
briey describes the CBOT contracts and pricing conven- both the United States and Canada.

t 3 e 1
discovery in the 10-year bond and futures contracts in 422

423

c 07
tions.
-
393

394 Like other exchange-traded derivatives, Treasury fu- Seasonality and Announcement Effects 424

rr e 00
395 tures have two advantages: trading is highly liquid and

8-
Seasonality and announcement eects are intimately re- 425
396 marking-to-market minimizes counterparty risk. The
lated to the microstructure literature in that the latter seeks 426
397 CBOT open auction trading hours are 7:20 am to 2:00 pm,
to explain how markets with heterogeneous agents react to 427
Central Time, Monday through Friday; the CBOT elec-

o
398
the release of information.
399 tronic market functions from 6:00 pm to 4:00 pm, Central
2 428

c
400 Time, Sunday through Friday. All Treasury contracts have
401 a MarchJune-SeptemberDecember cycle.

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 6 le-tex

6 The Microstructure of the US Treasury Market

429 Seasonality and Macroeconomic Announcements termined by future expected dividends. Macro announce- 479

ments can have little or no eect on stock prices if their ef- 480

430 The earliest studies considered the issue of daily seasonal- fects on expected dividends and discount rates oset each 481

431 ity in Treasuries. Flannery and Protopapadakis [34] docu- other. 482

432 ment diering day-of-the-week patterns in Treasuries and Several studies used more sophisticated econometric 483

433 stock indices. The patterns in the prices of Treasuries se- procedures to evaluate the impact of announcements on 484

434 curities vary by maturity and dier from those found in persistence in volatility in a full model. Jones, Lamont 485

435 stock indices. They conclude that no single factor explains and Lumsdaine [44] examine volatility patterns in the 486

436 seasonal patterns across asset classes. In contrast to this 5-year Treasury market around US announcements. Daily 487

437 day-of-the-week eect in spot T-bills, Johnston et al. [45] volatility from an ARCH-M does not persist for days after 488

438 nd day-of-the-week eects in government national mort- announcements and the authors interpret this as indicat- 489

439 gage association (GNMA) securities, T-note, and T-bond ing that agents rapidly incorporate announcement infor- 490

440 futures, but not in T-bill futures. The fact that day-of-the- mation into prices. Weekly volatility displays a U-shaped 491

441 week eects exist in spot T-bills but not in T-bill futures pattern; the largest price changes occur on Mondays and 492

442 points up the importance of futures settlement rules. Fridays. Further, Jones, Lamont and Lumsdaine [44] nd 493

443 Later studies began to consider the eects of macro a risk premium in returns on days of announcements. 494

444 announcements on price changes, volatility, volume and Bollerslev, Cai, and Song [24] also consider the interac- 495

445 spreads. Macroeconomic announcements have been an es- tion of announcements and persistence in volatility with 496

446 pecially popular subject of study because they occur at reg- 5-minute US Treasury bond data. Modeling the intraday 497

447 ular intervals that can be anticipated by market partici- volatility patterns and accounting for announcements re- 498

448 pants. The existence of survey expectations about upcom- veals long-memory in bond market volatility. 499

449 ing macro announcements permits researchers to identify An important issue in microstructure is the determi- 500

450 the shock component of the announcement, which al- nation of bid-ask spreads. Balduzzi, Elton, and Green [22] 501

451 lows them to investigate the dierential eects of antici- use intraday GovPX data to look at the eects of macro an- 502

452 pated and unanticipated news releases of dierent magni- nouncements on volume, prices and spreads. Conrming 503

453 tudes. previous ndings, prices adjust to news within one minute 504

454 Ederington and Lee [32,33] did the seminal modern while increases in volatility and volume persist for up to 505

455 work with intraday data on macro announcement eects 60 minutes. Spreads initially widen but then return to nor- 506

456 in bond markets. They found that volatility increases be- mal after 5 to 15 minutes. News releases explain a substan- 507

457 fore the announcement and remains elevated for some tial amount of bond market volatility. Importantly, Bal- 508

458

459

460
time afterwards. The employment, PPI, CPI and durable
goods orders releases produce the greatest impact of the
9 signicant announcements, out of 16 studied. Eder-
duzzi, Elton, and Green [22] argue that the dierential
impact of news on long and short bond prices indicates
that at least two factors will be needed for models of the
509

510

511

o f
461

462

463
ington and Lee [50] follow up on their earlier studies
by linking the literatures on seasonality and announce-
ments in the bond market. Comparing the contributions
(jumps) will be important in modeling bond prices.

r
Some recent papers have relaxed the restrictive as- o
yield curve. They also present evidence that discontinuities 512

513

514

464

465
of past volatility, seasonality and announcements in pre-
dicting intraday volatility bond futures data and exchange
sumption that announcements inuence Treasury mar-

P
ket variables in a linear, symmetric fashion. For example,
515

516

d
466 rates, these authors argue that announcements account for ChristieDavid, Chaudhry, and Lindley [29] allow the ef- 517

467 much of the apparent seasonality in interest rate volatility. fects of announcement shocks to depend on the size and 518

468

469
One of the earliest important results was that bond
market prices react more strongly to macro announce-

t e 1
sign of the shock. They measure these nonlinear eects on
the intraday 10- and 30-year Treasury futures from 1992

3
519

520

c 07
ments than do equity markets. Fleming and Remolona [37, to 1996.
-
470 521

471 38] examined the 25 largest price changes in the GovPX Most studies of the eects of volatility have mea- 522

472

473

474
data and related them all to macroeconomic announce-
ments. Fleming and Remolona [37] note: In contrast to
stock prices, US Treasury security prices largely react to
the arrival of public information on the economy. Flem-
e
sured such variation with some function of squared re-

rr 00 8 -
turns. One can use the volatility implied by options prices,
however, to measure expected volatility over longer hori-
zons. Heuson and Su [41], for example, show that implied
523

524

525

o
475 526

476 ing and Remolona [36,38] attribute the relative sensitivity

2
volatilities from options on Treasuries rise prior to macro 527

c
477 of bond markets to the fact that bond prices depend only announcements and that volatilities quickly return to nor- 528

478 on expected discount rates while stock prices are also de- mal levels after announcements. Beber and Brandt [23] use 529

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 7 le-tex

The Microstructure of the US Treasury Market 7

530 intraday, tick data from 1995 to 1999 to determine that tainty about the dates of target changes and the eect of 577

531 macro announcements reduce the variance of the option- learning by market participants. 578

532 implied distribution of US Treasury bond prices. The con- Poole and Rasche [56] also decompose federal funds 579

533 tent of the news and economic conditions explain these target changes into expected and unexpected compo- 580

534 changes in higher-order moments. The study attributes nents but use a later contract month than Kuttner [55] to 581

535 the results to time-varying risk premia rather than relative avoid problems associated with computation of the con- 582

536 mispricing or changing beliefs. tract payo. They nd that interest rates across the ma- 583

537 In a comprehensive study of the impact of US macroe- turity spectrum fail to respond to the anticipated compo- 584

538 conomic announcements across asset markets, Andersen, nents of the changes in the intended funds rate. 585

539 Bollerslev, Diebold and Vega [21] study the reaction of Poole, Rasche and Thornton [57] consider how 586

540 international equity, bond and foreign exchange markets. changes in FOMC procedures aect the impact of target 587

541 They conrm that US macroeconomic news drives bond changes on interest rates. This study rst succinctly de- 588

542 prices, as well as those of the other assets. scribes the changes in FOMC procedures in the 1990s. The 589

FOMC began to contemporaneously announce policy ac- 590

tions in 1994 and adopted this as formal policy in 1995. 591

Starting in August 1997, each policy directive has included 592

543 Monetary Policy Announcements
the quantitative value of the intended federal funds rate. 593

544 Researchers have carefully investigated the eects of the And since 1999, the FOMC has issued a press release af- 594

545 Federal Reserves actions on the Treasury market. While ter each meeting with the value for the intended federal 595

546 the literature has examined the eect of a wide variety funds rate and, in most cases, an assessment of the bal- 596

547 of monetary policy behavior and communications e. g., ance of risks. After describing such procedural changes, 597

548 open market operations, FOMC news releases, speeches, Poole, Rasche and Thornton [57] go on to consider the 598

549 etc. on many aspects of Treasury market behavior, a large response of the Treasury yield curve to funds rate target 599

550 subset of these papers deal with one specic topic: The ef- changes both before and after the FOMC began contem- 600

551 fect of federal funds target changes on the Treasury yield poraneously announcing target changes in 1994. In doing 601

552 curve. so, these authors account for measurement error in expec- 602

tations and uncertainty about the dates of target changes 603

553 Federal Funds Target Changes and the Treasury Yield and even whether market participants understood that the 604

554 Curve The expectations hypothesis of the term struc- Federal Reserve was targeting the funds rate prior to 1994. 605

555 ture motivates research on how the short- and long-end They assess the markets knowledge of targeting by exam- 606

556

557

558
of the Treasury yield curve react to unexpected changes in
the federal funds target rate. That is, if the FOMC increases
overnight interest rates, how does this change short- and
ining news reports. While short-rates respond similarly in
both subperiods, long rates do not respond as strongly to
funds rate target changes after 1994. The authors inter-
607

608

609

o f
ro
559 long-term rates? pret their results as being consistent with the Feds greater 610

560 Using data on 75 changes in the federal funds tar- transparency about long-run policy in the second subsam- 611

561 get from September 1974 through September 1979, Cook ple. With long-run expectations more rmly anchored, 612

562

563
and Hahn [53] nd that these target changes caused larger
movements in short-term rates than in intermediate- and on long rates.
P
unexpected changes in the funds target have smaller eects 613

614

d
564 long-term Treasury rates. A diculty with interpreting the One puzzle that has emerged from this literature is that 615

565 Cook and Hahn [53] results is that ecient markets pre- the average eect of changes in the federal funds target 616

566

567
sumably can often anticipate most or all of a target change
and such expectations are already incorporated into the

t e 1
on the yield curve is modest, despite the facts that such
changes should be an important determinant of the yield

3
617

618

c 07
yield curve. To confront this problem, Kuttner [55] de- curve and that yields are highly volatile around FOMC an-
-
568 619

569 composes target changes into anticipated and unantici- nouncements. Fleming and Piazzesi [35] claim to partially 620

570

571

572
pated components, nding unsurprisingly that Trea-
sury rates respond much more strongly to unanticipated
changes and that the results are consistent with the expec-
tations hypothesis of the term structure. That is, the antic-
e
resolve this puzzle by illustrating that such yield changes

rr 00 8 -
depend on the shape of the yield curve.
This literature on the reaction of the Treasury market
to monetary policy has become progressively more sophis-
621

622

623

o
573 624

574 ipated component of an interest rate change does not af-

2
ticated in assessing market expectations of Fed policy and 625

c
575 fect expectations. Hamilton [54] carefully reexamines the modeling institutional features of the futures market and 626

576 work of Kuttner [55], showing that it is robust to uncer- Fed operations. Nevertheless, the underlying conclusion 627

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 8 le-tex

8 The Microstructure of the US Treasury Market

628 that unanticipated target changes lead to large price in- consider whether these Federal Reserve events merely cre- 678

629 creases on short-term Treasuries and smaller changes on ate noise or transmit information about the future policy 679

630 the prices of long-term Treasuries has been remarkably ro- decisions or the state of the economy. They conclude that 680

631 bust. such events may reduce welfare by overwhelming private 681

information, creating herding behavior. 682

632 Other Federal Reserve Behavior and the Treasury Mar-

633 ket There has been a substantial literature analyzing how
634 other types of Federal Reserve behavior have inuenced
Announcements and Liquidity Variation 683
635 the Treasury market. The literature has considered open
636 market operations, FOMC statements, Congressional tes- The literature on variation in liquidity and price eects 684

637 timonies, and FOMC member speeches. overlaps with the literature on macroeconomic announce- 685

638 Open market operations are similar to macroeco- ments. The seminal work of Amihud and Mendelson [2] 686

639 nomic announcements in that they are potentially impor- showed that yields on short-time-to-maturity Treasuries 687

640 tant bond market events, occurring at regularly scheduled vary inversely with liquidity. That is, more liquid assets 688

641 times. Harvey and Huang [40] used intraday data from have lower yields/higher prices. Harvey and Huang [40] 689

642 1982 to 1988 to examine how Federal Reserve open market discovered elevated volatility in interest rate (and for- 690

643 operations inuenced foreign exchange and bond markets. eign exchange) futures markets, in the rst 6070 min- 691

644 The paper nds that Treasury market volatility increases utes of trading on Thursdays and Fridays. Ederington and 692

645 during open market operations, irrespective of whether Lee [32] conrmed Harvey and Huang [40]s speculation 693

646 they add or drain reserves. Oddly, volatility increases even that major macroeconomic announcements especially 694

647 more during the usual time for open market operations if the employment report, the PPI, the CPI, and durable 695

648 there are no such transactions. The authors interpret this goods orders create the intraday and intraweek patterns 696

649 nding as indicating that open market operations actually in the volatility of Treasury bond futures. Volatility is very 697

650 smooth volatility. high after announcements and remains elevated for hours. 698

651 Early studies made the simplifying assumption that the Fleming and Remolona [38] extend this work to show 699

652 eect of macro announcements on the Treasury market that the 25 greatest surges in activity in the 5-year on- 700

653 was constant over time. This is not necessarily the case, of the-run bond market came on macroeconomic announce- 701

654 course. For example, the eect of macro announcements ment days, within 70 minutes of the announcement. The 702

655 on the Treasury market might depend on monetary pol- most important announcements for trading surges were 703

656 icy priorities. Kearney [46] characterizes the changing re- employment reports, fed funds targets, 30-year auctions, 704

657

658

659
sponse of daily 3-month Treasury futures to the employ-
ment report over 1977 to 1997 and relates it to the chang-
ing importance of employment in the Feds reaction func-
10-year auctions, the CPI, NAPM surveys, GDP, retail
sales, and 3-year auctions. Releases that aect prices also
matter for trading activity. Fleming and Remolona [38]
705

706

707

o f
ro
660 tion. observe that timeliness, the degree of surprise in the an- 708

661 de Goeij and Marquering [30] also considers how both nouncement and market uncertainty also increase an- 709

662 macro announcements and monetary policy events af- nouncements impact on trading. 710

663

664
fect the US Treasury market. Using daily data from 1982
to 2004 de Goeij and Marquering [30] nd that macro
P
Researchers continued to explore the impact of vari-
ation in liquidity caused by other events. For example,
711

712

d
665 news announcements strongly aect the daily volatility of Fleming [28] exploits exogenous variation in Treasury is- 713

666 longer-term Treasury instruments while FOMC events af- suance to show that securities that are reopened the 714

667

668
fect the volatility of shorter-term instruments.
Some studies have explored more esoteric components

t e 1
Treasury sells additional quantities of existing securities
have greater liquidity, lower spreads, than comparable as-

3
715

716

c 07
of information about monetary policy. Boukus and Rosen- sets. Paradoxically, this higher liquidity does not produce
-
669 717

670 berg [25], for example, use Latent Semantic Analysis to lower yields for the reopened securities. 718

671

672

673
decompose the information content of FOMC minutes
from 1987 to 2005. They then relate the information con-
tent to current and future economic conditions. Chirinko
and Curran [28] argue that Federal Reserve speeches, tes-
e
More recent papers have explored variation in liquid-

rr 00 8 -
ity and volatility across markets. Chordia, Sarkar and Sub-
rahmanyam [14] estimate a vector autoregression (VAR)
in liquidity and volatility variables in stock and bond mar-
719

720

721

o
674 722

675 timonies, and meetings increase price and trading volatil-

2
kets. They nd that common factors make the variables 723

c
676 ity on the 30-year bond market. FOMC meetings are the innovations highly correlated. Volatility shocks predict 724

677 most important of the events considered. They go on to liquidity variables. 725

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 9 le-tex

The Microstructure of the US Treasury Market 9

726 End-of-the-Year Patterns in One-Month Treasury Bills Dungey, McKenzie, and Smith [31] estimate jumps 774

and cojumps (simultaneous discontinuities in multiple 775

727 The previous sets of papers studied daily and intraday sea-
markets) in the term structure of US Treasury rates. They 776
728 sonality, often as caused by macroeconomic or Federal Re-
nd that the middle of the yield curve often cojumps 777
729 serve announcements. Short-term Treasury bills also ex-
with one of the ends, while the ends of the curve exhibit 778
730 hibit year-end seasonality, however. Market participants
a greater tendency for idiosyncratic jumps. Macro news 779
731 consider Treasury market instruments of 30 days or less
is strongly associated with cojumps in the term structure. 780
732 to be highly liquid, close but not perfect substitutes
Using BrokerTec data from 20032005, Jiang, Lo, and 781
733 for cash. The fact that short-term Treasuries are not per-
Verdelhan [66] extend this work by focusing on the role of 782
734 fect substitutes for cash is presumably what allows the New
liquidity shocks estimated from the limit order book 783
735 York Desk to use open market operations to manipulate
in jumps and the relation of jumps to order ow and price 784
736 short-term interest rates through a liquidity eect. A pe-
discovery. 785
737 culiar year-end pattern in one-month Treasury yields re-
Lahaye, Laurent and Neely [47] examine jumps and 786
738 inforces this evidence that such Treasuries are not perfect
cojumps across foreign exchange, stock, gold and 30-year 787
739 substitutes for cash.
Treasury futures. Discontinuities in bond futures prices 788
740 Following on related work of Griths and Win-
were larger but less frequent than those in foreign ex- 789
741 ters [19] in repos, Griths and Winters [18] nd that
change rates and smaller and about as frequent as those 790
742 yields on one month T-Bills (and other one-month secu-
in equity markets. News announcements appear to cause 791
743 rities) increase signicantly at the beginning of December,
many cojumps of bond prices with prices of other types of 792
744 remain high during December, and return to normal a few
assets. 793
745 days before the year-end. This pattern does not exist in
746 three-month T-bills. Neely and Winters [20] nd similar
747 patterns in the one-month LIBOR futures market. Order Flow in the US Treasury Market 794

748 Griths and Winters [17,18,19] explain this Decem-

The eect of order ow on prices has been a popular re- 795
749 ber eect by asserting that a year-end preference for liq-
cent topic in microstructure. Several papers have explored 796
750 uidity drives the year-end surge in short-term interest
the impact of order ow on prices and the ways in which 797
751 rates. Debt holder (lenders in the money markets) start to
macro/monetary announcements inuence these impacts. 798
752 liquidate their one-month securities in the last few days
Huang, Cai, and Wang [42] use intraday 1998 GovPX 799
753 of November to meet cash obligations at the end-of-De-
spot data on the 5-year Treasury note to characterize trad- 800
754 cember. This preference for liquidity drives up one-month
ing patterns of primary dealers, announcement eects and 801

f
755 interest rates for most of December. Liquidity demand re-
volatility-volume relations. The paper nds that both pub- 802
756 turns to normal at the end of December as investors repur-

o
lic information (i. e., announcements) and dealer inven- 803
757 chase short-term instruments, and interest rates return to
tory/order ow aect trading frequency. 804
normal levels.

ro
758
Green [39] uses the Madhavan, Richardson, and Roo- 805

mans [48] model to study the impact of GovPX trading in 806

5-year around announcements. Order ow has its largest 807

Discontinuities in the US Treasury Market

P
759
price impact after larger macro surprises, times of greater 808

760 The literature on discontinuities (or jumps) in Treasury uncertainty about the announcement, and times of high 809

d
761 prices is closely related to the literature on announce- liquidity. Green [39] concludes that order ow does reveal 810

762 ments, as announcements are obvious candidates to ex- information about riskless rates. 811

763

764
plain jumps. Three recent papers have looked at discon-
tinuities in US Treasury prices. Huang [43] estimates daily

t e 1
Brandt and Kavajecz [26] nd that order ow im-
balances can explain up to 26% of the day-to-day varia-

3
812

813

c 07
jumps with bi-power variation on 10 years of 5-minute tion in yields on non-announcement days. In contrast to
-
765 814

766 data on S&P 500 and US T-bond futures to measure the Green [39], they nd that order ow has its strongest im- 815

767

768

769
response of volatility and jumps to macro news. He iden-
ties a major role for payroll news in bond market jumps
by analyzing their conditional distributions and regress-
ing continuous and jump components on measures of dis-
e
pact at times of low liquidity. Brandt, Kavacejz, and Un-

rr 00 8 -
derwood [27] extend the work of Brandt and Kavajecz [26]
to control for trader type and macroeconomic announce-
ments in explaining the impact of bond market order ow
816

817

818

o
770 819

771 agreement and uncertainty concerning future macroeco- on futures prices.

2 820

c
772 nomic states. Huang [43] also nds that the bond market Menkveld, Sarkar, and Van der Wel [49] conrm ear- 821

773 is relatively more responsive than the equity market. lier conclusions that announcements have signicant ef- 822

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 10 le-tex

10 The Microstructure of the US Treasury Market

823 fects on 30-year Treasury yields and they also nd that cus- 88% lower in the more liquid eSpeed ECN market. Flem- 867

824 tomer order ow is much more informative on announce- ing and Mizrach [58] nd further reductions in market im- 868

825 ment days than on non-announcement days. They go on pacts on the BrokerTec ECN for 2005 and 2006. 869

826 to investigate the prots that dierent types of traders

827 make on announcement and non-announcement days. Price Discovery 870
828 At high frequencies, order ow is highly autocorre-
A crucial issue in the market microstructure literature is 871
829 lated. A dynamic analysis of the market resilience requires
price discovery. This is the process by which prices embed 872
830 modeling this formally. We turn to empirical modeling of
new information. In the Treasury market, price discovery 873
831 the Treasury market order book in the next section.
occurs in both the secondary spot market and in the fu- 874

tures markets at the Chicago Board of Trade (CBOT). The 875

832 Modeling the Limit Order Book degree to which each market contributes to price discovery 876

833 A purchase or a sale of a Treasury bond inuences prices is a natural issue to address. 877

834 directly as trades work their way up the supply or demand To investigate relative price discovery in these two 878

835 curves. We would like to know whether these eects are Treasury markets, Mizrach and Neely [12] follow Has- 879

836 large and long-lasting. To address this question, we must brouck [11] and assume that the price series have a unit 880

837 introduce a dynamic model of the limit order book. root, are cointegrated, and have an rth order VAR repre- 881

838 Hasbrouck [11] proposed to study intra-day price for- sentation, 882

839 mation with a standard bivariate vector autoregressive

p t D 1 p t
1 C 2 p t
2 C    C r p t
r C u t : 883
840 (VAR) model. Time t here is measured in 1-minute in-
841 tervals. Let rt be the percentage change in the transaction It follows that the N returns, 884
842 price and x 0t be the sum of signed trade indicators ( +1 for 2 3
843 buyer initiated, 1 for seller initiated) over minute t. Trea- p1;t  p1;t
1
sury market data sets typically indicate trade initiation as 6 :: 7
844
rt D 4 : 5 D
p t ; (3) 885
845 a yit 1 or a take +1.
p N;t  p N;t
1
846 The bivariate vector autoregression assumes that
847 causality ows from trade initiation to returns by permit- have the convenient EngleGranger [59] error-correction 886
848 ting rt to depend on the contemporaneous value for x 0t , but representation, 887
849 not allowing x 0t to depend on contemporaneous rt . The
850 model for returns is specied as follows
p t D z t
1 C A1
p t
1 C    C A r
p t
r
1 C u t ; (4) 888

851

852


rt
x 0t

D
X5
iD1
" P15


ar;i
a x;i

r t
i
#
where zt is an error-correction term of rank N  1.
We analyze price discovery using the moving average
representation of our return process (3),
889

890

o f
ro

 891

iD0 br;i ur;t

853 C P15 x 0t
i C : (1)
p t D (L)" t : (5)
iD1 b x;i
u x;t 892
854

855

856
Mizrach and Neely [4] use 5 lags of the return series and
15 lags of the signed trades. The market impact is then de-
P
The disturbances are mean zero and serially uncorrelated,
E[" i;t ] D 0 and cov[" i;t ; " i;t
r ] D 0, but they may be con-
temporaneously correlated, cov[" i;t ; " j;t ] 0.
893

894

d
895
857 ned as the dynamic eect of a buy shock to the return
The information share is related to the long run im-
P
896
series,
pulse responses, (1) D 1

e 1
858
jD0 (L ), the permanent ef-
j 897

859
@r tCn
@x t
: (2)
markets, t
fect of the shock vector on the Treasury prices. Cointe-

c 07 3
gration makes the long run multipliers common across all
-
898

899

900

860

861

862
Mizrach and Neely [4] provide 15 minute market impact
estimates from the GovPX market in 1999. The 2-year note
is most resilient with prices only 0:0042% higher following
a buyer initiated trade. The 30-year bond is the least liquid,
2
6
(1) D 4 ::: e
1     N

rr 00 8
1     N
-
3
:: 7 :
: 5 (6) 901

o
863

864 with prices rising 0:0229% following a buy order. Mizrach

2
c
865 and Neely also report 2004 estimates for the Cantor elec- To eliminate contemporaneous correlation
 among the er- 902

866 tronic limit order book. Market impacts range from 45 to ror terms in (5), we decompose D E " t "0t , the N
N 903

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 11 le-tex

The Microstructure of the US Treasury Market 11

904 covariance matrix, to nd a lower triangular matrix M, Future Directions 952

905 whose i; jth element we denote mi j , such that M M 0 D . This article has reviewed the microstructure of the US 953
906 The Hasbrouck [11] information share for market j is de- Treasury market. The Open Market Desk at the Federal 954
907 ned as Reserve Bank of New York plays a uniquely important role 955
908
in the Treasury market by using transactions in those se- 956
909 Hj D curities to adjust the level of bank reserves. Primary deal- 957
hP i2
n
i mi j ers are key players in both Treasury auctions and the Feds 958
iD j
910 Pn 2 P n 2 ; open market operations. The Treasury market consists of 959
2
iD1  i m i1 C iD2  i m i2 C    C ( n m nn ) several phases: when-issued, primary, on-the-run and o- 960

911
(7) the-run. Two ECNs, eSpeed and BrokerTec, intermediate 961

the most active trading, during the on-the-run phase. The 962
912 where the  i s are the elements of row i of the long-run Treasury futures market at the CBOT complements trad- 963
913 multipliers in (6). Because the Choleski decomposition is ing in the spot market. 964
914 not unique, the information share will vary with the order Treasury markets exhibit end-of-year, daily and in- 965
915 of the equations in the VAR. traday seasonality. Macro and Federal Reserve announce- 966
916 Mizrach and Neely [12] pair spot and maturity ments are responsible for a substantial part of the daily and 967
917 matched futures for the 2-year, 5-year and 10-year on-the- intraday seasonality. The literature studying the impact of 968
918 run spot notes. This calculation requires us to adjust fu- order ows on Treasury prices has also considered how 969
919 tures prices according to the on-the-run spot instruments macro news and Federal Reserve actions inuence such 970
920 with which we compare them. The CBOT provides adjust- impact. 971
921 ment factors for each instrument. These adjustments typi- The futures markets in Chicago play an important 972
922 cally make a single bond the cheapest to deliver (CTD), but role in price discovery, and a discussion of Treasury mi- 973
923 the CTD is typically o-the-run. Nevertheless, the CTD crostructure needs to take this into account. Both spot and 974
924 o-the-run bonds and the most liquid on-the-run bonds futures markets are quite resilient and recent research on 975
925 are very close substitutes their daily returns are highly the Treasury ECNs suggest that the market continues to 976
926 correlated so it is reasonable to examine price discovery become more liquid. Fleming and Mizrach [58] report that 977
927 between futures prices and on-the-run bonds, despite the volume has increased almost 5 times since 2001. This in- 978
928 fact that they are not identical. crease in trading volume accompanies a decline in the im- 979
929 Mizrach and Neely [12] nd that information shares portance of the primary dealers. Beales and Titt reported 980
930 rise with the growth of the GovPX market, but fall as the

f
in the Financial Times in March 2007 that hedge funds 981
931 ECNs take market share from GovPX voice markets. The now account for 80% of trading activity in the Treasury 982

o
932 spot market share is highest for the 2-year note, reach- market with only a 20% share for the primary dealers. One 983
933 ing 86%, while the 10-year spot market share never ex- large fund alone, Citadel, accounts for 10% of the trad-

ro
984
934 ceeds 50%. In addition, relative market liquidity mea- ing volume on eSpeed and BrokerTec. It was perhaps in- 985
935 sures like spreads, trades and volatility each strongly ex- evitable that trading by the millisecond would come to the 986
936 plain daily relative price discovery shares. Mizrach and Treasury market as it did to equities and foreign exchange.

P
987
937 Neely [12] compute both upper and lower bound estimates Perhaps we should only be surprised that it took so long. 988
938 of the information shares. They also report estimates based The Treasury market plays a central role in the credit 989

d
939 on the Harris, McInish and Wood [13] methodology. market. Times of nancial crisis highlight the Treasury 990
940 Campbell and Hendry [51] nd similar results for the markets role as a safe haven for investors both in the US 991
941

942
Canadian government bond market. They nd that the
information share in the 10-year spot note is below 50%
e 1
and overseas. Treasury securities also serve as benchmarks

t
for complex derivatives like mortgage backed securities
3
992

993

c 07
in nearly all their sample of several months between 2002
-
943
and structured loans like collateralized debt obligations. 994
944 and 2004. Upper and Werner [65] nd that price discov- The microstructure of the US Treasury market is funda-

rr e200
995
945 ery in the German Bund is dominated by the futures mar- mental to our understanding of the global nancial mar-

8-
996
946 ket, and in times of stress, like the 1998 Long Term Capi- kets. 997
947 tal Management Crisis, the spot market information share
falls to essentially zero. Upper and Werner [65], however,

o
948

949 compare the futures market to the relatively illiquid, CTD

c
950 bonds. This might explain their nding that the spot mar-
951 ket does very little price discovery.

n
U
i i
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 12 le-tex

12 The Microstructure of the US Treasury Market

998 References 22. Ederington LH, Lee JH (2001) Intraday volatility in interest-rate 1058
and foreign-exchange markets: ARCH, announcement, and 1059
999 1. Akhtar MA (1997) Understanding open market operations. seasonality effects. J Futur Mark 21:51752 1060
1000 Public Information Department, Federal Reserve Bank of New 23. Elton EJ, Green TC (1998) Tax and liquidity effects in pricing 1061
1001 York TS2 http://research.stlouisfed.org/aggreg/meeks.pdf government bonds. J Finance 53:15331562 1062
1002 2. Amihud Y, Mendelson H (1991) Liquidity, maturity, and the 24. Engle R, Granger C (1987) Co-integration and error correction 1063
1003 yields on US treasury securities. J Finance 46:141125 representation, estimation and testing. Econometrica 55:251 1064
1004 3. Andersen TG, Bollerslev T, Diebold FX, Vega C (2007) Real-time 276 1065
1005 price discovery in stock, bond and foreign exchange markets. 25. Fabozzi FJ, Fleming MJ (2005) US treasury and agency secu- 1066
1006 J Int Econ 73:25177 rities. In: The Handbook of Fixed Income Securities, 7th edn. 1067
1007 4. Balduzzi P, Elton EJ, Green TC (2001) Economic news and bond McGraw Hill, New York TS5 1068
1008 prices: Evidence from the US treasury market. J Financial Quant 26. Federal Reserve Bank of New York, Markets Group (2007) 1069
1009 Anal 36:52343 Domestic open market operations during 2006. Federal Re- 1070
1010 5. Barclay MJ, Hendershott T, Kotz K (2006) Automation versus in- serve Bank of New York. http://www.ny.frb.org/markets/omo/ 1071
1011 termediation: Evidence from treasuries going off the run. J Fi- omo2006.pdf TS6 1072
1012 nance 61:23952414 27. Flannery M, Protopapadakis A (1988) From T-bills to common 1073
1013 6. Beber A, Brandt MW (2006) The effect of macroeconomic news stocks: Investigating the generality of intra-week return sea- 1074
1014 on beliefs and preferences: Evidence from the options market. sonalities. J Finance 43:431450 1075
1015 J Monetary Econ 53:19972039 28. Fleming M (2002) Are larger treasury issues more liquid? Evi- 1076
1016 7. Bernanke BS (2005) Implementing monetary policy, remarks dence from bill reopenings. J Money Credit Bank 34:707735 1077
1017 at the redefining investment strategy education symposium. 29. Fleming M (2003) Measuring treasury market liquidity. Fed Re- 1078
1018 Dayton TS3 serv Bank New York Econ Policy Rev 9:83108 1079
1019 8. Bollerslev T, Cai J, Song FM (2000) Intraday periodicity, long 30. Fleming M, Mizrach B (2008) The microstructure of a US 1080
1020 memory volatility, and macroeconomic announcement effects treasury ECN: The BrokerTec platform. http://ssrn.com/ 1081
1021 in the us treasury bond market. J Empir Finance 7:3755 abstract=1117975 TS7 1082
1022 9. Boukus E, Rosenberg JV (2006) The information content of 31. Fleming M, Piazzesi M (2005) Monetary policy tick-by-tick. 1083
1023 FOMC minutes. Working paper, Federal Reserve Bank of New Working paper, Federal Reserve Bank of New York TS4 1084
1024 York TS4 32. Fleming M, Remolona EM (1997) What moves the bond mar- 1085
1025 10. Brandt MW, Kavajecz KA (2004) Price discovery in the us trea- ket? Fed Reserv Bank New York Econ Policy Rev 3:3150 1086
1026 sury market: The impact of orderflow and liquidity on the yield 33. Fleming M, Remolona EM (1999) Price formation and liquidity 1087
1027 curve. J Finance 59:26232654 in the US treasury market: The response to public information. 1088
1028 11. Brandt MW, Kavajecz KA, Underwood SE (2007) Price discovery J Finance 54:19011915 1089
1029 in the treasury futures market. J Futur Mark 27:10211051 34. Fleming M, Remolona EM (1999) What moves bond prices? J 1090
1030 12. Campbell B, Hendry S (2007) Price discovery in canadian and Portf Manag 25:2838 1091
1031 US 10-year government bond markets. Working Paper 0743, 35. Garbade KD, Ingber JF (2005) The treasury auction process: 1092
Bank of Canada TS2 Objectives, structure, and recent adaptations. Fed Reserv Bank

f
1032 1093
1033 13. Chirinko RS, Curran C (2006) Greenspan shrugs: Formal pro- New York Curr Issues Econ Finance 11:111 1094
nouncements, bond market volatility, and central bank com-

o
1034 36. Goldreich D, Hanke B, Nath P (2005) The price of future liq- 1095
1035 munication. Presented at: The American Economic Association uidity: Time-varying liquidity in the US treasury market. Rev Fi- 1096

o
1036 Meetings nance 9:132 1097
14. Chordia T, Sarkar A, Subrahmanyam A (2005) An empirical 37. Green TC (2004) Economic news and the impact of trading on

r
1037 1098
1038 analysis of stock and bond market liquidity. Rev Financial Stud bond prices. J Finance 59:12011233 1099
1039 18:85129 38. Griffiths M, Winters D (1997) On a preferred habitat for liquidity 1100

P
1040 15. Christie-David R, Chaudhry M, Lindley JT (2003) The effects of at the turn-of-the-year: Evidence from the term-repo market. J 1101
1041 unanticipated macroeconomic news on debt markets. J Finan- Financial Serv Res 12:2138 1102
1042 cial Res 26:31939 39. Griffiths M, Winters D (2005) The turn-of-the-year in money 1103

d
1043 16. Cook T, Hahn T (1989) The effect of changes in the federal markets: Tests of risk-shifting window dressing and preferred 1104
1044 funds rate target on market interest rates in the 1970s. J Mon- habitat hypotheses. J Bus 78:13371364 1105
1045
1046
1047
etary Econ 24:331351
17. de Goeij P, Marquering W (2006) Macroeconomic announce-
ments and asymmetric volatility in bond returns. J Bank Fi-
t e 1
40. Griffiths M, Winters D (2005) The year-end price of risk in a mar-
ket for liquidity. J Invest Manag 3:99109

-3
41. Hamilton JD (2007) Assessing monetary policy effects using
1106
1107

c 07
1108
1048 nance 30:26592680 daily fed funds futures contracts. Fed Reserv Bank St. Louis Rev 1109
18. Duffie D (1996) Special repo rates. J Finance 51:493526 (forthcoming) TS8

rr e 00
1049 1110
19. Dungey M, McKenzie M, Smith V (2007) News, no-news and 42. Harris F, McInish T, Wood R (2002) Security price adjustment

8-
1050 1111
1051 jumps in the us treasury market. Cambridge University (un- across exchanges: An investigation of common factor compo- 1112
1052 publ) nents for dow stocks. J Financial Mark 5:277308 1113
1053 20. Ederington LH, Lee JH (1993) How markets process informa- 43. Harvey CR, Huang RD (2002) The impact of the federal reserve 1114

o
tion: News releases and volatility. J Finance 48:116191 banks open market operations. J Financial Mark 5:22357
2
1054 1115
1055 21. Ederington LH, Lee JH (1995) The short-run dynamics of the 44. Hasbrouck J (1991) Measuring the information content of 1116
1056
1057
price adjustment to new information. J Financial Quant Anal
30:11734

TS2 Please provide publisher location.

n c
stock trades. J Finance 46:179207 1117

TS3
TS4
TS5
TS6
Please provide publisher

U
Please provide publisher location and the ID-number of the working paper.
Please provide editors and volume.
Please provide access date.
i TS7 Please provide publisher and access date. i
TS8 Please update, if possible.
i i
Meyers: Encyclopedia of Complexity and Systems Science Entry 635 2008/7/31 16:13 page 13 le-tex

The Microstructure of the US Treasury Market 13

1118 45. Heuson AJ, Su T (2003) Intra-day behavior of treasury sector 67. Upper C, Werner T (2002) Tail wags dog? Time-varying infor- 1177
1119 index option implied volatilities around macroeconomic an- mation shares in the bund market. Working Paper 24/02, Bun- 1178
1120 nouncements. Financial Rev 38:16177 desbank, Frankfurt 1179
1121 46. Huang RD, Cai J, Wang X (2002) Information-based trading 68. Vayanos D, Weill P (2007) A search-based theory of the on-the- 1180
1122 in the treasury note interdealer broker market. J Financial In- run phenomenon. J Finance (forthcoming) 1181
1123 termed 11:269296 69. Warga A (1992) Bond returns, liquidity, and missing data. J Fi- 1182
1124 47. Huang X (2006) Macroeconomic news announcements, finan- nancial Quant Anal 27:60517 1183
1125 cial market volatility and jumps. Duke University (unpubl)
1126 48. Jiang GJ, Lo I, Verdelhan A (2007) Why do bond prices jump?
1127 A study of the US treasury market. Eller College of Manage-
1128 ment, University of Arizona (unpubl)
1129 49. Jones CM, Lamont O, Lumsdaine RL (1998) Macroeconomic
1130 news and bond market volatility. J Financial Econ 47:315337
1131 50. Johnston E, Kracaw W, McConnell J (1991) Day-of-the-week ef-
1132 fects in financial futures: An analysis of GNMA, T-bond, T-note,
1133 and T-bill contracts. J Financial Quant Anal 26:2344
1134 51. Kamara A (1994) Liquidity, taxes, and short-term treasury
1135 yields. J Financial Quant Anal 29:403417
1136 52. Kearney AA (2004) The changing impact of employment an-
1137 nouncements on interest rates. J Econ Bus 54:415429
1138 53. Kuttner KN (2001) Monetary policy surprises and interest rates:
1139 Evidence from the fed funds futures market. J Monetary Econ
1140 47:523544
1141 54. Lahaye J, Laurent S, Neely CJ (2007) Jumps, cojumps and
1142 macro announcements. Working Paper 2007032A, Federal
1143 Reserve Bank of St. Louis, TS2
1144 55. Madhavan A, Richardson M, Roomans M (1997) Why do securi-
1145 ties prices change? A transaction-level analysis of NYSE stocks.
1146 Rev Financial Stud 10:10351064
1147 56. Menkveld AJ, Sarkar A, Van der Wel M (2006) The informa-
1148 tiveness of customer order flow following macroeconomic an-
1149 nouncements: Evidence from treasury futures markets. (un-
1150 publ) TS9
1151 57. Meulendyke AM (1998) US monetary policy & financial mar-
1152 kets. Federal Reserve Bank of New York. http://research.
stlouisfed.org/aggreg/meulendyke.pdf TS4

f
1153
1154 58. Mizrach B, Neely CJ (2006) The transition to electronic commu-

o
1155 nication networks in the secondary treasury market. Fed Re-
1156 serv Bank St. Louis Rev 88:527541

o
1157 59. Mizrach B, Neely CJ (2007) Information shares in the US trea-
sury market. J Bank Finance (forthcoming) TS8

r
1158
1159 60. Neely CJ, Winters DB (2006) Year-end seasonality in one-
1160 month LIBOR derivatives. J Derivatives 13:4765

P
1161 61. Nyborg KG, Sundaresan S (1986) Discriminatory versus uni-
1162 form treasury auctions: Evidence from when-issued transac-
1163 tions. J Financial Econ 42:63104
1164
1165
1166
62. Poole W, Rasche RH (2000) Perfecting the markets knowledge
of monetary policy. J Financial Serv Res 18:255298
63. Poole W, Rasche RH, Thornton DL (2002) Market anticipations
d
e 1
1167
1168
1169
of monetary policy actions. Fed Reserv Bank St. Louis Rev
84:6594
64. Seligman J (2001) Is last money dearest? An analysis of us trea-
sury short term finance. Working paper, University of Califor-
t
c 07 - 3
e
1170
nia, Berkeley TS10
-
1171

rr 00
65. Simon DP (1991) Segmentation in the treasury bill market: Ev-
8
1172
1173 idence from cash management bills. J Financial Quant Anal
1174 26:97108

o
66. Simon DP (1994) Further evidence on segmentation in the
2
1175
1176 treasury bill market. J Bank Finance 18:139151

TS9 Please provide an institute.

n c
TS10 Please provide the ID-number of the working paper.

U
i i
Editors or typesetters annotations (will be removed before the nal TEX run)