The Response of Term Rates to Fed Announcements *



Similar documents
A Comparative Study of Linear and Nonlinear Models for Aggregate Retail Sales Forecasting

Fortified financial forecasting models: non-linear searching approaches

I S T H E A U S T R A L I A N F O R E X M A R K E T E F F I C I E N T? A T E S T O F T H E F O R W A R D R A T E

The Identification of the Response of Interest Rates to Monetary Policy Actions Using Market-Based Measures of Monetary Policy Shocks

2.4 Network flows. Many direct and indirect applications telecommunication transportation (public, freight, railway, air, ) logistics

Usefulness of the Forward Curve in Forecasting Oil Prices

Equity Valuation Using Multiples. Jing Liu. Anderson Graduate School of Management. University of California at Los Angeles (310)

Calculation of variable annuity market sensitivities using a pathwise methodology

Heat demand forecasting for concrete district heating system

11/6/2013. Chapter 14: Dynamic AD-AS. Introduction. Introduction. Keeping track of time. The model s elements

Policies & Procedures. I.D. Number: 1071

Duration and Convexity ( ) 20 = Bond B has a maturity of 5 years and also has a required rate of return of 10%. Its price is $613.

PROFITS AND POSITION CONTROL: A WEEK OF FX DEALING

A Note on Using the Svensson procedure to estimate the risk free rate in corporate valuation

Cross-sectional and longitudinal weighting in a rotational household panel: applications to EU-SILC. Vijay Verma, Gianni Betti, Giulio Ghellini

Empirical heuristics for improving Intermittent Demand Forecasting

SPEC model selection algorithm for ARCH models: an options pricing evaluation framework

Small and Large Trades Around Earnings Announcements: Does Trading Behavior Explain Post-Earnings-Announcement Drift?

INTEREST RATE FUTURES AND THEIR OPTIONS: SOME PRICING APPROACHES

Imagine a Source (S) of sound waves that emits waves having frequency f and therefore

NASDAQ-100 Futures Index SM Methodology

How Much Can Taxes Help Selfish Routing?

Long Term Spread Option Valuation and Hedging

The naive method discussed in Lecture 1 uses the most recent observations to forecast future values. That is, Y ˆ t + 1

II.1. Debt reduction and fiscal multipliers. dbt da dpbal da dg. bal

Term Structure of Prices of Asian Options

THE FIRM'S INVESTMENT DECISION UNDER CERTAINTY: CAPITAL BUDGETING AND RANKING OF NEW INVESTMENT PROJECTS

The Maturity Structure of Volatility and Trading Activity in the KOSPI200 Futures Market

Formulating Cyber-Security as Convex Optimization Problems

ANALYSIS AND COMPARISONS OF SOME SOLUTION CONCEPTS FOR STOCHASTIC PROGRAMMING PROBLEMS

Chapter 13. Network Flow III Applications Edge disjoint paths Edge-disjoint paths in a directed graphs

Chapter 7. Response of First-Order RL and RC Circuits

TEMPORAL PATTERN IDENTIFICATION OF TIME SERIES DATA USING PATTERN WAVELETS AND GENETIC ALGORITHMS

Performance Center Overview. Performance Center Overview 1

Chapter 8: Regression with Lagged Explanatory Variables

VOLATILITY DYNAMICS OF NYMEX NATURAL GAS FUTURES PRICES

Vector Autoregressions (VARs): Operational Perspectives

Nikkei Stock Average Volatility Index Real-time Version Index Guidebook

Chapter 6: Business Valuation (Income Approach)

Market Liquidity and the Impacts of the Computerized Trading System: Evidence from the Stock Exchange of Thailand

Journal Of Business & Economics Research September 2005 Volume 3, Number 9

Option Put-Call Parity Relations When the Underlying Security Pays Dividends

Predicting Stock Market Index Trading Signals Using Neural Networks

The Greek financial crisis: growing imbalances and sovereign spreads. Heather D. Gibson, Stephan G. Hall and George S. Tavlas

Morningstar Investor Return

OPTIMAL BATCH QUANTITY MODELS FOR A LEAN PRODUCTION SYSTEM WITH REWORK AND SCRAP. A Thesis

The Transport Equation

Optimal Path Routing in Single and Multiple Clock Domain Systems

GUIDE GOVERNING SMI RISK CONTROL INDICES

Chapter 8 Student Lecture Notes 8-1

How To Solve An Uncerain Daa Problem

Trading Strategies for Sliding, Rolling-horizon, and Consol Bonds

Alternative Settlement Methods and Australian Individual Share Futures Contracts. Donald Lien and Li Yang * (Draft: September 2003)

Hedging with Forwards and Futures

Formulating Cyber-Security as Convex Optimization Problems Æ

Risk Modelling of Collateralised Lending

Central Bank Communication: Different Strategies, Same Effectiveness?

Distributing Human Resources among Software Development Projects 1

PROFIT TEST MODELLING IN LIFE ASSURANCE USING SPREADSHEETS PART ONE

Why Did the Demand for Cash Decrease Recently in Korea?

The Twin Agency Problems in Corporate Finance - On the basis of Stulz s theory -

Chapter 1.6 Financial Management

Individual Health Insurance April 30, 2008 Pages

The yield curve, and spot and forward interest rates Moorad Choudhry

On the degrees of irreducible factors of higher order Bernoulli polynomials

AP Calculus BC 2010 Scoring Guidelines

Returns to defaulted corporate bonds

4. International Parity Conditions

Nanocubes for Real-Time Exploration of Spatiotemporal Datasets

CALCULATION OF OMX TALLINN

Measuring macroeconomic volatility Applications to export revenue data,

Credit Index Options: the no-armageddon pricing measure and the role of correlation after the subprime crisis

Markit Excess Return Credit Indices Guide for price based indices

MACROECONOMIC FORECASTS AT THE MOF A LOOK INTO THE REAR VIEW MIRROR


Mathematics in Pharmacokinetics What and Why (A second attempt to make it clearer)

Cointegration: The Engle and Granger approach

Principal components of stock market dynamics. Methodology and applications in brief (to be updated ) Andrei Bouzaev, bouzaev@ya.

Markov Chain Modeling of Policy Holder Behavior in Life Insurance and Pension

Banking, Inside Money and Outside Money

Appendix D Flexibility Factor/Margin of Choice Desktop Research

Transcription:

Revied: June The Repone o Term Rae o Fed Announcemen Abrac In February 4, 994 he Federal Reerve began he pracice o announcing change in he argeed level or he ederal und rae immediaely aer uch deciion were made. Thi paper inveigae o wha exen he policy o he announcemen aeced a key ingredien in he moneary ranmiion mechanim: he erm rucure o nominally rik-ree, Treaury ecuriie. We ind ha erm rae reac much more in union during announcemen day han a any oher ime. Moreover, he pracice o circumcribing almo all change in he ederal und rae arge o Federal Open Marke Commiee (FOMC) meeing dae regimen he ormaion o marke expecaion in he overnigh rae and he price dicovery proce o erm rae, hu aciliaing he Fed goal o conrolling long-erm rae. JEL Claiicaion: E4, E5, C5 Keyword: Announcemen Eec, Term Srucure, Raional Expecaion Selva Demiralp Board o Governor o he Federal Reerve Diviion o Moneary Aair Wahingon, DC 55 e-mail: elva.demiralp@rb.gov Òcar Jordà U. C. Davi Deparmen o Economic One Shield Ave. Davi, CA 9566-8578 e-mail: ojorda@ucdavi.edu Our paper beneied rom many ueul uggeion rom eminar paricipan a he Federal Reerve Bank o New York, Simon Fraier Univeriy, U. C. San Diego, U. C. Davi and he Federal Reerve Board. We are paricularly indebed o Charle Evan, Sherry Edward, Jame Hamilon, Kevin Hoover, Kenneh Kuner, David Lindey and Daniel Thornon. Special hank go o Kenneh We and wo anonymou reeree or heir uggeion. The opinion expreed are hoe o he auhor and do no necearily relec he view o he Board o Governor o he Federal Reerve Syem or oher member o i a.

Chairman Greenpan: Look, he main iue here i ha, a ar a I am concerned, I would like u o be couned. We are he cenral bank and we are making a major move, Trancrip o he Federal Open Marke Commiee Meeing, February 3-4, 994.. Inroducion The Federal Open Marke Commiee (FOMC) meeing held on February 3-4, 994 marked he ir change in he ederal und rae arge (or arge or breviy) ince Sepember 4, 99. During ha meeing, a reaonable porion o he dicuion cenered on he need o make he commiee inenion clear o he public aer all, rae had remained unchanged or over a year and i had been ive year ince he Federal Reerve la ook a ighening ance. The meeing concluded wih he mo igniican procedural change ince Chairman Volcker experimened wih a nonborrowed reerve arge in 979-98: he 5 bai poin increae in he ederal und rae arge would be made public immediaely a he meeing concluion. Since ha meeing in February 3-4, 994, he Federal Reerve (Fed) ha coninued he pracice o publicly dicloing any change in he argeed level or he ederal und rae. In hi paper we will argue ha hi procedural change ha added a new operaional ool o he Fed arenal: he announcemen eec. The pracice o announcing each change in he arge aec he moneary ranmiion mechanim in wo way: () by increaing he eecivene wih which he Fed manage ederal und rading around he argeed rae, and () by regimening he ormaion o expecaion and he price dicovery proce o nominally rik-ree, Treaury ecuriie. The ir o hee eec i analyzed in he companion paper by Demiralp and Jordà () wherea he econd eec coniue he cenral opic o hi paper. The holy grail o moneary policy i o aec long-erm rae, and ulimaely economic aciviy, by adequaely manipulaing expecaion. We will how ha by announcing when he arge i changed, erm rae o Treaury ecuriie end o move in union and in a manner broadly conien wih he raional expecaion hypohei during announcemen day, relegaing mo o he variaion in erm rae ouide announcemen day o Fiherian eec. However, a change in he arge, paricularly

when hi change revere he general rend o previou arge change, caue a ubanial reviion o long-erm expecaion on he ederal und rae. Evidence on he Fed abiliy o aec long-erm rae by managing he hor-erm rae ha been eluive a be. Kuner () perhap provide he mo provocaive reul in recen ime by emphaizing he diincive repone o erm rae o expeced and unexpeced movemen in he ederal und rae arge. Hi analyi largely mirror Cook and Hahn (989) inveigaion o one-day bond rae repone o change in he und rae arge or he period 974 o 979. Modern cenral banking i characerized, a a minimum, by public announcemen o an inere rae arge (uch a he ederal und rae in he U.S.), omeime, by addiionally dicloing an inlaion arge (uch a he Bank o Auralia, and he Bank o England), and in he exreme, by making he parameer o he policy reacion uncion publicly available, uch a he Reerve Bank o New Zealand. Wih regard o New Zealand, Guhrie and Wrigh () inveigae a phenomenon ha i relaed o he one under cruiny in hi paper and which hey label open mouh operaion. However, unlike New Zealand, mo change in he ederal und rae arge are widely anicipaed (we preen ormal evidence on hi poin), and while reerve depoi in he U.S. do no pay inere, he mechanim ha give rie o open mouh operaion in Guhrie and Wrigh () depar omewha rom he announcemen eec we inveigae. The ormal heoreical characerizaion o he announcemen eec in he U.S. can be ound in Taylor (). In a previou empirical udy, Thornon () doe no ind evidence in uppor o he announcemen eec or he period 974-979; 984-989; and 989-997. In conra, we will emphaize he dierence ha exi beween he 989-994 and he 994- period by aking he February 3-4, 994 policy change a our break poin. The remainder o he paper i organized a ollow. Secion examine ormal aiical evidence in uppor o he February 4, 994 hi in policy. Secion 3, decribe Kuner () marke-baed expecaion o a arge change while Secion 4 evaluae he qualiy o hee meauremen. Secion 5 examine he evoluion o erm rucure repone o arge change aer 994 and Secion 6 urher inveigae wheher hee repone are eniive o he FOMC chedule. Secion 7 conclude. 3

. I February 4, 994 a Special Dae? Cook and Hahn (989) inveigaed he naure o erm rucure relaion by examining he one-day repone o Treaury ecuriie o change in he ederal und rae arge. Denoe he rae o hee ecuriie generically a R and le denoe he argeed ederal und rae, he regreion R = α β u () wa eimaed on a ample coniing o 75 day in which he Fed changed he und arge over he period 974 hrough 979. In hi paper we inveigae how he naure o hee relaionhip changed in 994 wih he February 4 deciion o publicly dicloe change in he und rae arge. While here i lile dipue on he ingular imporance o hi dae rom he iniuional poin o view (e.g. ee Barolini, Berola, and Prai, ; and Taylor, ), we provide below ormal aiical evidence ha demonrae erm rae reponded dierenly hereaer. The ample ha we analyze conain 54 day in which he und rae arge wa changed over he period May 8, 989 o February,. The ar o our ample i deermined by he dae o ir reliable availabiliy o ederal und uure daa. Becaue hee daa are cenral o he empirical analyi in oher par o he paper, we preer o mainain hi ample hroughou he paper o be conien. The erm rae ha we inveigae include he hree- and ix-monh Treaury Bill rae and he wo-, ive-, and en-year Treaury noe rae (we reer he reader o Secion 5 and he daa appendix or deailed inormaion on he ource and adjumen made o hee daa). I February 4, 994 a pecial dae? To explore hi poibiliy, we will rely on everal muliple rucural break change e propoed in Bai and Perron (998) and There are 56 arge change in acualiy. However, he change recorded or 8//989 correponded o a.65% decreae (raher han he ypical.5%). The change recorded or 9/7/ ha been omied becaue i wa largely moivaed by he even o Sepember, raher han wih acual moneary policy. See he appendix or more deail on daa adjumen. The ederal und uure marke oicially opened in 988, however rading in hi marke wa raher hin a he beginning. Thi characeriic i explored in more deail in Secion 5. 4

companion paper, Bai and Perron (, and ). 3,4 The keyone o he Bai and Perron (998) e i he eimaion o a priori unknown, rucural break dae. Thi eimaion problem i olved wih an eicien algorihm ha compue he global minimizer o he um o quared reidual. In paricular, we conider he regreion in () wih m break (m regime): R = α β u = T,..., T j j j j () or j =,, m. The break poin T,, T m are explicily reaed a unknown (by convenion, T = and T m = T). The goal i o eimae he unknown coeicien (α,, α m ; β,, β m ; T,, T m ) auming α j α j and β j β j ( j m ). For each m-pariion, he aociaed lea quare eimae are obained by minimizing he um o quared reidual, S T = m T j [ R j β j ] j= = T j α (3) Then, he eimaed break poin are uch ha, Tˆ,..., Tˆ ) = arg min S ( T,..., T ) (4) ( m T,..., Tm T m where he minimizaion i aken over all pariion (T,, T m ) uch ha T j T j- q. The elecion o he number o rucural change can hereore be viewed a a peciicaion problem or which inormaion crieria can be ued (ee Liu, Wu, and Zidek, 997, henceorh LWZ when we reer o heir inormaion crierion). However, imulaion reul in Perron (997) ugge ha aympoically conien crieria, uch a Schwarz (978) inormaion crierion (SIC) and LWZ, end o perorm badly in mall 3 We hank Juhan Bai or graciouly providing u he GAUSS code ha accompanie he Bai and Perron (998) paper. 4 We will briely decribe Bai and Perron (998) mehod below alhough he reader i advied o conul he original reerence or a deailed explanaion. 5

ample (hee crieria end o undereimae he number o break). Aympoically eicien mehod, uch a Akaike (97, 974) FPE (inie predicion error) and AIC end o perorm beer in mall ample bu hey have eriou drawback when he number o ied parameer i a moderae o large racion o he ample ize, a iuaion ha arie in he preen conex a m increae. To overcome hee horcoming Bai and Perron (998) develop up Wald ype e or he null hypohei o no change veru an arbirary number o change, a well a he null hypohei o l change again he alernaive o l change. Their recommended raegy i o ue he ir ype o up Wald e o deermine wheher or no break are preen and hen ue a equenial applicaion o he econd ype o up Wald e o deermine he number o break. In addiion o he raegy ju decribed, we will explore a bia correcion o AIC (AIC C henceorh) inroduced by Hurvich and Tai (989), which i an approximaely unbiaed eimaor o he expeced Kullback- Leibler inormaion. 5 Thi correcion i paricularly well uied or he characeriic o he preen inveigaion and Hurvich and Tai (989) how ha i provide beer order model choice han any oher aympoically eicien mehod. For compleene, we alo repor reul wih he generalized cro validaion crierion (GCV) by Craven and Wahba (979) and he crierion by Rice (RICE, 984). 6 The reul o he muliple rucural change e are repored in Table - below. Becaue we have a relaively mall number o obervaion, we conrain he maximum number o break m o one and e he minimum ize o a pariion, T j T j, o % o he ample or obervaion. For each mauriy, Table conain he up F(l ) e or he null hypohei o no break again he alernaive o l break; and he double maximum e UDmax and WDmax. All o hee e are explained in deail in Bai and Perron (998). A he end o Table, we repor he dae eleced by he global 5 Speciically, = AIC [ ( q )( m ) ][( q )( m ) ] [ T ( q )( m ) ] AIC C where q i generically he dimenion o he vecor o regreor, m i he number o regime and T i he ample ize. 6 GCV = ( SSR ) RICE = ( SSR ) ( )( ) q m T T ( ( ) ) q m T T and 6

minimizer. Table conain a baery o inormaion crieria including: AIC C, RICE, GCV, SIC, and LWZ. The combinaion o he Bai and Perron (998) up Wald e and he baery o inormaion crieria ha we repor ugge a break in he daa or Treaury ecuriie o mauriy up o ive year. Similarly, boh mehod indicae ha he en-year T-noe rae i able over he ample period. We only ind diagreemen wih regard o he ive-year T- noe rae while he up Wald e ugge a break in he daa, none i deeced by he dieren inormaion crieria. Thee reul conirm he inding in Perron (997) who noe he poor perormance o SIC and LWZ in mall ample. Aympoically eicien crieria provide more accurae elecion and in paricular, he bia correcion in AIC C igniicanly ameliorae he horcoming experienced wih Akaike (974) original AIC ormulaion. Inereingly, he break dae eleced by he global minimizer o he um o quared reidual in expreion (4) are very imilar acro all he erie and uppor our prior regarding February 4, 994. Thu, boh he wo- and ive-year T-noe rae achieve heir minimum um o quare a exacly hi dae, wherea he hree- and ix-monh T- Bill rae achieve heir minimum and obervaion prior o February 4, 994 repecively, and he en-year T-noe rae (alhough aiically able over he whole ample) obervaion aer. Alogeher, hee dicrepancie are wihin accepable limi o ample variaion around February 4, 994 given a ample o 54 obervaion. Thee reul hu coniue rong upporive evidence ha he well documened iniuional change we are inveigaing meri urher inveigaion. 3. Meauring Marke Expecaion on Targe Change The announcemen o a change in he arge i quie oen an expeced even. Thereore, o make ene o he inluence ha hee announcemen have on erm rae, we need o eparae hi expecaional componen rom he rue elemen o urprie. One could conider regreion baed eimae o each o hee componen, however, hee eimae depend on he choice o peciicaion and on adequae availabiliy o variable 7

8 ha appropriaely ummarize he va array o inormaion available o marke paricipan. Inead, we preer o exploi available daa on he marke or ederal und uure and he mehod decribed in Kuner (). Under he naural aumpion ha hi marke i eicien, hee daa provide a reaonably cloe, marke-baed proxy o marke expecaion. We reer he reader o Kuner () or addiional deail and concenrae here on he meauremen raegy. Speciically, we rewrie equaion (4) in Kuner (), which decribe he pomonh uure rae on day and o monh a,,,,, i i i i m m m E m m m E m m m µ µ = = > > (5) where i denoe he average daily ed und rae up o dae, i he ederal un rae a ime, E i he uual condiional expecaion operaor, and, µ i a rik premium, which in an eicien marke wih rik neural inveor would be zero. Thereore noe ha, ( ) ( ) ( ),,,, i i E E m m E m µ µ = (6) The acual ederal und rae,, can be een a he um o he arge plu a mean zero argeing error, η. Hence, he ir erm in expreion (6) i a caled one ep-ahead oreca error o a arge change he expecaional componen we are inereed in meauring. The econd erm in (6) repreen he dierence in he average argeing error expeced o prevail in he remainder o he monh. Noe ha ince ) ( = E η or > hen a, p i > η, o ha he dierence i i E E will be virually equal o zero excep or he la ew day o he monh a which ime hee meauremen error may become non-negligible. In addiion o hee end-o-he-monh eec, one ha

o conider he poible impac o he mainenance period eaonaliy on he argeing error o he la ew day o he monh. Finally, he hird erm in expreion (6) conain he dierence beween he rik premia a day and. Thi dierence i likely o be zero i he rik premium remain relaively conan rom one day o he oher, nowihanding he ac ha i may be mall o begin wih. Kuner () hereore ugge ha he urprie componen o a arge change can be conruced by appropriaely recaling he dierence in expreion (6), ( ) m ε =,, (7) m o ha hi urprie componen ruly relec he number o day aeced by he change. Hence, he expeced componen o a arge change can be calculaed a, ( ) = E ε (8) Beore we begin our analyi however, we urher inveigae he naure o he meauremen error ha we dicued above o deermine wheher i migh explain ome o he reul ha we ind wih reerence o he 994 break poin. 7 Figure diplay he mainenance period eaonaliy o he argeing error η, broken down ino hree regime: 5/8/989 o /3/994; /4/994 o 8/3/998; and 8/4/998 o //. The ir wo regime correpond o he naural diviion a /4/994 and he hird regime relec he re-inroducion o lagged reerve accouning (LRA), which virually eliminaed he uncerainy urrounding he la ew day o he mainenance period and ha had characerized he ederal und marke up o ha poin. A he igure how, here are no ubanive dierence acro regime. The la wo day o he mainenance period diplay well-documened paern o exce volailiy (ee Hamilon, 997 and Cloue and Dow, ), which are largely diipaed wih he inroducion o he LRA in 998. 9

Similarly, Figure diplay he erm E ( ) conruced a in equaion (8), a a uncion o he number o buine day le in he monh. Once again, here are no ubanial dierence acro period, each diplaying omewha higher volailiy in he la hree day o he monh and in agreemen o our previou dicuion. Thi volailiy promp Kuner () o ue he one-monh uure rae inead when he arge change occur wihin hree day o he end o he monh. The dae ha require uch adjumen are: 7/7/989; /9/99; 4/3/99; /3/99; /3/996; 9/8/998; 6/3/999; /3/; and 6/7/. Table 3 diplay he arge change ha ook place during hee dae along wih he expeced change a meaured by expreion (8). A can be een rom hi able, here are no unuual paern, perhap wih he excepion o he /9/99 and 4/3/9 change, which were no well anicipaed. However, wheher or no hi i an ariac o an end-o-he-monh eec i diicul o deermine. In addiion o hee correcion, Kuner () adju wo obervaion in he ample ha correpond o dae when he Federal Open Marke Commiee (FOMC) made an announcemen aer he ederal und uure marke had already cloed. Thee dae are December 8, 99 (he Fed announced a 5 bai poin cu in he dicoun rae a 3:3 pm) and Ocober 5, 998 (when he Fed had an iner-meeing move and announced he move a 3:5 pm). Alhough here have been hree addiional inermeeing move in our ample correponding o /3/; 4/8/; and 9/7/, he new were releaed o he public a :3pm, :54am, and 8:am repecively and hereore, well beore he cloe o he uure marke. 8 7 Incidenally, we ake hi opporuniy o remind he reader ha our rucural break e purpoely avoid he meauremen iue by direcly looking a he repone o erm rae o change in he arge raher han o each o he expecaional and urprie componen. 8 For urher deail on he daa and he adjumen made in Kuner () we recommend he reader o conul and addendum prepared by Kenneh Kuner available a: hp://www.newyorked.org/rmaghome/economi/kuner/daanoe.pd

4. Evaluaing Marke Baed Expecaion o Change in he Targe Thi ecion examine he properie o he expecaional meaure E ( ) decribed in expreion (8) wih paricular reerence o i abiliy o predic, one day in advance, change in he ederal und rae arge. Furhermore, becaue we are inereed in any dierence emanaing rom he February 4, 994 break poin, we will pli he ample a ha dae. In addiion o evaluaing predicive abiliy, we alo conider end-o-monh eec, end-o-mainenance-period eec, and FOMC eec or he purpoe o aeing ome o he meauremen iue dicued in he previou ecion. Thu, we monh deine d = i all in any o he la hree buine day in he monh, oherwie; d = i all in any o he la wo day o he mainenance period, oherwie; and mp d = i all on he day in which an FOMC meeing conclude, oherwie. omc The und rae arge i ypically modiied in incremen o 5 bai poin o ha, bu or a ew change in 989 and he November 5, 994 increae o.75%, he range o value hee change ake belong o he e o repone {-.5, -.5,,.5,.5}. Thu, i i naural o model hee daa wih an ordered repone model uch a he ordered probi model. The peciicaion o he model i a ollow. We hypoheize he exience o an unoberved laen variable, uch ha γ β β β (9) monh mp omc = E ( ) md E ( ) pd E ( ) d E ( ) u where ~ i. i.. N(, ). The dicree change in he arge are relaed o he laen u d proce according o.5 i (, c ].5 i ( c, c] = M.5 i ( c4, ) ()

' wih c < c < c 3 < c 4. Denoe expreion (9) in compac noaion by = w π u, hen noice ha he probabiliy ha he arge i changed by 5 bai poin ay, i given by P( =.5 w φ( c 4 w ) = P( c ' 3 π ) φ( c < w 3 ' w π u ' π ) c 4 ) = () where φ(z) denoe he probabiliy ha a andard Normal variable ake on a value le han or equal o z. From expreion () i i raighorward o conruc he likelihood uncion, which can be eimaed by convenional numerical echnique. Table 4 conain he maximum likelihood eimae o hi ordered probi model on buine-week, daily daa rom May 8, 989 o February, pli in February 4, 994. The ir ub-ample conain 9 obervaion while he econd ub-ample conain 89. Overall, he eimae or he econd ub-ample are more precie a wineed by he igher igh o he model (he peudo R increae rom.8 o.7 and he average likelihood per obervaion improve rom.85 o.9). Noice ha he ir ub-ample only conain reducion in he arge and hereore he model eimae only wo limi poin, c and c o ha i r > c hen {,.5,.5}. While eimae a boh ub-ample ugge ha day when he FOMC mee renghen he marke-baed ignal E ( ), hi dierence i only igniican aer he Fed decided o announce arge change a he end o he FOMC meeing. Wih repec o he end-omonh and end-o-mainenance-period eec, i appear ha only he ormer are igniican and in he direcion one would expec when he ignal carrie a more ubanial meauremen error. However, he dierence beween regime are no aiically igniican. The boom hal o Table 4 repor a coare predicive exercie ha coni on gauging wha even ha he mo probabiliy o occurring. Thu, he column labeled Acual reer o he number o even correponding o he caegorie in Value, he column Pred. reer o he even wih he maximum prediced probabiliy, and he column Error repor he dierence beween Acual and Pred. The evidence

provided in hi able hereore ugge ha much o he improvemen in i repored above i alo releced in he accuracy o he predicion. Overall, he reul ugge ha he marke-baed meaure E ( ) carried a more accurae ignal aer 994 bu ha he dierence in hi accuracy are unrelaed o end-o-monh eec or end-o-mainenanceperiod eec. 5. Term Srucure Repone o Fed Announcemen The decompoiion o arge change ino heir expecaional and urprie componen (decribed in expreion 7 and 8), allow u o ak wheher or no he policy o announcing hee change ha made any dierence in he manner erm rae reac. One way o proceed i o ak wheher rae repond a all o predicable policy acion, in a manner imilar o Cook and Hahn (989) and Kuner (). In paricular, conider regreing a Treaury rae on he wo componen o a arge change, namely = α β E ( ) ) β R ( ε v () where R denoe he yield on he hree-, and ix-monh Treaury Bill, and wo-, ive-, and en-year Treaury noe and he remaining variable have been deined in expreion (7) and (8). Expreion () i eimaed on day in which he arge wa changed only, over he ample 5/8/989 /3/994 and /4/994 //. Hence, he index reer o day in which he arge i changed, - will reer o he arge change ha preceded he h arge change, and he noaion (-) denoe he day beore he h arge change. Thee regreion can be een a complemening he regreion repored in Table 3 in Kuner () and are repored here in Table 5. We make excepion o noe ha he obervaion or January 3, and Sepember 7, are paricularly unuual. The la arge change in wa a 5 bai poin increae on May 6. The January 3, change wa a 5 bai poin decreae and ook place ouide an FOMC meeing during he ir week marke opened 3

or he year. Similarly, he Sepember 7, change ollowed he even o Sepember,, aer he marke had been cloed or one week. Thi change more likely relec he Fed repone o an increaed ene o rik averion and inabiliy in inancial marke raher han o policy conideraion. We ruggled wih hi iue and aer reviewing he daa or oher inluenial obervaion, we decided ha i wa be o include a dummy variable in he regreion o neuralize he unuual eec o hee even. Hence, wih hee adjumen and depie having a mall number o degree o reedom (here are 4 arge change in he 989 994 ample and 3 in he 994 ample), he parameer are relaively preciely eimaed. The obviou paern ha emerge rom Table 5 are a ollow. We replicae wo general reul in Kuner (), namely ha, a he mauriy o he Treaury ecuriy increae, he reacion o he urprie componen o a arge change diminihe. However, noice ha while hee coeicien are raher cloe o he canonical value o one or virually all mauriie during he ir ample, he econd ample diplay much maller coeicien eimae and a more dramaic decay, paricularly or mauriie o wo-year and beyond. In ac, he urprie componen i no aiically dieren rom zero or he en-year T-noe, much along he line o wha i uually ound in he empirical erm rucure lieraure. Secondly, he anicipaed componen o a arge change ha a aiically inigniican eec excep or he pre-94 ample, hree- and ix-monh mauriie only. One would expec ha in an eicien inancial marke, hee erm would be zero and hence, we inveigaed he caue o hi anomaly. Trading a he incepion o he marke or ederal und uure wa iniially hin, hereore, we experimened wih a ample ha ared in January 99 inead, and redid he pre-94 regreion in Table 5. 9 We ound ha indeed, he coeicien on he anicipaed componen or all mauriie were aiically inigniican and numerically cloe o he canonical value o zero, uggeing ha he anomaly cied above ha o do wih he iniial novely o hi inancial derivaive. The nex ecion provide an explanaion or he lower eniiviy o erm rae o urprie policy announcemen baed on he manner he FOMC chedule regimen expecaion. 9 We hank Ken Kuner or uggeing hi experimen. 4

Thi eaure alo explain he beer predicabiliy o arge adjumen ha we repored in Secion 4. 6. The Timing Hypohei and he FOMC Schedule The inal experimen ha we conider conemplae wo iue imulaneouly. Fir, we explore he manner in which he FOMC chedule regimen he ormaion o expecaion abou uure arge change. In oher word, do change in he arge execued ouide regularly cheduled FOMC meeing carry more weigh becaue hey are unuual? Since 994 and up o he end o our ample (//), he Fed ha only changed he arge ouide an FOMC meeing our ime. Secondly, we examine wha Kuner () label a he iming hypohei, which we explain uccincly a ollow. Rudebuch (995) documen ha he Fed ypically change he arge in a erie o 5 bai poin increae/decreae in he ame direcion and ha he Fed rarely revere he general rend o hee change. Conequenly, when we calculae he repone o erm rae (paricularly ecuriie wih longer mauriie) o he urprie componen o a arge change, i i enirely poible ha he marke imed he arge change incorrecly bu neverhele, widely anicipaed ha change omeime in he near uure. Thu, he main advancemen or poponemen o anicipaed rae change will have a maller eec on erm rae han acion ha ruly indicae a reveral in he policy ance. Thi hypohei, a priori, could explain he low coeicien eimae or he repone o erm rae o he urprie componen o a arge change during he po-994 ample period. A imple way o explore boh o hee iue imulaneouly coni in eimaing a regreion on a ample ha conain dae in which he arge wa changed excluively, imilar in naure o he regreion in expreion (). However, in addiion we conider he ollowing wo variable: he dummy variable d FOMC which i deined above and indicae wheher he h change in he arge correponded wih an FOMC meeing or no; and he dummy variable d SWITCH which ake he value o one i he arge change a ime revere he direcion on he arge change a ime -, oherwie 5

(noe ha he ime index now indicae dae o arge change excluively raher han calendar day). Ineracing hee wo addiional variable muliplicaively in expreion () yield R β d = α α d SWITCH E ( ) FOMC ( α d ) d SWITCH SWITCH β E ( ) β ε β d ( ) β d E ( ) (3) FOMC ε β d FOMC SWITCH ( ) ε v Thu, hi paramerizaion meaure he repone o he erm rae R o, among oher eec, he urprie componen o a arge change realized: () a an FOMC meeing (meaured by ˆ ˆ β β ); () ouide an FOMC meeing (meaured by ˆβ ); (3) a an FOMC meeing which ignal a reveral in he policy ance (meaured by ˆ β ˆ β ˆ β ); and (4) ouide an FOMC meeing in addiion o ignaling a reveral in he policy ance (perhap he ronge ignal he Fed can end, meaured by β β ). Table 6 repor he eimae o expreion (3) over he enire ample 5/8/989 // (noe ha here are an inuicien number o arge change ouide FOMC dae and policy ance reveral o aord uicien degree o reedom in eimaing expreion (3) over pre and po 994 ample). The reul o hi experimen are quie riking and go a long way a juiying why i i oen hard o ind ha he ederal und rae ha any eec on long-erm rae. A arge change execued a a regularly cheduled meeing ha i conien wih he general direcion o he policy ance, ha no aiically igniican impac on Treaury rae a any mauriy excep or he hree- and ix-monh T-Bill. However, i ha ame arge change i announced ouide an FOMC meeing, ha change will have a igniican impac, alhough hi impac ubanially decline a he mauriy o he Treaury ecuriy increae (rom a value o abou.78 o.34, alhough in all cae igniican). In conra, when he Fed revere he direcion ollowed in previou arge change (a reveral o he policy ance), Treaury ecuriie diplay a very izeable repone, oen ime urpaing he canonical value o. In paricular, when he reveral ˆ ˆ The dae o hee change are /5/998; /3/; 4/8/; and 9/7/. 6

come a an FOMC meeing, a he mauriy increae, he repone i approximaely.5 ime he ize o he arge change. A imilar paern emerge i he reveral i done ouide he FOMC meeing alhough he repone are even ronger, approximaely doubling he repone o ha arge change. Thee reul hed ligh on a number o imporan iue. Fir, hey demonrae he imporance o he FOMC chedule in he manner marke orm expecaion on uure change in he ederal und rae. Second, hey how ha erm rae immediaely adju o new inormaion (characerized here by policy ance reveral and ouide o FOMC announcemen) in a manner conien wih he raional expecaion hypohei. However, once hi inormaion i incorporaed ino he erm rucure, ubequen, regularly cheduled, and anicipaed announcemen, appear o have lile impac on erm rae, explaining ome o he paern we repored in Table 5. Thi obervaion perhap explain he poor reul commonly obained in he lieraure on e o he raional expecaion hypohei (ee Rudebuch, 995 or a urvey). 7. Concluion The February 3-4, 994 FOMC meeing concluded wih one o he mo igniican procedural change in he Federal Reerve in recen ime he deciion o publicly announce change in he und rae arge immediaely aer he even. Thi procedural modiicaion coniue a igniican deparure rom he recalciran ecreive behavior ha had characerized radiional cenral banking. Among he reaon ha would juiy hi revoluion, i i naural o conjecure ha he new policy aord beer and more precie conrol o he overnigh rae and beer communicaion on uure policy move: in eence, beer conrol o long rae by manipulaing expecaion more adroily. Ulimaely he goal o moneary policy require ha he Fed be ucceul in guiding he marke expecaion o uure policy move, and hu, ha here be ome ynchroniciy in longer mauriy rae. The evidence we preen in hi ron demonrae We remind he reader ha expreion (3) conain wo dummy variable or he /3/ and 9/7/ obervaion a explained in ecion 5. They are omied here or breviy. 7

ha ince 994, he announcemen and he FOMC chedule allow marke o beer anicipae he iming and he naure o uure policy move. To be clear, we are no arguing ha prior o 994, he Fed wa unable o ignal he marke wha i preerred und rae objecive wa. However, he evidence ha we preen rongly ugge ha erm rae reac rongly when he Fed iniiae a policy ance reveral a repone ha i enirely conien wih he raional expecaion hypohei and wih a beer underanding on how he Fed implemen policy. 8

Reerence Akaike, Hirougu (97) Saiical Predicor Ideniicaion, Annal o he Iniue o Saiical Mahemaic,, 3-7. Akaike, Hiougu (974) A New look a he Saiical Model Ideniicaion, IEEE Tranacion on Auomaic Conrol, 9, 76-73. Bai, Juhan and Pierre Perron (998) Eimaing and Teing or Muliple Srucural Change in Linear Model, Economerica, 66, 47-78. Bai, Juhan and Pierre Perron () Compuaion and Analyi o Muliple Srucural Change Model, mimeo, Boon College. Bai, Juhan and Pierre Perron () Muliple Srucural Change Model: A Simulaion Analyi, mimeo, Boon College. Barolini, Leonardo, Giueppe Berola and Aleandro Prai () Day-o Day Moneary Policy and he Volailiy o he Federal Fund Rae, Journal o Money, Credi and Banking, 34(), 37-59. Cloue, J. A. and J. P. Dow () A Compuaional Model o Bank Opimal Reerve Managemen Policy, mimeo, Federal Reerve Board. Cook, Timohy and Thoma Hahn (989) The Eec o Change in he Federal Fund Rae on Marke Inere Rae in he 97, Journal o Moneary Economic, (4), 33-5. Craven, P. and G. Wahba (979) Smoohing Noiy Dae wih Spline Funcion, Num. Mah. 3, 377-43. Guhrie, Graeme, and Julian Wrigh () Open Mouh Operaion, Journal o Moneary Economic, (46)-, 489-56. Demiralp, Selva and Òcar Jordà () The Announcemen Eec: Evidence rom Open Marke Dek Daa, Economic Policy Review, proceeding o he conerence Financial Innovaion and Moneary Tranmiion, Federal Reerve Bank o New York, 8(), 9-48. Hamilon, Jame D. (997) Meauring he Liquidiy Eec, American Economic Review, vol. 87, no., 8-97. Hurvich, Cliord M. and Chih-Ling Tai (989) Regreion and Time Serie Model Selecion in Small Sample, Biomerika, 76, 97-37. 9

Kuner, Kenneh N. () Moneary Policy Surprie and Inere Rae: Evidence rom he Fed Fund Fuure Marke, Journal o Moneary Economic, orhcoming. Liu, J., S. Wu, and J. V. Zideck (997) On Segmened Mulivariae Regreion, Saiica Sinica, 7, 497-55 Perron, Pierre (997) L eimaion de modèle avec changemen rucurel muliple, Acualié Economiqué, 73, 457-55. Rice, J. Bandwidh Choice or Nonparameric Kernel Regreion, Annal o Saiic,, 5-3. Rudebuch, Glenn D. (995) Federal Reerve Inere Rae Targeing, Raional Expecaion and he Term Srucure, Journal o Moneary Economic, 35, 45-74. Erraum: December 995. Schwarz, G. (978) Eimaing he Dimenion o a Model, Annal o Saiic, 6. Taylor, John B. () Expecaion, Open Marke Operaion, and Change in he Federal Fund Rae, Federal Reerve Bank o S. Loui, Review, orhcoming. Thornon, Danniel L. () The Relaionhip beween he Federal Fund Rae and he Fed Federal Fund Rae: I i Open Marke or Open Mouh Operaion? Federal Reerve Bank o S. Loui, working paper 999-B.

Daa Appendix Federal Fund Rae Source: Releae H. 5, Seleced Inere Rae, bi-weekly ending on a Wedneday. Federal Reerve Board. Uni: in percen. Deiniion: weighed average o repored rae a which dieren amoun o he day rading hrough New York broker occur. Treaury Bill Rae Source: Releae H. 5, Seleced Inere Rae, daily daa, econdary marke rae. Uni: in percen Variable: 3- and 6- monh Treaury bill. Bond Yield Source: Bloomberg, Governmen ecor, on-he-run Treaury yield. Uni: in percen Variable: -, 5-, and - year Treaury ecuriie. Fed Fund Fuure Price Source: CI/FINANCIALS daabank o DRI, cloing price o Fed und uure conrac. Variable: Spo monh and -monh uure conrac. Timing Adjumen Following Kuner (, ), we apply wo iming adjumen or hoe cae when he arge change ook place aer he H.5 daa were colleced and he uure marke wa cloed (around 3.5 PM) o enure ha he policy change in he daa line up wih policy move. For hee dae (/8/99 and /5/998), Treaury bill rae are aligned wih he nex day in order o relec he marke rue percepion o he policy change. In paricular, he rae change on /8/99 i compued a he dierence beween /8/99 and /9/99, and he rae change or /5/998 i compued a he dierence beween /5/998 and /6/998, uing H.5 releae. The marke urprie i compued in a imilar ahion or hee dae, aking he dierence beween he cloing po price on he day o policy move and he opening rae on he ollowing day (a oppoed o he po price change on he day o policy move). No adjumen are applied o longer erm yield, a Bloomberg erie do relec policy change (ee Kuner () or more deail).

Table Srucural Break Te Mauriy SupF(l ) UDMax WDMax 3-monh.7.7.7 6-monh 7.8 7.8 7.8 -year.3.3.3 5-year.4.4.4 -year 7.34 7.34 7.34 Noe: () denoe igniican a he 9%(95%) conidence level. Mauriy Dae o break 3-monh July, 99 6-monh Sepember 4, 994 -year February 4, 994 5-year February 4, 994 -year April 8, 994 Noe: July, 99 and Sepember 4, 99 are he econd o la and la arge change beore he February 4, 994 change in policy. Similarly, April 8, 994 i he econd arge change aer February 4. Table Value or Dieren Inormaion Crieria or he Break Dae Deeced wih he Bai and Perron (998) Te. Mauriy Break ACI C RICE GCV SIC LWZ 3-monh -97.9-4.78-4.7-4.64-4.75-97.87-4.75-4.73-4.55-4.5 6-monh -8.8-5.8-5.8-5. -5. -6.68-5.5-5.6-5.8-5.5 -year -96.54-4.68-4.68-4.6-4.7-96.7-4.69-4.7-4.53-4.5 5-year -9. -4.58-4.59-4.5-4.63-9. -4.57-4.58-4.4-4.38 -year -3.87-4.8-4.8-4.75-4.86 -.65-4.8-4.8-4.64-4.6 Noe: indicae he minimum value.

Table 3. End-o-Monh Eec Dae Acual Change Marke Expecaion 7/7/989 -.5 -.5 /9/99 -.5.6 4/3/99 -.5 -.8 /3/99 -.5 -. /3/996 -.5 -.8 9/9/998 -.5 -.5 6/3/999.5.9 /3/ -.5 -.5 6/7/ -.5 -.5 3

Table 4. Ordered Probi Eimae Coeicien 989 994 994 4.77 7.9 E ( ) (.76) (.) A FOMC 6. 9.35 (4.3) (.) End o Monh -4.45-7.75 (.94) (.3) End o Mainenance Period.74.6 (.4) (.7) Avg. Log-Likelihood -.85 -.9 Peudo R.8.7 Obervaion 9 89 Noe: Eimae o limi poin omied. () indicae igniican a he 99%(95%) conidence level. Prediced Frequencie 989 994 994 Value Acual Pred. Error Value Acual Pred. Error -.5 3 -.5 8 - -.5 -.5 9 9 7 3-3 58 67-9.5.5 9 7.5.5 5 5 Eimae baed on buine daily daa. Speciicaion o he condiional mean i given by r monh mp omc = E ( ) β md E ( ) β pd E ( ) β d E ( ) ε γ (9) Noe ha E ( ) i compued a decribe in expreion (8). 4

Table 5. The Repone o Term Rae o he Componen o a Targe Change = α β E ( ) ) β R ( ε v () β β R Repone o Anicipaed Repone o Unanicipaed Mauriy 989 994 994 989 994 994 3-monh.33 (.5).7 (.4).9 (.3).8 (.) 6-monh.9 (.3).4 (.4).88 (.).63 (.) -year. (.7).4 (.5).74 (.4).5 (.5) 5-year.6..6.39 (.6) -year.5 (.6) (.6) -.3 (.5) (.3).43 (.3) (.7).5 (.5) Mauriy Dummie 3-monh 6-monh -year 5-year -year January 3,.8 (.8).4 (.8).3 (.).39 (.).37 (.) Sepember 7,.4 (.8). (.8).8 (.).6 (.).8 (.) Noe: expreion () i eimaed only on day in which he arge wa changed. There are 4 obervaion or he pre-994 (5/8/989 /3/994) ample, and 3 obervaion or he po-994 (/4/994 //) ample. Sandard error in parenhei. // indicae igniican a he 99% /95%/9% conidence level. 5

Table 6. The Timing Hypohei and he FOMC Schedule R β d = α α d SWITCH E ( ) FOMC ( α d ) d SWITCH SWITCH β E ( ) β ε β d ( ) β d E ( ) (3) FOMC ε β d FOMC SWITCH ( ) ε v Mauriy Timing 3-monh 6-monh -year 5-year -year A FOMC.63.8.3.3. (.8) (.5) (.) (.) (.9) Ouide FOMC.78.6.64.49.34 (.7) (.7) (.8) (.9) (.8) A FOMC SWITCH.4.68.73.76.4 (.5) (.44) (.56) (.6) (.54) Ouide FOMC SWITCH.58.9.6..74 (.53) (.47) (.6) (.64) (.58) R.75.69.6.5.48 Noe: Sandard Error in parenhei. () indicae igniican a he 99%(95%) conidence level. Calculaion o he Coeicien: A FOMC: β β Ouide FOMC: β A FOMC SWITCH: β β Ouide FOMC SWITCH: β β β Mauriy Dummie 3-monh 6-monh -year 5-year -year January 3,.48 (.3).7 (.).9 (.5).48 (.7).94 (.4) Sepember 7,. (.7).7 (.6).3 (.8).6 (.8).9 (.7) Noe: Sandard Error in parenhei. () indicae igniican a he 99%(95%) conidence level. 6

Figure. Mainenance Period Seaonaliy 4 % 3 5/9/989 - /3/994 - - 3 4 5 6 7 8 9 Day o he Mainenance Period 4 % 3 /4/994-8/3/998 - - 3 4 5 6 7 8 9 Day o he Mainenance Period 4 % 3 8/4/998 - // - - 3 4 5 6 7 8 9 Day o he Mainenance Period 7

Figure. End-o-he Monh Eec 3 % 5/9/989 - /3/994 - - -3 4 8 6 Day Le in he Monh 3 % /4/994-8/3/998 - - -3 4 8 6 Day Le in he Monh 3 % 8/4/998 - // - - -3 4 8 6 Day Le in he Monh 8