Anticipations of Monetary Policy in Financial Markets

Author(s): Joe Lange, Brian Sack and William Whitesell

Source: Journal of Money, Credit and Banking, Vol. 35, No. 6, Part 1 (Dec., 2003), pp. 889-909
Published by: Ohio State University Press
Stable URL: http://www.jstor.org/stable/3649863 .
Accessed: 05/02/2015 22:23

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

Ohio State University Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of
Money, Credit and Banking.

http://www.jstor.org

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE
BRIAN SACK
WILLIAMWHITESELL

Anticipationsof MonetaryPolicy in Financial

Markets

In recent years, financial marketsappearbetter able to anticipateFederal

Open Market Committee (FOMC) policy changes. Beginning in the late
1980sandearly 1990s, longer-terminterestratesandfuturesrateshavetended
to incorporate movements in the federal funds rate several months in
advance, in contrast to the largely contemporaneousresponse typically
observedbefore thattime. Afteridentifyingthese emergingtrends,the paper
parses the enhanced predictabilityinto a componentthat can be attributed
to the autoregressivebehavior of the funds rate and a nonautoregressive
component.The paperconsidersinstitutionaldevelopmentsin FOMCpolicy
making that may have contributedto each of these components, including
gradualismin adjustingthe federalfundsratetargetandtransparencyregard-
ing the setting of the target and future policy intentions.

THROUGH ADJUSTING the supply of reserves, a centralbank

can effectively control a single asset price-the short-terminterestrate. In forward-
looking financialmarkets,however, expectationsregardingthe futurepath of short-
term rates can importantlyinfluence longer-terminterest rates and stock prices,
which have a considerable effect on private spending decisions. Therefore, the
effectiveness of monetary policy depends in part on the speed and extent of
the transmissionof policy decisions and intentions to other asset prices. Indeed,
Woodford (1999) shows that the forward-lookingnature of financial markets can
have importantimplicationsfor determiningthe optimal setting of monetarypolicy.
The degree to which futuremonetarypolicy actions are anticipatedand built into
broader financial market prices has been widely studied. One of the important

We thankparticipantsof the MonetaryAffairs lunch workshopat the FederalReserve Board, as well

as the editor and two anonymous referees, for their comments. The views expressed are those of
the authorsand not necessarily those of the Board of Governorsof the FederalReserve System or other
members of its staff.
JOELANGEis an economist at Cornerstone Research. BRIAN SACKis a senior economist
in the Division of Monetary Affairs of the Board of Governors of the Federal Reserve
System. WILLIAMWHITESELLis the Deputy Associate Director of Division of Monetary
Affairs of the Board of Governorsof the Federal ReserveSystem.E-mail: william.c.whitesell
@frb.gov
Journal of Money, Credit, and Banking, Vol. 35, No. 6 (December 2003, Part 1)
Copyright 2003 by The Ohio State University

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
890 : MONEY, CREDIT,AND BANKING

contributionsto this literatureis Mankiw and Miron (1986), who found that the
yield curve has had little predictive power for future changes in short-termrates
after the inception of the Federal Reserve in 1914. They argue that this result
arises because the central bank tends to smooth interest rates, inducing random
walk behavior that makes future rate changes largely unpredictable.The lack of
predictabilityhas been supportedby numerousstudies, including Shiller,Campbell,
and Schoenholtz(1983) and Campbelland Shiller (1991). However, otherresearch-
ers, including Fama (1984), Mishkin (1988), Hardouvelis (1988), and Longstaff
(2000), have found that the yield curve does contain some informationregarding
future interestrate changes over particularhorizons.
The variationin the findings of these papers partly reflects differences in their
empirical approaches,specifications, and data. Nevertheless, a general conclusion
seems to have emergedfrom this literature,as articulatedby Cook and Hahn (1988)
and Rudebusch(1995): The very shortend of the yield curve displays some ability
to predict changes in short-terminterest rates, but this predictive power fades
fairly quickly as the horizon lengthens.1Rudebusch(1995), for example, finds that
the yield curve has no informationfor short-ratechanges beyond a month, after
which the random walk behavior noted by Mankiw and Miron takes effect.
However, disagreementsremain regardingthe extent to which interest rates have
predictive content and the horizon over which any predictive power exists.
We considerfurtherthe degreeto which financialmarketsanticipatefuturechanges
in short-terminterest rates, focusing on the years subsequent to the shift in the
FederalReserve'soperatingregime from nonborrowedto borrowedreservetargeting
in the early 1980s.2 Our results indicate that an importantshift occurredin the late
1980s and early 1990s in the ability of financial markets to anticipatemonetary
policy actions. Through most of the 1980s, market prices had predictive power
for policy actions only about a month ahead and responded substantially to
contemporaneousmovements in the federal funds rate. More recently, however,
market yields have become much better predictors of monetary policy moves
severalmonthsin advance,while the responseto contemporaneouspolicy moves has
diminished.
We review a numberof hypothesesthat could accountfor such results, including
changes in marketefficiency, in the natureof the shocks hitting the economy, and
in the behaviorof the Federal Open MarketCommittee(FOMC).While we cannot
separate the impact of all the possible explanations, we can parse the enhanced
predictabilityinto a componentassociatedwith the autoregressivepropertiesof the
funds rate and a component associated with other factors. Regarding the first of
these, it appearsthat the serial correlationin first differences of the federal funds
ratehas increasedof late, likely contributingto the improvedpredictabilityof policy
changes. This shift could reflect changes in the natureof the shocks to which the
FOMC has responded or greater gradualismin the strategy of implementationof
policy by the FOMC.
But the dynamics of the funds rate account for only part of the improvedability
of financial markets to anticipatemonetary policy. A considerable portion of the

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAMWHITESELL : 891

improvementinstead appearsto reflect otherfactors. Among those potentialfactors,

severalinstitutionalchangeshave mademonetarypolicy moretransparent,including
the shift away from the borrowedreserve operatingregime and towardstrictfederal
funds targeting in the latter part of the 1980s, as well as the Federal Reserve's
provisionof more explicit informationregardingpolicy targetsandtheirrationalesin
recent years.
In Section 1 of the paper,we presentevidence thatthe ability of financialmarkets
to anticipateFOMC policy changes improved over the period 1983-2000. Section
2 reviews a range of hypotheses that could possibly explain these results. Section 3
examinesempiricalevidence on whetherthe greateranticipatorybehaviorof financial
marketscan be explainedby a shift in the autoregressivepropertiesof the federalfunds
rate,or whetherotherfactors,such as increasedtransparency,may also have contrib-
uted by making FOMC policy choices more predictable.Section 4 concludes.

1. FINANCIAL MARKETS AND MONETARYPOLICY

In investigating the response of financial markets to monetary policy changes,

we first focus on the behavior of Treasuryand private market interest rates. The
simple cross-correlationsshown in Figure 1 provide a first glimpse of the dynamic
relationshipbetween marketrates and the federal funds rate. The figure shows the
correlationof changes in the federal funds rate with changes in yields on six-month
Treasurybills, two- andten-yearTreasurynotes, andBaa-ratedcorporatebonds.3The
correlationscompare a funds rate change at time 0 with changes in marketrates in
preceding months (negative numbers along the horizontal axis) and subsequent
months (positive numbers).Three subsamplesare considered,each with at least five
years of data. (The specific dates dividing the subsamples are discussed later in
the paper.) Between 1983 and 1989, market rates tended to move more or less
contemporaneouslywith changes in the fundsrate.Indeed,the correlationcoefficient
on contemporaneouschanges approaches0.75 for the six-month Treasurybill and
exceeds 0.5 even for long-termyields. The correlationsremainhigh for the movement
of market rates in the month following a change in the funds rate. Moreover,
anticipatorybehavior was fairly limited over that sample. Changes in the funds
rate are positively correlatedwith the previous month's changes in marketrates, but
the correlationsfall off quickly for changes in marketrates furtherin advance.
The patternof correlationssince February1994, however,has been quite different
from the earlier period. Changes in the funds rate have been substantiallyless
correlatedwith either contemporaneousor subsequentmovements in marketrates,
and more highly correlatedwith changes in market rates two and three months
beforehand.The shift in behavior apparentlytook place over time between the late
1980s and early 1990s, as shown by the partialshift in the correlationsin the inter-
mediate sample. Moreover,the shift is statisticallysignificant;while contemporane-
ous and one-month-aheadmovements in marketyields are significantly correlated
with funds rate changes in each of the three subperiods,movementsin marketyields

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
892 : MONEY, CREDIT, AND BANKING

Six-Month Treasury Bill Two-Year Treasury Note

0.75
0.75 - - - 1983:01 - 1989:03
- -1989:04 - 1994:01 I

- I
1994:02 - 2000:10

0.5 -0.5

'-
0.25
' \ \
0.25
",

000

-0.25 -0.25
-6 -5 -4 -3 -2 -1 0 1 2 -6 -5 -4 -3 -2 -1 0 1 2
Monthsrelativeto funds rate change Monthsrelativeto funds rate change

Ten-Year Treasury Note Baa Corporate Bond

0.75 0.75

0.5 0.5

0.25 0.25

-0.25 -0.25
-6 -5 -4 -3 -2 -1 0 1 2 -6 -5 -4 -3 -2 -1 0 1 2
Monthsrelativeto funds rate change Monthsrelativeto funds rate change

FIG.1. Correlationsbetween changes in the marketrate in month indicatedrelative to changes in the federal funds
rate at time 0. (Negative numbersindicate the numberof months in advance of federal funds rate changes)

two and three months ahead only became significantly correlatedwith subsequent
funds rate movements in the two more recent subperiods.4

1.1 The Predictive Power of the YieldCurve

While the above results indicate that longer-term interest rates have adjusted
increasingly in advance of changes in the federal funds rate, they do not address
whether market rates were moving to levels that were consistent with the future
path of the federal funds rate. To look at this question, we add some additional
structureto the empirical investigation.

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAMWHITESELL : 893

The forward-lookingnatureof longer-terminterestrates is perhapsmost simply

described by the expectations hypothesis, which states that the interest rate on a
longer-termn-period pure discount bond (in) is a simple average of future short-
term (one-period)interestrates, plus a constantterm premium.If we take the one-
period interest rate to be the federal funds rate (ff), the following equationresults,
with Et signifying expectations as of the end of period t:
1
ti Etfft+i + c. (1)
n i=l

In our data set, ff,+i is equal to the average effective federal funds rate over the
month t + i, and in is the marketrate at the end of month t. Any differencesin tax
treatment,risk, or liquidity are assumed to be capturedby the term premium,c.
The above equationhas been tested in variousforms by many researchers,includ-
ing Campbelland Shiller (1991) and the references they cite. Under one approach,
they note that the yield spreadbetween the longer-termrate and the short-termrate
should predictfuturechanges in the shortrate. This relationshipis obtainedsimply
by subtractingthe currentlevel of the short-termratefromboth sides of Equation(1):
I
in _- ffd - c - Efft+i
- fftend, (2)
fftend
F/i=I

where ffend is the funds rate at the end of the month (as opposed to the month-
average funds rate, fft).5 In words, the spreadbetween the longer term rate and the
federal funds rate at the end of the currentmonth, less a term premium, should
equal the expected change in the funds rate from its currentlevel to its average
level over the maturityof the longer-terminstrument.6
To test this equation,one can replacethe expected federal funds rate in the right-
hand side of Equation(2) by its actuallevel to determinethe yield spreadthatwould
be realized under perfect foresight. That value can then be regressed on the actual
yield spread (from the left-hand side of Equation2), as in the following equation:
n
- ftend= end + .
fft+ i o + (i~ (3)
n i=l

The errorterm will contain an expectation error,which will be uncorrelatedwith

the actual yield spreadunder rationalexpectations.7According to the expectations
hypothesis, the regression coefficient, P, should be unity, and the longer-termrate
should embody all predictablefuture movements in the short rate. The R2 statistic
from the regressionthen provides a measure of the portion of the futurechange in
the federal funds rate that is anticipatedby the longer-terminterest rate, which is
the primaryfocus of this paper.
We first estimate Equation (3) for the three-monthTreasurybill using monthly
data over our entire sample period of January 1983 through October 2000. The
results indicate a very low value for the R2 statistic, which is typical for such tests.
But we are more interestedin determiningif market expectations have improved,

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
894 : MONEY, CREDIT,AND BANKING

or whether the predictive power of this equation has increased over time. We
therefore split the sample at varying break points and compute the difference in
the R2 statisticbetween the earlier and later period for each break date. The results
areplottedin Figure2. As the breakdatemoves into the late 1980s, the improvement
in R2 increases considerably,rising from 0% to 27% for a break date at April 1989.
After leveling out for a while, the improvementagain rises substantially,reaching
a peak of around60% in early 1994.
These results indicate that the ability of three-monthTreasurybills to predict
federal funds rate changes increased remarkablyover the sample. Moreover, the
improvementappearsto have come over a period of time between the late 1980s
and early 1990s. To capturethis pattern,we introducebreakpoints in April 1989 and
February 1994 to create three distinct subsamples. We do not suggest that these
precise dates were associated with any sharp change in market behavior;rather,
they allow us to divide our sample period into three relatively equal segments and
also serve reasonably well in helping to characterizethe differences in market
behavior that have emerged graduallyin recent years.
The top portion of Table 1 reportsthe results from estimating Equation(3) over
the three subsamples. In the 1980s, the regression has a very poor fit, with an R2
statisticnot far above 0. Moreover,the coefficientestimatedover that periodis well
below the value of unity predictedby the expectationshypothesis.As emphasizedby

1.0-

Apr 1989 Feb 1994

0.8-

0.6-

0.4-

0.2

0.0

-0.2J
oo t00 ow ow w 00 w mm)
m 0 0 ) 0) D 0) D0) 00) 0
0) m
cn mm) 0) 0) 0 0 0)D 0 0)
D 0) 0) 0) c0 0
TTI-- Cl

FIG.2. Increasein the R2 statistic from the earliersubsampleto the later subsamplewhen the sample is divided at
the date indicated

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAMWHITESELL : 895

TABLE 1
PREDICTING FEDERAL FUNDS RATE CHANGES WITH TREASURY BILL YIELDS (Campbell-Shiller
Regressions)

Specification A: 1 - nd + (i nd) et
/•i=l +i 0 -

SpecificationB: 1 fft
- o+ - i _)? Et
mi=1 n i=l f+i t(it
•
Coefficients
Number of
Observations Constant Spread R2 Statistic

Spreadof Three-MonthTreasuryBill over Funds Rate (SpecificationA, n = 3)

1983:1-1989:3 75 0.191 0.230 0.025
(1.63) (1.44)
1989:4-1994:1 58 -0.087 0.263 0.177
(-2.01) (4.37)
1994:2-2000:10 81 0.146 0.556 0.587
(6.57) (7.52)
Spreadof Six-Month over Three-MonthTreasuryBill (Specification B, n = 3, m = 6)
1983:1-1989:3 75 0.128 -0.296 0.010
(0.65) (-0.57)
1989:4-1994:1 58 -0.336 0.752 0.038
(-4.45) (1.33)
1994:2-2000:10 81 -0.147 1.092 0.382
(-2.53) (5.99)
NOTES: This specificationcorrespondsto Table 2 from Campbelland Shiller (1991). All standarderrorsare correctedfor heteroskedasticity
and serial correlationusing the Newey-West procedure.

Mankiw and Miron (1986) and others,this finding likely reflectsthe downwardbias
in the coefficient that arises when some of the movements in the yield spreadare
driven by a time-varying term premium. In later intervals, however, the fit of
the equationhas improvedconsiderably,with the R2 statisticjumping to nearly60%
in the period since 1994. In addition, the estimated coefficient has moved much
closer to one, presumablybecause a greater share of the movements in the yield
spread reflects expected funds rate changes rather than movements in the term
premium.
In the bottom portion of the table, the spreadbetween the six- and three-month
Treasuryyields is tested for its ability to predictfuturechanges in the federal funds
rate. In particular,we write Equation (3) separately for the three- and six-month
Treasurybills and then take the differenceof these equations.The resultingequation
predicts changes in the federal funds rate from its averagelevel over the next three
months to its average level four to six months ahead. No relationship between
the spread and federal funds rate changes over that horizon is evident in the two
earlierperiods,butthe recentintervalshows some predictiveability,with a significant
coefficient around unity and an R2 of nearly 40%. Overall, the results strongly
indicate that there has been a considerable improvement in recent years in the

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
896 : MONEY, CREDIT, AND BANKING

abilityof Treasurybill yields to predictfederalfundsratechanges over a three-month

or slightly longer horizon.
An importantconsiderationin interpretingthese results is the potentialfor small-
sample bias. Bekaert, Hodrick, and Marshall(1997) argue that the coefficient 0 in
Equation(3) is biased upwardin small samples as a result of the persistence of the
short-terminterestrate. As a result,the R2 statisticsfrom the regressionswould also
be biased upward. However, it turns out that the bias is rather small under the
dynamic process of the short-terminterest rate estimated below (Section 3) for
the relatively short-termmaturitiesthat we consider. Moreover, the bias does not
appearto have shifted significantlyover the sample, and so it does not account for
the sharprise in the R2statisticsin the lattersubsamples-the focus of this paper.The
impact of the small sample bias on the results is explored in further detail in
the Appendix.

1.2 The Predictive Power of Futures Rates

Over recentyears, privateshort-terminstrumentsand futurescontractshave come
to rival, and even surpass, Treasurybills in terms of liquidity. There are two sets
of futurescontractsthat may be particularlyuseful for predictingfederal funds rate
movements: federal funds futures contracts that trade on the Chicago Board of
Trade, and eurodollarfutures contracts that trade on the Chicago Mercantile Ex-
change. We use these contractsto derive equationsto predict changes in the federal
funds rate that are similar to the ones based on Treasurybill yields above.
Federal funds futures contracts have a payout that is explicitly based on the
averageeffective federalfundsrateover a particularmonth.As a result,the following
equation should hold:

futff(t + i) - fftend- c = - fftend (4)

Etfft+i
where fft+iis again the average federal funds rate in month t + i, fftendis the funds
rate at the end of the currentmonth, and fut4(t + i) is the federal funds futuresrate
for the contract i months ahead taken at the end of the currentmonth. Again, c
representsa term premium, which is assumed constant across time for contracts
with the same maturity.
As with Equation (2), we replace the right-handside of Equation (4) with the
actual change in the federal funds rate and regress it on the change predicted by
the futuresrates. That is, we estimate the following regression:

(fft+i - fftend) = ( + 3(fut(t + i) - ffend) + Et , (5)

where again 0 should equal unity and the R2 statisticmeasuresthe predictabilityof

changes in the federal funds rate over the next i months. Unfortunately,federal
funds futures contracts were not introduceduntil the late 1980s, so our data are
monthly beginning in May 1989. Our results are thus limited to the two later
subperiodsidentifiedabove. Also, we restrictourselves to predictionsof up to three
months ahead, as federal funds futurescontractshave been less liquid beyond that

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAMWHITESELL : 897

horizon. The results, shown in the upperpartof Table2, indicate a markedimprove-

ment since 1994 in the proportionof federal funds ratemovementsthatare predicted
by futures prices. As was the case for Treasury bills, the R2 statistics have
increased sharply for all three contracts. These futures contracts have explained
around70% of the variance of two- and three-monthchanges in the federal funds
rate in recent years.8
Eurodollarfutures are liquid at longer horizons than those on the federal funds
rate. At its expirationdate, which is aroundthe middle of the last monthof a quarter,
the value of a eurodollarfuturescontractis basedon thethree-montheurodollardeposit
rate at thattime, which is heavily influencedby the outlook for the federalfunds rate
over the subsequentthree months. Adjusting for the horizon of the contracts, an
equation similar to that for federal funds futures should hold:

TABLE 2
PREDICTING FEDERAL FUNDS RATE CHANGES WITH FUTURES RATES

SpecificationA: (fft+i - ff~nd) = po + pl(futt(t + i) - fftend)) + et

SpecificationB: (fft+i- fft+i-1) = 3o+ Pl(futt(t + i) - futt(t+ i - 1)) + Et
Coefficients
Number of
Observations Constant Spread R2 Statistic

Federal Funds Futures

One month ahead (SpecificationA, i = 1)
1989:5-1994:1 57 -0.030 0.232 0.035
(-1.53) (1.09)
1994:2-2000:10 81 -0.019 0.797 0.375
(-2.21) (5.33)
Two months ahead (Specification A, i = 2)
1989:5-1994:1 57 -0.121 0.487 0.131
(-3.19) (2.79)
1994:2-2000:10 81 -0.038 0.915 0.706
(-2.27) (8.27)
Three months ahead (SpecificationA, i = 3)
1989:5-1994:1 57 -0.209 0.592 0.170
(-3.20) (2.73)
1994:2-2000:10 81 -0.043 0.868 0.666
(-1.50) (8.43)
EurodollarFutures
First contract(Specification A, i = 1)
1988:1-1993:4 19 -0.224 0.177 0.038
(-3.74) (0.87)
1994:1-2000:3 27 -0.119 0.627 0.402
(-2.26) (3.97)
Second contract(Specification B, i = 2)
1988:1-1993:4 19 -0.283 0.240 0.026
(-2.49) (0.69)
1994:1-2000:3 27 -0.002 0.529 0.238
(-0.04) (3.41)
NOTES: Regressionspredictthe change in the federalfunds rateover the horizonindicated.All standarderrorsare correctedfor heteroskedas-
ticity and serial correlationusing the Newey-West procedure.For eurodollarfutures,the firstcontractpredictsthe averagefederal funds rate
over the coming threemonths, and the second contractpredictsthe change from that average to its average from four to six months ahead.

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
898 MONEY, CREDIT, AND BANKING

futed(t + i) _ fftend- c = Et ff+i - fftend. (6)

Here, time is measuredin quartersratherthan months, with observationstakenjust
before the expirationdates of the futures contracts.The eurodollarfutures rate on
the ith contractobserved on day t is denoted futed(t + i), and the funds rate on that
day is ffpnd.The first futurescontract(i = 1) is just about to expire, so that its rate
will essentiallyequal the three-montheurodollardepositrate,which reflectsthe aver-
age federal funds rate expected over the three months beginning on the date of
expirationof the contract,denotedffq+1.The second contract(i = 2) would similarly
reflect the expected federal funds rate over the three months beginning at the
expirationof the second contract,ffq+2. The differencein the futuresrates can then
be used as a predictorof the change in the federal funds rate from its average level
over the next three months to its average level four to six months ahead.
The regression results, using data from 1988 throughthe third quarterof 2000,
are shown in the lower part of Table 2. Again, a remarkableimprovementin the
R2 is evident since early 1994, and the regressionexplains 40% of the variationin
federal funds rate changes over a three-month horizon. In addition, as found
above for Treasurybills, some predictiveability has even emergedfor federal funds
rate changes four to six months ahead, which was nonexistent before 1994.

1.3 Policy Surprises

The results from Tables 1 and 2 provide evidence that marketinterest rates are
building in a large portionof FOMC policy actions several months in advance. As
a consequence,marketsshouldnow be less surprisedat the time of the announcement
of a monetarypolicy action and, therefore, should react less than in the past. We
test for this, measuring the surprise component of the FOMC's policy decisions
by the change in the currentmonth federal funds futuresrate on the day the policy
decision became known, as in Kuttner(2001).9Because the futurescontractis based
on the average federal funds rate over the month, the amountby which the futures
rate reacts to an unexpectedpolicy change depends on its timing within the month.
A measure of the policy surprisethat accounts for this timing is:10

Policy Surprise- Afut", (7)

D-d
where Afutf is the change in the currentmonth federal funds futures rate on the
day that the federal funds rate change became known, d is the day of the month of
the policy action, and D is the total numberof days in the month.
The average size of this statistic over differentsubsamples,reportedin Table 3,
reveals a substantialdecline in the unanticipatedcomponent of policy moves over
the more recent period. In the period before 1994, the average size of a federal
funds rate change was 27 basis points, of which 16 basis points, or 59%, came as
a surprise to the market. Since 1994, however, while the average size of policy
moves has increased to over 32 basis points, only 8 basis points of these moves,
or 24%, has been a surprise."11

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAM WHITESELL : 899

TABLE3
MONETARY POLICY SURPRISES DERIVED FROM FEDERAL FUNDS FUTURES

PolicySurprise= D-d
D Afutf

Average Absolute Value (in Basis Points)

Change in Intended Surpriseas

Funds Rate Policy Surprise Percent of Change

1989:5-1994:1 27.3 16.1 58.9

1994:2-2000:10 32.5 7.8 24.0

NOTES: In the definitionof policy surprises,d is the day of the month of the policy action, and D is the total numberof days in the month.
See the text for more details.

2. POSSIBLEINTERPRETATIONS

The results of the previous section suggest that a significant shift has occurred
in the relationshipbetween financial market prices and movements in the federal
funds rate. A wide range of interpretationsof these developments is possible,
includinghypothesesregardingchanges in the efficiencyof markets,in the natureof
the economy, or in the behavior of the FOMC.
Some observershave arguedthatthe deepening of futuresmarketsin eurodollars
and federal funds since the late 1980s has reducedtransactioncosts and facilitated
arbitrage.Hatzius (1999) arguedthat the enhancedliquidity of futuresmarketshas
made other marketinterest rates more responsive to informed views on near-term
expectations for monetarypolicy. We believe any such effect to be ratherlimited,
however, as improved predictabilityof short rates has been evident also in the
Treasurybill market,which has been liquid enough throughoutthe last two decades
to provide ample opportunityfor position-takingand arbitrageactivity based on
policy expectations.
An alternativehypothesis is that the improved anticipatorybehavior of financial
markets has arisen in part from changes in the policy-setting practices of the
FederalReserve or in the transparencyof its intentions.Indeed, a numberof institu-
tional changes have takenplace over the past two decades. One of the most obvious
of these changes was the move toward a regime of explicit federal funds rate
targeting. Recall that from October 1979 through September 1982, the Federal
Reserve had employed a nonborrowedreserve operatingprocedure,directing the
Open MarketDesk to adjustnonborrowedreserves consistent with an intendedpath
for growth of the monetaryaggregates,within a wide range of permissiblefederal
funds rates. Throughmost of the rest of the 1980s, a borrowed reserve operating
regime prevailed,which was intermediatebetween nonborrowedreserveand federal
funds rate targeting.The FOMC's policy stance duringthis regime was expressed
in terms of a target level of discount window borrowingthat was associated with
an expected level of money growth and an expected range for the federal funds

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
900 MONEY, CREDIT, AND BANKING

rate. However, with an unannounced operating target for borrowing, the true
intentions of the FOMC were not always evident to the markets.
In the mid-1980s, the Federal Reserve tended to follow borrowing targets. On
occasions when the FederalReserve detecteda changein the relationshipof discount
window borrowingto the spreadbetweenthe federalfunds rateand the discountrate,
it would typically make "technicaladjustments"in the expected rangeof the federal
fundsrateratherthanchangetheborrowingassumption.No clearbreakpointexists for
the shift from borrowedreserve targetingto strict federal funds rate targeting,but
a few key datesin the transitioncan be mentioned.Following the stock marketcrash
in October 1987, the FOMC's directives specified that greaterflexibility should be
used, implying that the Desk should give priority to keeping the federal funds
rate close to expectations, rather than keeping borrowing at its target level. In
November 1988, in the face of another evident shift in the borrowing demand
function, the FOMC made a substantial technical adjustmentto the borrowing
assumptionitself ratherthan allowing the expected federal funds rate to change.
While the Desk had used a point value for the intended funds rate ratherthan a
rangeat certaintimes, it droppedany furtheruse of rangesafterJune 1989. Neverthe-
less, the marketmisinterpretedreserveadjustmentsmade aroundThanksgiving1989
as a policy change, and the Desk thereafterbegan to target the federal funds rate
even more closely and to signal that target more clearly to the markets.12Hitting
the borrowing assumption no longer took precedence over targeting the federal
funds rate in the implementationof policy, and the target federal funds rate was
readily discernableby marketparticipants.
The policy process has become more transparentin a numberof other ways as
well, which may have also contributedto an increasein the predictabilityof federal
funds rate movements.An importantstep towardincreasedtransparencytook place
in February1994, when the FOMC began announcingpolicy changes on the day
of its meetings and included reasons for those decisions that often held hints about
future policy intentions.13In July 1995, the FOMC began including explicit refer-
ences to the intendedfederal funds rate in the press releases following its meetings,
andin August 1997, it beganreferringto an explicit fundsratetargetin its directivesto
the Open Market Desk. Another major step took place in May 1999, when the
FOMCbegan releasingchanges in its policy "bias"at the conclusion of its meetings.
In January2000, the Federal Reserve announcedthat statementswould be released
afterevery FOMCmeeting and thatthe bias statementhad been modifiedto convey
the FOMC's views regarding the balance of risks in the economic outlook.14
Since that time, the press release has describedthe rationalefor the policy decision
at each meeting and the factors underlyingthe balance of risks. Lastly, increased
transparencymay also be evident in the natureof the speeches and public comments
of FOMC membersin recent years, which some marketparticipantsbelieve to be
more informativeabout future policy moves than in the past.
Partlyas a result of receiving clearerinformationabout actual policy setting and
the factorsunderlyingpolicy decisions,marketparticipantshave likely improvedtheir
understandingof the determinationof monetary policy over the period. Indeed,

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAM WHITESELL : 901

therehas been a rapidlyexpandingacademicliteratureon estimatingand evaluating

monetarypolicy rules, following the importantcontributionsof Taylor (1993). One
of the conclusions from thatliteratureis thatU.S. monetarypolicy can be reasonably
approximatedby simplepolicy rulesthatprescribesystematicresponsesto macroeco-
nomic data.This understanding,or perhapsan increasein the reliabilityof the policy
reactionfunctionin recent years, may have also improvedthe anticipatoryability of
the markets.15
Another possible explanation for the increased ability of financial markets to
anticipatemovements in the federal funds rate is that the dynamic behaviorof the
federalfunds ratehas shifted in a way thatmakes its changes more predictable.This
would be the case if, for example, changes in the federal funds rate have become
more serially correlated.Such a shift could reflect the natureof the shocks affecting
the economy, which may have become more persistent or serially correlatedover
recent years.16Alternatively,it could reflect the deliberateintention of the FOMC
to move the funds rate in incremental steps. Many estimated policy rules find
evidence of partial adjustmentin the target federal funds rate. As suggested by
Woodford(1999), such a strategymay allow a centralbankto reducethe volatilityof
short-terminterestrates, while anticipatoryadjustmentsof longer-termrates effec-
tively stabilize the economy. In fact, because long-term rates are now adjusting
fartherin advance of eventual policy moves, more gradualbehaviorby the FOMC
may have become optimal.17
In the results that follow, we do not attempt to empirically separate all of the
differing hypotheses that might account for the observed change in anticipatory
behavior.Instead,we parse out the improvementin the market'spredictiveabilities
into two components:one that is attributableto the autoregressivepropertiesof the
federal funds rate, and one that is attributableto nonautoregressivefactors such as
the increase in policy transparency.

3. DECOMPOSINGTHE IMPROVEMENTIN PREDICTIVEPOWER

As discussed in the previous section, the improved ability of financialmarkets

to anticipatemonetarypolicy over the latterpart of the sample could partlyreflect
a shift in the dynamics of the federal funds rate. While a numberof authorshave
investigated the univariate properties of the federal funds rate (e.g., Rudebusch
1995, andreferencestherein),we focus on the period since the onset of the borrowed
reserve operatingregime in the early 1980s and look for changes in behavior in
more recent years.
To investigatepossible alterationsin dynamic properties,we regress the monthly
change in the federalfunds rate on threeof its own lags over our three subsamples.18
As shown in Table 4, only the first lagged dependentvariable is significantwhen
the regression is estimated over the earliest subsample from 1983 to 1989, but
additionallags become increasinglysignificantin latersubsamples.The lengthening
of the autocorrelationfunction implies that informationabout funds rate changes

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
902 : MONEY, CREDIT,AND BANKING

TABLE4
DYNAMIC PROPERTIES OF THE FEDERAL FUNDS RATE

Afft
= p + 3lAfft-1+ 22Afft-2+ 3Afft-3+ Et
Coefficients

Constant Afft_1 Afft-2 Afft3 R2 Statistic

1983:1-1989:3 0.010 0.389 0.008 -0.039 0.148

(0.27) (3.36) (0.05) (-0.38)
1989:4-1994:1 -0.073 0.270 0.130 0.005 0.144
(-3.40) (2.04) (1.20) (0.03)
1994:2-2000:10 0.169 0.205 0.258 0.131 0.204
(1.09) (1.96) (2.47) (1.10)
NOTES:T-statisticsshown in parentheses,and all standarderrorsare correctedfor heteroskedasticityand serialcorrelationusing the Newey-
West procedure.

became availablefurtherin advance,althoughthe cumulativepredictiveinformation

in lagged rate changes, measuredby the R2 statistic, rose only in the most recent
period and by a fairly moderateamount.19
The above results suggest that changes in dynamic propertiesmay have made
only a limited contributionto the increasedpredictabilityof the funds rate in recent
periods. To assess this more carefully, we return to the Campbell-Shiller type
regressions from Equation (4) above, focusing on the three-monthTreasurybill.
For each subperiod,we first try to capturethe extent to which autoregressiveterms
help to predict the change from the end-of-month funds rate target to its three-
month-aheadaverage:We regress the dependentvariablein Equation(4) on lagged
changes of the funds rate and show the results as "SpecificationA" in Table 5. We
then addas an explanatoryvariablethe end-of-monthspreadbetweenthe three-month
Treasurybill andthe federalfundsratetarget;this regressionis labeled"Specification
B." Differences in fit between the two regressions indicate the extent to which
there is predictive power in market prices beyond what can be accounted for by
the autoregressivepropertiesof the funds rate.
As is evident from the R2 statisticof SpecificationA, the predictabilityassociated
with the dynamics of the federal funds rate improves only slightly in the middle
period and by 10%in the most recent subperiod.Moving to SpecificationB results
in virtually no improvementin fit in the first subsample, but it boosts the R2 by
12% in the middle period and by 40% in the most recent period. Thus, the results
suggest that the greater predictabilityof funds rate changes in the more recent
subperiodsowes predominantlyto factors other than the dynamic behavior of the
funds rate, perhapsreflecting the increase in policy transparencydiscussed above.
Indeed, as seen by comparingTables 1 and 5, the coefficientson the interestrate
spread are not significantlydifferentwhen lagged funds rate changes are included
in the regression. This is consistent with the conclusion that the predictive power
of the term spreadlargely reflects factors other than the dynamics of the funds rate
itself.20Note also thatlagged fundsrate changes sometimesremainsignificantwhen

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAM WHITESELL : 903

TABLE 5
PREDICTING FUNDS RATE CHANGES: AUTOREGRESSIVE COMPONENT AND THE THREE-MONTH TREASURY
BILLCOMPONENT
1
-end t- --
SpecificationA: 1 ft+i-
ff
nd
PO + 1
lAfft + PAfft- t-1 + 3Afft-
fft-2 + t

B:
Specification f t = o + PlAfft+ + ffnd) + e
3 i=1 3Afft-1 3Afft-2+ 4(i3-

Coefficients

Constant Afft Afft-1 Afft-2 _ ffd R2 Statistic

1983:1-1989:3
SpecificationA 0.090 0.483 -0.065 -0.096 - 0.120
(1.20) (2.60) (-0.51) (-0.50)
SpecificationB 0.159 0.461 -0.053 -0.055 0.168 0.132
(1.17) (2.34) (-0.39) (-0.25) (0.84)
Improvementin R2 0.012
1989:4-1994:1
SpecificationA -0.090 0.502 0.034 -0.177 - 0.150
(-1.89) (2.43) (0.41) (-0.96)
SpecificationB -0.052 0.389 0.069 -0.107 0.225 0.270
(-1.22) (2.00) (0.79) (-0.74) (3.88)
Improvementin R2 0.120
1994:2-2000:10
SpecificationA 0.027 0.305 0.324 0.073 - 0.221
(1.28) (2.78) (2.69) (0.61)
SpecificationB 0.124 0.119 0.158 0.018 0.499 0.622
(6.50) (1.10) (2.46) (0.18) (6.75)
Improvementin R2 0.401

NOTES: T-statisticsshown in parentheses,and all standarderrorsare correctedfor heteroskedasticityand serial correlationusing the Newey-
West procedure.

the spreadis included in the regression,probablybecause marketparticipantswere

unlikely to have fully built in the autoregressivepropertiesthat we estimate, given
that we fit those models with data that were not available in real time. For this
reason, the results may attributetoo large a share of the R2 to the autoregressive
component.
A similar exercise was performedusing federal funds futures (correspondingto
the upper panel of Table 2), with horizons of one-, two-, and three-monthsahead.
Table 6 reports on regressions involving lagged funds rate changes either with or
without futures marketquotes over the differentsubsamples(SpecificationsB and
A, respectively). The futures marketquotes add only a modest improvementin fit
over the 1989-1994 period, but a substantialamount in the 1994-2000 period.21
The decompositionthatwe have madebetween autoregressiveand nonautoregres-
sive componentsis subject several caveats. For example, the relative importanceof
the two componentsis influencedto some extent by the assumedspecificationof the
autoregressiveprocess.22Also, one shouldbe cautiousabouttime series findingsfrom
such short samples. That said, however, the results under alternativespecifications

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
904 : MONEY, CREDIT,AND BANKING

TABLE 6
PREDICTING FUNDS RATE CHANGES: AUTOREGRESSIVE COMPONENT AND THE FUTURES RATE
COMPONENT

SpecificationA: (fft+i- fffnd) = + PlAfft + 2Aft-1 + 3Afft-2

+
Et
f0o
SpecificationB: (fft+i - fftnd) = 1o? lff + f32Afft-1 + 3Afft-2 + f4(futt+i - ffnd) + Et
R2 Statistics

One Month Two Months Three Months

Ahead (i = 1) Ahead (i = 2) Ahead (i = 3)

1989:5-1994:1
Specification A 0.06 0.14 0.17
Specification B 0.08 0.20 0.24
Improvementin R2 0.02 0.06 0.07
1994:2-2000:10
Specification A 0.08 0.20 0.23
Specification B 0.42 0.71 0.69
Improvementin R2 0.35 0.51 0.46

do suggest that the nonautoregressivecomponent has been an importantfactor in

the improvedpredictive power of marketinterest rates.

4. CONCLUSIONS

This paper has found fairly strikingevidence of an improved ability of financial

marketsto predict future changes in policy by the FOMC. The shift in behavior
appearedto take place over a period ranging from the late 1980s to early 1994. A
caveatto the findingsis that,becauseof ourfocus on the importanceof recentchanges,
we are unavoidably left with short sample periods. We have been careful to
check that our resultshave not been drivenby small sample bias or by the influence
of individualepisodes. Moreover,the increasein predictabilityis remarkablystrong
andis evidentin a numberof differentmarkets,giving morecredenceto the findings.
We show that a portion of the improvedpredictabilityof FOMC actions can be
associatedwith strongerserial correlationin changes in the federalfunds rate,which
could reflecteithera deliberatestrategyof a slower phasing-inof policy adjustments
or changes in the underlyingnatureof the economy. However, the results indicate
thatfactorsotherthanthe autoregressivepropertiesof the federalfundsrateappearto
have played a more importantrole in enhancingthe predictabilityof FOMCactions.
Among these other factors, the Federal Reserve has implemented a number of
changes thatmay have improvedthe transparencyof the process of settingmonetary
policy. The Federal Reserve gradually shifted away from the borrowed reserve
operating procedure that had been in place in the mid-1980s in favor of strict
targetingof the federalfunds rate.Moreover,in 1994, the FOMCbegan announcing
policy changes on the day of its meetings, and offering some explanationof those
decisions. In addition,the speeches and public comments of FOMC membersmay

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAMWHITESELL 905

now better preparethe markets for future policy moves. Finally, perhaps in part
because of these changes, marketparticipantsand academic researchershave im-
proved theirunderstandingof the systematicresponses of policy to macroeconomic
developments. Empirical evaluations of sharperdistinctions among these specific
factors is an interestingarea for additionalresearch.

APPENDIX

Bekaert,Hodrick, and Marshall(1997), hereafterBHM, reportsignificantsmall-

samplebias in severalregression-basedtests of the expectationshypothesis,including
the Campbell-Shiller specificationestimated in Table 1:

n + 1 -(Afft+i) +
= X +(in - fft) + Ft. (Al)
i=1 fl

To be consistent with BHM, here we write the equation in terms of changes in the
month-averagefederal funds rate. As BHM explain, the persistence of short-term
rates creates an upward bias in the slope coefficient of Equation (Al) in small
samples. However, the size of the bias is heavily influenced by the data-generating
process of the short-terminterest rate. In particular,negative correlationbetween
short rates and the yield spread tends to amplify the upward bias on the slope
coefficient.BHM initially considerthe case in which the short-terminterestratefol-
lows an AR(1) process. There is a sizable small-sample bias in that case because
the correlationbetween the short-terminterestrate and the yield spreadis precisely
-1 under the expectations hypothesis, as innovations to the short-termrate are
expected to begin unwindingimmediately.BHM also considera moregeneralcase in
which the path of the short-terminterest rate is generatedby a second-orderVAR
that allows for heteroskedasticity.In that case, the negative correlation between
the short rate and the 12-monthyield spreadis about -0.5, and the small-sample
coefficient bias is substantiallyreduced.
We consider a specificationthat, in terms of generality,falls between these two
cases. In particular,as in Table 4, we describe the behavior of the month-average
federal funds rate in recent years by a third-orderautoregressiveprocess in changes.
Therefore,we investigatethe biases thatresultwhen the federalfundsrateis assumed
to follow thatprocess, where the parametersof the process are estimatedseparately
for the differentsubsamples.As a check of the robustnessof our findings, we also
report results under the assumption that the federal funds rate follows an AR(1)
process. While that specificationcan be rejectedin the data, it allows us to compare
results to the worst case considered by BHM.
Note thatBHM also adjustfor small samplebiases in the estimatesof the dynamic
process of the short-terminterestrate. In our case, the adjustmentto the first-order
autoregressive process in levels would make the federal funds rate look like a
randomwalk, which is an uninterestingcase because the yield spreadshould then
never change underthe expectationshypothesis.Because the third-orderdifferenced

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
906 : MONEY, CREDIT, AND BANKING

TABLE 7
AVERAGE PCOEFFICIENT FOR EQUATION (Al) (1000 Monte Carlo Replications)

Number of Observations

1983:1-1989:3 n = 75 n = 5000
ff, = 0.185 + 0.978 fft-1 3.33 1.03
Afft = 0.439 Afft-1 + 0.030 Afft-2 - 0.048 Afft-3 0.90 1.00
1989:4-1994:1 n = 58 n = 5000
fft = 0.097 + 0.983 fft,_ 6.14 1.05
Afft = 0.372 Afft-5 + 0.200 Afft-2 + 0.060 Afft-3 0.81 1.00
1994:2-2000:10 n = 81 n = 5000
fft = 0.445 + 0.924 fft-1 1.59 1.01
Afft = 0.221 Afft_1+ 0.274 Afft-2 + 0.148 Afft-3 0.86 1.00

process that we consider implicitly nests the random walk case, we do not employ
an adjusted levels process. Small-sample biases in the differenced process are
very small, and thus no adjustmentis needed for that specification.
Table 7 reports average point estimates of P with both first-orderin levels and
third-orderin changes autoregressivegeneratingprocesses for the short-terminterest
rate based on 1000 Monte Carlo simulations. The estimates in the large sample
(n = 5000) approximatethe asymptoticvalue of the coefficient,which is 1.0. Under
the first-orderautoregressiveprocess, the small sample bias pushes the coefficients
significantlyabove thatvalue in the small sample (n = 75, 58, and 81, respectively).
However, the bias is much less severe underthe third-orderautoregressivechanges
process, and the coefficientsare in fact below unity.The reasonis thatthe correlation
between innovations to the federal funds rate and changes in the yield spread is
small and slightly positive underthe more flexible dynamic process, as opposed to
the strongnegative correlationunderthe AR(1) process. This is consistent with the
finding in BHM that the AR(1) model overstates the bias (for shorter maturities)
relative to a more general VAR specification.
The bias in the slope coefficient in small samples also affects the R2 statistic of
the regression,as seen in Table 8. However, similarto the results for the coefficient

TABLE 8
AVERAGER2 STATISTICFOR EQUATION(Al) (1000 Monte Carlo Replications)

Number of Observations

1983:1-1989:3 n = 75 n = 5000
Ift = 0.185 + 0.978 fft- 0.11 0.03
Afft = 0.439 Afft-1 + 0.030 Afft-2 - 0.048 Afft-3 0.13 0.14
1989:4-1994:1 n = 58 n = 5000
fif,= 0.097 + 0.983 ft-1 0.14 0.02
Afft= 0.372 Afft-, + 0.248 Afft-2 + 0.132 Afft-3 0.23 0.32
1994:2-2000:10 n = 81 n = 5000
ff,= 0.445 + 0.924 fft-1 0.15 0.10
Afft= 0.221 Afft_1+ 0.274 Afft-2 + 0.148 Afft-3 0.27 0.34

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOELANGE,BRIANSACK,ANDWILLIAM
WHITESELL: 907

bias reportedin the above table, the upwardbias on the R2 statistic is fairly small
when the short-terminterest rate is assumed to follow a third-orderautoregressive
process in changes. Moreover, even under the first-orderprocess, the positive bias
on the slope coefficientand the R2 statistic appearsto be smaller in the more recent
sample-opposite the patternof the R2 statistic that we find in Table 1. Under the
third-orderdifferenced process, the bias tends to pull down the R2 statistic in
the more recent subsamples,again pushingin the opposite directionof our findings.
Based on thisevidence,it appearsveryunlikelythatthe smallsamplebiasis accounting
for the improvementof the regressionfit over the sample,therebystrengtheningour
claim that short rates have become better predictorsof longer-termrates in recent
years.

NOTES

1. Longer-termyields also seem to have some informationregardingfuture short-ratemovements,

which may be associated with mean reversion, as found by Fama and Bliss (1987).
2. Some studies, including Simon (1990), have found differences in behavior after the shift in the
Federal Reserve's operatingregime from nonborrowedto borrowedreserve targetingin October 1982.
3. The data are monthly averages from January1983 to October2000.
4. The significanceof these correlationswas tested using a regressionapproachsimilarto a Granger
causality test, where changes in the federal funds rate were regressedon currentand lagged changes in
marketyields.
5. To avoid a time-aggregationproblem,as pointed out by two anonymousreferees, the federal funds
rate for period t is end-of-month, rather than month-average. Because daily values of the effective
funds rate can be noisy, we use the FOMC's target funds rate.
6. Campbell and Shiller rewrite the right-handside of Equation(2) as a weighted average of future
changes in the funds rate, but doing so makes the notation cumbersomein our case given the necessary
use of both month-averageand month-enddata.
7. We use Equation(4) to predict changes in the federal funds rate ratherthan changes in very short-
dated Treasurybill yields because of our focus on the ability of financialmarketsto anticipatemonetary
policy actions and because bills at maturitiesshorterthan three months are less liquid.
8. Although the series of policy tightenings between February1994 and February 1995 were well
anticipatedby the market,the overall improvementin fit in the post-1994 period is not driven solely
by this short episode. Startingthe latter subsampleafter June 1995 produces similar regression results.
9. Before 1994, changes in the intended funds rate were not announcedon the day of change. The
market typically learned of such decisions through implicit signaling by the choice of open market
operationthe following day. However, even before 1994, discount rate changes were announcedon the
day of change, and the marketcould then typically interpretthe implicationsfor the funds rate that day.
For these reasons, a two-day change in the currentmonth federal funds futures rate is used to measure
the surprisein the policy action before 1994.
10. The equationis only approximatelycorrect,because it does not account for the small change in
the term premium on the day of the policy action or the amount by which the effective funds rate
deviated from the target on the day of the policy action. However, these effects should be small.
11. While the magnitudeof the surprisesin policy actions has fallen in recent years, the response of
marketinterestratesto a policy surpriseof a given size appearsto have remainedunchanged,as explored
in detail by Bomfim and Reinhart(2000) and Kuttner(2001).
12. For signaling methods used by the Open Market Desk, see Feinman (1993).
13. February1994 correspondsto the beginningof our final subsample.Bomfim and Reinhart(2000)
found that the shift to policy announcementsin February1994 did not seem to increase the response
of markets to a surprisein the federal funds rate of a given size. Instead, we focus more closely on
the impactof policy announcementson the magnitudeof the policy surpriseand the timingof movements
in marketinterest rates.
14. At that time, the traditional language regarding asymmetry in the chances for an easing or
tightening of policy was replaced with language regarding the balance of risks facing the economy,
which the FOMC intended to be less closely linked to imminent policy moves.
15. Fuhrer(1996) has shown that the form of the policy rule can have importantimplicationsfor the
behavior of the term structure.

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
908 : MONEY, CREDIT, AND BANKING

16. A numberof studies have found changes in the stabilityof the economy in the period since 1980,
but it may be prematureto assess whether there have been changes in the shocks or internaldynamics
of the economy as a whole since the mid-1990s.
17. Another hypothesis is that markets may appearmore anticipatorybecause policy has become
more responsive to them. It is unclear how the data could disentangle this proposition from the idea
that policy makers now make greater attempts to preparemarkets for future policy actions. The latter
view could be interpretedas a featureof a general trendtowardincreasedtransparencyin policy making.
18. We arenot takinga strongstandon whetherthe interestrateis nonstationary.Assuming stationarity
by adding a lagged level of the rate as an additionalregressordoes not affect the results strongly.The
coefficient on the lagged level is not significant over the entire sample, but has become significantin
the most recent subsample.
19. The traditionalR2 measure indicates the proportionof the variance in the dependent variable,
relativeto its mean,thatis accountedfor by explanatoryvariablesin the regressionotherthanthe constant
term. This measure does not help assess the degree of investor awareness of trend changes in policy
during easing and tightening phases, as might be indicated by the constant term in short samples. As
shown in Table 4, the constant term became significant in the middle period, which was dominatedby
policy easings. To measure the predictabilityof funds rate changes relative to a zero-change baseline,
which would capturethe impact of the estimated constant, we examined uncenteredR2 statistics from
the regressions.They were 0.15, 0.41, and 0.26 over the threesubperiods,respectively.Thus, the centered
R2 statistics reportedin the table may be understatingthe improvementin predictability,particularlyin
the middle period.
20. We thank the editor for pointing this out.
21. The differences in uncenteredR2 statistics between Specifications A and B on Tables 5 and 6
were comparableto those reportedfor centeredR2values. Note thatthe constanttermfor each specification
B reflects a term premium as well as possible trends in policy changes over the subsample.
22. Increasedmeanreversionis evidentin the fundsratein the mostrecentsubsample,forinstance,which
is not capturedby our specification.In tests that allow for this effect, the increasein the predictivepower
was more evenly balanced across the autoregressiveand nonautoregressivecomponents.

LITERATURE CITED

Bekaert, Geert, Robert J. Hodrick, and David A. Marshall (1997). "On Biases in Tests of
the Expectations Hypothesis of the Term Structure of Interest Rates." Journal of Financial
Economics 44, 309-348.
Bomfim, Antulio N., and Vincent R. Reinhart (2000). "Making News: Financial Market
Effects of Federal Reserve Disclosure Practices." Financial and Economics Discussion
Series No. 14.
Campbell,JohnY., andRobertJ. Shiller (1991). "YieldSpreadsandInterestRate Movements:
A Bird's Eye View." Review of Economic Studies 58, 495-514.
Cook, Timothy, and Thomas Hahn (1988). "The Effect of Changes in the Federal Funds Rate
Target on Market Interest Rates in the 1970s." Journal ofMonetary Economics 24, 331-351.
Fama, Eugene F. (1984). "The Information in the Term Structure." Journal of Financial
Economics 13, 509-528.
Fama, Eugene F., and Robert R. Bliss (1987). "The Information in Long-Maturity Forward
Rates." American Economic Review 77, 680-692.
Feinman, Joshua N. (1993). "Estimating the Open Market Desk's Daily Reaction Function."
Journal of Money, Credit,and Banking 25, 231-247.
Fuhrer, Jeffrey C. (1996). "Monetary Policy Shifts and Long-Term Interest Rates." Quarterly
Journal of Economics 111, 1183-1209.
Hardouvelis,Gikas A. (1988). "The Predictive Power of the Term StructureDuring Recent
MonetaryRegimes." Journal of Finance 43, 339-356.
Hatzius, Jan (1999). "Can the Bond Market Do the Fed's Dirty Work?"Goldman Sachs
Global Economic Research (November 12, 1999), 4-6.
Kuttner,KennethN. (2001). "MonetaryPolicy Surprisesand InterestRates: Evidence from
the Fed Funds FuturesMarket."Journal of MonetaryEconomics 47, 523-544.

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions
 JOE LANGE, BRIAN SACK, AND WILLIAMWHITESELL : 909

Longstaff,FrancisA. (2000). "TheTerm Structureof Very Short-TermRates: New Evidence

for the ExpectationsHypothesis."Journal of Financial Economics 58, 397-415.
Mankiw, N. Gregory, and Jeffrey A. Miron (1986). "The Changing Behavior of the Term
Structureof InterestRates." QuarterlyJournal of Economics 101, 211-228.
Mishkin,FredericS. (1988). "TheInformationin the TermStructure:Some FurtherResults."
Journal of Applied Econometrics3, 307-314.
Rudebusch,GlennD. (1995). "FederalReserveInterestRate Targeting,RationalExpectations,
and the Term Structure."Journal of MonetaryEconomics 35, 245-274.
Shiller,RobertJ., John Y. Campbell,and KermitL. Schoenholtz (1983). "ForwardRates and
Future Policy: Interpretingthe Term Structureof Interest Rates." Brookings Papers on
Economic Activity 1, 173-224.
Simon, David P. (1990). "Expectationsand the TreasuryBill-FederalFundsRate Spreadover
Recent MonetaryPolicy Regimes." Journal of Finance 45, 567-577.
Taylor, John B. (1993). "Discretion versus Policy Rules in Practice." Carnegie-Rochester
ConferenceSeries on Public Policy 39, 195-214.
Woodford,Michael (1999). "OptimalMonetaryPolicy Inertia."ManchesterSchool Supple-
ment 67, 1-35.

This content downloaded from 128.235.251.160 on Thu, 5 Feb 2015 22:23:50 PM

All use subject to JSTOR Terms and Conditions