Tests for Parameter Instability in Regressions with I(1) Processes

Author(s): Bruce E. Hansen

Source: Journal of Business & Economic Statistics, Vol. 10, No. 3 (Jul., 1992), pp. 321-335
Published by: American Statistical Association
Stable URL: http://www.jstor.org/stable/1391545 .
Accessed: 21/06/2014 23:43

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 support@jstor.org.

American Statistical Association is collaborating with JSTOR to digitize, preserve and extend access to Journal
of Business &Economic Statistics.

http://www.jstor.org

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
? 1992 AmericanStatisticalAssociation Journalof Business & EconomicStatistics,July 1992, Vol. 10, No. 3

Tests for Parameter Instability in

Regressions with 1(1) Processes
Bruce E. Hansen
Departmentof Economics,University
of Rochester,Rochester,NY 14627

Thisarticlederivesthe large-sampledistributions of Lagrangemultiplier(LM)tests forparameter

instabilityagainstseveralalternativesof interestinthe contextof cointegratedregressionmodels.
The fullymodifiedestimatorof Phillipsand Hansen is extended to cover general models with
stochastic and deterministictrends. The test statistics considered include the SupF test of
Quandt,as well as the LMtests of Nyblomand of Nabeya and Tanaka. It is found that the
asymptoticdistributionsdepend on the natureof the regressor processes-that is, if the re-
gressors are stochastic or deterministictrends. The distributionsare noticeablydifferentfrom
the distributions
whenthe dataare weaklydependent.Itis also foundthatthe lackof cointegration
is a special case of the alternativehypothesisconsidered(an unstableintercept),so the tests
proposedhere may also be viewed as a test of the nullof cointegrationagainstthe alternative
of no cointegration.The tests are appliedto three data sets-an aggregate consumptionfunc-
tion, a present value model of stock prices and dividends,and the term structureof interest
rates.

KEYWORDS: Cointegration;
Fullymodifiedestimation;Quandtstatistic;Parameterconstancy;
Structuralchange.

One potential problem with time series regression
models is that the estimated parametersmay change
over time. A formof model misspecification,parameter
nonconstancy,may have severe consequences on in-
ference if left undetected. In consequence, many ap-
plied econometriciansroutinelyapply tests for param-
eter change. The most commontest is the sample split
or Chow test (Chow 1960). This test is simpleto apply,
and the distributiontheory is well developed. The test
is crippled,however, by the need to specifya priorithe
timing of the (one-time) structuralchange that occurs
under the alternative.It is hard to see how any non-
arbitrarychoice can be madeindependentlyof the data.
In practice, the selection of the breakpointis chosen
either with historicalevents in mindor aftertime series
plots have been examined.This impliesthat the break-
point is selected conditionalon the data and therefore
conventionalcriticalvalues are invalid. One can only
conclude that inferencesmay be misleading.
An alternative testing procedure was proposed by
Quandt(1960),who suggestedspecifyingthe alternative
hypothesisas a single structuralbreakof unknowntim-
ing. The difficultywith Quandt'stest is that the distri-
butional theory was unknown until quite recently. A
distributional theoryfor this test statisticvalidfor weakly
dependent regressorswas presented independentlyby
Andrews (1990), Chu (1989), and Hansen (1990). Chu
consideredas well the case of a simplelineartime trend.
Another testing approachhas developed in the sta-
tistics literaturethat specifiesthe coefficientsunderthe
alternativehypothesisas randomwalks. Recent expo-
321
sitions were given by Nabeya and Tanaka (1988), Ny-
blom (1989), and Hansen (1990).
The precedingworks did not consider models with
integrated regressors.This article makes such an ex-
tension. The test statistics mentioned previously are
examined here in the context of cointegratingregres-
sions,makinguse of the fullymodifiedestimationmethod
of Phillips and Hansen (1990). The asymptoticdistri-
butionsof the test statisticsare found to dependon the
stochasticprocessdescribingthe regressors.It emerges
as an importantconclusionthat it is necessaryto know
the stochasticprocessof the regressorsbefore one can
apply the tests consideredhere.
An additionalfinding is that, since the alternative
hypothesisof a randomwalk in the interceptis identi-
cal to no cointegration, the test statistics are tests of
the null of cointegrationagainst the alternativeof no
cointegration.
A relatedresearcheffortby ZivotandAndrews(1992)
and Banerjee, Lumsdaine,and Stock (1992) developed
a distributionaltheory for the test of the unit-roothy-
pothesis employed by Perron (1989). Perron specified
the alternativeto be a single structuralbreakof known
timing,but the aforementionedarticlesspecifythe time
of the break as unknown.These articlesaddressa dif-
ferent question (testing the unit-root hypothesis), al-
though using similarmethods.
Section 1 sets up the structureof the model, allowing
for quite generalstochasticand deterministictrendsin
the regressors.This model builds on and extends the
setup used by Phillipsand Hansen (1990) and Hansen

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
322 Journalof Business & EconomicStatistics,July 1992

(1992a). Section 2 describesthe method of fully mod- hand,if a time trendis requiredin the levels regression,
ified estimationfor the model. Section 3 describesthe then x1, = k1, = k, = (1, t)' and there is no k2t. Another
test statisticsfor parameterinstability.Section 4 gives commonspecificationis that there are no trendsin the
the asymptoticdistributiontheory for the test statistics system, so k1t = kt = 1 and there is no k2t.
under the null hypothesis. Section 5 discussesthe al- Some applied researchers have considered using
ternative of no cointegration.Section 6 reports three breakingtrendfunctionsin addition to the simple in-
simple applicationsof the tests. Section 7 concludes. tegerpowersof time consideredhere. Althoughin prin-
Proofs are in the Appendix. ciple it is straightforwardto define the estimatorsand
A GAUSS programthat implementsthe estimation test statisticswith these more general trend functions,
methods and test statisticsis availableon requestfrom this extension will not be consideredin this articlefor
the author. two reasons. First, the restrictionto powers of time
simplifiesthe asymptotictheory. Standardizedpowers
1. A COINTEGRATED
REGRESSIONMODEL of time converge uniformlyto limiting functions, but
We will be workingwith a fairly standardmultiple- this is not true of discontinuoustrend functions. See
regressionmodel under cointegration.The cointegrat- Zivot and Andrews(1992) for an econometricexample
ing equationsare of weak convergencewith discontinuousfunctions.Sec-
ond, breakingtrends may only make sense in a prob-
Yt = Ax, + ul,, t= 1, . . ., n,
abilitymodel if the timingand magnitudeof the break
where the process x, = (xt, x2t)' is determined by the is allowed to be random.This simplyreintroducesI(1)
equations or 1(2) components into the system that are already
capturedin system (1)-(2).
Xlt = klt The followingnuisanceparametersplay an important
H klt + Hk2t + X2t
role in the formulationof the statisticswe will be con-
x2t=
1 2 sidering.Define the matrices
n n
X2t = + (2) 1
Xt-1 U2t
l = lim - E E(uju')
Define the vectors n---,c t=l j=1

1 n t
u = ( ut U2 ); k = (kt k2t). A = lim- E E E(uju) (3)
mi m2 + P2 Pl P2 n->oc n t=l 1j=

{u} is a sequenceof mean zero randomvectors, but the partitionedin conformitywith u:

elements of kt are nonnegativeintegerpowersof time.
Note that x1, has p, elements and x2thas m2 + P2 ele- =
n12' A = All A12
ments. More specific assumptionsregardingu, and kt (21 A21 A22/
will be made in Section3. This model places the trends
When the vector ut is weakly stationary,fl is propor-
k1, directlyin the regressionequation (I will speak of tional to the spectraldensity matrixevaluated at fre-
kt as trends althoughit may contain a constant). The
trends k2tdeterminethe behaviorof the stochasticre- quency 0. It is sometimes referredto as the long-run
covariancematrix.
gressorsx2tbut are excludedfrom the regression.If kt We also define
containsa constant,we assumethat it is an element of
kit and thus enters the levels regression. n1-2 = 11, - n12,,22 21 (4)
The notation of (2) may appearconfusingat first. It
and
turns out to be especiallyconvenientfor development
of a full theory for the cointegratedregressionmodel A2+ = A21 - A2222ll'21
(1). In earlier work on cointegratedmodels, such as One may loosely call l.2 the long-runvarianceof u1,,
that of Johansen(1988), Park and Phillips(1989), and
conditional on u2,. AI represents the bias due to en-
Phillips and Hansen (1990), allowing for general de-
terministictrends was intentionallyexcluded to allow dogeneityof the regressorsafterthe fully modifiedcor-
rectiondiscussedin Section 2.
the clean developmentof a large-sampledistributional
theory. Although theoreticallyelegant, this approach ESTIMATION
2. FULLYMODIFIED
excludes many applicationsof interest.
In most applications,it seems most reasonable to The constancytests we will discuss require an esti-
specify the stochasticregressorsx2, as I(1) with a de- mate of A in (1) that has a mixturenormalasymptotic
terministictrend. In this case, kt equals a constantand distribution.Forconcreteness,we will considerthe fully
a linear time trend. If y, and x2t are deterministically modified(FM) estimatorof PhillipsandHansen(1990).
cointegrated,using the terminologyof Ogaki and Park Alternative estimatorswith the same asymptoticdis-
(1990), then the levels regressionneed only contain a tribution include the maximum likelihood estimator
constant, so x,t = kit = 1 and k2t = t. On the other (MLE) of Johansen(1988) or the "leads and lags" es-

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
 Hansen:TestforInstability
with1(1)Processes 323

timator of Saikkonen (1991) and Stock and Watson In most applications,it appears that the choice of
(1991). (See PhillipsandLoretan[1991]for an informa- kernel is much less importantthan the choice of band-
tive review.) width parameter.For currentconsistencyproofs (see
Hansen 1992b), it is required that M - ooat some rate
2.1 Estimation of Covariance Parameters slower than sample size. This, however, does not pro-
The semiparametricmethod of Phillipsand Hansen vide a useful guide for the selection of the bandwidth
(1990) is a two-step estimator in which the first step parameterin a particularapplication.In a recentarticle,
estimates the covarianceparametersQ1.2 and A1I de- Andrews(1991)providedsome usefulguidelines(based
fined in (4). Our suggestion is to use a prewhitened on the minimizationof asymptotic truncated mean
kernel estimator with the plug-in bandwidthrecom- squarederror) to their selection. He recommendeda
mended by Andrewsand Monahan(in press). We out- plug-inbandwidthestimator.For the Bartlett, Parzen,
line the procedurein this subsection. and QS kernels, the choices are
First, estimate (1) by ordinaryleast squares (OLS), Bartlett: M = 1.1147(c()n)l/3
yielding the parameterestimates A and the residuals
u1, = Yt - Axt. Second, estimate (2) by OLS in differ- Parzen: M = 2.6614(a(2)n)l/5
ences: Ax2, = lAk1,t + I2Ak2t + u2t, yielding the re- QS: M- = 1.3221(a(2)n)15,
siduals u2t.Set u = (uAl,u2t).
The covariancematricesfl and A could be estimated where a(1) and a(2) are obtainedfrom approximating
directlyfrom the residualsut via a kernel. In most ap- parametricmodels. A particularlyattractivechoice sug-
plications,the cointegratingresidualsu1,have a signif- gestedby Andrewsis univariate(autoregressive)AR(1)
icant degree of serial correlation. In this event, the models for each element, eat (a = 1, . . ., p) of e,. Let
kernel estimate will be highly biased, unless a large (pa,6-2) denote the autoregressive and innovation vari-
bandwidthparameteris used, whichincreasesthe vari- ance estimatesfor the ath element. Then
ance of the estimator.In suchcases, an estimatorbased P A P
W2 Ca
on prewhiteningis often preferablein moderatesample i(5(1)
=.4PaOa
--
tawi a1 (1 - a)6( + Pa)2 1 (1 - Pa)
sizes.We suggestusinga (vectorautoregressive) VAR(1),
althougha higher order VAR could also be used. We and
first fit a VAR to the residualsut: ut = ia,_- + et. A p 2 2
/
kernel estimatoris then appliedto the whitenedresid- a^ Pa4a
(2)
uals et. These take the form () = (1 a- p)8 a ( - Pa)
n
I
The use of a plug-inbandwidthparameterhas several
- e t
e tjt
w(jlM) E
n t=j+l
advantages.First, it removes the arbitrarinessassoci-
j=o
ated with the choice of bandwidth.Many applied re-
and searchershave been frustratedwith the semiparametric
n n branchof the unit-rootliteraturebecause the test out-
e = fie
~ w(jlM)-
w(j/M) et_jet , comes sometimesdepend on the choice of bandwidth.
j=-n n t=j+l e et'
Second, simulationresults of Park and Ogaki (1991)
demonstratedthat its use can dramaticallyimprovethe
where w(.) is a weight function (or kernel) that yields
mean squared error of semiparametricestimates of
positive semi-definiteestimates and M is a bandwidth
cointegratingrelationships.
parameter.The estimatorQe can be seen as a scaled
estimate of the spectraldensityof e, (when e, is covari-
ance stationary)and has its origin in the literatureon 2.2 Estimation of the Regression Parameters
spectraldensityestimation,whichdates backto Parzen
Partition A and ft as A and fi, set il1.2 =l -
(1957).
The covarianceparameterestimates of interest can i12i2'f1i21 and A2 = A21 - fikP2lh21 and define
be obtained by recoloring: Q = (I - <)-l'e(I - the transformed dependent variable y+ = y -
)- 1 and A = (I - )-Ae(I - <)-1 - (I
- li"2j2l12t. The FM estimatorof A is then given by
A)--1j , where t = (1/n) Z=1it^;.
These estimatesrequirea choice of kernel and band- A+ y;- .
(0 A1i'))) (xt;)
t=1
width parameter.Any kernel that yields positive semi- t=l
definiteestimatescanbe used.Theseincludethe Bartlett, Associated with these parameterestimates are the
Parzen, and quadraticspectral(QS) kernels. Andrews residualsau = y+ - A +x,. One interestingfeatureof
(1991) recommendedthe QS kernel, which takes the the FM estimates that will be importantfor our later
form
developmentsis that
n
1
w(x) = 122(sin(6(rx/5) - cos(6rx/5)) x
w()=127T2X26=rx5 n t=i w21

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
324 Journalof Business&Economic
Statistics,July1992

althoughin OLS regressionthe sum of the productsof martingaledifferencesis 0: Ho: 62 = 0. One alternative
the regressorsand residualsis identically0. Thus the hypothesis is H3 : 82 > 0, G, = (fi.2 0 V,,)-l; tln E
scores of the problemare the variables 9C,with test statistic
1
S, = (XtU -
(5)
MeanF =
nY tlne
F, where n* = >
tln 3
1.
(21))
which satisfy Sn=1 , = 0.
The final alternativeis H4: 62 > 0, Gt = (f1.2 ?
Mnn)-1, with test statistic
3. TESTS FOR PARAMETER INSTABILITY
Hansen (1990) outlined a general theory of testing Lc = tr{Mnn1 E St? S)}.
for parameterinstabilityin econometricmodels. The
test statisticscan be derivedas Lagrangemultiplier(LM)
The F,, test (fixed t) is computationallysimple, cor-
tests in correctlyspecifiedlikelihoodproblems.In this
section, we describethese test statisticsin the context respondingto the classical Chow test or sample split
test. The test statisticis computationallyequivalentto
of fully modifiedestimationof cointegratedregression
models. estimatingA1 andA2 on the two subsamplesandtesting
their equivalenceusing a Wald test, usingthe variance
We can modify(1) to incorporatepossibleparameter
estimatefor the full-sampleestimates.Thiscanbe easily
instabilityby allowingA to depend on time: seen if we consider the special case of least squares
Yt = A,xt + u,. (1)' estimation on a single equation (ml = 1). Then note
that
For all of the tests, the null hypothesisis that the coef-
ficientA, in (1)' is constant,althoughthe tests differin
the treatmentof alternativehypotheses. Mnt Snt = (E XiXi Xii

The first two tests model At is obeyinga single struc-

tural break at time t, where 1 < t < n: = ( xx) I - (- xix ) xA = A, - A;

Ai = A1, i < t
that is, the score from the first part of the sample,
= A2, i>t. evaluatedat the estimate from the full sample, is pro-
The null hypothesisis Ho: A1 = A2. For the first test, portionalto the difference between the estimates ob-
the timing of the structuralbreak is known under the tainedfromjust the firstpartof the sampleand the full
alternativeH : A1 + A2, t known.A test for Hoagainst sample.It follows(witha little algebra)thatourstatistic
H1 is given by the statistic Fntis essentiallyequivalent to the Wald statistic that
tests the equivalenceof At and A. The only difference
Fn = vec(Snt)' (Q1.20 Vnt)-vec(Snt ) arises due to the choice of the varianceestimates.It is
= tr{SntVntl,SntL1},
well knownthatthisWaldstatisticis algebraicallyequiv-
alent to the classic Chow statistic, which is based on
where the differencebetweenthe estimatesobtainedfromthe
t two subsamples.For example,see SnowandIm (1991).
Sn = S i, (6) The distributionaltheory developed for this test
i=1
(asymptoticchi-squared)is only valid when t can be
chosenindependentlyof the sample.Thisis a restrictive
Vnt = Mnt- MntMnnl'Mn, (7)
assumptionin practiceand may be valid only when t is
and chosen in an arbitraryway, such as t = n/2. In this
t event, the test mighthave low power againstmany al-
Mnt = xiX'. (8) ternativesof interest.
i=l The SupF test dates back to Quandt(1960). Several
For the secondtest, the timingof the structuralbreak recentworkshave exploredthe distributionaltheoryin
is treatedas unknown:H2 : A1 4 A2, [tin] E -, where severalcontexts-those of Andrews(1990),Chu (1989),
5- is some compact subset of (0, 1), and [.] denotes and Hansen (1990). The only difficultyin implemen-
tation is the choice of the region S-. As pointedout by
"integerpart." This test statisticis simply
Anderson and Darling (1952) and emphasizedby An-
SupF = sup Fnt. drews (1990), the region 2- must not include the end-
tln E-
points 0 and 1; otherwisethe test statisticwill diverge
The thirdand fourthtests model the parameterA, as to infinityalmostsurely.The fix suggestedby Andrews
a martingale process: A, = A,_1 + E,; E(E,t-- 1) = 0, is to select T-= [.15, .85]. Although a reasonableap-
E(E,E') = 82G,. In this context, the null hypothesis can proach,this introducesan element of arbitrarinessthat
be written as the constraintthat the variance of the dilutes the appeal of the test.

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
 Hansen:TestforInstability
with1(1)Processes 325

The MeanF test statistic is derived from a different 7. k, = (1, t, t2, . . ., tP)', p = pi + P2 - 1; and
hypothesis structure but is seen to be simply the average 8. M51n = 0(1).
Fnt test. Although in principle the averaging can include
For the random sequence {ut, the assumptions im-
all values of r for which Fn,can be computed, in practice
some trimming will be required (since F,, will not be pose weak dependence through fairly mild conditions
on the strong mixing coefficients. The moment condi-
defined over all t). Thus the arbitrariness associated
tions are only slightly stronger than square integrability.
with the SupF test is not completely avoided. For the
- Assumption 5 implies that the elements of xt are not
remainder of the article, we set = [.15, .85] as for
the SupF test. mutually cointegrated, that x2tdoes not contain a lagged
The Lc test has a long history in statistical theory, dependent variable, and that xt and the error u1, are
not multicointegrated (see Granger and Lee 1990). As-
although it has not been fully understood until quite
sumption 6 says that the vector k2 plays a role in the
recently. It was first proposed by Gardner (1969) as a
asymptotic behavior of x2t. Assumption 7 restricts the
Bayes test for structural change. It was later indepen- vector kt to integer powers of time.
dently proposed by Pagan and Tanaka (1981), Nyblom Set Yt = jt=l uj. Our assumptions are sufficient for
and Makelainen (1983), and King (1987). These works
the following results:
all concerned tests on a single coefficient in a Gaussian
linear regression model. First attempts at a large-sample
=
distributional theory were made by Nyblom and Make- (1/Vn)Y,,[nr] > B(r) BM(Q); (9)
lainen (1983), Nabeya and Tanaka (1988), and Ley-
[nr] rf
bourne and McCabe (1989). A fairly complete theory
(l/n) YVtu+l > J0 Bdb' + rA; (10)
for maximum likelihood was given by Nyblom (1989) t=l
and was extended to general econometric estimators by
and
Hansen (1990). It has the advantage that it is much
easier to compute than the SupF and MeanF tests and A --- 1A'
, --> n. (11)
requires no decisions for trimming, hence excluding any Here and elsewhere, ">" denotes weak convergence
form of arbitrariness.
of the associated probability measures with respect to
The three proposed tests-SupF, MeanF, and Lc-
the uniform metric, and BM(fl) denotes a Brownian
are all tests of the same null hypothesis but differ in
motion with covariance matrix f. The invariance prin-
their choice of alternative hypothesis. In practice, all
of the tests will tend to have power in similar directions, ciple (9) was shown by Herrndorf (1984). Convergence
to the matrix stochastic integral (10) was shown by Han-
so the choice may be made on the computationalgrounds
sen (1992c). Consistent covariance parameter estima-
that Lc is much easier to calculate. But the appropriate
tion (11) was shown by Hansen (1992b).
test statistic for a particular application should also de-
We need to find a sequence of weight matrices that
pend on the purpose of the test. If the desire is to will appropriately standardize the regressors x, and the
discover whether there was a swift shift in regime, then
estimates A +. We adopt a method from Hansen (1992a).
the SupF test is appropriate. On the other hand, if one
Set n = diag(l, n- , n-2, .. n-P) and k(r) = (1,
is simply interested in testing whether or not the spec-
r, r2, . . ., rP)'. Thus
ified model is a good model that captures a stable re-
lationship, the notion of martingale parameters is more 8nk[nr]
-> k(r) as n - oc (12)
appropriate, since it captures the notion of an unstable
model that gradually shifts over time. If the likelihood uniformly in r. Partition 8, = diag(8n,, 82n) and k(r) =
of parameter variation is relatively constant throughout (k1(r), k2(r))' in conformity with k.
the sample, then the Lc statisticis the appropriatechoice. Equation (2) specifies that the stochastic regressors
x2t are driven by the processes k1,, k2t, and xt. Since k1,
is also in the levels regression, least squares will project
4. DISTRIBUTIONALTHEORY x2, orthogonal to k1,, leaving only k2tand x4. We would
like to isolate the effects of the stochastic trends from
The assumptions we require for the asymptotic dis-
the deterministic trends. Construct an (m2 + P2) x m2
tribution theory are summarized in the following. Let
matrix HI in the null space of I2. The matrix II will
{a,} denote the a-mixing (strong-mixing) coefficients
for {u,}. then annihilate the remaining deterministic component
k2t from x2t. Now define the weight matrix for x2t,
Assumptions. For some q > 3 > 5/2,
1. E(u,) = 0; =2n
2n
( 2J2(n
2 -1 1tf2
)112I)-1 2
\
1/2 )
\(1l/V)(n2*'22n2I)
2. am are of size -ql3/[2(q - /3)];
3. sup,-l EUtlq < o0; and the weight matrix for xt,
4. H as defined in (3) exists with finite elements;
5. H22 > 0 and H1.2 > 0; 8ln 2nO
r = -(
6. rank(I2) = P2; Fn F2nl r2J .

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
326 Journalof Business & EconomicStatistics,July 1992

To see that this is a good choice for weightingmatrix, We are now in a position to analyze the asymptotic
note that distributionof these processes.
r =
-= 8nklt
+ X2t)
Theorem 2.
rnxt = Vr2(I2k2t

/ 1nklt \
rrnsn(T)SS*(T)11.2,
(a) ~nFrSn(r)
(a) 1r)2
32nk2t + 32n(I2II2) II2x2 1, (13)
-
(1I/v i)(rl 'f1
k(1/x/~)(II2 2
f22II2) 11/2fI* X
12 X2t] (b) Fn(r) > F(T)on rE
T ;
so by (9) and (12), where S*(r) = S(T) - M(r)M(1)-1S(1), S(T) =
/kl(r) fO5XdW,F(r) = tr(S*(T)'V(T)-1S*(r)),and W1 =
= k2(r) = X(r), say, (14)
B1/2B. BM(Iml), independent of X.
rnX[nr]
W2(r) The process S*(T) is a tied-downversionof the pro-
cess S(r), whichis a continuoustime martingale.Con-
where W2(r) = (11'122112)-/2I* 'B(r) - BM(Im2). ditional on 9; = 0 - r - 1), the sigma field
-r(X(r):
SincefoXX' > 0 for all T> 0 (PhillipsandHansen1990,
generatedby the processX(r), both are Gaussianpro-
lemma A.2), F, is an appropriateweightingmatrixfor cesses. Theirconditionalcovariancefunctionsare given,
the processx,. Equation (14) says that x, is asymptoti- for r1 < T2:
cally dominatedby the trend processeskl(r) and k2(r)
and an m2-dimensionalstochastictrend (W2). E(vec(S(Tr)vec(S(T2))'j!~) = Im2 M(r1)
The test statisticsof Section3 are functionsof partial = Im20 (M(rT)
E(vec(S*(rO))vec(S*(T2))'1;)
samplesums.It willbe convenientto expressthese sums
as functionsof the space [0, 1]. Specifically,define - M(r)M(1)-'M(T2)).
[ni]
The distributionaltheoryfor the limitprocessof S('r)
Mn(r) = Mnln = Z XXi is analogousto the distributionaltheory that arises in
modelswithouttrends;for example,see Nyblom(1989)
and V"(r) = Vn[n = Mn(r) - Mn(r)Mn(1)-lMn(r). or Hansen (1990). In models withouttrends, we find
We can now findthe function-spacedistributionallimits
of these randomfunctionals.
n Sn(T) -= B*(r) = B(r) - (M)M-1B(l)
Theorem1.
= B(T) - TB(1),
(a) lrnMn()r)' M(T)= XX';
a Brownianbridge. In this expression,M > 0 is a con-
stant matrix. The difference between this result and
(b) - rnVn(T)rn> V(T)= M(T) - M(T)M(1)-'M(T); Theorem 2 arises because, in models without trends,
samplecovariancematricesconvergeto constantmatri-
ces. In modelswithstochastictrends,samplecovariance
(c) irnx tu'j XdBl +r A2) matricesare randomvariablesthat change over time.
where ui = ul - ii1u2,;
Thus the expressionfor S*(r) depends on the matrix
processM(r), representingthe samplecovariancestruc-
ture of the regressors.
(d) Vnr 21P (A
.( We can now give expressionsfor the asymptoticdis-
tributionof the test statisticsfrom Section 3.
+- A)rI
(e) Vn-(A dB.2X'' xx , Theorem3.
where B1.2 BM(fi1.2) is independent of X(r), and (a) F-*d i, b = (1 + p + m2)mi;
A21 = (I'22n) -21/2I'2 A21-
(b) supF -d su F(T);
All of the test statisticsconsideredin Section 3 were
functionsof the stochasticprocessSnt.Threeof the tests
were also functionsof the process Fnt.It will be con- (c) MeanF-d f F(r)dr;
venient to write these also as functionals on [0, 1].
Define
Inr] (d) Lc -->
Jd tr(S*(T)'M()- lS*(Tr)).
= =
Sn(T) Sn[n] E Si
i=1 The asymptoticdistributionof the standardF test is
chi-squared.This test, however, as suggested earlier,
Fn(r) = Fn[nr.
has limitedapplicabilitydue to the restrictivenatureof
= vec(Sn(T)) '(V11.2 ) Vgn(T))
-vec(S(T)). the alternativehypothesisinvolved. The other test sta-

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
 Hansen:Test for Instabilitywith 1(1)Processes 327

tisticsare nonstandardand dependon the natureof the
trends in X (i.e., p, ml, and m2).
Theorem 3 shows that it is importantto know the
trend propertiesof the regressorsbefore a parameter
constancy test can be mounted. Since the asymptotic
distributionsonly depend on a few parameters,appro-
priate critical values can be tabulated. If X contains
only deterministictrends, then analytic methods are
available. Nabeya and Tanaka (1988) derived the
asymptoticdistributionof Lc for x, = k, (in our nota-
tion) and iid errors u1. In this case, they have found
expressionsfor the characteristicfunctionof the limiting
distribution.Theirmethoddoes not immediatelyextend
to stochastic trends, so here I resort to simulation
methods.
The asymptotic distributionsare approximatedby
draws from samples of size 1,000 using iid normal
pseudorandomnumbers. The calculationswere made
in GAUSS386 using its random-numbergenerator.
Criticalvaluesfor the three tests are tabulatedin Tables
1, 2, and 3 for the single equation setting (ml = 1).
The tables includep = 0, 1, 2 and m2 = 0, 1, 2, 3, 4.
For m2 = 0, 1, 2, 25,000 replicationswere made. For
m2 = 3, 4, 10,000replicationswere made. The critical
values are noticeablydifferentfrom those for the case
of weakly dependent data. (For the supF statistic,see
Andrews [1990, table 1]; for the MeanF and Lc statis-
tics, see Hansen [1990, table 1].
The criticalvalues of Table 1 are useful but require
applied researchersto frequentlylook up tables when
making computations.It is more convenient to have
computer packages produce p values along with test
statistics.What we want is a functionp value = p(x),
which maps an observedtest statisticx into the appro-
priatevalue in the range[0, 1]. Thisis of specialinterest
when the p value falls into the range [0, .2]. Suppose
that we can well approximatethe functionp(x) by a
low-order polynomial: p(x) - aO+ alx + a2x2 + a3x3.
Then if we can obtain the parametersa = (ao, a1, a2,
a3), we can easily generate approximateasymptoticp
values automaticallyin the course of calculatingthe
statistic.
To calculate the parameters,I evaluated 38 upper
percentilesfrom the Monte Carlo distributions,from
.20 to .015 in steps of .005. Then I regressedthe per-
centiles on a third-orderpolynomial in the associate
criticalvalues. For all cases, the fit was very good over
this region. Experimentationwith extendingup to the
.010 percentile indicateda worsened fit, so it was not
done. The estimatedparametersare reportedin Table
1. On their own, they are not interesting. But when
incorporatedinto a computerprogram,they reducethe
need to use tables. The polynomials should only be
viewed as approximationsthat can produce p values
over the region[.20, .015].Thisis not a majorhandicap,
since a p value below .20 is rarelytermed"significant."
The distributionaltheory of this section is asymp-
totic. An investigationof the behaviorof the test sta-
tisticsin finite sampleswas undertakenby Gregoryand
Nason (1991). These authorsassessed the testing pro-
ceduresdescribedin this articleby applyingthe tests in
the context of a linear-quadraticmodel. Their Monte
Carlodesigninvolvedsamplesizes of 100,200, and500.
They found that the tests exhibitedvery little size dis-
tortionin these samples.They also found that the tests
had good power againstsimple structuralbreaksat the
first,second,andthirdquarterof the sample.The power
of the tests depended on a cost-of-adjustmentparam-
eter, which induces serial correlationinto the cointe-
gratingerror. As the degree of serialcorrelationin the
errorincreases,the powerdecreases.Thisis not entirely
surprising,because a highly seriallycorrelatederroris
close to a randomwalk, whichis equivalentto a random
walk in the intercept.The abilityof the test to discrim-
inate between these two cases breaks down, and the
power falls. Overall, Gregoryand Nason's study casts
a favorable light on the finite-sampleperformanceof
the test statisticsadvocatedhere.

Table 1. AsymptoticCriticalValuesfor SupF

Significancelevel p-value coefficients

m2 p 1% 5% 10% ao a, a2 a3
0 1 16.4 12.9 11.2 1.954 - .373 .0245 -.00055
0 2 20.0 15.8 14.1 2.487 -.400 .0219 - .00041
1 0 16.2 12.4 10.6 1.960 -.350 .0213 -.00044
1 1 19.0 15.2 13.4 2.666 -.424 .0230 -.00043
1 2 22.0 17.8 15.9 3.480 -.505 .0250 -.00042
2 0 18.6 14.8 13.0 3.182 - .491 .0258 - .00046
2 1 21.4 17.3 15.3 3.652 -.511 .0243 -.00039
2 2 23.9 19.7 17.7 4.003 - .508 .0219 - .00032
3 0 21.0 17.2 15.3 2.882 -.403 .0193 -.00031
3 1 23.9 19.3 17.3 3.248 - .397 .0163 - .00023
3 2 26.0 21.4 19.4 4.488 -.523 .0206 -.00027
4 0 23.6 19.0 17.1 3.522 - .449 .0194 - .00028
4 1 25.2 21.2 19.1 4.030 -.472 .0187 -.00025
4 2 28.0 23.2 21.0 5.341 -.594 .0224 -.00029
NOTE: These tablesareforthe singleequationmodel(ml = 1). Criticalvaluesare calculatedby MonteCarlosimulationusingsamples
of size 1,000. 25,000 replicationswere made form2 = 0, 1, 2. 10,000 replicationswere made form2 = 3, 4.

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
328 Journalof Business & EconomicStatistics,July 1992

Table2. AsymptoticCriticalValuesfor MeanF

Significancelevel p-value coefficients

m2 p 1% 5% 10% ao a, a2 a3
0 1 6.83 4.48 3.73 1.080 -.511 .0843 -.00478
0 2 8.85 6.22 5.11 1.595 -.613 .0818 -.00374
1 0 6.78 4.57 3.73 1.008 -.470 .0773 -.00438
1 1 8.61 6.22 5.20 1.386 -.501 .0629 -.00271
1 2 10.4 7.76 6.50 1.641 -.479 .0479 -.00163
2 0 8.50 6.17 5.18 1.477 -.557 .0729 -.00326
2 1 10.3 7.69 6.58 1.818 - .556 .0607 - .00223
2 2 11.9 9.12 7.88 2.121 -.550 .0489 -.00148
3 0 10.1 7.68 6.66 1.448 - .397 .0370 -.00117
3 1 12.0 9.21 7.89 2.22 - .580 .0520 - .00159
3 2 13.4 10.4 9.15 2.640 -.609 .0480 -.00128
4 0 11.7 9.08 7.87 2.162 -.563 .0505 -.00154
4 1 13.3 10.6 9.28 2.440 -.551 .0426 -.00113
4 2 15.0 11.9 10.4 3.287 -.702 .0512 -.00127
NOTE: See Note to Table 1.

5. A TEST OF COINTEGRATION
AGAINST
NO COINTEGRATION
Many applied econometriciansbelieve that it is im-
portant that an econometricmodel be able to survive
statisticaltests of the assumptionsunderlyingthatmodel.
In the case of estimatinga cointegratingrelationship,
a naturalhypothesisto test is thatof cointegrationitself.
In contrast, most cointegrationtests, such as those of
Engle and Granger(1987), Stock and Watson (1988),
Johansen(1988), and Phillipsand Ouliaris(1990), take
the null to be no cointegration.The one notable ex-
ception is the spuriousregressortest of Park, Ouliaris,
and Choi (1988).
The specificationtests developed in Section 3 are
clearlytests of the model of cointegrationproposedby
Granger(1981) and developed by Engle and Granger
(1987). It is, of course, possible to generalizethe def-
inition of cointegrationto allow a nonstationarylinear
relationshipbetween the variables,but this would be a
radicaldeparturefrom the idea Grangeroriginallyput
forward. But does the model of no cointegration, con-
ventionallydefined, fall into the set of alternativescon-
sidered by the specificationtests consideredhere?
For simplicity,rewriteModel (1) as
Yt = A1 + A2X2t + ut; (15)
that is, assume that k1t is simply a constant. Assume
that y, and x2, are not cointegrated.This is equivalent
to the statementthat the error u, is 1(1). Now we can
decompose ut as ut = W, + v,, where W, is a random
walk (AWtis white noise) and vt is stationary.Thus
Equation(15) can be writtenas
Yt = Alt + A2x2t + vt, (16)
where A1t = A1 + Wt. Equation (16) is a special case
of (1)', which is our model of cointegrationwith non-
stationarycoefficients.Specifically,we can see that no
cointegrationis equivalent to one coefficient, the in-
tercept, followinga randomwalk. This is a specialcase
of the alternativehypothesisfor which the Lc statistic
is an LM test statistic.We concludethat Lc is a test of
the null of cointegrationagainst the alternativeof no
cointegration.

Table3. AsymptoticCriticalValuesfor Lc

Significancelevel p-value coefficients

m2 p 1% 5% 10% ao a, a2 a3
0 1 .723 .468 .361 .927 -3.536 4.771 -2.225
0 2 .758 .480 .382 1.120 -3.644 4.155 -1.628
1 0 .898 .575 .450 .769 -3.432 5.471 -3.041
1 1 .959 .623 .497 .996 -3.493 4.311 -1.834
1 2 .999 .654 .520 1.171 -3.421 3.507 -1.240
2 0 1.03 .690 .556 .855 -3.829 6.085 -3.342
2 1 1.13 .778 .625 1.074 -3.658 4.358 - 1.778
2 2 1.19 .814 .666 1.263 -3.511 3.404 -1.133
3 0 1.18 .834 .680 1.247 -3.393 3.235 -1.066
3 1 1.29 .901 .752 1.430 -3.623 3.185 -.959
3 2 1.33 .954 .793 1.496 -3.636 3.075 -.894
4 0 1.31 .934 .780 1.451 -3.515 2.942 -.841
4 1 1.45 1.03 .866 1.694 -3.835 2.992 -.795
4 2 1.51 1.10 .922 1.726 -3.729 2.792 -.716
NOTE: See Note to Table 1.

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions

The SupF and MeanF statistics are not specifically
targetedfor the alternativeof random-walkcoefficients,
but they will have asymptoticpower againstthis alter-
native as well.
Interestingly,there is a connectionbetweenthe SupF
test, the MeanFtest, and the spuriousregressortest of
Park et al. (1988). For each t, F,, is the F test for the
significanceof sample-splitslope dummies.In the ter-
minology of Park et al. (1988), these are "spurious
trends."By the same argumentused by these authors,
each F,, statisticis consistentagainstthe alternativeof
no cointegration.Therefore, the SupF and MeanFsta-
tistics will be also.
There is another intuitive concept linking no coin-
tegrationand parameterinstability.Under the null of
cointegration,regressioncoefficientestimatesconverge
uniformlyin differentparts of the sample space to the
cointegratingrelationship.Under the alternativeof no
cointegration,however, the regressionestimates con-
verge to randomvariables,whichwill take on different
values in different samples. See Phillips (1986). Thus
sequential parameterestimates will display apparent
parameterinstability.This simple observationimplies
an importantmessage. Rejectionof the null of constant
parametersdoes not implythe particularalternativethe
test was designed to detect. There are many possibili-
ties. If the SupF test rejects, for example, it would be
quite inappropriateto conclude (on this piece of evi-
dence alone) that there were two cointegratingregimes,
which shifted at a particularpoint in the sample. The
only statisticallyjustifiedconclusionis that the standard
model of cointegration,includingits implicit assump-
tion of long-runstabilityof the cointegratingrelation-
ship, is rejected by the data.
6. APPLICATIONS
We now apply this testing method to three applica-
tions. In each example, the fully modified estimation
method is used. The covarianceparametersare esti-
matedusing a QS kernelon residualsprewhitenedwith
a VAR(1). (All of the reportedregressionswere also
estimated using the Parzen and Bartlett kernels, and
the results were nearly identical.) The bandwidthpa-
rameterwas selectedaccordingto the recommendations
of Andrews (1991), using univariateAR(1) approxi-
mating models. In all regressionsreported, the esti-
mates and standarderrors are the fully modified esti-
mates of Phillips and Hansen (1990). The estimated
plug-in bandwidthparameter(M) is reported. All of
the SupF and MeanF statisticsare calculatedusing the
trimmingregion [.15, .85]. The constancytest statistics
are reported along with their asymptoticp values (in
parentheses), which are calculated according to the
method of Section 4.
6.1 Aggregate Consumption Function
The notion of an aggregaterelationshipbetweencon-
sumptionand income has a long history in macroeco-
This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM
All use subject to JSTOR Terms and Conditions
Hansen:Test for Instabilitywith 1(1)Processes 329
nomics. In a recent article, Campbell (1987) showed
that a strict infinite-horizonpermanent-incomemodel
yields a cointegratingrelationshipbetween aggregate
consumption and aggregate disposable income. We
can now test the constancyof this cointegratingrela-
tionship.
The data, from Blinder and Deaton (1985), are sea-
sonallyadjustedaggregatequarterlyU.S. consumption
and total disposable income (DI,) in real per capita
units, for the period 1953:2-1984:4. Campbell esti-
mated the equation for both total consumption(TC,)
and nondurablesand servicesconsumption(NDS,). In
the following regressions,a constantand a time trend
were included:
TCt = -113 + -1.02t + .982DIt
(194) (1.83) (.088)
M = .89
SupF = 12.3 (.15)
MeanF = 6.2 (.05)
.51
Lc= (.09);
NDS, = 518 + 2.96t + .526DIt
(103) (.97) (.047)
AM= .97
SupF = 8.7 (>.20)
MeanF = 3.0
(>.20)
Lc = .14 (>.20).
First, examine the fully modified estimates. It ap-
pears that the correctionsare having an importantef-
fect. The OLS estimates of the first equation, for ex-
ample, yield a coefficientfor disposableincome of .85
ratherthan the economicallymore plausible .98. Note
that the estimatedbandwidthparameterfor both equa-
tions is less than 1, indicatingthat nearlyall of the serial
correlationin the residualswas capturedby the pre-
whiteningprocedure.
The tests when applied to the first equation do not
yield clear results, with p values rangingfrom .05 to
.15. Although the evidence suggeststhat the relation-
ship may indeed be unstable, the data are not suffi-
ciently informativeto be able to reject the null of sta-
bility. On the other hand, the second equation (for
nondurablesand services consumption)does not sug-
gest instabilityat all, since none of the test statisticsare
significantat the 20% level. These results (for both
equations)are robustto the choiceof kernelandwhether
the equationsare estimatedafter taking logarithms.
It is informativeto visuallyexaminethe sequence of
F statisticsfor structuralchange.Figures1 and2 display
these sequences for each regression, along with 5%
criticalvalues for its largest value (SupF), its average
value (MeanF), and for a fixed known breakpoint.
330 Journalof Business & EconomicStatistics,July 1992

I I I I I I I I I I I

4-1
0
U)
Ll-

C0

0 ? ? ? ? ?
1958 1962 1966 1970 1974 1978 1982

Figure 1. Total Consumption, 1953-1984: - , F Statistic Sequence; --, 5% Critical, SupF;...... 5% Critical, MeanF; ---, 5% Critical,
KnownBreak.

6.2 Present-Value Model

Campbell and Shiller (1987) argued that a standard
rational-expectations model of asset markets implies
that real stock prices and dividends should be cointe-
grated. Using price (P,) and (D,) indexes for the period
1871-1986, they found evidence to support this claim.
In a later series of articles, Campbell and Shiller (1988a,b)
argued for a logarithmic approximation that implicitly
assumes that the logarithms of the price and dividend
indexes are cointegrated. Using their data, we can test

OD
T- I II I I _I I I I I
' I

C0

0
v--

0
4-J

U) 00
d-..,

0 I I I I I

1958 1962 1966 1970 1974 1978 1982

Figure 2. NDS Consumption, 1953-1984: - , F Statistic Sequence; --, 5% Critical, SupF;...... 5% Critical, MeanF; ---, 5% Critical,
KnownBreak.

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
 Hansen:Test for Instabilitywith1(1)Processes 331

the stabilityof each specification: be unity, but the point estimate is significantlyabove
this value. The SupFand MeanFtests statisticssuggests
P, = -.15 + 32.1D,
that the relationshipis not stable. The plot of the se-
(.06) (3.5) quence of F statisticsis displayedin Figure4.
A = .78 Thisevidencesuggeststhatstockpricesanddividends
are indeed cointegrated,but the logarithmicapproxi-
SupF = 6.9 (>.20) mationused by Campbelland Shiller(1988a,b)may be
MeanF = 3.3 (.14) misspecified.
Lc = .30 (>.20);
6.3 Term Structure of Interest Rates
ln(P,) = 4.44 + 1.33 ln(D,) The theory of the term structureof interest rates
(.51) (.12) suggests that, if interest rates can be characterizedas
M = 1.07 1(1) processes,then they shouldbe cointegrated.Stock
and Watson (1988), for example, tested for cointegra-
SupF = 11.7 (.06) tion amongthree postwarU.S. interestratesand found
MeanF = 5.1 evidence of two cointegratingvectors (i.e., only one
(.03)
commontrend). They used monthlydata fromJanuary
Lc = .35 (.18). 1960to August 1979, presumablyto exclude a possible
regime shift in the term structuredue to the change in
The levels equation yields estimates very close to the FederalReserve'soperatingproceduresin 1979.We
those fromOLS (whichgivesa slope coefficientof 31.1). now test the hypothesisthat these relationshipsare sta-
Thiscorrespondsto a long-runreal-interestrateof 3.1%, ble over the entire period from January1960to March
which, as noted by Campbelland Shiller (1987), is be- 1990. We use the same series-the federal funds rate
low the sample mean returnof 8.2%. The relationship (FF), the 90-daytreasury-billrate (TB3), and the one-
appearsvery stable, however, with no significanttest year treasury-billrate (TB12)-and obtainedthe series
statistics. The plot of the sequence of F statistics is from the Citibasedata base.
displayedin Figure3. We report the results of two fully modified regres-
The logarithmicspecification does not perform as sions, TB3 on TB12 and TB3 on FF (the regressionof
well. The modelpredictsthatthe slope coefficientshould TB12 on FF yields resultsvery similarto the regression

00 -

0
I.s._

1880 1900 1920 1940 1960 1980

Figure 3. Stock Prices and Dividends, 1871-1986: , F Statistic Sequence; --, 5% Critical, SupF;...... 5% Critical MeanF; ---, 5%
Critical,KnownBreak.

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
332 Journalof Business&Economic
Statistics,July1992

in
1-

0
1880 1900 1920 1940 1960 1980

Figure 4. Logged Stock Prices and Dividends, 1871-1986: -- , F Statistic Sequence; --, 5% Critical SupF;...... 5% Critical, MeanF;
---, 5% Critical, Known Break.

of TB3 on FF): to have affected the relationship between the federal-

funds rate and the treasury-bill rates but not the rela-
TB3, = -.62 + 1.06TB12,
tionship between the treasury-bill rates of different
(.22) (.03) maturities.
M = 2.54
7. CONCLUSION
SupF = 3.6 (>.20) As shown by example in Section 6, in some appli-
MeanF = 1.6 (>.20) cations the three test statistics (SupF, MeanF, and Lc)
.21 may appear to be in conflict. There is no reason why
Lc= (>.20); all three tests should reject (or not reject) at a particular
TB3, = .49 + .83FF, level of significance in a particular sample. The tests
are looking in different directions and will have more
(.14) (.02) power against some alternatives than others. All of the
M = 2.51 tests, however, will have asymptotic power against the
same set of alternatives. The possibility of conflicting
SupF = 22.8 (.01) test statistics is not new to applied economists. There
MeanF = 8.4 (.01) are many tests for heteroscedasticity, for unit roots, for
cointegration, and so forth. The same care needs to be
Lc = .45 (.06). exercised in the present context. Calculation of all three
Over the entire period, it appears that the two treasury- test statistics seems the most judicious suggestion at this
bill rates are cointegrated with a stable relationship, time.
with a near-unity slope coefficient. This is strong sup- The tests were describedhere usingthe Phillips-Hansen
port for the theory of the term structure. In contrast, fully modified estimator. This is not the only possibility.
the relationship of the treasury-bill rate with the federal- It is quite straightforward to calculate the test statistics
funds rate appears unstable, with the SupF and MeanF for other asymptotically efficient estimates of cointe-
statistics highly significant. Figures 5 and 6 display the grating vectors, such as the MLE due to Johansen (1988)
sequences of F statistics for the two regressions. The or the "leads and lags" estimator of Saikkonen (1991)
sequence for the second regression crosses the 5% SupF and Stock and Watson (1991). Since the estimators are
critical value several times, achieving its maximal value asymptotically equivalent, the test statistics would have
approximately in 1980. This supports the conjecture the same asymptotic distributions as those tabulated in
that the change in the Federal Reserve's operating pro- this article. It is quite likely, however, that the asymp-
cedures altered the relationship between some interest totic proofs would be more difficult.
rates. It is interesting that this regime shift only appears This article only discussed joint tests on all of the

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
 Hansen:Test for Instabilitywith 1(1)Processes 333

00

4-

LL

0 -

0 I I I I I.

1960 1964 1968 1972 1976 1980 1984

Figure 5. Three-Monthand One-YearTreasury-BillRates, 1960-1990: , F Statistic Sequence; --, 5% CriticalSupF; ....... 5%

Critical MeanF; ---, 5% Critical, Known Break.

regression parameters in a cointegrating regression. It
should be possible to extend these results to tests on a
subset of the parameters as well. This will be left to
future research.
ACKNOWLEDGMENTS
An earlier version of this article was entitled "Testing
for Structural Change of Unknown Form in Models
With Non-stationary Regressors." My thanks go to Don
Andrews, Bill Brown, David Hendry, Peter Phillips,
Adrian Pagan, Jim Stock, the associate editor, the ref-
eree, and seminar participants at the University of
Rochester, Indiana University, and UCLA for helpful
discussions and comments on earlier drafts. This research
was supported by the National Science Foundation.
APPENDIX: PROOFS OF THE THEOREMS
Proof of Theorem 1
(a) The finite-dimensional result is immediate from
(14). Weak convergence follows from the continuous

0-

Co

Ic
U- r _

0
1960 1964 1968 1972 1976 1980 1984
and Federal Funds Rates, 1960-1990: --
Figure 6. Three-MonthTreasury-Bill , F StatisticSequence; --, 5% Critical,SupF; .....
5% Critical, MeanF; ---, 5% Critical, Known Break.

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions
334 Journalof Business & EconomicStatistics,July 1992

mapping theorem (see Billingsley 1968, p. 30) since + r(

XdB, T
M(r) is a continuousfunctionof r and X(.). a l 2 )
*21/

(b) This follows from part (a) and continuity.

(c) By (13) and theorem 4.1 of Hansen (1992c), - r X(f xx)\ XdB.-20- (
( nn 1
1 l-n-
RV f22 fl2 op)u+ = S(T)fl12 - M(T)M(1)- S(1)I'1.2
=
,nnCnx,ul'
1 '2 2t It
n"i
= S*(r)fi.2
f/ kdB.2
by Theorem 1 and the continuousmappingtheorem.
f
(lnf2n* 2 nr (frB2dB2 + tA+y (b)
Fn(T) = tr{Sn(T)'Vn(T)-'S(Tr)l21}

= tr{ S(T)'( rnvn(),T) , rsT)1}

J;WdB2+A
By (11),A
~\W2(d)
(d) By (11), > tr{lS2ST*(r)'
V(T)-S*(T)ll'.2ln.2 }
= tr{S*(T)'V(T)-lS*(T)}.

-
r2nA2+1 = | -1/2 2 (A21 - &22f221 21l) Proofof Theorem3
l(2*' 2\2 /2*)
Part(a) followsfromTheorem2 (b) and the fact that
0O conditional on 9x, for any r E ST,vec(S*(T)) = N(0,
II(n*' 1i22n 1/2
1.2 ( V(r)), and therefore
n1/

F(T)l -- X2b

=(2?) Since this distributionis independentof 9, it is the

unconditionaldistributionas well.
(e) Part (b) follows from Theorem2(b) and the contin-
n (A+ - A)F-1 = - V (OA+')rF uous mappingtheorem.
( 1v 1 lti2xr'
Part (c) follows from Theorem2(b), the continuous
mapping theorem and the fact that fyF(r)dr is well
-
(-12122
-
,121-22) n I 2t,Xt) defined.
Part (d) follows from Theorem l(a), Theorem2(a),
and (11).
\n x xn
[ReceivedJanuary1991. RevisedJanuary1992.]

=> dB ,.2X' + (0 A+')

REFERENCES

(OA,))(1 ) Anderson,T. W., and Darling,D. A. (1952), "AsymptoticTheory

of Certain 'Goodness of Fit' CriteriaBased on StochasticPro-
cesses," TheAnnals of Mathematical Statistics,23, 193-212.
- f dB2X'(f xx') Andrews, D. W. K. (1990), "Tests for ParameterInstabilityand
StructuralChange With Unknown Change Point," Discussion
Paper 943, Yale University,Cowles Foundationfor Researchin
by parts (a), (c), and (d), and (11). Economics.
and AutocorrelationConsistent
(1991), "Heteroskedasticity
Proof of Theorem 2 CovarianceMatrixEstimation,"Econometrica,59, 817-858.
Andrews,D. W. K., and Monahan,J. C. (in press), "An Improved
(a)
Ilnr Heteroskedasticityand AutocorrelationConsistent Covariance
1 1 Inl1 1 MatrixEstimator,"Econometrica,60.
-r,sn(r) = - rn St ,- r, alt
Banerjee,A., Lumsdaine,R. L., andStock,J. H. (1992),"Recursive
=
n- 7n tt = N/n 1=1 \ (A21))
andSequentialTestsof the Unit Root andTrendBreakHypothesis
1 [n'" Theory and InternationalEvidence. Journalof Business& Eco-
nomic Statistics,10, 271-287.
I , t= Billingsley, P. (1968), Convergenceof ProbabilityMeasures,New
York: John Wiley.
n t=l Blinder, A. A., and Deaton, A. S. (1985) "The Time Series Con-
sumptionFunctionRevisited,"BrookingsPaperson EconomicAc-
[n_] o0 tivity,465-511.
Vn r2A21J Campbell,J. Y. (1987), "Does SavingAnticipateDecliningLabor

This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM

All use subject to JSTOR Terms and Conditions

Income?An AlternativeTest of the PermanentIncome Hypoth-
esis," Econometrica,55, 1249-1273.
Campbell,J. Y., and Shiller,R. J. (1987), "CointegrationandTests
of PresentValueModels,"Journalof PoliticalEconomy,95, 1062-
1088.
(1988a), "Stock Prices, Earningand Expected Dividends,"
TheJournalof Finance,43, 661-676.
(1988b),"The Dividend-PriceRatio and Expectationsof Fu-
tureDividendsandDiscountFactors,"Reviewof FinancialStudies,
1, 195-228.
Chow, G. C. (1960), "Testsof EqualityBetweenSets of Coefficients
in Two LinearRegressions,"Econometrica,28, 591-605.
Chu, C.-S. J. (1989), "New Tests for ParameterConstancyin Sta-
tionaryandNonstationaryRegressionModels,"unpublishedman-
uscript,Universityof Californiaat SanDiego, Dept. of Economics.
Engle, R. F., and Granger,C. W. J. (1987), "Co-integrationand
Error-correction: Representation,EstimationandTesting,"Econ-
ometrica,55, 251-276.
Gardner,L. A., Jr. (1969), "On DetectingChangesin the Mean of
NormalVariates,"TheAnnalsof Mathematical Statistics,40, 116-
126.
Granger,C. W. J. (1981), "Some Propertiesof Time Series Data
and Their Use in EconometricModel Specification,"Journalof
Econometrics,16, 121-130.
Granger,C. W. J., and Lee, T.-H. (1990), "Multicointegration,"
Advancesin Econometrics,8, 71-84.
Gregory,A. W., and Nason, J. M. (1991), "Testingfor Structural
Breaks in CointegratedRelationships,"unpublishedmanuscript,
Queen's University,Dept. of Economics.
Hansen, B. E. (1990), "LagrangeMultiplierTests for Parameter
Instabilityin Non-linearModels," unpublishedmanuscript,Uni-
versityof Rochester,Dept. of Economics.
(in press a), "EfficientEstimationand Testing of Cointe-
gratingVectorsin the Presenceof DeterministicTrends,"Journal
of Econometrics,52.
(in press b), "ConsistentCovarianceMatrixEstimationfor
DependentHeterogeneousProcesses,"Econometrica,60.
(in press c), "Convergenceto StochasticIntegralsfor De-
pendentHeterogeneousProcesses,"EconometricTheory,8.
Herrndorf,N. (1984), "A FunctionalCentral Limit Theorem for
WeaklyDependentSequencesof RandomVariables,"TheAnnals
of Probability,12, 829-839.
Johansen,S. (1988), "StatisticalAnalysisof CointegrationVectors,"
Journalof EconomicDynamicsand Control,12, 231-254.
King, M. L. (1987), "An AlternativeTest for RegressionCoefficient
Stability,"TheReviewof Economicsand Statistics,69, 379-381.
Leybourne,S. L., and McCabe, B. P. M. (1989), "On the Distri-
bution of Some Test Statisticsfor CoefficientConstancy,"Bio-
metrika,76, 169-177.
Nabeya, S., and Tanaka,K. (1988), "AsymptoticTheoryof a Test
for the Constancyof RegressionCoefficientsAgainstthe Random
Walk Alternative,"TheAnnals of Statistics,16, 218-235.
This content downloaded from 62.122.76.45 on Sat, 21 Jun 2014 23:43:47 PM
All use subject to JSTOR Terms and Conditions
Hansen:Test for Instabilitywith 1(1)Processes 335
Nyblom, J. (1989), "Testingfor the Constancyof ParametersOver
Time," Journalof the AmericanStatisticalAssociation,84, 223-
230.
Nyblom, J., and Makelainen,T. (1983), "Comparisonsof Tests for
the Presence of Random Walk Coefficientsin a Simple Linear
Model,"Journalof the AmericanStatisticalAssociation,84, 856-
864.
Ogaki, M., and Park, J. Y. (1990), "A CointegrationApproachto
EstimatingPreferenceParameters,"Paper236, RochesterCenter
for EconomicResearch.
Pagan,A. R., andTanaka,K. (1981), "A FurtherTest for Assessing
the Stabilityof RegressionCoefficients,"unpublishedmanuscript.
Park,J. Y., andOgaki,M. (1991),"Inferencein CointegratedModels
UsingVAR Prewhiteningto EstimateShortrunDynamics,"Work-
ing Paper281, RochesterCenterfor EconomicResearch.
Park,J. Y., Ouliaris,S., and Choi, B. (1988), "A New Approachto
Testing for a Unit Root," mimeo, Cornell University,Dept. of
Economics.
Park, J. Y., and Phillips,P. C. B. (1989), "StatisticalInferencein
RegressionsWithIntegratedProcesses,Part1," EconometricThe-
ory, 4, 468-497.
Parzen, E. (1957), "On ConsistentEstimatesof the Spectrumof a
StationaryTime Series,"TheAnnalsof Mathematical Statistics,28,
329-348.
Perron,P. (1989), "The Great Crash,the Oil Price Shock, and the
Unit Root Hypothesis,"Econometrica,57, 1361-1401.
Phillips, P. C. B. (1986), "UnderstandingSpuriousRegressionsin
Econometrics,"Journalof Econometrics,33, 311-340.
Phillips,P. C. B., and Hansen, B. E. (1990), "StatisticalInference
in InstrumentalVariablesRegressionWithI(1) Processes,"Review
of EconomicStudies,57, 99-125.
Phillips, P. C. B., and Loretan, M. (1991), "EstimatingLong-run
EconomicEquilibria,"Reviewof EconomicStudies,58, 407-436.
Phillips,P. C. B., and Ouliaris,S. (1990), "AsymptoticProperties
of Residual Based Tests for Cointegration,"Econometrica,58,
165-193.
Quandt,R. (1960), "Testsof the HypothesisThat a LinearRegres-
sion SystemObeysTwo SeparateRegimes,"Journalof theAmer-
ican StatisticalAssociation,55, 324-30.
Saikkonen,P. (1991), "AsymptoticallyEfficientEstimationof Coin-
tegratingRegressions,"EconometricTheory,7, 1-21.
Snow, M. S., and Im, E. I. (1991), "The Equivalenceof Two Test
Statisticsfor Testing the Constancyof RegressionCoefficients,"
EconometricTheory,7, 419-420.
Stock, J. H., and Watson, M. W. (1988), "Testingfor Common
Trends,"Journalof theAmericanStatisticalAssociation,83, 1097-
1107.
(1991),"A SimpleEstimatorof Cointegrating Vectorsin Higher
OrderIntegratedSystems,"unpublishedmanucript.
Zivot, E., andAndrews,D. W. K. (1992), "FurtherEvidenceon the
GreatCrash,the Oil-PriceShock,andthe Unit-RootHypothesis,"
Journalof Business& EconomicStatistics,10, 251-270.