Professional Documents
Culture Documents
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.
Taylor & Francis, Ltd., American Statistical Association are collaborating with JSTOR to digitize,
preserve and extend access to Journal of Business & Economic Statistics
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
? 1994 American Statistical Association Journal of Business & Economic Statistics, October 1994, Vol. 12, No. 4
Wayne A. FULLER
Department of Statistics, Iowa State University, Ames, IA 50011
During the past 15 years, the ordinary least squares estimator and the corresponding pivotal
statistic have been widely used for testing the unit-root hypothesis in autoregressive processes.
Recently, several new criteria, based on maximum likelihood estimators and weighted symmetric
estimators, have been proposed. In this article, we describe several different test criteria. Results
from a Monte Carlo study that compares the power of the different criteria indicate that the
new tests are more powerful against the stationary alternative. Of the procedures studied, the
weighted symmetric estimator and the unconditional maximum likelihood estimator provide the
most powerful tests against the stationary alternative. As an illustration, the weekly series of
one-month treasury-bill rates is analyzed.
that p = 1. Different estimators and test criteria are obtained depending on what is assumed about Y1. Test criteria
are typically constructed using likelihood procedures under
the assumption that the et's are normally distributed. The
asymptotic distributions of the test statistics are, however,
valid under much weaker assumptions on the distribution of
suggested tests for unit roots based on these estimators. Elliott and Stock (1992), Elliott, Rothenberg, and Stock (1992),
1.1 Y1 Fixed
root less than 1 with much higher power than the test criteria
When YI is considered fixed and et ,, NI(0, a2), maximizing the log-likelihood function is equivalent to minimizing
(1993).
We summarize the new approaches, introduce a new test,
t=2
our conclusions.
1. TEST CRITERIA
in
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
OLS
n(I,M - 1) G-'1
I,ML G-1/2t
1,ML
G-'I2
n(P7,GLS - 1) G-
77,GLS G-1/2
En ( ,-;....) .0_.
Weighted symmetric
and
S [Y - Pl,ML - P,ML)
t=2
- 1,MLYt-1] 2 (1.8)
This is the most used test in practice because the pivotal ap-
G-lI may be found in table 8.5.1 of Fuller (1976). The pivotal and the likelihood-ratio-type statistics for testing p = 1
are
n n 1/2
and
G = W2(t)dt, H = W(t)dt,
and W(t) is a standard Brownian motion on [0, 1]. Empirical
n[p^(j) - 1] - G-'1 and the estimator has the same largesample behavior as the estimator obtained by regressing Yt
-Y1 on Yt-1 - Y1, which was suggested by Dickey and Fuller
(1979). See also Chan and Wei (1987) and Chan (1988).
Remark 1.2. Based on Remark 1.1, several approximations to the ML estimator are possible. We consider the
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
on the iteration number i. Even though the asymptotic distribution of n[f) - 1] obtained for a finite i is not the same
where V{(iU} is the variance of iju computed from the estimated information matrix. The limiting distribution of Tj
was given by Gonzalez-Farias (1992).
i=O
t=l
n-1
and
n-1 n -1/2
Y1 + (1 - p) _21Yt + Yn
2 + (n - 2)(1 - p)
equation (1.15) and the cubic equation can be solved iteratively. If the iterations converge, the estimators converge to
the simple symmetric and the weighted symmetric estimators treat observations at the beginning of the sample period
in the same way as observations at the end of the sample
period.
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
The simple symmetric estimator was included in our original Monte Carlo simulations. Because the power of the WS
statistic
25
50
100
250
oo
estimator.
NOTE: Except for (DOLS and 4N1.ML, reject Ho for values of the statistic less than the
critical values.
In this section, we study the empirical power of the statistics described in Section 2.1. Critical values for 5%-level
tests are given in Table 2, Section 2. The percentiles for finite
samples are estimated by generating et's from a standard nor-
Our model is
case (1) than for case (2) for all test statistics. The tables
contain only the pivotal statistics and 4ES. Except for OLS,
the pivotal statistics generally have power that is comparable
(1979).
We consider the test criteria based on the following.
5. WS -- n('ws - 1), ws
to or greater than that based on n(p'- 1). For OLS, the power
of n(p',oLS - 1) is much greater than that of T,OLS but is less
than that of n(&ws - 1).
(1) Y, s N(0, 1)
The powers of the pivotal statistics ,OLS, T7,GLS, and ?ws
are plotted in Figure 1, page 455. From the tables, we observe
the following:
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
Statistic
.98
.95
.93
.90
.85
.80
.70
Y, - N(0, 1)
Y, N[O, (1 -2)-'
4. For all sample sizes, the criteria based on the unconditional likelihood and WS estimators have higher power than
the true model, fP,ML and ir,ML, have higher power than those
slightly higher power than T7,GLS, and these two test criteria have higher power than the remaining criteria. For values
lower power than the criteria based on the unconditional likelihood and WS estimators.
(2).
3. EXTENSIONS
The test statistics presented here can be extended to higher-
Table 4. Empirical Powers for 5%-Level Test Criteria (n = 50, 5,000 replications)
p
Statistic
.98
.95
.93
.90
.85
.80
.70
Y,r N(0, 1)
Y, , N[O, (1-p2)-'
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
Statistic
.98
.95
.93
.90
.85
.80
.70
Y N (, 1)
Gonzalez-Farias (1992) gave an extension of the unconditional ML estimation for higher-order AR processes. Under
some regularity conditions on the roots, the ML estimation
tribution of n(I,GLS - 1), '7,GLs, and IES under the assumption that et satisfy some stationary and mixing conditions.
The limiting distributions depend on a nuisance parameter
Table 6. Empirical Pbwers for 5%-Level Test Criteria (n = 250, 5,000 replications)
p
Statistic
.98
.95
.93
.90
.85
.80
.70
Y N(O, 1)
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
(0
Co
CO
0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.70 0.75 0.80 0.85 0.90 0.95 1.00
rho
.o
rho
0GLS
""\'.\
.0
OLS
.. .....WLS
o--
GLS
0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.70 0.75 0.80 0.85 0.90 0.95 1.00
rho
rho
Figure 1. Powers of Pivotal Statistics When YI Is Generated From N(O, 1): , OLS;......., WLS; -.-.-., GLS.
12
t=p+1
i=2
t=p+1
i=21
given in Table 7.
In Section 3.3, we present simulation results that compare
the WS pivotal statistic with the OLS estimator. For the case
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
O
o
o
O
,
r
T-\
0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.70 0.75 0.80 0.85 0.90 0.95 1.00
rho
rho
n= 00,Yof Piv1 otalStationary n=250,Ys Generated From N[1 OLS; .........WLStationary; -----o
oo
"1
OLS
S--- GLS
0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.70 0.75 0.80 0.85 0.90 0.95 1.00
rho
rho
Figure 2. Powers of Pivotal Statistics When Y1 Is Generated From N[O,(1 -p2)-'J: OLS;.......... WLS; .-...., GLS.
where
testing 01 = 1, is
K = 2 tw(t)dt - w(t)dt.
Dependent
1- w2 Y, Y2 -Z3 . Zp+I
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
Statistic
OLS,p,A,
Tws,p,IA
= -.5, .5. For the cases IpI < 1, the initial observations are
100
-2.89
-2.60
oo
-2.86
-2.50
from those given in Table 2. The last column, for the infinite
much higher than the nominal level for the trend case when
1.00
5.8
5.3
5.4
4.6
AR(2)
=
-.5
.5
1.00
5.8
.95 7.9
.90 17.9
.80 44.7
.70 65.8
5.5
5.0
4.5
ARMA(1, 1)
=.5
71.4
1.00 6.0
.95 16.0
.90 39.2
.80 83.2
.70 94.3
85.8
6.8
24.8
51.4
85.8
94.6
48.2
57.2
10.7
14.7
26.8
69.1
89.4
15.0
30.5
67.3
88.8
9.9
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
AR(1)
AR(2)
Phi=-0.5
o.
0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.70 0.75 0.80 0.85 0.90 0.95 1.00
rho
rho
1".
0o
"'.
'
o?.
'
....
OLSmn
WLSmn
"""OLSt
to
- WLSt
0.70
0.75
0.80
0.85
0.90
0.95
1.00
rho
0.70
0.75
0.80
0.85
0.90
0.95
1.00
rho
Figure 3. Powers of Pivotal Statistics in Higher Order Processes for Both the Mean and the Trend Cases. Sample size is 100. OLSmn
= TOLS,p,, (- ); WLSmn = ws,p, (........); OLSt = OLS,pA, (-. ); WLSt = ws,p,(- -).
ARMA(1,1) model with 0 = .5, and 4.8 for the ARMA(1, 1)
model with q = -.5. The average ofws, is 3.9 for the AR(1)
model, 4.8 for the two AR(2) cases, 5.8 for the ARMA(1, 1)
model with 0 = .5, and 4.9 for the ARMA(1,1) model with
the day of the week. In particular, for our Friday series, the
buyer of the bill on Friday would normally take delivery the
0 = -.5.
4. EXAMPLE
7TOLS,
7w,
-2.49
-2.67
-2.56
-2.75
10
-2.44
-2.64
15
-2.02
-2.23
20
-1.76
-1.99
30
-1.60
-1.84
39
-1.61
-1.89
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms
1063.
15, the WS test fails to reject at the 10% level. The OLS
(1981), "Likelihood Ratio Tests for Autoregressive Time Series With a Unit
Root," Econometrica, 49, 1057-1072.
test fails to reject the unit root at the 10% level for all of the
et al. (1992) are very powerful under the model in which the
initial observations is N(0, a2) but are not the most powerful
against the alternative in which the process is a stationary AR
estimator are the most powerful. These tests also have good
power when the first observation is N(O, cr2). Hence they are
ACKNOWLEDGMENTS
We thank David Dickey, an associate editor, and the editor
for their comments. We are grateful to Cheng Su for helping
79, 355-367.
Autoregressive Unit Root:' unpublished paper presented at the NBERNSF Time Series Seminar, October 31, Chicago.
Elliott, G., and Stock, J. H. (1992), "Efficient Tests for an Autoregressive
Unit Root," unpublished paper, presented at the 1992 Joint Statistical
Wiley.
Park, H. J., and Fuller, W. A. (1993), " Alternative Estimators for the Pa-
Phillips, P. C. B., and Perron, P. (1988), 'Testing for a Unit Root in a Time
REFERENCES
Bansal, R., Gallant, A. R., Hussey, R., and Tauchen, G. (1992), "Nonparametric Estimation of Structural Models for High-Frequency Market Data,"
71,599-608.
SAS (1985), SAS Basics User Guide (Version 5), Cary, NC: SAS Institute.
This content downloaded from 200.3.149.179 on Wed, 06 Apr 2016 23:40:52 UTC
All use subject to http://about.jstor.org/terms