{smcl}
{* 14apr2003}{...}
{hline}
help for {hi:johans}, {hi:lrjtest}, {hi:wjtest}{right:[P.Joly]}
{hline}
{title:Johansen's maximum likelihood cointegration rank test}
{p 8 27}
{cmd:johans}
{it:depvarlist}
[{cmd:if} {it:exp}]
[{cmd:in} {it:range}]
[{cmd:,}
{cmdab:l:ags(}{it:#}{cmd:)}
{cmdab:noc:onstant}
{cmdab:t:rend}
{cmdab:ex:og}({it:varlist})
{cmdab:non:ormal}
{cmdab:stand:ard}
{cmdab:r:egress}
{cmdab:l:evel(}{it:#}{cmd:)}
]
{p 8 27}
{cmd:lrjtest}
{it:varlist}
{bind: }
[{cmd:,}
{cmdab:c:ir(}{it:#}{cmd:)}
{cmdab:r:estrict}
]
{p 8 27}
{cmd:wjtest}
{it:varlist}
{bind: }
[{cmd:,}
{cmdab:c:ir(}{it:#}{cmd:)}
]
{p}
{cmd:johans} is for use with time-series data. You must {cmd:tsset}
your data before using {cmd:johans}; see help {help tsset}.
Your version of Stata must also be up-to-date.
Type {cmd:update query} to find out whether updating is required; see
help {help update}.
{p}
{cmd:johans} may be used with panel data if the estimation sample is
restricted to a single panel. Observations for which the panel
variable is missing are omitted from the sample.
{p}
{cmd:by} {it:...} {cmd::} may be used with {cmd:johans}; see help
{help by}.
{p}
{it:depvarlist} and {it:varlist} may contain time-series operators;
see help {help varlist}.
{title:Description}
{p}
{cmd:johans} calculates the eigenvalues and the maximal
eigenvalue and trace statistics for the VAR defined by {it:depvarlist}.
These statistics can be used to test for the number of
cointegrating vectors in the system.
If option {it:normal} or {it:standard} is specified, maximum
likelihood estimates of the cointegrating vectors and of the matrix of
weights are displayed.
{cmd:johans} uses Johansen's method for computing the estimates and
test statistics.
See the references for more information on the method.
{p}
{cmd:lrjtest} and {cmd:wjtest} are used after {cmd:johans} to
test the null hypothesis that one or more of the variables in the VAR
do not enter the cointegrating relationship(s).
{cmd:lrjtest} calculates the likelihood-ratio test.
{cmd:wjtest} calculates the Wald test.
{title:Options}
{p 0 4}
{cmd:lags(}{it:#}{cmd:)} specifies the number of lags in the original
VAR (in levels). The default is 1.
{p 0 4}
{cmd:noconstant} suppresses the constant in the first stage VAR.
{p 0 4}
{cmd:trend} includes a trend term in each equation of the VAR.
{p 0 4}
{cmd:exog(}{it:varlist}{cmd:)} specifies exogenous variables that
enter each equation of the VAR.
{p 0 4}
{cmd:nonormal} suppresses the display of the normalized Alpha and
Beta' matrices.
{p 0 4}
{cmd:standard} requests the display of the standardized Alpha and
Beta' matrices instead of the normalized matrices.
{p 0 4}
{cmd:regress} indicates that the VAR estimates are to be displayed.
{p 0 4}
{cmd:cir(}{it:#}{cmd:)} is for use with {cmd:lrjtest} and {cmd:wjtest}
and indicates the number of cointegrating relationships in the system.
The default is 1.
{p 0 4}
{cmd:restrict} is relevant only with {cmd:lrjtest} and requests the
display of the restricted estimates of the eigenvalues and of the
Alpha and Beta' matrices.
{p 0 4}
{cmd:level(}{it:#}{cmd:)} specifies the confidence level, in percent,
for confidence intervals of the test statistics.
{it:#} must be one of 50, 80, 90, 95, 97.5, or 99; 95 is the default.
{title:Methods & Formulas}
{p}
Consider a vector autoregression of order {it:k}, VAR({it:k}),
{p 12}{it:k}{p_end}
{p 4 12}
{bind:{hi:x}_{it:t} = }
{bind:SUM {hi:B}_{it:i} * {hi:x}_{it:t-i}}
{bind: + {hi:f} * {hi:D}_{it:t}}
{bind: + {it:{hi:m}_0}}
{bind: + {it:{hi:m}_1} * {it:t}}
{bind: + {it:{hi:e}_t}}{p_end}
{p 11}{it:i}=1
{p 4 12}({it:t} = 1,...,{it:T} )
where
{p 4 18}{hi:x}_{it:t}{bind: }({it:p} x1) vector of {it:p} stochastic variables;{p_end}
{p 4 18}{hi:B}_{it:i}{bind: }({it:p} x{it:p} ) matrix of coefficients;{p_end}
{p 4 18}{hi:D}_{it:t}{bind: }(optional) matrix of zero-mean seasonal dummies with corresponding coefficient vector {hi:f}; and{p_end}
{p 4 18}{hi:m}_{it:0}, {hi:m}_{it:1}{bind: }(optional) ({it:p} x1) vectors of coefficients on intercept and trend terms, respectively.
{p}
Any such VAR({it:k}) may be re-parameterized as
{p 13}{it:k}-1{p_end}
{p 4 12}
{bind:d.{hi:x}_{it:t} = }
{bind:SUM {hi:G}_{it:i} * d.{hi:x}_{it:t-i}}
{bind: + {hi:P} * {hi:x}_{it:t-1}}
{bind: + {hi:f} * {hi:D}_{it:t}}
{bind: + {hi:m}_{it:0}}
{bind: + {hi:m}_{it:1} * {it:t}}
{bind: + {hi:e}_{it:t}}{p_end}
{p 13}{it:i}=1
{p 8}
{hi:P} = {hi:ab'}
{p}
where "d." represents the first-difference operator and {hi:G}_{it:i},
{hi:P}, are ({it:p} x{it:p} ) matrices of coefficients.
As shown above, matrix {hi:P} can be further decomposed into the product
{hi:ab'} where {hi:b'} represents the matrix of ({it:r}) cointegrating
vectors and {hi:a}, the matrix of weights on the vectors in each
equation of the system.
Both {hi:a} and {hi:b} are of dimension ({it:p} x{it:r} ).
(The notation {hi:P} = {hi:ab'} is a departure from the convention of
using uppercase (lowercase) letters for matrices (scalars) but has
become embedded in the cointegration literature so we stick
to it.)
{p}
The Johansen cointegration rank test is a test of the rank of the {hi:ab'}
matrix.
If, after inference, the rank is deemed to be ({it:r}), then there are
({it:r}) cointegrating relationships or vectors in the system.
The rank of any matrix can be determined by the number of non-zero
eigenvalues for that matrix.
{p}
Two features of the Johansen procedure are worth noting.
First, the test is really a sequence of tests.
The null hypothesis of rank (r)=0 (i.e. no cointegrating relationship)
is first tested and, if rejected, subsequent null hypotheses (Ho: r=1, Ho: r=2, etc.)
are tested until a null can no longer be rejected.
This is eminently sensible since we must first enquire whether any
cointegrating relation exists at all, and only in the affirmative are we
interested in finding out exactly how many can be identified.
{p}
The second distinguishing feature of the procedure is that it provides
two ways (two test statistics, that is) of testing the same null
hypothesis.
The difference lies in the alternative hypotheses implied by each.
The {it:maximum-eigenvalue} statistic tests the null hypothesis that
there are {cmd:at most} (r) cointegrating vectors against the
alternative of (r+1) vectors.
The {it:trace} statistic tests the same null hypothesis against the
alternative that there are at most {it:p} cointegrating relations
(where is {it:p} is always equal to the number of variables in the
{hi:x}_{it:t} vector).
It is important to note that the two statistics may lead to
conflicting results.
{p}
However, before making inference using the max-lambda and trace
statistics, certain assumptions regarding the deterministic components
(intercept, trend terms) must be made.
These assumptions will determine which table of critical values should
be used.
Following Osterwald-Lenum's notation (see {hi:references}),
let {hi:y}_{it:t} represent a subset or portion of the
re-parameterized VAR (the 2nd equation above), specifically, let
{p 4 12}
{bind:{hi:y}_{it:t} = }
{bind:{hi:ab'} * {hi:x}_{it:t-1}}
{bind: + {hi:m}_{it:0}}
{bind: + {hi:m}_{it:1} * {it:t}}
{p}
The precise form of this {hi:y}_{it:t} term depends on the assumptions
regarding the presence of intercept or trend terms and whether such
terms appear {it:within} or {it:outside} the cointegrating
equation(s) (CE).
By {it:outside} it is meant that a deterministic term lies in the
{it:remainder} of the VAR, that is, that portion of the VAR other
than the CE.
The possible assumptions or cases are:
{ul:Table/Case 0:} {it: no intercept, no trend}
{hi:y}_{it:t} = {hi:ab'} * {hi:x}_{it:t-1}
{ul:Table/Case 1*:} {it: intercept in CE only}
{hi:y}_{it:t} = {hi:a}({hi:b'}, {hi:b}_0) * ({hi:x'}_{it:t-1}, 1)'
{ul:Table/Case 1:} {it: intercept in VAR}
{hi:y}_{it:t} = {hi:ab'} * {hi:x}_{it:t-1} + {hi:m}_{it:0}
{ul:Table/Case 2*:} {it: intercept in VAR, trend in CE only}
{hi:y}_{it:t} = {hi:a}({hi:b'}, {hi:b}_1) * ({hi:x'}_{it:t-1}, {it:t})' + {hi:m}_{it:0}
{ul:Table/Case 2:} {it: intercept in VAR, trend in VAR}
{hi:y}_{it:t} = {hi:ab'} * {hi:x}_{it:t-1} + {hi:m}_{it:0} + {hi:m}_{it:1} * {it:t}
{p}
where {hi:b}_0 and {hi:b}_1 are of dimension ({it:r} x 1).
The case number refers to the table identifier in Osterwald-Lenum.
The corresponding critical values are reported by {cmd:johans}.
This may seem confusing but it shouldn't be.
Recall that a constant (trend) in an equation in difference-form
implies a trend (quadratic trend) in levels.
Therefore, cases 0 and 1* are appropriate when the series in
{hi:x}_{it:t} do not exhibit some apparent trend (when plotted in
levels); otherwise cases 1 and 2* should be used.
Case 2 is rare and only consistent with explosive series.
Case 1 is the most common and is the default in {cmd:johans}.
{p}
For instance, using the first of the examples provided in the
{hi:Examples} section below,
{com}. use http://fmwww.bc.edu/ec-p/data/macro/wgmacro.dta, clear
{txt}(Quarterly SA West German macro data, Bil DM, from Lutkepohl 1993 Table E.1)
{com}. johans investment income consumption, l(4) nonormal
{txt}Johansen-Juselius cointegration rank test{right: {txt}Sample: {res}1960q4 {txt}to {res}1982q4}
{right: {txt}Number of obs = {res}88}
{txt}{col 25}{* c |}{col 41}H1:
{txt}{col 18}H0:{col 25}{c |} Max-lambda{col 47}Trace
{txt} Eigenvalues rank<=(r){col 25}{c |} statistics{txt}{col 45}statistics
{txt} (lambda){col 19}r{col 25}{c |} (rank<=(r+1)){txt}{col 44}(rank<=(p={res}3{txt}))
{txt}{dup 24:{c -}}{c +}{dup 32:{c -}}
{res} .21617042 {txt} 0 {txt}{c |} {res} 21.433602 33.527577
.11390069 {txt} 1 {txt}{c |} {res} 10.64151 12.093974
.01636981 {txt} 2 {txt}{c |} {res} 1.4524647 1.4524647
{txt}{col 5}Osterwald-Lenum Critical values (95% interval):
{txt}{col 5}Table/Case: {res:1*}
{txt}{col 5}(assumption: {res:intercept in CE})
{txt}{col 18}H0:{col 25}{c |} Max-lambda{col 47}Trace
{txt}{col 14}{dup 11:{c -}}{c +}{dup 32:{c -}}
0 {txt}{c |} {res} 22.00 34.91
{txt} 1 {txt}{c |} {res} 15.67 19.96
{txt} 2 {txt}{c |} {res} 9.24 9.24
{txt}{col 5}Table/Case: {res:1}
{txt}{col 5}(assumption: {res:intercept in VAR})
{txt}{col 18}H0:{col 25}{c |} Max-lambda{col 47}Trace
{txt}{col 14}{dup 11:{c -}}{c +}{dup 32:{c -}}
0 {txt}{c |} {res} 20.97 29.68
{txt} 1 {txt}{c |} {res} 14.07 15.41
{txt} 2 {txt}{c |} {res} 3.76 3.76
{txt}
{p}
If we believe there is an intercept term in the VAR but not in
the CE, we refer to the quantiles in Table 1 of Osterwald-Lenum.
Using the Max-lambda statistic, we test the null hypothesis that (r)
= 0 against the alternative that (r) is at most 1.
Our test statistic of 21.43 exceeds the critical value of 20.97 which
leads to the rejection of the hypothesis of no cointegrating
relationship.
The Trace statistic of 33.53 also exceeds its corresponding critical
value of 29.68 which is consistent with the result using the
Max-lambda statistic.
{p}
We now move on to testing the null hypothesis that the rank of
{hi:ab'} is 1.
In this instance however, the Max-lambda statistic of 10.64 is smaller
than the critical value of 14.07 and we cannot reject the null
hypothesis.
The Trace test leads to the same conclusion.
Therefore, regardless of which statistic is used, we cannot reject the
hypothesis that we have 1 cointegrating vector.
{title:Examples}
{p 8 12}{inp:.} {stata "use http://fmwww.bc.edu/ec-p/data/macro/wgmacro.dta, clear":use http://fmwww.bc.edu/ec-p/data/macro/wgmacro.dta, clear}
{p 8 12}{inp:. johans investment income consumption, l(4)}{p_end}
{p 8 12}{inp:. lrjtest investment, c(1)}{p_end}
{p 8 12}{inp:. wjtest investment, c(1)}{p_end}
{p 8 12}{inp:. lrjtest income, c(1)}
{p 8 12}{inp:. * generate zero-mean (centered) seasonal dummies}{p_end}
{p 8 12}{inp:. forvalues i = 1/4 { gen byte q`i' = 0 }}{p_end}
{p 8 12}{inp:. forvalues i = 1/4 { replace q`i' = 1 if quarter(dofq(qtr))==`i' }}{p_end}
{p 8 12}{inp:. forvalues i = 1/4 { replace q`i' = q4-q`i' }}{p_end}
{p 8 12}{inp:. johans investment income consumption, l(4) exog(q1 q2 q3)}
{p}With longitudinal data, for each unit of a panel,
{p 8 12}{inp:.} {stata "use http://fmwww.bc.edu/ec-p/data/Greene2000/TBL15-1, clear":use http://fmwww.bc.edu/ec-p/data/Greene2000/TBL15-1, clear}
{p 8 12}{inp:. tsset firm year, yearly}{p_end}
{p 8 12}{inp:. by firm: johans i f c, l(3)}
{p 8 12}{inp:. johans i f c if firm==3, l(3)}{p_end}
{p 8 12}{inp:. lrjtest i, c(1) restrict}{p_end}
{p 8 12}{inp:. lrjtest c, c(1) restrict}
{title:References}
{p 0 2}
Johansen, S. (1988). "Statistical Analysis of Cointegration Vectors".
{it:Journal of Economic Dynamics and Control}. {cmd:12}. 231-254.
{p 0 2}
Johansen, S. and K. Juselius. (1990). "Maximum Likelihood Estimation
and Inference on Cointegration - With Applications to the Demand for
Money". {it:Oxford Bulletin of Economics and Statistics}. {cmd:52}.
169-210.
{p 4 4}
For inference, the most comprehensive set of critical values can be
found in:
{p 0 2}
Osterwald-Lenum, M. (1992). "A Note with Quantiles of the Asymptotic
Distribution of the Maximum Likelihood Cointegration Rank Test
Statistics". {it:Oxford Bulletin of Economics and Statistics}.
{cmd:54}. 461-472.
{title:Acknowledgements}
{p}
Thanks to Richard Sperling for suggestions on the displayed output.
{title:Author}
{p 0 2}Ken Heinecke and Charles Morris, Federal Reserve Bank of Kansas City
{p 0 2}Patrick Joly, Industry Canada{p_end}
{p 0 2}pat.joly@utoronto.ca
{title:Also see}
{p 0 19}
On-line: help for
{help vecar} (if installed),
{help vececm} (if installed),
{help dfuller},
{help pperron}
{p_end}