Johansson cointegration, ECM C0integration test (first we chec k stationary of data if all variable are stationary at first difference, then we will check cointegration ,if we find cointegration then we can go for ECM ,following chart also describing this procedure
Remember, the cointegration test is only valid if you have non-stationary series! The purpose of the cointegration test is to determine whether several non-stationary time
series are cointegrated or not. The presence of a Cointegrating relation forms the basis of the ECM(Cointegrating tells about the long run relationship, existence) To perform Johansen cointegration test, first open the series: or quick-group statistics Suppose all the variables are stationary at first difference and now we are going for
cointegration
Muhammad Saeed Meo, superior university Page 1
Muhammad Saeed Meo, superior university Page 2
Simple enter and in a dialog box simple write name of variable like lgdp lexport and enter. When
you
will
inter
a
window
will
be
open
like
Explanation of this dialog box Deterministic trend assumption of test Practical guides: • use case 1 only if you know that all series all series have zero mean (unusual in empirical studies);
• case 5 may provide a good fit in-sample in-sample but will produce implausible forecasts out-ofsample.;
• use case 2 if none of the series appear to have a trend; • use case 3 if series are trending and you believe all trends are stochastic;
• use case 4 if se ries are trending and you believe some of them are trend stationary;
• use case 6 if you are not certain which trend assumption to use (Eviews will help you determine the choice of the trend assumption). Now we select 3 and enter and results are these.
Muhammad Saeed Meo, superior university Page 3
How to know either here is cointegration of not.so read this line and this will show how much cointegration equations exist, remember at least, there must be one cointegration equation exist in this example we have two cointegration equations that is good. Sometime trace test indication cointegration, while
max-eigen don’t show don’t show cointegration between variables, so not so serious prob, in this case you can take trace test as bench mark and conclude that there is a cointegration. But it is better both of test show cointegration.
The first two tables report results for testing the number of Cointegrating relations. Two types of test statistics are reported: trace statistics and the maximum eigenvalue statistics. For each tab le, the first column is the number of Cointegrating relations under t he null hypothesis, the second column is the ordered eigenvalues of the matrix Π, the third column is the test statistic, and the last two columns are the 5% critical values
ECM error correction model (in above pages we saw cointegration among variable so now we will go for Ecm
Why we use error correction model? There are some problems in first difference these are following that’s why we apply ECM.
In first difference even we find that spurious relation(meaningless relation ,mathematically relation exist but logically does not ex ist like relationship between GDP and Corruption there must be negative relation but due to trend here positive positive relation exist) does not exist but we eliminate constant Out correlation occur in error term(autocorrelation leads insignificant effects)
Muhammad Saeed Meo, superior university Page 4
We can do over estimate or under estimate due to first differencing Difficult to find static equilibrium How to apply ECM? Drag file on Eviews note if you are using eveiw 5 then save excel file in file type 2003( suppose I have checked already stationary level and cointegration ) Now you’re variable appear in eveiw then simple go to quick estimate equation. How to write equation of ECM simple first dependent variable then independent like d(lgdp) c d(lexport) lgdp(-1) lexport(-1) and lexport(-1) and enter. Here GDP is our dependent variable and export is independent all are in logged form and space in c and variables .(pre-condition I have already told in crux part means all variables must be stationary at first difference)
Muhammad Saeed Meo, superior university Page 5
Now explanation of ECM D tells about the short run relation and prob value tell about the significant or insignificant relation,, and coefficient tell about the positive or negative relationship means if let suppose D LEXPORTS coefficient values comes in negative and probability value comes less than five percent then we will explain that there is short run negative and significant association between exp and gdp . Same explanation for positive association and (-1) tells about long run relation Here u can conclude that in short run and long run explored don’t have relationship with gdp R square values tells,, explanatory power of u model,, means how much ur dependent variable is effecting due to independent variable like in this case our dependent variable is effecting 99 percent .you note results are telling
Durbin Watson this value tells about the auto correlation in ur
that no relationship but R-square show much strength the reason is that I have not chosen good statistical test because all variables are not stationary at first difference. I just apply this model for your understanding.
variables which is not good, if ur values comes near 2 then there is
F-STATISTIS, values tells about the good fit of model. Means either jointly all variables effect or not dependent variables .F-statics value must be more
no correlation,, and D-W values
than F-STATISTSITCS problity vlaue.(note selection of right statistical test is
comes between 2-4(near two
most important to avoid from spurious results)
means 1.6,7,8,9,) but much less means harmful like in above results
d indicate short run relationship between variable like in this table export and dgp having no significant relationship in short run value.91 and in long run .94 probability value is value that is more than 5% which is insignificant so no relationship in long run .the variables which have values(-1) indicate that how much quickly in extraordinary situations dependent variable convert
Muhammad Saeed Meo, superior university Page 6
toward own actual level normally values comes equal to one or less than one if values increase more than one it means dependent variable will not come again in actual level.
Muhammad Saeed Meo, superior university Page 7
