Cointegration test eviews

[economistpk]

Hello,
I am working on Error correction modeling, three variables and I need help as I got clear results for ADF test on level and first difference but while applying cointegration I got the message as; singular matrix but when i click the option for no intercept and trend, it gives me 1 cointegration equation.
I am using Eviews 4 and am so worried as I am on last stage of my thesis now I cannot change the topic.
Reply me soon

Thanks

[sanda]

Cointegration refers to a scenario where linear combination of nonstationary variables is stationary. For these non-stationary time series variables, there is a possibility of estimation by differencing in cases where the differences are stationary as already mentioned above under stationarity and unit root tests. For estimation of the co-integrating relationship to be undertaken, it requires that all the time series variables in the model be integrated of order one I(1) . The next step after recognizing the order of integration of the variables as I (1) or above is to test whether the variables in question can co-integrate or not.

There are two common methods for testing co-integration and estimating the relationship among cointegrated variables namely the Engle-Granger (1987) Two Step Procedure and Johansen’s (1988) maximum likelihood method. In the Engle-Granger two-step procedure, variables entering the co-integrating vector are tested for integration of the order, I (1). Thus, the first step in this procedure is pre-testing the variables for their order of integration. The second step is estimating the long-run equation relationship and obtaining the residual. The third step is testing whether the residual is stationary. If the residual is stationary, then the variables are said to be co-integrated, i.e., they do have long run relationship. The final step is estimation of the error correction model (ECM) including the lagged value of the residual as the explanatory variable. The ECM model is estimated to see the short run relationship between the variables.
The Johansen maximum likelihood method is an alternative to the Engle-Granger Two Step Procedure. This procedure is a multivariate generalization of the Dickey-Fuller test

Regards

[sanda]

I would like to know, after performing a unit root test then you find that two out of five independent variables are integrated of order I(1) while the rest including the dependent variable are stationery at level i.e I(0).do you proceed to test for co integration and if so what is the procedure?.how do you then test for the long run and the short run relationship?

[economistpk]

Thank you so much for explaining in detail, but my problem is still there.

I am using both Engle Granger and Johansons, but while applying cointegration test I got message of singular matrix, but when I omit the trend means click the option for no deterministic trend it gives 1 cointegrating equation, what to do now in Eviews 4. I have three variables, means 2 independent and 30 years data,
any one please help ´for this.

Thanks

[economistpk]

also For unit root, I have got the correct results like at level all the 3 variables are non-stationay and at level they are stationary, means the variables are integrated of order 1, now the cointegration technique can be applied as all the series are integrated of order 1.

[economistpk]

sorry I mistakenly wrote "at level" instead of writting "at first difference", hope you understand, at first difference the series is stationary.