System Specification Identification Issue

[Szi]

I am trying to estimate a VECM in Eviews 6 using a system specification. I started with the automatic VECM estimation (“quick…estimate VAR”), but as this wouldn't allow me to do coefficient tests involving both the cointegrating relation and the error correction parts, I had to then estimate the model using a system.

:series: Please see the attached Workfile

In this system specification there are 5 variables – so 5 equations each with 3 cointegrating relations. I have removed the intercept terms from the cointegrating relations for identification purposes. I have been trying to estimate this using Eviews’ Full Information Maximum Likelihood (FIML) option, but keep getting this error message:

WARNING: Singular covariance - coefficients are not unique

It seems to me that the system should be able to be estimated (all the coefficients identified), but Eviews doesn't seem to be able to.

Is this an issue with Eviews, or is there something wrong with my system specification?

[trubador]

You cannot estimate Vector Error Correction models with System Estimation Methods, since they work differently than that of VAR models. You have to identify the number and the form of long-run cointegrating relationships first. Otherwise, system estimation techniques will fail to differentiate the multiplication of two constants (e.g. c(1)*c(2)). Moreover, you have only 48 observations and you will quickly run out of degrees-of-freedom, if you estimate 5-variable VEC model.

[Szi]

Ok thanks,
So if I want to perform a wald test on a number of VECM coefficients (including the cointegration terms) - how can I do that in Eviews? It doesnt let you in the VECM specification, and the system specification won't work..
:(