Cointegration, VECM and Impulse Response Function

[upani]

Hi,

I am using RATS 8.3 and CATS 2.0. I want to estimate the Cointegration between two non-stationary series. I am facing problem of estimation. The problem follows.

1. I am not sure whether to estimate VECM in RATS or CATS.

2. I am not sure about how to conduct the weak exogeneity test in CATS. No bivariate example is given in the CATS instruction manual.

3. How to estimate the Impulse response function?

Please help me.

With regards,
Upananda

[TomDoan]

Hi,

I am using RATS 8.3 and CATS 2.0. I want to estimate the Cointegration between two non-stationary series. I am facing problem of estimation. The problem follows.

1. I am not sure whether to estimate VECM in RATS or CATS.

With just two variables, it's probably simplest to do that in RATS. But you can certainly use CATS to examine the model carefully before you do the final estimates in RATs.

2. I am not sure about how to conduct the weak exogeneity test in CATS. No bivariate example is given in the CATS instruction manual.

Why does that matter? Do the test for weak exogeneity and look at the rank=1 line (which will be the only one of interest). The CATS manual focuses on a larger number of series because that's where CATS is most useful---when you have more than one cointegrating vector so you have to figure out a useful set of restrictions and normalizations to make it possible to interpret the cointegration space.

3. How to estimate the Impulse response function?

ECT.RPF does FEVD's. Impulse responses would be done the same way but using the IMPULSE instruction instead.

[upani]

Dear Sir,

Thanks for your reply. I got an idea. Would you suggest me from which resources, i should read how do it in case of RATS. I want to know what is the correct procedure?

With sincere regards,
Upananda

[TomDoan]

The section on "Cointegration and Error Correction Models" in the User's Guide. Once you've defined a model with a ECT instruction, ERRORS, IMPULSE and most of the procedures work the same way as they do for a non-cointegrated VAR.

[sanjeev]

Hi Tom,
Could you please clarify if ARDL approach to cointegration can be applied if our main variable of interest(dependent variable) is I(0) while other variables are a mix of I(1) and I(0) variables.
Please reply soon!

Thanks.
Regards.

[TomDoan]

If there's an I(1) on the right-side and an I(0) dependent variable, then you might as well put the I(1) in as differences. Other than that, there's no problem.