The RATS Software Forum

Dear Rats Users,

In order to test causality between three series in a panel of several countries and 30 years, after testing cointegration (with pancoin.src) and using panelfm.src to estimate by FMOLS a existing long-run relationship, I have to construct a panel vector error correction model for which the error correction term is needed. Panelfm.src gives the estimated coefficients of the slopes and the corresponding t-statistics, however, I do not know how to get the estimated coefficient for the intercept and what is more important, I cannot recover the residuals of the estimation to have the error correction term.
Is there any way to recover those residuals?

Another question: How can I implement System GMM (Blundell and Bond, 1998) in a panel VAR or VECM?

Thanks a lot.

I can answer his first question, but still wonder about the second.

As for the intercepts, they can be seen with the command
display %%ibetas

But I also wonder how to recover the residuals. When I run the TABLE instruction after doing @panelfm, it shows me a "U" series with a zero mean and the right number of observations. But I haven't been able to figure out how to recover it as a panel data object. Thanks, JS

The most current version of @PANELFM computes %RESIDS as the residuals.

Hi Tom,

I want to estimate an ECM. Firstly, since my variables are I(1) I did estimate the cointegration relationship using DOLS. After that, I am trying to save the residuals from this regression in order to construct an ECM. I compute them using %RESIDS -as you suggested- but I cannot see them in the series window, and therefore I cannot use them to compute the ECM. I might be doing something wrong but unfortunately I cannot see where is the problem. Please see below the codes I am using:

@paneldols(dlags=5,lags=5,bvec=||1.0||,print=full,average=simple,$
ibetas=dolsbeta,itstats=dolststat)
# logfund logindex
COMPUTE Residuals=%RESIDS

Thank you very much in advance!

C.

%RESIDS isn't included in many of the series selectors because it can change easily (and perhaps unexpectedly), and so isn't a good choice for inclusion directly in a regression. To make a copy, use

SET residuals = %resids

not COMPUTE.

Hi Tom,

Thank you for your help. The problem is that I need the residuals from either FMOLS or DOLS in order to compute an ECM for each individual in my dataset. The ECM will be computed with the differences of the variables from the long-run regression but I need to include the lagged residuals from FMOLS or DOLS. I wonder if you see a different way to do the ECM.

Thank you in advance.

C

I'm not sure I understand what you're asking. %RESIDS is the series of residuals from the group mean estimate of the cointegrating vector. Just copy it over to a new series using SET (not COMPUTE the way you were trying) and use the series that you create.

Hi Tom,

It works now! Thank you very much for your help.

C.