Estimating a system of equations using pool data

[deich]

I have a pool set (104 time series (c12 variables); 13 cross-section). I have estimated a system of three equations (with coefficienct constraints) for one of the time series (using an ECM approach). I have estimated three single equations for the pooled data. Is it (a) possible, and (b) possible in EViews to estimate a system (with coefficient restrictions) using the pooled data? If so, how?

[EViews Gareth]

If you create a panel workfile from your pooled data, you can estimate a system of equations inside the panel, which should do what you want.

[deich]

I have had a look at the online manual and I am guessing that there has to be an easier way to create a panel workfile than the one I have been trying (procs; restructure/resize current page). I have the data in pool format (variablename_area) for quarterly data (XXXXqy). So I can easily estimate a single pool equation or a system for one of the areas. What is the easiest way to create the required panel workfile?

[EViews Gareth]

http://forums.eviews.com/viewtopic.php?f=7&t=74

[deich]

Excellent - that was very helpful and I managed to do it easily and quickly. I also managed to do a system estimation and to work out under which settings the result is the same as the single equations in a pool.

One question: in a pool estimation for a single equation, I can allow the cross section coefficients to vary - can I do the same in the panel and, if so, how?

[EViews Gareth]

Cross-section specific coefficients can't be done automatically in a panel:

http://forums.eviews.com/viewtopic.php?f=4&t=948

[deich]

Many thanks. I think that answers the question for a single equation in panel but I cannot see how to apply it to a SYSTEM of equation within a panel. My three equation system is:

dlnrrent=c(1)+c(2)*dlnrrent(-1)+c(3)*dlnrrets+c(4)*dlnsupply+c(5)*resr1(-1)+c(6)*avr(-1)
davr=c(11)+c(12)*davr(-1)+c(13)*dlnrrets+c(14)*dlnsupply+c(15)*resr1(-1)+c(11)/(c(1)/c(6))*avr(-1)
dlnsupply=c(21)+c(22)*dlnsupply(-1)+c(23)*resr1avg4(-12)

and I would like to vary the constants (in the first instance) in these equations. Is this possible?

[deich]

See my posting of 10 June - can anyone help?

[EViews Glenn]

Replace the single intercept coefficient with a set of dummy variables multiplied by coefficients.