EViews.com

Anyone knows how to specify and estimate a panel data model with variable coefficients (varying over cross sectional units) from a panel workfile?

I know that this can be done using a fixed or random coefficients model (as explained in chaper 6 of Hsiao, 2nd edition analysis of panel data) but I assume that it is not equivalent to the Eviews fixed or random effects options available for cross-sections or periods (functional forms with intercepts).

Any help would be more than welcome.

Thanks in advance.
MC

There is no built in way to do it. You'll have to generate the dummy variables manually.

Many thanks for the hint Gareth.

My problem is that though I know how to specify the dummy variables in the case that the coefficients of the dummy variables correspond to the intercepts (as demonstrated in chapter 15 of Using Eviews for Principles of Econometrics, Griffiths, Hill and Lim) I´m not sure on how to do it in the case that the dummies correspond to the n coefficients of the variables.
In my case I have 11 cross sections (or n individuals) and 2 independent variales. Any hint on how to do it?

Thanks in advance once more.
MC

Just multiply the variable by the dummy.

If I understand well, after specifying the dummies with "generate series by equation" using the cross section identifier, the equation specification shall look like:
dependent variable and 22 independent variables (11 dummies * independent variable 1 + 11 dummies multiplied by independent variable 2)
Is this correct?

thanks
MC

indeed.

many thanks Gareth!

By the way, if I want to specify a unrestricted model where the intercept and slope coefficients are allowed to vary across cross sections (individuals) then I will have:
dependent variable, 11 dummies, 22 independent variables (11 dummies * independent variable 1 + 11 dummies multiplied by independent variable 2)

Is this correct? (this unrestricted model is extremely useful to test poolability in panel data)

Thanks.
MC

Yep.

Although in a fully unrestricted model, you could also just have 11 different regressions, each with 3 variables.