Dummies for estimation of firm and industry effect

[eviewsnoob]

He everyone,

I could really use your help with the following, i am trying to estimate firm-effect, industry-effect and size effect on firm profitability. I have 256 firms in my sample and 123 observations (monthly), the 256 firms operate in 9 classified industries. For the firms en industries i need dummies. Size is the number of employees and doesnt need a dummy, therefore i still have a constant in my regression.

How do i calculate this in eviews 7.2? (i also have access to eviews 8 if thats easier)

The basic regression looks like : firm profitability = constant + firm *dummy + industry *dummy + size + errorterm

thanks in advance for the help.

Regards,

[EViews Gareth]

Depends on how your data is structured.

[eviewsnoob]

I am now working in eviews 8, and i added the workfile so you can see the structure of the data. ( i had to delete some of the variables i have, since the workfile is bigger than 2 MB)
I am not sure how to use the dummy function to generate the right dummies.

Basically i have a dataset of 256 firms with each 124 monthly observations and i structered it for each variable vertically and the colomn firmnumbers is to identify which company it is and the same principle for the industrynumbers. So the firmnumbers go from 1 to 256 and the industrynumbers differ (the possibilities are 10, 15, 20, 25, 30, 35, 45, 50, 55).

How would i for instance make an OLS regression with RoA as the dependent variable and the rest is pretty much the same idea as in my first post ( The basic regression looks like : firm profitability = constant + firm *dummy + industry *dummy + size + errorterm )

Thanks in advance for your help,
Regards,

[EViews Gareth]

Something like:

Code: Select all

roa c @expand(firmnumber, @droplast) @expand(industrynumber, @droplast) size

[eviewsnoob]

It gives a near singular error, regressors may be perfectly collinear.

And with the @droplast for both the firmnumber and the industrynumber, therefore going in to the constant, it is not possible to say anything anymore about the constant, or is there?

When doing ls roa c @expand(firmnumber, @droplast) size it is fine, also for only including the industrynumber part, but combining them, gives the error.

[startz]

If the firms never switch industries, then separate firm and industry effects can't be estimated.

[eviewsnoob]

Yeah i seem to have reached that same conclusion now.

I guess the firm and industry effects then have to be done with ANOVA testing or something like that, but not sure how to do that in eviews.