Perfect multicollinearity

[jekonomou]

Hi there

OK, I guess this will be a first...

I have been running some simple pooled OLS regressions where I have a variety of different dependent variables (financial risk) and the exact SAME specification of regressors on the right hand side (including country dummies, year dummies and industry dummies). Here's the "interesting" part: Regressions run fine overall and suddenly I get a non-singular matrix message when the dependent variable is firm beta. My question is, how is it possible to get perfect multicollinearity in the regressors according to what the regressand is??? It makes no sense at all.

The sample is the same at all times. Any ideas welcome as this is particularly perplexing for me. Many thanks.

[EViews Gareth]

Hard to say without seeing the workfile.

[jekonomou]

The file is about 50 MB and I don't seem to be able to upload it through "upload attachments". Is there a max size limit?

[EViews Gareth]

Try now.

[jekonomou]

Never mind Gareth, found the error. Very quirky and had to do with the data handling before uploading them on views.

Only constructive comment (which I'm sure you have heard plenty of times before): There has to be an update on eviews where dummies that lead to perfect multicol. are automatically dropped when the the regressions are run (as it happens with other software).

Many thanks for the immediate replies in any case and feel free to kill the thread.