"near singular matrix" when add dummy to fixed effects model

[treefarm]

Hello,

I am estimating a panel (balanced) least squares model with cross section fixed effects. My cross sections refer to census divisions and I would like to include a dummy variable that is equal to 1 for northern census divisions and 0 for all other census divisions. However, when I do this I get a "near singular matrix" error. I can include the dummy variable, but only if I turn off the fixed effects. Can anyone explain why this is?

Thanks

[EViews Gareth]

http://forums.eviews.com/search.php?key ... mit=Search

[treefarm]

Sorry, still a little confused...I thought the dummy variable trap was when your dummy variables sum to 1 (i.e. are mutually exclusive), as would be the case with cross section fixed effects unless one of the cross section dummies is removed (I understand EViews does this automatically), or when you have a dummy for each of the four seasons or something like that. The dummy variable I added to my fixed effects model is 1 for multiple cross sections and 0 for multiple cross sections, so it overlaps with the fixed effects dummies and I don't see how it results in dummy variables that sum to 1.

[EViews Gareth]

"Near singular matrix" is caused, in 99.99% of cases, by the dummy variable trap. In panel estimation with fixed effects and dummy variables, 99.99% of dummy variable traps are caused by interaction between the dummies and the fixed effects dummies.