Fixed Effects Least Squares

[lightpoetry]

I'm working on a project where I analyze the impact of the change in per capita income on gubernatorial and presidential election results from 1970 to 2004. Because it's panel data, I'm using a fixed effects least squares model, where each year and each county has its own dummy variable. The problem is that there are roughly 3000 counties in the United States.

My equation is as follows:

Vote_percentage_by_county = αIncumbency + βper_capita_change*incumbency +County Fixed Effects + Year Fixed Effects.

I get a near singular matrix error, that the regressors may be perfectly collinear if I use more than 7 of the 9 year dummy variables. I get the same error when trying to include N-1 (N=number of counties) of the county level dummy variables.

I'm not sure what to do to fix this. Any help would be very much appreciated.

[startz]

Check to see if you have a bunch of NAs in per_capita_change. Past that, you might post your workfile including the equation you're trying to estimate.

[lightpoetry]

The equation is

ls vote_per_dem c incumbency per_capita_change*incumbency yearx idx

where id represents the county id

[lightpoetry]

Not sure how to attach the eViews file. I've tried several times but it just resets to no file chosen.

[startz]

Here is the eViews file.

Not quite.

[lightpoetry]

Is there a limit to the size of the attachment? The eViews file is ~700 mb.

[startz]

There is a limit. Also, that's a very, very big EViews file for the data you're describing. You might want to go Options/General Options/Data storage, and be sure Use compression is turned on.