Alarmingly high R-squared value in Panel Data regression

[Antonb0]

Hi all,

I would need some suggestions for solving a problem of mine:

Me and my thesis partner currently writing our master's thesis and using Eviews 8 to conduct tests. We've got 1010 observations for 181 companies during 10 years, the dataset is unbalanced panel data. We're using the Fixed Effects model, with fixed cross section effects specification, with White's heteroskedastic robust standard errors. In our estimates, we get an alarmingly high R-squared, somewhere between 0,8-0,9. The litterature is not the best about this topic, but from what I've read is that many researchers working with panel data are happy if they get a R-squared above 0,1, which makes this even more unsettling. We've checked for near multicollinearity between the independent variables, and we haven't got a value over 0,65.

So to summarize what I am wondering about:
1. What are the possible reasons for this high R-squared value? Should we be worried or just leave it?
2. Is using the fixed cross section effects specification the right way to go, or should we use fixed period instead?

Might be wise to mention that we have brief experience when it comes to econometrics (one course during this fall for our master's).

Any help will be appreciated!

[EViews Glenn]

The EViews R2 compares to a model with a fixed intercept. Alternative definitions compare to a model with a full set of dummy variables. You can compute this R2 by estimating another model that just has the fixed effects (put C as the regressor, and use the fixed effects; then take the SSR and compute in the usual fashion).