Different R2 in EViews and Stata (Panel data)

[mofar]

Hello EViews experts,

I have faced a puzzle when I estimate panel data fixed country and time effects in EViews (v.7.2). While there is no significant different between point estimates in EViews and Stata ( for exactly the same specification, same sample period and number of cross section), but surprisingly I get very high R2 in EViews. Stata gives 3 different R2(within, overall, and between) but non of these R2 in Stata is similar to calculated R2 in EViews. I have seen this many times when I do panel country and time fixed effects. Is there any reason behind these different R2 squares? I appreciate any assistance to solve this issue.

Many thanks and best regards,
Reza

[EViews Glenn]

The R2 for EViews is against a fully restricted model with common intercept which differs from the variants that Stata provides (for example, the within is against a model with only fixed effects). Arguably some of the measures that they provide may be more interesting, but they require additional computation which can be costly in nonlinear specifications, and we've never thought it that important (we're not big fans of R2 in general). I'll put it on the list of things for us to consider offering...

[mofar]

Many thanks for your feedback. I am still not completely sure what to do with R2 when I report fixed effects results. As I mentioned I get R2 of above 95% for 2 way fixed effects while in Stata its around 50% or so. What is your suggestion? Can we report EViews results while reporting R2 from Stata? I sometimes prefer estimating panel in EViews because of more standard error robust options in it.

[EViews Glenn]

Provided they are the same specification and same results, the Stata R2 would be the one that we would compute (were we to do so).

[priyo]

hi Reza and Eviews' Expert

Now i am doing my paper and facing the same problem with you related to the different result of R-squared among EViews dan Stata. Exactly the same problem with yours, eviews gives extremely higher R2 compare to Stata's. And for sure, my Supervisor doesn't happy with this very high R2 for panel.
Do you already get the answer for this typical problem.
looking forward your good news.

Thanks,
priyo

[EViews Gareth]

Ask him which computation for R^2 he would like you to use.

[EViews Glenn]

It's pretty straightforward to compute whatever measure you wish based on the results for two different equations.