Panel 2SLS EViews and Stata

[berten]

I use panel two-stage least squares estimation with period and regional fixed effects and White period standard errors in EViews. Now I try to get the same results in Stata, and although it seems very simple, I don't get the same results for standard errors. The closest results are given by:

ivregress 2sls mp6 rgdpgr lrgdppc2 icrgave2 r1 r2 r3 yr* (L.kcont2 = L.bit_us L.eu), vce(cluster ifscode)

The coefficients are the same, but the standard error for kcont2 is 0.324050 in EViews, and .3189183 in Stata. They are close enough perhaps, but I wonder why they differ, and whether there is an option that would make them the same.

Thanks for your attention.

[EViews Gareth]

Probably a degree of freedom correction difference.

Divide the SEs in one package by those from the other package. If the ratio is constant across all variables, it is almost certainly DoF.

[berten]

Thanks for your reply. The ratios are fairly close to each other, but not exactly the same. They differ after 4th decimal number, e.g. 0.984163864 and 0.984136258. When I unselect the degrees of freedom correction, the SEs become a lot more similar, and the ratio becomes 0.999999059320208. But it's still not 1. I wonder if there is something else going on here. Could you say that this is all normal?
A different question is whether to use degrees of freedom correction or not. Why does Stata's current setting exclude such correction, while EViews does it?

[EViews Gareth]

Yeah, I'd say that's just degree of freedom correction.