Panel data dummy variables are being omitted

[Femke Versluis]

I tried to do a regression on the overestimation of companies, including two dummy variables: the sector that could be between 0 and 5 and the region that could be either 0 or 1. But when I try to regress this the dummy variables are being omitted. And I can't seem to figure out how I can fix it. Anyone that could help?

[William Lisowski]

Welcome to Statalist.

Because you have specified a fixed effects regression, any variable which does not vary over time within a company will be will be omitted.

[Clyde Schechter]

Adding to William Lisowski's spot-on diagnosis of the situation, at this point you need to think about why you wanted to include these variables in the model in the first place. There are two possiblities:

1. Your goal is to estimate the effects of these variables on your outcome variable (age). If so, you must abandon fixed-effects estimation. Consider using -xtreg, re- or -xtreg, be- instead.

2. Your goal is to adjust ("control") for their effects on the outcome, but they are not of direct interest. In that case, you have no problem and can just leave things as they are. Because these are fixed attributes of the company, their effects are automatically adjusted for in the fixed-effects model without having to be explicitly included.

[Carlo Lazzaro]

Femke:
just two sidelights to previous excellent advice:
- your T dimension in not negligible. Hence I would also consider -xtregar-;
- I would also check whether your model is well specified (ie, does it need more predictors?);
- eventually (but this is probably a copy and paste mishap) your dependent variable is -age-, not -overestimation-.

[Femke Versluis]

Thank you all for helping me out with what I did wrong. But sadly there's still a lot I don't understand. I was wondering if you could also help me with testing the endogeneity. I wanted to perform a Durbin–Wu–Hausman test and found that I needed the xtivreg since I am using panel data. But when I use this I get this:

If I don't use the xtivreg but the ivreg, I get this:

which I don't think is meant to happen, because if I include the dummy variables as i.sector and i.region I get that factor-variable operators are not allowed.
Could anyone help me with this problem?

[Carlo Lazzaro]

Femke:
the issue seems to be that you did not code endogenous regressor(s) and instrument(s) properly.
See -xtivreg- entry in Stata .pdf manual for further details.

[Rabia Ibrar]

Dear Carlo Lazzaro and Clyde Schechter i am using pooled OLS and two way fixed effect for net oil exporters and then i divide the countries into top net oil exporters and remaining. i created a dummy variable for this. D=1 for top net oil exporters and D=0 otherwise, but when i run the regression then there is no issue in pooled OLS but two way fixed effect gave me omitted result for this dummy variable. Can any one help me regarding this. i used this command xtreg ihs_GDPGrowthannual ihs_Realoilspotprice ihs_percapoilcons countrydummy yr2-yr26 , fe vce (cluster C_id)

[Carlo Lazzaro]

Rabia:
if your variable are time-invariant, the -fe- estimator will wipe them out (as expected).
On addition, please take a look at the FAQ to post more effectively. Thanks.