Hausman test solution for vce(cluster) options

[Salvatore Greco]

Hello Stata Community,
I am using panel data and I want to execute a Hausman test to test Fixed-Effects and Random-Effects. I made the two estimates using the command xtreg with the option vce(cluster id).
I have a problem with the command Hausman fixed random. Baically, it says I cannot run the test if I had previously used xtreg with those options. In particular, this is what Stata shows me.

. hausman fixed random
hausman cannot be used with vce(robust), vce(cluster cvar), or p-weighted data
r(198);

I searched on the forum and the command xtoverid, cluster(idcode) should be the command I may consider.
I installed it and ran it but it does not work. This is Stata message.

. xtoverid, cluster(id)
Error - must have ivreg2/ivreg29/ivreg28 version 2.1.15 or greater installed
r(601);

How can I run a Hausman test yo choose between FE e RE if I am using clustered robust standard errors for my panel data analysis?

Thanks in advance.
Regards,

[Joro Kolev]

Install the commands that the user written command is asking you to install, fit your model with one of the suggested commands, and try again.

[Salvatore Greco]

Hello Joro, thanks for your reply:
Are you sure that ivreg2 can be used with Fixed-Effects and Random-Effects as in the command ivreg2 description I can read only

Extended instrumental variables/2SLS, GMM and AC/HAC, LIML and k-class regression

?
Thanks.
Could you please helping me answering this question?
Salvatore

[Joro Kolev]

Apparently you do not even need to use ivreg2. -xtoverid- should work after Stata Corp -xt- commands. Here is from the help file of -xtoverid-:

(Equivalence of xtoverid statistic and standard Hausman fixed-vs-random effects test)

. webuse abdata

(Balanced panel)
. xtreg n w k if year>=1978 & year<=1982, re

(Artificial regression overid test of fixed-vs-random effects)
. xtoverid

. di r(j)

. est store re

. xtreg n w k if year>=1978 & year<=1982, fe

. est store fe

(In homoskedastic balanced panel case, Hausman test using sigma from FE estimation...)
. hausman fe re, sigmaless

(... is numerically equal to the artificial regression overid statistic)
. di r(chi2)

(Artificial regression overid statistic readily extends to non-homoskedastic case)

. xtreg n w k, re cluster(id)

. xtoverid

[Carlo Lazzaro]

Salvatore:
you could also go Mundlak.
See explanation and worked out example in Stata blog (Fixed effect or random effect: the Mundlak approach).

[Salvatore Greco]

To Carlo and Joro:
Can you please help me work out how the xtoverid command works?
These are my steps:
1) I run the command xtreg, re cluster (id);
2) I run the command xtreg, fe cluster (id);
3) I run the command xtoverid;

Stata shows

. . xtoverid
xtoverid not compatible with xtreg model fe
r(198);

I do not understand why it says that sure I am not able to read the instructions properly in the help section.
Please let me know.
Regards

[Carlo Lazzaro]

Salvatore:
the reference case for the community-contributed module -xtoverid- is -re- (or, put differently, the null for -xtoverid- is that -re- is the way to go).
Therefore, the correct sequence is:

xtreg, re cluster (id)
xtoverid

If the -xtoverid- outcome reaches statistical significance, you have to switch from -re- to -fe-.