areg for random effect model or hausman test

[Yasutake Homma]

Hello Statalist,

I am analyzing bond level data with issuer/country/currency with monthly time window.

An issuer has issued multiple bonds in a month but I would like to do fixed effect model with issuers.
Thus, I use the following command:

areg yield x1 x2 i.country i.currency i.time, a(issuers)

I believe this is so called fixed effect model which should show the same result with xtreg fe command.
But I would like to confirm whether fixed effect or random effect model is better by hausman test.

I thought i cannot use xtreg fe or re in this situation because issuers have issued multiple bonds in a same month.
But I cannot come up with random effect model by using "areg" command.

If you could give me any solutions, it would be highly appreciated.

Regards;
HOMMA

[Carlo Lazzaro]

Yasutake:
the trivial issue is that -areg- does not support -re- specification (see -xtreg,re- for N>T panel datasets).
In addition, the -areg- entry in Stata .pdf manual explains why, despite they give back identical point estimates, -areg- and -xtreg,fe- do differ as far as their assumption about sample increase is concerned.
For differences between the two commands in estimating the -vce- matrix, see

http://cameron.econ.ucdavis.edu/research/Cameron_Miller_JHR_2015_February.pdf

.

[Yasutake Homma]

Thank you very much Carlo. I think I understood the explanation. I really appreciate it.