hausman test to compare OLS, 2SLS, and SUR

[Constantin Alba]

Dear statalist members,

I am trying to compare the efficiency of OLS vs. 2SLS vs SUR to justify the selection of methodology for my study. I know how to use hausman to compare fixed vs. random effects, but cannot find a way how to do comparison between the three methods mentioned above.

specifically, my problems is as following:

1. I have 2 model that I test using SUR -

sureg (eq 1) (eq 2), corr

- both equation at once, so no problems to store estimates

2. now, to compare SUR outcome with OLS (for example) I need to combine eq1 and eq2 into a system of equations for OLS somehow... I was think to do it using SEM (gsem) but not sure if that would be correct.

3. no idea how to do the comparison with 2sls. I cannot

4. extra question that maybe you can help me with - the error terms do correlate (Breusch-Pagan test), but all my regressors are the same in both equations. I read several various explanations of SUR and I could not get a clear answer whether in this case SUR should be more efficient than OLS.

thank you in advance

C.

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a helpful answer by following the FAQ on asking questions - provide Stata code in code deimiters, Stata output, and sample data. You have make a lot of posts not to know this by now.

You don't use a hausman for this. If 2SLS indicates endogeneity, then you need that, 3SLS, or some other technique that allows for simultaneity. SUR will also test whether the errors are correlated.

[Constantin Alba]

You didn't get a quick answer. You'll increase your chances of a helpful answer by following the FAQ on asking questions - provide Stata code in code deimiters, Stata output, and sample data. You have make a lot of posts not to know this by now.

You don't use a hausman for this. If 2SLS indicates endogeneity, then you need that, 3SLS, or some other technique that allows for simultaneity. SUR will also test whether the errors are correlated.

Thank you for your response. There is no code/output to provide, as I am looking for a code or at least the direction s to which way/command would be appropriate here.

I would not usually use hausman for this purpose, but I came upon such usage in a recent article (Journal of Operations Management 2016) that applied hausman to compare effectiveness of OLS vs 2SLS vs SUR. As this is a top journal, I thought that this might an appropriate methodology. But I haven't found any information about it anywhere else, thus asked here...