Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • xtivreg2 versus xtabond2 for GMM estimation using panel data

    Dear all,

    I am a bit confused, if I should use xtivreg2 or xtabond2 for estimation of my panel data model, where some of the explanatory variables are potentially endogenous.
    Both commands seem to handle GMM estimation of such a model. So, I am wondering what's the advantage of one over the other?

    Thanks for any input!

    Best, Jan

  • #2
    You can obtain identical results with xtivreg2 and xtabond2. See for example the following Stata thread:
    http://www.stata-journal.com/article.html?article=st0159

    xtabond2 has more options to customize the GMM estimator but it is advised that you make yourself familiar with the background on GMM estimation before you play around with them. In particular, you might want to read How to do xtabond2 by David Roodman.
    https://www.kripfganz.de/stata/

    Comment


    • #3
      Hi Sebastian,

      thanks for very helpful answer! Indeed, I already read the paper by Roodman and used xtabond2 for dynamic panel data estimation.
      Now, I want to estimate a static panel data model (because I have highly unbalanced data with a lot of missing values especially for the dependent variable),
      such that I thought that xtivreg2 might be more appropriate. I also want to employ fixed effects, which seems to be easier done with xtivreg2.
      However, I don't see any differentiation between difference and system GMM in xtivreg2. Do you know why this is the case?

      Best, Jan

      Comment


      • #4
        If you specify xtivreg2 with the fd option, you can obtain a fist-difference GMM estimator. I do not think that you can use xtivreg2 for system GMM estimation.
        https://www.kripfganz.de/stata/

        Comment


        • #5
          Ok, thanks again!

          Comment

          Working...
          X