Announcement

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

  • Choice of covariance matrix, system GMM, xtabond2

    Hello,
    I am running system GMM on dynamic panel data. My question is what estimate for covariance matrix (h) should I prefer. H(1) assumes homoskedasticity, which is not present in my data. So, I am between h(2) and h(3) which give quite different results.
    Thank you!

  • #2
    You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

    You might also consider xtdpdml (user written) or xtivreg2. If you're using stabond2, you should find Roodman's paper on it. I don't know xtabond2 well enough to layout the reasons to use one or the other. Obviously, it is troubling if the results are extremely sensitive to this assumption.

    Comment


    • #3
      Asymptotically, the choice does not matter if you are using the twostep estimator. In finite samples, however, the choice might matter. There is no clear guidance about which matrix performs better.
      https://twitter.com/Kripfganz

      Comment

      Working...
      X