Announcement

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

  • Including lags of exogenous variables in xtabond2

    I am implementing system GMM using the xtabond2 command. I would like to generate instrumental variables from lags of an exogenous variable. Therefore, rather than introduce the instruments with the option ivstyle, I am considering using gmmstyle with laglimits(0 .) - to capture both the current value as well as lags. As a test I have used laglimits(0 0) and collapse to compare with ivstyle. The results were similar but not identical to ivstyle. Just checking if anyone else has introduced lags this way or have a better suggestion. Thank you, Dan

  • #2
    Do you have unbalanced data? Then you might have to specify the suboption mz for ivstyle() to obtain the same results to gmmstyle with laglimits(0 0).

    Another reason for the discrepancy might be if you did not specify the equation() suboption to ivstyle().

    If you are considering the instruments for the first-differenced model, also note that gmmstyle() does not transform the instruments, while ivstyle() by default applies a first-difference transformation to the instruments unless the suboption passthru is specified.

    More on GMM estimation of linear dynamic panel data models in Stata:
    https://twitter.com/Kripfganz

    Comment


    • #3
      Thank you very much Sebastian. That's great advice. I matched
      Code:
      ivstyle(var, mz equation(diff) passthru)
      with
      Code:
      gmmstyle(var, laglimits(0 0) collapse equation(diff))
      and then introduced lags for the exogenous variable via the laglimits().
      Thanks also for the very comprehensive materials. I'll work through your presentation - it'll enable me to better address the underlying issue of instrument proliferation I encountered in this query
      https://www.statalist.org/forums/for...abond2-command
      thanks again, Dan

      Comment

      Working...
      X