Announcement

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

  • Dynamic panel approach for large T

    Hello everybody,

    I have data for a small number of countries (12) and a large number of observations for each of them (500).Therefore I am facing the problem to have a dataset with large T and small N. My right hand side variables are expected to be endogenous so I am looking for a dynamic approach to solve this issue.I would like to know which is the best procedure in stata to proceed (xtreg with fixed effects, xtabond...)

    Thanks in advance

  • #2
    If your right-hand side variables are endogenous, I recommend to use xtivreg (or xtivreg2) with a choice of suitable instruments. xtabond is not suitable for such large time dimensions. xtreg is not suitable for endogenous variables.
    https://twitter.com/Kripfganz

    Comment


    • #3
      Thank you for your answer Sebastian. So, if I use xtivreg what are the best instruments, if there is some tip for them I mean, is it better to use lags, variables in differences time dummies....?

      BTW, Have you heard about xtscc? Because I have read it works with large T, but I don't know what they mean with large ...
      ​​​​​​
      Regards

      Comment


      • #4
        Marcos:
        I would follow Sebastian's helpful advice.
        As an aside, I do not think that the community-contributed command -xtscc- (-search xtscc-) can deal with endogeneity.
        Kind regards,
        Carlo
        (Stata 18.0 SE)

        Comment


        • #5
          Thank you professor Lazarro. So xtivreg works with large T and addresses endogeneity. Then my question would be about which are usually the best instruments to use (lags, differences, time dummies ...? And another question, if I run different models changing some variables for robustness,, can I use different instruments in different models? (I mean, as long as the over and underidentification tests are OK, the instruments are allright or the instruments should be exactly the same in all models? )

          Thank you again for your help

          Regards

          Comment


          • #6
            Marcos:
            usually, the choice of instruments (when available) is led by the literature in your research field (and the same holds for robustness, reminding that different models give different results).
            As an aside, please call me Carlo, just like all on (and many more off) the list do. Thanks.
            Kind regards,
            Carlo
            (Stata 18.0 SE)

            Comment

            Working...
            X