Announcement

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

  • The use of instrumental variables in PPML

    Dear Joao Santos Silva and Tom Zylkin
    Suppose we have a case of endogeneous regressor, the index of legal enforcement of property laws, hereafter denoted as (PLi,t). Further, lets suppose that, we are interested in estimating the impact of PLi,t on export flows of country i to country j at time t (Xij,t). It is imperative to mention that the index of PLi,t is time-variant measure of countrys' strength of property laws where higher (lower) indicate stronger (weaker) incidence of laws governing property. We control for multilateral resistance terms (MRTs) using Baier and Bergstrand (2009) transformation that uses first‐order log‐linear Taylor series expansion around a symmetric trade cost world. In a panel data context (like in our case), we use the recommendation in Baltagi et al. (2015) that applies Baier and Bergstrand (2009) method to panel data using period specific transformations.
    However, due to endogeneity in PLi,t, the estimated parmateres based on PPML are biased and inconsistent. We use dummy variables for former British and French colonies as the instruments for PLi,t . That is:
    BRITISH = 1 if country i is a British colony
    =0, otherwise
    similarily
    FRENCH =1, if country i is a French colony
    =0, otherwise
    Many reserachers have argued that colonizers affect the formation and successive development of their colonies' institutions governing property laws (PLi,t). These two instruments (BRITISH and FRENCH) are highly correlated with PLi,t and affect export flows via PLi,t. It is to be noted that PLi,t is country-specific time-varying variable and its potential instruments BRITISH and FRENCH dummies are time-constant.

    MY QUESTION:

    Can we use the time-constant variables (dummies for former BRITISH and FRENCH colonies) as an instrument for time-varying index of property laws (PLi,t) and estimate the model using ivppml or ivpoisson in case of panel data?

    If yes, are there any further recommendations i should be cautious about.

    Please get back to me, I shall be very thankful to you.

    (I am sorry for posting this longish query)

    Thanks and regards,
    (Ridwan)

    References:

    Baier, S. L. and J.H. Bergstrand. (2009). Bonus vetus ols: A simple method for approximating international trade-cost effects using the gravity equation. Journal of International Economics, 77(1):77-85.

    Baltagi, B. H., P. H. Egger, and M. Pfaffermayr (2015). Panel data gravity models of international trade. In B.H. Baltagi (Ed.), The Oxford Handbook of Panel data, pp. 608-642. Oxford University Press
    Last edited by Ridwan Sheikh; 12 Sep 2023, 09:34. Reason: Accidently posted without completing

  • #2
    Dear Ridwan Sheikh,

    This falls well outside my comfort zone, but let me give my view since you are not getting a lot of help. The instruments you are using do not seem ideal, but I realize that it is difficult to find better ones. So, I suggest you estimate with and without instruments and see if the results are sensible. In any case, I would not make any strong statements based on such results.

    Best wishes,

    Joao

    Comment


    • #3
      Thanks Joao Santos Silva for replying back to this query. I did get a lot of help and interesting suggestions from Tom Zylkin on twitter (X) and i am really thankful to you all.
      I agree the instrument validity might be a little concern here, but prior literature does suggest some defence in using such an instrument (at least theoretically). The endogeneity is not that serious issue in my case, but again i argue that it may not be ironclad, that is why using the IV version of poisson estimator. It seems a nice sugegstion to esimate the model with and without instrument- Thanks.

      A related question I want to ask is that I tried running ivpoisson or ivppml on the dataset from your website, it was not executing

      Code:
      sysuse auto.dta
      ivppml price mpg foreign trunk, inst(weight displacement foreign trunk)
      This throws an error message - command ivppml is unrecognized
      I tried installing from SSC but did'nt find the packages (ivpoisson and ivppml) on SSC. I am using STATA version 14.2. Can you answer why that may be a case?

      Thanks and regards,
      (Ridwan)

      Comment


      • #4
        Dear Ridwan Sheikh,

        The ivppml command was never released because it is still a bit rough. To run it, you have to download the ado file from my page and put it somewhere where Stata can find it.

        Best wishes,

        Joao

        Comment


        • #5
          Thank you Joao Santos Silva. Thanks for the suggestion, I will do that.

          Thannks and regards,
          (Ridwan)

          Comment

          Working...
          X