Announcement

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

  • PPML Gravity Coefficients

    Hello,

    I am running a ppml gravity estimation and would like to use the coefficients to calculate/predict trade.

    What "kind" of model do I use to calculate trade volume?

    Does a linear regression like y= b0 + b1*var1 + b2*var2 adequately represent ppml estimation? Or is ist another equation that I need to use for the coefficients?

    Thanks!

  • #2
    Dear Tobias Tan,

    A gravity equation is an exponential model, and that is what PPML estimates, not a linear model.
    ​​​​​
    Best wishes,

    Joao

    Comment


    • #3
      Dear Prof. Santos Silva,

      thank you for the quick answer and the clarification. I thought that because of the log-linearization I could then approximate trade flows by using the simple linear relationship and the coefficients.

      My aim is to use the coefficients I get from the ppml estimation and to set up an equation where I can calculate trade flows (and then compare them to "real" trade flows).

      Thus my question is whether I can estimate trade flows with my PPML coefficients I just by concatenating the term (e.g. ß0 + ß1*var1+...) in order to approximate a trade flow or whether I need another kind of equation (not linear)?

      Best wishes,
      Tobias

      Comment


      • #4
        Dear Tobias Tan,

        A gravity equation is an exponential model, so you need to take the exponential of the linear index you mention above. You can use the predict command to do that; make sure you read the help and choose the right option; for example, predict with the option xb will generally give you the linear index that you have to exponentiate, but there will be an option that will give you directly the exponentiated values. Depending on the command you use for estimation, these options vary, so make sure you use the right one.
        ​​​​​
        Best wishes,

        ​​​​​​​Joao

        Comment


        • #5
          Dear Prof. Santos Silva,

          many thanks for the explanation!

          Best,
          Tobias

          Comment

          Working...
          X