Announcement

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

  • 2SLS with ivreg2

    Dear all,

    I would like to perform two stage least square (2SLS) using ivreg2.
    My baseline model is as follows:

    Code:
    yt= xt-1 + controlst
    Given that my endogenous x-variable is lagged for one year (t-1), it seems illogical in the first stage to have control variables and instrumental variables (IVs) at period t to predict endogenous x-variable at period t-1. See below:

    Code:
    xt-1 =  controlst + IVt + Ɛ


    Hence, I thought of regressing endogenous x-variable at period t-1 on a set of control variables and IVs at period t-1 in the first stage of the 2SLS.

    Code:
    xt-1 =  controlst-1 + IVt-1 + Ɛ


    I then use the predicted value of lagged x-variable at period t-1 from the first stage to my second stage of 2SLS as follows (this is what I would like to achieve):

    Code:
    yt= x^t-1 + controlst
    I am not sure whether it is possible to do so in ivreg2 since the command restricts the ability to modify the model. If the command does not support such function, mind suggesting any other possible ways of doing it? I need to also report the partial R-square, partial F-value, over-identifying restriction test and Hausman test.
    In addition, please advise whether it is right to do so econometric-ally.

    My apologies if it is not clear or if it has been posted elsewhere (I cant find one), but please feel free to let me know.
    My sincere and advance thank you for your time.
    Regards,
    Ricky
    Last edited by Ricky Liu; 25 Sep 2018, 01:27.

  • #2
    I think ivregress lets you use different instruments without including all the other variables from the main model.

    Comment


    • #3
      Hi Phil. Thanks for the reply.
      As per ivreg2, ivregress doesn't allow me to have different set of control variables in the first stage.

      Code:
      ​ivregress estimator depvar varlist1 (varlist2 = varlistiv) if   in  weight, options

      Comment

      Working...
      X