Announcement

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

  • Using lagged instrumental variables for lagged endogenous variables

    Dear Statalist members,

    I am trying to use IV model with panel data (xtivreg) along with fixed effects (fe) and clustered errors. Thus, the sample Stata code is as following:


    (1) xtivreg Y (X = Z), fe vce(cluster ID) first


    Now, I want to check the lagged effects of the endogenous variable, L.X, on the dependent variable, Y, and also want to control for the endogeneity of the lagged endogenous variables with the lagged instrumental variables (L.Z). For example,


    (2) Y = _const + _beta * X + _beta * L1.X + _beta * L2.X + .... + _beta * Ln.X + e

    where

    X = Z and
    L1.X = L1.Z
    L2.X = L2 Z
    .....
    Ln.X = Ln.Z


    I thought about using 3SLS and other GMM-based dynamic linear panel regression models (xtabond, xtabond2, etc), but they focus more on including the lagged dependent variables into the regression model as an instrumental variable. Are there any other clear ways to use the lagged endogenous variables and their lagged endogenous variables in panel data setting?


    Many thanks in advance,

    Moon

  • #2
    Eagerly waiting for a response!!

    Comment


    • #3
      Eagerly waiting for a response!! Please!!

      Comment


      • #4
        You'll increase your chances of a useful answer by following the FAQ on asking questions.

        If you just want lagged x, then you can simply create lagged x and instrument it too:

        g Lx=L.x
        xtivreg y (x Lx=z L.z)

        While this does use z to instrument Lx (as well as L.z), as long as they're exogenous it is probably OK. User written xtivreg2 includes many diagnostics you might find helpful. Whether laggex x is endogenous is worth testing as is the exogeneity of the instruments.

        Comment

        Working...
        X