Announcement

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

  • System GMM in Stata

    Dear users,

    For my master's thesis I am trying to establish the most important determinants of inflation differentials in the EU-12 (12 countries) before and after the 2007 credit crisis. Following Licheron (2007), I'm using a dynamic panel model estimated by system-GMM in Stata. Having limited experience with Stata, I'm having difficulties with the xtdpsys commands in Stata.

    I first want to do the system-GMM estimations for the period 1999-2007 for the EU-12 countries. Based on the literature, my model includes the following variables:

    Dependent variable: infl (inflation)
    Independent variables:
    • Assumed to be endogenous: gap (output gap)
    • Assumed to be predetermined: l.infl (lagged inflation), fisc (fiscal position), prod_growth (productivity growth), l.ppp (lagged purchasing power parity)
    • Assumed to be exogenous: l.dneer (lagged change of nominal effective exchange rate)
    In a first attempt, following Licheron (2007), I want to include the endogenous variable (gap) lagged by 3 periods in the first-difference equation as an instrument and the once lagged first difference in the level equations. In addition, I would like to include the predetermined variables (l.infl, fisc, prod_growth and l.ppp) lagged by 2 periods in the first-difference equation as an instrument and the once lagged first difference in the level equations. Finally, I want to include exogenous variables (l.dneer) lagged by 1 period in the first-difference equation as an instrument and the once lagged first difference in the level equations.

    Moreover, I need to include time effects. My idea was to create time dummies for every year as below.

    Code:
    gen yr1999 = 0
    Code:
    replace yr1999 = 1 if year==1999
    Taking into consideration all of the above, what would my xtdpsys command look like? Something like this maybe...

    Code:
    xtdpdsys infl l.dneer yr1999-yr2007, end(gap, lag(3/3)) pre(l.infl fisc prod_growth l.ppp, lag (2/2)) twostep noconstant vce(robust)
    Having tried this, Stata tells me:
    factor variables and time-series operators not allowed
    r(101);

    I am aware that DP estimators are notorious for being very sensitive to model specification and instruments used, so I probably need to experiment with the amount of instruments used quite a bit. But I want to make sure I understand how to use the xtdpsys commands first (I read the help files on Stata.com).

    I included my excel file with data if it helps anyone to give me advice.

    Best regards,
    Sebastiaan





    Attached Files
Working...
X