Announcement

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

  • GMM Simultaneous linear equations and not concavity

    Dear Statalisters,

    I am having some issues using gmm to estimate two linear equations. My understanding is in such a setup the objective of the minimization is quadratic in the coefficients that I am trying to estimate giving me a fairly trivial concave problem. The fact is that Stata complains that the problem is not concave and this causes the optimization to be extremely slow and, in practice, unfeasible.

    Another thing that might help detect the problem is that if I do not force the weighting matrix to be identity (ie use the default option for the initial matrix) I get a non-invertibility problem.

    In particular, I am trying to estimate a model which is very close to that in Chiappori et al. (2002) - "Marriage Market, Divorce Legislation, and Household Labor Supply". This is the code I am running

    Code:
    gmm (hours_tot_sp -{wife:lwage_sp lwage lwage_inter mf_tot school_sp age_sp} - {const} ) ///
        (hours_tot -{husb:lwage lwage_sp lwage_inter mf_tot school age} -{const2}) if married==1 & year==2010 & child==0 & child5==0, winitial(identity) ///
        instruments(1: lwage mf_tot school_sp school2_sp age_sp age2_sp i.size_dummy) ///
        instruments(2:  lwage_sp mf_tot school school2 age age2 i.size_dummy) ///
        derivative(1/wife = -1) derivative(1/const = -1) ///
        derivative(2/husb = -1) derivative(2/const2 = -1)
    I am pretty sure I am making some trivial mistake.

    Thanks in advance for your help.

    Best,
    Davide
Working...
X