[Limdep Nlogit List] Latent class Frontier Problems
Noth, Felix
noth at finance.uni-frankfurt.de
Wed Aug 26 19:06:19 EST 2009
Dear Limdep-Users,
I am trying to fit a latent class stochastic frontier production function by using model=BC and explain production regime member ship as well as TE.
FRONTIER;lhs=lnY1;rhs=one,lnk1,lnl,lnKK,lnLL,lnKL,t,tsqu,lnkt,lnlt;Eff=u;wts=y2;rob;rh2=one,a1,a2,a3,t,tsqu;pds=ni;model=BC;$
FRONTIER;lhs=lnY1;rhs=one,lnk1,lnl,lnKK,lnLL,lnKL,t,tsqu,lnkt,lnlt;Eff=u;wts=y2;rob;rh2=one,a1,a2,a3,t,tsqu;pds=ni;model=BC;
LCM=one,b1,b2,b3;Pts=3;Group=RGM;CProb=RMP;maxit=200$
create;TE=exp(-u)$
dstat;rhs=TE$
Estimates look nice but I am not able to derive serious levels of TE.
I get either almost 100% technical efficient firms like
===============================================================================
Variable Mean Std.Dev. Minimum Maximum Cases Missing
===============================================================================
------------------------------------------------------------------------------
All observations in current sample
------------------------------------------------------------------------------
TE | .150860E-02 .385925E-01 .514820-130 .999841 37958 0
or no estimates of TE at all.
===============================================================================
Variable Mean Std.Dev. Minimum Maximum Cases Missing
===============================================================================
------------------------------------------------------------------------------
All observations in current sample
------------------------------------------------------------------------------
TE | No valid information found for this variable
I think that perhaps Iteration don't get far enough for plausible starting values and therefore firms don't "leave" the top-efficiency level, but the problem does not change by using e.g. maxit=500 or higher!
Any suggestions?
Thanks.
Felix
More information about the Limdep
mailing list