Dear all,
I’m currently trying to
estimate an application of the permanent income hypothesis for a household survey.
Mi specification runs as follows:
(Ct+1 – Ct)= B1*(Yt+1 –
Yt) + B2*X �
where X stands as a vector of
household characteristics.
In addition I have to sub samples,
beneficiaries of government programs (BEN) and non beneficiaries (NBEN). I’m
specifically interested in identifying the different B’s for both sub
samples, which would allow me to find the impact of government programs on the
capability of households to smooth consumption. So I ran a regression with the
following specification:
(Ct+1 – Ct)= B1*(Yt+1 –
Yt)*BEN + B2*(Yt+1 – Yt)*(1-BEN) + B3*X �
where BEN is a binary variable
that takes the value of 1 if households are beneficiaries and 0 if they are
not.
However, I’m worried that
my results are not correct since program beneficiaries have different characteristics
than non beneficiaries, and that would turn into a selection bias; also, that the
BEN variable could be endogenous to (Ct+1 – Ct); more vulnerable household
will apply to government programs more frequently.
So, my question is, is there a
way to fix my estimates in order to identify the impact of state programs on
B1?
I was thinking of a variation of
the heckman selection model that allowed me to correct for the selection bias
of program beneficiaries as well as the bias for �non beneficiares. Is this
possible? Or what would the appropriate procedure be?
Best Regards,
__________________________________________________
Andr�s Moya
Facultad de Econom�a
Universidad de Los Andes
Bogot�, Colombia
