Listers
Is there a good reference for how Stata implements the heckman model for the
common example of wages for union vs. nonunion members?
I have tried to review the (R)heckman and each example seems to demonstrate
the censored or truncated distribution aspect (that of a female's
reservation wage and there are only wage observations for those that work).
There is also a good FAQ on endogeneity and self-selection but I think
heckman should still work.
I have tried the following:
Assume the following model:
Eq. (1) y1 = y2 + x1 + x2 + u
y1 is a continuous variable
y2 is a dummy variable (i.e. union)
There is a question that y1 & y2 are endogenous. y1 may cause y2 and y2 may
cause y1.
Accounting for this, the system is now
Eq. (1) y1 = y2 + x1 + x2 + u
Eq. (2) y2 = y1 + v
It was suggested that I use a heckman 2-step to see if y2 is endogenous by
interpreting y2 as a "choice" variable instead of as a selection bias. (I
interpret the selection bias question to be where the number of observations
on y1--the dependent in equation (1)-- are fewer than those for all the
exogenous x's.
When I try to run:
heckman y1 x1 x2 (y2 x3 x4) twostep
I get the error:
Dependent variable never censored due to selection:
model would simplify to OLS regression
This I guess makes sense because I am not 'selecting' anything, that is, I
have observations for all variables.
Is this an inappropriate application of Stata's heckman commands?
Or should I run equation(2) independently, gather the predicted values and
use in place of the dummy y2 in equation (1) or try treatreg as the FAQ
suggests?
Thanks for help - SF
Steve Fraser
Univ of South Florida
(813) 974-9680
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
