st: Regressor with too many zeros

["joe jacob" <otharain@hotmail.com>]

Dear all,

I am estimating the effect of export orientation (exports/sales) by firms on 
their learning efforts from imported inputs. I use an -ivtobit- model due to 
the censored natured of the learning variable and the endogenous nature of 
the export intensity variable. There are a few other explanatory variables 
also, which capture the level of skill and technological capability of the 
firm. The problem is that there are a large number of non-exporting firms 
(hence a zero value for many observations of the export intensity variable). 
 My estimation did not produce a significant positive relation between 
export intensity and learning. BUT when I added an export dummy (1 for 
exporters and 0 for non-exporters) the variable became highly significant. 
(both export intensity and export dummy were instrumented) However, the 
export dummy became significant Negatively with a higher coefficient value. 
This latter result seems very counter-intuitive.

Put simply, the question is thus, when the key regressor has too many zeros 
(as my export intensity variable has), how do we tackle the problem of 
non-variability? Simply excluding zero-valued observations don't seem right?

Please comment.

Thanks,

Joe

_________________________________________________________________
The hottest things. The coolest deals. http://www.msn.co.in/Shopping/ Get 
them online!

*
*   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/