st: Estimating who selects into my sample?

[Jen Zhen <jenzhen99@gmail.com>]

Dear list members,

I have the following econometric problem:

There is a drug which policy makers thought should not be given to
those aged below 20, while it's perfectly fine for those aged above
20.

Yet some, albeit not all, sub-20s used to get it. Then there was a
policy change punishing sub-20s for getting the drug and sellers for
selling it to sub-20s.

I observe who demands the drug and who of those demanding it does get
it. For each demander I observe whether she is aged below or above 20.

 Now I would like to evaluate the effects of the punishment
introduction on the demand for and supply of the drug. My first shot
was to take a dummy for whether an applicant is aged below 20, then
regress that a dummy for the post-punishment-period plus a range of
controls.

I get a constant of 0.15 and a coefficient of interest of -0.10 and
interpreted this as saying: Before the introduction of the punishments
15% of applicants were aged below 20. Afterward only 5% were.

I'm not sure though whether I should worry about sample selection
since I always observe only those demanding the drug but not those not
demanding it. On the other hand I'm not analyzing some other outcome
for a sample subject to that selection, but am analyzing that
selection itself.

So I'm wondering whether the above approach would still be valid. If
not, how else could I analyze the impact of the policy change?

Thanks so much for your input and best regards,

JZ
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/