st: Differences-in-differences

[Bhavani Shankar <bshanx@gmail.com>]

Hi folks,

I'd be grateful for some advice, being a novice in econometric policy
evaluation. I have the following policy evaluation problem. I am
seeking to estimate the effects of a policy restricting advertising of
junk food to children in the UK that started in Jan 2008. I have
household food expenditure data for cross-sectional waves 2007 and
2008, ie before and after. Problem is, there is no geographically
defined control group available for D-in-D, as the program was rolled
out to the whole country. One potential strategy I have been
considering is as follows: use families with children as the treatment
group, and families without children as the control group, with the
dependent variable specifed as junk food expenditure per adult
equivalent. Do D-in-D estimation on this basis.

My question is simple: does this sound reasonable, or is there
something obviously wrong with this train of thought? Sorry if the
question sounds naive. I'd really appreciate any insights/comments!

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