Re: st: diff-in-diff in levels or in logs

[Maarten Buis <maartenlbuis@gmail.com>]

On Thu, May 2, 2013 at 3:18 PM, Jack Knife  wrote:
> I am estimating a difference in difference regression with stata...
> and I would like to know what would be the best approach:
> I estimate a difference in difference in levels: that is if my output
> variable is Y, treated individuals have the dummy D=1 and treatment
> period is in time T=2
>
> I run -reg Y T D T#D
>
> and the coefficient on T#D gives me the diff in diff estimates. But if
> Y is a count variable... how could I say that Y decreased by X% after
> the reform using the same approach?
> Should I regress log(Y) instead of Y?

You should not transform your dependent variable, instead you should
use a link function. However, for a fully saturated model like you
propose it just does not matter, either model gives you exactly the
same predicted values. Here is a discusion in terms of a binary
dependent variable:
<http://www.stata.com/statalist/archive/2012-02/msg00351.html>, and
here is an example in terms of a log link function:

*------------------ begin example ------------------
sysuse nlsw88, clear
reg wage i.union##i.married
predict yhat1

poisson wage i.union##i.married
predict yhat2

assert yhat1 == yhat2
*------------------- end example -------------------
(For more on examples I sent to the Statalist see:
http://www.maartenbuis.nl/example_faq )

Hope this helps,
Maarten

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany

http://www.maartenbuis.nl
---------------------------------
*
*   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/