Re: st: Difference-in-Difference on panel data without treatment and control group distinction

[Nils Braakmann <nilsbraakmann@googlemail.com>]

Well, just to comment on that idea, too... The only chance for a rdd
that I see with your question is using time as the forcing variable
and the exact timing (day, month) of the financial crisis as the
threshold . However, you would have to deal with several issues:
(a) You would need an exact threshold where treatment status changes.
In you case this could only be the onset of the financial crisis.
However, I am not sure whether there is any theoretically justified
single point in time from which on the financial crisis "is there".
(b) Probably more problematic: An rdd gives you a local effect
directly at the threshold, i.e. the difference in outcomes directly
before and after the threshold. With time as the forcing variable,
this means that you are essentially looking at the ultra-short run, in
an extreme case comparing cases one day before the financial crisis
with cases one day after the financial crisis. If your effects of
interest need more time to materialize, you'd have to widen the window
around the discontinuity threshold and are essentially back at
before-after-comparisons (which is essentially a rdd with a very wide
window around the discontinuity). Widening the time around the
discontinuity, however, makes it more likely that influences other
than the financial crisis interfere with your estimation.

Hope this helps,
Nils

On Tue, Apr 6, 2010 at 3:14 PM, Eirik Egeland Nerheim
<Eirik.Nerheim@stud.nhh.no> wrote:
> Thanks. I am going to look into regression discontinuity now:-)
>
> *
> *   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/
>

*
*   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/