Re: st: How to interpret time dummies in simple difference in difference regression

[Austin Nichols <austinnichols@gmail.com>]

Melisa Haytasingh <prettymelisa@gmail.com>:
Probably you want to define a new variable RelTim=time-1970 for time
relative to the discovery of resources, and then interact that with
dummies for each country. Graph your predictions to see that the coef
on the dummy for TnT will be the level jump in gdp_percapita and the
coef on TnTxRelTim will be the difference in time trends before and
after discovery. Using time instead of relative time makes it easy to
project your comparisons back to the nonexistent year zero, rather
then making comparisons at 1970. It's not clear to me why you would
want to assume constant linear growth in gdp_percapita in all
countries as your baseline (counterfactual), but linear time trends
are popular in these kinds of regressions.

On Mon, Nov 18, 2013 at 7:24 AM, Melisa Haytasingh
<prettymelisa@gmail.com> wrote:
> Hi guys,
>
> I'm doing a regression to analyse the difference between resource-rich
> Trinidad & Tobago and their resource-poor Caribbean neighbour,
> Barbados. I am treating resources (oil and gas) as the endogenous
> variable and regressed GDP on 5 variables: a time trend, dummy taking
> 1 for all observations for T&T before resources (before 1970), dummy
> taking 1 after discovery of resources, dummy taking 1 for obs of
> Barbados before 1970 and dummy taking 1 for obs of Barbados after
> 1970. The regression looks like this in Stata:
>
> -          regression 1:
>
> reg gdp_percapita time TnT_no_resources TnTresources
> barbados_before1970 barbados_after1970, nocons
>
> -          regression 2:
>
> reg inlfation time TnT_no_resources TnTresources barbados_before1970
> barbados_after1970, nocons
>
>
> I am having trouble figuring out how to interpret the results. I think
> I need to find the difference between the coefficients like this:
>  (barbados_before1970 - TnT_no_resources) and (barbados_after1970 -
> TnTresources) and then compare those figures with each other. But I am
> not sure. Also, I am not sure about how to interpret the coefficient
> itself on each of the dummies. I tried to look for similar regressions
> online and I got a better overall understanding but still confused
> about how to interpret. :(
>
>
> Thank you very much for any help.
>
>
> Sincerely,
>
> Melisa
*
*   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/