Triple difference and fixed effects - how to do-issues

[Akane Marie]

Dear everyone,

Im currently doing a rapport about developing countries' export to developed countries, where I (in stata) have to conduct a triple difference analysis combined with fixed effect. Im very new to Stata programme and therefore have issues with it which I hope anyone can help with me.

The data:
I have destination and origin countries (exporter and importer), products (97), and I have tariff data (Ad valorem equivalent, expressed like 0.001809) for these products for 2001, 2004 and 2008. So I have to analyze impact of the developed countries' trade scheme with tariff reduction on the exports from the developing countries.
When I convert the data from excel to Stata all the variables were in string, so I convert the all three tariffs into numeric by typing destring Tariff2001, replace. However the variables come out as Double type. Am I doing it right?
Furthermore, I have integers the destination and origin countries by using encode commend by using encode destination, generate(importer) and encode origin, generate(exporter) so that I have encoded versions of the origin and destination country. This comes out in two colums in the data.
I have to conduct a country-fixed effect (I guess i have to use both importer and exporter?), product-fixed effect (for the 97 products) and year/time-fixed effect (2001, 2004 and 2008) (with this one I don't have any rows or columns with the years). How to I proceed from her? Im a little bit lost on how to do the FE and thereby continue to the Triple difference.

Am I missing out some information? Let me know.

Thanks for the help
Regards, Akane Marie

[Christos Makridis]

Generally speaking, if something does not convert from string to numeric, it means there is one or more values that's string. To find those quick, use the "isreal" command on the variable.

I have never used encode and am not sure what purpose it serves, but it sounds like there's no reason for you to use it. You have numeric ids for the destination and origin, right? Write out your triple DD estimator. I would highly highly recommend reading Petra Moser's AER paper that uses one -- I've always referred to it for educational purposes (http://papers.ssrn.com/sol3/papers.c...act_id=1910247). Where is your variation coming from -- changes within country and within products?

[Akane Marie]

Hi Christos,
thanks for your answer.
I don't have numeric ids for the destination and origin - I actually changed the number to letters, like Belgium - BEL. Should I change it back to numbers? That won't be an issue.
Regarding variation, I have a separate firm level data from the developing countries for the same products, year and country. So I have data that changes across country, time and products. The purpose of the report is to the see the effect of the tariff changes on firms from developing countries. I have written this equation so far:
#firm_cpt = Beta0 + Beta1 *tariff_cpt + gamma_c + gamma_p + gamma_t

The gamma part is the fixed effect and the tariff_cpt is the difference-in-difference.