teffects with panel data?

[Thibaud Marcesse]

I am working with panel data and have used xthdidregress to do a difference in differences analysis using the reservation status of villages in rural India as a `treatment.’ I was considering using teffects as a robustness check for my estimates, and it appears to work with panel data. However, my treatment estimates are wildly different. Is this because teffects only computes an aggregated difference between two groups, and does not take into account the pre-treatment stage? Also, would anyone recommend using teffects with panel data or not?

Thanks for letting me know!

[Jeff Wooldridge]

Thibaud: You need to provide more information. Is this a staggered intervention? Exactly how are you applying teffects? If you use the aipw option with xthdidregress then it is effectively using teffects on the differenced outcomes (across two time periods). If you apply teffects to the levels of the outcomes, you are not fully controlling for unobserved heterogeneity. See the FAQs about showing us what you typed and what Stata produced. Also, using -dataex- is recommended.