Unconditional quantile regression / pooling together data

[Iur Domsky]

Dear Statalist members. Has anyone ever used unconditional quantile regression in stata (ivqte command; Firpo's two-spep 2007 estimator)? When I estimate the coefficient associated with the cross product of a time dummy variable (indicating 1 when the year is 1, 0 otherwise) and a qualitative dummy variable (that is not exclusively of year 1) with all the years/all the data, I get exactly the same results when I use the subset of the data comprising of year's 1 data points only, and my treatment effect is the qualitative dummy variable (and not the interaction term). Should I not be seeing some sort of efficiency gains when I pool together all the data? When I use different estimators (eg. ols), the coefficient estimates will be different. Maybe I am doing something wrong. Some information from someone that has already used this estimator would be good. Thanks

[FernandoRios]

Hi Iur,
A couple of notes on what you are doing.
1. What you are estimating is not an unconditional quantile regression. (see https://www.nber.org/papers/t0339) but rather quantile treatment effect under exogeneity)
2. I'm having problems visualizing your data. SO it may be wise to provide a small sample of what you are doing, (or at least the output you obtain) to better assess your problem.
HTH