Dear all,
I am currently using the mtreatreg-command, which has provided to be very useful!
However, I am stumbling upon one significant problem: when I change the base outcome, the results are inconsistent with each other.
My endogenous choice variable migtype_corop which indicates whether after graduating you migrate to i) your home region, ii) a new region, iii) you stay in the study region. I estimate the effect of this choice on hourly earnings: ln_hbiu. I have >17k observations in my dataset and estimate the following:
xi: mtreatreg ln_hbiu age sex foreign_parent grade i.pre_education post_degree experience i.industry i.year, mtreat(migtype_corop = age sex foreign_parent grade i.pre_education post_degree experience i.industry i.year, gdp_pc_homeregion unemployment_rate_homeregion pop_density_homeregion) basecat(1) sim(500) dens(normal) robust diff
I get highly significant results on the outcome equation and the lambda's. Now when I take basecategory 1, I find that those moving back earn more compared to those moving on, but when I take basecategory 2, I find the reverse, namely those moving on earn significantly more compared to those returning.
Do you have any idea what can be the problem?
My data is in wide form, and running the VIF-command after the OLS regression shows no multicolinearity issues.
Your help is very much appreciated!
Best regards,
Christian
I am currently using the mtreatreg-command, which has provided to be very useful!
However, I am stumbling upon one significant problem: when I change the base outcome, the results are inconsistent with each other.
My endogenous choice variable migtype_corop which indicates whether after graduating you migrate to i) your home region, ii) a new region, iii) you stay in the study region. I estimate the effect of this choice on hourly earnings: ln_hbiu. I have >17k observations in my dataset and estimate the following:
xi: mtreatreg ln_hbiu age sex foreign_parent grade i.pre_education post_degree experience i.industry i.year, mtreat(migtype_corop = age sex foreign_parent grade i.pre_education post_degree experience i.industry i.year, gdp_pc_homeregion unemployment_rate_homeregion pop_density_homeregion) basecat(1) sim(500) dens(normal) robust diff
I get highly significant results on the outcome equation and the lambda's. Now when I take basecategory 1, I find that those moving back earn more compared to those moving on, but when I take basecategory 2, I find the reverse, namely those moving on earn significantly more compared to those returning.
Do you have any idea what can be the problem?
My data is in wide form, and running the VIF-command after the OLS regression shows no multicolinearity issues.
Your help is very much appreciated!
Best regards,
Christian
