I together with expose the outcome playing with most other matching methods due to the fact good robustness examine
Since the previous books possess noticed that the ATT may vary with respect to the matching approach, we present our results for around three various other complimentary strategies: (1) matching which have a great kernel Epanechnikov and you can a data transfer regarding 0.1; (2) coordinating on around three nearby locals within a distance out-of 0.01; and you will (3) so you can restriction even more the newest comparison category, i matches procedures and you will control people in this metropolitan/rural, decades, and you will college or university attendance status (on the enough time-focus on estimates, i simply limit to urban/rural and you can ages).
The main answers are displayed using a propensity score detailed with linear, squares and interaction terms. New model making use of the MxFLS investigation spends 108 variables together with model by using the EMOVI studies makes use of 57 details. 18 New robustness section includes results for some other demands of your own propensity get. Plus, we establish robustness checks towards the record potential proportion since complimentary rating rather than the tendency rating. Overall, all of our results are secure round the specifications and you can coordinating tips.
We become 86 details when you look at the 2002 to your estimate of your own propensity score having MxFLS: years, years of education, university attendance, really works position, local code, experience with contraceptives, earlier in the day sexual passion, outlying status, and you may father missing throughout the family
I guess some other assessment so you can corroborate harmony regarding tendency score. Earliest, we provide visual evidence according to abilities by the Dehejia and you will Wahba (1999, 2002) both before and after coordinating so you can corroborate the balancing in addition to commom assistance presumptions. We have the brand new stratification test both before and after coordinating recommended by Dehejia and Wahba (1999, 2002). 19 Second, i include the standard bias measure suggested from the Rosenbaum and you will Rubin (1985) both before and after matching. 20 We statement just the average standardized bias.