Due to the fact past literature has actually realized that the fresh new ATT can differ according to matching means, we establish our results for around three various other coordinating measures: (1) matching which have a good kernel Epanechnikov and you can a data transfer out of 0.1; (2) matching on the around three nearby locals within a distance off 0.01; and you may (3) so you’re able to restriction even more the fresh new assessment classification, i matches procedures and you can manage individuals within metropolitan/outlying, many years, and university attendance status (for the much time-run rates, we just maximum to help you urban/rural and ages).
A portion of the results are showed playing with a propensity score complete with linear, squares and communications terms. This new model using the MxFLS analysis spends 108 details and design utilizing the EMOVI investigation utilizes 57 details. 18 The latest robustness area comes with results for some other specifications of your own tendency score. And, we present robustness checks into log possibility proportion as the complimentary score as opposed to the tendency rating. Generally, the answers are steady around the requirements and you will complimentary strategies.
I tend to be 86 variables inside 2002 for the estimation of propensity score for MxFLS: age, many years of education, university attendance, functions reputation, native code, knowledge of contraceptives, past sexual activity, outlying status, and you may father missing on household
We estimate other evaluating to help you corroborate harmony throughout the tendency score.