Hi, I am a new member, so apologies in advance for rookie mistakes.
My question is more methodological than implementational.
Data: Two waves of household survey panel data with women 18-45 years old in focus (asked many additional questions).
Outcome variable: whether or not a woman received pre-natal care during last birth (within three years of interview date).
Policy variation: A reproductive health intervention that started between two waves and had an exogenous roll out - across districts/counties.
Research question: Did the policy improve maternal health outcomes like pre-natal care?
Issue: Though it appears to be a case of straightforward diff-in-diff strategy , I face a dilemma.
Option 1. Create a balanced panel of women who have given birth within three years of interview date in BOTH ROUNDS, and then do a diff-in-diff to see women who were exposed to the policy had higher likelihood of getting care. (Again the policy varies at district level.)
pros: I can eliminate individual-level time invariant unobserved factors.
cons: There is a selection into women choosing to be pregnant before the second time. Also, sometimes people take less care in second pregnancy.
Option 2: Pool all recent mothers across two waves and do a standard D-i-D.
Pros. No such selection.
Cons: While the D-i_D can take care of district-level time invariant unobservables, it cannot do so at the individual level.
I am also wondering about how to implement #1 above in the most efficient way in Stata as my dependent variable is binary.
Thanks very much in advance!
My question is more methodological than implementational.
Data: Two waves of household survey panel data with women 18-45 years old in focus (asked many additional questions).
Outcome variable: whether or not a woman received pre-natal care during last birth (within three years of interview date).
Policy variation: A reproductive health intervention that started between two waves and had an exogenous roll out - across districts/counties.
Research question: Did the policy improve maternal health outcomes like pre-natal care?
Issue: Though it appears to be a case of straightforward diff-in-diff strategy , I face a dilemma.
Option 1. Create a balanced panel of women who have given birth within three years of interview date in BOTH ROUNDS, and then do a diff-in-diff to see women who were exposed to the policy had higher likelihood of getting care. (Again the policy varies at district level.)
pros: I can eliminate individual-level time invariant unobserved factors.
cons: There is a selection into women choosing to be pregnant before the second time. Also, sometimes people take less care in second pregnancy.
Option 2: Pool all recent mothers across two waves and do a standard D-i-D.
Pros. No such selection.
Cons: While the D-i_D can take care of district-level time invariant unobservables, it cannot do so at the individual level.
I am also wondering about how to implement #1 above in the most efficient way in Stata as my dependent variable is binary.
Thanks very much in advance!
