Dear Statalisters! I'm pretty new to Statalist and found a lot of posts regarding DID and PSM, but nothing that would fit my specific setting. So I apologize if the topic already exists.
I have a repeated cross-sectional dataset for Chile from 1990-2017 (Household surveys that were conducted every two to three years). I merged the surveys to be able to work with the data.
I want to investigate how the labor supply of women evolved after the provision of free childcare in 2006 for a certain subgroup of the population.
Eligibility for childcare is dependent on the following:
-mother is in the labor force (working, looking for a job or studying)
-her child is below the age of 5
-she is part of the lower 60% of the income distribution.
Treatment group: women who are in the labor force (working, looking for a job or studying), with the youngest child (must be <5) in childcare
Control group: Women who are in the labor force (working, looking for a job or studying) with children bellow 5 but who do not send them to childcare.
Control variables: age, education, education_partner, number of children, i.year, i.region
I have two outcome variables: hours worked weekly and a binary variable that answers whether the women worked last week or not (1,0).
Now I would like to use DID in order to measure the treatment effect of the intervention. Since it is not longitudinal data, I have some trouble finding the right specification. I ran DID for hours worked but it gives me insignificant results.
command: reg hrs_work treat time treat##time age educ male_educ nr_child i.year i.r
Do you see what I am missing?
My thought was that propensity score matching could help. But I would need to pscore by year which is not possible as far as I know?
Command:
global treatment treat
global ylist hrs_work
global xlist age educ male_educ nr_child
pscore $treatment $xlist, pscore(myscore) blockid(myblock) comsup
this does not take into account my time dimension. I am a bit lost.
So, to sum it up:
1. Question: How do I best proceed with my type of data for the continuous outcome variable of hours worked?
2. How can I combine PSM and DID that controls for the years and the fact that it is not panel data?
3. How do I proceed in the case of the binary variable? Probit/DID?
Thank you for your suggestions and time.
Kind Regards
Related Posts with Difference in Differences with propensity score matching for repeated cross sectional surveys
could not calculate numerical derivatives -- flat or discontinuous region encounteredDear all, I am trying to run the stochastic frontier model based on Battese and Coelli (1995), whic…
Summary Statistics for Categorical VariablesDear Statalist, I used the following command Code: sum influenced ibn.size and was wondering if …
Pooled Panel or Cross Sectional DataGood Morning I have a query regarding a data set I am handling. I am trying to see the out migration…
Creating two month intervalsHello, I would like to create a date variable that represent successive two month intervals - e.g., …
Pooled Panel or Cross Sectional DataGood Morning I have a query regarding a data set I am handling. I am trying to see the out-migration…
Subscribe to:
Post Comments (Atom)
0 Response to Difference in Differences with propensity score matching for repeated cross sectional surveys
Post a Comment