Dear Statalist Community,
I have a question about how I should cluster my standard errors when using a fuzzy RDD in time.
In my study, the year in which respondents finished mandatory schooling is my running variable, and the treatment is a 1940 territorial occupation that would affect the outcome through the exposure to a new schooling system (instrument: "school"=age when respondent left school).
I am aware that there is a new paper that says that one should not cluster SEs at the running variable level, but does that still hold for RDDs in time? I believe in this case the treatment level is something like "age respondents were when the occupation occurred", as the effect of the occupation via the new school system would occur at this level.
I really am confused about the correct way to go about this because I believe the level of the treatment is temporal in this case.
Thank you so much!
I have a question about how I should cluster my standard errors when using a fuzzy RDD in time.
In my study, the year in which respondents finished mandatory schooling is my running variable, and the treatment is a 1940 territorial occupation that would affect the outcome through the exposure to a new schooling system (instrument: "school"=age when respondent left school).
I am aware that there is a new paper that says that one should not cluster SEs at the running variable level, but does that still hold for RDDs in time? I believe in this case the treatment level is something like "age respondents were when the occupation occurred", as the effect of the occupation via the new school system would occur at this level.
I really am confused about the correct way to go about this because I believe the level of the treatment is temporal in this case.
Thank you so much!
Code:
rdrobust dv year_end_school, c(1940) all fuzzy(school) vce(cluster year_end_school)
