Hello,
I’m using data coming from an RCT experiment with two groups of individuals: a control group, and a treatment group (treat). Since not all treated individuals actually participated in the treatment, I want to perform a diff-in-diff with instrumental variable.
I instrument the variable indicating actual participation in the treatment (“digit“ in the following code) with the initial assignment to the treatment group (the “treat” dummy). Variables “digit” and "year" are also dummies in that case.
Here is my simplified specification:
I use “c.” instead of “i.” in front of dummies, otherwise, Stata drops some interaction terms.
I have two questions:
1) Can the instrumented variable be continuous if the instrument is a dummy? In my case, can I use the intensity of treatment use, rather than whether or not individuals actually used the treatment?
2) When specifying dummies interactions with “c.”, which option does Stata display? Is it the equivalent of ib1.digit #ib1.year?
Many thanks,
I’m using data coming from an RCT experiment with two groups of individuals: a control group, and a treatment group (treat). Since not all treated individuals actually participated in the treatment, I want to perform a diff-in-diff with instrumental variable.
I instrument the variable indicating actual participation in the treatment (“digit“ in the following code) with the initial assignment to the treatment group (the “treat” dummy). Variables “digit” and "year" are also dummies in that case.
Here is my simplified specification:
ivreg2 dep_variable ///
c.year ///
(c.digit c.digit #c.year = c.treat c.treat #c.year) , cluster (id) first
c.year ///
(c.digit c.digit #c.year = c.treat c.treat #c.year) , cluster (id) first
I have two questions:
1) Can the instrumented variable be continuous if the instrument is a dummy? In my case, can I use the intensity of treatment use, rather than whether or not individuals actually used the treatment?
2) When specifying dummies interactions with “c.”, which option does Stata display? Is it the equivalent of ib1.digit #ib1.year?
Many thanks,
