Hi all,
I am trying to understand if the effect of a policy differs for different subgroups. I am using a simple two-way fixed effects regression. I would like to 1) estimate the effect of the policy on the different subgroups and 2) test if the differences are statistically significant.
For example, say I am interested in a policy that was introduced in 2016 and only affected people with two or more children (treated group), and the subgroups I am interested in within the treated group are households with different age of the youngest child. I ran the following two regressions. Note: Treated==1 if children >=2, treatedpost=1 if treated==1 & year>=2016. There are three subgroups to the age of youngest child variable agey_group (categories 1,2,3, standing for 0-5,6-10,11-17). y is some outcome variable.
(1) reg y ib(last).agey_group treated i.agey_group#treated ib(last).agey_group#treatedpost ///
`demographics' i.year i.state [pweight= wt06], cluster(state)
(2) reg y treated ib(last).agey_group#treated ib(last).agey_group##treatedpost ///
`demographics' i.year i.state [pweight= wt06], cluster(state)
The difference is in 1) I have ib(last).agey_group and ib(last).agey_group#treatedpost and in 2) I have ib(last).agey_group##treatedpost, which also includes a main effect for treatedpost. When I ran the first the regression I get the following:
agey_group#treatedpost
1 1 | -.0001651 .0208628 -0.01 0.994 -.0420693 .0417391
2 1 | -.0151655 .0252823 -0.60 0.551 -.0659465 .0356155
3 1 | -.0389839 .0786064 -0.50 0.622 -.1968695 .1189018
|
When I ran the second regression I get the following
treatedpost | -.0389839 .0786064 -0.50 0.622 -.1968695
agey_group#treatedpost |
1 1 | .0374768 .0645595 0.58 0.564 -.0921947 .1671484
2 1 | .0022179 .0630853 0.04 0.972 -.1244926 .1289284
My current interpretation is that regression 1 gives me the estimates for the effect of the policy on the three groups and whether those effects are significant, and regression 2 gives the effect of the policy on the base group (group 3), and estimates for the interaction effect for the other groups and whether the interaction effects are significant.
Is this the correct interpretation?
Thanks!
Best,
Angela
I am trying to understand if the effect of a policy differs for different subgroups. I am using a simple two-way fixed effects regression. I would like to 1) estimate the effect of the policy on the different subgroups and 2) test if the differences are statistically significant.
For example, say I am interested in a policy that was introduced in 2016 and only affected people with two or more children (treated group), and the subgroups I am interested in within the treated group are households with different age of the youngest child. I ran the following two regressions. Note: Treated==1 if children >=2, treatedpost=1 if treated==1 & year>=2016. There are three subgroups to the age of youngest child variable agey_group (categories 1,2,3, standing for 0-5,6-10,11-17). y is some outcome variable.
(1) reg y ib(last).agey_group treated i.agey_group#treated ib(last).agey_group#treatedpost ///
`demographics' i.year i.state [pweight= wt06], cluster(state)
(2) reg y treated ib(last).agey_group#treated ib(last).agey_group##treatedpost ///
`demographics' i.year i.state [pweight= wt06], cluster(state)
The difference is in 1) I have ib(last).agey_group and ib(last).agey_group#treatedpost and in 2) I have ib(last).agey_group##treatedpost, which also includes a main effect for treatedpost. When I ran the first the regression I get the following:
agey_group#treatedpost
1 1 | -.0001651 .0208628 -0.01 0.994 -.0420693 .0417391
2 1 | -.0151655 .0252823 -0.60 0.551 -.0659465 .0356155
3 1 | -.0389839 .0786064 -0.50 0.622 -.1968695 .1189018
|
When I ran the second regression I get the following
treatedpost | -.0389839 .0786064 -0.50 0.622 -.1968695
agey_group#treatedpost |
1 1 | .0374768 .0645595 0.58 0.564 -.0921947 .1671484
2 1 | .0022179 .0630853 0.04 0.972 -.1244926 .1289284
My current interpretation is that regression 1 gives me the estimates for the effect of the policy on the three groups and whether those effects are significant, and regression 2 gives the effect of the policy on the base group (group 3), and estimates for the interaction effect for the other groups and whether the interaction effects are significant.
Is this the correct interpretation?
Thanks!
Best,
Angela
