Dear Stata users,
I am analysing a national policy. There are 2 pre-treatment periods and 4 post treatment periods. Data is repeated cross-section at individual level, but treatment is being considered at state level. The policy starts at t=0 for all states, but both groups become treated as the policy covers the entire country. States with high intensity quotas were assigned to the treatment group while states with low intensity quotas belong to the comparison group. Treatment is irreversible and no-one is treated before the policy. Groups are stable - no leavers or switchers after the first period. However, treatment varies across time for both groups, but much less for the control group.
References I am using:
de Chaisemartin, C, D'Haultfoeuille, X (2020a). Difference-in-Differences Estimators of Intertemporal Treatment Effects.
de Chaisemartin, C, D'Haultfoeuille, X (2020b). Two-way fixed effects estimators with heterogeneous treatment effects (2020)
My code below:
The variable "treated" is binary including only high intensity states. Wondering if I'm doing it right.
My questions:
Thank you in advance.
Sandra
I am analysing a national policy. There are 2 pre-treatment periods and 4 post treatment periods. Data is repeated cross-section at individual level, but treatment is being considered at state level. The policy starts at t=0 for all states, but both groups become treated as the policy covers the entire country. States with high intensity quotas were assigned to the treatment group while states with low intensity quotas belong to the comparison group. Treatment is irreversible and no-one is treated before the policy. Groups are stable - no leavers or switchers after the first period. However, treatment varies across time for both groups, but much less for the control group.
References I am using:
de Chaisemartin, C, D'Haultfoeuille, X (2020a). Difference-in-Differences Estimators of Intertemporal Treatment Effects.
de Chaisemartin, C, D'Haultfoeuille, X (2020b). Two-way fixed effects estimators with heterogeneous treatment effects (2020)
My code below:
Code:
did_multiplegt scores stateid year treated , robust_dynamic dynamic(3) placebo(1) breps(20)
| Estimate SE LB CI UB CI N Switchers
-------------+------------------------------------------------------------------
Effect_0 | .0123678 .0136192 -.0143259 .0390615 824638 174220
Effect_1 | .0502237 .0192586 .0124768 .0879706 915453 199351
Effect_2 | .0500739 .0209238 .0090633 .0910846 935257 198399
Effect_3 | .0698923 .0286472 .0137439 .1260408 988892 216437
Average | .0472202 .0193816 .0092323 .0852082 3664240 788407
Placebo_1 | .0205559 .0187101 -.0161158 .0572276 824638 174220
My questions:
- the "placebo" test gets the same number of observations as the ones used for the effect_0 which is the number of observations for the first period after treatment. The entire sample consists of over 5 million observations, and 3,66 million includes only the post-treatment period. Why doesn't did_multiplegt compute the number of pre-treatment observations? I'm sure I must be missing something.
- I've got the matrices with the estimates and I can save with save_results, but I was wondering if there is a way to store the estimates and send them to a table.
- may I say that the average results correspond to the DIDm Wald-TC estimates?
Thank you in advance.
Sandra
