Hi,
I am using four waves of repeated cross-sectional data to run a DiD with staggered treatment. Treatment is at the district level and outcomes are at individual. Districts can enter treatment from the 3rd wave. Once treated, their status never switches back. I am trying to estimate treatment effects using the csdid command, however no coefficients are getting estimated. The code I have used is the following:
Here, 'emp' is my dependent variable. I initially defined gvar as '0' for all the individuals in districts that are never-treated, '2010' for all individuals (across all waves) in districts that entered treatment in wave 3 and '2018' for all individuals in districts that entered treatment in wave 4. When I tabulated wave and gvar in my data, it looked like this:
tab year first_treat
| first_treat
Year | 0 2010 2018 | Total
-----------+---------------------------------+----------
2000 | 47,693 34,653 22,483 | 104,829
2005 | 52,894 31,859 21,650 | 106,403
2010 | 42,687 25,430 17,341 | 85,458
2018 | 39,878 26,337 17,498 | 83,713
-----------+---------------------------------+----------
Total | 183,152 118,279 78,972 | 380,403
When no estimates were produced this way, I defined gvar as '0' for non-treated, '2010' only for individuals observed in wave 3 within districts that entered treatment that year, and similarly '2018' for individuals in districts that entered treatment in 2018. Here's, a tabulation of wave and the new gvar:
| first_treat
Year | 0 2010 2018 | Total
-----------+---------------------------------+----------
2000 | 104,829 0 0 | 104,829
2005 | 106,403 0 0 | 106,403
2010 | 60,028 25,430 0 | 85,458
2018 | 66,215 0 17,498 | 83,713
-----------+---------------------------------+----------
Total | 337,475 25,430 17,498 | 380,403
However, in this case too, csdid produced no estimates of treatment coefficients. Where can the problem lie? If you need an example dataset, please let me know which gvar is correct and I will provide a representative sample of the data.
Thank you, in advance, for your help!
I am using four waves of repeated cross-sectional data to run a DiD with staggered treatment. Treatment is at the district level and outcomes are at individual. Districts can enter treatment from the 3rd wave. Once treated, their status never switches back. I am trying to estimate treatment effects using the csdid command, however no coefficients are getting estimated. The code I have used is the following:
Code:
csdid emp if female == 0, cluster(id) time(year) gvar(first_treat) method(dripw) notyet
tab year first_treat
| first_treat
Year | 0 2010 2018 | Total
-----------+---------------------------------+----------
2000 | 47,693 34,653 22,483 | 104,829
2005 | 52,894 31,859 21,650 | 106,403
2010 | 42,687 25,430 17,341 | 85,458
2018 | 39,878 26,337 17,498 | 83,713
-----------+---------------------------------+----------
Total | 183,152 118,279 78,972 | 380,403
When no estimates were produced this way, I defined gvar as '0' for non-treated, '2010' only for individuals observed in wave 3 within districts that entered treatment that year, and similarly '2018' for individuals in districts that entered treatment in 2018. Here's, a tabulation of wave and the new gvar:
| first_treat
Year | 0 2010 2018 | Total
-----------+---------------------------------+----------
2000 | 104,829 0 0 | 104,829
2005 | 106,403 0 0 | 106,403
2010 | 60,028 25,430 0 | 85,458
2018 | 66,215 0 17,498 | 83,713
-----------+---------------------------------+----------
Total | 337,475 25,430 17,498 | 380,403
However, in this case too, csdid produced no estimates of treatment coefficients. Where can the problem lie? If you need an example dataset, please let me know which gvar is correct and I will provide a representative sample of the data.
Thank you, in advance, for your help!
Comment