Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Csdid - never treated group not recognized

    Dear Statalisters,
    I have repeated cross section survey data from different countries. I want to estimate how the (staggered) abolition of one law has impacted the outcome of interest. In my setting, the law was based on year of birth cohort, thus in the same country we have individuals that were subject to it (not treated) and individuals not subject to it anymore (treated). In addition to countries that abolished this particular law, I also have a sample of countries that never abolished it, thus individuals are always not treated. My issue is the following: when using csdid (or csdid2) from ssc the command seems not to recognize the never treated group (set as zero) and perform the regression only on not-yet treated groups. Any idea why?

    Please see attached example. The variable cohort is my gvar and varies by country.

    Code:
    Code:
    tab yrbrn cohort if inrange(yrbrn,1970,1975)
    
    
    Year of |                   cohort
         birth |         0       1942       1948       1976 |     Total
    -----------+--------------------------------------------+----------
          1970 |       626        110         20        104 |     2,398 
          1971 |       562        109         10         87 |     2,214 
          1972 |       572         92          6         67 |     2,121 
          1973 |       492         78          6         90 |     2,041 
          1974 |       499         77          3         73 |     2,049 
          1975 |       471         70          7         64 |     1,977 
    -----------+--------------------------------------------+----------
         Total |     3,222        536         52        485 |    12,800 
    
    
    csdid AvgIndex, time(yrbrn) gvar(cohort) long2 cluster(country_wave)
    No never treated observations found. Using Not yet treated data

  • #2
    THis is actuallybecause your data is not adequate for staggered DID
    1) your last period 1976 happens AFTER the last observed year.
    2) the other two periods happen BEFORE the first observed year

    Because of this, there is no proper before and after data to do a proper DID

    Comment

    Working...
    X