Dear STATALIST community,
I would like to kindly ask your help. I am using the did2s command (but have also used the csdid) to run an event study. I have panel data and my unit is country and my time variable is year. My dependent variable corresponds to transitions to democracy, being always equal to zero, except when a transition to transition occurs (==1). My treatment (treat_massmov) is absorbing and corresponds to the momement from which there is a mass mobilization movement (moment in which it switches from 0 to 1).
I have used the did2s command with no problem to establish the effect of treatment on democratization:
However, when I use it to produce an event study, it produces strange results. I was wondering if it could be a problem that my dependent variable does not remain 1 during the time in which a country is a democracy, since I wanted to measure the effects of the treatment on a transition to democracy event. Furthermore, the fact that a country can democratize and revert to autocracy could also potentially be a problem, as the DV would switch to 1 multiple times in the same unit.
Should I just use a dependent variable that ==0 during a country's authoritarian spell, and ==1 during its democratic spells? And should this variable be binary? Or should it be continuous (e.g. democratization index -10 to 10)?
Thank you very much for your help.
PS: the way in which I was coding the event study, follows the did2s authors' help file for event studies:
I would like to kindly ask your help. I am using the did2s command (but have also used the csdid) to run an event study. I have panel data and my unit is country and my time variable is year. My dependent variable corresponds to transitions to democracy, being always equal to zero, except when a transition to transition occurs (==1). My treatment (treat_massmov) is absorbing and corresponds to the momement from which there is a mass mobilization movement (moment in which it switches from 0 to 1).
I have used the did2s command with no problem to establish the effect of treatment on democratization:
Code:
did2s transition if [aw=_webal], first_stage(i.country_id i.year controls) second_stage(treat_massmov) treatment(treat_massmov) cluster(country_id)
Should I just use a dependent variable that ==0 during a country's authoritarian spell, and ==1 during its democratic spells? And should this variable be binary? Or should it be continuous (e.g. democratization index -10 to 10)?
Thank you very much for your help.
PS: the way in which I was coding the event study, follows the did2s authors' help file for event studies:
Code:
*First, I create a time trend variable: *The variable first_treat corresponds to the year in which a country is first treated (has its first mass mobilization movement) bysort country_id (year): gen timetrend = year - first_treat replace timetrend=0 if treat_group==0 *Restrict it 20 years prior and after the treatment. gen timetrend2=timetrend replace timetrend2= . if timetrend>20 replace timetrend2= . if timetrend<-20 gen rel_year_shift = timetrend2 + 20 replace rel_year_shift = 100 if rel_year_shift == . *Run the model: did2s transition if [aw=_webal], first_stage(i.country_id i.year controls) second_stage(ib100.rel_year_shift) treatment(treat_massmov) cluster(country_id)
