Hello everyone,
I'm currently conducting my master thesis on the impact of conditional cash transfer (CCT) on various household welfare indicators. The panel dataset I'm using goes from 2006 to 2018. The CCT program was implemented in 2015, however, the effect of the program (asking interviewed households whether received the program or not) was captured in 2018 wave, hence there's no information on 2015.
I have created a dummy for treatment and time as follow:
* Time dummy
gen time = (year =2018) & !missing(year)
which takes 1 if happens in or after 2015; 0 otherwise
* Treatment dummy
gen treated = (received_takaful == 1)
which takes 1 if happens if recevied Takaful "name of the CCT program"; 0 otherwise.
One of my outcomees of interest is years of schooling. To check for the pre-period parralel trends assumptions, I ran the following command:
preserve
collapse (mean) yrs_of_schooling, by(treated year)
reshape wide yrs_of_schooling, i(year) j(treated)
graph twoway connect yrs_of_schooling* year if year < 2018
restore
However, this is the plot of the graph i receive: Only showing the control group with no existence of treated individuals.
Likewise, when I write the command:
lgraph yrs_of_schooling year, by(treated)
I also receive the following:
Can u please advise what could be the problem why the trends/plot for the treatment group is not depicted in the graph?
I have been thinking about it for a while and I reached to some conclusions:
I'm currently conducting my master thesis on the impact of conditional cash transfer (CCT) on various household welfare indicators. The panel dataset I'm using goes from 2006 to 2018. The CCT program was implemented in 2015, however, the effect of the program (asking interviewed households whether received the program or not) was captured in 2018 wave, hence there's no information on 2015.
I have created a dummy for treatment and time as follow:
* Time dummy
gen time = (year =2018) & !missing(year)
which takes 1 if happens in or after 2015; 0 otherwise
* Treatment dummy
gen treated = (received_takaful == 1)
which takes 1 if happens if recevied Takaful "name of the CCT program"; 0 otherwise.
One of my outcomees of interest is years of schooling. To check for the pre-period parralel trends assumptions, I ran the following command:
preserve
collapse (mean) yrs_of_schooling, by(treated year)
reshape wide yrs_of_schooling, i(year) j(treated)
graph twoway connect yrs_of_schooling* year if year < 2018
restore
However, this is the plot of the graph i receive: Only showing the control group with no existence of treated individuals.
Likewise, when I write the command:
lgraph yrs_of_schooling year, by(treated)
I also receive the following:
Can u please advise what could be the problem why the trends/plot for the treatment group is not depicted in the graph?
I have been thinking about it for a while and I reached to some conclusions:
- One of them could be because the number of treated individuals is way less than that of the control group, hence the control group is dominating in the graph? (Treated is 961 individuals; while control is more than 60,000).
- Maybe I'm lacking weights in the graph?
Comment