Hi,
I am trying to measure the effectiveness of minimum age marriage laws using triple differences but every time I run the regression, most of my variables are omitted due to collinearity. I've tried many different ways to do it using other Statalist questions and it still hasn't worked for me.
I have household data for many different countries and have created a dataset where there is an observation for each individual from ages 9 to 29. Each country implemented child marriage laws at different times and I am also using variation within ethnicities where the treatment group are ethnic groups who typically marry underage and the control group are ethnic groups who typically marry over 18.
I have a post reform variable which = 1 for each country where the current year is after their year the law was implemented. I have created a treatment variable that =1 if ethnic groups practice underage marriage across all countries and also individual treatment variables for each country as I am unsure which to use. I have a separate ethnicity variable for each country (came like this with my dataset). I also have individual country dummies which =1 for each country it is named after (e.g. for variable benin, =1 if country is benin; =0 otherwise)
I have set the dataset as a panel using xtset id age as this gave a strongly balanced panel (when i tried using xtset id currentyear, this came up as weakly balanced).
My equation is as follows:
Y is age of first marriage (this was originally supposed to be a dummy variable for whether an individual was married underage but I was unsure if this works with DDD)
The last code I have attempted is as follows. This had the least variables omitted but my DDD estimator was omitted so I still do not understand what to do.
Thank you
I am trying to measure the effectiveness of minimum age marriage laws using triple differences but every time I run the regression, most of my variables are omitted due to collinearity. I've tried many different ways to do it using other Statalist questions and it still hasn't worked for me.
I have household data for many different countries and have created a dataset where there is an observation for each individual from ages 9 to 29. Each country implemented child marriage laws at different times and I am also using variation within ethnicities where the treatment group are ethnic groups who typically marry underage and the control group are ethnic groups who typically marry over 18.
I have a post reform variable which = 1 for each country where the current year is after their year the law was implemented. I have created a treatment variable that =1 if ethnic groups practice underage marriage across all countries and also individual treatment variables for each country as I am unsure which to use. I have a separate ethnicity variable for each country (came like this with my dataset). I also have individual country dummies which =1 for each country it is named after (e.g. for variable benin, =1 if country is benin; =0 otherwise)
I have set the dataset as a panel using xtset id age as this gave a strongly balanced panel (when i tried using xtset id currentyear, this came up as weakly balanced).
My equation is as follows:
Y is age of first marriage (this was originally supposed to be a dummy variable for whether an individual was married underage but I was unsure if this works with DDD)
The last code I have attempted is as follows. This had the least variables omitted but my DDD estimator was omitted so I still do not understand what to do.
Code:
reg agefrstmar postreform##country country##dchildmar postreform##dchildmar postreform##country##dchildmar i.currentyear i.country i.ethnicitybj, cluster(id)
Attached Files
