Hello, i am stuck in assessing the best way forward for my analysis.
I have a long format dataset, of which i have run an Dif-in-Dif as a basemodel. The code for this is [reg share Immigrant postreform Immigrant_postreform, cluster(School_id)].
So [Immigrant] is my group sorting variable, [Postreform] the time-indicator, and lastly the interaction between the two. Since they are both coded 0-1, the interpertation becomes an DiD.
The data is school-based, which is why i adjust on that level. Each school appears in 8 observations (different years). Reshaping my data to long format, implicate that i also have a variable of [Schoolid_year], where each uniqe combination of [School_id] and [year] appears twice, one for [Immigrant=0] and one for [Immigrant=1].
I now wish to introduce another condition. I have an expectation that the interaction effect of [immigrant X postreform] will vary depending of an socioeconomic reference value (SES). This variable is assigned on school level every year. So each school_id will have 1 estimation of [SESreference] assigned per year - also meaning that since each [Schoolid_year] will appear twice (because of [Immigrant=0] [Immigrant=1] for each year), each set of [schoolid_year]will have the same value of [SESreference]. [SESreference] is a continiuos variable.
So how do i best estimate this? I have attempted to do a Fixed effects on [Schoolid_year], but i am simply not sure if this is the way to go, and unsure of the excat interpretation of the output.
I can do the following FE, and get the same outcome as my base DiD: [reghdfe share postreform Immigrant_postreform, absorb(immigrant) cluster(School_id)].
I then want attempt to use the command: [reghdfe share postreform Immigrant_postreform, absorb(Schoolid_year immigrant) cluster(School_id)]. it changes the outcome coeficient of immigrant_postreform marginally, but i am not sure how too interperet it.
I have also attempted to use another proxy variable for [SESreference], that is [schooltype] which is a dummy variable, so that way it is easier to interperet as a Dif-in-Dif-in-Dif. But that variable is not a very good proxy, so i would rather not have to use that.
code for that: [reg share Immigrant postreform Schooltype Immigrant_postreform Schooltype_Immigrant Schooltype_Postreform Immigrant_postreform_Schooltype, cluster(School_id)].
Is there another way that i can include this condition of SESreference on school_year basis? Thankful for any help
Best regards, Ninna
I have a long format dataset, of which i have run an Dif-in-Dif as a basemodel. The code for this is [reg share Immigrant postreform Immigrant_postreform, cluster(School_id)].
So [Immigrant] is my group sorting variable, [Postreform] the time-indicator, and lastly the interaction between the two. Since they are both coded 0-1, the interpertation becomes an DiD.
The data is school-based, which is why i adjust on that level. Each school appears in 8 observations (different years). Reshaping my data to long format, implicate that i also have a variable of [Schoolid_year], where each uniqe combination of [School_id] and [year] appears twice, one for [Immigrant=0] and one for [Immigrant=1].
I now wish to introduce another condition. I have an expectation that the interaction effect of [immigrant X postreform] will vary depending of an socioeconomic reference value (SES). This variable is assigned on school level every year. So each school_id will have 1 estimation of [SESreference] assigned per year - also meaning that since each [Schoolid_year] will appear twice (because of [Immigrant=0] [Immigrant=1] for each year), each set of [schoolid_year]will have the same value of [SESreference]. [SESreference] is a continiuos variable.
So how do i best estimate this? I have attempted to do a Fixed effects on [Schoolid_year], but i am simply not sure if this is the way to go, and unsure of the excat interpretation of the output.
I can do the following FE, and get the same outcome as my base DiD: [reghdfe share postreform Immigrant_postreform, absorb(immigrant) cluster(School_id)].
I then want attempt to use the command: [reghdfe share postreform Immigrant_postreform, absorb(Schoolid_year immigrant) cluster(School_id)]. it changes the outcome coeficient of immigrant_postreform marginally, but i am not sure how too interperet it.
I have also attempted to use another proxy variable for [SESreference], that is [schooltype] which is a dummy variable, so that way it is easier to interperet as a Dif-in-Dif-in-Dif. But that variable is not a very good proxy, so i would rather not have to use that.
code for that: [reg share Immigrant postreform Schooltype Immigrant_postreform Schooltype_Immigrant Schooltype_Postreform Immigrant_postreform_Schooltype, cluster(School_id)].
Is there another way that i can include this condition of SESreference on school_year basis? Thankful for any help
Best regards, Ninna
