Hi,
I have a DID research design that is as follows. The data I have covers the period of 2010-2016 which is by firm level. Treatment/control status is defined at the time of 12/31/2013 at county level. It means that, at the end of 2013, some of the counties are given intervention- in the treatment group and some are not- in the control group. The main outcome variable is measured in a time span. it often covers a year long period or so but the frequency of observation is quarterly- so each quarter is a point of observation but the data reflect the past year observation.
The complication arises when the data is at June of 2014- second quarter, which means that the outcome variable is measured between June 2013 and June 2014. So it covers the time of treatment and therefore such outcome variable value can not be assigned to either treatment or control group for this quarter In this case I assign the missing values to proceed, which i am not sure if it's acceptable. Following this method, I will have 4 points that are assigned missing in the year of 2014 (from Jan 2014 to Jan 2015) because the treatment status can not be defined.
Further complication is that not every county is treated at the same time, so for example if country A is treated at the end of 1st quarter of 2014, then I will also have to assign missing for that county for the data between March 2014 and March 2015. In this case I still have treatment/control and before/after
My question is, is it acceptable if I just proceed with these missing values in an DiD estimation at year-quarter level- considering I still have enough quarters before and after the intervention time for both treatment and control groups? I have a bad feeling about this but I do not know if there is literature on this. Also it's not often an outcome variable is measured between two dates.
Thanks a lot,
I have a DID research design that is as follows. The data I have covers the period of 2010-2016 which is by firm level. Treatment/control status is defined at the time of 12/31/2013 at county level. It means that, at the end of 2013, some of the counties are given intervention- in the treatment group and some are not- in the control group. The main outcome variable is measured in a time span. it often covers a year long period or so but the frequency of observation is quarterly- so each quarter is a point of observation but the data reflect the past year observation.
The complication arises when the data is at June of 2014- second quarter, which means that the outcome variable is measured between June 2013 and June 2014. So it covers the time of treatment and therefore such outcome variable value can not be assigned to either treatment or control group for this quarter In this case I assign the missing values to proceed, which i am not sure if it's acceptable. Following this method, I will have 4 points that are assigned missing in the year of 2014 (from Jan 2014 to Jan 2015) because the treatment status can not be defined.
Further complication is that not every county is treated at the same time, so for example if country A is treated at the end of 1st quarter of 2014, then I will also have to assign missing for that county for the data between March 2014 and March 2015. In this case I still have treatment/control and before/after
My question is, is it acceptable if I just proceed with these missing values in an DiD estimation at year-quarter level- considering I still have enough quarters before and after the intervention time for both treatment and control groups? I have a bad feeling about this but I do not know if there is literature on this. Also it's not often an outcome variable is measured between two dates.
Thanks a lot,
