Hello,
I am struggling with setting up/converting my current dataset set-up in order to run a DiD estimation.
I have annual data for a number of years (unbalanced), per entity. For each entity I have their respective method of financial reporting as the binary DV variable (GAAP=1;Non-GAAP=0), their pension funded ratio is a continues variable (IV) and a number of other control variables for each year observed.
In my original model, I run a logistic regression with DV as the method of reporting and IV of interests as the funded ratio + controls. In simple terms: I am testing whether higher funded ratio is associated with increased likelihood of following GAAP.
Because entities could switch/change their method of reporting from GAAP -> Non-GAAP and the other way around, it was suggested I also run a difference in difference model. But I am not sure how I can set it up for Stata to be able to attempt the dif-in-dif estimation.
As I see it, I essentially have 4 groups within my 11 years of data: those that are always GAAP, those that are never GAAP, those that switched to GAAP, and those that witched from GAAP. There is no specific point in time (or one-year regulation) when they have to make the switch, it could vary from entity to entity.
Specific Questions:
-How do I recode it for purposes of the “control” and “treatment” groups?
-How do I convert to great the “pre” and “post” time?
-What if my outcome variable is binary?
Any thoughts, suggestions or directions towards materials describing similar set-up are greatly appreciated.
I am struggling with setting up/converting my current dataset set-up in order to run a DiD estimation.
I have annual data for a number of years (unbalanced), per entity. For each entity I have their respective method of financial reporting as the binary DV variable (GAAP=1;Non-GAAP=0), their pension funded ratio is a continues variable (IV) and a number of other control variables for each year observed.
In my original model, I run a logistic regression with DV as the method of reporting and IV of interests as the funded ratio + controls. In simple terms: I am testing whether higher funded ratio is associated with increased likelihood of following GAAP.
Because entities could switch/change their method of reporting from GAAP -> Non-GAAP and the other way around, it was suggested I also run a difference in difference model. But I am not sure how I can set it up for Stata to be able to attempt the dif-in-dif estimation.
As I see it, I essentially have 4 groups within my 11 years of data: those that are always GAAP, those that are never GAAP, those that switched to GAAP, and those that witched from GAAP. There is no specific point in time (or one-year regulation) when they have to make the switch, it could vary from entity to entity.
Specific Questions:
-How do I recode it for purposes of the “control” and “treatment” groups?
-How do I convert to great the “pre” and “post” time?
-What if my outcome variable is binary?
Any thoughts, suggestions or directions towards materials describing similar set-up are greatly appreciated.
Comment