Dear Statalist users
I have a question with a potential selection-issue in an ongoing study.
Can you use different unit of analysis for the first and second stage of a Heckman model?
Put simply, we have a total sample of around 70 firms, where only 12 firms engage in philanthropic activities. The intended dependent variable is firm philanthropy, which is observed for a subsample of a larger population, causing concern for sample selection.
Therefore, sample selection issue may be happening at the firm-year level, where we have some firms self-selecting to engage in philanthropy in a given year.
However, our main interest is observing whether firms' certain activities at the county level in a specific year leads to philanthropy in the county. In other words, the unit of analysis becomes firm-county-year.
Conceptually, we would want to (1) in the first stage, estimate the likelihood of firms selecting into the philanthropy sample and (2) examine whether firms' activities in a county lead to philanthropy in that county.
The problem is, the unit of analysis would not be congruent in the first and second stage of a Heckman model.
I hope the question makes sense and I do apologize for my lack of knowledge in the matter.
If you have any advice or insight on the matter, any help would be greatly appreciated.
Thank you in advance.
I have a question with a potential selection-issue in an ongoing study.
Can you use different unit of analysis for the first and second stage of a Heckman model?
Put simply, we have a total sample of around 70 firms, where only 12 firms engage in philanthropic activities. The intended dependent variable is firm philanthropy, which is observed for a subsample of a larger population, causing concern for sample selection.
Therefore, sample selection issue may be happening at the firm-year level, where we have some firms self-selecting to engage in philanthropy in a given year.
However, our main interest is observing whether firms' certain activities at the county level in a specific year leads to philanthropy in the county. In other words, the unit of analysis becomes firm-county-year.
Conceptually, we would want to (1) in the first stage, estimate the likelihood of firms selecting into the philanthropy sample and (2) examine whether firms' activities in a county lead to philanthropy in that county.
The problem is, the unit of analysis would not be congruent in the first and second stage of a Heckman model.
I hope the question makes sense and I do apologize for my lack of knowledge in the matter.
If you have any advice or insight on the matter, any help would be greatly appreciated.
Thank you in advance.
