Hi all,
I am working with (unbalanced) nonprofit accounting panel data with organization-year observations. Based on the mission statement of an organization in a given year, I identified "religious" organizations.
Religiosity is my main independent variable of interest. I want to match religious organizations with secular ones to account for inherent differences between the two. I want to do so based on industry code, state, and asset (quartiles). So far so good.
Ideally, I match one secular organization with one religious organization over time. This should not be to difficult, as industry and state do not differ over years, and for assets I planned on simply using the mean over years.
However, as my treatment dummy is based on yearly mission statements, this can change over time for an organization, in both directions.
What are options in Stata to deal with something like this? Or is this an inherent problem to any matching whatsoever?
Would simply matching per organization-year observation be a valid solution? I thought that it is quite important to match the same organizations over time, but perhaps I am mistaken in that regard?
Another issue is that, even if religion would be constant over time, not all organizations are present in the dataset for every year. It could be very well that I match organizations that are not present in the same years. A possible workaround I thought of was to count the number of times an organization is present in the data, and match based on that. This is of course not equivalent to matching on years, but does make sure I do not match an organization occurring once with one occurring 10 times. My data spans from 2009 to 2020, and I thought perhaps I could simply control for years in the analysis, but ignore it in that manner for the matching procedure. Is that something that would make (methodological) sense?
I'm sorry if my questions are not relevant for this forum, as they are not directly Stata related but more methodological in nature.
Kind regards,
Johannes
I am working with (unbalanced) nonprofit accounting panel data with organization-year observations. Based on the mission statement of an organization in a given year, I identified "religious" organizations.
Religiosity is my main independent variable of interest. I want to match religious organizations with secular ones to account for inherent differences between the two. I want to do so based on industry code, state, and asset (quartiles). So far so good.
Ideally, I match one secular organization with one religious organization over time. This should not be to difficult, as industry and state do not differ over years, and for assets I planned on simply using the mean over years.
However, as my treatment dummy is based on yearly mission statements, this can change over time for an organization, in both directions.
What are options in Stata to deal with something like this? Or is this an inherent problem to any matching whatsoever?
Would simply matching per organization-year observation be a valid solution? I thought that it is quite important to match the same organizations over time, but perhaps I am mistaken in that regard?
Another issue is that, even if religion would be constant over time, not all organizations are present in the dataset for every year. It could be very well that I match organizations that are not present in the same years. A possible workaround I thought of was to count the number of times an organization is present in the data, and match based on that. This is of course not equivalent to matching on years, but does make sure I do not match an organization occurring once with one occurring 10 times. My data spans from 2009 to 2020, and I thought perhaps I could simply control for years in the analysis, but ignore it in that manner for the matching procedure. Is that something that would make (methodological) sense?
I'm sorry if my questions are not relevant for this forum, as they are not directly Stata related but more methodological in nature.
Kind regards,
Johannes
