I had wanted to ask this community for some advice on some potential methodologies I am developing.

Context: I want to evaluate the effectiveness of a component of this scholarship program, which involves visiting schools and talk about opportunities at universities, qualms and prejudices, and ways to finance studies through their scholarships - unfortunately, school visits are not done evenly across schools.

Primary research questions : Do school visits influence the socio-economic or gender composition of recommended students?

Data I have:

  • Panel dataset between of 11 admission cycles (2009-2020) and containing a bunch of SSE and financial variables linked to each school.

  • Dummy variable for if a school visit happened

  • Dummy if school only recommended one student

  • Dummy if school recommended as many as allowed

  • SES info of students recommended from school

For my primary research question, I was curious what model you would suggest? I was thinking about an IV approach, given my dataset suggests a drop-off in recommendations between the years due to a quota implementation - I was thinking that this could be a good IV variable but curious to hear thoughts!

  • $\begingroup$ I find it very unclear what it is you think are affecting what and how. Gender composition of what? Do each school select a set of students to recommend and it is the gender composition of this set you are talking about. How? Who is doing the visiting? And how is this thought to have any affect on composition? Before you can even begin to discuss the validity of an IV you need to have some idea of this mechanism. I think adding this to question would improve it. Before doing any IV I would try some DD even if you end up rejecting this due to some endogeneity probs. You cannot not do it. $\endgroup$ May 30, 2021 at 18:21
  • $\begingroup$ Hi, thank you, and I will edit the question further to reflect more details - answering your question, gender composition of the students selected and the people doing the visiting are essentially past alumni of the program who then encourage students to apply. Hmm interesting suggestion and I will think around a DD before but, out of curiosity, may you elaborate a bit on why you think this? - I was thinking around IV given the strict cut-offs of the scholarship program might def be a factor in change in composition $\endgroup$
    – Tara
    May 30, 2021 at 19:54
  • $\begingroup$ Not sure the DD will work for your data (now that I think I understand how the school visits works - still perhaps there is a subsample of schools that go from not having these to having - offcourse here there is major selfselection problem - but as a first step and to introduce what it is you are trying to estimate it may still work ... just something to have in an introduction maybe ... first slides of a presentation). $\endgroup$ May 30, 2021 at 20:06
  • $\begingroup$ I get why you think your IV strategy may work and I agree it is worth a shot. $\endgroup$ May 30, 2021 at 20:08
  • $\begingroup$ Yea, that makes sense! Just struggling a bit with re to modelling what I have ... $\endgroup$
    – Tara
    May 30, 2021 at 20:08


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.