Undergraduate Gender Diversity and the Direction of Scientific Research
Brief overview
Exploit switch to co-education among research-oriented universities in the US in 1960-90
Main findings:
- increase in undergraduate female student body
- induced interest in gender-related topics among researchers
Further clarifications about interpretation
Policy change vs exposure
Persistence of induced interest in gender topics
Spillover effects within the analysis sample
Employment-related incentives
Additional comments and suggestions
Analysis of results of gender-related papers
Intensities
Co-authorship patterns
Seniority
Topic scopes
More discussion about lower co-authorship with incumbent female researchers
Teacher information in course catalogues
Additional measure of quality based on conference presentations
Similar study by race
Editorial comments
Comparisons in descriptive plots
For example, female vs male, gender- vs non-gender related papers, elite vs rest, etc.Institutional context
To what extent could faculty influence university admission policies?
Extended comment: Policy change vs exposure
Could the change in admission policy alone, even if there were no increase in female enrolments, generated interest in gender-related topics?
Even if results cannot be disentangle the two, some discussion may be helpful.
Extended comment: Persistence of induced interest in gender topics
This is related to the first comment whether interest was induced by exposure or policy change. The switches to co-education analysed in the study period may have been special and different from prior switches because it included largest wave of switches and higher share of elite universities. Therefore, the switch events alone could have generated interest in the topic, which for most researchers probably finishes as abruptly as it started. Analogous to the boom in Covid research. Therefore, analysis of how long-lasting these effects were could provide additional information whether the effects could be attributed to exposure or policy change.
The existing results in Figures 6 and D3 do suggest that the effect start dying out after year 3, which can be consistent with the policy-change induced interest.
Extended comment: Spillover effects within the analysis sample
- One consideration is randomness of sample observations. Universities in the analysis sample are well-known, meaning their switch is more visible to everyone. So, if Brown switches to co-education, it might influence decision of Yale to follow.
- Second consideration are spillovers of treatment effect. In introduction, the authors claim that in-sample universities are more research-oriented than out-of-sample universities. So, one possibility is that Brown switching to coeducation, sparks interest among Yale researchers. In addition, co-authorship ties within in-sample universities are potentially stronger than between in- and out-of-sample universities1. So, Yale researchers may have been involved in gender-related projects by their treated Brown colleagues.
One potentially useful robustness analysis could be to progressively exclude universities switching around 1970 (donut-ring type) or progressively exclude universities depending on co-authorship strength with elite research universities.
Extended comment: Employment-related incentives
The analysis in Section 6 suggests that faculty composition shifted
towards having more female researchers (Table 4) and
towards researchers with prior interest in gender topics (Table A11).
This suggests that researchers may have been (informally) required to produce some gender-related research in order to find or keep academic positions.
Extended comment: Intensities
It may be helpful to include analysis exploiting intensities in addition to indicators. For example,
intensity of increased exposure to female student body
intensity of relatedness of paper to gender topics (female words mentioned once in abstract vs ten times)
intensity of interest in gender topics (publishing one gender-related paper vs ten papers)
References
Footnotes
Which is what the current spillover analysis focuses on.↩︎