Definition and estimation of intervention effects in complex systems: Gender equity in academia
Embargo until
Date
2016-05-02
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
Even though gender equity in academia has been extensively studied, female faculty are still consistently hired at lower ranks, paid lower salaries and promoted less frequently than men.
Previous work has focused on the individual faculty member as a study unit and, in most applications, on a single academic reward or representation outcome. However, existing approaches are insufficient to assess equity at institutional level for single-institution studies, from a causal inference perspective.
How do differential gender practices in awarding salaries and ranks affect institutional measures of prestige and investment? In this dissertation we developed a simulation-based approach to estimate and conduct inference for gender equity outcomes defined at institution level and investigate how gender disparities along individual careers contribute to institutional measures.
The statistical challenge in addressing these issues corresponds to the estimation of higher-level causal effects in complex systems for which only one observation of the outcomes of interest is available. The methods proposed combine faculty-level models that describe the academic career with a university-level, i.e., an aggregate-level, definition of causal effect.
We applied these methods to simulated data based on an existing university. We found that the simulated institution does not deviate significantly from gender neutrality in terms of departures from the institution and total time in higher ranks for female faculty in 2005-2013. However, under a counterfactual gender-neutral scenario, the total compensation paid to female faculty over these 9 years would have been 2.8\% higher (95\% CI [1.2\%, 4.4\%]). The main determinant of this disparity is the significantly lower initial salaries for female faculty, with women earning 6.0\% less on average at-hire than otherwise similar men.
This analysis aims to complement individual-level gender equity studies with an institutional perspective, to aid in the achievement of a more gender-neutral structure in academia. Furthermore, the methods proposed have wide applications to other complex systems and designs, such as health agencies networks, pharmaceutical market dynamics and transportation systems.
Description
Keywords
complex system, causal inference, simulation, regression model, gender, academia