Primary Submission Category: Causal Inference and Bias/Discrimination
Credible Evidence of Gender Discrimination Using Instrumental Inequality
Authors: Jiwoo Kim, Yongnam Kim,
Presenting Author: Jiwoo Kim*
Considering sex discrimination as a direct effect of sex on an outcome given a justifiable mediator, one of the significant challenges for a statistical approach to measuring sex discrimination is the presence of mediator-outcome confounding. As in Bickel et al.’s (1979) example of sex bias in UC Berkeley, which compares males’ and females’ acceptance rates in each department, the conventional approach is susceptible to potential confounders that may influence both department choice (mediator) and acceptance (outcome). This paper introduces a novel approach using the causal constraint of instrumental inequality to investigate sex discrimination. This method has the advantage of requiring neither the absence of mediator-outcome confounding nor specific functional forms or distributional assumptions. Applying the proposed approach to two sex discrimination cases in South Korea, this paper offers more credible statistical evidence of sex discrimination without relying on unrealistic assumptions.