Can Mental Illness Lead to Dismissal? From a Causal Machine Learning Perspective

Abstract

Causal inference has been used extensively in health, economics, policy research, and other fields. With the introduction of the Neyman-Rubin framework in 1974, more scholars began to realize that correlation between variables is not equivalent to causation, and therefore, relying too heavily on statistical correlation methods to model can lead to serious theoretical flaws. In this paper, we use data on the work of people with mental illness to analyze whether society treats people with mental illness equally, use propensity score matching (PSM) method to reduce the dimensionality of covariates, and estimate the causal effect of having a mental illness on hiring rates. Our study shows that the covariates can all be well balanced after the implementation of PSM and that employees with mental illness have a 5.8% greater likelihood of leading to dismissal compared to employees in the general population.