Systemic Discrimination in Multi-Stage Decision Processes: Theory, Measurement, and School Disciplinary Policy

• In Event: Measurement Challenges and Innovations

Abstract

Traditionally, quantitative research on discrimination has examined direct bias, net of factors that may themselves be shaped by prior discrimination. This approach assumes that biases are additive and linearly separable, suggesting that intervening on the most biased decision makers is the most effective path toward equity. However, emerging work reconceptualizes “systemic discrimination” as an outcome of interdependent systems, such as sequential decision processes, where the unjust effects of earlier decisions propagate to later decisions.

This paper shows how sequential decision-making amplifies the long-term impact of initial discrimination. It shows how the total inequity faced by disadvantaged groups can be underestimated by traditional measurement methods that fail to account for the interdependence of sequential biased decision making. It also derives measurable conditions under which it is optimal to intervene on an earlier stage actor, even when the later stage actor is far more biased. In effect, this research explores how to counteract systemic discrimination by leveraging its own structural dynamics.

The empirical evidence of this approach uses school disciplinary data on referrals and suspensions from the North Caroline Education Research Center. However, the results generalize to any multi-stage decision process, including that which constitutes the entire criminal justice system.

Author

• David McMillon, Emory University