Administrative Errors and Race: Can Technology Mitigate Inequitable Administrative Outcomes?

Abstract Scholars have long recognized the role of race and ethnicity in shaping the development and design of policy institutions in the United States, including social welfare policy. Beyond influencing the design of policy institutions, administrative discretion can disadvantage marginalized clientele in policy implementation. Building on previous work on street-level bureaucracy, administrative discretion, and administrative burden, we offer a theory of racialized administrative errors and we examine whether automation mitigates the adverse administrative outcomes experienced by clientele of color. We build on recent work examining the role of technological and administrative complexity in shaping the incidence of administrative errors, and test our theory of racialized administrative errors with claim-level administrative data from 53 US unemployment insurance programs, from 2002 to 2018. Using logistic regression, we find evidence of systematic differences by claimant race and ethnicity in the odds of a state workforce agency making an error when processing unemployment insurance claims. Our analysis suggests that non-white claimants are more likely to be affected by agency errors that result in underpayment of benefits than white claimants. We also find that automated state–client interactions reduce the likelihood of administrative errors for all groups compared to face-to-face interactions, including black and Hispanic clientele, but some disparities persist.