COVID-19 and race: Protecting data or saving lives?

Abstract

This article uses the COVID-19 pandemic to demonstrate how our understanding of ethnic inequalities could be improved by greater use of algorithms that infer ethnic heritage from people’s names. It starts from two inter-connected propositions: the effectiveness of many public sector programs is hampered by inadequate information on how differently different ethnic groups behave, and anxiety over how to discuss matters to do with race inhibits proper evaluation of methodologies which would address this problem. This article highlights four mindsets which could benefit from challenge: the officially sanctioned categories by which ethnic data are tabulated are too crude to capture the subtler differences which are required for effective communications; while self-identification should continue to drive one-to-one communications, it should not preclude the use of more appropriate methods of recording ethnic heritage when analyzing data for population groups; public servants often fail to recognize the limitations of directional measures such as the Index of Multiple Deprivation as against “natural” classifications such as Mosaic and Acorn; and in their quest for predictive accuracy statisticians often overlook the benefit of the variables they use being “actionable,” defining population groups that are easy to reach whether geographically or using one-to-one communications.