Paper short abstract:

Drawing on qualitative data, this paper examines the epistemic crisis that the US Census Bureau faces while trying to enroll stakeholders in conducting the census. These efforts are unexpectedly triggering an epistemic corruption of democracy's data infrastructure and the federal statistics project.

Paper long abstract:

Since 1790, the US has conducted a decennial census to reapportion political representatives. While this data has never been perfect--and while there have always been actors invested in manipulating the data-- mechanisms are innovated each decade to improve the data quality.

The Census Bureau created a "trusted messengers" program to improve data quality by involving community stakeholders. This effort simultaneously widened the range of actors that were invested in how the census works, triggering epistemic tensions. The bureau leverages statistical techniques to get the best count possible; community-based stakeholders expect the bureau to directly count everyone.

This tension came to a head during the 2020 census as partisan politics, scientific developments, and attempts to be resilient during a global pandemic collided. While the statistical quality of the 2020 census data was on par with earlier decades, the perceived quality was abysmal and the legitimacy of the operation was contested. Many stakeholders who served as "trusted messengers" for the census during the 2000 and 2010 cycles played a key role in delegitimizing the census in 2020. Moreover, they continue to do so. Disconnects between the Census Bureau and its stakeholders continue to magnify a "crisis in expertise" (Eyal).

Drawing on ethnographic fieldwork and interviews, this paper examines the epistemic crisis that the Census Bureau faces while trying to enroll stakeholders in creating democracy's data infrastructure. While "trust in numbers" (Porter) has always been illusory, this paper interrogates the consequences of epistemic corruption of the federal statistics project.