UC Berkeley

This dissertation applies quantitative methods to the study of how race and ethnicity are socially constructed in quantitative research. This reflexive application of method reveals the contribution to constructionist theory that is possible from thoughtfully applied quantitative reasoning. In part, this contribution is the elaboration of the specific ways in which racial and ethnic data are constructed, providing details to the broad constructionist rhetoric that has been critiqued for being overly grand and substantively shallow. The dissertation suggests that both theory and method can be improved through the integration of both - by demonstrating how quantitative methods can incorporate constructionist perspectives while also demonstrating specific ways in which race and ethnicity data are constructed.

The first substantive chapter tackles racial identification in survey interviews, demonstrating that complexities of specific interactions can shape how individuals identify their own race, revealing the many ambiguities present when relying on a single identification in a survey. The second chapter describes how missing responses are filled in, producing an apparent universality of racial self-identification in the decennial census and American Community Survey. By exploring the patterns in these allocations, this chapter demonstrates how statistical descriptions of communities are reliant upon decisions made by bureaucracies rather than simple aggregations of self-identifications. The final chapter explores the mathematical details of statistical models that have been applied to studying intersectional differences. This chapter reveals that the current gold-standard approaches, while well motivated, unintentionally incorporate assumptions which are anathema to their theoretical perspectives. Finally, the chapter suggests that the application of Gaussian Process models may have similarly favorable mathematical properties, while avoiding contradictory assumptions.