Workers are among an organization’s most important assets, yet organizations do not have complete control over the workforces they assemble. My research examines how organizational workforces take shape, from the internal and external environments that constrain who becomes available and considered, to the evaluative processes that determine who is ultimately selected.
Stream I
Organizations can only select from available and visible candidates. This pool is shaped by existing workforce structures, local labor markets, and competing employers. This stream examines how these interdependencies constrain or enable the workforces organizations can build.
Stream II
Selection, too, does not happen in a vaccum. Prior judgments, including promotions, ratings, credentials, and algorithmic assessments, become inputs into later decisions. This stream examines how evaluators interpret inherited judgments in light of the conditions under which they were produced.
I introduce Hierarchical Demographic Configurations (HDCs) to capture the racial composition of supervisors, peers, and subordinates surrounding a focal position. A conjoint experiment shows that job seekers interpret diversity differently across these relations of authority and dependence, and that HDCs predict choices better than aggregate diversity measures. Agent-based simulations then show that modest tier-specific preferences, repeated across workers and time, are sufficient to generate persistent segregation even without discriminatory gatekeeping.
Using EEO-1 and Census data for 27,812 establishments employing more than five million workers, I examine how the racial composition of an establishment's immediate neighborhood relates to its workforce beyond the available local labor pool. Matched comparisons across adjacent neighborhoods and within the same firm across locations show persistent local demographic patterns, especially in customer-facing settings and for professionals in healthcare.
I examine how segregation among incumbent employers interacts with local pool demographic composition to constrain a new organization's ability to assemble a diverse workforce. The project treats organizations as localized, interdependent competitors for workers rather than as actors with complete control over their hiring prospects, asking how field-level composition conditions the demographic possibilities available to individual firms.
We develop a positive-deviance approach to diversity research. Rather than asking whether diversity is beneficial on average, we focus on the mechanisms that produce unusually inclusive relationships, advancement pathways, and high-performing diverse groups. The chapter shifts attention to the organizational conditions under which diverse workforces function particularly well.
Using EEO-1 workforce reports linked with Compustat financial data, we examine how ethnoracial diversity and internal segregation relate to firm performance. The project extends my structural stream from the formation of demographic arrangements to their organizational consequences.
We develop the concept of context-amplified signals to explain why the same career achievement can carry different informational value depending on who earned it and where. Longitudinal promotion data and a complementary experiment show that women's prior promotions carry greater value when earned in male-dominated settings, where evaluators infer that success was harder won. Demographic context therefore travels with credentials and shapes how later evaluators read them.
I model evaluation as an ordered production process rather than a single act. The theory distinguishes eight loci — from specifying outcomes and criteria through curating signals, assessing qualities, ordering alternatives, and allocating outcomes. Through evaluative inheritance, each locus passes an altered evaluative product to the next. The framework helps explain how distinct mechanisms can produce similar disparities, why bias may appear at some stages but not others, and why improving judgments does not necessarily improve outcomes.
We extend evaluative inheritance to human–algorithmic handoffs in hiring. As first-round algorithmic assessments increasingly precede human review, later evaluators must decide whether to endorse or override a judgment they did not produce. In partnership with a large nonprofit, we are developing a field study of how AI assessments of nontraditional candidates are interpreted, accepted, and revised by subsequent human evaluators.