Randomized controlled trials have become one of economists’ most trusted tools for identifying discrimination. By holding everything constant except a single randomized factor—such as whether a résumé signals a Black- or white-sounding name—RCTs can cleanly measure unequal treatment in hiring, housing, lending, and other decisions. But in a new essay, economists Dania V. Francis, Rob Gillezeau, Damon Jones, and Angelino Viceisza argue that the same experimental clarity is far harder to achieve when the target is not an isolated act of bias, but systemic racial discrimination.
The authors begin by laying out a conceptual framework for systemic racial discrimination grounded in established definitions of systemic racism. In their account, “systemic” refers to patterns of disadvantage generated and reinforced by institutions, policies, and social structures—forces that operate across multiple domains and over time, and that can persist even when no single decisionmaker is overtly prejudiced. This distinction matters for methodology: RCTs are designed to manipulate well-defined treatments, yet systemic discrimination often looks like an interconnected set of mechanisms rather than a single “switch” researchers can flip.
With that model in mind, the essay evaluates where randomized experiments can realistically contribute—and where their reach is limited. One central issue is scale. To directly test the effects of systemic racial discrimination, an experiment would need to manipulate structural conditions broadly, something that is typically infeasible, ethically fraught, or both. The authors therefore consider smaller-scale experimental strategies that, while not recreating a society’s full institutional architecture, might still illuminate particular components of systemic processes.
Another approach is to run trials across different structural settings, comparing results across environments that vary in policy regimes, institutional practices, or other contextual features. Such designs can help researchers learn how the same intervention behaves under different structural constraints, shedding light on how systems shape outcomes even when the “treatment” is identical.
The essay also explores experiments that alter what participants notice rather than what structures do. Researchers may be able to randomize the salience of systemic forces—for instance, by making institutional barriers more visible—to see whether awareness changes behavior, decisions, or expectations. Relatedly, the authors discuss information-based interventions aimed at reducing systemic discrimination, testing whether better information can weaken discriminatory patterns embedded in institutions or markets.
The salience of racial identity itself can also be manipulated in experimental settings, allowing researchers to examine how identities interact with institutions and rules. Finally, the authors highlight the importance of strategic responses: people and organizations adapt to systemic constraints, and RCTs can sometimes measure those adaptations—how individuals navigate barriers, how firms respond to incentives, or how institutions change practices in the face of scrutiny.
Even as the essay maps out these possibilities, it underscores the boundaries of what randomization can reveal. Because systemic racial discrimination is complex, multi-layered, and historically rooted, RCTs may miss key dynamics or fail to capture long-run and cross-domain effects. The authors point to the need for other empirical strategies to complement experiments and address the gaps RCTs leave behind.
The work was supported financially by the Alfred P. Sloan Foundation through J-PAL North America. The authors also acknowledge feedback and assistance from multiple scholars and research staff, and they note that the views expressed are their own and do not necessarily represent those of the National Bureau of Economic Research.
Why it matters
Policymakers and researchers increasingly look to “what works” evidence, often privileging RCTs, to guide anti-discrimination efforts. This essay clarifies that while experiments can reveal important mechanisms, they may be poorly suited to testing the full machinery of systemic racism. The message is pragmatic: to understand and reduce structural inequities, evidence will likely need to come from a broader toolkit than randomization alone.