Regime-Dependent Distributional Effects: A Causal Framework for Aggregate Counterfactuals

• In Event: Causal Inference for Sociology

Abstract

Sociologists routinely make causal claims about how individual-level experiences contribute to inequality, heterogeneity, and polarization – claims about distributions rather than averages. Yet causal inference in sociology overwhelmingly targets individual-level average treatment effects. I argue that aggregate claims embed a second, rarely formalized counterfactual: how would the population distribution differ under an alternative regime of treatment allocation? I introduce the Effect of Treatment Regime on Aggregate Distributions (ETRAD), a regime-centered distributional estimand that formalizes this question. The ETRAD treats distributional functionals such as variance or polarization indices as outcomes and defines counterfactuals over both treatment prevalence and assignment mechanisms. Under standard causal assumptions, these quantities are identified via the distributional g-formula. An application to college completion and moral attitudes shows that meaningful individual-level effects need not translate into substantial distributional change; aggregate consequences are regime-dependent rather than mechanically implied by average effects. The ETRAD clarifies how micro-level causal effects scale, or fail to scale, into macro-level structure and encourages sociologists to treat allocation rules themselves as objects of causal inquiry, rather than taking observed regimes as given.

Author

• Kevin Kiley, North Carolina State University