🤖 AI Summary
This paper addresses causal inference under network interference in a finite population. We propose an “agnostic” causal parameter—the Average Partial Causal Effect (APCE)—that requires no structural model assumptions and is identifiable and consistently estimable from a single network randomized experiment. Methodologically, we develop a structural-model-agnostic identification framework based on exposure mapping invariance, integrate inverse-probability weighting with design-based finite-population inference, and leverage local smoothness of equilibrium models to interpret APCE as a derivative-weighted average response—extending the Local Average Treatment Effect (LATE) paradigm. Our key contributions are: (i) the first definition of a globally interpretable, identifiable, and estimable intervention response measure under minimal interference assumptions; (ii) a departure from structural modeling, achieving both robustness and interpretability; and (iii) exact recovery of a derivative-weighted average of the response function under smooth equilibrium conditions, establishing a novel paradigm for network causal inference.
📝 Abstract
In contrast to problems of interference in (exogenous) treatments, models of interference in unit-specific (endogenous) outcomes do not usually produce a reduced-form representation where outcomes depend on other units' treatment status only at a short network distance, or only through a known exposure mapping. This remains true if the structural mechanism depends on outcomes of peers only at a short network distance, or through a known exposure mapping. In this paper, we first define causal estimands that are identified and estimable from a single experiment on the network under minimal assumptions on the structure of interference, and which represent average partial causal responses which generally vary with other global features of the realized assignment. Under a fixed-population, design-based approach, we show unbiasedness and consistency for inverse-probability weighting (IPW) estimators for those causal parameters from a randomized experiment on a single network. We also analyze more closely the case of marginal interventions in a model of equilibrium with smooth response functions where we can recover LATE-type weighted averages of derivatives of those response functions. Under additional structural assumptions, these"agnostic"causal estimands can be combined to recover model parameters, but also retain their less restrictive causal interpretation.