Structural Nested Mean Models Under Parallel Trends with Interference

2024-05-20

AI Summary

Conventional difference-in-differences (DiD) methods rely on the “no-interference” assumption and thus yield biased estimates in the presence of interference—such as spillovers, network dependence, or cluster-level contagion. Method: We propose DiD-SNMMs, the first extension of structural nested mean models (SNMMs) to DiD settings with time-varying interference. DiD-SNMMs jointly identify dynamic direct and indirect effects that depend on both treatment history and covariates. The approach integrates causal graph modeling, explicit representation of cluster- or network-based interference structures, and inverse-probability weighting for estimation. Contribution/Results: Simulations demonstrate that DiD-SNMMs robustly estimate multilevel intervention effects, substantially improve identification of true spillover mechanisms, and enhance statistical efficiency—thereby overcoming fundamental theoretical and practical limitations of standard DiD in interference-prone settings.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Content-based information diffusion

Abstract

Despite the common occurrence of interference in Difference-in-Differences (DiD) applications, standard DiD methods rely on an assumption that interference is absent, and comparatively little work has considered how to accommodate and learn about spillover effects within a DiD framework. Here, we extend the so-called `DiD-SNMMs' of Shahn et al (2022) to accommodate interference in a time-varying DiD setting. Doing so enables estimation of a richer set of effects than previous DiD approaches. For example, DiD-SNMMs do not assume the absence of spillover effects after direct exposures and can model how effects of direct or indirect (i.e. spillover) exposures depend on past and concurrent (direct or indirect) exposure and covariate history. We consider both cluster and network interference structures an illustrate the methodology in simulations.

Problem

• Extends DiD-SNMMs to handle interference in time-varying settings
• Estimates spillover effects without assuming absence after direct exposures
• Models how direct and indirect exposures depend on exposure history

Innovation

• Extends DiD-SNMMs to handle interference
• Models spillover effects without assuming absence
• Considers cluster and network interference structures