Synthetic Blips: Generalizing Synthetic Controls for Dynamic Treatment Effects

TL;DR

This paper extends causal inference for panel data to settings with dynamic, sequential treatments under unobserved confounding by introducing a low-rank latent factor model. It generalizes synthetic control methods to dynamic contexts through synthetic blip effects, enabling a backwards-induction identification strategy that expresses unit-specific mean outcomes as additive combinations of treatment blips across time. The authors develop linear time-varying (LTV) and linear time-invariant (LTI) latent-factor formulations, derive identification results, and propose the SBE-PCR estimator with consistency guarantees. An empirical application using Korean firm data demonstrates individualized dynamic treatment effects and derives both retrospective and prospective optimal treatment allocation rules for export-finance programs. The framework unites synthetic-control-type approaches with dynamic treatment-effect models, offering practically scalable identification in adaptive observational settings and actionable policy insights for sequential interventions.

Abstract

We propose a generalization of the synthetic control and interventions methods to the setting with dynamic treatment effects. We consider the estimation of unit-specific treatment effects from panel data collected under a general treatment sequence. Here, each unit receives multiple treatments sequentially, according to an adaptive policy that depends on a latent, endogenously time-varying confounding state. Under a low-rank latent factor model assumption, we develop an identification strategy for any unit-specific mean outcome under any sequence of interventions. The latent factor model we propose admits linear time-varying and time-invariant dynamical systems as special cases. Our approach can be viewed as an identification strategy for structural nested mean models -- a widely used framework for dynamic treatment effects -- under a low-rank latent factor assumption on the blip effects. Unlike these models, however, it is more permissive in observational settings, thereby broadening its applicability. Our method, which we term synthetic blip effects, is a backwards induction process in which the blip effect of a treatment at each period and for a target unit is recursively expressed as a linear combination of the blip effects of a group of other units that received the designated treatment. This strategy avoids the combinatorial explosion in the number of units that would otherwise be required by a naive application of prior synthetic control and intervention methods in dynamic treatment settings. We provide estimation algorithms that are easy to implement in practice and yield estimators with desirable properties. Using unique Korean firm-level panel data, we demonstrate how the proposed framework can be used to estimate individualized dynamic treatment effects and to derive optimal treatment allocation rules in the context of financial support for exporting firms.

Figures (9)

• Figure 1: DAG that is consistent with the exogeneity conditions implied by the definition of $\mathcal{I}^{\bar{d}^T}$.
• Figure 2: DAG that is consistent with the exogeneity conditions implied by the definition of $\mathcal{I}^d_t$. From time step $t + 1$, the action sequence $(D_{n, t + 1}, \dots, D_{n, T})$ can be adaptive, i.e., dependent on the observed outcomes $\{Y_{n, t}\}_{t \in [T]}$ (depicted by the red arrow).
• Figure 3: DAG that is consistent with the exogeneity conditions implied by the definition of $\tilde{\mathcal{I}}^d$. From time step $2$, the action sequence $(D_{n, 2}, \dots, D_{n, T})$ can be adaptive, i.e., dependent on the observed outcomes $\{Y_{n, t}\}_{t \in [T]}$ (depicted by the red arrows). Hence, there is no non-adaptive period for these units.
• Figure 4: Dynamic Treatment Effects on Export Values
• Figure 5: Potential Export Values Under Front-, Even- and Back-Loading Treatment Schedules

Theorems & Definitions (61)

• Definition 1: Latent factors
• Definition 2: SI donor units
• Theorem 1: SI Identification Strategy
• Remark
• Proposition 1
• Proposition 2
• Theorem 2
• Remark
• Theorem 3
• Theorem 4