TEA-Time: Transporting Effects Across Time

2026-03-07

AI Summary

This study addresses the limited temporal generalizability of treatment effects estimated in randomized controlled trials, which are typically confined to specific time points and difficult to extrapolate. The authors propose a time-transfer framework grounded in the assumption of separable time effects, decomposing the transfer average treatment effect into an observable average treatment effect and a time ratio. Identification is achieved through either repeated trials or a shared treatment group across time periods. This work establishes the first systematic theory for causal effect transportability along the time dimension and introduces two identification strategies accompanied by corresponding doubly robust, semiparametrically efficient estimators. Theoretical analysis and Monte Carlo simulations confirm the nominal coverage of these estimators, while empirical A/B testing demonstrates that, under strong assumptions, the shared-treatment-group strategy substantially improves estimation efficiency—albeit with a trade-off between bias and variance.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Intelligent Robots: State Estimation

Application Category

• User Modeling, Personalization and Recommendation: Studies of user behavior, including longitudinal effects of personalized systems
• Security and Privacy: Large-scale security measurements
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Treatment effects estimated from randomized controlled trials are local not only to the study population but also to the time at which the trial was conducted. We develop a framework for temporal transportation: extrapolating treatment effects to time periods where no experiment was conducted. We target the transported average treatment effect (TATE) and show that under a separable temporal effects assumption, the TATE decomposes into an observed average treatment effect and a temporal ratio. We provide two identification strategies -- one using replicated trials comparing the same treatments at different times, another using common treatment arms observed across time -- and develop doubly robust, semiparametrically efficient estimators for each. Monte Carlo simulations confirm that both estimators achieve nominal coverage, with the common arm strategy yielding substantial efficiency gains when its stronger assumptions hold. We apply our methods to A/B tests from the Upworthy Research Archive, demonstrating that the two strategies exhibit a variance-bias tradeoff: the common arm approach offers greater precision but may incur bias when treatments interact heterogeneously with temporal factors.

Problem

• temporal transportation
• treatment effect
• time extrapolation
• randomized controlled trials
• transported average treatment effect

Innovation

• temporal transportation
• transported average treatment effect
• doubly robust estimation
• common treatment arms
• separable temporal effects