When are time series predictions causal? The potential system and dynamic causal effects

2026-03-20

AI Summary

This study addresses the nonparametric identification and estimation of dynamic causal effects of an intervention at a specific time point on future outcomes in time series settings with covariates. To this end, the paper introduces a novel “potential system” framework that integrates the potential outcomes paradigm with a dynamic systems perspective, thereby unifying nonparametric cross-sectional causal inference and time series causal analysis for the first time. This approach provides a rigorous causal foundation for widely used methods such as local projections, impulse response analysis, and time series experiments. Moreover, it establishes a general and identifiable framework for causal time series analysis, offering theoretical support and enabling effect estimation across a broad range of existing and emerging methodologies.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Action, Change, and Causality

Application Category

• User Modeling, Personalization and Recommendation: Studies of user behavior, including longitudinal effects of personalized systems
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

The potential system is a nonparametric time series model for assessing the causal impact of moving an assignment at time $t$ on an outcome at future time $t+h$, accounting for the presence of features. The potential system provides nonparametric content for, e.g., time series experiments, time series regression, local projection, impulse response functions and SVARs. It closes a gap between time series causality and nonparametric cross-sectional causal methods, and provides a foundation for many new methods which have causal content.

Problem

• causal inference
• time series
• dynamic causal effects
• nonparametric methods
• potential outcomes

Innovation

• potential system
• nonparametric time series
• causal inference
• dynamic causal effects
• time series experiments