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

TL;DR

This paper introduces the potential system, a nonparametric time series model that assesses causal effects of interventions over time, bridging the gap between time series causality and nonparametric causal inference methods.

Contribution

It provides a new nonparametric framework for causal inference in time series, applicable to various existing methods like impulse responses and SVARs, and establishes a foundation for future causal methods.

Findings

Provides a nonparametric model for causal effects in time series

Bridges time series causality with nonparametric causal inference

Lays groundwork for new causal analysis methods in time series

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.