Invariant Causal Prediction for Sequential Data with Application to Monetary Policy

ETH Zurich researcher Niklas Pfister presents a methodology for inferring causal predictors from sequentially ordered data without requiring prior knowledge of data environments. The work provides statistical confidence bounds and asymptotic detection results for the novel causal inference technique. An application to monetary policy in macroeconomics demonstrates the method's utility.

Use Cases

• Detect instantaneous causal relations in multivariate linear time series based on the described invariance principle.
• Build models that remain invariant under interventions using only causal predictors, as discussed in the description.
• Apply the methodology to macroeconomic problems, such as analyzing monetary policy effects, as referenced in the description.

Strengths

• Methodology is supported by statistical confidence bounds and asymptotic detection results.
• Includes a concrete application to monetary policy in macroeconomics.
• The work is published as Open Access (green), facilitating reuse.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

ETH Zurich

Collection Method

Research methodology presented in an academic paper.