Assessment of resampling methods for causality testing: A note on the US inflation behavior

Area of Science:

• Time series analysis
• Causal inference
• Econometrics

Background:

• Granger causality is crucial for understanding time series relationships.
• Existing methods often struggle with conditional dependencies in multivariate settings.
• Partial Transfer Entropy (PTE) offers a model-free approach to assess causality.

Purpose of the Study:

• To assess various resampling methods for testing the null hypothesis of no Granger causality.
• To evaluate the performance of Partial Transfer Entropy (PTE) under different independence assumptions.
• To investigate the causal relationship between US CPI, money supply, and crude oil prices.

Main Methods:

• Utilized six resampling methods combining time-shifted surrogates and stationary bootstrap with varying independence levels.
• Examined the empirical null distribution, size, and power of PTE tests.
• Introduced a seventh method using contemporaneous stationary bootstrap for distribution approximation.
• Applied PTE and resampling methods to US CPI, money supply, and crude oil growth rates.

Main Results:

• Resampling methods become more conservative as independence in the surrogate/bootstrapped series increases.
• A consistent causal link was found: crude oil price changes influence inflation, conditional on money supply.
• This oil-inflation link post-1986 is not explained by traditional cost-push theories.

Conclusions:

• The choice of resampling method impacts the conservatism of Granger causality tests.
• Partial Transfer Entropy is a robust tool for uncovering conditional causal relationships in economic time series.
• Crude oil prices exert a significant influence on US inflation, beyond monetary factors, in the post-1986 era.