Paper: arXiv 2306.01511

Authors: Jozef Barunik, Lukas Vacha

Abstract

We propose a novel framework for modeling time-varying persistence in economic time series, allowing for smoothly evolving heterogeneity in shock dynamics. We leverage localized regression techniques to flexibly identify changes in persistence over time, offering a data-driven alternative to traditional parametric models. We applied this methodology to U.S. inflation and stock market volatility data and found substantial persistence variations that align with key macroeconomic events and market conditions. The results reveal previously undetected pockets of predictability and provide significant increases in out-of-sample forecast accuracy. These findings have important implications for economic modeling, forecasting, and policy analysis.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces a solid non-parametric approach to time-varying shock persistence using localized regression, bridging advanced econometrics with practical market applications. Its empirical evaluation on US inflation and equity volatility demonstrates clear out-of-sample forecasting gains, making it both mathematically sound and empirically useful.

Research Flowchart

  flowchart TD
    A[Goal: Model Time-Varying Persistence in Economic Time Series] --> B[Data Inputs: U.S. Inflation & Stock Market Volatility]
    B --> C[Methodology: Localized Regression Techniques]
    C --> D[Identify Dynamic Persistence & Macro Event Alignments]
    D --> E[Outcomes: Hidden Predictability & Higher Forecast Accuracy]