2024-09-29-美联储-菜单成本经济中的非线性动力学_来自美国数据的证据（英）_20页_502kb

Summary

Executive Summary

This paper questions the ability of standard menu cost models to explain two key empirical features of inflation and price dynamics: the large dispersion in price changes and the increase in the fraction of price changes during periods of high inflation. We test three variants: (1) the canonical Golosov and Lucas (2007) model with a fixed menu cost; (2) a Calvo-plus model with probabilistic free price changes and a fixed cost (NS); and (3) an economy with a uniform distribution of menu costs (Uniform). The GL model captures the increase in the fraction of price changes with inflation but fails to match the dispersion in price change sizes, predicting nearly neutral money. The NS and Uniform models better match price dispersion but predict a nearly constant fraction of price changes, reducing nonlinearity. We conclude that standard menu cost models cannot simultaneously reproduce the dispersion in the size of price changes and the increase in the fraction of price changes over the business cycle. An important challenge is to develop models that reconcile these two features.

Methodology

• Three Model Variants:

  • GL (Canonical): Fixed menu cost.
  • *NS: Probabilistic free price changes (1−λ) plus a fixed cost.
  • *Uniform: Free price changes plus a uniform distribution of costs (0 to ξ̄).

• Data: Uses U.S. inflation data and fraction of price changes from 1979–2014.

• Calibration: Targets average price change frequency (0.105), median price change size (0.075), and inflation statistics (mean 3.4%, standard deviation 2.6%).

Key Findings

1. Fraction of Price Changes

• GL Model: Predicts significant but underwhelming increases in the fraction of price changes with inflation, capturing about ~32% of the observed rise.
• NS and Uniform: Predict a nearly constant fraction of price changes, with fluctuations below 12%.

2. Dispersion in Price Changes

• GL: Produces a bi-modal distribution of price changes, missing the observed unimodal distribution with higher kurtosis (e.g., 2.4–3.3 vs. 1.5 data kurtosis).
• NS and Uniform: Better match the distribution (e.g., Uniform generates a unimodal distribution with lower kurtosis).

3. Nonlinear Dynamics

• GL: Predicts significant nonlinearity (e.g., small monetary shocks have CIRs ~14–8% of Calvo baseline), but small real effects due to poor dispersion fit.
• NS/Uniform: Predict linear output dynamics despite larger real effects (CIRs ~36–58% of an otherwise similar Calvo model), as the fraction of price changes remains largely unchanged.

4. Monetary Policy Implications

• GL:Nearly neutral money due to poor price change dispersion.
• NS/Uniform:Stronger monetary non-neutrality (output responses ~36–58% of Calvo) but reduced nonlinearity.

Conclusion

Standard menu cost models fail to reproduce both empirical regularities. Directions for future research include models with multi-product firms and low "misallocation within the firm," as proposed by the authors in complementing work (Blanco et al., 2024a). The paper highlights the trade-off between matching price change dispersion and the fraction-price change relationship.