2015年-IMF国际货币组织全球_A_Volatility_and_Persistence_19页_430kb

Summary of "A Volatility and Persistence-Based Core Inflation"

Core Content

This working paper introduces a new approach to constructing core inflation measures by simultaneously considering volatility and persistence of price changes, rather than relying solely on excluding volatile items or using only persistence or expenditure weights.

Main Views

• Core inflation is traditionally defined as headline inflation excluding volatile items such as food and energy.
• However, volatility and persistence are not equivalent. Some volatile items are also highly persistent, and excluding them may lead to the loss of important inflationary information.
• The paper argues that persistence in price changes is crucial for understanding inflationary pressures and potential second-round effects.
• A new methodology is proposed that combines both volatility and persistence in the weighting of price items, leading to a more accurate measure of core inflation.

Key Information

1. Volatility and Persistence Relationship

• Volatile items are often assumed to be temporary and hence not informative, but this is not always the case.
• Figure 2 shows that volatility and persistence are weakly correlated, indicating that they are distinct dimensions of price changes.
• Items like "Tubers, Roots and Legumes" and "Green Vegetables" are highly volatile but low in persistence, suggesting they should be excluded.
• Conversely, some items traditionally excluded (e.g., oil) are highly persistent, implying they should be included in core inflation measures.

2. Core Inflation Measures Proposed

• The paper proposes three types of core inflation measures:
  • Single-weighted:
    • SW-V: Based on volatility
    • SW-P: Based on persistence
  • Double-weighted:
    • DW-VE: Combines volatility and expenditure
    • DW-PE: Combines persistence and expenditure
  • Triple-weighted:
    • TW-PVE: Combines volatility, persistence, and expenditure
    • TW-RMSE: Uses root mean square error from an AR model to adjust weights

3. Data and Estimation Strategy

• The study uses Brazilian IPCA data for 52 items from January 1995 to December 2012.
• Volatility is calculated using a 48-month moving window.
• Persistence is measured using the sum of autoregressive coefficients from an $\mathrm{AR}(q)$ model, with $q$ selected by the Schwarz criterion.
• Expenditure weights are used to compute the headline inflation index.
• The weights are normalized to sum to 1 for each period.

4. Performance of Core Inflation Measures

• The volatility-persistence-based measures outperform traditional measures in terms of tracking the trend of headline inflation.
• The RMSE and MAD (mean absolute deviation) of the new measures are significantly lower than those of other core inflation measures, indicating better accuracy.
• IPCA-EX, IPCA-DP, and IPCA-MS are compared against the new measures, with the latter showing superior performance.
• The TW-PVE and TW-RMSE measures, which incorporate all three dimensions (volatility, persistence, and expenditure), are among the top performers.

5. Forecasting Ability

• The paper tests the predictive power of the core inflation measures using a multi-step forecasting model (equation 13).
• R² values are low, indicating that both headline and core inflation over the previous quarters are not strong predictors of future inflation.
• However, the TW-PVE and TW-RMSE measures show significant predictive ability, outperforming traditional measures in all forecast horizons (2, 3, and 4 quarters ahead).

Conclusion

The paper concludes that traditional core inflation measures, which exclude volatile items or rely only on persistence or expenditure, fail to capture the full picture of underlying inflation. A volatility and persistence-based core inflation measure provides a more accurate and informative alternative, especially in the context of Brazil, where economic shocks are frequent and commodity prices are volatile yet persistent.

The proposed triple-weighted measure (TW-PVE) is highlighted as the most robust and informative among the new measures, offering better performance in both tracking inflation trends and forecasting future inflation.

References

• Cecchetti (2006)
• OECD (2005)
• Bryan and Cecchetti (1993)
• Gavin and Mandal (2002)
• Cutler (2001)
• Demarco (2004)
• Babestskii et al. (2007)
• Bilke and Stracca (2008)
• Rangasamy (2009)

Figures and Tables

• Figure 1: Shows the long-lasting price movements of oil, indicating its high persistence.
• Figure 2: Displays scatter plots of persistence and volatility for IPCA items, showing weak correlation and identifying items that should be excluded.
• Table 1: Descriptive statistics of various core inflation measures, showing biases and volatility.
• Table 2: Deviations from trend inflation, showing the relative performance of core inflation measures.
• Table 3: Forecasting results using a multi-step model, indicating the predictive accuracy of the new core inflation measures.