Alternative measures to evaluate core inflation

1
Alternative measures to evaluate core inflation

• Banco Central do Brasil seminar on core inflation
  and price indices harmonization

2
Conceptions of core inflation

• Persistent element of inflation, excluding
  transient supply shocks.
• Generalized element of inflation excluding
  relative price shocks.
• Empirical issue is generalized inflation
  persistent?

3
Alternative approaches

• Variety of approaches discussed in Roger (1998)
  and Wynne (1999).
• Empirical focus has been primarily on estimating
  the generalized component of CPI inflation.
• Time-series versus cross-sectional approaches to
  generalized inflation.

4
The stochastic approach

• Price movements in each period viewed as core
  inflation plus relative price shocks.
• If relative price shocks distributed normally,
  sample mean inflation is unbiased, lowest
  volatility estimate of generalized, core
  inflation.
• But distribution rarely normal typically high
  kurtosis and right-skewed.

5
High kurtosis, right skewness
6
Implications of high kurtosis

• Mean inflation rate an inefficient (volatile)
  estimator of generalized inflation.
• 3 main approaches to trying to get a better
  estimator
• Exclusion-based measures (e.g. CPI ex volatiles)
• Volatility-adjusted measures (neo-Edgeworthian)
• Robust measures (e.g. trimmed means).

7
Skewness/asymmetry

• Right-skewness indicates asymmetry extreme price
  rises larger than extreme price declines.
• Can lead to undesirable downward bias to core
  measures if symmetry assumed.
• Alternative sources of positive skewness
• Arithmetic versus log-normal price change
• Infrequent price adjustment.

8
Log-normal price changes

• If price changes are log-normal, percentage
  changes will show chronic right-skewness.
• Skewness independent of inflation rate.

9
Infrequent price adjustment

• Reasons for infrequent price adjustment
• Seasonal supply
• Infrequent measurement of prices
• Infrequent adjustment of administered prices
• Stickiness in private sector prices.
• Theory has focused on 4th factor, but evidence
  mainly for 1st 3rd factors.
• Important consequence asymmetry and bias no
  longer independent of level of inflation.

10
Dealing with asymmetry

• Log-normality
• Fix the CPI. Increase use of geometric averaging
  in CPI construction.
• Infrequent price adjustment
• Seasonal adjustment of CPI components.
• For residual asymmetry, use of asymmetric
  trimming.

11
Evaluation of measures

• For public accountability purposes
• All 3 types of measures available on timely
  basis.
• All 3 transparent verifiable.
• Minimal bias vs. CPI feasible.
• Exclusion-based easier to explain than others.
• For internal policy formulation purposes
• Volatility forecasting properties very
  important - Marques et al (2000) criteria, not
  Bryan Cecchetti.
• Performance in structural models forecasting.

12
Which measures are best?

• Evidence from Asia-Pacific suggests
• Exclusion-based measures always show poor
  performance.
• Volatility-adjusted measures smooth, but exclude
  signal as well as noise.
• Trimmed means only work well if over 20.
• Efficiency gains over CPI about as expected from
  Roger (2000) 30 to 50 percent reduction in
  volatility of changes in quarterly inflation.

13
Observations

• Construction of core measures
• Inter-temporal aggregation should be linked to
  policy decision-making
• Cross-sectional aggregation should be linked to
  external accountability needs
• Consider use of seasonally-adjusted data
• Interpolation between discrete points may be
  useful in trimmed means.

14
Observations

• Use of core measures
• Variety of measures for public accountability
• Best technical measure for policy formulation.
• Generalized, transient shocks
• VAT type shocks relatively easy to remove
• Exchange rate shocks how and whether to remove
  a policy issue, not just a technical one.