Evaluating alternative measures of Core Inflation for Argentina Laura D’Amato Juan Sotes Paladino and Lidia Sanz BCRA VIII Annual Seminar, Central Bank of Brazil August 2006
Motivation Argentina: return to active monetary policy after more than ten years under a hard peg Have an adequate assessment of inflation trend crucial for monetary policy decisions But CPI inflation influenced by transitory movements on relative prices Core Inflation as a monetary phenomenon (Silver, 2006) How to separate signal of inflation from temporary noise
Motivation Problem: dramatic change in RER due to large depreciation of the peso in January 2002 Change in relative prices rather permanent Strong initial impact of nominal depreciation on “core inflation” followed by very different paths of tradable and non-tradable goods prices The cross section distribution of CPI components’ price changes suffered sharp alterations due to abrupt depreciation of the currency
Tradable and Non-tradable goods prices
Goods vs Services Prices
Cross-section distribution of CPI components’ price changes ( )
Cross-section distribution of CPI components’ price changes ( )
Cross-section distribution of CPI components’ price changes ( )
Outline 1.Core Inflation as a signal extraction problem 2.Exclusion-based methods 3.Limited influence estimators 4.Evaluation 5.Conclusions
1.Core Inflation as a signal extraction problem Persistence-based reweighting Blinder (1997): Core Inflation as the persistent component of CPI Cutler (2001) reweigh the CPI using persistence based weights Two weighting vectors: Jan-93/Jan04 and Jan-93/Oct-00
1.Persistence based Core Inflation Index (CPIP)
1.Core Inflation as a signal extraction problem Persistence based reweighting: Results Food items, usually excluded, have high weight (meat, diary products and eggs) Regulated Services lose weight due to sporadic changes in prices Items with high imports component are excluded Have to revise persistence weights
1.Core Inflation as a signal extraction problem Volatility-based reweighting Marques, Neves y Sarmento (2000): weights based on standard deviation of price changes of CPI components relative to mean CPI inflation
1.Volatility-based Core Inflation Index (CPIV)
2.Exclusion-based methods Systematic criteria: Exclude Items simultaneously less persistent and more volatile and according to: i.Variance of relative price changes (whole period) ii.Marginal contribution of each item to the volatility of CPI inflation iii.Ordering according to Principal Components Analysis
2.Exclusion-based methods Systematic criteria: Criterion i : 11.7% of CPI components excluded, basically items with high imports component Criteria ii and iii: exclusion of almost all Regulated Services Items (14% of CPI)
2.Exclusion-based methods: Ex F&E and Core INDEC
3.Limited influence estimators Brian and Cechetti (1997): Cross-section distribution of CPI components’ price changes non-normal and with heavy tails Strong case for the use of robust estimators of the first moment of cross-section distribution Cross-sectional distribution of 65 components of CPI Truncate at the 7.5%. 15% and 20% level Underestimates CPI inflation Frequency of exclusion informative about volatility of CPI components’ price changes. Can use as exclusion criterion
3.Limited influence estimators
4.Evaluation If interested on Core Inflation for prediction: Granger Causality Poor results: only volatility-based index performs adequately Marques et al (2001): Core inflation as an atractor In Argentina CPI is I(0) only persistence- based index performs rather well
5.Conclusions Persistence and volatility based indicators better performance for prediction Exclusion indicators and limited influence estimators can be useful to have an assessment of current “core inflation” Trimmed means probably useful when cross section distribution of price changes is unstable Correct bias of trimmed means (Silver 2006) Having more observations of new monetary regime more work needed. Recalculate persistence weights Principal component analysis for post-crisis period
Recent evolution of Core Inflation measures