Some history

1. The Canadian Experience with Core InflationThérèse Laflècheat the VIII Annual Seminar, Central Bank of Brazil11 August 2006

2. Some history • Targets for reducing inflation were adopted in February 1991 • The target was defined in terms of year-over-year growth rate of total CPI • The operational guide for policy was expressed in terms of CPI excluding food and energy prices, adjusted for the effects of indirect tax changes (CPIXFET) • In 2001, CPIX replaced CPIXFET as the official core inflation measure used as an operational guide to policy

3. Definition of CPIX • Measure that excludes 8 of the most volatile components from total CPI… • Fruit • Vegetables • Gasoline • Fuel Oil • Natural gas • Intercity transportation • Tobacco • Mortgage-interest costs • …and the effect of changes in indirect taxes on the remaining components 3 energy components out of 4 are excluded and only 2 food components out of 8 are excluded Larger coverage: CPIX includes 84% of the total CPI basket compared to 74% for CPIXFET

4. Why a core inflation measure ? • Core inflation is a better indicator of current underlying inflationary pressures than total CPI • Total CPI can be misleading because some of its components are volatile • Short-run fluctuations caused by temporary shocks tend to reverse themselves fairly quickly • It would be neither feasible nor desirable for monetary policy to counter such temporary movements • Attempting to do so would increase economic volatility

5. Why a core inflation measure ? • Core inflation is a useful short-term operational guide for the conduct of monetary policy • Given lags in the effects of interest rate changes on output and inflation, monetary policy must be forward-looking • Core inflation isolates the underlying trend to which total inflation will return • Many indicators are used to estimate future inflation; core inflation is one relatively simple indicator, straightforward to measure on a timely basis

6. Why a core inflation measure ? • Core inflation is a good communication tool • Analyzing and comparing the evolution of core and total inflation in the Monetary Policy Report helps the general public understand and assess the monetary policy decisions of the Bank

7. Core inflation is only a guide • As long as core and total inflation share a common long-term trend, focusing on core inflation is consistent with targeting total inflation • The Bank wants to provide an environment with low, stable and predictable inflation which helps consumers make the best decisions • Total CPI is the best measure to estimate the evolution of the cost of living for Canadian households. Thus, the Bank targets inflation measured with total CPI • Focusing on core inflation does not mean that the Bank is not concerned about inflation in the components excluded from the core measure • Core inflation is simply a convenient guide to achieving its objective of controlling total inflation

8. Alternative measures of core inflation monitored by the Bank • Order statistics : MEANSTD, WMEDIAN • 54 components of the CPI • Cross-sectional distribution of the 54 price changes (12-month percentage) at each month over a specific time period • Order statistics : calculated using the ordered distribution • WMEDIAN is the weighted median of this distribution • That is, the value that separates the ordered distribution in two parts, with the sum of the weights of each part being equal to 50 per cent • No component is excluded from this measure

9. Alternative measures of core inflation monitored by the Bank • MEANSTD is the weighted average of the distribution where values above and below the average by at least 1.5 times the standard deviation are eliminated • The components excluded from MEANSTD differ from one month to the next < 1.5 S.D. from de mean > 1.5 S.D. from de mean mean

10. Alternative measures of core inflation monitored by the Bank • Double weighted measure : CPIW • Each of the 54 components is assigned a weight inversely proportional to its variability • The variability is defined as the standard deviation of the change in relative prices (the difference between the price change of a component and the inflation rate as measured by total CPI) • The weight is the reciprocal of this measure of variability and is applied to a component in addition to its initial weight, which represents the importance of this component in consumer spending No component excluded from this measure

11. Alternative measures of core inflation monitored by the Bank • Why the Bank monitors several measures of core inflation: • CPIX was chosen because it had some advantages over the other measures, but it remains an imperfect estimate of underlying inflation • When all measures convey the same message, it is reasonable to assume that the Bank has a relatively good estimation of underlying inflationary pressures • When the measures diverge, however, there is more uncertainty about the trend, requiring a close examination of the reasons for the disparity

12. Episode of large uncertainty about inflation trend Total CPI and core inflation (CPIX)

13. Episode of large uncertainty about inflation trend CPIX and alternative measures of core inflation

14. Episode of large uncertainty about inflation trend Two temporary supply shocks

15. Episode of large uncertainty about inflation trend CPIX excluding electricity price (CPIX9) and CPIX excluding electricity price and auto insurance premiums (CPIX10)

16. Re-evaluation of the various core inflation measures • Do the criteria used to select CPIX in 2001 still favour the measure today ? • Updated period (extended up to December 2005) • Measures evaluated • CPIX, CPIXFET, CPIW, MEANSTD, WMEDIAN • Two new measures : • CPIX9 (CPIX excluding electricity) • CPIX10 (CPIX9 excluding auto insurance premiums)

17. No unique concept or definition of core inflation • Basic concept of core inflation • Total inflation can be separated into two components: • The core part, representing the underlying trend of inflation as shaped by the pressure of aggregate demand against capacity • The non-core part, which reflects price movements caused by temporary shocks or relative price changes

18. Criteria for evaluating the measures of core inflation • Component volatility • A component subject to temporary shocks is more volatile than one that is not • Volatility of the core measures • If core inflation represents the underlying trend in inflation, it should be less volatile than total inflation • Absence of bias between the core measures and total inflation • Over time, the core measures and total inflation should have followed the same trend • Ability to predict total inflation • In the short run, total inflation can diverge temporarily from its trend, but it should return to it in the long term

19. Criteria for evaluating the measures of core inflation • All these empirical criteria are time sensitive • The component volatility and the volatility of the core measures are dependent of the time period over which they are calculated • Assessments of the bias and the predictive power of the core measures may yield different results over different time periods

20. Criteria for evaluating the measures of core inflation • In addition to the empirical criteria mentioned above, the central bank must take into account some practical criteria • A core inflation measure should be timely. • It is also desirable that core measure not be revised. Revisions would reduce its usefulness both as decision and communication tool. • For communication purposes, a core measure should be credible, easy to calculate and to explain.

21. Component volatility • Component volatility • In 2001, the eight components excluded from CPIX were among the most volatile, both in terms of their standard deviation and their frequency of exclusion from MEANSTD. • At the end of 2005, the same eight components were still among the most volatile components. However, two other components, electricity prices and auto insurance premiums, have exhibited more volatility and, hence, became potential candidates for exclusion • Volatility itself is not sufficient: there must also be some theoretical foundations for the exclusion of a component

22. Component volatility • Theoretical foundations for excluding the eight components from CPIX • Fruit and vegetables are frequently affected by weather-related disturbances to crop. The other food components are not often exposed to temporary shocks and, hence, are not very volatile • Prices of gasoline, natural gas and fuel oil are clearly influenced by the world price of oil, which is very volatile and determined on foreign markets • Intercity transportation includes airfares, which are significantly influenced by oil prices

23. Component volatility • Theoretical foundations for excluding the eight components from CPIX • Prices of tobacco vary substantially with changes in excise taxes, which are clearly supply shocks • Mortgage-interest costs are a special case: a rise in interest rates from monetary policy aimed at reducing inflation would raise mortgage-interest costs, adding temporarily to inflation. This would send the wrong signal about the underlying short-run trend of inflation. Many central banks exclude this component from their core measure for this reason

24. Component volatility • The case of electricity prices • Changes in Alberta and Ontario electricity markets have made these prices more volatile in 2002 and 2003 • Since that time, electricity prices have returned to a more stable path • An eventual deregulation of provincial electricity markets could lead to more volatility in these prices • For now, the exclusion of electricity prices would be premature. However, the bank will continue to monitor CPIX9 closely

25. Component volatility • The case of auto insurance premiums • Automotive insurance premiums rose by 30 per cent between January 2002 and January 2003 • This increase is explained by a combination of rising claim costs and the bursting of the technological bubble in 2001: rising claims were no longer offset by solid investment returns for insurance companies • Following this surge, auto insurance premiums declined because many provinces imposed price reductions on insurance companies

26. Component volatility • The case of auto insurance premiums • Because of the way insurers adjust their premiums, this kind of event observed in the past may recur. • However, auto insurance premiums movements are not volatile but could rather be characterized as subject to infrequent persistent shifts in their trend. • CPIX10 is biased relative to total CPI. • It is not appropriate at this time to exclude auto insurance premiums from the official core measure. The Bank should continue to monitor CPIX10.

27. Volatility of the core measures • The core measures should be less volatile than total inflation • We look at : • standard deviation • coefficient of variation (standard deviation divided by the mean) • mean of the absolute monthly change in year-over-year inflation • Based on the standard deviation and the coefficient of variation, CPIX is the less volatile measure • Based on the mean of absolute changes, CPIW is the less volatile measure • The order statistics exhibit the largest volatility, no matter how volatility is measured.

28. Absence of bias • To identify a bias, we compare the unconditional means of the various core measures with that of total CPI • The means of CPIX, CPIW and CPIX9 are in the same range as the mean of total CPI • The means of CPIXFET and MEANSTD are lower than the others, but the differences are not statistically significant • WMEDIAN and CPIX10 have the lowest means and they are significantly statistically different from the mean of total CPI. • Conclusion : WMEDIAN and CPIX10 are biased relative to total CPI; all of the other measures, including CPIX, have followed the same trend as total inflation over the last 14 years.

29. Ability to predict total inflation • Predictive power • Given that a core measure captures inflation’s underlying trend, deviations between core inflation and total inflation should be reversed in the future, with total inflation coming back to underlying inflation : pt+iTotal – ptTotal = a + b (ptcore – ptTotal) + mt (1) pt+icore –ptcore = a + B (ptTotal – ptcore ) + vt (2) • If core inflation is above total inflation, total inflation must have been hit by a shock that will be reversed. Total inflation should therefore be expected to increase in the future (b > 0), but core inflation should be unaffected (B = 0). • If the restriction a = 0 and b = 1 holds, equation (1) collapses to : pt+iTotal = ptcore + mt and in that case, core inflation is an unbiased predictor of total inflation Cogley’s equations

30. Ability to predict total inflation • Predictive power • All the core measures are unbiased predictors of total inflation • Deviations between core and total inflation are not persistent and total inflation moves toward core rather than vice versa • CPIW has the best overall performance among the core measures

31. Practical criteria • Timeliness: all the core measures monitored by the Bank can be calculated the day inflation data are released by Statistics Canada • Only CPIW is subject to historical revisions. One of the weights of this measure is the reciprocal of the standard deviation of the change in relative prices. The standard deviation is calculated over a specific time period and hence, would change with the time period selected. • However, the extension of the period on which the CPIW weights are calculated did not lead to significant historical revisions of the time series. Moreover, it would be possible to update the weights every few years and link the new series to the old one in order not to change the history

32. Practical criteria • To be credible, a measure must be understood and accepted by the public. • The exclusion-based measures, like CPIX, are the most easy to understand • The order statistics require some statistical knowledge to be understood; moreover, their movements are not as easy to explain as those of CPIX • CPIW is also more difficult to understand than CPIX. However, its concept is actually quite simple; the difficulty lies in the calculation of the special weight based on the volatility of the components

33. Main Conclusions • CPIX still satisfies all of the empirical and practical criteria. No core measure significantly outperforms it. Moreover, CPIX is familiar to the public and well accepted. • Although CPIW is more sophisticated than the official core measure, it does slightly better than the other core inflation measures on purely statistical grounds. Therefore, CPIW merits closer study. • All the core measures contain information on the underlying trend in inflation and are particularly useful for identifying the source and nature of persistent but temporary shocks that affect inflation and push it away from the target.

34. Main Conclusions The Bank will retain CPIX as its official measure of core inflation, but will continue to monitor the other core inflation measures closely, in particular CPIW.