The Missing Links:
Superlative CPI Inflation and Inflation Expectations in Canada
Department of Economics
C.D. Howe Institute Policy Conference
Canada’s Monetary Policy Regime After 2011
4 November 2008
Missing Links between the Instrument and the Goal:
1. Accurate measurement of inflation: the target.
2. Accurate forecasting of inflation: the operational guide.
1. Target the inflation rate in a new chained, superlative CPI.
2. Use a measure of expected/forecasted inflation (not core) as an operational
1. The Target
Inflation-targeting central banks use some version of a CPI.
The Canadian CPI is a Laspeyres index and so uses weights from the past.
Weights are updated every 4 years. The last update was in May 2007 using 2005
Appealing features: monthly frequency; not revised; just a 3-week delay.
Biases in the CPI
A. Substitution bias
B. Quality changes
C. Omissions in the user-cost measure for housing.
How large is the bias? The Bank estimates an average of 0.6%.
A sawtooth pattern is possible in the bias as weights become stale.
1. The Bank could revisit and report on bias on a regular timetable.
2. Statistics Canada could update its CPI basket more frequently than every 4
3. Statistics Canada could improve its treatment of owner-occupied housing.
Substitution bias can be reduced significantly with a superlative price index such
as the geometric mean of Paasche and Laspeyres, named Fisher’s ideal price index.
The challenge is to forecast or estimate Paasche weights.
Models to follow: US C-CPI-U and Canadian CPIC
Figure 1: Inflation in the US CPI-U and C-CPI-U
US CPI-U and C-CPI-U Inflation
2002 2004 2006 2008
Source: Bureau of Labor Statistics, www.bls.gov/cpi/data.htm
Figure 2: Inflation in the CPI and CPIC
CPI and CPIC inflation
2002 2004 2006 2008
Source: Bank of Canada Banking and Financial Statistics, July 2008, tables H3 and H8.
Do delays, revisions, and frequency matter for the target index?
4. Statistics Canada could estimate a monthly, chained superlative index, with a
5. The Bank could target the 12-month superlative CPI inflation rate.
Under recommendation 5, the Bank would wait an extra month. But 12-month
inflation is not difficult to forecast 1-month ahead. And 0.6—1.0% is a large bias
as a proportion of the 2% inflation target.
Should the Bank target a stability index rather than a cost-of-living index?
Central banks have targeted the price of gold or the price of a US dollar.
Recent research suggests they should seek price stability (to minimize price-
adjustment costs) yet allow changes in relative prices.
If sector A has flexible prices and sector B has sticky ones, then target inflation in
sector B and let sector A adjust.
Generally, place more weight on sticky sectors in the target index. e.g. wages.
The catch: these designs are based on volatility outcomes not necessarily frictions.
2. The Operational Guide
The Bank needs an operational guide because of the delayed effects of changes in
Core inflation is designed, in part, to be easy to explain, by omitting some volatile
The goal is a core index that forecasts future headline inflation.
But in Canada core and headline inflation may not have a common trend ...
Figure 3A: CPI and CPIX Figure 3B: Headline and Core Inflation
headline and core inflation rates
CPI and CPIX
2002 2004 2006 2008 2002 2004 2006 2008
In designing an inflation forecast you would not tie your hands to reweighting or
excluding elements of the CPI.
No one claims that core is the only thing that forecasts headline inflation.
(Though predictable departures of headline inflation from 2% are rare under
inflation targeting, and may imply insufficient policy reaction.)
Yet when you add other things as bases for forecasting, then different signals can
An arms-length measure of expected or forecasted inflation might be better. That
aggregates information to forecast.
Inflation expectations also partly determine the path of inflation. And they
respond to changes in the overnight interest rate.
Measuring Expected Inflation
A. Indexed bonds
Like the UK break-even inflation rate or the Cleveland Fed’s TIPS-based measure.
But in Canada real return bonds have a 30-year maturity and are not very liquid.
Like the SPF in the US or the survey used by the Banco Central do Brasil.
Forecasts must be for fixed horizons not calendar years.
The Bank of Canada Business Outlook Survey could be refined to give one index.
Prompt availability, no revisions, and monthly frequency do matter for the
Expected or forecasted inflation can be constructed with these properties.
It plays a large role in the consensus approach to monetary policy and in TOTEM.
1. The Government of Canada could issue real-return bonds with maturities of
1, 2, and 5 years.
2. We could survey professional forecasters. A CIRANO-JDI project ...