Statistical Implications of Inflation Targeting
Chapter

2 The Bank of Canada’s Approach to Inflation Targeting

Author(s):
Carol Carson, Claudia Dziobek, and Charles Enoch
Published Date:
September 2002
Share
  • ShareShare
Show Summary Details

The fundamental appeal of a monetary framework that directly targets inflation as the single monetary policy objective stems from its apparent simplicity and clarity. Surely the general level of prices is an easy thing to measure, and thus it must also be an easy matter—even for the person in the street—to have a clear view of whether the central bank is or is not achieving its objective. Unfortunately, as is the case in most areas of economics, an inflation-targeting framework is not quite as simple to understand or to operate as it seems at first glance.

This chapter deals with Canada’s experience with the informational and statistical requirements for operating a direct inflation-targeting regime. We will argue that, while the basic objective of the direct inflation-targeting approach is relatively simple and transparent, the informational, statistical, and econometric requirements for implementing such a regime are quite demanding. These demands turn out, however, to be a virtue of inflation targeting, since they impose a great deal of discipline on the process by which a central bank formulates its monetary policy and undertakes its monetary operations. Over time, this discipline can help both to improve the implementation of monetary policy and to strengthen the credibility of the inflation-control target itself.

In what follows, we will argue these propositions by building on Canada’s experience with inflation targeting. The chapter first outlines the reasons that lay behind Canada’s adoption of an inflation-targeting regime back in 1991 and sketches the basic statistical elements of an inflation-targeting regime. A subsequent section describes in detail the monetary transmission mechanism. The second half of the chapter then turns to statistical aspects that are particularly important to an inflation-targeting country, including the information required to assess policy developments, how to measure the policy stance, market expectations, the credibility of monetary policy, and accountability over the longer term. The final section provides some concluding thoughts.

Why Is Canada a Direct Inflation Targeter?

Canada’s monetary regime has two elements: a freely floating exchange rate and a monetary policy that seeks to achieve an explicit, quantitative inflation-control target as the key nominal anchor of the system. These features are mutually consistent—indeed, they are two sides of the same coin.

First, the floating exchange rate. Although the North American economy is highly integrated, Canada’s production structure is quite different from that of the United States—or indeed, most other advanced industrial countries. In particular, because primary commodities are much more important in Canada’s production and export mix, our terms of trade exhibit quite a strong negative correlation with those of the United States, or other G-10 industrial countries taken as a group (Table 2.1 and Roger, 1991).1 Thus, economic shocks typically affect the Canadian economy differently from the way they impact on other industrial economies (hereafter called asymmetric shocks). A floating exchange rate provides Canada with a highly sensitive and flexible economy-wide price that fosters a smoother adjustment to economic shocks than would occur if the exchange rate were fixed. If the exchange rate were not allowed to “flex,” the adjustment to terms of trade or other shocks would have to be accomplished by movements in domestic nominal wages and prices that are much less responsive to shocks in the short run. Mainly as a result, Canada has operated a floating exchange rate for all but 8 of the past 50 years.

Table 2.1.Group of 10: Absolute Terms of Trade Correlation1
1969–20001969–19791979–19891989–2000
G-102U.S.G-10U.S.G-10U.S.G-10U.S.
Canada−0.478−0.458−0.676−0.627−0.054−0.215−0.607−0.571
United States0.5031.0000.5061.0000.5271.0000.0611.000
Japan0.6580.6630.6570.6570.7140.7180.2930.517
Germany0.6680.5350.7180.7280.8780.482−0.0970.425
France0.8610.6990.8800.9320.9590.5430.5850.340
Italy0.8350.6760.8220.7770.9730.5110.4250.537
United Kingdom0.1610.2400.3710.591−0.961−0.7620.1130.101
Netherlands0.5510.5540.7650.6560.4090.6140.0380.004
Sweden30.5800.5160.5570.6340.7950.418−0.2710.232
Switzerland40.6800.5410.7740.6620.6100.477
Australia0.1580.2100.2370.2400.0730.289−0.633−0.321

We thank our colleagues A. Ghent and J. Murray for making this table available.

Calculated as the trade-weighted average terms of trade of the other G-10 countries plus Switzerland (except for 1989–2000, where weights were calculated excluding Switzerland). Weights used for each period were derived from bilateral trade shares for 1974–75, 1984–85, 1994–95, and average of these three periods for 1969–2000 time period.

To 1999 only.

To 1987 only.

We thank our colleagues A. Ghent and J. Murray for making this table available.

Calculated as the trade-weighted average terms of trade of the other G-10 countries plus Switzerland (except for 1989–2000, where weights were calculated excluding Switzerland). Weights used for each period were derived from bilateral trade shares for 1974–75, 1984–85, 1994–95, and average of these three periods for 1969–2000 time period.

To 1999 only.

To 1987 only.

The second element of Canada’s monetary regime is inflation targeting. After experimenting with other monetary policy targets, Canada moved to a direct inflation-targeting regime a decade ago, in February 1991, with a joint announcement by the Bank of Canada and the Government of Canada. The basic reason why Canada became, and remains, a direct inflation-targeting country is that we needed a nominal anchor. Furthermore, we have found that the transparency and credibility of this nominal anchor tends to enhance the efficiency with which a flexible exchange rate operates to absorb asymmetric macroeconomic shocks.

Canadian policymakers learned this lesson the hard way during the 1970s and 1980s. In particular, the experience of the 1980s, when there appeared to be a “vicious circle” of exchange rate depreciation leading to higher inflation and further depreciation, eventually taught us that a flexible exchange rate system works well only if all transactors know that—whether the exchange rate floats up or down—the central bank remains committed to maintaining low, stable inflation and possesses the instruments, the technical expertise, and the economic data needed to do so.

Although the Bank of Canada allows first-round pass-through of exchange rate movements, business people know that under the inflation-target regime, they cannot pass on second-round increases in imported input prices owing to an exchange rate depreciation, or grant wage increases beyond productivity gains in the hope that inflation and currency depreciation will help maintain their profit margins. They realize that any appreciation in the nominal exchange rate is equivalent to a decrease in competitiveness, so that they must cut other costs to maintain profit margins. Hence, one key element in making a flexible exchange rate work efficiently is to implement a monetary policy that successfully delivers low and stable inflation, even when the exchange rate is moving up or down to maintain external balance in the face of shocks.

Broadly speaking, Canada’s experience over the past decade shows that an inflation target provides an effective nominal anchor for monetary policy by reducing and stabilizing inflation expectations. The average rate of inflation in the 20 years preceding the introduction of the inflation targets in February 1991 was over 7 percent. As Figure 2.1 shows, since the introduction of the inflation target, the 12-month increase in the consumer price index (CPI) excluding the effects of changes in indirect taxes has been within the target range for most of the period. The overall CPI inflation rate has been within the band almost as much, except when it has been affected by large price shocks or large changes in the rates of indirect taxation.

Figure 2.1.Canada: Total CPI and CPI Adjusted for the Effect of Changes in Indirect Taxes1

(Year-over-year percentage change)

Sources: Bank of Canada; and Statistics Canada.

1CPI, consumer price index.

In sum, at the Bank of Canada we believe that direct inflation targeting makes it clear to market participants that the long-run objective of monetary policy is to maintain low, stable, and predictable inflation. Because the current and future price levels are known with a higher degree of certainty, the exchange rate can operate more effectively (that is, nominal exchange rate movements are real movements) so that it adjusts the Canadian economy in the face of asymmetric shocks. In our view, this is the fundamental rationale for Canada’s regime of a floating exchange rate combined with direct inflation targeting. Moreover, low inflation itself is a good thing since it leads to better economic outcomes.

Basic Statistical Elements of Inflation Targeting

Several aspects of the direct inflation-targeting framework indicate some of the statistical requirements for operating such a regime. These include (1) the decision regarding the specific price index to target; (2) the decision regarding the numerical target; (3) the range (if any) around the target level; and (4) the decision on whether there is any role for an operational target that differs from the main target.

Which Statistical Price Index Should Be the Basis for Target Inflation?

The three obvious contenders for the statistical index to be used as the basis for the target in a direct inflation-targeting regime are the CPI or one of its variants, the GDP deflator, or the price deflator for consumer expenditures. The GDP deflator is the broadest measure of prices and, as such, would seem to be the ideal candidate. Movements in the GDP deflator, however, may be influenced by factors that are not relevant to households. For example, Canada’s GDP deflator is strongly influenced by changes in the terms of trade for the primary and semiprocessed commodities that make up a relatively large share of our output. Such commodity price movements, while they affect production decisions, typically have little connection with the changes in the overall domestic price level that are relevant to domestic households.2 This makes the GDP deflator inappropriate as the price index to target in Canada’s inflation-control regime. In principle, the chain-price index for consumer expenditures on goods and services is a good summary measure of the overall price level that is relevant to households. But, like the GDP deflator, it is derived implicitly from nominal and real estimates of spending, is published only at quarterly intervals with a considerable lag, and is often revised significantly.

One measure of prices that we have not seriously considered at the Bank for the target is one based on industrial producer prices. There is such an index available on a monthly basis in Canada. And one would expect that it should share some common links with consumer prices. Nevertheless, empirical work conducted at the Bank indicates that changes in the producer price index are not closely related to those in consumer prices.

There are solid statistical reasons for preferring the CPI as the appropriate measure of price changes in an inflation-targeting country. It is typically available on a monthly basis, so it can be tracked at regular and frequent intervals; it is published in a timely fashion with short lags; the initially published data are rarely revised; and it is widely known, both for the previous reasons and for its use in adjusting various contracts for inflation. Thus, on grounds of transparency and timeliness, the CPI is the preferred price index to target. Of course, as a measure of movements in the general price level it has the disadvantages of any Laspeyres (base-weighted) index. But in Canada the weights of the index are usually revised on the basis of a detailed household expenditure survey every four years. Thus, for all of these reasons the CPI has been chosen as the statistical index that the Bank of Canada targets. For similar reasons, virtually all other inflation-targeting countries use a CPI-based target as well.

What Is the Appropriate Inflation Rate to Target?

When the inflation-target regime began in 1991, the first quantitative objectives were introduced as inflation-reduction targets, with the ultimate quantitative definition of what constitutes price stability to be determined at a later date. The initial goal was to reduce inflation (defined as the year-to-year change) to progressively lower levels—first to 3 percent by the end of 1992, then to 2.5 percent, then to 2 percent by the end of 1995—to ensure a climate favorable for long-lasting economic growth. The time horizon for achievement of the initial target, 22 months, reflected the understanding at that time of the lags in the monetary policy transmission process. The targets have been extended three times, to the end of 1998, to the end of 2001, and to the end of 2006. The latest agreement maintains the 2 percent midpoint of a 1 to 3 percent target range.

A key factor to consider in choosing the quantitative target for inflation is the size of the bias in the chosen statistical measure relative to movements in the true cost of living. The CPI, for example, has at least five main sources of bias: base weighting (commodity substitution bias), formula bias, imperfect adjustment for quality changes, new goods (sometimes divided up into new products and new brands), and outlet substitution bias. It is a difficult empirical exercise to obtain a tight estimate of the size of this bias. At the Bank of Canada, we have conducted considerable research into this area, and our best statistical estimate for Canada at this time on the upper bound of the bias in the all-items CPI is 0.7 percent, with the mean being 0.5 percent (Crawford, 1998).

With respect to other factors, considerable research needs to be undertaken to determine the appropriate target for the rate of change of the price index. For example, the factors examined leading up to the recent extension of the inflation target in Canada included downward rigidity in nominal wages, the inability of nominal interest rates to go below zero, and the potential risk of a costly deflation. The Bank concluded that the weight of the evidence for each of these issues did not suggest that they would create problems if the target for inflation in Canada remained in the 1 to 3 percent range (see Amirault and O’Reilly, 2001; Bank of Canada, 2001b; Crawford, 2001a; and Crawford and Wright, 2001). In addition, the Bank has examined the evidence for a lower target but found that it continues to be difficult to quantify the longer-run benefits of such a change (see O’Reilly, 1998; and Ragan, 1998).

What Are the Roles of the Point Target and the Target Range?

One of the decisions that has to be made when implementing an inflation-targeting regime is whether to focus on a point target—say, 2 percent—or on a range around a point target—for example, 2 percent plus or minus 1 percent. The role of the point target is clear: to provide an anchor for expectations around a specific number. Indeed, if the central bank achieves this point target on average over time, then expectations of inflation over the medium term will become anchored on this objective. This may also lead to less uncertainty about the future price level. But the difficulty in forecasting the future rate of inflation of the targeted price index, the lack of precision in predicting the effects of monetary policy actions on measured inflation, and the fact that large relative price changes can affect the CPI—even over the medium term—suggest that a range needs to be set around the target if the target is the total CPI (Longworth and Freedman, 2000). Choosing the width of such a range requires careful consideration of the statistical evidence and an assessment of the risks associated with choosing a wider or narrower band than justified by the historical data on changes in the index chosen to be targeted. Ultimately a judgment has to be made, but it should be informed as much as possible by an understanding of how the economy is likely to evolve under inflation targeting—an understanding gained by analysis of a wide range of statistical information.

The practical trade-off confronted by central bankers in initially choosing the width of the range is clear: the wider it is, the higher is the probability that the central bank can keep the actual inflation rate within the band, but the less useful is the target in altering expectations of the long-run inflation rate and economic behavior more generally. And as will be noted below, the establishment of credibility—in the sense that expectations of the medium-term inflation rate converge on the chosen target—is a key element in enhancing the effectiveness of monetary operations.

In Canada, a range of 1 percent above or below the midpoint of the target range has been established to increase credibility. Thus, the target range is 1 to 3 percent, with a midpoint of 2 percent. While this range is, in fact, smaller than the variability of the overall CPI before the inflation targets were introduced, it encompasses outcomes for inflation that are considered the most likely to occur.3 The variability of inflation after the introduction of the target, however, has been quite low. The range should not be interpreted as the Bank’s indifference to the outcome, but rather as a reflection of the short-term uncertainty of outcomes stemming from unpredictable shocks to the price level. However, the boundaries of the range are “soft-edged.” If the rate of inflation moves outside the range, the Bank of Canada takes action to bring it back to its target, but the return will normally be gradual. Since the cause of a given shock is not always apparent, a gradual approach tends to reduce the variability in output and inflation outcomes.

In the event that the measured rate of increase in headline CPI falls outside the target range, the Bank has also committed itself to make use of its regular publications—the Monetary Policy Reports and Updates—to explain, using relevant statistical information, why inflation has deviated from the target midpoint, whether steps are needed to ensure that inflation moves back to the target midpoint, and when inflation is likely to return to the target midpoint. Of course, the speed with which inflation can be returned to target, absent shocks, depends on the various lags in the monetary policy transmission process and on the emphasis given to other factors like interest rate smoothing or short-term output stabilization.

The Distinction Between “Headline” and “Core” CPI

To emphasize trend movements in inflation rather than short-term fluctuations, our inflation target focuses on the year-to-year increase in the CPI. Although the inflation-control target is specified in terms of the “headline” (total) CPI, and the Bank’s responsibility to the public is for the change in the total CPI, for operational purposes the Bank of Canada uses as the key index for its monetary policy actions a measure that we refer to as “core” CPI. This measure excludes items whose prices have been shown statistically to be the most volatile, as well as the effect of changes in indirect taxes on the remaining components. We believe that this core index provides a better measure of the underlying rate of inflation for practical purposes, especially when there are relative price shocks that exert large transitory effects on the volatile components of the index.

Because it takes time for monetary actions to have any significant effect on inflation, focusing on core inflation is helpful for looking through short-run factors and focusing on the underlying trend that is likely to persist in the future (Macklem, 2001). For example, in early 2001 headline CPI inflation was at the top of the target band, and core inflation was at the midpoint—but a weakening pace of activity suggested an easing in the underlying trend of inflation, so we lowered the policy interest rate.

Another reason to focus on core inflation is to see through the first-round effects of changes in indirect taxes. They increase the price level in proportion to the tax increase.4 This raises the inflation rate temporarily, but once the price level has reached its new level, the rate of inflation is no longer affected. The Bank therefore operates its monetary policy so as to accommodate first-round effects of a change in indirect taxes. But it has indicated that it does not accommodate second-round effects that might occur from indirect tax hikes, such as allowing them to feed through into inflation expectations, or into subsequent changes in wages and prices of other goods and services.

Since, over an extended period of time, the “headline” CPI and “core” CPI measures have tended to move in a very similar fashion, and are likely to continue to do so in the future, achieving the target for headline CPI over the medium term can be done through an operational target such as core CPI (Figure 2.2). Nevertheless, if there were a sustained divergence between the movements in these two indices, it is changes in the headline CPI that would provide the basis for targeting over the longer term. In such a situation, the Bank would adjust its desired target for core CPI inflation so that the expected trend in headline CPI inflation would be centered in the inflation-control range (Macklem, 2001). Headline CPI inflation would remain the target, but core inflation would continue to guide policy actions to keep headline (or total) CPI inflation on target. Again, early 2001 is informative, where the divergence between headline and core inflation was interpreted as being temporary, and the pressure on underlying inflation as being in an easing direction.

Figure 2.2.Canada: Total CPI and Core CPI

(Year-over-year percentage change)

Sources: Bank of Canada; and Statistics Canada.

1Excludes eight most volatile components and the effect of indirect tax changes on remaining components.

From 1991 to 2001, for operational purposes the Bank used a statistical definition of core inflation that included the total CPI less the prices of food and energy, and the effects of indirect taxes on the remaining components.5 Over this period, the Bank has undertaken considerable analytical and empirical work to refine its statistical measure of core inflation, and with the renewal of the inflation-control target in May 2001, the Bank announced a new measure of core inflation. The previous measure was replaced with the CPI excluding the eight most volatile components as well as the effects of indirect taxes on the remaining components (Figure 2.3).6 Even though both measures show a similar pattern, the new measure of core inflation has several advantages over the former:

Figure 2.3.Canada: Two Measures of Core Inflation

(Year-over-year percentage change

Source: Bank of Canada.

1CPI excluding food, energy, and the effect of changes in indirect taxes.

2CPI excluding eight most volatile components and the effect of indirect tax changes on remaining components.

  • It includes 84 percent of the total consumer expenditure basket versus 74 percent in the old measure of core.

  • It excludes only the five food and energy components that are the most volatile. The previous measure excluded all food and energy components, even though not all of them were volatile. For example, the new measure now excludes only those most directly affected by supply shocks: fruit, vegetables, gasoline, fuel oil, and natural gas.

  • It excludes tobacco prices, which are volatile because of changes in tobacco taxes and other factors.

  • It excludes the costs of mortgage interest, which tend to rise when monetary policy is tightened and thus give a misleading signal of coming inflation trends.

  • It is less volatile than the old measure.

  • The new measure is better at explaining future changes in total CPI adjusted for the effects of changes in indirect taxes (see Macklem, 2001).7

Although the list of items that have the most volatile price movements has proven to be relatively stable in Canada over quite long samples (Hogan, Johnson, and Laflèche, 2001)—thus implying that the composition of the new core measure is also stable—changes in market structure or regulations could affect the behavior of price indices and could potentially change items that comprise the high-volatility group. This could affect our statistical measure of core inflation. Continued analysis and research is therefore required on the measurement of underlying inflation.

Time Horizon for Achieving the Target Inflation Rate and the Monetary Policy Transmission Process

Over what time horizon should a central bank seek to achieve its direct inflation target? This is one of the most difficult empirical questions for a country that wishes to operate such a regime.8 It is a truism of monetary theory that the lags in the effect of monetary policy on inflation are “long and variable.” A policy action taken today is transmitted gradually through the economy and may not exert its full, cumulative impact on the price level until many months or even a few years later. To have a credible inflation-targeting policy, it is essential to be able to give the public a clear idea—backed up by solid statistical analysis—of the time period over which these lags operate. If a given policy action takes two years to affect the CPI, a central bank cannot promise to achieve a particular target rate for inflation a year from now unless that rate is consistent with policy actions that were taken a year ago. Obviously, making the judgment on the nature of these time lags is essential to putting in place a credible inflation targeting strategy. But in view of the multiple channels of influence, this can involve more statistical and economic work than any other aspect of such a framework. Moreover, the nature of the channels through which we believe monetary policy operates has profound implications for the statistical data on which we concentrate our analysis in formulating our monetary policy actions.

To illustrate how complicated this is, consider the nature of the monetary policy transmission process. In a common “textbook” version of the monetary policy transmission process, there is a well-defined demand function for real money balances as a function of real economic activity, wealth, and the opportunity costs of holding real money balances—including the expected inflation rate minus the “own rate of return” on the relevant monetary asset. The long-run inflation rate then depends on the rate of growth of the nominal money stock relative to growth in the demand for real money balances.

This analysis implies that quantitative targets for the growth of some measure of the money stock may be useful to stabilize inflation, at least over the medium term. In this model, the rate of growth of the money stock depends on the growth of base money, which can be controlled by the central bank, and changes in the “money” multiplier. The central bank achieves the targeted rate of total money growth by injecting or withdrawing liquidity from the system to cause base money to grow at the appropriate rate.

Canada, and many other advanced industrial countries, tried to use quantitative targets for money growth as an inflation-control device in the 1970s and 1980s. While having these quantitative targets, however, the Bank of Canada regarded the short-term interest rate as the main channel for the transmission for monetary policy, with an M1 target used as a gauge for the appropriate level of interest rates (Duguay and Longworth, 1998). Unfortunately, beginning in the early 1980s monetary velocity functions became unstable in virtually all the industrial countries, including Canada, mainly owing to financial deregulation and innovation.9 In these conditions, central banks found themselves unable to achieve their inflation targets even when they achieved their money growth targets.10

In Canada, the paradigm now followed by the Bank of Canada is well known and contains three major linkages (see Longworth, 1999). The first is from the Bank’s target for the overnight interest rate to other financial variables: the term structure of market interest rates, rates set by financial institutions on their deposits and loans, and the exchange rate. The second linkage runs from financial variables to aggregate demand and the output gap. And the third set of linkages runs from the output gap, the exchange rate, and inflation expectations to inflation. These linkages are illustrated in Figure 2.4 and discussed further below.

Figure 2.4.Canada: Monetary Policy Transmission Mechanism

Source: Longworth (1999).

Financial Variables

The term structure of interest rates in Canada is heavily influenced by the world real term structure of interest rates, risk premiums related to Canadian government debts and deficits, and domestic inflation expectations. In general, however, a rise in nominal short-term rates (more recently, the target for the overnight rate) translates into a rise in nominal domestic rates on Government of Canada bonds all along the term structure, with typically smaller effects the further along the maturity spectrum one is looking at (see Muller and Zelmer, 1999). In addition, this is usually translated into corresponding movements in nominal rates on corporate securities of similar maturities and in the interest rates that are set by Canada’s chartered banks and other deposit-taking financial institutions on loans and deposit (see Clinton and Howard, 1994). It is very useful to know if the long-run relationship is one in which the changes in institution-determined rates move one-for-one with changes in similar market rates and how rapid the adjustment is. In Canada, it is quite rapid—typically much more rapid than in the United States, Canada’s largest trading partner.

In an open economy like Canada’s, knowledge about how the exchange rate responds to interest differentials with world interest rates or with the rates of major trading partners is also required to understand the transmission mechanism. Canada’s exchange rate responds in an important way to (unexpected) changes in interest rate differentials between Canada and the major industrial countries. Given Canada’s trade patterns, the United States has a very high weight (about 86 percent) in our trade-weighted exchange rate index.

Impact on Financial Variables and Aggregate Demand

For most inflation-targeting countries, an important part of the transmission mechanism is likely to be one in which aggregate demand depends on (1) the level of real interest rates (relative to equilibrium real interest rates), which affects consumption spending (especially on durable goods), and which also affects decisions to invest in housing and non-housing capital; and (2) the level of the real exchange rate, which affects exports and imports in the usual ways. Empirical evidence for Canada suggests that it is short-term (and perhaps short- to medium-term) interest rates that matter the most for aggregate demand.

The force with which interest rates and exchange rates impact on aggregate demand builds up slowly over time. In Canada, the maximum effect is reached after six to eight quarters. Direct credit, money channels, or both may be important in affecting aggregate demand in some countries, but there is little evidence yet for Canada. In addition to knowing how monetary policy affects aggregate demand, it is important to identify the major areas where shocks to demand will arise. Likely major sources of shocks are foreign demand, domestic fiscal policy, and the real world price of commodities (especially for countries that are major net exporters of commodities and thus have large terms of trade effects when real commodity prices change).

Impact on Inflation from the Output Gap, Expectations, and the Exchange Rate

A key part of the monetary transmission mechanism in Canada—and indeed in all industrial countries—is the effect of excess demand or supply in product markets on the rate of inflation. Thus, changes in monetary conditions work through their effect on the “output gap”—the level of output relative to capacity—to influence the inflation rate over time. Since the output gap requires a variable that cannot be directly measured, potential output, and a variable that is often subject to ongoing revision, actual output, construction of it is often difficult, and the resulting estimate subject to much debate. As a result, we review a large amount of related data for consistency with the estimate of the gap and use other approaches to assess the robustness of the implications of particular estimates of the output gap for the setting of monetary policy (see Longworth and Freedman, 2000).

Inflation expectations also play a very important role in the inflation process. Since there is no direct measure of expectations, they have often been proxied by using lagged inflation, surveys of inflation expectations, and the long-term nominal bond yield relative to the real yield on a real-return bond. Even though the backward-looking component in these expectations would likely remain important for some time after the introduction of an inflation-targeting regime, it would seem reasonable to expect that, over time, the achievement of the inflation target should help to condition the formation of the forward-looking components of price expectations. The challenge for the monetary authority is to determine, using available statistical information, when economic agents have moved to putting more weight on the inflation target when forming their expectations.

The Bank of Canada has considered inflation to be mainly influenced via the output gap (including the induced effect that the change in actual inflation thus produced has through the backward-looking element of inflation expectations). During the initial years of the inflation-targeting-regime, the Bank neither relied on expectations being influenced by the target nor on exchange rate pass-through to consumer prices, as was done in New Zealand,11 to guide inflation to its target. In recent years, as inflation expectations have become somewhat influenced by the targets themselves in Canada, exchange rate changes are now seen as affecting the price level, not the ongoing rate of inflation.

In such a framework, it has not proved very fruitful to try to distinguish the separate roles of money and credit in the monetary policy transmission mechanism. If there is a policy-induced increase in the short-term interest rate relative to the expected rate of inflation, then financial intermediaries will raise the interest rates on their lending and on their deposit liabilities, and borrowers from the financial system will reduce their demand for credit and enhance their holdings of money balances. These two effects will tend to slow the growth of aggregate demand for domestic output. In turn, slower growth of aggregate demand will reduce the rate of output growth relative to its capacity level, thereby tending to cause the rate of price inflation to fall. A reduction in the short-term interest rate will have the opposite effects.

Two important points about this mechanism follow:

  • First, in such a world, it is important to develop econometric models of all the main channels of the monetary transmission process in order to get some idea of the time it takes for a given change in the policy interest rate to impact on aggregate demand and inflation.

  • Second, the control of inflation via adjustments in the short-term interest rate will work better when long-term inflation expectations are well anchored by the credibility of the direct inflation-targeting mechanism itself.

Following Svensson (1997), the Bank can be characterized as adjusting its policy instrument—the overnight interest rate—to minimize a loss function.12 The components of this loss function are the gap between inflation and target inflation, and between output and potential output. The emphasis given to output stabilization in monetary policy decisions is reflected in the size of the coefficient on the output gap term. As Svensson has shown, the larger this weight, the more gradual the return to the inflation target when actual inflation has moved away from it.

The key point to note here is that the choice of the forecast horizon for inflation must be fundamentally consistent with the speed of return of inflation to its target, and both are functions of the structure of the economy and the weight given to output stabilization (the λ in the period loss function in footnote 12). Effectively, the central bank chooses its reaction function to minimize the loss from the increased output variability associated with achievement of the inflation target over time. Given the uncertainty about which model best represents the economy, robustness across models has much to recommend it. This is why at the Bank of Canada we examine a number of alternative statistical and econometric approaches for modeling the monetary transmission process.

The linkages by which a change in the policy interest rate is transmitted through to final demand for domestic output, to activity, and to prices are complex and involve long and variable time lags. Furthermore, given ongoing structural changes in the economy as well as the changing behavior of private market participants—particularly as experience under an inflation-targeting regime is gained—these lags must be constantly analyzed and reassessed. Moreover, for the inflation-targeting process to be credible, the horizon over which the target is to be achieved must be congruent with the lags in the monetary transmission process that are found by empirical means.

In Canada’s case, based on a careful assessment of all the available empirical work, it is the judgment of the Bank of Canada staff that, given the lags between changes in interest rates and their impact on the output gap, and lags between changes in the output gap and its impact on inflation, it seems reasonable to return to the inflation target in six to eight quarters. This is equivalent to deciding policy on the basis of the loss function for a given λ. Furthermore, in its strategy and communications, the Bank must be clear that it is setting its policy instrument now with a view to keeping its forecast of the future inflation rate as near as possible to the middle of its target range.

The Bank of Canada’s Framework for Evaluating Statistical Information

We come now to the specifics of the question: the statistical requirements for operating an inflation-targeting regime. The degree of sophistication of the statistical techniques depends on the data available. But it also depends on an assessment of how much statistical information is actually needed to get a handle on the country-specific features of the monetary policy transmission mechanism and other features of the economy. To make its monetary policy decisions, the Bank of Canada must develop its views on the likely future movements in aggregate demand and supply in the Canadian economy and of future inflation prospects, given its knowledge of the monetary policy transmission mechanism. To do this, the Bank makes use of a wide range of statistical information, econometric estimates and projections, as well as careful monitoring. In particular, the Bank uses staff monitoring, model-based projections, regional staff input, and analysis of financial market developments. While these approaches are as rigorous as possible, judgment has to be used in all of them, and more so in some.

Staff monitoring

The Bank of Canada uses a wide variety of quantitative and qualitative information to monitor economic developments and their implication for the performance of the Canadian economy. Much of the information derives from data published by Statistics Canada and relates to the demand side of the economy. Although Statistics Canada publishes a vast amount of information related to the economy, staff at the Bank of Canada typically examine a much narrower subset.13 The most important piece of data is the national accounts, which are released quarterly, contain detailed information on expenditures and income, and are used as the basis of the projection process. In addition, on a monthly basis, several pieces of data are examined by staff to assess how the economy is evolving relative to the projection. The most important of these are data on the CPI, merchandise trade, retail sales, housing starts and resales, motor vehicle sales, manufacturing orders and shipments, labor force statistics, and GDP at basic prices (an industry-based measure). Because much of this is demand related, the data are complemented by a wide variety of other indicators on inflation and capacity pressures, as outlined in Table 2.2. They are derived from Statistics Canada and other data sources, and include capacity utilization rates, estimates of the stock-to-sales ratio, information on skills shortages, and various indicators related to wages and prices. Over time, the range of indicators that has been used in analyzing economic developments has not changed much. What has changed is the presentation of these indicators and a move to a more systematic manner in which they are examined. For example, as Table 2.2 shows, the indicators of capacity pressures are grouped under various themes that help in their interpretation. One set of indicators that has, however, taken on increased prominence over time is that related to expectations. One reason is that under inflation targeting, we attempt to anchor expectations of future inflation, and various measures of inflation expectations help us to assess how well we are doing in this regard.

In the monitoring process, staff judgment is supplemented with information provided by the use of partial-equilibrium econometric models such as Phillips curves, aggregate demand (IS) curves, and artificial neural network models (Tkacz and Hu, 1999). The staff consider the statistical data as well as output from the various models as a whole to come up with a view on the monitoring of real GDP and inflation, since it is often the case that individual indicators do not necessarily paint a consistent picture of the likely evolution of the Canadian economy. Moreover, the accuracy of the models, as well as the usefulness of various data series in monitoring economic developments, need to be periodically assessed. Structural breaks, for example, could limit the use, or change the interpretation, of some measures. Thus, these models are updated frequently (for example, testing for structural breaks); they are respecified, sometimes incorporating new econometric techniques; and new models are developed from time to time to help in the analytic work.

Table 2.2.Canada: An Example of Useful Statistical Indicators
199920002000200020002001200120012001
Q4Q1Q2Q3Q4Q1Q2Q3Q4
Inflation
Core CPI1.41.31.21.11.51.82.32.41.8
CPI excluding food, energy and
indirect taxes1.61.51.31.41.71.91.92.01.6
CPIW11.71.61.41.61.81.92.42.31.8
MEANSTD21.41.41.31.51.82.22.82.41.8
CPI inflation2.42.72.42.73.12.83.62.71.1
Chain price index for GDP2.93.83.93.53.43.71.90.2
Chain price index for consumption1.71.91.91.92.21.82.62.2
Chain price index for consumption
excluding food and energy1.11.11.11.01.31.31.71.9
Product market
Output gap0.31.31.01.31.00.70.1−0.8−1.6
Capacity utilization rate
Nonfarm goods
(1988:Q1 peak = 87.2)84.786.586.686.485.083.081.979.4
Manufacturing
(1988:Q1 peak = 84.6)85.186.886.786.884.881.580.077.4
Unfilled orders/shipments-
Manufacturing excluding aerospace products and parts (1998:Q3 peak = 0.81)0.740.730.710.700.710.710.670.67
Aggregate stock-to-sales ratio (C$)
National accounts0.670.670.660.660.660.660.660.68
Labor market
Unemployment rate7.06.86.76.96.97.07.07.17.6
Participation rate
(2001:Q4 equilibrium value = 65.8)65.665.865.865.966.166.166.165.966.1
Growth in employment (annual rates)
(average growth of trend
labor force in 2001 = 1.3%)3.13.61.71.03.00.91.1−0.50.4
Skilled labor shortage (% firms)
Manufacturing (Statscan survey)
(1989:Q2 peak = 13)8.06.08.08.08.08.08.07.07.0
All sectors
(Bank of Canada regional survey)43.043.056.059.051.048.035.035.0
Wages and costs
Labor Force Survey—average hourly
earnings (unweighted)
All employees2.93.03.13.13.03.43.72.93.1
Permanent employees3.03.23.33.53.43.43.82.73.0
Labor Force Survey—average
hourly earnings (fixed-weight)
Permanent employees3.03.23.53.43.13.53.93.33.9
Wage settlements-Private sector3.82.82.41.92.22.42.93.2
Compensation per hour-Business sector1.45.64.33.33.62.02.53.5
Unit labor costs-Total economy
(labor income/GDP, market prices)1.02.01.31.62.73.21.52.5
Unit labor costs-Business sector−0.53.42.21.63.22.22.13.3
Bank of Canada commodity
price index20.927.720.713.713.98.02.7−9.4−24.2
Energy54.776.557.548.652.040.511.4−18.1−41.6
Nonenergy9.612.07.1−1.2−3.5−8.5−2.0−3.8−11.3
Real estate market
New housing price index1.51.92.32.42.42.52.82.92.8
Resale housing price-Royal Lepage6.26.44.04.64.63.33.93.83.8
Nonresidential property price-Russell1.91.10.61.13.23.23.23.41.8
Vacancy rate-Apartments (%)
(April 1986 trough = 1.4)2.61.61.61.1
Vacancy rate-Offices (%)
(1987 trough = 8.2)8.78.07.46.96.67.07.28.08.4
Vacancy rate-Industrial market (%)
(1986 trough = 4.6)5.04.44.44.24.33.94.14.14.2
Expectations
Percentage of firms expecting price
increases over the next
six months to be:
1% or less18.08.013.07.09.020.05.024.056.0
2% or less75.053.057.049.052.066.055.072.094.0
3% or less94.092.092.092.092.097.093.096.097.0
Survey of forecasters
CPI Inflation
20022.12.12.11.8
2–3 years1.81.82.02.02.0
6–10 years1.81.92.12.02.1
CPI excluding food and energy
20011.82.11.91.82.02.02.1
20021.92.01.81.7
Expectations of implicit real/
nominal bond spread2.12.12.01.92.12.32.32.21.9
Sources: Bank of Canada; Statistics Canada; Conference Board; Consensus Economics; Royal Lepage; and Human Resources Development Canada.

Average growth of a double-weighted measure. Initial weights are multiplied by a second weight that corresponds to the reciprocal of the standard deviation of relative price change.

Weighted mean of the distribution of the year-over-year growth rates of the 54 components, trimmed to exclude components with year-over-year growth rates (in absolute value) greater than 1.5 standard deviations from the mean growth rate.

Sources: Bank of Canada; Statistics Canada; Conference Board; Consensus Economics; Royal Lepage; and Human Resources Development Canada.

Average growth of a double-weighted measure. Initial weights are multiplied by a second weight that corresponds to the reciprocal of the standard deviation of relative price change.

Weighted mean of the distribution of the year-over-year growth rates of the 54 components, trimmed to exclude components with year-over-year growth rates (in absolute value) greater than 1.5 standard deviations from the mean growth rate.

Model-based projections

The key model at the Bank of Canada is the Quarterly Projection Model (QPM), a macroeconometric model with a well-defined steady state, which embodies the Bank staff’s mainstream paradigm for projection and policy analysis (see Coletti and others, 1996; Longworth and Freedman, 2000). With each quarterly release of the national accounts, a new projection is made for the Canadian economy using QPM. This projection contains a number of facets, but the ultimate objective is to ensure that the Bank achieves its inflation target six to eight quarters ahead. Since published data for the most recent quarter only appear with a lag, the starting point for the projection (that is, the first two quarters) comes from the detailed staff monitoring. In addition, the Bank also uses an Ml Vector Error Correction Model (Ml-VECM), a small model in which the monetary aggregate Ml plays a key role (see Hendry, 1995; Armour and others, 1996; and Maclean, 2001).

Regional staff input

Canada is a very large country with distinct economic regions and less-than-perfect factor mobility. While monetary policy can only be made for the country as a whole, it is nevertheless useful to have additional anecdotal information from these regions to complement national data. In this area, the main statistical information comes from a quarterly economic survey undertaken by the staff of the Bank’s five regional offices. In preparation for each quarterly projection, they survey approximately 120 enterprises across Canada and include questions on output growth, employment growth, capacity utilization, shortages of skilled workers, etc. The results of the questions are presented as a balance of opinion—for example, the percentage of firms expecting increases less those expecting decreases (see Amirault and Lafleur, 2000). The regional representatives also present their judgmental forecast of output growth in the Canadian economy, built up separately from the Bank’s knowledge of developments in the regions, aided by surveys, regional contacts, and the views of private sector forecasters.

Analysis of financial market developments

This analysis, about which more is presented below, uses derivatives prices to obtain market expectations about future developments in interest rates.

Inflation targeting is most effective if the process for assessing all of the incoming statistical and anecdotal information is clear and widely understood. Helping to make it so requires a close relationship between the decision-making process and the availability of statistical information.

Since December 5, 2000, we have moved to a system for implementing monetary policy in which changes to the midpoint of the target range for the overnight rate (the policy rate) are announced on eight prespecified dates each year. The scheduling of the dates is based on the timing of the flows of economic information that the Bank uses to gauge the economic situation in Canada (Thiessen, 2000b). Four of these dates are linked to the staff economic projection that follows the release of the quarterly national accounts. The other four dates occur about halfway between these dates, such that meetings are always between five and eight weeks apart. The choice of these dates is related to the release of other important pieces of information, such as the CPI. In addition, each of the dates is separated from U.S. policy action dates by a week or more.

There were several reasons why the Bank chose to introduce the system of fixed announcement dates. First, it reduces uncertainty in financial markets about when the Bank might change interest rates and, hence, allows market participants to more fully prepare for changes. Second, it allows the Bank to report to the public more frequently about the economy and its conduct of monetary policy. In addition, since the dates are different from those of the U.S. Federal Reserve Board, it allows more attention to be focused on Canadian economic circumstances. Third, it allows the Bank to emphasize the medium-term perspective of monetary policy. Fourth, and most important from the standpoint of this chapter, these dates conform closely to the flow of information that is received from Statistics Canada and other sources. Finally, regular announcements enhance the transparency of monetary policy and help financial markets to better understand and anticipate the Bank’s actions.

Prior to each fixed announcement date, the Governing Council of the Bank receives a presentation by staff that encompasses an assessment of the external situation; an update of the Canadian outlook for the current and next quarter stemming from the staff monitoring; a macroeconomic model-based forecast of Canadian economic developments further out using the Bank staff’s mainstream economic model on the four occasions when the dates follow the release of the national accounts; a set of risk scenarios also generated using the mainstream model; a judgment-based forecast using information obtained by the Bank’s regional offices; a money- and credit-based outlook using in part the model based on a key money aggregate; several measures of the monetary stance; and a review of the expectations of financial market participants. Other important inputs include alternative measures of capacity and inflation pressures, a last-minute update of the short-run monitoring for inflation and output, and a discussion of the balance of risks.

Using this information and other information available to it, the Governing Council decides on its view of the economy and the appropriate setting of monetary policy. The council operates on a consensus basis and comes to a common view on the likely future path for the economy, the pressures on underlying trend inflation, and the paths for interest rates and monetary conditions that would result in inflation remaining at, or moving to, its target six to eight quarters in the future.

A key output of this process is the press release associated with the fixed announcement date. In it, the Bank not only explains the key factors underlying the decision, it also sets out the Bank’s view on the prospects for output and inflation in the period ahead. This is intended to help the public, the media, and the markets better understand and anticipate the direction of monetary policy. Indeed, given the lags between monetary policy actions and the impact on inflation, improving awareness of how inflation is likely to develop over the future contributes to a better understanding of monetary policy actions.

In addition to the press release on the fixed announcement date, the Governing Council communicates its assessment of recent data and its outlook for output growth (a band) and inflation relative to target to the Canadian public in the semiannual Monetary Policy Report, semiannual Monetary Policy Report Update, and speeches and presentations. Other staff, including senior officers and senior regional representatives, participate in the communication effort by making presentations as well. By keeping the Canadian public regularly informed in this manner, the Governing Council is working to increase credibility for the inflation target and its commitment to achieving it.

Statistical Measures of Monetary Stance

As indicated above, if long-term inflation expectations are well grounded by a credible inflation-targeting strategy, then the Bank’s control of the overnight interest rate is sufficient to allow it to alter the term structure of interest rates to endeavor to keep future inflation in the target range at the six- to eight-quarter target horizon. Thus, clearly, the term structure of interest rates is one useful statistical measure of the ease or tightness of the Bank’s current policy stance (Figure 2.5). Of course, given the various channels through which monetary policy influences the economy and inflation, several other statistical measures of the stance of monetary policy are also useful. At the Bank, we have found three other measures that appear to be particularly helpful in providing additional insight. Along with the term spread or yield gap, the Bank examines the monetary conditions index (MCI), the real 90-day commercial paper rate, and the so-called money gap.

Figure 2.5.Canada: Term Spread

(90-day paper – over 10-year Government of Canada Bond)

Source: Bank of Canada.

The MCI is a weighted average of the level of short-term interest rates and the exchange rate for the Canadian dollar (Figure 2.6).14 It recognizes that in a small open economy like Canada, the exchange rate is also a key element of monetary conditions, in addition to interest rates (Freedman, 2001b). However, while the Bank can control short-term interest rates, it has little control over day-to-day movements in the exchange rate. For example, movements in the exchange rate may reflect important supply-side shocks (for example, real nonenergy commodity price shocks). As such, the exchange rate will adjust to help buffer the Canadian economy in the face of adverse shocks. In these conditions, a change in the MCI would be appropriate. On the other hand, when the exchange rate has moved because of portfolio shifts between currencies, unchanged macroeconomic conditions would imply unchanged desired monetary conditions and thus the need to adjust interest rates to keep the MCI on its old track (Longworth, 1999). One drawback of examining only the MCI is that it is difficult to interpret movements in it. The index is based on 1987:Q1 = 100, but given structural adjustments over time (including changes in underlying inflation and relative domestic and foreign prices), a comparison of the MCI today with its level in 1987 is of limited use. Rather, it should be compared with its recent level taking into account the interpretation of exchange rate movements.

Figure 2.6.Canada: Monetary Conditions Index

Source: Bank of Canada.

Another measure of monetary stance is the real 90-day commercial paper rate, which gives an idea of the evolution of short-term real interest rates over time and thus the relative tightness of monetary conditions (Figure 2.7). If the long-run average of this rate is considered to be close to the neutral interest rate—one that achieves the target rate of inflation—then interest rates above the long-term average indicate relatively tighter policy. Conversely, those below it reflect looser monetary conditions (this is true only if not trying to disinflate).

Figure 2.7.Canada: Real 90-Day Commercial Paper Rate

(Nominal - year-over-year core CPI)

Source: Bank of Canada.

The final measure of policy stance is the so-called money gap (Figure 2.8).15 It draws on the “buffer stock” approach to money demand analysis and indicates the extent to which money is estimated to be in “disequilibrium” (Maclean, 2001). The current approach of the Bank is to derive it from the Ml-VECM model mentioned above. When there are persistent deviations of money from the level consistent with long-run money demand, the resultant money gap is associated with significant changes in inflation. The interpretation of this measure is, therefore, somewhat different from the others. A money gap that is becoming more positive (or more negative) indicates when inflationary pressures are rising (or falling) and thus the need for tighter (or looser) monetary conditions. Note that a money gap that is actually positive indicates excessive inflation pressures.

Figure 2.8.Canada: Stance Measure Based on Adjusted M11

Source: Bank of Canada.

1 Shows the difference between the growth rate of money predicted by the Ml-VECM model over the following year if the interest rate is fixed and the growth rate over the same period that is consistent with inflation reaching 2 percent, eight quarters ahead. Therefore, positive numbers indicate excessive inflation pressure.

It is interesting to illustrate the information conveyed by these separate measures during recent periods. As Figures 2.5 through 2.8 indicate, in the late 1990s and 2000, the MCI, the yield gap, and the real 90-day commercial paper rate all pointed to modestly tighter monetary conditions. This was consistent with the buildup in inflation pressures as suggested by the Ml stance measure and thus the need for tighter policy. In 2001, however, while the first three measures pointed to a substantial easing in monetary conditions, the Ml stance measure suggested that inflation pressures remained in the economy—albeit more muted as 2001 progressed—and thus indicated the need for either a further tightening in monetary conditions or no further easing. This stemmed from the relatively rapid pace of Ml growth that has continued since 1999. It was also in contrast to indicators that point to an easing in demand pressures in the economy, such as slowing real GDP growth. Ultimately, what is important is the assessment of all of these indicators as a whole for the implications of future inflation.

Measures of the stance of monetary policy have to be interpreted in the context of a wide range of other information, including current and projected developments in output and inflation and other indicators of inflation pressures. Indeed, while the Bank of Canada has a relatively direct effect on the first three measures through its control of the overnight rate, its control over the money gap is indirect, with a change in interest rates subsequently affecting the demand and supply of money. The projection model, QPM, provides a path for interest rates that can then be compared with the first three measures of the policy stance for monitoring purposes, especially for fixed announcement dates that come between projection exercises. Another key source of information in Canada, given its fully developed financial markets, is what financial markets anticipate money policy actions will be.

Empirical Analysis of Market Expectations of Future Policy Actions

In advanced economies such as the United States and Canada, futures contracts for short-term financial instruments (Fed funds futures in the United States and Banker’s Acceptances and Forward Rate Agreements in Canada) provide information about the market’s expectation of future movements in policy-determined interest rates.

Figures 2.9 through 2.12 provide examples of the evolution of financial market participants’ views on prospective interest rate developments over the course of two fixed announcement date periods: October 23 and November 27, 2001. After the fixed announcement date in August when the policy rate was lowered to 4 percent, markets had quickly priced in an additional 25 basis points decline for the October 23 date to 3.75 percent.16 In the intervening period to September 10, with the release of further information that suggested additional weakness in the economy, market participants subsequently revised down their expectations to a target for the overnight rate of 3.5 percent (Figure 2.9).

Figure 2.9.Canada; Expected Three-Month Commercial Paper Rates, October 23 Fixed Announcement Date

Source: Bank of Canada.

Note: Vertical lines mark fixed announcement dates.

Figure 2.10.Canada: Expected Three-Month Commercial Paper Rates, November 23 Fixed Announcement Date

Source: Bank of Canada.

Note: Vertical lines mark fixed announcement dates.

Figure 2.11.Canada: Evolution of Expected Three-Month Commercial Paper Rates, up to the October 23 Fixed Announcement Date

Source: Bank of Canada.

Note: Vertical lines mark fixed announcement dates.

Figure 2.12.Canada: Implied Overnight Versus Overnight Target

Source: Bank of Canada.

In the aftermath of the events of September 11, the Bank of Canada made the decision to lower interest rates outside the normal schedule of announcement dates, by 50 basis points on September 17. The reason was to underpin business and consumer confidence in the wake of the terrorist attacks and to provide further support for economic growth. This move was not expected by financial market participants. Once taken, however, financial markets revised their expectations for the announcement date of October 23. With the lowering of rates to 3.5 percent on September 17, they gradually revised down their expectations to a rate of about 3 percent on the October 23 fixed announcement date, or another 50 basis points decline. In part, this reflected the belief that economic conditions had already worsened prior to the events of September 11 and that this justified additional easing. In the event, the Bank of Canada lowered rates by 75 basis points (25 basis points more than expected), and in the press release indicated that a more pronounced weakening trend in economic activity had become evident and that the terrorist attacks of September 11 appeared to have exacerbated this trend. The level of economic activity was, therefore, likely to be further below the economy’s potential output than previously expected, placing additional downward pressure on inflation.

Despite the greater-than-expected decline in rates, markets quickly shrugged off the surprise and formed expectations for the November 23 fixed announcement date, anticipating a further 50 basis points decline in the target for the overnight rate to 2.25 percent (Figure 2.10).17 In the intervening period to the November fixed announcement date, the Bank of Canada had released its semiannual Monetary Policy Report, which undoubtedly helped to influence expectations. It set out the Bank’s two working assumptions: that there would be no further major escalation in terrorism, and that business and consumer confidence would return to normal levels in the second half of 2002. Developments since that time appeared consistent with these assumptions. On November 23, the Bank lowered its target for the overnight rate to 2.25 percent, in line with market expectations.

The process of expectations formation by market participants does not have to be smooth. Expectations of future interest rates can move abruptly based on unanticipated changes in data and unforeseen economic events. This process of adjusting expectations is evident in Figures 2.11 and 2.12, which show the evolution of expectations up to the October 23 fixed announcement date. After the Bank lowered interest rates on September 17, markets quickly revised down their expectations of future rates. This happened predominantly within a week. After that, expectations of what level rates would be set at on October 23 remained more or less constant.

Another way to examine the impact of the move to the system of fixed announcement dates is to look at the volatility in short-term money markets. Since the adoption of the system, uncertainty and volatility in money markets should have been lowered relative to the previous regime, because market participants would know when the Bank plans to change rates.18 A recent paper (Aba, 2001), which examines the period immediately following the move to fixed announcement dates, notes that the standard deviation in 30-day prime commercial paper yields was the same in the period following the shift to the fixed announcement dates (the same pattern holds true for three-month treasury bills). In the United States, however, the standard deviation rose because of unexpected reductions in interest rates by the Federal Reserve Board (in January and April). Canadian market participants did not expect these interest rate declines to be followed in Canada, suggesting that they were paying more attention to domestic developments than previously.19

Of course, as discussed above, the central bank does not have to validate market expectations, suggesting that there are circumstances in which the central bank might do something other than what the market expects. If the central bank expects the economy to evolve in a certain way on the basis of current policies and has a certain view of the transmission mechanism, it will take a particular action at the next decision date.

If the market has the same view as the authorities about

  • (1) the future evolution of the economy;

  • (2) the nature of, and time lags in, the monetary policy transmission mechanism;

  • (3) the authorities’ ultimate monetary policy target; and

  • (4) the time interval over which this target is to be achieved;

then the market should correctly anticipate the policy action that the authorities will actually take at the next decision date.

In many circumstances, one would expect the market and the Bank to interpret the economic situation in a similar way and to have similar views about the appropriate path for policy. However, new information that becomes available can change the Bank’s view regarding the outlook for output and inflation. If the market interprets this information differently from the Bank, the Bank can consider to what degree it should endeavor to condition the market to its current view on the outlook before taking a policy decision, or use the announcement date itself as a signaling device.

Provided that the market has confidence in the policy target (owing to transparency), the authorities will only need to “send a signal” by surprising the markets when market expectations of factors (1) and (2) in the list above do not correspond to those of the authorities. For example, in the face of a negative commodity price shock, the exchange rate would be expected to depreciate to temper the effects of the shock on the Canadian economy. The market, however, may view the nature of the shock differently from the central bank and attempt to bid up interest rates as the exchange rate depreciates. In such a situation, the Bank would be expected to resist the upward pressure on interest rates if it has concluded that the shock is indeed a negative commodity price shock. Sometimes markets also differ in their view about how long a shock might last. If the central bank expects that a shock will last less time than its ability to react to it—as noted earlier, about six to eight quarters—then it will not usually respond to the shock even if the markets feel that one is justified, which they manifest through expected reductions (or increases) in interest rates. Such shocks could encompass a wide variety of temporary demand or supply shocks.20

Despite the possibility of different interpretations of shocks leading to different conclusions, financial markets seem to have accepted that the Bank’s policy actions are firmly directed toward achieving the inflation target. In this vein, monetary policy seems to have credibility.

Measuring the Credibility of Monetary Policy

Central banks that are implementing an inflation-targeting approach also need to be able to assess whether their strategy is credible. There are several types of statistical data that can be examined to make inferences about the extent to which inflation and inflation expectations are firmly anchored and, hence, about the degree of credibility of the inflation target. These data can be categorized into those that provide relatively direct evidence and those that provide evidence on behavior more likely to occur when there is more confidence that the future path of price inflation will be low and stable. On the former, obvious sources of information are the standard deviation of inflation around the inflation target, survey measures of expected inflation, and, where available, the difference between the yield on nominal bonds and comparable maturity real-return bonds. On the latter, statistical indicators more likely to be associated with a greater degree of confidence in low, stable growth in price inflation are the dispersion of inflation forecasts across forecasters, the average deviation of inflation forecasts from the inflation target, and the length of various types of contracts.

A focus on preserving a low and stable inflation environment should lead to a reduction in the variability of inflation. A reduction in inflation variability tied to a low-inflation environment allows people to take a longer view with respect to their planning and to better perceive the trade-offs that they face in their transactions. This in turn leads to a better allocation of economic and financial resources. In Canada, the standard deviation of core inflation, the year-to-year change in the CPI excluding the eight most volatile components as well as the effects of indirect taxes on the other components, relative to the target for inflation has fallen to 0.5 percentage points in the second half of the 1990s, from between 2 and 3 percentage points historically.

With respect to survey measures of expected inflation, there is work assessing the experience of inflation-targeting countries, including Canada (see Perrier and Amano, 2000). Johnson (1998), for example, uses individual forecasts of inflation by professional forecasters from 1984 to 1995 to examine the credibility of monetary policy across 18 industrial countries, including Canada. He concludes that, among inflation-targeting countries, Canada and New Zealand have the most credible targets. Perrier (1998) uses survey data on inflation expectations of professional forecasters from the Conference Board of Canada. He concludes that these data suggest that the Bank of Canada has developed increasing credibility over the inflation-targeting period.

These studies are based on formal regression analysis, but the flavor of their results can be captured by looking at the behavior of survey measures of inflation in the inflation-targeting period as portrayed in Figure 2.13. Also shown in Figure 2.13 is the difference between the yield on Government of Canada long-term real-return bonds and comparable maturity nominal bonds. The survey measures are both based on surveys of forecasters’ outlooks for inflation, but for different time horizons. As the chart makes clear, all of these measures, as well as the spread between real and nominal spreads, have been solidly anchored on the midpoint of the target range for some time. The bond yield differential shows somewhat more variance because it is subject to investor preferences and the amount of liquidity in the real-return bond markets.

Figure 2.13.Canada: Comparison of Selected Inflation Expectations

Sources: Bank of Canada; Conference Board; and Consensus Economics.

1Difference between nominal and real-return 30-year bonds.

Another way that the data from forecasters might be used to assess whether credibility has increased is to examine the dispersion of their forecasts for inflation. To the extent that there has been an increase in credibility, it would be expected that the dispersion of forecasts of inflation should be less than they were before inflation targeting was adopted. As Figure 2.14 shows, there has been a decline in the dispersion of inflation forecasts in Canada in the inflation-targeting period.

Figure 2.14.Canada: Long-Term Dispersion of Inflation Expectations

Source: Stuber (2001).

Note: The inflation uncertainty measures are based on the dispersion of forecasts for inflation 6 to 15 years ahead by particpants in a survey of economists and portfolio managers published by KPMG.

Statistical evidence of increased confidence in a low, stable growth path for the price level can be found in the labor, housing, and financial markets. In Canadian labor markets, there has been a reduction in the person-days lost due to strikes (reflecting, in part, less confusion about future inflation), a lengthening of the life of collective wage agreements, and a decline in the number of such agreements containing cost-of-living adjustment (COLA) clauses (Table 2.3). As to the housing market, the proportion of mortgages with five-year terms is now higher than it was in the mid-1980s, and many financial institutions have been offering 7- to 10-year mortgages (on the former, see Montplaisir, 1996-97; and Hewitt, 1997). In a related manner, the use of long-term financial instruments by firms fell over the course of the 1970s and recovered as inflation moved down (Figure 2.15).21

Table 2.3.Canada: Selected Indicators Pertaining to Wage Negotiations
PeriodAverage Life of Wage

Settlements (Months)
Proportion of Wage

Settlements with COLA

Clauses (Percent)1
Work Stoppages

(Percent of Working

Time Lost to Strikes)
1978 to 199024.223.40.23
1991 to 199525.514.90.06
1996 to 200034.010.50.09
Source: Human Resources Development Canada.

COLA, cost-of-living adjustment

Source: Human Resources Development Canada.

COLA, cost-of-living adjustment

Figure 2.15.Canada: Ratio of Long-Term Business Credit to Total Business Credit

Source: Bank of Canada.

Note: Data for 2001 represent the average of January to November.

Once inflation targets have been in place for a period of time, one of the simplest ways to assess credibility is to look at what happens to measures of inflation expectations when actual inflation is moving away from the midpoint of the target band. In Canada, medium- and long-term expectations have remained anchored on the midpoint of the inflation target band even when headline inflation has stayed in the upper part of the target range, as it has since 1999 (see Figure 2.1 and Figure 2.13).

On balance, a wide range of statistical evidence for Canada suggests, individually and cumulatively, that the current inflation target has a high degree of credibility. An open question for inflation-targeting central banks is whether they should say something about the implications of their approach for the evolution of the price level in the long run.

Central Bank Accountability for the Evolution of the Price Level in the Longer Term

In discussing the fundamental goal of monetary policy in Canada, we have indicated that the Bank of Canada takes action to ensure the targeted inflation rate will be achieved 18 to 24 months in the future. But what about the price level in the longer term?

Over the period 1991 to now, as a whole, Canada has moved from using direct inflation targets as a means of disinflation to nearly six years of targeting low and stable inflation in the 1 to 3 percent range. The need to renew, in 2001, the agreement between the government and the Bank of Canada on the long-run target for monetary policy raised some interesting questions about the responsibility for the evolution of the price level over the longer term.

The reason, of course, is that inflation targeting of the type we have been discussing can involve base-period drift. If there is a jump in the price level in one period—say, because of a very large change in a relative price—then the inflation target is missed for one period (bygones), but the price level would be out of line forever. To make the point more starkly, note that over a 10-year period the price level that would result from continuously attaining an inflation rate of 3 percent would be 22 percent higher than the price level that would result if the inflation rate had risen continuously at 1 percent a year over the same period. As discussed further below, however, if the central bank targets the midpoint of the target range, such an outcome will be extremely unlikely.

People must make lifetime saving plans based on views about future real consumption and the real rate of return on the assets that they accumulate as a result of their saving plans. Even small errors in estimating the real return on assets associated with a given nominal return—say, owing to mistakes in predicting the inflation rate—can lead to large differences between desired and actual asset accumulation over long periods of time. Therefore, governments, as the elected representatives of the people, have a responsibility to try to reduce the amount of “noise” that results from unanticipated movements in inflation rates and price levels. In principle, the government and, of course, the central bank should be accountable in some way for such differences. At present, however, no simple procedure exists for incorporating a price level commitment in the inflation-targeting framework. Moreover, whether targeting a price level reduces or increases the volatility of output depends on a number of factors, particularly the extent to which price expectations are forward looking, and the extent to which there are important permanent relative price shocks in the economy.

One way a central bank could meet this responsibility is by making inflation more predictable over longer time periods. Thus, it could continue to target a reasonable range for the rate of inflation of the overall price level (perhaps adjusted for changes in indirect taxes), but to place even more emphasis on targeting the midpoint of the range. Going forward, a possible accountability framework would then be one in which the central bank could show that, over some well-defined past period, it had on balance achieved a path for the price level exclusive of the short-run effects of changes in tax rates that puts it reasonably close (that is, within a narrow range) to what would have been achieved if it had consistently been aiming at the middle of its inflation-control target range. To help make inflation more predictable over longer periods, the renewal of the inflation target in May 2001 contained two particular features: a longer-term agreement (five years instead of the three-year term of the two previous agreements), and an affirmation that monetary policy will be directed to moving inflation to the 2 percent midpoint of the target range over a six- to eight-quarter horizon.

One indicator that the Bank of Canada plans to focus on as time passes is the evolution of the CPI relative to the midpoint of the target range from the date that the new targets were implemented (May 2001). As to why this makes sense, Figure 2.16 shows that, over time, average inflation (adjusted for the effect of changes in indirect taxes) has become more firmly centered around 2 percent as the length of the averaging period increases. This result occurs since the probability of achieving an inflation rate close to 2 percent increases with the length of the averaging period because there is a greater probability that temporary random shocks will average close to zero (Crawford, 2001b). Average inflation therefore becomes more predictable over longer time horizons.22 Indeed, the annualized rate of core inflation over periods of two years is within 1 percent of the target midpoint approximately 90 percent of the time (Crawford, 2001b).

Figure 2.16.Canada: CPI Adjusted for the Effect of Changes in Indirect Taxes

(Average percentage change from December 1994 at annual rates)

Source: Bank of Canada.

Conclusions

In Canada, we believe that it takes 18 to 24 months for a change in the policy interest rate to exert its full effect on inflation. Given the various forms of uncertainty facing the policymaker, it is important to rely on a variety of information sources, monitoring, and model projections in as systematic a manner as possible. However, judgment inevitably plays an important role in decision making. Indeed, monetary policy is a combination of “scientific” assessment of economic statistics and judgment. And this mix of the best available statistics, good economic analysis, and a dose of judgment will undoubtedly continue to be required, given the changing economic landscape and the need to interpret those changes.

We have benefited from comments by our colleagues, especially David Longworth, Chuck Freedman, Ron Parker, John Murray, Tiff Macklem, and Nicolas Parent. We would also like to thank participants at the IMF seminar on the Statistical Implications of Inflation Targeting for their comments. Annie de Champlain provided excellent research assistance. We are solely responsible for any errors. The views expressed in this chapter are ours alone and should not be attributed to the Bank of Canada.

Canada’s exports, for example, contain a much higher proportion of energy, paper, and wood products, as well as nonferrous metals and related products, telecommunications equipment, and automotive products than do those of the United States.

Commodity prices have been on a general downward trend, and since they play a relatively more important role in production than consumption, growth in the GDP deflator has diverged from that of Canada’s CPI. Another factor placing relative downward pressure on growth in the GDP deflator has been falling machinery and equipment prices, especially computer prices (Freedman, 1996).

As Crawford (2001b) notes: “Although economists typically use confidence intervals of 95 percent or 99 percent, the target range was never described as a range that would contain actual inflation with such a high level of confidence. Indeed, empirical work done by Bank staff in 1990–91 suggested that inflation would be inside the range only about two-thirds of the time.”

This is based on the simplifying assumption that such tax changes are passed through immediately and one-for-one into consumer prices. Whether this actually happens depends on a number of factors such as the competitive structure of the industry, the point in the economic cycle, etc.

The Bank of Canada has defined the measure of core inflation. Prior to the adoption of the new core measure, the Bank of Canada calculated core inflation as well as the indirect tax adjustment. Now, with the move to the new measure of core inflation, Statistics Canada calculates the measure of core inflation. The Bank of Canada, however, makes the adjustment for indirect tax effects on the remaining components.

The eight excluded items are fruit, vegetables, gasoline, fuel oil, natural gas, intercity transportation, tobacco, and mortgage-interest costs. These excluded components were found to be more than 1.5 standard deviations from the mean of the cross-sectional distribution at least 25 percent of the time over the period 1986 to 2001 (Macklem, 2001). This was not the case for other items that were excluded under the former measure of core, such as meat, fish and other seafood, dairy products, etc.

The better performance is judged by comparing regressions in which the future change in the 12-month CPI (adjusted for the effect of indirect taxes) is explained by the current difference between a core measure of inflation and CPI inflation (once again adjusted for the effect of changes in indirect taxes). Movements in the new measure of core, CPIX, better explain future CPI movements than do movements in the former measure of core, the CPI excluding food and energy and the effects of changes in indirect taxes, CPIXFET. This better performance holds whether the future change is over 12 or 18 months and whether the period starts in 1986 or 1992.

In the Canadian case, the modest changes in interest rates that were required in the short run to keep Ml within its target band were not enough to have much impact on output or prices, given the high interest elasticity of the demand for money (Thiessen, 2000a).

Furthermore, between 1992 and 1994, Canada phased out the legal requirement for chartered banks to hold minimum cash reserve ratios on their deposit liabilities at the central bank. From 1994 forward, banks would have to maintain appropriate capital and liquidity for prudential purposes, but they were no longer required to hold a fraction of their liabilities as deposits with the Bank of Canada. In a system without fractional reserves, “base money” is no longer a useful measure of the stance of monetary policy, reenforcing the role of the short-term interest rate as the main channel for the monetary policy transmission mechanism. Note, however, that the Bank of Canada has never used base money or monetary aggregates as the instrument for monetary policy (although Ml was the intermediate target between 1975–82), a role that has always been played by the short-term interest rate (see Longworth, 2002).

As Freedman (2001a, footnote 24) notes, “The difference in behavior between the two central banks in the first half of the decade may have reflected the fact that the inflation target range in New Zealand was perceived as hard-edged while that in Canada was soft-edged. The Reserve Bank of New Zealand therefore actively adjusted nominal interest rates to bring about movements of the exchange rate, in order to take advantage of the pass-through from the exchange rate to prices, a mechanism that operates more quickly than the mechanism operating through aggregate demand to prices.”

In Svensson’s formulation the period loss function is and the intertemporal loss function is L(πτ,γτ)=1/2[(πτπ*)2+λγt2], and the intertemporal loss function is

where Δ is the discount factor, πτ, is actual inflation, π* is the target rate of inflation, γτ is real output relative to capacity output, and λ is the weight given to output stabilization.

See Kozicki (2001) for an overview of why central banks monitor so many inflation indicators.

The relative weights are 3 to 1, for example, a 3 percent change in the exchange rate is equivalent to a 1 percentage point change in its effect on output.

The money gap is the difference between the growth rate of money predicted by the Bank’s Ml-VECM model over the following year if the interest rate is fixed and the growth rate over the same period that is consistent with inflation reaching 2 percent eight quarters ahead. A positive number indicates excessive inflation pressure.

Figures 2.9 through 2.11 use Forward Rate Agreements to calculate expectations. These are priced by using forward agreements to enter into a three-month Banker’s Acceptance contract, In the first month, the expected three-month commercial paper rates are derived from the three-month forward contract in that month. For subsequent months, the three-month forward contract beginning the following month (or a 1 x 4 contract) is used, followed by a 2 x 5 contract, and so on. Expectations are calculated assuming a constant term premium for each forward contract (that is, the premium is the same for all 1 x 4 contracts, but may differ from the premium for a 2 x 5 contract, and so on).

Expectations in Figure 2.12 are derived using price information for each day from Banker’s Acceptances, and not Forward Rate Agreements as used in Figures 2.9 through 2.11.

For example, in the past, when the U.S. Federal Reserve Board changed interest rates, market participants in Canada often then expected the Bank of Canada to follow suit the next day. In the period up to 9:00 a.m.—when rates are set—trading activity would dry-up as participants waited for the Bank’s action, introducing volatility into the system.

A simple regression model by Aba (2001) finds that the forecasted standard deviation was higher than the one actually observed over the fixed announcement date period, suggesting that the move to the fixed schedule reduced volatility in Canadian money markets.

This will of course depend on the credibility of the central bank. In a situation of low credibility, the central bank may be forced to react to temporary shocks if they feed through into inflation expectations.

For a more detailed discussion, see Jenkins and O’Reilly (2001).

If there is no autocorrelation from year to year (as appears to be the case in the targeting period for total CPI inflation, abstracting from the effect of large indirect tax changes), the statistical formula for the variance of an average implies that the width of the interval will be inversely proportional to the square root of the length of the period over which the average is being calculated.

    Other Resources Citing This Publication