VIII Inflation Targeting and Macroeconomic Volatility

Vladimir Klyuev, Martin Mühleisen, and Tamim Bayoumi
Published Date:
October 2007
  • ShareShare
Show Summary Details
Tamim Bayoumi and Vladimir Klyuev 

Canada was one of the pioneers of inflation targeting (IT). The IT regime that was introduced in February 1991 was the second in the world, with New Zealand having adopted a similar framework in 1989. Despite its novelty, the IT regime was a great success from an early stage and has been widely credited as a key element in reducing inflation, anchoring private sector inflationary expectations, and reducing macroeconomic volatility (Ragan, 2005).

The IT framework has also been allowed to evolve over time, particularly around the renewals of the IT agreement between the Bank of Canada and the federal government that have occurred about every five years. Early in the process, these changes included lowering the inflation target, but this has been fixed at 2 percent since December 1994. On occasion, they involved fine-tuning the way core inflation is measured. More radical changes, such as lengthening the time horizon over which the inflation target should be achieved in response to certain types of shocks, have at least been contemplated and discussed within the Bank of Canada. Analyzing the link between IT and macroeconomic volatility is an important ingredient in determining the role of the framework in shaping the Canadian economy and in evaluating the potential benefits and costs of proposed regime changes.

This section uses a small dynamic monetary model to assess the link between IT and macroeconomic volatility. The notion that the monetary policy framework itself and the credibility it enjoys are important for macroeconomic stability is a common one among policymakers.1 This analysis, which follows the approach in Bayoumi and Sgherri (2004a, 2004b), suggests that the most important link goes through the impact of monetary credibility on the forward-lookingness of private sector expectations and presents evidence that IT lowered macroeconomic volatility through these “credibility” effects. This implies that changes to the framework should be approached in a careful manner so as to avoid eroding the credibility of the overall regime.

The Framework

Analysts often use small models to assess the impact of monetary policy on the economy. A typical closed economy model involves three equations of the following type:

where y is the output gap (the difference between actual and potential output), r is the real interest rate (current nominal rate less expected inflation), π is annualized inflation, i is the nominal interest rate, εs are error terms, superscript e represents expectations, and other Greek letters reflect parameters. This type of model highlights two main monetary transmission channels: the real interest rate channel, through which the central bank affects the spending decisions of the private sector; and an expectation channel, where the monetary authority influences the private sector’s expectations by conveying information about the future course of monetary policy.2

The monetary reaction function equation (1) follows the approach of Clarida, Galí, and Gertler (1999). It comprises a Taylor rule in which the nominal interest rate responds to expected future inflation and the lagged output gap (assuming the current gap is not observable), augmented by a lagged dependent variable to reflect interest rate smoothing, while the unobserved inflation target and natural rate of interest are subsumed in the constant term.

Aggregate demand is determined in equation (2), which links the output gap to the short-term real interest rate, as well as forward- and backward-looking weights on the output gap itself. The weights add up to the subjective discount rate, taken to be 0.99 in quarterly data. This type of relationship comes from a Euler equation for consumption augmented to allow for habit persistence.

The expectation-augmented Phillips curve given in equation (3) describes the model’s supply side. It relates current inflation to a weighted average of future and past inflation, with the weights adding up to one. This type of specification is typical in empirical work,3 although its theoretical justification has been a source of contention.

Modern inflation theory suggests that more predictable monetary policy can improve macroeconomic stability by making the Phillips curve more forward-looking. Recent theoretical work has established that the inflationary expectations become more forward-looking as uncertainty about future demand is reduced.4 In particular, modern versions of the Lucas “islands” model of nominal rigidities that take into account uncertainties about how people assume others will react to a given event imply that the degree to which expectations in the Phillips curve are backward-looking should rise with uncertainty about aggregate demand. As discussed in Bayoumi and Sgherri (2004a), this implies that the level of nominal inertia should be negatively related to monetary credibility, as this is an important determinant of uncertainty about aggregate demand. That paper finds a strong, statistically significant and correctly signed cointegrating relationship in the U.S. data between the uncertainty about real interest rates (the proxy for credibility in this model) and the weight on backward-looking expectations in the Phillips curve, with the former Granger causing the latter and no evidence of reverse feedback. This research motivates us to investigate whether the introduction of the IT regime, by lowering uncertainty about monetary policy, has made the inflation process more forward-looking, thereby producing a more flexible economy and reducing macroeconomic instability.

It should also be recognized that policymakers in Canada were among the first to recognize this potential link and relied on it as a rationale for introducing inflation targets. For example, in March 1995, Gordon Thiessen, then Governor of the Bank of Canada, observed that “[b]y making its inflation-control objectives more explicit, the Bank hoped not only to influence inflation expectations, but also reduce uncertainty in the economy and the financial markets” (Thiessen, 1995). He went on to say that “with credible targets, inflation expectations, and therefore inflation, are less likely to react to temporary demand and supply shocks.” In addition, inflation targets impose discipline on the Bank, which “makes monetary actions more predictable and less a source of uncertainty for others as they make economic decisions.”

Empirical Results

Empirical estimates suggest that the introduction of IT may indeed have made the Canadian Phillips curve more forward-looking. The model was estimated using quarterly data over 1982–89 (the period before the introduction of inflation targeting), and the IT period of 1992–2005—the years 1990–91, being a period of transition, were eliminated from the sample.5 Estimation used the generalized method of moments (GMM) with the first four lags of inflation, the output gap, and interest rates to instrument the forward-looking variables in the equation. The results produce little evidence of instability over time in the IS curve or the coefficient on the output gap in the Phillips curve, but do show a statistically significant shift in the coefficient on forward-looking inflation in the Phillips curve after the introduction of IT in addition to more aggressive response to inflation and greater interest rate smoothing in the monetary reaction function. In particular, the preferred specification—reported in Table 8.1—involves a rise in the coefficient on forward-looking inflation of almost one-third, from 0.54 to 0.71, as IT reduced uncertainty about the monetary reaction function.

Table 8.1.Estimates of Monetary Model1
1982–89Both Periods1992–2005
IS curve
Expected output gap0.507 (.006)*
Lagged real interest rate−0.008 (.001)*
Standard error of regression0.4920.226
Phillips curve
Expected inflation0.54 (.130)*0.71 (.039)*
Lagged output gap0.030 (.024)
Standard error of regression1.4872.110
Monetary reaction function
Lagged interest rate0.699 (.024)*0.833 (.026)*
Constant term6.965 (.480)*0.986 (.711)
Long-run coefficient on inflation0.813 (.088)*1.532 (.317)*
Long-run coefficient on output gap0.610 (.074)*−0.230 (.279)
Standard error of regression1.0160.851
Source: IMF staff calculations.

More forward-looking inflation expectations enhanced the expectations channel, reducing the need for aggressive responses to macroeconomic shocks. Figure 8.1 reports impulse-response functions from the estimated model with regard to disturbances to the output gap, inflation, and interest rates over the two samples.6 As shown in the first two columns of the figure, after the introduction of IT, more limited monetary responses were needed to stabilize the economy in response to macroeconomic shocks, largely reflecting the speedier impact of interest rate changes on private sector behavior. In addition, the increased importance of the expectation channel implied by more forward-looking inflation expectations also allowed monetary policymakers to respond to shocks in a more gradual manner.

Figure 8.1.Impulse-Response Functions

Source: IMF staff calculations.

The results suggest that the increase in Canada’s macroeconomic stability since 1991 largely reflects a decline in inflation inertia. To illustrate this further, the first four rows in Table 8.2 report the standard deviation of inflation and the output gap and the interest rate before and after the introduction of IT and compares them with the volatility implied by our estimated monetary model. While the model does not very accurately replicate the standard deviations observed historically, it does capture the reduction in macroeconomic volatility after the introduction of IT, including in the interest rate as IT made policy responses more predictable.

Table 8.2.Standard Deviation of Inflation, Output Gap, and Interest Rate before and after Introduction of Inflation Targeting (IT)(In percent)
EquationShocksInflationOutput GapInterest Rate
A ActualPre-IT2.12.22.0
B ActualIT1.81.51.6
C Pre-ITPre-IT4.81.54.1
D ITIT3.20.81.6
E ITPre-IT2.41.61.8
F Pre-ITIT6.50.84.8
G IT MR; Pre-IT PCIT6.11.24.5
H IT PC; Pre-IT MRIT3.20.71.8
Source: IMF staff calculations.Note: Rows A and B: actual standard deviations in pre-IT and IT periods; C and D: asymptotic standard deviations based on estimated model (1) for the two periods; E and F: mixing estimated model parameters from one period with estimated standard deviations of shocks from the other; G and H: model-based asymptotic standard deviations assuming the structure of the economy and the shocks are as estimated for the IT period, except that row G uses the pre-IT period Phillips curve (PC) and row H uses the pre-IT period monetary reaction (MR) function.

Further analysis suggests that the change in the behavior of the central bank and the private sector clearly contributed to the decline in macroeconomic variability in Canada since 1991, but that the difference in the magnitude or nature of shocks had an ambiguous effect. We explore the reasons behind the fall in model volatility, distinguishing between the impact of changes in the shocks, the monetary policy reaction function, and private sector behavior. Row E in Table 8.2 shows the implied level of macroeconomic volatility if the economy having the 1990s structure had been subject to shocks that prevailed in the 1980s, while the next row reverses the experiment, looking at the volatility of an economy with a 1980s’ structure buffeted by 1990s-like shocks. The results suggest that the shocks in the 1990s imposed less output variability but more inflation and interest rate volatility. On the other hand, the change in the structure of the economy helped reduce the variability of inflation and nominal interest rates without a noticeable effect on output volatility.

More forward-looking inflation expectations played a key role in reducing macroeconomic variability. Simply replacing the pre-IT monetary reaction function with the IT-period rule without any attendant change in the forward-looking nature of the Phillips curve (row G) produces only a small reduction in inflation and interest rate volatility at a cost of higher output gap variability. By contrast, the rise in the expectations component of the Phillips curve (row H) explains almost all of the implied reduction in macroeconomic volatility.

The model helps evaluate the impact of possible changes to the IT framework. One particular modification currently under discussion is an extension of the time over which the central bank aims to achieve its target. In our framework, such an extension is equivalent to reducing the coefficient on inflation in the monetary reaction function. As reported in Table 8.3, assuming no change in the Phillips curve, lowering the coefficient on inflation in the monetary response function from 1.5 to 1.2 would slightly reduce the volatility of interest rates and increase the volatility of the output gap and inflation by negligible amounts.

Table 8.3.Standard Deviation of Inflation, Output Gap, and Interest Rate under Hypothetical Regimes
EquationParametersInflationOutput GapInterest Rate
(In percent)
ITλ=.71; Yθ1=
Longer reactionλ=.71; Yθ1=
Longer reaction; less credibility;λ=.68; Yθ1=1.23.540.841.62
Source: IMF staff calculations.

A key issue is whether a modification of the framework, such as delaying or muting response to shocks, might degrade the regime’s credibility. Should the markets perceive the change as a weakening of the Bank’s commitment to the inflation target, they would adjust their expectation-formation process. That adjustment would likely result in an increase in inflation inertia, as expectations became less forward-looking. As a consequence, the reduction in macroeconomic volatility over the past 15 years might be partially reversed, as the private sector would respond more to demand and supply shocks and less to policy action. As the last row of Table 8.3 demonstrates, a reduction in the coefficient on expected inflation in the Phillips curve as small as one-fifth of the earlier gain would completely negate the benefits to interest rate volatility and increase perceptibly the volatility of inflation.

Conclusions and Policy Implications

This analysis suggests that the reduction in macroeconomic volatility in Canada after the introduction of inflation targeting is largely attributable to the reaction of the private sector to the establishment of a credible monetary policy framework. With greater confidence in the central bank’s commitment to price stability, the private sector started forming inflation expectations in a more forward-looking manner, thereby reducing considerably the degree of nominal inertia in the Phillips curve. This has attenuated private sector reaction to demand and supply shocks, thus lessening the volatility of inflation and muting the business cycle.

An implication of this analysis is that further refinements of the IT framework should be carried out in a gradual manner that does not compromise monetary credibility. The private sector’s expectation-formation process and reaction to shocks are crucial determinants of macroeconomic stability. Potential benefits from adjustments in the regime should be weighed carefully against the possibility that a loss of credibility from a perceived waning of commitment to the inflation target would worsen the macroeconomic environment. This also underscores the importance of communicating carefully the rationale for any proposed changes.

See, for example, Dodge (2005).

In the implementation, π was annualized quarterly change in the CPI, in percent; i was the target for overnight rate set by the Bank of Canada, in percent; and y was the output gap calculated by the Bank of Canada as percentage of potential real GDP.

The coefficient below one on inflation in the pre-IT monetary policy rule violates the Taylor principle and makes the model dynamically unstable. In simulations, a value of 1.2 was imposed on that coefficient. The negative estimated coefficient on the output gap in the IT monetary policy rule was insignificantly different from zero and was replaced with zero in simulations.

    Other Resources Citing This Publication