RDP 2002-01: Inflation Targeting and the Inflation Process: Some Lessons from an Open Economy 2. Which Inflation Rate to Target in an Open Economy?

An important issue that confronts an inflation-targeting central bank in an open economy is that changes in the exchange rate can have a significant effect on inflation outcomes via the prices of traded goods. If the central bank is pursuing a strict inflation target, the policy responses required to offset the effects of exchange-rate-induced changes in inflation may be damaging to the non-traded sector of the economy, and generate a large degree of volatility in output.

Consequently, Ball (1998) and Svensson (1998) raised the possibility that it may be preferable that a central bank target a measure of inflation that abstracts from these direct exchange rate effects. This section reviews their argument and outlines the main considerations that might affect the choice of which rate of inflation to target.

These issues can be illustrated by the following simple model similar to that in Ball and Svensson.

where y is the output gap, r the real policy interest rate, e the exchange rate and π inflation.

The first equation is an aggregate demand equation, where monetary policy is assumed to affect output with a one-period lag. A depreciation of the exchange rate also leads to an expansion in output with a one-period lag, through its effects on net exports. The second equation is an open-economy Phillips curve. Changes in the exchange rate are assumed to be passed through immediately into the prices of imported goods in the consumer price basket. For the moment, inflation expectations are assumed to be backward-looking, depending on past values of the aggregate inflation rate; this will be discussed further below. Note that exchange rate changes affect inflation more rapidly than they do output. The third equation implies that the exchange rate appreciates in response to a rise in domestic interest rates.

The central bank is assumed to have an objective function of the standard form:

where it sets its policy instrument to minimise deviations of inflation from its target, and minimise the output gap. When λ is zero, the central bank can be characterised as a strict inflation targeter where output considerations are always secondary to minimising inflation variability.

Consider a temporary depreciation of the exchange rate that results from a portfolio realignment that lasts for only one period (that is, a decline in ε3). The depreciation will generate an immediate increase in inflation as imported goods prices rise. If a rigid inflation target is in place, the increase in inflation can be counteracted by a rise in interest rates to offset the downward pressure on the exchange rate from the portfolio shift. This policy change is reversed in the following period when the downward pressure on the exchange rate dissipates. The policy-maker can thus successfully stabilise the inflation rate, but at the cost of inducing additional volatility in output, as output responds to the shifts in interest rates.

If instead the policy-maker were targeting non-traded inflation rather than aggregate inflation in Equation (4), the policy response would be considerably muted. A more muted response would also occur in a more flexible inflation-targeting regime. Output variability, in both these cases, would be correspondingly less but at the expense of greater volatility in the aggregate inflation rate. Some policy response would still be necessary to reduce the volatility resulting from the effect of the depreciation on output and to the extent that non-traded goods prices or inflation expectations are also affected by movements in the exchange rate.

Hence targeting aggregate inflation as against non-traded inflation presents a choice between inflation variability and output variability. Responding to exchange-rate-induced fluctuations in inflation increases output variability, ignoring them increases aggregate inflation variability.

Ball (1998, 2000) argues that responding to a measure of ‘long-run’ inflation ‘purged of the transitory effects of exchange-rate fluctuations’ is the optimal strategy for a central bank in an open economy. To further bolster his argument, he raises the possibility that, in practice, the increased output variability from responding to aggregate inflation may destabilise inflation in the medium term (although such a result is not possible in his simple framework).

However, to make such an assessment, one needs to be able to assess the relative costs of inflation and output variability. While trade-off curves can be drawn illustrating the various combinations of output and inflation variability that correspond to different objective functions or rules for the central bank, the paucity of knowledge about the relative costs to society of inflation and output variability prevents an easy comparison of these combinations. The coefficient λ in the objective function (Equation (4)) is a critical but unknown variable. The general assumption is that some degree of inflation variability should be permitted, the question is how much?

One also needs to know which measure of inflation enters the objective function. The aggregate consumer price index is designed to be representative of the average consumption basket, so would appear to be the most obvious measure of inflation to use. However, various sectors of the economy, most notably producers in the non-traded sector, may face considerably different price baskets and obviously would be relatively disadvantaged if an aggregate measure were targeted, rather than a non-traded measure.

Nevertheless, curves showing trade-offs between output variability and the variability of various measures of inflation can be generated, and presented as a menu of options to policy-makers. The rest of this section discusses some of the key features of the economy that affect the shape and position of these trade-off curves in an open economy, and hence the relative merits of targeting aggregate or non-traded inflation.

Firstly, the nature of the shocks hitting the economy will be important, both in terms of their source and their persistence. Bharucha and Kent (1998) examine this with a calibrated model similar to that above. They demonstrate that if the shocks occur primarily to the exchange rate (Equation (3)), then a non-traded inflation target may be preferable in the sense that output variability is substantially lower. If, on the other hand, the shocks primarily occur in the non-traded sector of the economy, then a non-traded inflation target will place much of the burden of adjustment on the traded goods sector.

The persistence of the shocks is also an important consideration. Temporary changes in the exchange rate that are likely to return to equilibrium within a short period do not necessarily warrant a policy response. The inflationary impulse from an exchange rate temporarily below equilibrium should be offset by the disinflationary effect of the subsequent appreciation back to equilibrium. On the other hand, if changes in the nominal exchange rate are expected to be permanent, monetary policy needs to ensure that the resultant inflationary pressures do not lead to a permanent increase in the inflation rate.

While it is easy to state this principle, its practical implementation is particularly problematic. As Ball (2000) notes, it would be useful to find an alternative measure of inflation that simplified this problem in practice. In the next section, we examine whether movements in unit labour costs may serve as a useful proxy.

A second element that affects the nature of the trade-off is the extent to which aggregate and non-traded inflation are affected by movements in the exchange rate.

The aggregate inflation rate will be affected directly according to the extent of import penetration of the consumer goods market. However, exchange rate changes may still have a significant direct impact on non-traded inflation to the extent that the non-traded sector is dependent on imported inputs in production.

The speed and extent of the pass-through of exchange rate changes to final goods prices is also important. A more protracted pass-through reduces the impact of a given exchange rate change on the inflation rate and thereby reduces the size of the necessary policy response to it. Some evidence of a change in the pass-through of exchange rate changes in the 1990s is discussed in Section 4.

Third, the inflation expectations process will play a critical role. An important aspect of inflation targeting is to maintain stability in inflation expectations, and thereby anchor ongoing inflation. Therefore, the appropriate inflation-targeting strategy will depend on how inflation expectations are formed, the degree to which they are forward-looking, and how well anchored they are.

If inflation expectations are primarily backward-looking and are dependent on movements in the aggregate inflation rate, exchange rate changes will tend to have a larger and more persistent impact on both aggregate and non-traded inflation, as they get built into expectations. If on the other hand, inflation expectations are forward-looking, temporary exchange rate changes (which are recognised as being so by wage and price-setters) will not lead to much movement in inflation expectations. This is a key part of the process that affects the extent to which exchange rate changes lead to a temporary boost to inflation rather than a permanent pick-up.

Similarly, if the inflation target is perceived as credible, inflation expectations will be better anchored on the target inflation rate and again will not respond as much to temporary deviations in the actual inflation rate. In such circumstances, the credibility of the inflation target is somewhat self-fulfilling. Shocks to the inflation rate, from changes in the exchange rate for example, would not be expected to lead to a prolonged deviation of inflation from target. Because this belief is held, expectations do not adjust, and the inflation rate is more stable.