Content uploaded by Jose De Gregorio
Author content
All content in this area was uploaded by Jose De Gregorio on Mar 24, 2016
Content may be subject to copyright.
Implementation of Inflation Targets in Emerging
Markets
José De Gregorio*
Central Bank of Chile
June 2008
Abstract
This paper looks at some issues surrounding the implementation of
flexible inflation targets, with a particular focus on emerging markets. It
starts by clarifying how to define an inflation target and how this is related
to balancing the tradeoff between inflation and output. Flexible inflation
targets, which define the objective of monetary policy in terms of a range
for inflation and a time horizon for reaching the target when inflation
deviates from the range, balance the benefits of price stability with the
costs of achieving it. The paper also discusses the need for transparency
and effective public communication to ensure the accountability and
effectiveness of monetary policy. It also discusses the role of the exchange
rate in an inflation targeting framework.
JEL Classification numbers: E50, E52, E58, E61.
Keywords: Central bank objectives, inflation target, exchange rate,
intervention.
2
1. Introduction
The main objective of the great majority of central banks in the world is controlling
inflation. In some cases financial stability is an added objective; in others, employment or
economic development objectives are also included. In the case of Chile there are two
explicit objectives, namely, price stability and the “normal functioning of external and
internal payments.” The latter objective, taken from the central bank’s charter, corresponds
to financial stability,1 which involves two dimensions. The first is the stability of the
domestic financial system, which in simple terms may be described as avoiding financial
distress and dealing with it when it occurs. The second is the normal functioning of the
system of payments to and from the rest of the world, which in simple terms means
avoiding balance of payments crises and difficulties of access to international financial
markets.
To address the price stability objective, Chile follows an inflation targeting scheme.
Chile is part of a worldwide trend in which a large and growing number of countries have
adopted this approach to conducting their monetary policy (International Monetary Fund,
2005, chap. 4). In an inflation targeting regime, the central bank publicly announces a
numerical objective regarding inflation, which may be either a specific number or a range.
Although the target range itself is quite clearly defined, generally the percentage of the time
that the central bank expects inflation to lie within the range is not made explicit, and, of
course, one cannot expect inflation always to stay within the range, since inflationary or
deflationary shocks will inevitably lead to deviations.2 However, the central bank does
generally set an explicit horizon within which the target is to be met and deviations are to
be corrected.
In this paper I intend to clarify some issues regarding the definition of inflation
objectives and the conduct of monetary policy under inflation targeting schemes. In
particular, I will address the widespread but incorrect perception that such schemes imply
that the central bank does not consider unemployment or output growth in conducting its
policy. Although it is well understood in academic circles that the central bank does take
these into account, it is less well understood among policymakers, and even less among the
general public. I also discuss some issues concerning the use of inflation targets in practice,
in particular the role of public communication and the exchange rate. To summarize, I
attempt to show that:
• The inflation objective can be described in terms of a desired distribution for
inflation. This may be thought of as defining an average value for inflation and its
variability (variance). But in practice the target is defined by a mean value or a
range.
• Setting the target as a mean and a variance is equivalent to defining the target in
terms of a range and the percentage of the time one expects inflation to be in the
range. This is comparable to setting the target around an inflation projection for the
future, where the future time frame, or “policy horizon,” depends on the variance of
3
the inflation target. The greater the fraction of time inflation is sought to be within
the range, the shorter the policy horizon must be.
• A flexible inflation targeting scheme, in which the target is defined along with a
time horizon, reflects an objective function of the central bank that values not only
price stability but also output and employment stability. In particular, a direct
relationship also exists between the policy horizon and the central bank’s tolerance
of deviations of inflation from the target, on the one hand, and the importance
attributed by the authorities to output deviations, on the other.
• Exchange rates are an important element in an inflation targeting regime in an open
economy as long as fluctuations in the value of the currency have effects on
inflation. Although many countries with floating exchange rates and low inflation
have seen a decline in the pass-through from exchange rate fluctuations to changes
in domestic prices, large exchange rate swings may still have effects on inflation,
and hence should be considered when making monetary policy decisions.
2. Defining the Range for the Inflation Target and the Policy Horizon
The inflation target is fixed over a stated time horizon because it is recognized that inflation
cannot be controlled in the short term, since monetary policy acts with a lag. Furthermore,
and as discussed further below, a gradual adjustment of inflation when it deviates from the
target avoids the costs, in terms of reduced economic activity, that would be incurred if
inflation were returned to its target immediately. In other words, even if monetary policy
did not operate with a lag, it would still be desirable to adjust gradually.3 Moreover, in
general, when the inflation target is specified in the projection horizon, explicit reference is
made to a precise point, which always corresponds to the center of the target range.
This section develops a simple framework that explicitly shows the different ways
of defining a given inflation target. In the first subsection, the central bank is assumed to
take inflation as given, and some equivalences are shown in the definition of the inflation
target that are useful for understanding its formulation. The next subsection complements
this discussion by adding economic structure and endogenizing the inflation process.4
2.1. A simple framework for understanding the inflation target
Consider a central bank whose target for inflation is defined as a range between
π
and
π
,
with its center equal to *
πππ
= ( + )/2 . Some central banks define the target in this way; for
example, in Canada, Israel, and New Zealand this range is from 1 to 3 percent, in South
Africa it is from 3 to 6 percent, and in Chile the range is from 2 to 4 percent.5 Other
countries define the target as a single number, without specifying a range; for example, the
United Kingdom sets a target of 2 percent, and Norway and Iceland 2.5 percent.
4
One can think of the inflation target and its range as corresponding to a probability
distribution for inflation, with the objective understood as an expected value and a
variance.6 However, in reality most central banks define a range rather than a variance,
because a range is more easily understood by the general public. Also, defining a
distribution rather than a range requires much more information and certitude than central
banks have in practice. As this section should make clear, in order for these two concepts,
range and variance, to be equivalent, one must define not only the range, but also the
fraction of time that inflation is expected to be within the range—in other words, its
probability of being within the range at a given time. I will denote this probability by x. In
practice the value of x is not defined, although, as shown here, when a policy horizon is
defined, the tolerated variability of inflation is given implicitly.
Once a target range is known, the first question to ask is what this range means.
Central banks are reluctant to be specific, but it is useful to think of the central bank as
wanting inflation to lie within the range x percent of the time. It suffices to specify these
two parameters—the target range and the percentage x—to establish both the center of the
range (that is, the expected value of inflation) and the variance. Specifically, given the
assumption of a known symmetric distribution (for example, a normal distribution), and
given the range and the percentage of the time inflation is expected to be within the range,
one can immediately determine the variance of inflation. The greater the value of x, the
lower the variance of inflation must be.
As shown by Svensson (1997), the inflation target may be operationalized by setting
the objective in terms of an inflation projection over a given horizon, which in practice is
usually between four and eight quarters. One reason for such a relatively long horizon is
that monetary policy affects inflation with a lag. A second is that adjusting inflation rapidly
to its target would entail undesired costs in terms of reduced economic activity and high
unemployment, even if inflation were perfectly controllable. In other words, inflation
targeting models do take unemployment into account. In fact, in the following section it is
assumed, for the sake of simplicity, that the central bank determines inflation
instantaneously, yet the optimal adjustment is still gradual.
Another relevant equivalence is that between the variance of inflation from its target
and the policy horizon. Suppose that inflation follows a first-order autoregressive, or
AR(1), process given by:7
1,
*( *)
ttt
π
πρπ π ε
−
−= − + (1)
where t
ε
is an independent and identically distributed (i.i.d.) random shock with a mean of
zero and a variance of 2
ε
σ
, and
ρ
is the autocorrelation coefficient, which is between zero
and one. The expected value of inflation is *
π
and its unconditional variance is:8
2
2
2
1
ρ
σ
σε
π
−
=. (2)
5
In making its monetary policy decisions, the central bank projects inflation into the
future. The central bank observes t
ε
, but from t + 1 forward, the best it can do is to assume
that this shock will be zero. The inflation projection one period ahead, conditional on all of
the information available at time t, will thus be (1 ) *
t
ρ
πρπ
+
−, and the projection T
periods ahead will be
(1 ) *
TT
ttT t
π
ρπ ρ π
+
Ε= +− . (3)
As the horizon lengthens (that is, as T rises), T
ρ
approaches zero and the inflation
projection approaches *
π
. Consider now the case where the central bank announces that it
wishes inflation to be around *
π
in period T. More precisely, it wants the forecast of
inflation to converge to *
π
. In this respect, the objective of the central bank is the
convergence of Ttt +
Ε
π
, where the relevant information set contains t
π
. Therefore the
conditional forecast (1 ) *
TT
t
ρ
πρπ
+− is the variable on which the operational objective of
the central bank is based.
Given that only as T goes to infinity does the projection fully converge to *
π
, it is
assumed that a tolerance margin is allowed, expressed as the variance of the conditional
forecast s. As a consequence, the variance of projected inflation that is obtained from
equation (3) is:
ρ
σπ
log2
loglog 2
−
=s
T, (4)
ρ
ρσ ε
log2
)1log(loglog 22 −+−
=s, (5)
In the latter expression it should be noted that, given that 2
π
σ
<s and 1<
ρ
, both the
numerator and the denominator are negative; accordingly, T is well defined, since it is
necessarily positive. It follows that the greater the variance of target inflation, 2
π
σ
, or in
other words, the greater the range of inflation for a given x, the longer the policy horizon
over which the conditional forecast is expected to converge toward *
π
. Likewise, the
greater the value of
ρ
, the longer the policy horizon, since the increased persistence of
inflation slows down its convergence to the center of the target range.
In brief, it has been established that defining an inflation objective in terms of its
mean and variance is equivalent to defining a range within which inflation is expected to
remain during a given percentage of the time. This, in turn, is directly related to the
projection horizon over which inflation is expected to converge toward its expected value.
6
Therefore, if one knows the inflation distribution, and assuming inflation follows an AR(1)
process like that described in equation (1), we have established that all three definitions of
the target shown below are equivalent:
i. The inflation target has an expected value of *
π
and a variance of 2
π
σ
.
ii. The inflation target is given by the range [
π
,
π
] in which it is expected to be
x% of the time.
iii. Projected inflation is expected to be around *
π
with a variance of s over a
horizon of T periods ahead.
If one defines all the parameters of any one of these three definitions, one can then
determine the parameters of the other two. Therefore, if one knew the economy’s behavior
exactly, it would be impossible to separate the inflation targeting decision from the policy
horizon. However, in reality this behavior is not known with accuracy, and this explains the
lack of numeric precision in all parameters of the objective function. Moreover, it may be
argued that specifying these parameters may lead to inconsistency, precisely because of the
uncertainty that exists with respect to the actual structure of the economy. For example, the
target may be defined in terms of ii. or iii., but the value of T could be inconsistent with the
target specified in ii., simply because the economy’s structure is not fully known.9
As an alternative to defining all of the inflation target’s parameters precisely, central
banks have moved to increase transparency and provide public explanations of their
deviations from the target in their periodic inflation reports, also called monetary policy
reports. For example, whenever inflation in the United Kingdom deviates from its target,
the governor of the Bank of England writes a formal letter to the Chancellor of the
Exchequer to give an account of why the deviation has occurred. All these arrangements
replace a more mechanical and explicit behavior with respect to the inflation target, in a
world with much more uncertainty than is assumed in the models, with a public and
transparent rendering of accounts. This practice recognizes that there are risks and
contingencies that central banks’ projection models cannot predict, nor can all policy
responses to more complex scenarios than simple deviation of inflation from its target—
particularly those associated with financial stability—be anticipated. It reflects the need to
balance a well-defined rule, by which the central bank’s performance can be evaluated,
with proper flexibility in a real world that contains much uncertainty.
2.2. Is inflation all that matters?
Whereas the previous section assumed that the central bank takes the inflation process as
given, this section goes further and adds structure to the economy, to understand where
inflation comes from and how it relates to the output gap. This is done by deriving equation
(1) above from the fundamental parameters of the economy, which in this case are given by
preferences between unemployment and inflation along a Phillips curve. The value of
ρ
is
determined by the monetary authorities, who gradually adjust inflation so as to reduce the
cost of that adjustment in terms of output. The possibility of demand shocks is ignored.
7
Here I will use the model presented in De Gregorio (1995), which allows the
optimal course of inflation to be derived from a social loss function, where the losses derive
from inflation and output gaps, plus a Phillips curve that incorporates indexation of
inflation. I will assume that the central bank has determined an optimal inflation rate *
π
,
but that it adjusts inflation gradually toward this rate in order to reduce welfare losses.
The social loss function is given by10
22
()(*),Layy
ππ
=−+− (6)
where y is GDP and
y
its full-employment level. It should be noted that here there is no
inflationary bias as in Barro and Gordon (1983), since the central bank’s preferences are
socially optimal (Rogoff, 1985).11
Inflation is determined by the following Phillips curve:
vyy
tttt
+
−
+Ε−+= −− )()1( 11
δ
π
α
α
π
π
. (7)
The term v corresponds to an i.i.d. inflationary shock with zero mean and variance 2
v
σ
. This
Phillips curve incorporates persistence of inflation through the term 1−t
α
π
, which may be
interpreted as the result of indexation of prices and salaries. A simple case is that of certain
regulated utility prices, which are indexed to past inflation. The persistence term could also
represent the outcome of overlapping decisions on prices and salaries as in the extension of
Taylor (1980) proposed by Fuhrer and Moore (1995).12 The parameter α can also be
interpreted as related to the credibility of the inflation target. If the public is confident of
the authorities’ commitment to the inflation target, expectations will be more forward
looking than if credibility is lacking, in which case the public may assume that past
inflation will tend to be more persistent. The Phillips curve’s slope is δ and, to simplify the
notation, its inverse is defined as θ.
Solving for the output gap in the Phillips curve and replacing it in the objective
function, we have that the first-order condition for the central bank’s optimization is given
by the following (subscript t is eliminated, and instead subscript -1 is used for a one-period
lag):
222
11
2
1
*[(*)(1)(*)]
1t
aaav
a
π
παθππαθππθ
θ
−−
−= −+− Ε−+
+. (8)
Taking expectations from the above expression to solve for rational expectations of
inflation, and replacing this expression in the same first-order condition, we obtain the
following expression for optimal inflation:
8
1
1
*(*)
11
v
oo
ππ π π
α
−
−= − +
++
//
, (9)
where
αθ
2
1
a
o≡
/. (10)
Optimal inflation has the same form assumed in equation (1), where the autocorrelation
coefficient and the error depend on the fundamental parameters of the model and on the
inflationary shock. That is,
2
1 and
11 1
av
oa o
θ
α
ρε
θ
αα
== =
++ +
//
(11)
It should be noted that expected inflation is equal to the central value of the target
range, *
π
, and the variance is
22
2
2
22
11
v
aa
π
σ
θ
σ
θ
ρ
⎛⎞
=⎜⎟
+−
⎝⎠. (12)
From these equations it can be easily verified that
ρ
and 2
π
σ
are increasing functions of a,
α, and θ, and that 2
π
σ
is increasing in the variance of the inflationary shock )(2
v
σ
. In section
2 it was shown that increasing the variance of target inflation is similar to extending the
policy horizon or widening the target range, all else equal. An increase in the variance of
inflation produces the same results. Since an increase in any of the three parameters (a, α,
θ) increases both 2
π
σ
and
ρ
, one can conclude that increases in those parameters also
lengthen the policy horizon T, as can be seen from equation (5).
These results can be interpreted as follows:
• When the central bank is not concerned about unemployment (that is, a = 0), the
value of
ρ
will be zero, and expected inflation will adjust to *
π
in each period.
Therefore the policy horizon collapses to zero: the central bank attempts to meet the
inflation projection in each period. In this case inflation would equal *
π
, because
monetary policy would fully offset the effect of any inflationary shock. As a
increases, the policy horizon lengthens, or, similarly, the inflation target variance
increases.
• The greater the volatility of inflationary shocks, the greater the variance of target
inflation, which in turn generates a longer policy horizon.
9
• Something similar occurs when the degree of backward-lookingness, measured by
α, increases, as this also produces a slower adjustment and greater variability of the
inflation target—the target range increases.
• When the slope of the Phillips curve decreases (δ falls and θ rises), the output gap
has a smaller impact on inflation. Therefore the central bank will accept a greater
inflation variance, or a longer policy horizon, because it does not want to vary the
output gap too much to offset inflationary shocks.
Why does all this happen? Because even though the central bank defines its
objective in terms of an inflation target, it also considers the costs, in terms of
unemployment, of attaining the target. In other words, having an inflation target does not
mean that unemployment costs are disregarded. It should be noted that although, in this
model, monetary policy operates without lags, the inflation target is not intended to be met
in the short term.
To sum up, the persistence of inflation declines, and the policy horizon shortens, in
response either to a decline in a, that is, an increase in inflation aversion; or to a decline in
α, the degree of indexation; or to an increase in the slope of the Phillips curve δ = 1/θ, that
is, a decline in θ. The decline in inflation persistence should also be accompanied by a
decline in the variance of inflation, as long as the variance of the inflation shock )(2
v
σ
remains constant.
The advantage of specifying the macroeconomic stability objective in terms of an
inflation target is that it avoids the inconvenience of defining two objectives that might be
inconsistent. For example, defining a limit for output variation along with an explicit
inflation target could make both objectives incompatible with the economy’s structure.
There are additional complications when the target is specified in terms of the
activity level, since full-employment output is not known with certainty, and therefore
pursuing very low unemployment could lead to the traditional inflationary acceleration, or,
conversely, underestimation of full-employment output could lead to too rapid a
deceleration of inflation or even deflation.
However, the fundamental reason for choosing an inflation targeting regime is that
otherwise inflation would be undetermined. Monetary policy deals with prices and
inflation. Defining an inflation target provides an anchor for inflation.
3. Issues in Inflation Targeting: Institutions and Exchange Rates
In this section I will address two relevant issues for the implementation of inflation targets.
The first is the use of public communication to report on the fulfillment of the inflation
target. This is particularly relevant given that central banks do not have full control over
inflation. I also discuss the limits to transparency, as well as the role of the exchange rate in
10
the inflation targeting regime. There is a natural place for exchange rate policy, and even
exchange rate stabilization, in an inflation targeting regime, once one takes into account the
inflationary consequences of exchange rate fluctuations. I also discuss the role of foreign
exchange intervention as an exceptional measure.
3.1 Communication and transparency as substitutes for precision
Given the results presented above, inflation targeting central banks should define their
objectives by stating a target range for inflation and indicating the percentage of the time
they intend inflation to lie within that range. This is equivalent to saying that they should
announce the mean and variability of inflation. Additionally, convergence of inflation to the
target can be defined as an intermediate objective, in a policy horizon that takes into
account both the costs of attaining the target and the lags with which monetary policy
operates. In general, this horizon is defined as from one to three years. However, the
definition of the target is not precise. It is not completely clear what fraction of the time (x
in the model) the central bank expects inflation to be within the tolerance range, nor is the
central bank precise about the distance from the center of the band it expects the forecast to
be (s in the model).
In practice, there is uncertainty regarding the transmission mechanisms of monetary
policy and the structure of the economy: the “true” parameters of the model in section 2 are
unknown. This is particularly valid in emerging market economies, because of their
frequent structural and policy changes. This makes it difficult to be precise in defining all
of the parameters of the inflation target. When monetary policy is anchored by a fixed
exchange rate or by targeted monetary aggregates, monitoring is very simple. It suffices to
see whether the exchange rate is at the announced level or whether the monetary aggregates
follow the announced path. With this information in hand, the public can evaluate whether
the monetary policy objectives are being met. However, exchange rate and monetary
anchors are less and less frequently used, since they are less efficient as a means of
conducting monetary policy. Currently, most central banks instead set interest rates and
pursue an implicit or explicit inflation target. Monitoring inflation is easy, but explaining
deviations from the target is a more difficult task.
Inflation usually does deviate to some degree from the target. In Chile this has been
the case since 2007, a period that has witnessed a sharp increase in food prices and a
continuing rise in the oil price. But such deviations can have different causes: they may be
due to some genuinely uncontrollable event or influence (such as a food price shock), or
they may be due to the central bank failing to adopting a monetary policy consistent with its
inflation target. For this reason, public communication and accountability are particularly
important under an inflation targeting scheme. This is why most central banks that have
adopted inflation targets prepare periodic inflation reports to inform the public about actual
inflation performance and explain its consistency with monetary policy decisions. This
increases the credibility of monetary policy and strengthens commitment to the inflation
target.
But predictability and transparency can also improve the effectiveness of monetary
policy. Predictability is desirable because it avoids abrupt adjustments in asset prices. Also,
11
it allows that the monetary policy and its expected future path be transmitted to the entire
yield curve. In fact, monetary policy determines the interbank interest rate quite accurately,
but aggregate demand depends on longer-term rates. Working capital depends on rates
ranging from three months to one year, and investment on even longer-term rates. To the
extent that monetary policy signals and strategy provide useful information for the market
as it forms its expectations, the whole yield curve will be affected, and consequently
monetary policy will be more effective.
Transparency also has its limits, and these are precisely related to the effectiveness
of monetary policy, its decision-making independence, and the need for monetary
policymakers’ deliberations to be as frank and productive as possible. For example, in
many countries the minutes of monetary policy meetings do not indicate which members
take one position or another. This is reasonable, because what is more important is the tenor
of the discussion, and revealing names could discourage members from speaking candidly
at meetings, thus affecting the central bank’s capacity to act with genuine independence.
Full transcripts of the discussions are good for transparency and accountability, but these
can be provided after a relatively long lag: the Federal Reserve Board, for example, releases
transcripts after five years. How transparency is implemented in each country is specific to
each case and its institutional tradition, but effective monetary policy in an inflation
targeting framework invariably requires a high degree of transparency.
Finally, adoption of an inflation target in emerging market economies poses certain
institutional requirements. The first is an independent central bank that is empowered to
establish a policy horizon that extends beyond the political and electoral horizons. Also
necessary is a solid fiscal situation, to avoid subordination of monetary policy to fiscal
policy and so eliminate the possibility of inflationary financing, which undermines the
credibility of the inflation target.
3.2 The role of the exchange rate and asset prices
Another key element in an inflation targeting regime is the exchange rate regime. If the
economy is financially fully open, one cannot simultaneously adopt an independent
monetary policy (that is, set the policy interest rate) and control the exchange rate —this is
the well-known “impossible trinity.” Therefore a floating exchange rate regime is a
prerequisite for an effective inflation targeting scheme. However, exchange rate
fluctuations, in particular periods of persistent appreciation or depreciation, may create
additional risks to macroeconomic stability that may require additional policy action.
After the bursting of the dot-com bubble in the United States some years ago, there
was considerable discussion of whether the Federal Reserve should have raised interest
rates preemptively, even in the absence of signs of rising inflation, so as to prevent the
bubble from continuing to grow. This question has become even more urgent with the
bursting of the housing bubble and the financial crisis that followed. Those who believe
that preemptive action should have been taken argue that such action could have prevented,
or at least attenuated, the financial crisis and economic slowdown that followed the bursting
of the bubbles. Those on the other side maintain that monetary policy would have been
neither effective nor necessary, because a slowdown would have followed even if the
12
bubble had deflated gradually, nor is there any certainty that an increase in interest rates
could have prevented the bubbles.
This is certainly a second-order problem compared with the dilemma most emerging
market economies face, namely, what to do about the exchange rate. Most countries have a
preference for a “competitive” real exchange rate, one that gives an advantage to the
country’s exports in foreign markets. This preference is natural in the wake of the many
disastrous experiences with massive currency overvaluation under schemes of exchange
rate inflexibility. Many considerations justify a flexible exchange rate regime, and this is
not the place to discuss them.13 I will focus here only on the role of the exchange rate in an
inflation targeting regime.
There has been much discussion of whether the exchange rate should or should not
affect monetary policy decisions, that is, the setting of the policy interest rate.14 In the
context of a Taylor rule, the question is whether or not the exchange rate should be an
argument in the rule. In a more general context, it is important to remember that, rather than
any mechanical rule for setting the interest rate, what needs to be considered is what
trajectory of interest rates is consistent with the inflation target. To the extent that the
exchange rate affects the inflation outlook, it is a relevant variable to consider when
deciding monetary policy. Its effect on consumer price inflation is much more evident than
its effect on the prices of other assets, such as house or stock prices. Suppose, for instance,
that the currency appreciates sharply and unexpectedly. Even if the pass-through coefficient
is relatively low, as it is particularly in countries that have an inflation target set by an
independent central bank, such an appreciation should reduce inflationary pressure and
therefore create room for a monetary easing, which should in turn reduce exchange rate
pressure and provide a better guarantee of achieving the inflation target.15
Thus the exchange rate does affect monetary policy decision-making, but it does so
because of its effect on inflation, without additional considerations of competitiveness or
exchange rate volatility. Therefore when the exchange rate has an effect on inflation
projections, monetary policy should take it into account. For example, if the currency
depreciates and the depreciation begins to affect inflation, an increase in the interest rate
will contribute to attaining the inflation target both through its impact on investment
expenditure and consumption of durable goods, and through its effect on the exchange rate.
In addition, such a move on the part of monetary policy should stabilize the exchange rate.
In any event, it is worth stressing the importance of analyzing the reasons behind exchange
rate movements. If the depreciation is due to an adverse external scenario, it may have less
of an impact on the inflation outlook, and, accordingly, the monetary policy reaction would
not necessarily be an increase in the interest rate.
Nevertheless, the direct effects of exchange rate fluctuations on inflation are not the
only way the exchange rate can affect monetary policy decisions. Most central banks also
have the objective of safeguarding financial stability. In Chile this objective is described in
the central bank charter as the “normal functioning of internal and external payments.” This
is interpreted as meaning that the central bank must seek to must avoid financial and
exchange rate crises and extreme turbulence, as well as financial and exchange rate
imbalances that could jeopardize macroeconomic stability. Undoubtedly one way to help
13
ensure financial stability is through the adoption of a flexible exchange rate, but this does
not preclude the possibility of exchange rate bubbles threatening financial stability.
An exchange rate bubble is more complicated to deal with than an asset price
bubble. If the authorities are certain that, for example, a housing bubble exists, they can
increase interest rates in order to prick the bubble. But in an emerging market economy, if
the currency is experiencing a bubble, leading to a severe appreciation, raising interest rates
could worsen the problem by encouraging an increase in carry trade. However, central
banks can use other tools to stabilize the foreign exchange market. In particular, they can
intervene directly by increasing or reducing their holdings of foreign exchange reserves.
But this must be an exceptional measure, linked to changes in the position of international
liquidity, and undertaken within a predefined period of time.
More concretely, current experience, in Chile in particular, 16 suggests that three
main conditions must be met when implementing foreign exchange intervention in the
context of an inflation targeting regime with a commitment to control inflation. First, it
cannot pursue a specific value for the currency, because then the inflation target would be
subordinated to the exchange rate objective, thus falling victim to the “impossible trinity.”
Second, the depreciation that is expected to occur in the wake of the intervention must be
consistent with the inflation outlook; otherwise credibility will be undermined.
Implementing a mechanical rule for intervention, independent of the value of the currency,
preserves monetary autonomy to use interest rates to control inflation. Third, the
intervention must be completed within a predefined period: as already stated, intervention
is an exceptional measure, and announcing a date for its termination strengthens the
commitment to keep the intervention transitory. Experience also confirms that pursuing a
mechanical rule for intervening, with some degrees of technical freedom, may be desirable
to avoid speculation against the central bank.
Finally, a word about the objective of having a “competitive” real exchange rate that
promotes exports. This discussion is similar to that on the natural unemployment rate, in
that the fundamental determinants of the exchange rate are beyond the scope of monetary
policy. If the authorities tried to hold the exchange rate permanently to a level inconsistent
with its fundamentals, the result could be inflation, which would restore the real exchange
rate to its equilibrium level. As argued before, when the authorities are relatively certain
that the exchange rate has overreacted, and provided there are no inconsistencies with the
inflation target, transitory intervention may provide some relief. But a permanent attempt to
avoid an appreciation will lead to inflation, so that a real appreciation takes place after all,
through higher inflation and not through a nominal exchange rate adjustment. In addition,
as experience shows, such an attempt may encourage the entry of short-term capital,
generating additional exchange rate pressure.
Monetary policy cannot affect the real exchange rate in the medium or the long run,
since it is determined by its fundamentals, such as the degree of trade openness,
productivity growth, fiscal policy, net international assets, and the terms of trade. Interest
rate changes induce exchange rate fluctuations, but as long as they are incorporated
properly in the context of an inflation targeting regime, the movements of interest rates
should be stabilizing. The exchange rate, in turn, functions as a shock absorber.
14
4. Concluding Remarks
As I hope to have shown, targeting inflation on the basis of a range within which inflation
is expected to stay most of the time is similar to fixing an objective for projected inflation
over a given policy horizon, or indicating an expected value and a variance for inflation.
Either way, the definition actually used by central banks is not quite accurate, since the
structure of the economy is not known with sufficient certainty to allow precise definition
of the target parameters. Moreover, some margin of flexibility must be allowed to address
situations that are impossible to anticipate.
On the other hand, defining the monetary authority’s objective in terms of an
inflation target does not mean that the business cycle, particularly unemployment, is
irrelevant to it. This is reflected in the fact that the target is not intended to be met always
and under all circumstances, and in the fact that the target is established in the context of a
policy horizon generally of one to three years.
This paper has used a simple analytical model to clarify these points but has omitted
some relevant aspects of monetary policy practice, although these should not change the
conclusions. The economy is subject to many kinds of shocks besides inflationary ones.
This paper has not considered in detail the credibility of the central bank, but the very
decisions it makes and the formulation of its objective reveal information about its ability to
contain inflation, as well as about its commitment to the target. Incorporating these aspects
adds much more complexity, but generally their implications point in the direction of
rigorously meeting the target, since the more credibility the central bank enjoys, the less
costly are the adjustments to achieve the inflation target. In terms of the model presented
here, credibility can be thought of as reducing inertia and the degree of indexation, thus
enabling a faster return of inflation to the target range when deviations occur. It can also be
shown that increased credibility on the inflation objective allows a reduction in the
variability not only of inflation, but also of output (De Gregorio, 2007). Indeed, this is what
the world has witnessed with the Great Moderation.17
In this paper the need for optimal gradualism has been justified by inflation
persistence. Excessive activism, in the sense of having a very short policy horizon, can be
viewed as leading to greater volatility of interest rates and asset prices, which could lead to
instability in the financial system. In a more general dynamic stochastic model, one could
conceive of an optimal monetary policy whose horizon varies depending on the nature and
magnitude of the shocks it encounters. In actual practice, the problem may be solved with
escape clauses that allow for deviations from the target in exceptional situations, for
example when financial stability is threatened.
The analysis of the model was presented in the context of a closed economy.
Extension to an open economy and interactions with the exchange rate should not change
the main conclusions of this discussion, but they certainly add new sources of fluctuation.
This issue has been treated informally here in the light of accumulated experience in
15
emerging markets. In any event, incorporating elements of an open economy could provide
an additional reason to adopt a medium-term horizon. If the central bank adopted a very
short horizon or a very narrow target range, the principal mechanism of monetary policy
pass-through to inflation would be the exchange rate rather than aggregate demand. This, in
turn, could generate deviations of the exchange rate that might affect the external
equilibrium of the economy–an important concern for emerging market economies subject
to strong fluctuations in external financing. I have also discussed the role of the exchange
rate in the context of an inflation targeting regime. As long as changes in exchange rates
affect inflation, the exchange rate should be one of the variables that central banks take into
account when deciding the stance of monetary policy. Doing so may also have a stabilizing
effect on exchange rate fluctuations. However, in emerging markets there may be reasons
to use additional instruments in periods of excessive fluctuation that threaten financial
stability—financial market turbulence like that stemming from the U.S. subprime mortgage
crisis is a pertinent example. In such a case, intervention in the foreign exchange market
may be advisable as long as it is consistent with the inflation target. This has a parallel with
discussions in industrial countries on preventing bubbles from generating financial fragility.
No distinction has been made here between core inflation, which is calculated for a
subset of goods in the consumer price index (CPI), and so-called headline inflation. In
general, inflation targets refer to headline rather than core inflation, although the latter
tends to be more stable. Although theory tends to prefer core inflation, there are reasons
(which are not fully rational from an analytical standpoint) why central banks prefer
headline inflation. In the first place, use of any of the various alternatives to headline
inflation would pose a problem of credibility and public understanding. In addition, if one
were to eliminate goods whose prices are more volatile, such as fuels, from the target
measure of inflation, the result would only be to reinforce the second-round effects of a
shock on such prices. In other words, if the central bank wishes to minimize second-round
effects, it should be willing to respond to cost shocks, even if the response is not
immediate. Accordingly, fixing the target in terms of headline inflation provides an anchor
to all prices.
Central banks do not directly control inflation, as assumed in the model presented
here, but do so through the interest rate, which in turn affects aggregate demand and output.
To generalize the results, it would be sufficient to specify aggregate demand, which could
be affected through the interest rate so as to meet the expected inflation trajectory.
However, the main results would not be altered, because it is inflationary shocks that
generate the tradeoff between inflation and unemployment.
Finally, it could be argued that a flexible inflation targeting policy is inconsistent
with a central bank mandate that is limited to price stability and financial stability. In fact,
given that this mandate does not explicitly include output stability, as considered in the
model presented here, one could conclude that the horizon should be kept at its minimum.
This argument, however, contains two deficiencies. First, a monetary policy that targets
inflation in each period, disregarding its effects on output, may undermine financial
stability. Such a policy would lead to extreme volatility of interest rates and of output. This,
in turn, could increase financial vulnerability, for example by inducing abrupt changes in
the financial positions of firms and households. Second, central banks also have more
16
general objectives relating to the overall welfare of the population. In the case of the
Central Bank of Chile, its charter says explicitly that “When passing its resolutions, the
Board shall consider the general orientation of the Government’s economic policies.” This
“general orientation” has to do with improving the welfare of the population, which
necessarily implies taking into account the costs both of inflation and of output and
employment fluctuations when pursuing monetary policy. In short, although inflation may
seem to be the only concern of monetary policy in an inflation targeting regime, this paper
has shown that this is not the case. Output and unemployment fluctuations are implicitly
incorporated by allowing inflation to gradually adjust to its objective.
Consequently, a policy of flexible inflation targeting does consider full employment
among its objectives. However, it is still preferable to organize monetary policy around a
flexible inflation target. Doing so provides a nominal anchor to the economy, which in turn
strengthens monetary policy credibility, a crucial element to minimize the costs of attaining
price stability.
References
Amano, R. (2007), “Inflation Persistence and Monetary Policy: A Simple Result,”
Economic Letters, Vol. 94, pp. 26-31.
Ball, L. (1999), “Policy Rules for Open Economies,” in J. B. Taylor (ed.), Monetary Policy
Rules, Chicago, Ill.: University of Chicago Press.
Blanchard, O. and J. Simon (2001), “The Long and Large Decline in U.S. Output
Volatility,” Brookings Papers on Economic Activity, no. 1, pp. 135-164.
Barro, R. and D. Gordon (1983), “A Positive Theory of Monetary Policy in a Natural-Rate
Model,” Journal of Political Economy, Vol. 91, No. 4, pp. 589-610.
De Gregorio, J. (1995), “Policy Accommodation and Gradual Stabilizations,” Journal of
Money, Credit and Banking, Vol. 27, No. 3, pp. 727-741.
De Gregorio, J. (2007), “Defining Inflation Targets, the Policy Horizon and the Output-
Inflation Tradeoff,” Working Paper No. 415, Central Bank of Chile.
De Gregorio, J. (2008), “The Great Moderation and the Risk of Inflation: A View From
Developing Countries,” Documento de Política Económica No. 24, Central Bank of
Chile.
De Gregorio, J. and A. Tokman (2007), "Overcoming Fear of Floating: Exchange Rate
Policies in Chile,” in N. Batini (ed.), Monetary Policy in Emerging Markets and Other
Developing Countries, New York: Nova Science Publishers Inc.
Edwards, S. and E. Levy Yeyati (2005), “Flexible Exchange Rates as Shock Absorbers,”
European Economic Review, Vol. 49, pp. 2079-2105.
17
Fuhrer, J. and G. Moore (1995), “Inflation Persistence,” Quarterly Journal of Economics,
Vol. 110, pp. 127-160.
International Monetary Fund (2005), World Economic Outlook, Washington D.C.:
International Monetary Fund, October.
Kim, C.-J. and C. Nelson (1999), “Has the U.S. Economy Become More Stable? A
Bayesian Approach Based on a Markov Switching Model of the Business Cycle,” The
Review of Economics and Statistics, Vol. 81, No. 4, pp. 608-616.
Larraín, F. and A. Velasco (2001), “Exchange Rate Policy for Emerging Markets: The Case
for Floating,” Essays in International Economics No. 224, Princeton University.
Rogoff, K. (1985), “The Optimal Commitment to an Intermediate Monetary Target,”
Quarterly Journal of Economics, Vol. 100, No. 4, pp. 1169-1189.
Svensson, L. (1997), “Inflation Forecast Targeting: Implementing and Monitoring Inflation
Targets,” European Economic Review, Vol. 41, pp. 1111-1146.
Svensson, L. (1999), “Inflation Targeting: Some Extensions,” The Scandinavian Journal of
Economics, Vol. 101, No. 3, pp. 337-361.
Taylor, J. (1980), “Aggregate Dynamics and Staggered Contracts,” Journal of Political
Economy, Vol. 88, No. 1, pp. 1-23.
Walsh, C. (2003), Monetary Policy and Theory, 2nd ed., Cambridge, Mass.: MIT Press.
Woodford, M. (2003), Interest and Prices, Princeton, N.J.: Princeton University Press.
18
* I am grateful to Luis Felipe Céspedes, Eduardo Engel, Jordi Galí, and Andrea Tokman for their valuable
discussions and comments, and to Christopher Neilson for superb research assistance.
1 In fact, Title III of the Organic Constitutional Law explicitly refers to the central bank’s authority to regulate
the financial system and the capital market, as well as to its powers to safeguard the financial system’s
stability.
2 Schemes of this type are known as “flexible inflation targets,” as opposed to “strict inflation targets,”
because inflation is returned gradually to the target range when deviations from the conditional forecast occur.
See Svensson (1999).
3 This argument is valid for supply shocks, which are the kind I analyze here. A more general model would
allow for demand shocks, which require a different policy response. For the sake of simplicity, this paper
omits demand shocks, as they do not change the conclusions.
4 A more general presentation of the model with additional discussion can be found in De Gregorio (2007), on
which this part of the paper is based.
5 The exact phrasing of the target may differ across countries. For example, in some countries the target is
stated as a range, whereas in others it is the center of the target plus or minus a deviation.
6 In what follows, I assume that inflation has a known symmetric distribution—specifically, a normal
distribution—which is fully defined by its expected value and its variance.
7 I assume that
ρ
is known, and in the next section it is treated as endogenous. However, if
ρ
is uncertain,
the value estimated by the central bank will affect the variability of inflation, and this effect will depend on
whether
ρ
is under- or overestimated (Amano, 2007).
8 It suffices to take the variance of both sides of equation (1), where the unconditional variances of inflation
and past inflation are the same and equal to 2
π
σ
.
9 One might argue that estimating equation (1) is easy, and that from there the target may be determined with
accuracy, but the relationship between inflation and monetary policy should be also known. The assumption
of an AR(1) process is made for expository purposes, but in reality the univariate process could be more
complicated. Furthermore, defining the target based on the estimation of reduced-form equations is a prime
example of the Lucas critique.
10 This is a simplification of a more general loss function that could be more formally derived following
Woodford (2003, chap. 6). Where indexation is present, the utility function will be somewhat different, but
the main qualitative results presented in this paper should not change.
11 Strictly speaking, from a welfare point of view the relevant objective is to minimize the present value of
losses rather than the value in each period. The solution to that problem is significantly more complex; the
details are presented in De Gregorio (2007). The assumption of a static loss function implicitly assumes that
the central bank has no ability to commit to future policies, and so it optimizes period by period.
12 For more details, see Walsh (2003), chapter 5.3. The existence of indexation is what complicates the
solution of the problem when an intertemporal loss function is assumed. In the event that a = 0, the static and
the intertemporal solutions are the same.
13 See Larraín and Velasco (2001) and Edwards and Levy Yeyati (2005).
14 See, for example, Ball (1999).
15 Chile has several times confronted abrupt changes in the exchange rate with inflationary and monetary
policy implications, which were explicitly mentioned in the statements after the relevant monetary policy
meetings (December 2005 and February, March, and April 2008).
16 Since Chile started floating the exchange rate in 1999, there have been three intervention episodes. Those of
2001 and 2002 occurred in the presence of major turmoil in the region and strong depreciation pressures (see
De Gregorio and Tokman, 2007). The latest, which is ongoing at the time of this writing, began in the
presence of a rapid appreciation in the context of high uncertainty in world financial markets due to the U.S.
subprime mortgage crisis, and with the purpose of increasing international reserves in the face of a reduction
in international liquidity during the last few years.
17 The Great Moderation was first discussed by Kim and Nelson (1999) and Blanchard and Simon (2001). For
a discussion of the various hypotheses offered to explain the Great Moderation and, in particular, the role of
better monetary policy, see De Gregorio (2008).
