Inflation targeting and relative price variability: what difference does inflation targeting make?
Choi, Chi-Young ; Kim, Young Se ; O'Sullivan, Roisin 等
1. Introduction
Variability in relative prices is known to be a major channel
through which inflation can induce welfare costs by impeding an
efficient allocation of resources in the economy. Consequently,
substantial effort has been devoted in the literature to examining the
link between relative price variability (RPV) and aggregate inflation.
Although much of the existing theoretical and empirical literature
points to a positive monotonic relationship, newer contributions suggest
that the relationship between inflation and RPV is more complicated,
particularly in terms of its sensitivity to the inflation regime. (1)
The primary purpose of this study is to investigate whether the
connection between inflation and RPV is influenced in an important way
by the monetary policy framework chosen by a central bank. Specifically,
this article focuses on exploring whether the adoption of an inflation
targeting (IT) framework exerts any significant impact on RPV as
measured by the standard deviations of sectoral inflation rates relative
to the aggregate rate. Since it was first implemented in New Zealand
more than two decades ago, the popularity of IT has spread, with some
twenty-five countries worldwide implementing the framework to date
(Freedman and Laxton 2009). The literature is now replete with studies
pointing to reductions in both the level and the volatility of inflation
in countries that have adopted IT (e.g., Mishkin and Schmidt-Hebbel
2007). (2) While studies of the impact of IT on aggregate inflation
performance are plentiful, little attention has been paid to the impact
of IT on RPV.
The question of whether and indeed how IT affects RPV is an
important one for several reasons. First, exploring the potential
connection between IT and RPV is a worthwhile exercise given the
popularity of IT as a monetary framework and the centrality of RPV to
the current generation of macromodels. The importance of RPV is
recognized in standard New Keynesian Dynamic Stochastic General
Equilibrium (DSGE) models, where the variance of relative prices is
viewed as a useful summary statistic. As noted by Amano, Ambler, and
Rebel (2007), for example, in DSGE models, the optimal rate of target
inflation and the optimal variability of inflation relative to output
depend on the quantitative effects of price dispersion on macroeconomic
equilibrium. Second, answering the question helps us identify the
driving force behind the change in RPV, distinguishing between IT
adoption itself and its subsequent impact on inflation. If the
relationship is monotonically positive, as is often believed in the
literature, one should expect that IT adoption would bring about a
decline in RPV in the same way as it has led to a decline in inflation.
If the relationship is more complex, however, the effect of IT adoption
on reducing RPV may hinge on the change in inflation regimes after IT
adoption. Third, our answer to the question also sheds additional light
on the empirical evidence for the relative effectiveness of IT across
different stages of development. While there is strong evidence that
developing countries benefit more from IT than industrial countries in
combating inflation and its volatility (e.g., Petursson 2004: Lin and Ye
2009), we are aware of no empirical research that has assessed this
issue with respect to RPV.
To address the question, we consider a data set of twenty
industrial and developing countries consisting of 12 targeters and eight
nontargeters during the so-called great moderation period starting in
the mid-1980s. We first find that IT adoption brings about a downward
shift in mean inflation in all countries under study, consistent with
the literature. The more interesting findings relate to the link between
IT adoption and RPV. For countries with initially high inflation rates,
(3) we find that a fall in mean inflation is associated with a similar
decline in RPV after the adoption of IT. In countries with initially low
inflation rates, however, RPV changed little and even increased after IT
adoption. This result, as in the recent findings by Choi (2010) and Choi
and Kim (2010), suggests that the nature of the connection between
inflation and RPV is not monotonic but instead hinges on inflation
regimes, with a linear positive relationship at high trend inflation and
a U-shaped relationship in low or moderate inflation environments. (4) A
similar story is evident for nontargeting countries, with RPV falling
with mean inflation only in the high-inflation countries. Combined, it
seems that what matters most for the structural changes in RPV is not
the adoption of IT per se but the initial inflation environment prior to
adopting IT.
Once the structural shift in inflation is taken into account, our
regression analysis based on a range of econometric techniques,
including semiparametric regression, parametric regression, and rolling
regression, suggests that IT adoption has brought about a tighter
connection between inflation and RPV. Put differently, the same shocks
to inflation lead to a larger dispersion of relative prices under IT,
probably because a stronger commitment to a numerical target for
inflation results in a higher degree of nominal rigidity via the
sluggish response of inflation expectations. Prices might also have
become more rigid after IT adoption because of a fall in average
inflation, consistent with the ample empirical evidence on the inverse
relationship between the degree of price rigidity and the inflation
regime (e.g., Kiley 2000; Nakamura and Steinsson 2008). If firms set
their prices more flexibly in high-inflation settings, while maintaining
stickier prices in low-inflation environments, an increase in the degree
of price rigidity could be associated with a larger dispersion of
relative prices (e.g., Ball, Mankiw, and Reis 2005). Since the tighter
link between inflation and RPV is also observed in most nontargeters,
however, the increased rigidity in price adjustment is posited to be
driven more by the fall in mean inflation than by the change in the
monetary policy framework itself.
We also find that the underlying relationship between inflation and
RPV takes a U-shaped profile in most cases under study, in line with the
recent findings by Choi (2010) and Fielding and Mizen (2008). While the
U-shaped profile is found in low-inflation countries regardless of IT
adoption, it is observed in high-inflation targeters only after IT
adoption. However, no such shift to a U-shaped relationship is observed
in the high-inflation nontargeters under study, suggesting that IT makes
a major difference in high-inflation countries but not in low-inflation
countries. The U-shaped relationship implies the presence of a point at
which RPV is minimized, which we denote as [[pi].sup.*] throughout this
article. According to our empirical results, [[pi].sup.*] is positive
and significantly different from zero in most countries, indicating that
the inflation-RPV relationship is U-shaped around a positive inflation
rate. RPV, therefore, changes not with the inflation rate per se, as
widely believed in the literature, but with the deviation of the
inflation rate from [[pi].sup.*] in either direction. In this context,
[[pi].sup.*] is conceptually related to the central bank's
numerical target for inflation (e.g., Ireland 2007) or the inflation
target level perceived by the public (e.g., Kozicki and Tinsley 2005).
Given that IT purports to reducing uncertainty about future price
developments by strengthening the anchoring of inflation expectations
toward a numerical objective, dispersion of relative prices would
increase with any departure of inflation from the targeted level.
In fact, we find that the estimates of [[pi].sup.*] are well within
the announced target range of inflation in most targeting countries and
that [[??].sup.*] has declined over time as trend inflation did.
[[??].sup.*] is also informative for nontargeters in identifying the
public's perception of inflation. Although nontargeters do not
announce quantitative inflation objectives, market expectations are
formed anyway by what the market believes the unannounced inflation
target to be. This is particularly the case for the nontargeters that
are widely recognized as implicit targeters, where [[??].sup.*] is found
to match well with the implicit target levels of inflation reported by
other researchers. In this vein, it is fair to argue that targeters have
no clear superiority over implicit targeters with the reputation and
commitment for pursuing low inflation when it comes to the anchoring of
the public's inflation expectations to a certain intended target
level.
The story, however, changes somewhat significantly when we examine
the effectiveness of IT in countries with high initial inflation rates.
While targeters with high initial inflation could effectively stabilize
market expectations of inflation around the targeted level of inflation,
there is no clear evidence of stabilizing inflation expectations in
their nontargeting counterparts. Given that one of the major criteria
for the success of IT is the level of control it exerts on the
public's inflation expectations, the potential gains from adopting
IT are more pronounced in countries with high inflation rates. Our
findings therefore lend credence to the view that adoption of IT is more
beneficial to developing countries with typically high inflation rates.
The remainder of this article is structured as follows. Section 2
describes and presents a preliminary analysis of the data. Section 3 is
devoted to a discussion of the econometric analysis of the relationship
between RPV and inflation in targeting and nontargeting countries. The
robustness of our regression results is also examined in that section.
Section 4 discusses the implications of the U-shaped relationship
between RPV and inflation with a focus on [[pi].sup.*] and its
relationship to explicit/implicit target inflation rates. Section 5
concludes this article. The appendices contain detailed descriptions of
the data.
2. Data and Preliminary Analysis
The Data
Our data set comprises monthly (quarterly for Australia) indices of
national consumer prices and their subaggregates for 12
targeters--Australia (AUS), Brazil (BRA), Canada (CAN), Hungary (HUN),
Israel (ISR), Korea (KOR), Mexico (MEX), Norway (NOR), the Philippines
(PHL), Sweden (SWE), the United Kingdom (UK), and South Africa
(ZAF)-along with eight nontargeters Argentina (ARG), Switzerland (CHE),
Germany (GER), Hong Kong (HK), Italy (ITA), Japan (JPN), Turkey (TUR),
and the United States (US). (5) The number of subaggregate items varies
across countries, from five in TUR to 17 in ZAF. Data limitations for
these subaggregate price indices led us to set the starting year of the
sample period at 1984, which marks the onset of the so-called great
moderation period, when the volatility of aggregate economic variables,
including inflation, declined significantly in most industrial
countries. While the starting point is slightly different for some
countries (GER, HUN, TUR, and UK), the end point of the data range is
2009:M2 (2009: I for AUS) in all countries. The sources of the
underlying data are listed in Table A. 1 in the Appendix, to which
further details on the data have been relegated.
Table 1 presents the twenty countries that are categorized based on
their initial inflation regime and their adoption of IT. (6) Although it
is customary to sort countries into groups of industrial versus
developing nations, it is more appropriate here to classify them by
their initial inflation regime in view of its potential importance in
the inflation-RPV nexus (e.g., Bick and Nautz 2008; Choi 2010).
Throughout the article, high-inflation countries are defined as those
with average annual inflation rates greater than 10% in the pre-IT
period, which encompasses BRA, HUN, ISR, PHL, MEX, and ZAF for targeters
and ARG and TUR for nontargeters, as listed in Table 1. Our sample
therefore comprises 12 low-inflation economies and eight high-inflation
countries. Inflation is measured in a standard way by calculating
annualized percentage changes in the consumer price index. Unless noted
otherwise, we concentrate on the deseasonalized month-to-month inflation
rates, where the price indices are seasonally adjusted using the Census
X12-method. RPV is then constructed by calculating the standard
deviation (SD) of the disaggregate inflation rates, (7)
[RPV.sub.t] = [square root of ([N.summation of (i=1)]
[[omega].sub.i][([[pi].sub.it] - [[pi].sub.t]).sup.2])]
where [[pi].sub.it] = ln[P.sub.it] - ln[P.sub.i,t-1], [[??].sub.t]
= [[summation].sup.N.sub.i=1] [[omega].sub.i] [[pi].sub.it],
[[omega].sub.i] denotes the fixed expenditure weight of the ith product
that sums to unity, and [P.sub.it] represents the price index of ith
good at time t.
Preliminary Data Analysis
Table 2 presents summary statistics on average inflation and RPV
for each country for two subsample periods, where the full sample is
split by a certain break point. For targeters, the onset of their IT
regime is used as the break point, (8) whereas the break points for
nontargeters are determined by Bai and Perron's (1998) multivariate
structural break tests for their inflation series, as shown in Table 3.
A couple of observations can be made from Table 2. First, there
exists a notable decline in average inflation after the break point in
all the countries considered, regardless of IT adoption. This
observation accords well with the findings by some earlier studies
(e.g., Cecchetti and Debelle 2004; Levin and Piger 2004). Not
surprisingly, the fall in average inflation is more significant in the
high-inflation countries, from double- or triple-digit annual inflation
to single digit annual inflation. By contrast, no such universal decline
is observed in RPV, with a marked decline seen only in high-inflation
countries. It is a country's initial inflation regime rather than
its IT status that appears to be important. In countries with low
initial inflation rates, a shift in mean inflation is not associated
with any comparable reduction in the cross-sectional variation of
relative prices. Average RPV has actually increased after the break
point in some low-inflation countries, including targeters CAN, NOR, and
the UK and nontargeters CHE, GER, HK, and US. This finding casts some
doubt about the validity of the well-established positive relationship
between inflation and RPV.
An essentially similar picture is painted in Figure 1, which
portrays the empirical densities of inflation and RPV before (solid
line) and after (dashed line) the break point. As can be seen from the
plots, there is a remarkable difference between inflation and RPV in
their empirical densities. While the distribution of inflation clearly
shifts leftward in most countries, reflecting the decline of mean
inflation, the distribution of RPV barely shifts after the break, except
for the high-inflation countries. The structural connection between
inflation and RPV captured by the comovement of the empirical densities
can be found only in the high-inflation countries, regardless of IT
adoption.
To shed additional light on this issue, we run the Bai-Perron
structural break test on the RPV series and report the results in Table
3 along with those for inflation. While the outcomes of these tests
point strongly to the presence of structural changes in the inflation
rates of almost all countries, for RPV, evidence of a structural shift
is found mainly in the high-inflation countries. Table 3 also reports
the estimated dates for the structural breaks in inflation and RPV.
Among the eight countries that exhibit structural changes in both
inflation rates and RPV, the timing of the decline in inflation roughly
matches that of RPV only in the high-inflation countries. In some
targeters, such as CAN, HUN, KOR, MEX, and UK, the estimated break
points in inflation rates are close to the official adoption dates of
IT, lending support to the use of the IT adoption date as the break
point. (9) In the other targeters, the timing of the decline in mean
inflation is a bit earlier than the formal announcement dates of IT
adoption. Such a time lead, however, makes intuitive sense if those
countries stabilized inflation prior to making an official announcement
of IT adoption. Overall, the results from the BaiPerron test generally
corroborate those from Table 2 and Figure 1.
3. Econometric Analysis
Our discussion in the previous section suggests that a mean shift
in inflation is accompanied by a similar structural change in RPV in the
high-inflation countries but not in the low-inflation countries. This
seemingly loose structural connection between inflation and RPV in the
low-inflation countries, however, does not necessarily imply a collapse
of the link between inflation and RPV, especially when the two variables
of interest are suspected to undergo some different structural changes.
One might then reasonably ask to what extent (if at all) the adoption of
IT has impacted RPV once the structural change in the inflation rate is
properly taken into account. To investigate this, the current section
utilizes various econometric techniques to carry out a series of
regression analyses. We first implement a semiparametric regression
technique to identify the underlying functional form of the relationship
between inflation and RPV without imposing any prior assumptions. Based
on the information obtained regarding the functional form, we then apply
a parametric regression technique to two subsamples split by the
aforementioned break points. As a sensitivity analysis, we also conduct
a rolling regression analysis to check the robustness of our regression
results to the choice of break points.
[FIGURE 1 OMITTED]
Underlying Functional Form and Semiparametric Regression Analysis
In the literature, the empirical evidence on the positive link
between inflation and RPV is built largely on regression analysis,
typically with inflation as the causal factor. A common feature of this
existing literature is that the studies focus on linear relationships,
although the linearity restriction is often called into question (e.g.,
Parks 1978; Hartman 1991). In the absence of any concrete guidance from
economic theory, a useful strategy to identify the underlying functional
form is to utilize a semiparametric approach that involves combining the
attractive features of both parametric and nonparametric models,m
Following Fielding and Mizen (2008) and Choi (2010) on which this
section largely draws, we consider a partially linear regression model
as follows:
[RPV.sub.t] = [X'.sub.i][beta] + g([[pi].sub.t]) +
[[epsiolon].sub.t], (1)
where [X.sub.t] is a (p + q) x 1 vector of the regressors that
includes the lagged terms of RPV and inflation and [X'.sub.t] =
{[RPV.sub.t-1], ..., [RPV.sub.t-p], [[pi].sub.t-1], ...,
[[pi].sub.t-q]}. g(.) is an unknown smooth differential function that
captures a contemporaneous effect of inflation on RPV and determines the
underlying functional form of the relationship between inflation and
RPV. The g(.) function in Equation 1 is estimated semiparametrically, as
illustrated by Choi (2010), with particular emphasis on the estimation
of g'(.), the first derivatives of g'(.).
Figure 2 plots the semiparametric estimates of the g'(.)
function (solid line) along with the dotted horizontal line that
captures g'(.) = 0. Of interest is the point where the estimated
g'(.) function crosses the dotted horizontal line, which
corresponds to the RPV-minimizing inflation [pi] rate, denoted as
throughout this article. If the inflation rate is below [[pi].sup.*],
then g'(.) < 0 and g(.) is downward sloping, while g'(.)
> 0 and g(.) is upward sloping if the inflation rate is above
[[pi].sup.*]. In most cases considered, the fitted g'(-) function
is approximately linear and upward sloping, and the transition of
g'(.) from negative to positive values indicates that g(-) has a
quadratic form. This is particularly the case for the countries with low
initial inflation regardless of IT adoption. In those countries, the
point where g(-) intersects the dotted horizontal line, or [??]*, is
lower in the second subsample, implying that the U-shaped relationship
shifts leftward as mean inflation falls.
Albeit overwhelming, the evidence of a U-shaped relationship is not
ubiquitous. In the countries with high initial inflation, the fitted
g'(.) function does not cross the dotted horizontal line but
remains consistently above or below it, implying that the g(.) function
is not quadratic but more likely monotonic. This is the case for the
high-inflation targeters (BRA, HUN, ISR, and MEX) before their adoption
of IT and for the high-inflation nontargeters (ARG and TUR) in both
subsample periods. Notice that the underlying functional form between
these two groups of high-inflation countries is quite different in the
second subsample. While it switches from monotonic to U-shaped in the
targeting high-inflation countries, no such a transition is observed in
the nontargeting high-inflation countries. As is discussed in more
detail in section 4, this result may reflect the difference that IT
adoption makes for the countries with high initial inflation.
The U-Shaped Relationship and Parametric Regression Analysis
Our semiparametric analysis suggests that a well-specified
parametric model of the inflation-RPV nexus should incorporate two
features: (i) a structural change in the underlying model and (ii) a
quadratic U-shaped profile. To accommodate the first feature, the full
sample is split into two subsamples based on the aforementioned break
points. To capture the second feature, we employ the following
parametric model:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
where the lag lengths (p, q) are chosen by the BIC rule. (11) This
parametric specification can be seen as general because it nests both
linear and quadratic models. If [[beta].sub.2] in Equation 2 approaches
zero, the functional form collapses to linear, and hence the overall
relationship between RPV and inflation is determined solely by
[[beta].sub.1]. If [[beta].sub.2] is positive, the relationship is U-
shaped, and the minimum point of U-shape occurs at [[pi].sup.*], where
RPV takes on its lowest value. As shown by Choi (2010), the minimum
point can be estimated by [??]* = -[[??].sub.1]/2[??]2.
[FIGURE 2 OMITTED]
An important question regarding the U-shaped relationship is
whether it is around zero ([[pi].sup.*] = 0) or around a nonzero
inflation rate ([[pi].sup.*] [not equal to] 0). If the association is
U-shaped around zero inflation, RPV would monotonically increase with
inflation (or deflation), and hence higher inflation causes a larger
dispersion of relative prices as documented by a large number of earlier
studies. If, instead, the relationship is U-shaped around a nonzero
inflation rate, RPV rises not with the inflation rate but with the
deviation of inflation from [[pi].sup.*]. The farther away a shock
drives inflation from [[pi].sup.*], the more cross-sectionally dispersed
relative prices become, x* is also useful in tracking the stability of
the U-shaped relationship between inflation and RPV by looking at the
time-varying behavior of [[pi].sup.*].
The parametric regression results reported in Table 4 warrant
several comments. First, in most cases under study, the relationship is
U-shaped around a positive inflation rate that is significantly
different from zero. (12) As can be seen from the third and fourth
columns of the upper panel of Table 4, the impact of inflation
volatility on RPV is nonnegative ([[??].sub.2] [greater than or equal
to] 0) in all cases, while that of the inflation level is negative
([[??].sub.1] < 0) in the vast majority of cases, indicative of a
U-shaped relationship between inflation and RPV. As presented in the
lowerleft panel of Table 4, however, [[??].sub.1] is positive in the
prebreak period in both high-inflation nontargeting countries and two of
the high-inflation targeters, BRA and MEX. In these countries, the
corresponding [[??].sub.2] is very close to zero, and thus [pi]x* is not
properly defined, suggesting that the underlying relationship is more
likely to be positive linear. Apart from them, [??]* is positive and
significantly different from zero in all countries, as the lower bound
of the 95% confidence interval for [??]* is consistently above zero.
This result implies that the RPV-related welfare cost of inflation is
minimized when the inflation rate is above zero rather than zero.
Second, the underlying relationship between inflation and RPV is
not stable over time but instead varies across inflation regimes in a
systematic manner. This time variation is particularly noticeable in the
high-inflation targeting countries, where the underlying relationship
appears to switch from a positive monotonic relationship ([[??].sub.1]
> 0 and [[??].sub.2] [equivalent] 0) in the prebreak period to a
U-shaped profile ([[??].sub.1] < 0 and [[??].sub.2] > 0) in the
postbreak period. In the targeters with low initial inflation where the
evidence of a U-shaped profile is found in both subsample periods, we
note a decrease in the value of [??]*, reflecting a leftward shift of
the Ushaped relationship. A broadly similar story is told for
nontargeters that have maintained low and stable inflation during the
great moderation period.
Third, the link between inflation and RPV has become stronger after
IT adoption in most targeters when judged by a larger value of
[[??].sub.2] in the post-IT period. In CAN, for example, the value of
[[??].sub.2] has increased almost sixfold, from 0.26 to 1.55. This
increase in [[??].sub.2] signifies a steeper curvature of the U-shape
and hence a larger response of RPV to the same inflation shock that
leads to a deviation of inflation from [??]*. (13) This result is
posited to be driven by an increase in the degree of price rigidity
under IT. (14) When inflation expectations are anchored to an announced
inflation target, economic agents react more sluggishly to a temporary
shock that drives inflation away from the targeted inflation rate,
causing a larger dispersion of relative prices. This stronger response
of RPV to inflationary shocks in the second subperiod can be seen also
in nontargeters, however, especially in those that maintained an
implicit but credible commitment to low inflation.
Robustness Check Using Rolling Regression Analysis
To ensure that our results in the previous section are not driven
by the choice of specific sample periods, we appeal to the rolling
regression approach that does not impose any prior restrictions on the
timing of break points. This is an attractive feature when the
full-sample estimates are vulnerable to time variation in the
conditional mean of the inflation process.
Figure 3 presents the estimates of [[beta].sub.1] and
[[beta].sub.2] in Equation 2 from a sequence of rolling samples. Each
point in the plot exhibits [[??].sub.1] (thin line) and [[??].sub.2]
(heavy line) at t that are obtained using data from t-120 (t-40 for AUS)
to t with a window of 10 years. (15) The numbers on the horizontal axis
therefore represent the beginning year of each ten-year window. For
instance, 1984 captures the subsample period of 1984-1993 and so on. As
anticipated, the rolling estimates of [[beta].sub.2] are consistently
positive, while those of [[beta].sub.1] are negative in most cases,
indicating that the relationship between inflation and RPV is U-shaped
around a positive [[pi].sup.*]. A notable exception, however, can be
found in some high-inflation countries (BRA, MEX, ARG, and TUR) where
[[??].sub.2] is close to zero in the early part of the sample period
with [[??].sub.1] being positive. This suggests that the relationship is
positive linear during the corresponding sample period. In MEX, for
example, a high-inflation targeter, [[??].sub.2] is close to zero until
around 1994, while [[??].sub.1] is positive, implying that the
relationship between inflation and RPV in MEX is positive linear until
the subsample period of 1994-2003. After 1994, however, [[??].sub.2] in
MEX switches to positive, and [[??].sub.1] swings to negative,
indicative of a U-shaped profile. (16)
The rolling estimates for [[??].sub.1] and [[??].sub.2] also
display a significant variation over time. Our visual inspection
suggests that the timing of structural changes in the two coefficient
estimates roughly coincides with the changes in the monetary policy
regime reported in Table 3. For instance, the timing of the structural
change in [[??].sub.2] is very close to their official adoption dates of
IT in some targeters, such as AUS, CAN, KOR, and UK.
4. [[pi].sup.*] and Target Inflation
Our discussion so far suggests that the relationship between
inflation and RPV is U-shaped around a nonzero inflation rate in most
countries, especially after IT adoption. A central implication of the
U-shaped relationship is that RPV changes not with inflation per se but
with the deviation of inflation from [[pi].sup.*]. Questions then
naturally arise regarding how to interpret the nonzero [[pi].sup.*] and
how [[pi].sup.*] is related to the inflation target. This line of
inquiry is pursued in the current section.
Interpretation of [[pi].sup.*] and the Inflation Target
In empirical macromodels, inflation rates are often partitioned
into two parts: (i) its perceived equilibrium attractor or the perceived
central-bank target for inflation and (ii) deviations from the
equilibrium (e.g., Kozicki and Tinsley 2008). Since IT purports to
reducing uncertainty about future price developments by anchoring
inflation expectations toward a numerical objective, the dispersion of
relative prices would be minimized if the actual inflation rate were
equal to the inflation target. Because RPV would rise with any deviation
of actual inflation from the targeted inflation rate in either
direction, [[pi].sup.*] can be viewed as conceptually related to the
target level of inflation perceived by the public (e.g., Kozicki and
Tinsley 2005) or the central bank's inflation target (e.g., Ireland
2007).
In light of the fact that trend inflation is usually pinned down by
a central bank's target in general equilibrium models, [[pi].sup.*]
is also related to trend inflation (e.g., Sbordone 2007; Cogley and
Sbordone 2008). (17) In this context, it would be instructive to examine
whether the estimate of [[pi].sup.*] is close to the announced target
level of inflation.
Table 4 presents the estimates of [[pi].sup.*] before and after the
break point, along with the explicit/implicit numerical targets for
inflation. The results in Table 4 illustrate a couple of interesting
points with regard to [??]*. First, [??]* appears to have fallen into
the target range of inflation after IT adoption in eight out of 12
targeters, probably because a strong commitment to an announced target
helps the public form expectations for the policy outcome (e.g.,
Woodford 2004). In the remaining four targeters (AUS, HUN, MEX, and
PHL), [??]* stays outside the target range but not far from the upper
end of the target. Aside from AUS, this may be because more time is
needed for these countries to build credibility around the relatively
recently adopted new monetary policy framework. (18)
Second, [??]* is useful for nontargeters in identifying the market
perception of the unannounced inflation target, particularly in the
nontargeters that are generally regarded as implicit targeters. Although
nontargeters do not announce any quantitative inflation objectives,
market expectations are still anchored by what the market believes the
inflation target is. For example, some nontargeters, such as CHE, GER,
JPN, and US, are widely recognized as de facto targeters because their
commitment to low inflation or price stability has been deeply embedded
in their monetary policy framework (e.g., Truman 2003). (19) As shown in
Table 4, the estimates of [[pi].sup.*] for these implicit targeters are
well within the implied targets ranges reported by other researchers.
[FIGURE 3 OMITTED]
Time Variation of [[pi].sup.*] and Trend Inflation
Recently, growing evidence has emerged on the shifts in trend
inflation over time (e.g., Amano, Ambler, and Rebel 2007: Ireland 2007:
Stock and Watson 2007; Cogley and Sbordone 2008). Based on a macromodel
with a time-varying inflation trend, for instance, Cogley and Sbordone
(2008) maintain that trend inflation in the United States has been
nonzero and varied over time. A similar conclusion is reached by Stock
and Watson (2007) based on an unobserved component trend-cycle model
with stochastic volatility. Using the Kalman filter technique, Leigh
(2008) also documents that the Fed's implicit target is not
constant but instead has varied significantly over time, from near 3% in
the early 1980s, to 34% in the late 1980s and early 1990s, and to 1-2%
after the 1990-1991 recession, before rising to 2-3% during 2001-2004.
Similarly, Ireland (2007) reports that the Fed's inflation target
has increased from 1.25% in 1959 to more than 8% in the late 1970s,
followed by a gradual reduction to below 2.5% in 2004. This time
variation of trend inflation is often explained by the central
bank's updating of its policy rule when it learns more about the
structure of the economy. (20) For example, Levin and Piger (2004)
assert that movements in the mean of inflation reflect shifts in private
agent perceptions of the policy target for inflation. Since the central
bank's inflation target bears particular relevance for the
inflation expectations of the public, it would be of interest to examine
how closely the estimates of re* match the expected inflation of
economic agents.
[FIGURE 4 OMITTED]
Given the availability of a direct measure of inflation
expectations for the United States, we use that country as a case study.
Figure 4 reports the result of this exercise by plotting the evolution
of [??]* in the United States for a 10-year rolling window sample,
together with the long-horizon inflation expectations from the Survey of
Professional Forecasters. (21) As can be seen from the figure,
[[pi].sup.*] fits quite well the survey measures of long-horizon
inflation expectations in the United States, confirming our prior
intuition that [[pi].sup.*] is closely related to the inflation
expectations of the public. The time-varying patterns of both [??] and
the inflationary expectations of the public are believed to be commonly
driven by the changes in the central-bank target for inflation.
Unfortunately, data for long-term inflation expectations are not
available in many other countries under study. This led us to utilize
the period average inflation as a rough substitute for the market
expectation of inflation. As depicted in Figure 4, the period average
inflation rates (dotted line) are closely linked to the expected
inflation rates in the United States.
[FIGURE 5 OMITTED]
Figure 5 plots the evolution of [??]* (heavy solid line) over time,
together with the sample average inflation rates (dotted line) and the
targeted level/band of inflation (thin solid line) for all the countries
under study. Since [??]* is not properly defined in the high-inflation
countries before IT adoption with [??]2 close to zero, we concentrate on
the post-IT period for those high-inflation countries. A couple of
important features emerge from Figure 5.
First, [??]* has steadily declined over time in most countries, in
a similar pattern to that exhibited by period average inflation. In some
countries (especially US, JPN, and HK), [??]* moves in sync with the
period average inflation rates, consistent with the recent empirical
evidence on time-varying trend inflation. In the two nontargeting
high-inflation countries (ARG and TUR), however, [??]* diverges from the
period average inflation rate, most likely because of a weak anchoring
of inflation expectations in those countries.
Second and more important, [??]* is already within the announced
target range of inflation in all the low-inflation targeters, with the
exception of AUS and NOR, where [??]* recently moved out of the target
range. This may mirror the indirect evidence of IT's effectiveness
in reducing inflation expectations toward the announced target after IT
adoption. Some targeters managed to contain inflation expectations
within a prescribed narrow band in a relatively short time, although it
appears to have taken a bit longer in others. In CAN, for instance,
[??]* fell rapidly below the targeted inflation level right after the
adoption of IT, indicative of a quick adjustment of the public's
inflation expectations after the adoption of the new monetary policy
framework. A broadly similar pattern is observed in the nontargeting
countries that are widely known as de facto targeters. In CHE, GER, ITA,
JPN, and US, for example, [??]* is well within the implicit target range
for inflation estimated by other researchers. Our result therefore
supports the finding by Ball and Sheridan (2005) that targeters do not
necessarily entertain a clear advantage in anchoring inflation
expectations compared to the nontargeters that have maintained low and
stable inflation without explicitly adopting IT.
The story, however, changes somewhat significantly when we look at
the countries with high initial inflation rates. Although [??]* in some
high-inflation targeters (HUN, MEX, and PHL) is yet to fall within the
target range, it seems to be moving toward it after IT adoption. This
may be because they are still building the credibility of the new
monetary policy framework, having adopted IT relatively recently. By
contrast, no such pattern of moving toward a certain level of inflation
can be seen in their nontargeting counterparts. Interestingly, a
significant difference exists between the two nontargeting
high-inflation countries. While [??]* consistently deviates from the
sample average inflation rate in ARG, the gap between [??]* and the
sample average inflation has diminished steadily over time in TUR. This
difference may rest on the fact that TUR has adopted IT in 2006 but is
considered here as a nontargeter. These results, therefore, support our
prior intuition that IT serves to reshape inflation expectations,
particularly in initially high-inflation countries without well-defined
numerical inflation objectives.
Overall, our results highlight the informativeness of [??]*
regarding inflation expectations formed by economic agents in both
targeters and nontargeters. [??]* is instrumental in identifying the
public's expectations of inflation for nontargeters and in
assessing the effectiveness of IT in establishing a credible nominal
anchor for targeters. Given that one of the major criteria for the
success of IT is the level of control it exerts on the public's
inflation expectations, our results offer qualitative support for the
view that IT is more beneficial to countries with initially high
inflation rates.
5. Concluding Remarks
Inflation targeting has become a popular monetary policy framework
in the past two decades, largely for its success in reducing both
inflation and inflation volatility. This article investigates whether
and how the adoption of IT exerts a significant influence on the
variability of relative prices as it did on inflation. Grappling with
this question is crucial not only in evaluating the effectiveness of IT
beyond its impact on aggregate inflation but also in understanding the
transmission mechanism of inflation as a key element of the standard New
Keynesian DSGE models.
By examining 12 targeting countries and eight nontargeting
countries, we first find that what matters for RPV is not IT adoption
per se but rather the inflation regime prior to the adoption of IT. RPV
has fallen with mean inflation rates only in the countries with high
initial inflation rates regardless of whether they targeted inflation.
Once the structural change in inflation is accounted for, however, our
regression analysis suggests that the connection between inflation and
RPV has become tighter after IT adoption, with the same shocks to
inflation leading to a larger dispersion of relative prices. This
tighter relationship in the later subsample is not unique to inflation
targeters, however.
We also find that the relationship between inflation and RPV takes
a U-shaped profile around a nonzero inflation rate in most of the
countries under study. An important implication of this is that RPV
changes not with the inflation rate, as widely accepted in the
literature, but with the deviation of inflation from [??]* at which RPV
is minimized. Insofar as economic agents anchor their expectations of
inflation around the target level, relative prices would become more
dispersed as the public responds more sluggishly to the shocks that
drive actual inflation away from the targeted level. In this vein, [??]*
can be viewed as conceptually related to the inflation target level
perceived by the public and to trend inflation, which is often pinned
down by a central bank's target. Our estimates of [??]* are also
informative about the market perception of the unannounced inflation
target in nontargeting countries. For low-inflation nontargeters, we
find that the estimates of [??]* match quite well with the implicit
inflation targets reported by other researchers.
When it comes to the anchoring of inflation expectations as
measured by [??]*, therefore, targeters seem to have no clear advantage
over the implicit targeters with a comparable commitment to low
inflation. The effectiveness of IT, however, stands out when countries
with high initial inflation rates are considered. While IT serves to
anchor inflation expectations in the targeters with high initial
inflation rates, no clear signs of stabilization of inflation
expectations are observed in their nontargeting counterparts. Our
findings therefore support the argument that adoption of IT is
potentially more beneficial to developing countries with typically high
inflation rates.
Appendix: Data Description
Table A.1. Data Description
Country Data Span Data Source
AUS 1984:I-2009:I Australian Bureau
{993:II} of Statistics (ABS)
BRA 1984:M1-2009:M2 Brazilian Institute
{1999:M6} of Geography and
Statistics (IBGE)
CAN 1984:M1-2009:M2 Statistics Canada
{1991:M2}
HUN 1992:M1-2009:M2 Hungarian Central
{2001: M6} Statistical Office
(KSH)
ISR 1984:M1-2009:M2 Central Bureau of
{1997:M6} Statistics
KOR 1984:M1-2009:M2 Korea National
{1998:M4} Statistical Office
(NSO)
MEX 1984:M1-2009:M2 Bank of Mexico
{2001:M1}
NOR 1984: M1 2009: M2 Statistics Norway
{2001:M3}
PHL 1984:M1-2009:M2 Philippines National
{2002:M1} Statistical Office
SWE 1984:M1-2009:M2 Statistics Sweden
{1993:M1}
UK 1988:M1-2009:M2 National Statistics
{1992:M10}
ZAF 1984:M1-2009:M2 Statistics South
{2000:M2} Africa
ARG 1984: M1-2009: M2 National institute of
{1991:M4} Statistic and
Censuses
(INDEC)
CHE 1984:M1-2009:M2 Federal Statistical
{1993:M6} Office
GER 1991:M1-2009:M2 Federal Statistical
{1992:M10} Office Germany
HK 1984:M1-2009:M2 Census and Statistics
{1997:M3} Department
ITA 1984:M1-2009:M2 National institute
{1995:M11} of Statistics
JPN 1984:M1-2009:M2 Statistics Bureau
{1993:M8}
TUR 1986:M1-2009:M2 Central Bank of the
{2002:M3} Republic of
Turkey (CBRT)
US 1984:M1-2009:M2 Bureau of Labor
{1999:M9} Statistics (BLS)
Country Subaggregate Items
AUS [8] Food (20.7): alcohol and tobacco (9.2); clothing and
footwear (5.2): housing (26.2); household contents and
services (12.9); transportation (17.6); communication
(4.4); education (3.6)
BRA [7] Food products and beverages (23.6); housing (16.7);
household articles (6.4): apparel (8.4); transportation and
communication (20.1); health and personal care (10.3);
personal expenses (14.6)
CAN [8] Food (17.0): shelter (26.6); household operations and
furnishings (I1.1); clothing and footwear (5.4);
transportation (19.9); health and personal care (4.7):
recreation, education, and reading (12.2); alcoholic
beverages and tobacco products (3.1)
HUN [7] Food (23.7): alcoholic beverages and tobacco (9.6);
clothing and footwear (5.6); consumer durable goods
(7.3); electric _as and other fuels (8.5); other goods,
including motor fuels and lubricants (16.8); services (28.5)
ISR [10] Food, excluding vegetables and fruit (14.8); vegetables
and fruit (3.6); housing (20.7); dwellings maintenance
(10.6); furniture and household equipment (3.8); clothing
and footwear (3.2); health (5.2); education, culture, and
entertainment (12.5); transport and communication
(21.1); miscellaneous (4.5)
KOR [12] Food and nonalcoholic beverages (14.0); alcoholic
beverages and cigarettes (1.5); clothing and footwear
(5.8); housing, water, and fuels (17.0); furnishings and
household equipment (4.2); health (5.2); transportation
(10.9); communication (6.0); culture and recreation (5.6);
education (11.1); eating out and accommodation (13.3);
miscellaneous (5.4)
MEX [8] Food, beverages, and tobacco (22.7); clothes, footwear,
and accessories (5.6); housing (26.4); furniture and
domestic accessories (4.9); health and personal care (8.6);
transportation (13.4); education and entertainment
(11.5); miscellaneous (6.9)
NOR [121 Food and nonalcoholic beverages (11.2); alcoholic
beverages and tobacco (2.7); clothing and footwear (5.9);
housing, water, electricity, gas, and other fuels (29.5);
furnishings, household equipment, and routine
maintenance (6.3); health (2.7); transport (17.9);
communications (2.1); recreation and culture (12.0);
education (0.3); restaurants and hotels (3.4);
miscellaneous goods and services (6.0)
PHL [6] Food, beverages, and tobacco (50.0); clothing (3.0);
housing and repairs (16.8); fuel, light, and water (6.9);
services (15.9); miscellaneous (7.3)
SWE [11] Food and nonalcoholic beverages (13.2); alcoholic
beverages and tobacco (3.7); clothing and footwear (5.4);
housing, water, electricity, gas, and other fuels (26.7);
furnishings and household goods (5.5); health (3.2);
transport (14.6); communication (3.5); recreation and
culture (11.9); restaurants and hotels (6.8); miscellaneous
goods and services (5.4)
UK [12] Food and nonalcoholic beverages (11.8); clothing and
footwear (5.7); alcoholic beverages, tobacco, and
narcotics (4.4); housing, water, and fuels (12.6);
furnishings, household equipment, and routine repair of
house (6.6); health (2.2); transport (15.1); communication
(2.3); recreation and culture (14.5); education (2.1);
hotels, cafes, and restaurants (12.8); miscellaneous goods
and services (9.9)
ZAF [17] Food (21.0); nonalcoholic beverages (1.1); alcoholic
beverages (1.4); cigarettes, cigars, and tobacco (1.1);
clothing and footwear (3.3); housing (22.1); fuel and
power (3.5); furniture and equipment (2.5); household
operation (4.8); medical care and health expenses (7.2);
transport (14.8); communication (3.0); recreation and
entertainment (3.3); reading matter (0.4); education (3.5);
personal care (3.7); other (3.3)
ARG [9] Foods and beverages (31.3); apparel (5.2); housing and
basic services (12.7); household equipment and
maintenance (6.5); medical attention and health care
expenses (10.0); transportation and communication
(17.0); leisure (8.7); education (4.2); miscellaneous goods
and services (4.4)
CHE [12] Food and nonalcoholic beverages (10.9); alcoholic
beverages and tobacco (1.7); clothing and footwear (4.2);
housing and energy (19.7); furnishings, household
equipment, and routine household maintenance (4.6);
health (14.3); transport (11.0); communication (2.8);
recreation and culture (10.9); education (0.9); restaurants
and hotels (9.8); miscellaneous (10.2)
GER [12] Food and nonalcoholic beverages (10.4); alcoholic
beverages and tobacco (3.9); clothing and footwear (4.9);
housing, water, electricity, gas, and other fuels (30.8);
furnishings and household equipment (5.6); health (4.0);
transport (13.2); communication (3.l); recreation and
culture (11.6); education (0.7); restaurants and hotels
(4.4); miscellaneous goods and services (7.4)
HK [9] Food (26.9); housing (29.2); electricity, gas, and water
(3.6); alcoholic drink and tobacco (0.9); clothing and
footwear (5.9); durable goods (5.5); miscellaneous goods
(4.8); transport (9.1); miscellaneous services (16.2)
ITA [8] Food, beverages, and tobacco (21.8); clothing and
footwear (9.7); housing, water, electricity, gas, and other
fuels (11.0); furnishings and household equipment (9.7);
health (9.0); transport and communication (20.2);
education, recreation, and culture (9.8); miscellaneous
(9.0)
JPN [10] Food (24.6); housing (26.3); fuel, light, and water
charge (5.8); furniture and household utensils (2.9);
clothes and footwear (4.9); medical care (4.1);
transportation and communication (10.2); education
(4.7); reading and recreation (11.1); miscellaneous (5.5)
TUR [5] Foodstuffs (36.6); heating and lighting articles (9.8);
clothing and house furniture (9.4); house rent and
maintenance (11.7); miscellaneous (36.4)
US [6] Apparel (4.2); food (17.9); other (3.8); housing (49.4);
medical (7.3); transportation (17.4)
Dates in {} refer to the IT adoption date for targeters and the break
point of the inflation series estimated by the Bai-Perron (1998)
method for nontargeters. The numbers in square brackets represent the
number of subaggregates, and the entries inside the parentheses denote
the weight of each subaggregate.
References
Amano, Robert. Steve Ambler, and Nooman Rebel. 2007. The
macroeconomic effects of nonzero trend inflation. Journal of Money,
Credit and Banking 39:1821-38.
Bai, Jushan, and Pierre Perron. 1998. Estimating and testing linear
models with multiple structural changes. Econometrica 66:47-78.
Ball, Laurence, N. Gregory Mankiw, and Ricardo Reis. 2005. Monetary
policy for inattentive economies. Journal of Monetary Economies
52:703-25.
Ball, Laurence, and Niamh Sheridan. 2005. Does inflation targeting
matter? In The inflation targeting debate, edited by Ben S. Bernanke and
Michael Woodford. Chicago: University of Chicago Press, pp. 249-82.
Becker, Sascha S., and Dieter Nautz. 2009. Inflation and relative
price variability: New evidence for the United States. Southern Econamic
Journal 76:146-64.
Bernanke, Ben. S., Thomas Laubach, Frederic S. Mishkin, and Adam S.
Posen. 1999. Inflation targeting." Lessons from the international
experience. Princeton, N J: Princeton University Press.
Bick, Alexander, and Dieter Nautz. 2008. Inflation thresholds and
relative price variability: Evidence from U.S. cities. International
Journal of Central Banking 4:61-76.
Cecchetti, Stephen, and Guy Debelle. 2004. Has the inflation
process changed? Mimeograph, Bank for International Settlements.
Choi, Chi-Young. 2010. Reconsidering the relationship between
inflation and relative price variability. Journal of Money. Credit and
Banking 42:769-98.
Choi, Chi-Young, and Young S. Kim. 2010. Is there any asymmetry in
the effect of inflation on relative price variability? Economics Letters
108:233-36.
Clarida, Richard H., Jordi Gali, and Mark L. Gertler. 1998.
Monetary policy rules in practice: Some international evidence. European
Economic Review 42:1033-67.
Cogley, Timothy, Giorgio E. Primiceri, and Thomas J. Sargent. 2010.
Inflation-gap persistence in the U.S. American Economic Journal:
Macroeconomics 2:43-69.
Cogley, Timothy, and Argia M. Sbordone. 2008. Trend inflation,
indexation, and inflation persistence in the New Keynesian Phillips
curve. American Economic Review 98:2101-26.
Fielding, David, and Paul Mizen. 2008. Evidence on the functional
relationship between relative price variability and inflation with
implications for monetary policy. Economica 75:683-99.
Freedman, Charles, and Douglas Laxton. 2009. Why inflation
targeting? IMF Working Paper No. 09-86.
Goodfriend, Marvin. 2003. Inflation targeting in the United States.
NBER Working Paper No. 9981.
Grier, Kevin B., and Mark J. Perry. 1996. Inflation, inflation
uncertainty, and relative price dispersion: Evidence from bivariate
GARCH-M models. Journal of Monetary Economics 38:391-405.
Hartman, Richard. 1991. Relative price variability and inflation.
Journal of Money, Credit and Banking 23:185-205.
Ireland, Peter N. 2007. Changes in the Federal Reserve's
inflation target: Causes and consequences. Journal of Money, Credit and
Banking 39:1851-82.
Kiley, Michael T. 2000. Endogenous price stickiness and business
cycle persistence. Journal of Money, Credit and Banking 32:28-53.
Kozicki, Sharon, and Peter A. Tinsley. 2005. Permanent and
transitory policy shocks in an empirical macro model with asymmetric
information. Journal of Economic Dynamics and Control 29:1985-2015.
Kozicki, Sharon, and Peter A. Tinsley. 2008. Term structure
transmission of monetary policy. North American Journal of Economics and
Finance 19:71-92.
Kuzin, Vladimir. 2006. The inflation aversion of the Bundesbank: A
state space approach. Journal of Economic Dynamics and Control
30:1671-86.
Lastrapes, William D. 2006. Inflation and the distribution of
relative prices: The role of productivity and money supply shocks.
Journal of Money, Credit and Banking 38:2159-98.
Leigh, Daniel. 2008. Estimating the Federal Reserve's implicit
inflation target: A state space approach. Journal of Economic Dynamics
and Control 32:2013-30.
Levin, Andrew T., and Jeremy M. Piger. 2004. Is inflation
persistence intrinsic in industrial economies? ECB Working Paper No.
334.
Lin, Shu, and Haichun Ye. 2009. Does inflation targeting really
make a difference in developing countries? Journal of Development
Economics 89:118-23.
Mishkin, Frederic S., and Klaus Schmidt-Hebbel. 2007. Does
inflation targeting make a difference? NBER Working Paper No. 12876.
Nakamura, Emi, and Jon Steinsson. 2008. Five facts about prices: A
reevaluation of menu cost models. Quarterly Journal of Economics
123:1415-64.
Parks, Richard W. 1978. Inflation and relative price variability.
Journal of Political Economy 86:79-95.
Parsley, David C. 1996. Inflation and relative price variability in
the short and long run: New evidence from the United States. Journal of
Money, Credit and Banking 28:323-41.
Petursson, Thorarinn G. 2004. The effects of inflation targeting on
macroeconomic performance. Central Bank of Iceland Working Paper No. 23.
Sbordone, Argia M. 2007. Inflation persistence: Alternative
interpretations and policy implications. Journal of Monetary Economics
54:1311-39.
Stock, James H., and Mark W. Watston. 2007. Why has U.S. inflation
become harder to forecast? Journal of Money, Credit and Banking 39:3-33.
Truman, Edwin M. 2003. Inflation targeting in the worm economy.
Washington D.C.: Institute for International Economics.
Vining, Daniel R., Jr., and Thomas C. Elwertowski. 1976. The
relationship between relative prices and the general price level.
American Economic Review 66:699-708.
Woodford, Michael. 2004. Inflation targeting and optimal monetary
policy. Federal Reserve Bank of St. Louis Review 86:15-41.
(1) Since the work of Vining and Elwertowski (1976), a sizable
literature documents a positive relationship between inflation and the
cross sectional variability of relative prices for many countries and
for various time periods. See Lastrapes (2006) and Becket and Nautz
(2009) for more recent empirical evidence on the positive relationship.
Choi (2010), however, recently reports empirical evidence that the
relationship not only is nonmonotonic but also exhibits significant
variation over inflation regimes.
(2) For an opposing view, see Ball and Sheridan (2005), who find
little evidence that IT improves macroeconomic performance for a group
of OECD countries.
(3) As is formally defined in section 2, we consider high inflation
as an average annual inflation rate higher than 10% before the adoption
of IT.
(4) Using simulation experiments, Choi (2010) shows that a modified
Calvo model, which embeds sectoral heterogeneity in price rigidity, can
explain this feature of the data.
(5) The selection of countries was guided mainly by the
availability of sufficiently long continuous data series for
subaggregate consumer price indices. Turkey adopted IT in January 2006
but is counted as a nontargeter in our study because it was a
nontargeter for most of our sample period.
(6) We follow much of the literature in categorizing targeters
based on de jure rather than de facto targeting.
(7) Since inflation data is used for RPV, the term relative
inflation variability is more appropriate. Throughout this article,
however, we follow the convention in the literature and call this
measure relative price variability (RPV), the term that has been
ingrained in the literature since the gold-standard era. In fact, the
literature reports largely similar results for the inflation-RPV nexus
when price-level data are used for constructing RPV (e.g., Parsley
1996).
(8) The exact timing of IT adoption varies with the definition of
targeting. While some authors (e.g., Bernanke et al. 1999) date the
start of targeting at the point when targets were first announced,
others (e.g., Bali and Sheridan 2005) date on the basis of actual
implementation, which often lags the announcements. Here we stick to the
former approach, but the difference between the two dates is not
consequential in most countries under scrutiny.
(9) Because of the substantive heterogeneity observed in the break
points across countries, panel data analysis is of reduced merit here,
as the subsample panels would become "'unbalanced" or
"'incomplete." This either imposes a serious limitation
on the efficiency gain or makes the estimation infeasible. This problem
is known to be exacerbated in dynamic panel models containing lagged
terms of dependent variables as in our case.
(10) By combining the easy interpretability of the parametric
approach with some of the flexibility of the nonparametric approach, the
semiparametric approach is known to get around the so-called
curse-of-dimensionality problem while allowing for flexibility in
functional form.
(11) Inclusion of the square term of inflation ([pi]2 t) is also
consistent with the findings by some earlier studies (e.g., Parks 1978;
Hartman 1991) that inflation volatility is a significant explanatory
variable of RPV.
(12) This is consistent with the point made by Sbordone (2007) that
trend inflation, no matter how it is defined, has rarely been zero over
past decades, even when a 1% upward bias is allowed for in the measured
inflation rate.
(13) As shown in Choi (2010), the marginal effect of a deviation of
inflation from [[pi].sup.*] can be approximated by
([DELTA][RPV.sub.t]/[DELTA] [[pi].sup.d.sub.t]) [approximately equal to]
2[[beta].sub.2] [[pi].sup.d.sub.t], which depends solely on
[[beta].sub.2], where [[pi].sup.d.sub.t] = ([[pi].sub.t] - [[pi].sup.*])
denotes the inflation deviation. While the table includes only the
contemporaneous effect of inflation on RPV, we find qualitatively
similar results for the cumulative effect regarding the greater impact
in the postbreak period. We also find that adding country characteristic
variables as regressors does not alter much the conclusions reached in
this article. These results are not reported here to preserve space but
are available from the authors on request. The authors are grateful to
an anonymous referee for bringing these issues to our attention.
(14) The potential impact of nominal rigidities on the
inflation-RPV nexus can be analyzed separately from that of sectoral
productivity or demand shocks within a VAR framework as in Lastrapes
(2006). This important issue, however, is left for future research, as
it goes beyond the scope of this article.
(15) Similar results are obtained using rolling windows of eight
and 12 years.
(16) This coincides with the timing of the Mexican crisis that
brought an end to its fixed exchange rate regime and marked the
beginning of Mexico's path to inflation targeting.
(17) The deviation of inflation from [[pi].sup.*] is also similar
in spirit to the inflation gap described by Cogley, Primiceri, and
Sargent (2010) and Sbordone (2007). Cogley, Primiceri, and Sargent
(2010) define the inflation gap as deviations of inflation from a
time-varying inflation trend. Alternatively, the deviation can be viewed
as the gap between actual inflation and expected inflation as in Grier
and Perry (1996).
(18) According to the Reserve Bank of Australia's statements
on monetary policy, AUS experienced a large deviation of inflation from
the target level in the early 2000s mainly because of higher oil prices
and tax changes. A goods and service tax introduced in July 2000 to
replace the existing wholesale sales tax has led to large increases in
price indices between June 2000 and September 2001.
(19) Although the Fed has never officially stated a target range
for inflation, many observers suggest that the Fed has in fact practiced
implicit inflation targeting during the Volcker-Greenspan era (e.g.,
Goodfriend 2003). Clarida, Gall, and Gertler (1998) report that the
estimate of the Federal Reserve's unobserved implicit inflation
target is around 4%. According to Kuzin (2006), the German
Bundesbank's implicit inflation target was more than 4% in 1975 but
declined to near 2% in 1998. In Japan, it is broadly recognized that the
Bank of Japan views 0 2% as the appropriate level of inflation in the
medium to long run (http://www.boj.or.jp/en/type/release/zuiji_new/k060309b.htm).
(20) As pointed out by Ireland (2007), transitory movements in the
measured rate of inflation can be driven by various shocks, but large
and persistent movements in inflation cannot occur without the help of
monetary policy.
(21) The sequence of 10-year-ahead inflation expectations is a
widely reported measure of long-run inflation expectations and was
downloaded from the website of the Federal Reserve Bank of Philadelphia
(http://www.phil.frb.org).
JEL Classification: E31, E52. E58
Chi-Young Choi, Department of Economics, University of Texas at
Arlington, Arlington, TX 76019, USA; E-mail cychoi@ uta.edu:
corresponding author.
Young Se Kim, Department of Economics, Sungkyunkwan University,
Seoul 110-745, Korea.
Roisin O'Sullivan, Department of Economics, Smith College,
Northampton, MA 01063, USA.
The authors are grateful to coeditor Kent Kimbrough and two
anonymous referees for constructive comments that helped to improve this
article. The authors also wish to thank Steve Cecchetti, Kang Liu,
Shin-Ichi Nishiyama, Margie Tieslau, Taka Tsuruga, and the seminar
participants at the Academia Sinica, National Chung Cheng University,
National Sun Yat-Sen University, National Taiwan University, Texas Tech
University, and the University of Texas at Arlington for helpful
comments and Vikas Kakkar for providing CPI data for Hong Kong. Any
remaining errors are the authors'.
Received January 2010: accepted June 2010.
Table 1. Country Classification
ITers Non-ITers
Low inflation [6] AUS, CAN, KOR, [6] CHE, GER, HK,
NOR, SWE, UK ITA, JPN, US
High inflation [6] BRA, HUN, ISR, [2] ARG, TUR
MEX, PHL, ZAF
Numbers in square brackets represent the number of countries in the
group. High-inflation countries are defined as those with average
annual inflation rates greater than 10% before IT adoption (for ITers)
or break points (for non-ITers).
Table 2. Summary Statistics for Inflation and RPV
Month-to-Month
Inflation
Pre-IT Post-IT
AUS 1.42 [0.84] 0.67 [0.56]
BRA 12.14 [11.34] 0.56 [0.41]
CAN 0.34 [0.27] 0.15 [0.26]
HUN 1.34 [0.66] 0.42 [0.34]
ISR 3.06 [4.66] 0.21 [0.47]
KOR 0.47 [0.42] 0.24 [0.30]
MEX 2.61 [2.26] 0.37 [0.20]
NOR 0.32 [0.25] 0.15 [0.43]
PHL 0.88 [1.10] 0.44 [0.42]
SWE 0.51 [0.51] 0.12 [0.29]
UK 0.46 [0.33] 0.16 [0.17]
ZAF 0.90 [0.52] 0.49 [0.45]
ARG 15.64 [15.45] 0.58 [1.33]
CHE 0.27 [0.24] 0.08 [0.20]
GER 0.25 [0.28] 0.12 [0.20]
HK 0.61 [0.38] -0.03 [0.56]
ITA 0.51 [0.21] 0.19 [0.11]
JPN 0.13 [0.30] 0.00 [0.23]
TUR 4.67 [2.11] 1.63 [1.34]
US 0.27 [0.18] 0.21 [0.34]
Month-to-Month
RPV
Pre-IT Post-IT
AUS 1.11 [0.48] 0.89 [0.44]
BRA 2.26 [1.86] 0.47 [0.26]
CAN 0.40 [0.27] 0.45 [0.33]
HUN 0.71 [0.59] 0.57 [0.45]
ISR 1.48 [1.02] 0.81 [0.37]
KOR 0.63 [0.38] 0.60 [0.31]
MEX 0.86 [0.78] 0.36 [0.19]
NOR 0.41 [0.26] 0.65 [0.50]
PHL 0.58 [0.52] 0.41 [0.28]
SWE 0.72 [0.65] 0.54 [0.33]
UK 0.38 [0.19] 0.44 [0.19]
ZAF 1.02 [0.76] 0.83 [0.47]
ARG 2.86 [2.07] 0.64 [0.54]
CHE 0.41 [0.30] 0.42 [0.35]
GER 0.37 [0.40] 0.45 [0.36]
HK 0.56 [0.30] 0.91 [0.96]
ITA 0.32 [0.22] 0.23 [0.12]
JPN 0.39 [0.26] 0.33 [0.24]
TUR 2.48 [1.30] 1.43 [0.85]
US 0.25 [0.16] 0.46 [0.47]
12-Month
Inflation
Pre-IT Post-IT
AUS 5.73 [2.69] 2.69 [1.42]
BRA 151.35# [117.56] 6.80 [2.93]
CAN 4.08 [1.15] 2.06 [1.15]
HUN 16.24# [5.49] 5.46 [1.79]
ISR 35.90# [48.97] 2.73 [2.83]
KOR 5.53 [2.01] 3.16 [1.51]
MEX 31.16# [24.28] 4.59 [0.99]
NOR 3.82 [2.12] 1.90 [1.34]
PHL 10.48# [9.20] 5.35 [2.49]
SWE 6.24 [2.47] 1.60 [1.35]
UK 6.04 [1.37] 1.92 [0.76]
ZAF 11.07# [3.72] 5.83 [3.19]
ARG 193.93# [115.27] 10.45 [20.16]
CHE 3.19 [1.68] 1.80 [0.47]
GER 2.86 [1.45] 1.54 [0.67]
HK 7.49 [2.35] -0.13 [3.09]
ITA 5.96 [1.84] 2.41 [0.75]
JPN 1.58 [1.13] 0.06 [0.86]
TUR 57.28# [11.40] 21.67 [13.60]
US 3.23 [1.12] 2.79 [1.03]
12-Month
RPV
Pre-IT Post-IT
AUS 2.86 [1.22] 2.45 [0.88]
BRA 11.18 [6.62] 3.18 [1.01]
CAN 1.60 [0.66] 1.93 [1.01]
HUN 4.28 [1.30] 3.57 [1.24]
ISR 5.90 [2.50] 3.10 [1.05]
KOR 3.42 [1.00] 2.78 [0.75]
MEX 5.27 [3.51] 1.87 [0.58]
NOR 2.07 [0.71] 3.17 [1.04]
PHL 3.40 [1.37] 2.29 [1.29]
SWE 3.15 [1.55] 2.60 [0.85]
UK 2.16 [0.55] 2.71 [0.77]
ZAF 4.80 [1.85] 4.53 [1.43]
ARG 11.18 [4.44] 3.76 [2.78]
CHE 1.04 [0.77] 1.76 [0.85]
GER 1.78 [0.78] 2.18 [0.85]
HK 2.73 [0.51] 3.42 [1.79]
ITA 1.51 [0.60] 1.18 [0.47]
JPN 1.54 [0.59] 1.31 [0.49]
TUR 8.24 [2.69] 5.87 [1.88]
US 1.40 [0.67] 1.86 [1.02]
The entries are the mean values of inflation rates and RPV during the
corresponding period, and the numbers in square brackets denote their
standard deviations. For non-ITers, the break points are determined
by the Bai-Perron's (1998) multivariate structural break tests for
inflation series. Boldface indicates the high-inflation countries and
their average annual inflation rates during the pre-IT period.
Note: Boldface indicates the high-inflation countries and
their average annual inflation rates during the pre-IT period is
indicated with #.
Table 3. Results of the Bai-Perron Test
Inflation
IT Adoption
Date Break Date CI
AUS 1993:II 1990:IV [90:I-91:IV]
BRA 1999:M6 1994:M6 [94:M6-00:M9]
CAN 1991:M2 1991:Ml [89:M8-91:M7]
HUN 2001:M6 1997:M7 [97:M5-98:M4]
2001:M7 [00:M10-02:M6]
ISR 1997:M6 -- --
KOR 1998:M4 1998:M3 [97:M1-02:M12]
MEX 2001:M1 1988:M4 [87:M12-89:M12]
1999:M3 [99:M3-05:M7]
NOR 2001:M3 1989:M7 [88:M8-90:M3]
PHL 2002:M1 --
SWE 1993:M1 1991:M12 [91:M7-92:M11]
UK 1992:M10 1992:M1 [91:M11-92:M8]
ZAF 2000:M2 1993:M5 [92:M5-93:M9]
ARG 1991:M4 [91:M4-96:M2]
CHE 1988:M12 [88:M1-89:M9]
1993:M6 [93:M3-93:M12]
GER 1992:M10 [92:M8-94:M4]
HK 1997:M3 [96:M9-97:M5]
ITA 1995:M11 [95:M9-96:M7]
JPN 1993:M8 [91:M12-96:M2]
TUR 2002:M3 [02:M2 02:M9]
US 1999:M9 [97:M12-00:M10]
RPV
Break Date CI
AUS --
BRA 1992:M3 [92:M3-95:M5]
1996:M2 [96:M12-96:M6]
2003:M8 [03:M1-06:M12]
CAN --
HUN -- --
ISR 1987:M9 [87:M7-90:M6]
1993:M4 [92:M10-94:Ml1]
KOR --
MEX 1990:Ml [89:M12-91:M1]
NOR 2001:M1 [97:M8-01:M9]
PHL -- --
SWE -- --
UK 2005:M12 [03:M8-06:M7]
ZAF -- --
ARG 1991:M6 [91:M6-93:M2]
CHE --
GER -- --
HK -- --
ITA 1991:M10 [90:M8-94:M7]
JPN -- --
TUR 2000:M12 [00:M9-01:M12]
US -- --
"IT adoption" represents the month (quarter for AUS) and year of the
inflation target announcement as explained in section 2. Entries
represent the occurrence of break points in the year and month
estimated by the sequential procedure estimation method of Bai and
Perron (1998, 2003). In brackets are the 90% confidence intervals (CI)
for the end dates. We consider a partial structural change model of
[y.sub.t]] = [x'.sub.t] + [z'.sub.t] [[delta].sub.j] + [u.sub.t] with
t = [T.sub.j-1] + 1, ..., [T.sub.j], by setting RPV or inflation as
[y.sub.t], the lagged terms of the dependent variable as [x.sub.t],
and the constant term as [z.sub.t] such that the coefficients for the
constant term are allowed to shift. By adding lagged terms of the
dependent variable as regressors, no serial correlation is assumed in
the errors terms, {[u.sub.t]}. Following the guidelines from Bai and
Perron, we assume that the break does not occur during the initial
15%r or the final 15% of the sample period in testing for structural
breaks. The maximum number of breaks is set to five and minimum regime
size to 5% of the sample. Robust standard errors are used based on a
quadratic spectral kernel HAC estimator with AR(1) prewhitening
tillers. An entry of "--" indicates that the series does not exhibit a
statistically significant break.
Table 4. Parametric Regression Results
Full Sample
(p, q) [[??].sub.1] (SE) [[??].sub.2] (SE)
AUS (1.0) -0.44 (0.02) 0.21 (0.00)
BRA (1.0) 0.09 (0.00) 0.00 (0.00)
CAN (1.0) -0.44 (0.00) 0.54 (0.00)
HUN (1.0) -0.15 (0.01) 0.19 (0.00)
ISR (2.9) 0.04 (0.00) 0.01 (0.00)
KOR (1.0) -0.40 (0.01) 0.60 (0.00)
MEX (1.2) 0.27 (0.00) 0.01 (0.00)
NOR (1.0) -0.54 (0.00) 0.86 (0.00)
PHL (7.0) -0.20 (0.00) 0.22 (0.00)
SWE (3.0) -0.23 (0.00) 0.48 (0.00)
UK (3.0) -0.16 (0.00) 0.25 (0.00)
ZAF (1.0) -0.42 (0.02) 0.39 (0.00)
ARG (5.2) 0.02 (0.00) 0.00 (0.00)
CHE (3.0) -0.54 (0.01) 1.32 (0.02)
GER (1.0) -0.25 (0.02) 1.06 (0.02)
HK (1.0) -0.35 (0.00) 0.45 (0.00)
ITA (1.0) -0.32 (0.01) 0.72 (0.02)
JPN (1.0) -0.15 (0.00) 0.60 (0.00)
TUR (1.0) 0.31 (0.00) 0.00 (0.00)
US (4.0) -0.53 (0.00) 1.17 (0.00)
Full Sample
[??] * [5%, 95%]
AUS 4.3 ([double dagger]) [3.1, 5.4]
BRA 351.6 ([double dagger]) [242.6, 460.6]
CAN 4.9 ([double dagger]) [3.7, 6.11
HUN 4.7 ([double dagger]) [-0.3, 9.7]
ISR -40.2 [-156.5, 76.21
KOR 4.0 ([double dagger]) [3.1, 4.9]
MEX -270.0 [-702.6, 162.51
NOR 3.8 ([double dagger]) [3.1, 4.4]
PHL 5.4 ([double dagger]) [3.5, 7.4]
SWE 2.8 ([double dagger]) [1.5, 4.2]
UK 3.7 ([double dagger]) [1.5, 5.9]
ZAF 6.5 ([double dagger]) [4.0, 8.9]
ARG 653.2 ([double dagger]) [225.9, 1080.5]
CHE 2.4 ([double dagger]) [1.9, 3.01
GER 1.4 ([double dagger]) [0.2, 2.7]
HK 3.0 ([double dagger]) [2.0, 3.9]
ITA 2.6 ([double dagger]) [1.7, 3.6]
JPN 1.5 ([double dagger]) [0.6, 2.4]
TUR 1714.4 [-9371, 12,800]
US 2.7 ([double dagger]) [2.4, 3.1]
Numerical Target
(%)
AUS 2-3 (since 1993)
BRA 2.5-6.5 (since 2005))
CAN 1-3 (since 1994)
HUN 2-4 (since 2006)
ISR 1-3 (since 2003)
KOR 2-4 (since 2003)
MEX 3 (since 2003)
NOR 2.5 (since 2001)
PHL 2.5-4.5 (since 2009)
SWE 1-3 (since 1993)
UK 2 (since 1997)
ZAF 3-6 (since 2006)
ARG n.a.
CHE 2
GER 2
HK n.a
ITA 2
JPN 0-2
TUR 2-6
US 2-3
Before Break
(p, q) [[??].sub (SE) [[??].sub (SE)
AUS (1.0) -0.64 (0.09) 0.26 (0.01)
BRA (1.0) 0.07 (0.00) 0.00 (0.00)
CAN (1.0) -0.10 (0.03) 0.26 (0.01)
HUN (1.0) -0.02 (0.04) 0.15 (0.00)
ISR (2.9) -0.01 (0.00) 0.01 (0.00)
KOR (1.0) -0.56 (0.01) 0.68 (0.00)
MEX (1.2) 0.29 (0.01) 0.01 (0.00)
NOR (1.0) -0.33 (0.02) 0.56 (0.02)
PHL (7.0) -0.18 (0.00) 0.21 (0.00)
SWE (3.0) -0.21 (0.02) 0.47 (0.00)
UK (1.0) -0.10 (0.02) 0.22 (0.00)
ZAF (1.0) -0.25 (0.04) 0.30 (0.01)
ARG (1.1) 0.04 (0.00) 0.00 (0.00)
CHE (1.3) -0.47 (0.02) 1.11 (0.04)
GER (3.0) -0.42 (0.04) 1.11 (0.02)
HK (1.0) -0.38 (0.00) 0.39 (0.00)
ITA (1.0) -0.09 (0.19) 0.46 (0.17)
JPN (1.0) -0.18 (0.01) 0.56 (0.01)
TUR (1.0) 0.30 (0.01) 0.00 (0.00)
US (4.0) -0.85 (0.01) 1.43 (0.02)
Before Break
[??] * [5%, 95%]
AUS 5.0 ([double dagger]) [3.6, 6.4]
BRA 359.8 ([double dagger]) [165.0, 554.5]
CAN 2.4 [-4.3, 9.0]
HUN 0.7 [-14.5, 15.9]
ISR -16.7 [-138.9, 105.5]
KOR 4.9 ([double dagger]) [4.0, 5.9]
MEX -283.7 [-888.0, 320.7]
NOR 3.6 ([double dagger]) [2.0, 5.2]
PHL 5.4 ([double dagger]) [3.0, 7.9]
SWE 2.7 [-0.2, 5.7]
UK 2.8 [-3.9, 9.6]
ZAF 6.3 ([double dagger]) [2.8, 9.8]
ARG 796.9 ([double dagger]) [175.4, 1418.4]
CHE 2.5 ([double dagger]) [1.6, 3.5]
GER 2.3 ([double dagger]) [0.4, 4.1]
HK 5.4 ([double dagger]) [4.1, 6.6]
ITA 1.2 [-7.8, 10.3]
JPN 1.9 ([double dagger]) [0.6, 3.1]
TUR 1236.8 [-7141, 9614]
US 3.6 ([double dagger]) [3.2, 4.0]
After Break
(p, q) [[??].sub.1] (SE) [[??].sub.2] (SE)
AUS (1, 0) -0.34 (0.03) 0.19 (0.00)
BRA (2, 0) -0.10 (0.01) 0.22 (0.00)
CAN (1, 0) -0.51 (0.00) 1.55 (0.01)
HUN (1, 0) -0.58 (0.02) 0.81 (0.01)
ISR (1, 0) -0.06 (0.01) 0.12 (0.00)
KOR (1, 0) -0.25 (0.01) 0.75 (0.03)
MEX (1, 0) -1.75 (0.05) 2.36 (0.09)
NOR (4, 0) -0.30 (0.01) 0.76 (0.00)
PHL (1, 0) -0.49 (0.01) 0.50 (0.00)
SWE (1, 0) -0.21 (0.01) 0.66 (0.02)
UK (1, 0) -0.34 (0.01) 1.50 (0.05)
ZAF (1, 0) -0.88 (0.02) 0.87 (0.01)
ARG (8, 0) -0.02 (0.00) 0.04 (0.00)
CHE (3, 0) -0.51 (0.01) 1.94 (0.08)
GER (1, 0) -0.32 (0.02) 2.02 (0.13)
HK (1, 0) 0.40 (0.06) 0.61 (0.02)
ITA (1, 0) -0.76 (0.02) 2.40 (0.12)
JPN (1, 0) -0.03 (0.00) 1.92 (0.06)
TUR (1, 0) -0.28 (0.01) 0.10 (0.00)
US (1, 0) -0.43 (0.00) 1.18 (0.00)
After Break
[??] * [5%, 95%]
AUS 3.7 ([double dagger]) [1.4, 5.9]
BRA 2.8 [-2.0, 7.7]
CAN 2.0 ([double dagger]) [1.5, 2.5]
HUN 4.3 ([double dagger]) [2.7, 5.9]
ISR 2.4 [-5.4, 10.1]
KOR 2.0 ([double dagger]) [0.6, 3.4]
MEX 4.4 ([double dagger]) [4.1, 4.8]
NOR 2.4 ([double dagger]) [1.3, 3.5]
PHL 5.7 ([double dagger]) [4.4, 7.0]
SWE 1.9 ([double dagger]) [0.1, 3.7]
UK 1.3 ([double dagger]) [0.8, 1.9]
ZAF 6.0 ([double dagger]) [4.2, 7.1]
ARG 3.3 [-4.5, 11.0]
CHE 1.6 ([dagger]) [0.9, 2.2]
GER 1.0 ([double dagger]) [0.2, 1.7]
HK -1.9 ([double dagger]) [-3.3, -0.5]
ITA 1.9 ([double dagger]) [1.5, 2.3]
JPN 0.1 [-0.3, 0.5]
TUR 17.3 ([double dagger]) [9.1, 25.5]
US 2.2 ([double dagger]) [1.6, 2.8]
Regression equation is [RPV.sub.t] = [[alpha].sub.0] + [p.summation
over (h=1)] [[alpha].sub.h] [RPV.sub.t-h] + [[beta].sub.1][[pi].sub.t]
+ [[beta].sub.2][[pi].sup.2.sub.t] + [q.summation over (j=1)]
[[sigma].sub.j][[pi].sub.t-j] + [[epsilon].sub.t], where the lag
lengths (p, q) are selected by the BIC rule. Break points are IT
adoption date for ITers and structural break points of inflation for
non-ITers. [??] * is the annualized monthly inflation rate, obtained
from [??] * = [(-[[??].sub.1]-2[[??].sub.2]) x 12]; 5% and 95% inside
the square brackets represent the lower and = numerical inflation
upper bounds of the 95% confidence interval of [??] *, which is
calculated using the delta method. ([double dagger]) represents that
[H.sub.0] : [[pi].sup.*] = 0 can be rejected at 5%. The targets are
obtained from the IMF's Monetary Bulletin (2007-2) and central banks'
websites.
