Money-Based Interest Rate Rules: Lessons From German Data: February 2007
Money-Based Interest Rate Rules: Lessons From German Data: February 2007
Money-Based Interest Rate Rules: Lessons From German Data: February 2007
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/23755091
CITATIONS
READS
25
3 authors:
Christina Gerberding
Franz Seitz
Deutsche Bundesbank
Hochschule Amberg-Weiden
SEE PROFILE
SEE PROFILE
Andreas Worms
Deutsche Bundesbank
25 PUBLICATIONS 1,050 CITATIONS
SEE PROFILE
Franz Seitz
(University of Applied Sciences Amberg-Weiden, WSB Poznan, Poland)
Andreas Worms
(Deutsche Bundesbank)
Discussion Paper
Series 1: Economic Studies
No 06/2007
Discussion Papers represent the authors personal opinions and do not necessarily reflect the views of the
Deutsche Bundesbank or its staff.
Editorial Board:
Heinz Herrmann
Thilo Liebig
Karl-Heinz Tdter
Abstract:
The paper derives the monetary policy reaction function implied by money growth
targeting. It consists of an interest rate response to deviations of the inflation rate from
target, to the change in the output gap, to money demand shocks and to the lagged
interest rate. In the second part, it is shown that this type of inertial interest rate rule
characterises the Bundesbanks monetary policy from 1979 to 1998 quite well. This
result is robust to the use of real-time or ex post data and to the consideration of serially
correlated errors. The main lesson is that, in addition to anchoring long-term inflation
expectations, monetary targeting introduces inertia and history-dependence into the
monetary policy rule. This is advantageous when private agents have forward-looking
expectations and when the level of the output gap is subject to persistent measurement
errors.
Keywords:
JEL-Classification
Nicht-technische Zusammenfassung
Die Geldpolitik wird in Makromodellen blicherweise ber eine geldpolitische
Reaktionsfunktion fr die kurzfristigen Zinsen beschrieben. Besondere Popularitt
erlangte dabei wegen ihrer Einfachheit und ihrer angeblichen Robustheit die sog.
Taylor-Regel. Danach soll der kurzfristige Realzins ber sein Gleichgewichtsniveau
hinaus erhht werden, wenn die Inflation ihr Ziel berschreitet und/oder wenn das
Niveau des realen Outputs ber seinem Potenzialwert liegt. In jngster Zeit wurden
diese Regeln allerdings wegen der Nicht-Bercksichtigung von Datenunsicherheit
kritisiert. So konnte z.B. gezeigt werden, dass die Bundesbankpolitik bei Verwendung
derjenigen Daten, die den Entscheidungstrgern zum Zeitpunkt der Entscheidungen
tatschlich vorlagen (sog. Echtzeitdaten), durch eine Taylor-Regel nicht adquat
beschrieben werden kann. Vielmehr reagierte die Bundesbank, die als eine der
erfolgreichsten Zentralbanken gilt, nicht auf das Niveau der Produktionslcke, sondern
auf dessen Vernderung. Zudem war ihre Zinspolitik durch ein hohes Ma an Persistenz
gekennzeichnet. Beide Charakteristika werden von der modernen makrokonomischen
Theorie als wnschenswerte Eigenschaften robuster geldpolitischer Regeln angesehen.
In dem vorliegenden Papier zeigen wir zunchst, dass die Reaktion einer Zentralbank
auf
das
Outputwachstum
relativ
zum
Potenzial
direkte
Konsequenz
einer
Geldmengenorientierung ist, wie sie die Bundesbank von 1975 bis 1998 verfolgte. Um
dies zu zeigen, formulieren wir ein kleines Modell und leiten die Zinsregel ab, die sich
im Rahmen dieses Modells aus einer Geldmengenorientierung ergibt. Eine derartige
geldmengenbasierte
Zinsregel
beinhaltet
als
weitere
Feedback-Variablen
den
verzgerten Zins, Abweichungen der Inflation vom Inflationsziel und unter Umstnden
kurzfristige Geldmengenbewegungen. Aus dieser Ableitung folgt auch, dass die
Geldmengenorientierung eine zustzliche und wnschenswerte Abhngigkeit der
Zinsentscheidungen von der wirtschaftlichen Entwicklung der Vorperioden (history
dependence) impliziert. Mit dieser Zinsregel kann die Bundesbankpolitik von 1979 bis
1998 gut nachvollzogen werden.
Im Anschluss an die Prsentation der Schtzergebnisse werden die Argumente der
geschtzten Reaktionsfunktion einer nheren Untersuchung und Plausibilittsprfung
unterzogen. Zum einen wird gezeigt, dass die Konzentration auf die Vernderung der
Produktionslcke die Geldpolitik weniger anfllig fr Messfehler und -ungenauigkeiten
macht. Zweitens stellen wir fest, dass die fr die Bundesbank geschtzte Zinsregel
starke hnlichkeit mit der optimalen zeitinvarianten Zinsregel hat, die sich im Rahmen
makrokonomischer
Standardmodelle
bei
vorausschauendem
Verhalten
des
Privatsektors ergibt. Drittens zeigen wir, dass der groe Einfluss des verzgerten
Zinssatzes auf die aktuellen Zinsentscheidungen tatschlich ein Zinsglttungsmotiv
reflektiert und nicht wie in der Literatur manchmal unterstellt - Messfehler beim
Zielzins oder die Nicht-Bercksichtigung wichtiger erklrender Variablen. Darber
hinaus knnen wir begrnden, warum die Zinspolitik der Bundesbank vor dem
Hintergrund der angelegten mittelfristigen Perspektive und dem Hauptziel der
Preisstabilitt vorausschauend war.
Die empirische Evidenz fr Deutschland und neuere theoretische Erkenntnisse zur
optimalen regelgebundenen Geldpolitik drften auch fr das Euro-Whrungsgebiet
relevant sein. Vor dem Hintergrund der erhhten Unsicherheit seit Beginn der
Whrungsunion scheint die hervorgehobene Rolle der Geldmenge in der geldpolitischen
Strategie des Eurosystems ein sinnvoller Ansatz zu sein.
Contents
1 Introduction
13
4.1
13
4.2
15
4.3
17
5 Conclusions
References
18
20
25
Table 1b:
25
Table 2a:
26
Table 2b:
26
Table 3:
27
setting
Figure 1
1)
24
Introduction
There is an extensive literature on optimal and estimated monetary policy reaction
functions. These range from the "classic" Taylor rule (Taylor, 1993) and numerous
variants of it (e.g. Clarida et al., 1998; Mehra, 2001; Christiano and Rostagno, 2001;
Gerlach-Kristen, 2003; Chadha et al., 2004) to nominal income rules (e.g. McCallum
and Nelson, 1999; Rudebusch, 2002a) and different specifications of speed limit
policies (Orphanides, 2003b; Walsh, 2004; Bernhardsen et al, 2005; Gerberding, Seitz
and Worms, 2005). In the last decade, the most prominent monetary policy rules were
those in the spirit of Taylor (1993). According to these rules, the short-term real interest
rate should be raised if inflation increases above target and/or if the level of real output
rises above trend. The popularity of such rules stems from their simplicity and their
good performance across a wide array of macroeconomic models. In addition, the case
for Taylor rules has been strengthened by the claim made by Clarida et al. (1998) and
others that both the Feds monetary policy under Paul Volcker and Alan Greenspan and
the Bundesbanks monetary policy during the era of monetary targeting (1979-1998)
can be well captured by a forward-looking variant of the Taylor rule.
However, one shortcoming of these studies is that they abstract from the
measurement problems which policymakers face with respect to key variables entering
the Taylor rule like the Equilibrium level of the real interest rate and the level of the
output gap. For the US, Orphanides (2001, 2003b) has convincingly demonstrated that
the use of real-time information can considerably change the outcome of an analysis of
past monetary policy decisions. In particular, he finds that a Taylor rule based on realtime estimates of inflation and the output gap tracks the Feds monetary policy in the
1970s quite closely and thus would not have been helpful in avoiding the policy
mistakes of that era which can be identified today with the advantage of hindsight. In a
*
E-mail: [email protected], [email protected], [email protected]. We thank seminar participants at the OENB, the University of Stuttgart-Hohenheim and H.
Herrmann for helpful comments. The views expressed in this paper are those of the authors and should
not be interpreted as those of the Deutsche Bundesbank.
similar vein, Gerberding et al. (2004, 2005) have shown that the use of real-time data
for Germany considerably changes the assessment of the Bundesbanks monetary policy
reaction function. According to that analysis, the Bundesbank did not respond to the
level of the output gap as suggested by the Taylor rule, but rather to the change in the
output gap as well as to deviations of (expected) inflation and money growth from their
respective target values. Furthermore, the results suggest that the monetary policy of the
Bundesbank was characterised by a high degree of interest rate inertia.
Interestingly, targeting the rate of change rather than the level of the output gap
has recently been advocated by a number of authors, such as Orphanides (2003a) and
Walsh (2003, 2004). One argument in favour of such an approach is that estimates of
the level of the output gap are subject to much greater uncertainty than estimates of its
change. Another advantage is that targeting the change in the output gap makes
monetary policy more history-dependent, which is an important component of an
optimal commitment policy in forward-looking models (Woodford, 1999). However,
the latter argument has been put forward only recently, and thus does not answer the
question why the Bundesbank might have looked more at changes than at the level of
the output gap.
In the present paper, we argue that the Bundesbanks focus on output growth
rather than on the level of real output (relative to potential) was a direct consequence of
its use of money as an intermediate target and indicator variable. To demonstrate this,
we first derive the interest rate reaction function implied by monetary targeting (part 2).
In our model, money-based interest rate rules feature a response to the lagged interest
rate, to deviations of inflation from target, to the change in the output gap and possibly,
but not necessarily, to short-run movements of money in addition. In the third part of
the paper, we show that this type of inertial interest rate rule characterises the
Bundesbanks monetary policy from 1979 to 1998 quite well. Furthermore, we
demonstrate that this result is robust to the use of ex post or real-time data. In section 4,
we discuss the economic reasoning and consequences of all the arguments incorporated
in the Bundesbanks interest rate rule. Here, we also consider the question whether the
large weight of the lagged interest rate in the estimated policy rule represents a genuine
feature of the Bundesbanks monetary policy or simply picks up the influence of serially
correlated errors (Rudebusch, 2002b, 2005). Section 5 summarises and concludes.
growth and at least according to its own descriptions based monetary policy
decisions on deviations of actual money growth from these targets. Contrasting this
view, recent empirical studies argue that its interest rate decisions can be described
sufficiently well by a standard forward-looking Taylor rule which relates the short-term
nominal interest rate to deviations of expected future inflation from target,
)]
(1)
Here, it is a short-term nominal interest rate and it* is its long-run equilibrium
level (which is equal to the sum of the long-run equilibrium real rate of interest plus the
trend rate of inflation). The parameters , and capture the degree of interest rate
smoothing and the strength of the interest rate response to the inflation gap and the
output gap, respectively. E is the expectation operator and t is the information
available to the central bank at the time it sets interest rates (t). In order to test the
relevance of money for the Bundesbanks interest rate decisions, these studies simply
add a measure of the money gap (e.g. the deviation of money growth from target) to the
list of explanatory variables:1
(1a)
one step further and show that depending on the central banks success in identifying
shocks to money demand, the interest rate rule implied by monetary targeting need not
include a monetary variable at all. To clarify this issue, it proves useful to reformulate
monetary targeting in terms of an interest rate reaction function (see also Orphanides,
2003b, 990ff.). In fact, when explaining its strategy to the public, the Bundesbank
always made the point that money growth did not serve as an instrument of monetary
policy, but as an intermediate target and indicator variable in order to achieve the
ultimate goal of price stability. Interpreted in this sense, monetary targeting can be
represented by an interest rate reaction function of type (1) which links the short-term
interest rate to deviations of money growth from the announced target value (all
variables except interest rates are in logarithms):
it = (1 ) it* + (1 ) E ( mt t ) mtT + it 1 ,
(2)
where captures the strength of the interest rate response to the money growth
m t = y t + p t v t
Under the assumption that the long-run growth rates of output and velocity are
exogenous to monetary policy, the trend rate of inflation moves one to one with the
trend rate of money growth. Accordingly, the Bundesbank derived its monetary
target, mtT , from its target value for the rate of change in prices (the so-called price
assumption or price norm), ptT , plus the growth rate of the production potential,
yt* , minus the trend rate of change in the velocity of circulation vt* . The latter two
(4)
This procedure disregards the fact that a strategy of monetary targeting is only a means to reach the final goal of
price stability.
Combining Eq. (3) and (4), we can decompose the money growth gap
( m
mtT into an inflation gap pt ptT , a real output growth gap yt yt*,est ,
which is equivalent to the change in the output gap, and a velocity growth gap
(v
vt*,est :
) (
) (
(5)
Finally, we can replace vt with the help of a standard money demand function
that relates real money holdings (mt pt ) to output yt (transactions variable) and the
interest rate it (opportunity cost variable) in a stable and predictable way:3
(mt pt ) = 1 yt 2 it + t
(6)
The term captures all deviations of actual money holdings from the level
explained by fundamentals, especially short-run dynamics and money demand shocks.
Taking first differences of (6) and inserting in (3) yields
vt = (1 1 ) y t + 2 it t
(7)
In the long-run, output growth equals potential output growth and = 0, so that
the trend change in velocity is a function of potential output growth, y*, and the trend
change in the steady-state (equilibrium) nominal interest rate, i*:
vt* = (1 1 ) yt* + 2 it*
(8)
In order to estimate the magnitude of v*, the central bank has to estimate the
parameters 1 and 2, as well as y*, and i*:
vt*, est = (1 1 ) yt*, est + 2 it*, est
(8a)
According to the Fisher equation, the long-run nominal interest rate can be
decomposed into the long-run (natural) real rate of interest and the long-run rate of
inflation, that is it* = rt* + pttr , so that any trend change in i* must be due to an upward
or downward trend in the real rate of interest and/or to a change in trend inflation. While
it may be argued that the successive lowering of the Bundesbanks price norm from 5 %
3
The money demand function in (6) is a very standard one to concentrate on the main arguments. A more
sophisticated forward-looking specification within a similar spirit may be found in Kajanoja (2003)
in 1975 to 2 % in 1985 did in fact lead to a decrease in the trend rate of inflation, the
Bundesbank did not take this into account when deriving its money growth targets but
assumed that the nominal interest rate is constant in the long-run.4 Setting i*,est equal to
zero and substituting (7) and (8a) into (5) yields:
(5a)
Eq. (5a) can now be used to rewrite Eq. (2) in a way that facilitates comparison
with the class of Taylor rules described by Eq. (1). Provided that the central banks
estimate of 1 is unbiased, the term ( 1 1 )yt*,est can be subsumed into the error term,
which leaves us with:
(9)
+ (1 ) 2
1 + (1 ) 2
At first sight, Eq. (9) looks rather similar to the Taylor rule (1) in that it includes
the rate of inflation, the output gap and the lagged interest rate as feedback variables.
However, a closer inspection reveals some important differences:
(a) Monetary targeting as expressed by (9) implies a policy response to the difference
between the growth rate of actual output and the (estimated) growth rate of potential
output growth whereas the Taylor rule (1) includes a response to the estimated level
of the output gap. Hence, monetary targeting introduces history dependence into the
policy rule which is an important component of an optimal commitment policy
when agents have forward-looking expectations (see Woodford 1999).5
(b) Monetary targeting implies a response to the true values of p and y (which
determine money demand) while at the same time introducing measurement errors
in money growth, t, into the policy rule. By contrast, interest rate rules with a direct
feedback from prices and output - such as Taylor-type rules or nominal income
targeting - are vulnerable to measurement errors in these variables, but do not suffer
from measurement errors in money growth. The relative performance of these rules
4
See Deutsche Bundesbank (1992, p. 27f). One reason for ignoring an expected downward trend in the nominal
interest rate due to a trend decline in inflation is that it would imply an upward correction of the money growth
target which would in turn decrease the speed at which the trend rate of inflation is brought down.
(10)
with 1 > 2 0 . A positive value of 2 would either indicate that the staff made
systematic mistakes in estimating the long-run relationships or that, despite the
medium-term orientation of their approach, policymakers still showed some response to
short-run fluctuations in money demand (even if these were correctly identified).
5
6
Eq. (10) shows that simply amending the Taylor rule (1) by a money (growth) gap
in order to check whether monetary policy actually reacted to monetary aggregates does
not do justice to the medium-term nature of monetary targeting. To see this more
clearly, we can use Eq. (5a) to substitute for t:
+ ' 'it1 + t
(10a)
with ' ' = ( ' 2 ) /(1 2 ) . Note that Eq. (10a) only corresponds to our initial
representation of monetary targeting, Eq. (2), if 1 equals 2, i.e., if policymakers do not
filter actual money growth figures but react to any deviation of money growth from
target (which is at odds with the Bundesbanks own explanations of its approach). If,
however, 2 is smaller than 1, the interest rate rule implied by monetary targeting bears
some resemblance to the variants of nominal income growth targeting or natural
growth targeting advocated, among others, by McCallum and Nelson (1999) and
Orphanides (2004). However, one potential advantage of money growth targeting over
these two monetary policy strategies is that money growth may contain useful
information on the unobserved true value of current output growth whereas central
banks which target output growth directly have to rely on noisy estimates of this
variable. Of course, this advantage hinges critically on the relative magnitude of the
measurement errors in money and output data as well as on the central banks ability to
identify money demand shocks in real time and to separate short-run from long-run
influences on money demand (see Coenen et al., 2005).
The crucial difference between (10) and a Taylor rule is whether monetary policy
reacts to the change or to the level of the output gap. But whether monetary policy
indeed reacted to one or the other (or to both) can obviously only be tested empirically
by estimating a reaction function which contains both arguments. More generally,
nesting the ingredients of the forward-looking Taylor rule (Eq. (1)) and the feedback
variables implied by Eq. (10a) into one model leads to the following interest-rate rule (n
1):
(11)
)) + i
t 1
+ t
with p2 = (1 2 ), y = 1 (1 2 ), m = 2
If the estimated values of p1 and y were significantly positive and p 2 , y as
well as m turn out to be insignificant, then this would be evidence in favour of the
claim made by Clarida et al. (1998) that the Bundesbank preached monetary targeting,
but in fact followed a forward-looking Taylor rule. If, however, the estimated values of
studies on monetary policy reaction functions were based on finally revised data.
Influential examples include Clarida and Gertler (1997) or Clarida et al. (1998, 2000).
However, Orphanides (2001, 2003c) has pointed out that ex post data on key macro
variables may differ considerably from the information available to policymakers at the
time the decisions are made. This so-called real-time data problem stems from the fact
that some potential determinants of monetary policy suffer from considerable
measurement problems and are often substantially revised over time. Indeed, with the
advantage of hindsight we now know that measurement problems are particularly
pronounced for the level of the output gap, which plays a prominent role in interest rate
rules of the Taylor type. Interestingly, this is not specific to the US but seems to be an
international phenomenon (see Gerberding et al. (2005) for Germany, Gerdesmeier and
Roffia (2005) for the Euro Area, Kamada (2004) for Japan, Nelson and Nikolov (2001)
for the UK and Orphanides (2001) for the USA). For the purpose of practical monetary
policy, estimating reaction functions on revised data is hence inappropriate a priori
since this ignores serious measurement errors, leading to biased estimates (and test
statistics).
However, more recently, the argument has been put forward that the available
real-time data sets do not fully reflect the information set available to policymakers
when they took decisions. For instance, the analysis of a broad set of indicators may
have enabled policymakers to implicitly circumvent the measurement problems
underlying real-time estimates. If this were true, policymakers own (implicit) estimates
of key macro variables may differ from those contained in real-time data sets (which are
usually based on published data and staff estimates).
As the outcome of this debate is still open, our approach is to use ex post data as
well as real-time data to estimate the Bundesbanks reaction function. Looking at both
sets of results seems particularly appropriate in our context since the interest rate
representation of monetary targeting derived in Section 2 includes a response to the
true rate of inflation and the true growth rate of output. As our benchmark ex post
series, we match the last available vintage of official Bundesbank estimates of the
production potential (dating from Jan. 1999) with the March 1999 vintages of all other
data.7 By allowing the horizon of the inflation gap, n, to vary from zero to six, the two
variables measuring current and future price pressure can be subsumed into one term
and we are left with the following version of Eq. (11):
(12)
where the measurement error in money growth, t, has been set equal to zero, and
the superscript a indicates rates of change over the previous four quarters.
Furthermore, it* has been replaced by the sum of a constant and the price target, ptT+ n .
An important issue is the method used to generate the forecasts. Since we do not know
policymakers true forecasts of inflation, the output gap and the change in the output
gap, we follow the standard practice of using the realized values as proxies. Therefore,
the error term t is a linear combination of the forecast errors of inflation and output and
the exogenous disturbance term t. In order to avoid simultaneity problems, the RHSvariables are instrumented by a vector of variables It which belong to the central banks
information set at the time it sets interest rates and which are orthogonal to t. As we
7
10
Compared to the approach taken by Gerberding et al. (2005), we have reduced the number of instruments from 29
to 17. We do this to avoid the potential biases associated with using too many instruments, but the results are
largely unchanged.
11
smoothing. Moreover, the parameters of the change in the output gap and the inflation
gap are not too far apart from each other.9 In fact, for n=0, the point estimate of y is
even slightly above p, which is perfectly in line with the parameter restrictions implied
by monetary targeting in the case of 1 > 1 (see Eqs. (9) and (10)).10
These results prove to be quite robust to changes in the forecast horizon n
(1n6), the exact timing of the inflation and output variables, the concrete
specification of the money gap (annual growth rates, annualised 6-month growth rates,
level specifications), and to the choice of alternative proxies for the unobserved
forecasts of inflation (consumer prices, output deflator, consensus forecasts).11
However, what is perhaps most surprising is that the results based on real-time
data differ only slightly from the results in the ex-post setting. An obvious explanation
for this congruence is that (in contrast to other central banks) policymakers at the
Bundesbank focussed their attention on indicator variables which were exposed to
measurement error only to a comparatively small extent. Figure 1 illustrates that this is
indeed the case. First of all, as shown in Figure 1(a), the measurement errors regarding
the change in the output gap were much smaller and much less persistent than the
measurement errors regarding the level of the output gap. Second, when splitting up the
change in the output gap into actual output growth and potential output growth (Figure
1(b)), we find that the measurement errors in output growth and the change in the output
gap follow very similar patterns, while the measurement errors regarding potential
output growth are smaller, but more persistent. Finally, as illustrated by Figure 1(c),
revisions in consumer prices and in money growth were even smaller in size throughout
the sample period, with money growth figures being hardly ever revised at all. While
this may not be true for other countries over different sample periods, Coenen et al.
(2005) reach very similar conclusions with respect to euro-area data since 1999.
Rudebusch (2002a) shows that nominal income targeting performs well when inflation is forward-looking.
The income elasticity of broad money demand is generally estimated to be greater than one in the case of the euro
area and in Germany, see e.g., Bruggemann et al. (2003) and Scharnagl (1998).
11
See Table 2 in Gerberding et al. (2004) and further calculations which are available upon request.
10
12
output gap turns out to be insignificant in almost all regressions. Instead, the significant
and sizable response to both, inflation and the output growth gap as well as the high
degree of interest rate smoothing suggest that the Bundesbank took its money growth
targets seriously. However, there are also two aspects in which the results deviate from
the money based interest rate rule derived in Section 2. First, by responding to expected
future inflation rather than to current inflation, policymakers at the Bundesbank seem to
have taken a more forward-looking approach than implied by the interest rate rules (9)
and (10). Second, beyond the feedback from the variables implied by monetary
targeting, there seems to have been an additional, independent response to money
growth. We will discuss each of these results in turn.
4.1
The strong and robust influence of the change in the output gap on interest rate
decisions points to an omitted variables bias in standard Taylor rule specifications of
the Bundesbank reaction function like the one estimated by Clarida et al. (1998). In this
sense, our results throw serious doubt on the widespread practice of using the Taylor
rule even if it does not accurately describe central banks real-time behaviour as a
reasonably accurate ex-post description of monetary policy which may be exploited, for
instance, in the estimation of DSGE models based on ex-post data.
From a normative point of view, targeting the change rather than the level of the
output gap can be advantageous for two different reasons. First, as demonstrated by
Orphanides et al. (2000), there may be a case for responding to the change in the output
gap rather than to its level if the measurement errors in the level of the output gap are
large and highly persistent.12 The measurement errors in the level of the output gap are
defined as (the tilde refers to real-time values):
(y
yt* ) ( ~
yt ~
yt* ) = ( yt ~
y t ) ( yt* ~
yt* )
(13)
13
This high degree of persistence implies that, e.g., high positive errors in period t usually
follow high positive measurement errors in t-1. However, given this high degree of
persistence, the measurement errors of the change of the output gap
(y
y t* ) (~
y t ~
y t* ) = [( y t ~y t ) ( y t* ~
y t* )] [( y t 1 ~
y t 1 ) ( y t*1 ~
y t*1 )]
(14)
are much smaller than that of the level. Therefore, in normative terms, it may be
preferable to focus on output growth (relative to trend growth) rather than on the level
of the output gap. Orphanides (2003a), Orphanides et al. (2000) and Walsh (2004) show
that in the presence of imperfect information about the level of potential output,
monetary policy strategies such as inflation and output growth targeting, difference
rules or speed limit policies outperform simple Taylor-type rules.
Second, responding to the change in the output gap may be welfare-improving
since it introduces history-dependence into the policy rule, thereby stabilising inflation
expectations and, via the expectations channel, stabilising also actual inflation. To fully
understand the argument, consider the following example.13 Assume that policymakers
care about stabilising inflation, output and the interest rate around target values. In this
case, the central banks period loss function takes the form14
Lt = (pt ptT ) 2 + y ( yt y*) 2 + i (it i*) 2 ,
(15)
where y and i are the relative weights attached to output and interest rate
stabilization and y* is assumed to be consistent with the steady-state level of output.
Assume further that the aggregate demand and supply equations are of the standard
New-Keynesian type.15 Under these assumptions, the first order conditions which
characterize optimal monetary policy under discretion can be transformed into
it = i * + p (pt ptT ) + y ( yt y*)
(16)
Eq. (16) can easily be interpreted as a policy rule of the Taylor type. However,
with forward-looking price setting and a short-run output inflation trade-off, there are
gains from commitment to a policy rule. Under commitment, the central bank takes the
effects of its own actions on private sector expectations into account. As a consequence,
13
14
14
optimal policy is not purely forward-looking, but history-dependent in the sense that it
implies systematic responses to the lagged interest rate, to the lagged change in the
interest rate and to the lagged output gap. Choosing the commitment solution that is
optimal from a timeless perspective, the interest rate rule takes the form:16
it = (1 1 )i * + 1it 1 + 2 it 1 + p (pt ptT ) + y (yt yt* )
(17)
Comparing Eq. (17) with Eqs. (9) and (10), we find that the optimal time-invariant
policy rule under commitment shares many features with the interest rate representation
of flexible monetary targeting derived in Section 2.
4.2
The estimated Bundesbank reaction function implies a high degree of interest rate
smoothing, as measured by the persistence parameter . Recently, some authors have
questioned whether the significance of the lagged interest rate in estimated policy rules
reflects "true" interest rate smoothing, arguing that it may be caused either by
measurement error in the target interest rate (Lansing, 2002; Apel and Jansson, 2005) or
by the omission of variables from the reaction function to which policymakers actually
did respond (e.g. Rudebusch, 2002b). If this were true, efforts to identify the monetary
policy rule would create the illusion of interest rate smoothing behaviour when, in fact,
there is none.
In order to investigate these two potential sources of misspecification, it proves
useful to re-write the central banks reaction function as:
it = (1 ) it + it 1 + t ,
(18)
where i is the target interest rate determined by the policy rule. If i is measured
with error, one might expect the coefficient (1-) to be biased downwards, thus creating
the impression of partial adjustment. One plausible reason why i might be subject to a
measurement error is the use of data that were not available to policymakers at the time
of their policy decisions (that is, ex post data). In our setup, we can easily investigate
this possibility by comparing the estimates of based on ex post data with the
15
To ensure long-run neutrality, the inflation variable entering the Phillips curve and the output equation should be
defined as inflation relative to its steady-state value. See McCallum and Nelson (2004), p. 44, footnote 3.
15
corresponding estimates based on real-time data. Since we do not find any significant
differences in the estimates of , the high degree of partial adjustment found in our
estimates of the Bundesbanks policy rule obviously is not driven by measurement error
in the data entering the reaction function.
However, a number of other considerations could give rise to the impression of
interest rate smoothing were in fact there is none. For instance, Rudebusch has
repeatedly made the point that the lagged interest rate may simply pick up the influence
of serially correlated errors which may be caused, for instance, by serially correlated
omitted variables (see Rudebusch (2002, 2005)). English et al. (2003) have
demonstrated that this hypothesis can be tested empirically by estimating a model which
combines the partial adjustment model of Eq. (18) with an AR(1)-model for the error
term (19): 17
t = t 1 + t
(19)
it = (1 ) it + (1 ) (1 ) it 1 it 1 + it 1 + t
(20)
Estimating Eq. (20) for our preferred specification of the reaction function (12)
yields direct estimates of the interest rate smoothing parameter and of the ARparameter . As shown in Table 3, we cannot reject the hypothesis = 0 for all forwardlooking specifications ranging from n = 0 to 6.18 In contrast, the estimated values of the
other coefficients, including , remain largely unchanged compared to the baseline
specification. Hence, we conclude that the Bundesbank's monetary policy rule is
characterized by a high degree of true interest rate smoothing. This result is in line
with our finding from Section 2 that monetary targeting introduces additional inertia
16
The advantages of focussing on this solution are explained in Woodford (2003, p. 464ff)
The critique of Welz and sterholm (2005) on size distortions of this test is not justified in our case, as we should
have taken into account all relevant systematic influences on German monetary policy in our sample. The
exchange rate, one often mentioned variable in this direction, should be captured by the other variables in our
setting.
18
n = 0 is already forward-looking as a first estimate of the output gap in t is only available in t+1. The cases n = 2, 3
are not shown in the table as the coefficients are nearly the same as for the other forecast horizons. They are
available upon request.
17
16
into the policy rule, and it matches the Bundesbank professed preference for conducting
a steady-as-she-goes interest-rate policy ("Politik der ruhigen Hand").
4.3
17
Conclusions
In the present paper, we have taken up the question raised by Gerberding et al.
(2005) of why the Bundesbank might have looked more at changes than at the level of
the output gap. To shed light on this issue, we have shown that monetary targeting taken
seriously implies an interest rate response to deviations of inflation from target, to the
change in the output gap, to the lagged interest rate and to deviations of money demand
from long-run equilibrium. The latter is usually seen as the principal drawback of
monetary targeting. However, we have argued that a central bank with a focus on
money will be aware of this problem and, depending on the staffs success in
identifying shocks to money demand, the interest rate response to such shocks will be
muted or even non-existent.
With their implied response to the lagged interest rate and to the change in the
output gap, money growth targets introduce inertia and history-dependence into
monetary policy. As shown by Giannoni and Woodford (2003), both features are
important components of optimal monetary policy in standard New-Keynesian models
with forward-looking expectations. In addition, responding to the change in the output
gap rather than to its level may be advantageous when the latter is subject to large and
persistent measurement errors as has historically been the case. Furthermore, as pointed
out by Nelson (2003), beyond the stabilisation concerns captured by short-run models,
central banks have to be concerned with pinning down the steady-state rate of inflation,
and this was certainly the main motivation behind the Bundesbanks commitment to a
money growth target.
19
20
18
Therefore, the lessons to be drawn from the Bundesbanks experience differ from
those provided by Rudebusch and Svensson (2002) who compare the relative
performance of inflation targeting and monetary targeting in a small empirical model
estimated on US data. In contrast to our analysis, they conclude that the reaction
function resulting from monetary targeting is quite unsuitable for stabilizing inflation
and the output gap, even if there are no shocks to money demand. One reason for this
negative verdict is that there are no expectation channels and consequently, no gains
from commitment in their purely backward-looking model. While this may or may not
be an adequate model for the US economy, Gal et al. (2001), Smets and Wouters
(2003), ECB (2005), and most recently Stracca (2006) present evidence that inflation in
the euro area is characterized by a relatively low degree of intrinsic inertia.21 Moreover,
Woodford (2005) has shown that commitment continues to be important for optimal
policy even if the assumption of model-consistent expectations is replaced by the
weaker assumption of near-rational expectations.
A second reason for Rudebusch and Svenssons negative verdict on monetary
targeting is that their analysis abstracts from the problem of data uncertainty. In fact,
they argue that it is not obvious that monetary targeting would be favoured under such
uncertainty since money data are also subject to important revisions.22 Again, while this
may be true for the US (see Amato and Swanson, 1999), Coenen et al. (2005, 982)
show that the ECBs preferred measure of the broad money stock, M3, is subject to only
small revisions after the first quarter and to negligible revisions in subsequent quarters.
Hence, the available empirical evidence suggests that the lessons from German
data, together with the insights from recent research on optimal monetary policy under
commitment, are more relevant for the euro area than the lessons from US data
presented by Rudebusch and Svensson. Having said this and against the background of
the increased uncertainty monetary policy makers in EMU are confronted with, the
Eurosystems prominent role for money seems to be a sensible approach. Taken
seriously, this orientation introduces the necessary ingredients of a robust and inertial
monetary policy rule. However, in order to arrive at more definite conclusions, the
present analysis needs to be complemented by further studies which take account of the
21
22
Benigno and Lpez-Salido (2006) have shown that this is also true for Germany
See Rudebusch and Svensson (2002), footnote 26
19
structural relationships as well as of the degree of model and data uncertainty currently
prevailing in the euro area. This is an important task for future research.
References
Apel, M. & Jansson, P. (2005): Some Further Evidence on Interest-Rate Smoothing:
The role of measurement errors in the output gap, Working Paper no. 178,
Sveriges Riksbank, March.
Baltensperger, E. (1998): Monetary Policy under Conditions of Increasing Integration
(1979-96), in: Deutsche Bundesbank (ed.), 50 Years of the Deutsche Mark,
Clarendon Press, p. 439-524.
Benigno, P. & Lpez-Salido, J.D. (2006): Inflation Persistence and Optimal Monetary
Policy in the Euro Area, Journal of Money, Credit and Banking 38, 587-614.
Bernanke, B.S. & Mihov, I. (1997): What does the Bundesbank Target?, European
Economic Review 41, 1025-1053.
Bernhardsen, T., Eitrheim, ., Jore, A.S., & Risland, . (2005): Real-time Data for
Norway: Challenges for Monetary Policy, North American Journal of Economics
and Finance 16, 333-350.
Bruggemann, A., Donati, P. & Warne, A. (2003): Is the Demand for Euro Area M3
stable?, in: Issing, O. (ed.), Background Studies for the ECB's Evaluation of its
Monetary Policy Strategy, Frankfurt, pp. 245-300.
Chadha, J.S., Sarno, L., & Valente, G. (2004): Monetary Policy Rules, Asset Prices, and
Exchange Rates, IMF Staff Papers 51, 529-552.
Christiano L.J. & Rostagno, M. (2001): Money Growth Monitoring and the Taylor
Rule, NBER Working Paper No. 8539, October.
Clarida, R. and M. Gertler (1997): How the Bundesbank Conducts Monetary Policy, in:
C. Romer and D. Romer (eds.), Reducing Inflation: Motivation and Strategy,
Chicago, pp. 363412.
Clarida, R., Gal, J. & Gertler, M. (1998): Monetary Policy Rules in Practice: Some
International Evidence, European Economic Review 42, 1033-1067.
Clarida, R., Gal, J. & Gertler, M. (2000): Monetary Policy Rules and Macroeconomic
Stability: Evidence and Some Theory, Quarterly Journal of Economics 115, 147180.
Coenen, G., A. Levin and V. Wieland (2005): Data Uncertainty and the Role of Money
as an Information Variable for Monetary Policy, European Economic Review 49,
975-1006.
Deutsche Bundesbank (1992): The Correlation Between Money Growth and Price
Movements in the Federal Republic of Germany, Monthly Report January, 20-28.
Deutsche Bundesbank (1998): Monetary Policy Strategies in the Countries of the
European Union, Monthly Report January, p. 33-47.
20
English, W. B., W. R. Nelson & Sack, B.P. (2003): Interpreting the Significance of the
Lagged Interest Rate in Estimated Monetary Policy Rules, Contributions to
Macroeconomics, vol. 3, no 1.
European Central Bank (2005): Price-setting behaviour in the euro area, Monthly
Bulletin, November.
Gali, J., M. Gertler and D. Lopez-Salido (2001): European Inflation Dynamics,
European Economic Review, 45 (7), 1121-1150.
Gerberding, C., Seitz, F. & Worms, A. (2004): How the Bundesbank really conducted
monetary policy: An analysis based on real-time data, Deutsche Bundesbank
Discussion Paper Series 1, No 25/2004.
Gerberding, C., M. Kaatz, F. Seitz &. Worms, A (2005): A Real-Time Data Set for
German Macroeconomic Variables, Schmollers Jahrbuch, Journal of Applied
Social Science Studies, Jg. 125 (2005), Heft 2, p. 337-346.
Gerberding, C., Seitz, F. & Worms, A. (2005): How the Bundesbank Really Conducted
Monetary Policy, North American Journal of Economics and Finance 16, p. 277292.
Gerberding, C., Scharnagl, M. (2005): Taylor Rules versus Monetary Targeting:
Lessons from the German Experience, mimeo.
Gerdesmeier, D. & Roffia, B. (2004): Taylor Rules for the Euro Area: The Issue of
Real-Time Data, Deutsche Bundesbank Discussion Paper Series 1, No 37/2004.
Gerlach-Kristen, P. (2003): Interest Rate Reaction Functions and the Taylor Rule in the
Euro Area, ECB Working Paper No. 258, September 2003.
Giannoni, M. and M. Woodford (2003): Optimal Interest Rate Rules: I. Theory, NBER
Working Paper 9419.
Hubrich, K. (1999): Estimation of a German Money Demand System - a long-run
analysis, Empirical Economics 24, p. 77-99.
Kajanoja, L. (2003): Money as an Indicator Variable for Monetary Policy when Money
Demand is Forward Looking, Bank of Finland Discussion Paper 9-2003.
Kamada, K. (2004): Real-Time Estimation of the Output Gap in Japan and its
Usefulness for Inflation Forecasting and Policymaking, Deutsche Bundesbank
Discussion Paper Series 1, No 14/2004.
Kara, H. (2003): Optimal Monetary Policy Rules under Imperfect Commitment:
Reconciling Theory with Evidence, Manuscript, March 2003.
Lansing, K.J. (2002): Real-Time Estimation of Trend Output and the Illusion of Interest
Rate Smoothing, Federal Reserve Bank of San Francisco Economic Review, 1734.
McCallum, B.T. & Nelson, E. (1999): Performance of Operational Policy Rules in an
Estimated Semiclassical Structural Model, in: Taylor, J.B. (ed.), Monetary Policy
Rules, The University of Chicago Press, Chicago and London, 15-54.
21
22
Stracca, L. (2006): A speed limit monetary policy rule for the euro area, ECB Working
Paper No. 576, April 2006.
Svensson, L. (1999): Inflation Targeting: Some Extensions, Scandinavian Journal of
Economics 101(3), p. 337-361.
Taylor, J.B. (1993): Discretion versus Monetary Policy Rules in Practice, in: Carnegie
Rochester Conference Series on Public Policy 39, 195-214.
Walsh, C.E. (2002): Speed Limit Policies: The output gap and optimal monetary policy,
manuscript, July 2002; a shorter version was published in the American Economic
Review, 93(1), March 2003, 265-278.
Walsh, C.E. (2004): Implications of a Changing Economic Structure for the Strategy of
Monetary Policy, in: Monetary Policy and Uncertainty, Jackson Hole Symposium
2003, Federal Reserve Bank of Kansas City, 297-348.
Welz, P. & sterholm, P. (2005): Interest Rate Smoothing versus Serial Correlated
Errors in Taylor Rules: Testing the Tests, Uppsala University, Department of
Economics, Working Paper 2005:14, March.
Wolters, J., Tersvirta, T. & Ltkepohl H. (1998): Modeling the Demand for M3 in the
Unified Germany, Review of Economics and Statistics 80, p. 399-409.
Woodford, M. (1999): Optimal Policy Inertia, NBER Working Paper 7261, August
1999.
Woodford, M. (2003): Interest & Prices - Foundations of a Theory of Monetary Policy,
Princeton University Press, Princeton and Oxford.
Woodford, M. (2005): Robustly Optimal Monetary Policy with Near-Rational
Expectations, NBER Working Paper No. 11896, Dec. 2005.
23
O u t p u t ga p *
C h a n ge i n o u t p u t ga p ( f - q ) *
6
4
2
0
-2
-4
74
76
78
80
82
84
86
88
90
92
94
96
98
8
R e a l o u t p u t gr o w t h ( f - q )
P o t e n t ia l o u t p u t gr o w t h ( f - q ) *
6
4
2
0
-2
-4
74
76
78
80
82
84
86
88
90
92
94
96
98
8
C P I ( r a t e o f c h a n ge , f -q )
M o n e y s t o c k ( r a t e o f c h a n ge , f - q )
6
4
2
0
-2
-4
74
1)
76
78
80
82
84
86
88
90
92
94
96
98
The measurement errors are defined as the differences between the ex post figures (March 1999 vintages)
and the initial figures.
* The calculation is based on Bundesbank estimates of potential output.
24
n=0
1.03**
(0.41)
n=1
1.43***
(0.42)
n=2
1.71***
(0.60)
n=3
2.92***
(0.77)
n=4
2.37**
(1.02)
n=5
2.41**
(1.17)
n=6
2.99**
(1.39)
0.55**
(0.23)
0.30
(0.18)
0.24
(0.30)
-0.14
(0.27)
0.19
(0.38)
0.41
(0.44)
0.44
(0.46)
1.25**
(0.53)
1.04**
(0.40)
1.34**
(0.54)
1.33***
(0.41)
1.74**
(0.68)
2.08**
(0.92)
2.33**
(1.06)
0.54***
(0.19)
0.46***
(0.16)
0.49**
(0.20)
0.25*
(0.15)
0.40**
(0.18)
0.52**
(0.23)
0.61**
(0.27)
0.83***
(0.04)
0.80***
(0.05)
0.84***
(0.05)
0.85***
(0.03)
0.88***
(0.03)
0.90***
(0.03)
0.91***
(0.03)
0.94
0.95
0.95
0.96
0.95
0.95
0.95
R
0.61
0.60
0.57
0.53
0.57
0.57
0.56
SEE
0.00
0.01
0.28
0.32
0.87
0.39
0.09
JB
0.61
0.56
0.65
0.72
0.59
0.53
J-stat 0.51
***(**/*) denotes significance at the 1% (5%/10%) level. Estimation period: 1979Q1 to 1998Q4,:
estimation method: GMM; HAC-robust standard errors in parentheses; for further notes see table 1.
Ex-post series as of March 1999.
Variables: left-hand-side variable: 3-month money market rate (end-of-quarter); right -hand-side
variables: inflation gap according to cpi; level and change in the output gap with Bundesbank's
own estimates of production potential. For further details on the data see Gerberding et al (2004)
To correct for extreme outliers in the residuals, it proved necessary to include a dummy variable in
the estimations which is one in the first quarter of 1981 and zero otherwise. The dummy captures
the jump in money market rates which occurred in February 1981 when the Bundesbank replaced
its normal lombard loans by a new special lombard facility which cost 3 percentage points more.
The instrument set includes the contemporary values of inflation and the price assumption (which
were known to policy makers at the end of each quarter) as well as two lags of each explanatory
variable. Pretesting suggests that this instrument structure is sufficient.
R: adjusted coefficient of determination; SEE: standard error of the regression; J-stat: p-value of
the J-statistic on the validity of overidentifying restrictions ; JB: p-value of the Jarque Bera test of
the normality of residuals.
Table 1b: Ex-post estimates of Equation (12) with y = 0
n=0
1.91***
(0.41)
n=1
2.03***
(0.29)
n=2
2.15***
(0.33)
n=3
2.69***
(0.52)
n=4
2.73***
(0.68)
n=5
3.17***
(0.95)
n=6
3.45***
(1.08)
2.22**
(0.92)
1.18***
(0.43)
1.30***
(0.45)
1.37***
(0.45)
1.78***
(0.65)
2.20**
(0.83)
2.07**
(0.83)
0.98***
(0.36)
0.53***
(0.20)
0.48**
(0.18)
0.29**
(0.14)
0.38**
(0.15)
0.50**
(0.20)
0.60**
(0.24)
0.86***
(0.05)
0.79***
(0.06)
0.81***
(0.05)
0.85***
(0.03)
0.87***
(0.03)
0.89***
(0.03)
0.90***
(0.03)
0.89
0.93
0.94
0.96
0.95
0.94
0.95
R
0.87
0.68
0.61
0.52
0.57
0.60
0.56
SEE
0.00
0.00
0.17
0.41
0.90
0.67
0.10
JB
0.74
0.63
0.71
0.79
0.64
0.57
J-stat 0.67
***(**/*) denotes significance at the 1% (5%/10%) level. Estimation period: 1979Q1 to 1998Q4,:
estimation method: GMM; HAC-robust standard errors in parentheses; for further notes see table 1.
Ex-post series as of March 1999. For further notes see Table 1a.
25
n=0
2.17***
(0.48)
n=1
2.19***
(0.36)
n=2
2.43***
(0.33)
n=3
3.05***
(0.45)
n=4
2.64***
(0.71)
n=5
2.73***
(0.81)
n=6
3.56***
(1.07)
0.06
(0.18)
0.01
(0.14)
-0.09
(0.11)
-0.31**
(0.15)
0.00
(0.23)
0.04
(0.25)
-0.16
(0.31)
2.41***
(0.77)
1.79***
(0.53)
1.53***
(0.43)
1.72***
(0.48)
2.57***
(0.87)
3.01***
(1.11)
3.57***
(1.19)
0.98***
(0.31)
0.61***
(0.21)
0.39**
(0.16)
0.17
(0.15)
0.60**
(0.23)
0.80***
(0.30)
0.91***
(0.34)
0.84***
(0.04)
0.82***
(0.04)
0.82***
(0.04)
0.85***
(0.03)
0.89***
(0.02)
0.91***
(0.02)
0.92***
(0.02)
0.90
0.93
0.95
0.96
0.95
0.95
0.94
R
0.82
0.66
0.56
0.53
0.57
0.60
0.61
SEE
0.00
0.14
0.67
0.45
0.91
0.58
0.04
JB
0.67
0.68
0.74
0.49
0.48
0.44
J-stat 0.68
***(**/*) denotes significance at the 1% (5%/10%) level. Estimation period: 1979Q1 to 1998Q4,:
estimation method: GMM; HAC-robust standard errors in parentheses.
Variables: left-hand-side variable: 3-month money market rate (end-of-quarter); right -hand-side
variables: inflation gap according to cpi; level and change in the output gap with Bundesbank's
own estimates of production potential. For further details on the data see Gerberding et al (2004)
To correct for extreme outliers in the residuals, it proved necessary to include a dummy variable in
the estimations which is one in the first quarter of 1981 and zero otherwise. The dummy captures
the jump in money market rates which occurred in February 1981 when the Bundesbank replaced
its normal lombard loans by a new special lombard facility which cost 3 percentage points more.
The instrument set includes the contemporary values of inflation and the price assumption (which
were known to policy makers at the end of each quarter) as well as two lags of each explanatory
variable. Pretesting suggests that this instrument structure is sufficient.
R: adjusted coefficient of determination; SEE: standard error of the regression; J-stat: p-value of
the J-statistic on the validity of overidentifying restrictions ; JB: p-value of the Jarque Bera test of
the normality of residuals.
Table 2b: Real-time estimates of Equation (12) with y = 0
n=0
2.30***
(0.33)
n=1
2.21***
(0.25)
n=2
2.26***
(0.25)
n=3
2.57***
(0.39)
n=4
2.64***
(0.58)
n=5
2.81***
(0.71)
n=6
3.07***
(0.88)
2.57***
(0.71)
1.82***
(0.49)
1.47***
(0.39)
1.74***
(0.49)
2.57***
(0.86)
3.07***
(1.09)
3.44***
(1.25)
1.05***
(0.30)
0.61***
(0.21)
0.39***
(0.14)
0.30**
(0.14)
0.60**
(0.23)
0.78**
(0.30)
1.04***
(0.39)
0.85***
(0.04)
0.82***
(0.04)
0.82***
(0.04)
0.86***
(0.03)
0.89***
(0.02)
0.91***
(0.02)
0.92***
(0.02)
0.90
0.93
0.95
0.96
0.95
0.95
0.94
R
0.82
0.66
0.56
0.52
0.56
0.59
0.60
SEE
0.00
0.15
0.78
0.49
0.91
0.61
0.06
JB
0.75
0.77
0.77
0.58
0.57
0.53
J-stat 0.76
***(**/*) denotes significance at the 1% (5%/10%) level. Estimation period: 1979Q1 to 1998Q4,:
estimation method: GMM; HAC-robust standard errors in parentheses; for further notes see table
2a.
26
n=0
2.12***
(0.33)
n=1
2.15***
(0.28)
n=4
2.35***
(0.43)
n=5
2.29***
(0.35)
n=6
2.51***
(0.35)
2.53***
(0.69)
1.74***
(0.53)
1.49*
(0.81)
1.14*
(0.67)
0.63
(0.46)
0.93***
(0.30)
0.56**
(0.22)
0.37*
(0.22)
0.33*
(0.18)
0.25**
(0.11)
'
0.84***
(0.04)
0.81***
(0.05)
0.85***
(0.05)
0.83***
(0.05)
0.80***
(0.06)
0.12
(0.09)
0.12
(0.12)
0.12
(0.11)
0.14
(0.10)
0.13
(0.12)
0.20
0.24
0.47
0.46
0.44
R
0.67
0.65
0.55
0.56
0.57
SEE
0.06
0.34
0.86
0.27
0.03
JB
0.68
0.62
0.47
0.41
0.73
J-stat
***(**/*) denotes significance at the 1% (5%/10%) level. Estimation period: 1979Q1 to
1998Q4,: estimation method: GMM; HAC-robust standard errors in parentheses.
Variables: left-hand-side variable: 3-month money market rate (end-of-quarter); right hand-side variables: inflation gap according to cpi; growth of the output gap with
Bundesbank's own estimates of production potential; money gap: annual growth rates
relative to target. For further details on the data see Gerberding et al (2004)
To correct for extreme outliers in the residuals, it proved necessary to include a dummy
variable in the estimations which is one in the first quarter of 1981 and zero otherwise.
The dummy captures the jump in money market rates which occurred in February 1981
when the Bundesbank replaced its normal lombard loans by a new special lombard
facility which cost 3 percentage points more. The instrument set includes the
contemporary values of inflation, the monetary target and the price assumption (which
were known to policy makers at the end of each quarter) as well as two lags of each
explanatory variable (except the monetary target). Pretesting suggests that this instrument
structure is sufficient.
R: adjusted coefficient of determination; SEE: standard error of the regression; J-stat: pvalue of the J-statistic on the validity of overidentifying restrictions ; JB: p-value of the
Jarque Bera test of the normality of residuals.
27
2006
2006
2006
Dieter Nautz
Christian J. Offermanns
lvarez, Dhyne, Hoeberichts
Kwapil, Le Bihan, Lnnemann
Martins, Sabbatini, Stahl
Vermeulen, Vilmunen
Robert Jckle
2006
2006
2006
2006
2006
2006
28
Wolfgang Lemke
Theofanis Archontakis
Antonio Bassanetti
Jrg Dpke, Roberto Torrini
Roberta Zizza
10
2006
11
2006
12
13
2006
2006
J. Dpke, D. Hartmann
C. Pierdzioch
Kai Christoffel
Tobias Linzert
Thomas A. Knetsch
Guntram B. Wolff
14
2006
15
2006
Johannes Hoffmann
Jeong-Ryeol Kurz-Kim
16
2006
17
2006
Kai Christoffel
Keith Kster
Tobias Linzert
18
2006
Boris Hofmann
19
2006
Kerstin Bernoth
Guntram B. Wolff
20
2006
Clemens Fuest
Thomas Hemmelgarn
Fred Ramb
29
21
2006
22
2006
Real-time forecasting and political stock market Martin Bohl, Jrg Dpke
anomalies: evidence for the U.S.
Christian Pierdzioch
23
2006
Christoph Fischer
Daniel Porath
24
2006
Sascha O. Becker
Marc-Andreas Mndler
25
2006
Badi H. Baltagi
26
2006
Atsushi Inoue
Lutz Kilian
Fatma Burcu Kiraz
27
2006
H. Pesaran, D. Pettenuzzo
A. Timmermann
28
2006
Kurt F. Lewis
Charles H. Whiteman
29
2006
30
2006
31
2006
Sandra Eickmeier
30
32
2006
33
2006
Christian Schumacher
Jrg Breitung
34
2006
35
2006
Mark Hallerberg
Guntram B. Wolff
36
2006
C. Moser
T. Nestmann, M. Wedow
37
2006
Kerstin Stahn
Thomas A. Knetsch
Hans-Eggert Reimers
Claudia Stirbck
38
39
2006
2006
40
2006
C. M. Buch, J. Dpke
H. Strotmann
41
2006
Kirsten H. Heppke-Falk
Jrn Tenhofen
Guntram B. Wolff
31
42
43
44
2006
2006
2006
Sandra Eickmeier
Christina Ziegler
Regionalwhrungen in Deutschland
Lokale Konkurrenz fr den Euro?
Gerhard Rsl
Nikolaus Bartzsch
45
2006
Rainer Frey
Katrin Hussinger
46
2006
Harald Stahl
Stephanus Arz
47
2006
48
2006
01
2007
Sascha O. Becker
Marc-Andreas Mndler
02
2007
03
2007
04
2007
Nikolai Sthler
32
05
2007
06
2007
C. Gerberding
F. Seitz, A. Worms
33
2006
J. Dpke, D. Hartmann
C. Pierdzioch
02
2006
03
2006
Klaus Dllmann
04
2006
Claudia M. Buch
Andrea Schertler
Natalja von Westernhagen
05
2006
Evelyn Hayden
Daniel Porath
Natalja von Westernhagen
06
2006
Christian Merkl
Stphanie Stolz
07
2006
W. Gerke, F. Mager
T. Reinschmidt
C. Schmieder
08
2006
Michael Koetter
09
2006
Klaus Dllmann
Nancy Masschelein
10
2006
E. Fiorentino
A. Karmann, M. Koetter
11
2006
F. Fecht, H. P. Grner
34
12
2006
Falko Fecht
Hendrik Hakenes
01
2007
Michael B. Gordy
Eva Ltkebohmert
02
2007
Michael Koetter
Daniel Porath
03
2007
Thomas Kick
Michael Koetter
04
2007
C. E. Bannier
F. Fecht, M. Tyrell
05
2007
A. Behr, A. Kamp
C. Memmel, A. Pfingsten
Christoph Memmel
Peter Raupach
06
2007
35
The Deutsche Bundesbank in Frankfurt is looking for a visiting researcher. Among others
under certain conditions visiting researchers have access to a wide range of data in the
Bundesbank. They include micro data on firms and banks not available in the public.
Visitors should prepare a research project during their stay at the Bundesbank. Candidates
must hold a Ph D and be engaged in the field of either macroeconomics and monetary
economics, financial markets or international economics. Proposed research projects
should be from these fields. The visiting term will be from 3 to 6 months. Salary is
commensurate with experience.
Applicants are requested to send a CV, copies of recent papers, letters of reference and a
proposal for a research project to:
Deutsche Bundesbank
Personalabteilung
Wilhelm-Epstein-Str. 14
60431 Frankfurt
GERMANY
37