Download Franz Seitz - OTH Amberg

Are the Effects of Monetary Policy Asymmetric?
Experience of the Bundesbank
and Lessons for the European Central Bank
January 2000
Franz Seitz
University of Applied Sciences Amberg-Weiden
Hetzenrichter Weg 15
D-92637 Weiden
Germany
Tel.: (+49) 961-382-172
Fax: (+49) 961-382-110
e-mail: [email protected]
Contents:
1. Introduction
2. Theoretical reasons for asymmetric effects of monetary policy in Germany
3. The strategy of monetary targeting and the reaction function of the Bundesbank
4. Empirical analysis of asymmetric effects of monetary policy shocks
4.1 The data
4.2 The econometric approach
4.3 Estimation results
5. Summary and conclusions
Appendix: The reaction function

I thank J. Clostermann, J. Kakes, H.-E. Reimers, S. Schich, K.-H. Tödter and K. Wesche for valuable
comments. This is a revised draft of a paper presented under different titles at the annual meeting of the European
Economic Association 1998 in Berlin, the HARLEM-Seminar 1998 in Berching, the Deutsche Bundesbank 1998,
the Ifo Institute for Economic Research and the University of Vienna.
Are the Effects of Monetary Policy Asymmetric?
Experience of the Bundesbank and Lessons for the European Central Bank
1. Introduction
Are there real effects of monetary policy? Or do monetary policy actions predominantly alter
prices? For a long time economists care about these queations. But no clear cut answer has
evolved. There is only a consensus view, if at all, that in the long run monetary policy
primarily affects the evolution of prices (McCandless/Weber 1995). Many economists also
agree that the real effects of monetary policy under rational expectations are mainly due to
unanticipated shocks (Barro/Rush 1980).1 But Mishkin (1982) found that anticipated
monetary policy does seem to matter, too.
While the analysis of monetary policy shocks usually proceeds in assuming a symmetric
response of prices and quantities, recent theoretical work (Ball/Mankiw 1994; Weise 1999)
and empirical studies for the USA (DeLong/Summers 1988, Cover 1992, Morgan 1993,
Ammer/Brunner 1995, Rhee/Rich 1995, Garibaldi 1997) have shown that the effects of
money-supply shocks on output are asymmetric.2 This asymmetry means that monetary
contractions generate or accelerate recessions while monetary expansions most of all increase
prices without generating an upswing. Or, to put it differently: Tight monetary policy slows
the economy more than easy monetary policy accelerates it. Similarly, for a panel of 18
European countries and 38 countries, respectively, Karras (1996a, 1996b) presents evidence
for asymmetries in an international context. Moreover, he does not only find asymmetric
effects on GDP but also on consumption and investment. In his analysis as well as in
Ammer/Brunner (1995) there is no evidence in favour of cyclical asymmetries. This means
that the "apparent" asymmetry is not due to the fact that monetary policy is more effective
during expansions than during recessions (see for a contrary result Weise 1999 and in the case
of Germany Kakes 1998). Furthermore, in the empirical analysis of Caballero/Engel (1992),
1
See also the panel discussion "Is there a Practical Core of Macroeconomics That we Should All Believe"
between O. Blanchard, A. Blinder, M. Eichenbaum, R. Solow and J. Taylor at the annual meeting of the
American Economic Association 1997. The remarks are published in American Economic Review, 1997, Vol.
87, pp. 230-246.
2
The idea of asymmetric real effects of monetary policy has a long tradition and was usually motivated with
downward rigidity of nominal wages and prices (see the historical overview in Morgan 1993). Recent research
puts this asymmetry on a more solid theoretical and empirical basis. For the USA Froyen et al. (1997) even
present results of asymmetric effects of political pressures on the Fed.
1
Rhee/Rich (1995) and Buckle/Carlson (1998) monetary policy displays asymmetric effects
which are exacerbated by increases in average inflation. And Macklem et al. (1996) show that
both anticipated and unanticipated monetary shocks may generate asymmetries. They also find
no robust evidence of asymmetric output effects of factors other than domestic monetary
policy shocks in Canada.3 For the case of Switzerland Lenz (1997) finds that only if net
exports enter the money-supply equation as an explanatory variable is there evidence for
asymmetric effects. Otherwise monetary shocks have no real effects at all.
In trying to identify asymmetric effects it is necessary to specify a central bank reaction
function to assess the stance of monetary policy. Many of the above mentioned papers identify
this stance using monetary aggregates and do not take due account of the concrete monetary
framework in the countries investigated. This may be misleading because not all changes in
monetary growth reflect changes in policy. Monetary developments are also largely demanddriven. Therefore monetary aggregates are only an imperfect measure of the stance of
monetary policy. It is better to use an indicator which is more under the direct control of the
central bank, e.g. the day-to-day money market rate or the monetary base.4
Besides the panel analysis of Karras (1996a; 1996b) there are only two papers on asymmetric
effects of monetary policy for Germany.5 De Bondt (1998), using the methodology of Morgan
(1993), finds no evidence of asymmetries on industrial production. Borio/Fritz (1995)
investigate whether there exists a possible asymmetric effect induced by an asymmetric
passthrough of policy and money market rates to bank lending rates. The only country in their
study where this is the case is Germany. There, the response is relatively fast with respect to
increases in interest rates. But there are several papers on the reaction function of the
Bundesbank without taking into account possible asymmetric effects (e.g. Bernanke/Mihov
1997, Chadha/Janssen 1997, Schächter 1998).6 Unfortunately, these studies do not take due
account of the strategy of monetary targeting applied by the Bundesbank. This means that the
adequate assessment of monetary developments is not possible without taking into account the
inflationary prospects or the inflation forecast of the Bundesbank. This paper tries to correct
3
One other reason for output asymmetries can be found in Acemoglu/Scott (1997).
For the USA Bernanke/Blinder (1992) for instance advocate using the federal funds rate. Macklem et al. (1996)
use the term spread as a measure for the stance of monetary policy. However, in Germany, this spread may also
change as a result of a multitude of (exogenous) influences ( Seitz 1996).
5
Karras (1996b) checks his panel results for each of his countries analyzed and got no hints of asymmetries for
Germany. But, as he mentioned, this estimation yields rather imprecise coefficients.
4
2
this insufficiency. The final new aspect in the present study is the consideration of possible
asymmetric effects not only on GDP (or GNP) but also on private consumption, investment,
exports and the unemployment rate. Besides aggregate output effects only two papers deal
with other real variables: For the US Karras (1996a) finds that asymmetry holds for both
consumption and investment, too, and Garibaldi (1997) gets evidence of asymmetric effects of
monetary policy on job destruction and job creation. I find that for Germany asymmetry is not
a general phenomenon, but mainly affects the German economy via its influence on exports.
The rest of the paper is structured as follows. Section 2 presents some theoretical arguments
for asymmetric effects of monetary policy and tries to shortly assess their validity for
Germany. Part 3 introduces the concept of monetary targeting and its implications for the
reaction function of the Bundesbank. Section 4 discusses the methodology employed and
presents the empirical analysis. To test for asymmetric reactions I use different regression
approaches. The task is to first identify the stance of monetary policy and then to test for
asymmetries. It is analyzed whether these asymmetries govern the effects of monetary policy
on GDP, consumption, investment and unemployment. Section 4 summarizes and draws some
tentative conclusions for the monetary policy of the European Central Bank (ECB).
2. Theoretical reasons for asymmetric effects of monetary policy in Germany
Monetary asymmetries are due to certain rigidities or frictions. The more recent literature
suggests three reasons why easy policy may be less effective than tight policy (Morgan 1993,
pp. 22f.).
One explanation why monetary policy might have asymmetric effects is that expectations of
economic agents are not symmetrically distributed over the business cycle. For example,
consumers and firms may be more pessimistic during recessions than they are optimistic
during booms. Or the outlook of economic agents may simply matter more during recessions
or downturns than during upswings. Although these psychological explanations may not be
fully convincing from an economist's point of view, the results of the business cycle survey of
the German Ifo-Institute of Economic Research and of consumers' surveys present some
justification for them. The different lengths of upswings and downswings (Oppenländer,
6
There is some empirical evidence that the Bundesbank reacts in an asymmetrical way, see Clarida/Gertler
(1997), Schächter (1998).
3
1995) and the non-stationarity of the output gap in Germany also give hints for the relevance
of this psychological factor. This line of reasoning may also be responsible for cyclical
asymmetries (see Kakes 1998, Weise 1999).
The second cause for asymmetries refers to imperfections of credit markets and credit
rationing. This line of reasoning focuses on the special role of banks in the transmission of
monetary policy. It states that in the course of a tight monetary policy banks reduce their credit
supply because of the informational asymmetries adverse selection and moral hazard. This
credit channel works over and above the traditional channels of monetary transmission
(Bernanke 1993). The reactions are due either to negative effects of a restrictive monetary
policy on the net worth or goodwill of business firms (balance sheet channel) or the reliance
on bank loans and the lack of alternative financial resources, especially for small firms and
households (bank lending channel). This tightens the credit constraint on some borrowers and
limits their spending. Hence credit supply, demand and output decline more sharply over the
short term than solely due to higher interest rates. On the other hand, easy policy in a
recession relaxes credit constraints by lowering interest rates. But this will not necessarily
boost borrowing and spending if a slowing economy has reduced the demand for credit. Thus
if the credit constraint is no longer binding before policy is eased, relaxing the constraint will
not augment easy policy.7 A number of studies have confirmed the existence of the credit
channel for the USA. By contrast, only few studies to date are available for Germany. And the
empirical evidence presented is not unequivocally in favour or against the credit channel or
the relevance of credit constraints (see e.g. Stöß 1996, Gonzalez Minguez 1997,
Guender/Moersch 1997 and Kremp/Stöß/Gerdesmeier 1999).
A third theoretical reason for asymmetries refers to the relative rigidity of prices. If prices are
less flexible downward than upward a restrictive monetary policy will reduce output with little
change in prices, while easy policy will cause prices to rise with little change in output. The
usual theoretical reason given for this rigidity are menu costs and their interdependencies with
the trend rate of inflation (Ball/Mankiw 1994). Caballero/Engel (1992) show that these price
rigidities are an international phenomenon. But the results of a recent business survey cast
doubt on an asymmetric price setting behaviour of German firms (Köhler 1996).
7
Further (credit) asymmetries may be due to different firm sizes (Oliner/Rudebusch 1996), the relative costs of
external finance to different firms (Kim/Ni/Ratti 1998) and the regional distribution of firms (Carlino/DeFina
1996, Hayo/Uhlenbrock 1999). Besides finance constraints capacity constraints may also be responsible for
asymmetric effects of aggregate demand on real economic activity.
4
There is one further and neglected explanation of possible asymmetries, the liquidity trap.
This concept, introduced by J.R. Hicks describes a situation in which conventional monetary
policies have become impotent, because nominal interest rates are near zero. This is the case
because bonds and money are viewed as perfect substitutes.8 Maybe the current situation in
Japan suits this situation well. But Germany only moved in this direction, but has never
reached such low interest rate levels (the 3-month money market rate at the end of 1998 was
about 3.5%).
Overall there is only indirect and ambiguous empirical evidence for the validity of these
arguments in the case of Germany. Therefore, a direct empirical test of monetary stimuli and
reactions of output as well as of other real variables is the task of the following sections. But
before turning to the empirical implementation we have to explain the monetary policy
strategy of the Bundesbank and its implications for the reaction function.
3. The strategy of monetary targeting and the reaction function of the Bundesbank
In trying to understand the strategy of monetary targeting it is useful to take as a starting point
the well-known equation of exchange. This equation allows us to derive some definitions for
the actual price level p, the equilibrium price level p* and the price target pT (in each case in
logarithms):
(1)
p  m  yr  v
(2)
p*  m  yr * v *
(3)
p T  m T  yr * v *
Equilibrium values are indexed by an asterisk (*), target values by a "T"; m is the money stock
(e.g. in the German case the logarithm of M3) and v is the velocity of circulation. The
equilibrium price level p* ("p-star") is defined as money per unit of real potential
(equilibrium) output yr* at the equilibrium level of velocity v* (see equation (2)). This
indicator intends to measure the price level which would occur at the actual money holdings if
8
Krugman (1998) makes the point that the liquidity trap involves a kind of credibility problem. Expansionary
monetary policy does not work just because the public does not expect it to be sustained, i.e. there are no
5
production and velocity were at their equilibrium levels. In contrast, the price target pT in (3)
refers to the money-supply target mT per unit of real potential (equilibrium) output yr* at the
equilibrium level of velocity v*. Combining (1) and (2) yields the so called price gap (p*-p)
which is composed of the output gap (yr-yr*) and the liquidity gap (v*-v):9
(4)
p *  p  ( yr  yr*)  (v *  v)
The price gap indicates inflationary pressures if the rate of capacity utilization is high (the
output gap is positive) and/or if there is a monetary overhang, i.e. if velocity is below
(liquidity holding is above) its long-run equilibrium level. Combining (2) and (3) shows that
deviations between the equilibrium price level p* and the price target pT correspond with
deviations between the money stock m and the monetary target mT.
(5)
( p *  p T )  (m  mT )
A "naive" strategy of monetary targeting would only concentrate on deviations from the
money supply target (5) and orientate its monetary policy instruments accordingly. But growth
targets for a monetary aggregate are finally used to stabilize price developments on a low
level. The derivation of the monetary target is based on an implicit or explicit "inflation
target". According to the above concept this refers to the GDP-deflator. If this "inflation
target" is not met a central bank which follows a strategy of monetary targeting has to react to
this even if the development of the money stock is on track.
(1) and (3) together with (4) yield
(6)
p  p T  (m  m T )  ( p *  p)
Equation (6) shows that deviations from the price target are made up of deviations from the
money supply target less the price gap. If the final target of monetary policy is price stability,
the stance of monetary policy has to be adjusted as long as the price target is violated. Only if
prices are in equibibrium (the price gap is zero) are deviations from the price target equivalent
inflation expectations.
9
For the p-star-concept and the price gap in the case of Germany see Tödter/Reimers (1994) and
Clostermann/Seitz (1999).
6
to deviations from the monetary target. This has to be taken into account when constructing a
reaction function.
The starting-point of the transmission of monetary stimuli is the day-to-day money market.
The corresponding money market rate (iday) is under the most direct control of central banks.
Therefore the thing to do is to relate the overnight money market rate to deviations in the
"inflation target". This is done in (7).
(7)
iday  f ( 4 ( p  p T ))  f ( pgap) ,
where pgap  4(p-pT) and 4p = pt-pt-4 is the inflation rate. The annual inflation rate is used
because the Bundesbank set annual monetary targets.
As mentionend above, the "inflation target" underlying the monetary targets refers to the
GDP-deflator. Nonetheless the CPI is used in this paper. It is in the center of the public
interest and obviously the Bundesbank also concentrated on this inflation measure (Deutsche
Bundesbank 1997a, pp. 82-92; 1997b, p. 10). Equation (7) is the basis for the reaction
function to be estimated in the following section. A theoretical rationale is given in the
appendix. The relationship is illustrated in figure 1 which shows the nearly parallel movement
of iday and pgap.
4. Empirical analysis of asymmetric effects of monetary policy shocks
4.1 The data
The data used are taken from the database of the Deutsche Bundesbank. They are seasonally
unadjusted. The sample period covers the post Bretton-Woods era from the first quarter of
1974 to the fourth quarter of 1998. Until the end of 1990 the data refer to West Germany.
From the first quarter of 1991 data for unified Germany were used. The day-to-day money
market rate (iday) is used as the operational variable of the Bundesbank. The transmission
process of monetary policy begins on the day-to day money market and the Bundesbank
exerted the fastest and most direct effect on this interest rate via repurchase operations. The
purpose then is to identify asymmetric effects of monetary-policy shocks on four variables: the
growth rates of real GDP (yr), real private consumption (cr), real investment in machinery and
equipment (ir) and real exports (exr), respectively, and the change in the unemployment rate
7
(u). All variables, except the interest rate and the unemployment variable are in logarithms.
The difference operator  relates to quarter-by-quarter changes, i.e. first differences. Therefore
three seasonal dummies s1, s2 und s3 enter the regressions. The effect of German unification
on certain time series is captured by a further dummy variable (dgu). It is one in the first
quarter of 1991, zero otherwise and enters the equations for the real variables because allGerman data are used from the first quarter of 1991.10
4.2 The econometric approach
The problem of identifying asymmetric effects of monetary policy can be divided into two
parts.
In the first part the stance of monetary policy is determined via (unexpected or unanticipated)
changes in the overnight rate. To do this I use the monetary policy reaction function implied
by monetary targeting and derived in section 3 (see equation (7)). The monetary targets are not
treated as a separate final goal (as e.g. in Bernanke/Mihov 1997 or Chadha/Janssen 1997) but
as a means to control inflation. Therefore, the reaction function does not consist separately of
(contemporary and lagged) growth rates of the money stock and inflation rates (and other
variables). Instead this paper applies the concept of a price gap (pgap).11 This framework has
been ignored in nearly all studies dealing with reaction functions of German monetary policy
(see e.g. Clarida/Gertler 1997). This may also be the adequate model for the monetary policy
of the European Central bank. Its interpretation of the new two-pillar strategy is very similar
to that of the "pragmatic" strategy of monetary targeting of the Bundesbank (European Central
Bank 1999).
In nearly all the papers which do not recognize these considerations no significant influence of
the money-supply growth on the setting of interest rates by the Bundesbank can be detected.12
As we will see this is not the case when the behaviour of the Bundesbank is modelled more
accurately.
10
All variables included in the estimations are I(0). The results of the stationarity tests are available from the
author upon request.
11
For a comprehensive overview of the monetary targeting strategy of the Bundesbank see Issing (1994).
12
Bernanke/Mihov (1997, p. 1049) e.g. jump to the conclusion that "the Bundesbank is better characterized as an
inflation targeter than as a money targeter".
8
The general form of the overnight rate reads13
(8)
M
N
j1
j 0
iday t      j  iday t  j    j  pgap t  j   t ,
where t is a white-noise error term. The variable pgap measures the difference between the
annual inflation rate, measured with the consumer price index, and the inflation target of the
Bundesbank which was (explicitly or implicitly) inherent in the money-supply targets (see
(7)). The deviations from the inflation target consist of deviations from the money growth
target and the difference between the growth rates of the equilibrium price level and the actual
price level. The latter is made up of the output gap and the liquidity gap (see (4)). Up to four
lags of pgap are considered. A positive unanticipated shock in monetary policy is defined as
(9)
easy t  min(0,  t ) ,
a negative shock as
(10)
tight t  max( 0,  t ) .
Thus, the variations in the day-to-day money market rate not explained by the variables taken
into account in (8), the residuals , are used to identify the stance of monetary policy. The
positive residuals represent tight policy, because the residuals measure how much the
overnight rate exceeds the level predicted by (8). And negative residuals (innovations)
represent an unanticipated easy policy because the day-to-day rate is lower than the predicted
level. This measure of the stance of monetary policy forms the basis for the second part of the
estimation procedure.
In this second stage the (logarithmic) growth rates of GDP, consumption, investment and
exportsas well as the change in the unemployment rate are regressed on the lagged
endogenous variable and the current and lagged positive and negative residuals of the first
stage. Moreover, the unification dummy dgu, the seasonal dummies and in one case a time
trend enter the equation.
13
A theoretical justification of this reaction function is given in the appendix.
9
(11)
O
Q
R
i 1
i 0
i 0
zt  a   bi  zt i   ci  tight t i   d i  easyt i
e  dgut  f 1  s1  f 2  s2  f 3  s3   t
with: zt  yr,cr,ir,exr,u
t  N(0,²)
Theoretically one would expect ci<0, di<0 for zt  yr,cr,ir,exr and ci>0, di>0 for zt  u. The
lags M, N, O, Q and R are reduced successively according to the usual significance levels
using the general-to-specific-methodology starting with a maximum lag number of 4. The
cumulative impact of monetary policy on the real variables is measured by the sum of the
significant coefficients on "tight" and "easy". There are asymmetric effects of monetary policy
if either only "tight" or "easy" exert a significant influence or if ci - di  0. These
hypotheses are tested using Wald tests. Further influences on the four real variables like the
effects of fiscal policy or exchange rates are not considered to concentrate on the effects of
monetary policy. Partly they are already incorporated in the first stage reaction regression.
Equations (8) and (11) are estimated by different econometric methods. First, a two-stage
OLS-regression procedure is used. This is common to all studies on asymmetric effects of
monetary policy and means that both equations are estimated separately by OLS. In the first
stage equation (8) is estimated and the residuals recovered; in the second stage "tight" and
"easy" are calculated and used to estimate (11). But Mishkin (1982) already pointed out the
advantages of a simultaneous estimation of the system (8) and (11). Thus, the second method
is the simultaneous estimation of (8) and (11). As the error terms in (8) and (11) may be
correlated a SUR estimator is also applied as a third alternative. Fortunately, the overall
results do not depend on the particular method applied.14 Therefore only the results of the
two-stage regression are shown. The results of the two other methods are available from the
author upon request.
14
This is common to all papers which use more than one estimation method.
10
4.3 Estimation results
The results of the first stage estimation are summarized in the following equation (12). The
numbers in brackets below the coefficients are the absolute t-values.15
(12)
iday t  0.75 1.16 iday t 1  0.30 iday t 3  0.14 pgapt   t
( 3.4)
( 20.0)
( 5.9)
( 2.2)
R² = 0.94; SE = 0.56; LM(1) = 0.53; LM(4) = 0.45;
ARCH(1) = 0.51; ARCH(4) = 0.14
The coefficients have the theoretically expected signs and are significant. More than 90% of
the variance of the overnight rate is explained. There are no hints of first or fourth order
autocorrelation (see the p-values of the LM-tests) or heteroskedasticity (see the p-values of the
ARCH-tests). The coefficients of the lagged day-to-day rate sum to +0.86. This means that the
Bundesbank prefered steady and smooth operating procedures to volatile money market rates.
This corresponds to the medium-term orientation of German monetary policy. As equation
(12) shows the Bundesbank reacts contemporary to deviations from the inflation target. If
price developments exceed the inflation target (or alternatively if money growth is too
excessive compared to the money-supply target) by one percentage point the overnight rate
rises by 0.14 percentage points in the short run to stem inflationary pressures. In the long run
the day-to-day rate increases by one percentage point. If instead of the price gap only
deviations from the money growth target are used in (12) the coefficient would not be
significantly different from zero. This would yield the conclusion of the other studies cited
above that the Bundesbank is no money targeter.
The residuals of this regression, divided into expansionary ("easy") and restrictive ("tight")
phases are illustrated in figure 2. The figure clearly shows the tight stance of monetary policy
at the end of the seventies and the beginning of the eighties followed by an easier orientation.
This was the time of the second oil price shock during which German monetary policy was
characterized by a high degree of uncertainty. The reputation of the Bundesbank at that time
was very low. Interest rates were on a rising trend at the end of the seventies (see figure 1).
This rising trend continued until the end of 1981, but was interrupted for a short time in the
15
Shortening the sample to 1988:1 - 1998:4 leads to only minor changes of the results. There were some hints
that the Bundesbank acted more in a forward-looking manner.
11
second half of 1980. Figure 2 also shows that in the course of German unification only
relatively moderate (unanticipated) monetary policy shocks occured.
The residuals of equation (12) enter the second stage, equation (11), as a measure of the stance
of the monetary policy shocks. The results are summarized in the following table. The
coefficients of the seasonal dummies and the time trend are not listed separately. Furthermore
only the significant parameters of "easy" and "tight" are shown. The concrete lags appear in
the footnotes to the table. To derive consistent estimates of the covariance matrix in the
presence of both heteroskedasticity and autocorrelation the Newey-West procedure is applied
(Newey/West 1987). The row "Stab" indicates whether in the nineties the CUSUM test
statistic is suggestive of parameter stability (stable) or instability (unstable) at the 5 percent
critical level. In the unstable case the exact period is indicated. The last three rows show the pvalues of Wald tests of the respective restrictions of symmetric or asymmetric effects of
monetary policy in Germany.
All equations have satisfactory statistical properties. With the exception of the equation for
real exports the coefficient of determination is about 0.9, i.e. 90 % of the variance of the
respective real variable is explained. The GDP equation, the consumption equation and the
unemployment equation are reasonably well specified, given the p-values of the diagnostic
statistics and the stability test. But there are problems with the assumption of normality and
hints of ARCH(1)-effects in the investment equation (the fourth column) as indicated by the
p-values of the Jarque-Bera (JB) and ARCH test . Furthermore, this relation shows hints of
instability in 1993.
There is unambiguous evidence of monetary asymmetries in the case of GDP, exports and the
unemployment rate:16 Restrictive monetary measures have a negative effect on economic
activity while easy monetary policy does not stimulate economic activity. According to our
analysis these effects run via asymmetries on exports. This is in contrast to the result in Karras
(1996b) and de Bondt (1998) who find no asymmetries in the case of Germany. The former
only used annual data, the money stock as monetary instrument and relatively few degrees of
freedom. And the latter does not incorporate adequately the strategy of monetary targeting. He
simply regresses a short-term interest rate on current and lagged values of production growth
and inflation.
12
Table: The effects of tight and easy monetary policy shocks
z
a
b1
yr1)
cr1)
ir1)
exr1)
u1)
2.68 (7.2)
5.64 (10.5)
13.91 (7.2)
7.74 (12.4)
0.04 (0.9)
-0.17 (1.8)
(-)
(-)
(-)
0.29 (4.2)
b3
(-)
(-)
(-)
0.15 (1.6)
(-)
b4
0.33 (5.7)
0.19 (3.0)
0.33 (4.4)
(-)
-0.38 (4.1)
ci
-0.91 (2.8)6)
-1.13 (3.4)6)
-2.52 (2.2)4)
-1.24 (1.9)6)
0.17 (1.9)4)
di
e
(-)
-1.01 (2.0)2)
-2.22 (2.7)5)
(-)
(-)
8.81 (25.3)
29.93 (73.9)
21.56 (22.1)
-5.51 (6.2)
(-)
test statistics
0.91
R²
0.96
0.97
0.65
0.86
SE
1.22
1.41
3.49
2.96
0.27
ARCH(1)
0.17
0.37
0.01
0.80
0.94
ARCH(4)
0.63
0.41
0.14
0.76
0.72
JB
0.64
0.61
0.00
0.30
0.91
Stab
stable
stable
93.1-93.4
stable
stable
ci=0
0.00
0.00
0.03
0.06
0.05
di=0
-
0.05
0.01
-
-
ci=di
-
0.85
0.85
-
-
(-): no significant influence; -: not relevant.
1) Newey-West-correction.
2) only contemporary.
3) only lag 1.
4) only lag 2.
5) only lag 3.
6) only lag 4.
An unanticipated increase of one percentage point in the money market rate (in the sense
described above) causes GDP and export growth to decline by 0.9 and 1.2 percentages,
respectively, and the unemployment rate to increase by about 0.19 percentage points. Looking
at the regressions for investment and private consumption, the third and fourth column of the
table, there are statistically significant negative individual effects of "tight" and "easy". But
according to Wald tests the hypothesis that these coefficients are equal in magnitude cannot be
rejected which means that there are no asymmetric effects. Thus, the most volatile part of
GDP, i.e. investment, and the major component of GDP, private consumption, may accurately
be characterized by the traditional symmetric effects of monetary policy shocks. But there are
some minor asymmetries in consumption and investment due to different time patterns of easy
and tight monetary policies. The conclusions of Karras (1996a) within a panel analysis that the
16
Hess/Iwata (1997) find that West German GDP displays no asymmetric persistence.
13
asymmetries in output govern the effects of money-supply shocks on consumption and
investment as well can therefore not be confirmed.
5. Summary and conclusions
The present paper has examined whether there were asymmetric real effects of monetary
policy in Germany. Compared to the previous literature it has several new features. It is the
first one dealing with the situation in Germany on an individual basis with quarterly data and
sufficient degrees of freedom. The second novelty is the specification of the reaction function
of the Bundesbank using the price gap concept inherent in the monetary targeting framework.
To capture the stance of monetary policy the day-to-day money market rate is a better measure
than monetary aggregates which were used in many other studies. Finally asymmetries are
tested for five real variables: GDP, private consumption, investment in equipment and
machinery, exports and the unemployment rate.
It has been shown that the empirical evidence for asymmetric effects of the monetary policy of
the Bundesbank is not as unambiguous as in the case of the USA. Nevertheless the detected
asymmetries point to a careful implementation of monetary policy. Negative money-supply
shocks have a significant effect on some real variables (see the results for GDP, exports and
the unemployment rate) while positive money-supply shocks predominantly increase prices.17
Consequently, an easy policy to generate an economic upswing is not an adequate policy if
overall output and the unemployment rate should be influenced. Such a policy might even be
counter-productive in that it fuels inflation which necessitates a more severe monetary
contraction later on. The policy of the Bundesbank in the last years which has often been
characterized as a policy of "very small steps" seems to be an adequate reaction to these facts.
The above analysis may be extended in several ways. First, the reaction function of the
Bundesbank might be modified to incorporate a forward looking manner, e.g. not only lagged
but also future values of pgap may appear in (8).18 Alternatively, the reactions of the
Bundesbank might also be tested for an asymmetric behaviour, i.e. whether the reactions of
the Bundesbank to deviations above the inflation target differ from those below the target
(Clarida/Gertler 1996, Clarida/Gali/Gertler 1998, Schächter 1998). Second, one might also
17
This is only indirect reasoning, no direct test of asymmetries in a money supply-price system.
14
look at the effects of anticipated monetary policy shocks. This could be done by including the
fitted values of the overnight rate in the equations for the real variables. Third, further real
variables might be examined, e.g. government expenditures or other components of
investment expenditures (investment in construction or stocks). Fourth, the money-pricesystem could be analyzed directly instead of inferring the price effects of monetary policy
indirectly from the real effects. And fifth, further regressors might be included in the
explanation of the real variables, e.g. the exchange rate or import prices.
What conclusions can be drawn for the policy of the Eurosystem? Under the Maastricht Treaty
the European Central Bank was set up on the lines of the Bundesbank. The Eurosystem’s
strategy consists of a quantitative definition of price stability and two further pillars: The first
pillar assigns money a prominent role. Therefore the Governing Council of the European
Central Bank has decidend to announce a reference value for the annual growth rate of M3 of
4½ %. The second pillar refers to a broadly based assessment of inflationary developments.
Assuming that the first pillar becomes more and more important and in light of the facts that
economic developments in Germany have been a major influence in the business cycle
dynamics in the European Monetary Union (its GDP share is about 33% of the overall EMUGDP), and that the Bundesbank has been notably successful in keeping inflation low, our
results suggest that the ECB should be careful in its monetary policy actions: Raising rates
may generate a downswing while lowering rates may only fuel inflation. But, on the other
side, as we have seen, the asymmetric effects of monetary policy influence the German
economy mainly via exports. The openness of the Euro area is much less than that of Germany
(as a percentage of GDP exports of the Euro area are about 13 % compared to nearly 30 % of
Germany). This means that dangers of asymmetries may be less important for the Eurosystem.
18
First efforts with a forward looking reaction function yielded no asymmetric output effects. This may be
interpreted as the Bundesbank trying to eliminate possible asymmetries in this way.
15
Appendix: The reaction function (9)
Assume the following quadratic loss function V(.) of the central bank
(A1)
V 

( p  p )
2
T
2


2
iday   2   iday  iday 1 2
2
The loss function is made up of three components: Deviations from the inflation target (ppT), deviations between the central bank rate or the short-term money market rate iday and the
long run average interest rate  and changes in the interest rates (iday-iday-1). The first
component reflects the fact that the final goal of a central bank is price stability. The second
component considers interest rate levels which exceed or fall short of a certain "normal"
level.19 While a "too-low" interest rate stimulates economic activity in the short run (with
rigid prices) the opposite holds true for a "too-high" interest rate. Therefore this component
stands for goals of monetary policy besides fighting inflation, e.g. business cycle motives. The
third part of the loss function is motivated by the fact that large interest rate fluctuations
destabilize financial markets and the real economy in increasing the uncertainty inherent in
monetary policy. The first-order condition for a minimum of the loss function in (A1) results
in
(A2)
iday 

p


( p  p T ) 

iday 1
   iday
 
 


If higher interest rates imply lower inflation (expectations), i.e. p/iday < 0, and this
relationship is constant over time (A2) may be written as
(A3)
iday   0  ß  iday 1    ( p  p T )
This is equivalent to the estimated equation (8) or (12), respectively, in the main text.
19
This presupposes that interest rates are stationary; see e.g. Seitz (1998).
16
Figure 1: The overnight rate and the price gap
16
12
8
4
0
-4
74 76 78 80 82 84 86 88 90 92 94 96 98
overnight rate
price gap
17
Figure 2: Positive and negative monetary policy shocks:
The residuals of equation 12
2
1
0
-1
-2
76
78
80
82
84
86
EASY
88
90
92
94
96
98
TIGHT
18
References
Acemoglu, D., Scott, A. (1997), Asymmetric Business Cycles: Theory and Time-Series
Evidence, Journal of Monetary Economics, Vol. 40, pp. 501-533
Ammer, J., Brunner, A. (1995), When is Monetary Policy Effective?, International Finance
Discussion Paper No. 520, Board of Governors of the Federal Reserve System,
Washington
Ball, L., Mankiw, G. (1992), Asymmetric Price Adjustment and Economic Fluctuations, The
Economic Journal, Vol. 104, pp. 247-261
Barro, R.J., Rush, M (1980), Unanticipated Money and Economic Activity, in: Fischer, S.
(ed.), Rational Expectations and Economic Policy, The University of Chicago Press, pp.
23-48
Bernanke, B. (1993), Credit in the Macroeconomy, Federal Reserve Bank of New York
Quarterly Review, Vol. 18, No. 1, pp. 50-70
Bernanke, B., Blinder A. (1992), The Federal Funds Rate and the Channels of Monetary
Transmission, The American Economic Review, Vol. 82, pp. 901-921
Bernanke, B., Mihov, I. (1997), What does the Bundesbank Target?, European Economic
Review, Vol. 41, pp. 1025-1053
Borio, C.E.V., Fritz, W. (1995), The Response of Short-Term Bank Lending Rates to Policy
Rates: A Cross-Country Perspective, BIS Working paper No. 27, May 1995
Buckle, R.A., Carlson, J.A. (1998), Inflation and Asymmetric Output Adjustments: Tests
Using Business Survey Data, in: Oppenländer, K.H., Poser, G. (eds.), Social and
Structural Change – Consequences for Business Cycle Surveys, Ashgate Publishers,
Aldershot, pp. 131-146
Caballero, R.J., Engel, E.M.R.A. (1992), Price Rigidities, Asymmetries, and Output
Fluctuations, NBER Working Paper No. 4091, Cambridge
Carlino, G.A., DeFina, R.H. (1996), Does Monetary Policy have Differential Regional
Effects?, Federal Reserve Bank of Philadelphia Business Review, March/April, pp. 17-27
Chadha, J.S., Janssen, N.G.J. (1997), What Monetary Authorities do - an Examination of
Reaction Functions for Germany, Japan, the UK and the US, Swiss Journal of Economics
and Statistics, Vol. 133, pp. 455-476
19
Clarida, R., Gertler, M. (1997), How the Bundesbank Conducts Monetary Policy? in: Romer,
C.D., Romer, D.H. (eds.), Reducing Inflation – Motivation and Strategy, The University
of Chicago Press, Chicago, pp. 363-406
Clarida, R., Gali, J., Gertler, M. (1998), Monetary Policy Rules in Practice: Some
International Evidence, European Economic Review, Vol. 42, pp. 1033-1067
Clostermann, J., Seitz, F. (1999), Der Zusammenhang zwischen Geldmenge, Output und
Preisen in Deutschland – ein modifizierter P-Star-Ansatz, Oktober 1999
Cover, J.P. (1992), Asymmetric Effects of Positive and Negative Money-Supply Shocks, The
Quarterly Journal of Economics, Vol. 107, pp. 1261-1282
De Bondt, G.J. (1998), Credit and Asymmetric Effects of Monetary Policy in Six EU
Countries: An Overview, DNB Staff Reports 23/1998, Amsterdam
DeLong, B.J., Summers, L. (1988), How Does Macroeconomic Policy Affect Output?,
Brookings Papers on Economic Activity, Vol. 2, pp. 433-480
Deutsche Bundesbank (1997a), Annual Report 1996, Frankfurt
Deutsche Bundesbank (1997b), Monthly Report September 1997, Vol. 49, No. 9, Frankfurt
European Central Bank (1999), The Stability-oriented Monetary Policy Strategy of the
Eurosystem, Monthly Bulletin January 1999, pp. 43-56
Froyen, R.T., Havrilevsky, T., Waud, R.N. (1997), The Asymmetric Effects of Political
Pressures on U.S. Monetary Policy, Journal of Macroeconomics, Vol. 19, pp. 471-493
Garibaldi, P. (1997), The Asymmetric Effects of Monetary Policy on Job Creation and
Destruction, IMF Working Paper No. 97/57, Washington
Gonzalez Minguez, J.M. (1997), The Balance Sheet Transmission Channel of Monetary
Policy: The Cases of Germany and Spain, Banco de Espana - Servicio de Estudios,
Documento de Trabajo No. 9713
Guender, A., Moersch, M. (1997), On the Existence of a Credit Channel of Monetary Policy
in Germany, Kredit und Kapital, Vol. 30, pp. 173-185
Hayo, B., Uhlenbrock, B. (1999), Industry Effects of Monetary Policy in Germany, ZEI
Working Paper B14
Hess, G.D., Iwata, S. (1997), Asymmetric Persistence in GDP? A Deeper Look at Depth,
Federal Reserve Bank of Kansas City Research Working Papers 97-02, May 1997
20
Issing, O. (1994), Die Geldmengenstrategie der Deutschen Bundesbank, Lecture at the
symposium "20 Years of Monetary Targeting in Germany" in Montabaur on December 9,
1994, reprinted in Deutsche Bundesbank, Press Excerpts, No. 91, pp. 1-11
Kakes, J. (1998), Is German Monetary Policy Asymmetric over the Business Cycle?,
University of Groningen, August 1998, mimeo
Karras, G. (1996a), Are the Output Effects of Monetary Policy Asymmetric? Evidence from a
Sample of European Countries, Oxford Bulletin of Economics and Statistics, Vol. 58, pp.
267-278
Karras, G. (1996b), Why are the Effects of Money-Supply Shocks Asymmetric? Convex
Aggregate Supply or "Pushing-on-a-String"?, Journal of Macroeconomics, Vol. 18, pp.
605-609
Kim, J., Ni, S., Ratti, R.A. (1998), Monetary Policy and Asymmetric Response in Default
Risk, Economics Letters, Vol. 60, pp. 83-90
Köhler, A. (1996), Nominale Preisrigiditäten auf Gütermärkten - Eine empirische
Überprüfung neukeynesianischer Erklärungsansätze, Ciret Studies 51, Munich
Kremp, E., Stöß, E., Gerdesmeier, D. (1999), Estimation of a Debt Function: Evidence from
French and German Firm Panel Data, in: Sauvé, A., Scheuer, M. (eds.), Corporate Finance
in Germany and France – A Joint Research Project of the Deutsche Bundesbank and the
Banque de France
Krugman, P.R. (1998), It's baaack! Japan's Slump and the Return of the Liquidity Trap,
Brookings Papers on Economic Activity, No. 2, pp. 137-187
Lenz, C. (1997), Asymmetric Effects of Monetary Policy in Switzerland, Swiss Journal of
Economics and Statistics, Vol. 133, pp. 441-454
Macklem, T., Paquet, A., Phaneuf, L. (1996), Asymmetric Effects of Monetary Policy:
Evidence from the Yield Curve, Working Paper No. 42, CREFE, Université du Québec à
Montréal
McCandless Jr., G.T., Weber, W.E. (1995), Some Monetary Facts, Federal Reserve Bank of
Minneapolis, Quarterly Review, Vol. 19, No. 3, pp. 2-11
Mishkin, F.S. (1982), Does Anticipated Monetary Policy Matter? An Econometric
Investigation, Journal of Political Economy, Vol. 90, pp. 22-51
21
Morgan, D.T. (1993), Asymmetric Effects of Monetary Policy, Federal Reserve Bank of
Kansas City Economic Review, 2nd quarter, pp. 21-33
Newey, W., West, K. (1987), A Simple, Positive Semi-Definite, Heteroskedasticity and
Autocorrelation Consistent Covariance Matrix, Econometrica, Vol. 55, pp. 703-708
Oliner, S.D., Rudebusch, G.D. (1996), Is there a Broad Credit Channel for Monetary Policy?,
Federal Reserve Bank of San Francisco Economic Review, No. 1, pp. 3-13
Oppenländer, K.-H. (1995), Zum Konjunkturphänomen, in: Oppenländer, K.H. (ed.),
Konjunkturindikatoren, Kap. 1.1, Oldenbourg Press Munich/Vienna, pp. 4-22
Ravn, M.O., Sola, M. (1996), Asymmetric Effects of Monetary Policy in the U.S.: Positive vs.
Negative or Big vs. Small?, mimeo
Rhee, W., Rich, R.W. (1995), Inflation and the Asymmetric Effects of Money on Output
Fluctuations, Journal of Macroeconomics, Vol. 17, pp. 683-702
Schächter, A. (1998), The Bundesbank's Strategy of Implicit Inflation Targeting. What Can be
Learned for the European Central Bank?, Bayerische Julius-Maximilians-Universität
Würzburg, Wirtschaftswissenschaftliche Fakultät, Wirtschaftswissenschaftliche Beiträge
des Volkswirtschaftlichen Instituts, July 1998
Seitz, F. (1996), Indicators for Monetary Policy from a Central Bank's Point of View, in:
Oppenländer, K.H., Poser, G. (eds.), Business Cycle Surveys: Forecasting Issues and
Methodological Aspects, Aldershot, Avebury, pp. 419-440
Seitz, F. (1998), Geldnachfrage, Zinsen und Zinsstruktur, in: Konjunkturpolitik, Vol. 44, pp.
256-286
Stöß, E. (1996), Enterprises' Financing Structure and their Response to Monetary Policy
Stimuli - An Analysis Based on the Deutsche Bundesbank's Corporate Balance Sheet
Statistics, Discussion paper 9/96, Economic Research Group of the Deutsche
Bundesbank, Frankfurt/Main
Tödter, K.-H., Reimers, H.-E. (1994), P-Star as a Link Between Money and Prices in
Germany, Review of World Economics, Vol. 130, pp. 273-289
Weise, C.L. (1999), The Asymmetric Effects of Monetary Policy: A Nonlinear Vector
Autoregression Approach, Journal of Money, Credit and Banking, Vol. 31, pp. 85-108
22
Abstract: Are the Effects of Monetary Policy Asymmetric? Experience of the
Bundesbank and Lessons for the European Central Bank
The paper analyzes whether German monetary policy has asymmetric effects on real variables.
At first some theoretical arguments in favour of asymmetries are presented. The following
section investigates the implications of the strategy of monetary targeting for the reaction
function of the Bundesbank. This may be relevant for the ECB, too, as money plays a
prominent role within the first pillar of her strategy. Section 4 presents the econometric
analysis from 1974 to 1998. The regression procedure first identifies the stance of
(unanticipated) monetary policy shocks. These results are then used in the second stage to
detect possible asymmetric effects. For GDP, exports and unemployment there are
unambiguous signs of asymmetries while for private consumption and investment this is not
the case.
Zusammenfassung: Wirkt die Geldpolitik asymmetrisch? Erfahrungen der
Bundesbank und Lehren für Europa
Das Papier untersucht die Frage, ob die deutsche Geldpolitik asymmetrische reale Wirkungen
erzeugt. Zunächst werden einige theoretische Argumente, die für Asymmetrien sprechen,
präsentiert. Anschließend werden die Implikationen der Geldmengenstrategie für die
Reaktionsfunktion der Bundesbank analysiert. In Abschnitt 4 erfolgt eine ökonometrische
Analyse für den Zeitraum 1974 bis 1998. Da die EZB im Rahmen der ersten Säule ihrer
Strategie der Geldmenge eine herausragende Rolle zuweist, dürfte diese Analyse auch für die
EZB von Relevanz sein. Zunächst wird der Restriktions- bzw. Expansionsgrad der Geldpolitik
festgelegt. Im zweiten Schritt erfolgt, darauf aufbauend, eine Untersuchung möglicher
asymmetrischer Effekte der Geldpolitik. Beim BIP, den Exporten und der Arbeitslosigkeit
können eindeutig Asymmetrien festgestellt werden, beim Privaten Konsum und den
Ausrüstungsinvestitionen dagegen nicht.
JEL: C22, E41, E43
Key words: asymmetry, monetary policy, reaction function, two-stage regression
23