Foreword
The European Summer Institute (ESI) has been established to promote and stimulate high-quality
analysis and public debate on economic, monetary and financial policy issues in Europe. The aim of
the ESI is trying to bring together leading researchers and senior policy-makers from central banks,
ministries of finance and economics and other institutions in the European Union, in order to create
both a meeting place and an informal network to stimulate the discussion of these key policy issues.
We hope that this report of last years conference, based on the authors’ summary of the papers they
presented at the conference, will widen the debate and stimulate further discussion in the academic
and policy communities. We are grateful to the authors for their work in preparing these summaries.
The founding members and sponsors of the ESI are the Deutsche Bundesbank, the
Nederlandsche Bank, the National Bank of Belgium, the Bank of Finland, the European Central Bank,
the Oesterreichische Nationalbank and the Banco de Portugal. The seventh conference ‘The Euro
Area as an Economic Entity’ was organized jointly with the Centre for Economic Policy Research
(CEPR), and was hosted by The Deutsche Bundesbank, Eltville (Frankfurt-am-Main) on
12-13 September 2003. We thank the Deutsche Bundesbank for their generosity in hosting and
supporting this meeting.
March 2004
Sylvester Eijffinger
Kees Koedijk
ii
Table of Contents
Foreword
ii
Table of Contents
iii
Introduction
1
Summaries of the papers presented at the conference:

Gert Peersman
4
What Caused the Early Millennium Slowdown? Evidence based on
Vector Autoregressions

Sylvia Kaufmann
11
The Business cycle of European Countries. Bayesian Clustering of
Country-Individual IP Growth Series

Lucrezia Reichlin
17
Business Cycles: A Comparison between the US and Europe

Jeffry Frieden
19
The Political Economy of International Institutions

Jeffrey Frankel and Menzie David Chinn
26
World Interest Rates

Ignazio Angeloni and Michael Ehrmann
38
Monetary Policy Transmission in the Euro Area: Any Changes after EMU?

Hermann Remsperger
45
Inflation Differentials in EMU: Causes and Consequences
iii
Introduction
Summarized by Fred Ramb
(The Deutsche Bundesbank)
The seventh annual conference of the Center for Economic Policy Research and the European
Summer Institute, which in 2003 was sponsored by the Bundesbank, was held in Eltville in
mid-September. The main purpose of these annual conferences is to enrich the dialogue between
universities and research institutions, on the one hand, and the central banks, on the other. At this
vibrant forum for facilitating an exchange of ideas between central bankers and academics topical
monetary policy issues are discussed, and the latest research results are presented. The theme of last
year’s conference was “The euro area as an economic entity”.
In his introductory remarks Bundesbank Vice President Jürgen Stark spoke about evaluating the
ECB’s monetary policy strategy, about the different economic cycles in the member states and the
associated differences in inflation rates, and the Stability and Growth Pact. He concentrated on the
pact and emphasized, among other things, the need for member states to coordinate their fiscal, labor
market and structural policies if an increasing degree of integration is to be achieved in Europe.
During his speech Stark handed over the prize for the best central bank research paper. Gert
Peersman of the Bank of England was presented with the award for his research on “What caused the
early millennium slowdown? Evidence based on vector autoregressions”. It is interesting to note that
the three best authors are working for central banks outside their respective home countries, a fact
that illustrates the growing exchange of staff within the European national central banks.
Otmar Issing (member of the Executive Board of the ECB) spoke about the increasing international
role of the euro. In the first quarter of 2003, for example, about 41% of the debt securities held
worldwide were denominated in euro. By way of comparison, 43% of debt securities were
denominated in US dollars and 5% in Japanese yen. He attributed this not least to the euros stability
resulting from the stability-oriented policy pursued by the ECB.
The research paper presented by Peersman examines the causes of the economic crisis in industrial
countries such as the United States and the euro-area countries at the beginning of the new
millennium. Four different shocks – a supply shock, a demand shock, a monetary policy shock and an
oil price shock – can be identified using a comparatively simple VAR model. In fact, the cause of the
cyclical downturn turns out to be an interaction of several shocks. However, the results are not
independent of the chosen method. Making alternative assumptions for the specification can lead to
different values for the respective impact of the identified effects. David Mayes of the Bank of Finland
presented the research paper “Asymmetries in the euro-area economy” jointly with Matti Virén
1
(Bank of Finland). In an empirical analysis of the period between 1987 and 2000 the authors
investigate the extent to which asymmetries are identifiable in the monetary policies applying within the
European Union. They conclude that inflationary shocks which substantially exceed a set inflation
target are associated with a greater policy response from the central bank than smaller shocks are.
This also applies to the reaction of monetary policymakers to deviations from the output gap. The two
authors therefore suggest that the ECB refrain from fine-tuning measures but, instead, should react
only to fundamental macroeconomic shocks. Sylvia Kaufmann of the Oesterreichische Nationalbank
presented the third research paper (“The business cycle of European countries; Bayesian clustering of
country-individual industrial production growth series”). This study uses panel data and Bayesian
methodology in an attempt to identify a possible congruence in the economic cycles within Europe and
vis-à-vis other industrial nations. Congruent economic cycles can be identified for the most part within
the European Union. Furthermore, it is possible, first, to distinguish between a European and an
overseas economic cycle and, second, to identify a difference in the timing of economic cycles in the
case of two European states (Finland and Ireland). Until the beginning of the 1990s these two
economies had economic cycles that were more or less in line with those of countries overseas
(Australia, Canada, Japan and the United States) but have since converged significantly towards a
European economic cycle.
The second day of the conference started with a paper from Lucrezia Reichlin from the University of
Brussels. She discussed the extent to which real-time data should be used for empirical analyses and
presented results from empirical studies on economic cycles in the European Union, the United States
and the United Kingdom in which real-time data had been used. In Reichlin’s view it is difficult to
conduct a timely economic analysis on the basis of GDP data. The reasons for this are that GDP data
are available only on a quarterly basis and with a three-month delay, entail comparatively sizeable
measurement errors and are therefore subject to adjustment. In the case of some European countries
this situation is seen as a major problem and has led to demands for better official statistics.
“One Europe, one vote?” This was the question posed by Jeffry Frieden of Harvard University in his
research paper. Taking a theoretical approach, the author discussed the extent to which different
forms of voting behavior in the EU member states influence negotiating powers. Optimal voting
behavior is a trade-off between the advantages of having considerable voting clout in a standardized
voting procedure and the disadvantages which arise from making compromises owing to
heterogeneous interests.
Menzie Chinn of the University of Wisconsin presented a paper jointly with Jeffrey Frankel of Harvard
University. The paper, entitled “World interest rates”, examines interest rate levels worldwide. By
means of a panel study based on data for the United States, Germany, France, Italy, Spain and the
United Kingdom between 1973 and 2003, the authors analyse how far short-term and long-term
interest rates were determined by monetary policy in the United States or in Europe. The United States
was found to dominate during the past three decades. Although long-term real interest rates in the
2
United States and Europe have converged significantly since 1999, no empirical analysis has so far
been made owing to methodological problems and the absence of a structural model.
The monetary transmission process since the beginning of European monetary union was the subject
of the paper by Ignazio Angeloni and Michael Ehrmann of the European Central Bank. In their
empirical analysis the authors focus on the banking channel, the interest rate channel and the capital
market channel. There are indications that the transmission channel via the banks has grown in
importance and has become more homogeneous within the member states. The interest rate channel
changed with the start of European monetary union and now has a largely uniform effect in the
member states. The authors come to a similar conclusion with regard to the capital market channel.
In his closing address Hermann Remsperger discussed the reasons for and the consequences of
inflation rate differentials in the euro-area countries. No one fundamental reason for inflation
differentials has been identified. We emphasize that we found several causes for inflation differentials.
Empirical studies indicate that may be due to different output gaps and different price movements
within the member states, and so regional differences within the euro area cannot be ruled out.
However, it must be assumed that a self-correcting mechanism will develop in the long term within any
monetary union with a single monetary policy. Yet persistent inflation differentials carry a
macroeconomic cost, not to mention the fact that they also result in interregional tensions. As the
ECB’s monetary policy cannot take account of regional aspects, it is the responsibility of the member
states to overcome shocks on both the supply and demand sides. Wage policy plays an important role
in this context. Finally, with regard to EU enlargement, it will be necessary for the new member states
to bring their respective inflation rates as closely into line with the existing inflation target as possible.
The eighth annual conference of the Center for Economic Policy Research and the European Summer
Institute next year will probably be convened in an east European city.
3
What Caused the Early Millennium Slowdown? Evidence
based on Vector Autoregressions
Gert Peersman
(Bank of England)
1. Introduction
Between the first quarter of 1994 and the second quarter of 2000 (1994Q1-2000Q2), industrialized
world (aggregate of 17 countries) real GDP grew at an average annual rate of 3.1 percent. This was
even 3.9 percent for the US. At the same time, annual inflation was historically very low: on average
less than 2 percent. Activity weakened at the end of 2000 and industrialized countries experienced
negative growth by the end of 2001. In the paper, we analyze the underlying sources of this slowdown
and the preceding expansion.
Since the seminal work of Sims (1980), vector autoregressions (VARs) are often used as a tool for
analyzing underlying shocks in explaining recessions, as in Blanchard (1993) and Walsh (1993), each
of whom analyze the 1990-1991 recession in the US. Blanchard (1993) estimates a VAR on the
components of GDP and finds that the recession was associated with large negative consumption
shocks. Walsh (1993) analyses aggregate supply, aggregate spending, money demand and money
supply disturbances. His results suggest that the downturn was due to restrictive monetary policy and
negative aggregate spending factors. In contrast to these papers, we focus on the recent slowdown.
Moreover, our analysis is done at the industrialized world level, and a comparison is made between
the US and the Euro area. Within this VAR framework, we identify four types of underlying
disturbances, i.e. an oil price, aggregate supply, aggregate demand (spending) and monetary policy
shock. In order to identify these shocks, we compare two strategies. The first is based on conventional
zero contemporaneous and long-run restrictions. As an alternative, we propose an identification
scheme based on more recent sign restrictions. The latter is the main methodological contribution of
the paper and allows us to check whether the identification strategy matters for the results.
2. A simple model for the industrialized world, US and Euro area
In the paper, we estimate a simple four-variables VAR for the sample period 1980Q1-2002Q2. The
variables that we include in the VAR are the first difference of oil prices, output growth, consumer
inflation and the short-term nominal interest rate. We identify four types of underlying disturbances,
respectively an oil price, aggregate supply, aggregate demand (spending) and monetary policy shock.
Because there is a continuous interaction among all variables in the system, it is not possible to
estimate the immediate impact of the shocks on the variables. This well known identification problem
has to be solved and is typically done by imposing a number of restrictions. Two different procedures
are discussed below. The first is based on traditional zero contemporaneous and long run restrictions,
and as an alternative, we provide results based on sign restrictions.
Traditional identification strategy
For the traditional identification strategy, we use an extended version of the Gali (1992) and Gerlach
and Smets (1995) approach. In order to identify oil price shocks, we assume that there is a
contemporaneous impact of an oil price shock on all the other variables in the system, but no
4
immediate impact of the other shocks on oil prices. The assumption of exogenous contemporaneous
oil price movements is commonly used in the empirical VAR-literature. For instance, studies examining
the monetary transmission mechanism assume no immediate impact of monetary policy shocks on
commodity or oil prices. As the oil price is a financial variable, this assumption is questionable and will
be relaxed in our alternative procedure.
Following Blanchard and Quah (1989), we rely on a vertical long-run Philips curve to assume that
demand and monetary policy shocks have no long-run impact on the level of real output. Supply
shocks are thus associated with permanent shocks to output. These restrictions are better justified by
economic theory. Nevertheless, some equilibrium growth models (for example many overlapping
generations models or models with hysteresis effects) allow for permanent effects of aggregate
demand and monetary policy shocks on output because they can affect the steady state level of
capital. Furthermore, relying on long-run conditions can be highly misleading from an empirical point of
view. Faust and Leeper (1997) show that substantial distortions are possible due to small sample
biases and measurement errors when using these type of restrictions. Again, this long run neutrality is
relaxed when we use sign constraints as an alternative.
In order to discriminate between aggregate spending and monetary policy shocks, we follow the
literature on the monetary transmission mechanism and use the restriction that monetary policy shocks
have no immediate effect on output, i.e. there is a lag in the transmission process. There is, however,
no theoretical reason to justify this restriction which is inconsistent with a large class of general
equilibrium monetary models. The latter problem is avoided with an procedure based on sign
restrictions.
Identification based on sign restrictions
In order to check the robustness of the results, we use an alternative approach that does not suffer
from these problems. Faust (1998), Uhlig (1999) and Canova and De Nicoló (2002) introduce sign
restrictions on the impulse response functions to identify a monetary policy shock. The advantage of
their strategy is that zero constraints on the contemporaneous impact or the long-run effects of the
shocks are not necessary. Instead, their approach only makes explicit use of restrictions that
researchers often use implicitly. In their analysis, researchers experiment with the model specification
until the results look reasonable. For example, according to conventional wisdom, a restrictive
monetary policy shock is expected to have a negative impact on prices and a temporary effect on
output. This a priori theorizing is made more explicit with sign restrictions, and at the same time, no
additional short and long-run restrictions are necessary. In contrast to the existing literature, we do not
only identify monetary policy shocks, but a full set of shocks (oil price, supply, demand and monetary
policy). The identification of additional shocks should also help to identify the monetary policy shock.
As already mentioned, because there is a continuous endogenous interaction between all the
variables in the system, it is not possible to estimate the effects of the shocks on the variables. In order
to decompose the interactions between the variables, a set of restrictions are typically imposed (see
above). There are, however, an infinite number of possible decompositions and the traditional
identification strategy only generates one of them. In our alternative procedure, we generate ALL
possible decompositions (in order to transform an infinite number of decompositions into a large but
5
finite number, we use a grid interval) and select the decompositions that are consistent with a number
of imposed sign conditions. Results are then presented for the median of all accepted decompositions
and certain percentiles.
The following sign restrictions are used in the paper to identify the shocks. These restrictions are
based on conventional wisdom and are consistent with traditional aggregate supply – aggregate
demand schemes.

After a positive oil price shock (increase of oil prices), there is no increase of output, no decrease
of prices and no decrease of the interest rate.

Output does not decrease after a positive supply shock. In addition, there is no increase of prices
and the interest rate.

A positive aggregate demand shock has no negative impact on output. There is no decrease of
prices, no decrease of the interest rate and no decrease of the oil price level.

An unexpected rise in the interest rate (monetary policy shock), results in no increase of output
and no increase of the price level. Moreover, there is no increase of the oil price level.
Are the impulse responses different for both strategies?
Because we impose the restrictions of ‘no increase’ instead of ‘decrease’, a majority of decompositions
based on traditional restrictions are part of the solutions obtained with sign restrictions. This allows us
to situate the traditional solution in the whole distribution of possible solutions obtained with sign
restrictions.
In general, impulse response functions to all shocks look very similar for both approaches, but there
are some interesting differences. For instance, the response of oil prices to a demand and monetary
policy shock is substantial larger when using sign restrictions. The magnitude is around three times as
high. The largest part of the effect is even instantaneously. This illustrates that contemporaneous zero
constraints for oil prices in the traditional approach is too stringent. Oil prices do react endogenously to
other shocks in the economy. Ignoring this implies that part of the demand and monetary policy shocks
are identified as oil price shocks. On the other hand, we do not find permanent effects of a monetary
policy and aggregate spending shock on the level of output. This restriction was imposed in the
traditional strategy, but is now a result obtained from the estimations. Whilst the immediate effect of a
monetary policy shock on output is restricted to be zero in the conventional literature, we find a
substantial effect with our alternative approach. More than 1/3 of the total impact is estimated to occur
within one quarter.
In order to illustrate the quantitative importance of the differences, Figures 1 and 2 show the results of
two topical simulations. The first, Figure 1, is an exogenous rise in oil prices of 10 percent. The black
lines are the medians of the responses based on sign restrictions, together with 84 th and 16th
percentiles error bands (dotted black lines). The grey line is the estimated response based on
traditional restrictions. The response of prices is very similar across both methods: the result of the
conventional approach lies within the confidence bands of the sign conditions. The response of output,
however, is estimated to be much smaller with traditional restrictions, and lies even outside the error
bands. This implies that the solution obtained with conventional restrictions lies in the tails
6
(below 16th percentile) of all possible solutions. The difference is also economically very important. A
10 percent increase in the oil price has a long-run negative effect on output of 0.22 percent with
traditional restrictions. The median response based on sign restrictions predicts an impact of 0.48
percent, which is more than double.
Figure 1: Responses to a 10 percent rise in oil prices
Output
Prices
0.1
0.8
0.0
0.7
-0.1
0.6
-0.2
0.5
-0.3
0.4
-0.4
0.3
-0.5
0.2
-0.6
0.1
-0.7
-0.8
0.0
0
4
8
12
16
20
0
4
8
quarters
12
16
20
quarters
The second simulation, shown in Figure 2, is an exogenous increase of the interest rate of 50 basis
points. The results are clearly different. The responses based on conventional restrictions lie mostly
outside the error bands of the distribution of sign constraints. The response of oil prices is much
smaller: 2.7 percent in the long-run, compared to a median response of 21 percent with sign
restrictions. More relevant for the monetary transmission mechanism is a much smaller effect on
output and prices with traditional restrictions. The maximum impact on output is -0.27 percent with
conventional restrictions, whilst the impact is estimated to be between -0.39 and -1.04 percent with
sign constraints, economically and important difference. The long-run impact on prices obtained with
traditional methods lies just inside the error bands (though not in the short-run). The difference with the
median estimate based on sign conditions is, however, very high. The latter is more than double: -0.81
percent compared to -0.32 percent using conventional restrictions. In sum, the differences between
both approaches for the impact of an exogenous oil price and monetary policy shock are substantial
and economically very relevant.
Figure 2: Responses to a 50 basis points increase in the interest rate
Output
Prices
1.0
Oil price
0.0
0.8
5
0
-0.2
0.6
-5
-0.4
0.4
-10
0.2
-0.6
-15
0.0
-0.8
-0.2
-20
-1.0
-0.4
-25
-0.6
-1.2
-30
-0.8
-1.4
-1.0
-1.2
-35
-40
-1.6
0
4
8
12
quarters
16
20
0
4
8
12
16
20
0
4
quarters
8
12
16
20
quarters
3. Decomposing output: an analysis of the early millennium slowdown
Based on the estimates, we can calculate the shocks and the cumulative effects of these shocks on
output. This means that output can be written as the sum of a deterministic component, the
contribution of current and past oil price shocks, current and past aggregate supply shocks, current
7
and past aggregate demand shocks, and current and past monetary policy shocks. The contributions
of the shocks to the level of industrialized world output (measured as a deviation from deterministic
output) are presented in Figure 3 for the period 1995Q1-2002Q2.1
Figure 3: Contribution of shocks to industrialized world output
Traditional restrictions
Sign restrictions
2.5
2.5
2
2
1.5
1.5
1
1
0.5
0.5
0
0
-0.5
-0.5
-1
-1
1995Q1 1996Q1 1997Q1 1998Q1 1999Q1 2000Q1 2001Q1 2002Q1
oil
supply
demand
monetary
1995Q1 1996Q1 1997Q1 1998Q1 1999Q1 2000Q1 2001Q1 2002Q1
oil
supply
demand
monetary
Starting in 1995, there is a continuous increase in the level of output due to positive supply shocks
(typical characterized as the new economy). These positive effects stagnate around 2000Q2, after
which there is a negative contribution of supply shocks to output until the end of 2002Q2. These
results are very consistent across both identification strategies. The fall in output is only a bit more
pronounced with the conventional approach. Annual growth was on average more than 0.3 percent
higher as a result of positive non-oil supply shocks between 1995 and 2000 for both methodologies.
Also the results for the contribution of demand shocks are very similar for both approaches. The
contribution to output growth was positive in 1998 and 1999, but turned negative in 2000 and 2001.
For the last three quarters of 2001, negative demand shocks made a substantial contribution to the
slowdown of around 1 percent. In 2002, the contribution to output growth is again positive.
The negative supply and demand shocks are accompanied by a negative impact of oil price shocks.
The result is, however, highly influenced by the methodology used. For both methods, we find that
declines in oil prices during the period 1997-1998 had positive effects on output afterwards. The
figures are, on the other hand, different for the increases of oil prices in 1999 and the first quarter of
2000. With conventional methods, as a result of a slow pass-through of oil price shocks, this had a
negative and highly significant impact on industrialized world output growth of 0.44 percent in 2001.
This finding is not consistent with the results obtained using alternative restrictions: the impact of oil
price shocks is estimated to be negligible.
The opposite is true for the impact of monetary policy shocks. Both methods find a significant positive
contribution of monetary policy shocks to output growth in 2000 as a result of easy monetary policy in
1
Tables with the contribution to output growth and inflation are included in the paper.
8
1999. The magnitude is much larger with sign restrictions. On the other hand, restrictive monetary
policy had an insignificant effect on output growth in 2001 using conventional restrictions, but
restrictive monetary policy played an important and significant role using sign conditions: industrialized
world output is estimated to have fallen by 0.38 percent in 2001.
In sum, we find a very important role for aggregate demand and aggregate supply shocks in explaining
the recent slowdown across both identification methods. With traditional restrictions, we also find a
considerable impact of oil price shocks, while restrictive monetary policy played a major role with sign
conditions. These results indicate that a lot of the effects of oil price shocks from the traditional
approach are picked up by monetary policy shocks using sign conditions in explaining the recent
slowdown, and illustrate that restricting the contemporaneous response of oil prices and output to
monetary policy shocks to be zero can have a substantial influence on the results and the conclusions.
4. Comparison between the United States and the Euro area
An extension of the analysis involves making a comparison between the United States and the Euro
area. The contributions of the shocks to output are presented in Figure 4 for respectively the traditional
approach and sign restrictions procedure. A first feature is that the contribution and volatility of the
shocks was much higher in the US over the past seven years (which is not the case for the whole
sample period). With respect to oil price shocks, the results are very similar for both areas and
consistent with the aggregate results. We find a negative effect on output with the traditional approach
and almost no effect with sign conditions (even positive effects for US in 2002). Whilst the effects of oil
price shocks had the same sign and magnitude in both areas, this is not the case for the other shocks.
Figure 4: Contribution of shocks to output in United States and Euro area
Euro area - traditional
US - traditional
5
5
4
4
3
3
2
2
1
1
0
0
-1
-1
-2
-2
1995Q1
1996Q1 1997Q1
oil
9
1998Q1
supply
1999Q1
2000Q1
demand
2001Q1
2002Q1
monetary
1995Q1
1996Q1 1997Q1
oil
1998Q1
supply
1999Q1
2000Q1
demand
2001Q1
2002Q1
monetary
US - sign
Euro area - sign
5
5
4
4
3
3
2
2
1
1
0
0
-1
-1
-2
1995Q1
-2
1996Q1 1997Q1
oil
1998Q1
supply
1999Q1
2000Q1
demand
2001Q1
2002Q1
monetary
1995Q1
1996Q1 1997Q1
oil
1998Q1
supply
1999Q1
2000Q1
demand
2001Q1
2002Q1
monetary
With conventional restrictions, supply shocks made an accumulated positive contribution to output of
4.4 percent over the period 1996-2000 for the US, while this was hardly 1.4 percent for the Euro area
for the same period. The ‘new economy’ idea was clearly more a US phenomenon. Moreover,
negative supply shocks led to a fall in output of 0.7 percent in 2001 in the US but only 0.1 percent in
EMU. This difference between both areas also emerges with sign restrictions.
The pattern of demand shocks was different across both areas. In the US, the contribution of demand
shocks to output between 1996 and 2001 is always above baseline with traditional restrictions and
most of the time with sign constraints. From the beginning of 2000 onwards, demand shocks became
mainly negative and the contribution to output turned below baseline in 2001. Accordingly, output
growth was respectively 1.07 and 1.26 percent lower in 2001 for both approaches. For the Euro area,
there were a number of negative demand shocks between 1995 and 1997 with corresponding effects
on output, after which there was a positive trend, though very small in magnitude, until the end of
2000. In 2001, Euro area output fell respectively 0.10 and 0.33 percent due to negative demand
shocks for our two methods.
Monetary policy was rather stimulating until the beginning of 2000. This reinforced the ongoing boom
in the US. Consistent with industrialized world aggregates, this effect is more pronounced with sign
conditions. According to the latter method, output growth is estimated to have been 1.11 percent
higher in 2000 as a result of weak monetary policy. Conversely, monetary policy became very
restrictive by the end of 2000. With conventional constraints, the contribution of tight monetary policy
to the recession in 2001 was very modest. Using sign restrictions, however, annual growth was 0.53
percent lower. In the Euro area, policy became stimulating in 1999, after the introduction of the euro.
European Central Bank interest rates were always below an average policy rule until the middle of
2000. The impact on output growth was significantly positive in 2000. From the middle of 2000
onwards, in contrast to the US, monetary policy was always relatively neutral in the Euro area. We
find, however, a significant negative effect on output growth in 2001Q3 and Q4 using sign restrictions.
This is mainly the result of reversed effects of past stimulating shocks following long-run neutrality of
monetary policy on the level of output.
10
Conclusions
In the paper, we have analyzed the underlying sources of the early millennium slowdown using a
simple four variables VAR for the industrialized world, US and Euro area. Within this VAR, four shocks
are identified, i.e. supply, demand, monetary policy and oil price shocks, based upon two different
identification strategies. One is based on traditional zero contemporaneous and long-run constraints,
and we propose an alternative based on more recent sign restrictions. We find that the recent
slowdown is caused by a combination of several shocks. Across both methodologies, we find an
important role for negative aggregate spending shocks. In addition, there were negative aggregate
supply shocks, negative effects of restrictive monetary policy in 2000 and a negative impact of oil price
increases in 1999. The magnitude of the latter two is significantly different between both approaches.
We find an important role for oil price shocks with conventional restrictions and for monetary policy
shocks using sign conditions. The effects of the shocks are more pronounced in the US than the Euro
area.
11
The business cycle of European countries. Bayesian clustering
of country-individual IP growth series. A shortcut with possible
extensions.
Sylvia Kaufmann (Oesterreichische Nationalbank)
Questions and some answers
The paper deals with the question whether the growth rate of industrial production has followed the
same or a similar business cycle pattern in euro area countries and in all European countries.
Moreover, the approach taken here allows to assess the relation of the European countries with
transatlantic or “overseas” countries, in particular Australia, Canada, Japan and the United States. The
focus lies on three different observation periods, a long-term historical perspective (1978-2001), a
medium-term (1990-2001) and a short-term perspective (1999-2001), which reveals whether the
synchronization of IP growth has changed among the countries in the course of increasing European
integration.
The results give answers to the following questions:
(i)
Is there a common growth cycle for the euro area countries?
Generally there is, as well in the long-term historical perspective as in the medium-term
perspective (during the 1990s). In the recent past (1999-2001), however, the countries
may be divided into two groups, whereby the largest euro area countries along with Spain,
Austria, Belgium and Greece experienced a shorter expansion period than the remaining
countries.
(ii)
Can we discriminate between a “European” and an “overseas” cycle?
Yes; when the analysis includes Australia, Canada and the US, these three countries
characterize one group over all time horizons, while most European countries fall into the
other group.
(iii)
Did the synchronization change over time?
The increased integration of European countries is reflected in the changing classification
of Finland and Ireland. While under the long-term historical perspective they follow more
closely the “overseas” cycle, they follow more closely the European cycle during the
nineties and during the recent past. There is also evidence for a changing synchronization
between the European and the overseas cycle. Until the early nineties, overseas
downturns were leading European downturns by half a year to about one year. During the
nineties, however, overseas countries went through a long-lasting expansionary period
while European countries experienced three full growth cycles.
12
(iv)
Which countries follow the “overseas” rather than the “European” cycle?
Noteworthy is that in contrast to previous studies, the UK, and also Japan, follow more
closely the European rather than the overseas cycle over the long-term historical
perspective. It is only during the recent three years, that both countries are classified into
one group with Canada and the US. Finally, the downturn in these countries is identified to
have begun already in the second half of 2000, before affecting all countries in 2001.
The model and the estimation method
The individual country series are analyzed within a panel data framework which enables us to enlarge
the focus from the euro area towards a European versus overseas perspective and/or to restrict the
analysis to the first three years of the common European currency. The model specification allows for
a time-varying growth rate that switches according to a latent state indicator which itself follows a
Markov switching process. Countries that follow the same or a similar switching pattern are grouped
together whereby the groupings are not set a priori; rather, for a given number of groups, the
classification is estimated along with all model parameters and the latent state indicator. For country i ,
the growth rate of industrial production (the first difference of the logarithmic level) evolves according
to
yit  iG  iR ( I it 1)  i yi ,t 1     p yi ,t  p   it ,
with
 it ~ i.i.d. N (0, 2 ) , t  1,, T ,
(1)
i  1, , N . For a single country, the model comes close to the
one estimated for US GNP in Hamilton (1989). The country-individual growth rate
latent state variable
 iG   iR
 it   G
 i
 it
depends on a
I it , which takes on the value 0 or 1:
iff I it  0
iff I it  1
The latent specification for
.
I it takes into account that the state prevailing in each period is usually not
observed with certainty. The process governing
I it is Markov switching of order one to capture in
addition the fact that recoveries and recessions may have a different duration:
P( I it  l | I i ,t 1  j )   ijl , with
1

l 0
i
jl
 1 , j  0,1 .
Having a cross-section of countries, the aim is to group those together which behave similarly over
time, i.e. which comove in the sequence of recoveries and slowdowns/recessions and which also
experience similar growth rates over time. Again, the appropriate grouping is usually not known with
certainty; therefore, a second latent variable, a group indicator, is introduced, which takes on the value
k out of K possible values, Si  k , k  1,, K . The country-individual growth rate  it is finally
specified as:
 kG   kR
iff S i  k and I kt  0
G
 k
iff S i  k and I kt  1
 it  
13
,
where the probabilities
Thus, within a group
P(Si  k ) are given by  kG , k  1,, K with the restriction kK1 kG  1 .
k , the countries are pooled to estimate the group- and state-specific parameters
(  kG ,  kR ) , and it is also the same state indicator that governs the switching process within the group,
i.e.
I it  I kt iff S i  k .
Note finally, that the autoregressive parameters in equation (1) are assumed to be group-and stateindependent. Likewise the variance of the error terms is also assumed to be group-independent,
 i2   2 .
The estimation of the model should yield an inference on the model parameters (gathered in the vector
 ) and additionally on the group and the (group-specific) state indicators. Note that, if we knew
S N  ( S1 ,, S N ) and I T  ( I1T , , I KT ) , where I kT  ( I kT , I k ,T 1 ,, I k1 ) , k  1,, K , we would
be left with a regression model that could be estimated by standard classical methods like GMM in the
present case. This would still be feasible if only one latent variable would be present, as the marginal
likelihood
L(Y T |  , S N ) or alternatively L(Y T |  , I T ) can be derived (and maximized). With both
latent variables unknown, the marginal likelihood
L(Y T |  ) is not feasible.2 In this respect, using the
hierarchical structure given within the Bayesian framework,3 the complexity can be reduced to obtain
the inference on the joint posterior distribution
 ( , S N , I T | Y T )
by drawing iteratively from each
conditional posterior distribution:
(i)
 (S N |  , I T ,Y T ) ,
(ii)
 (I T |  , S N ,Y T ) ,
(iii)
 ( | S N , I T , Y T ) .
The results briefly summarized above were obtained by iterating 10,000 over these sampling steps
and discarding the first 4,000 to remove dependence from initial values. The remaining simulations
were used to conduct the posterior inference, e.g. the mean and the standard error of the posterior
distributions were estimated by the mean and the standard deviation of the simulated values.
Comparisons to previous studies
The obtained results are broadly consistent with previous studies that analyzed the synchronization
and the correlation structure of European economies. Artis and Zhang (1997, 1999) find that the
contemporaneous correlation between European countries has increased during the ERM period while
at the same time the correlation with the US cycle has decreased. Forni and Reichlin (2001) analyze
regional output fluctuations in 9 European countries and find out that the European component
2
To derive L(Y T |  ) , we need to integrate out L(Y T | , S N ) over S N which itself depends on I T . The same
applies if we want to marginalize L(Y T |  , I T ) .
3
See Kaufmann (2003) for a detailed description of the hierarchical structure, the assumptions on the prior
distributions and the derivation of the posterior distributions for each sampling step.
14
explains almost 50% of output growth in most regions and that it is highly correlated among the
regions, which indicates a high degree of synchronization of output fluctuations in Europe. Finally, in a
recent study, Mitchell and Mouratidis (2002) find that average correlation of various business cycle
measures among the euro area countries has increased since the 1980s and continues to rise. The
exception appears to be the United Kingdom, which, according to Artis, Krolzig and Toro (1999),
follows more closely the US rather than the German cycle and, according to Forni and Reichlin (2001),
has a larger national than European component in output fluctuations. The present results are partly at
odds with this evidence as the UK follows the European countries more closely than the overseas
countries, in the long-term historical perspective (1978-2001) as well as in the medium-term
perspective (1990-2001). In the recent short-term perspective (1999-2001), however, it turns out that
the UK has been moving more closely than before with the US and Canada.
An additional byproduct of the estimation is obtained by using the inference on the posterior state
probabilities to date business cycle turning points. For the euro area, they follow quite closely the
dates for the 1990s published in the Monthly Bulletin of the ECB (2002) and lag the ones identified by
EuroCOIN (see Altissimo et al., 2001) by two to three quarters.
Implications of the model specification and extensions
As already mentioned, the model specification assumes the same error variance and group- and state
independent dynamics across all countries, i.e. group- and state-independent autoregressive
parameters, too. It is the case, however, that different countries in the panel may exhibit different
volatilities in their IP growth rates, whereby notably catching up countries (like Ireland or Portugal)
experienced a higher volatility in growth rates than “core” euro area countries like France, Germany
and Italy. To cope with this problem, all series are standardized prior to estimation, so that the
groupings are estimated according to a similar growth pattern that is not due to differences in business
cycle volatilities. Given the group- and state-independent dynamics in equation (1), the similar growth
pattern we estimate refers to periods where a group of countries experienced a recovery or a
slowdown broadly speaking at the same time, i.e. where growth rates were higher or lower at the same
time. Thus, with this specification group-specific steepness and/or deepness are not captured. 4
Moreover, the series are not demeaned, so that the estimated groupings do also catch similar longterm growth perspectives (of the standardized series). If one would remove the country-specific mean
growth rate, one first would have to decide whether to remove a constant or a time-varying trend
reflecting changing long-term growth perspectives. Group-specific deepness would then be reflected in
“asymmetric” state-specific growth rates, i.e. the above-average growth rate would be lower than the
below-average growth rate (in absolute terms).
Allowing for group- and state-specific autoregressive parameters would capture group-specific
steepness of the business cycle, whereby a larger autoregressive parameter during below-average
4
These features and the ones addressed subsequently will be incorporated in an extensively revised version of
the working paper.
15
growth periods would reflect the fact that growth deterioration and initial recovery phases are quite
strong and fast, while maturing recovery periods are quite smooth.
Finally, to avoid standardizing the series, one might generalize straightforwardly the variance process
to country-specific variances,  it ~ i.i.d.N (0, i2 ) , by introducing a weighting parameter i , which relates
the country-specific variance to the overall error variance of the panel,  i2  
2
i (an approach
pursued in Frühwirth-Schnatter and Kaufmann, 2003).
References
Altissimo, F., A. Bassanetti, R. Cristadoro, M. Forni, M. Lippi, L. Reichlin, and G. Veronese (2001),
EuroCOIN: A real time coincident indicator of the euro area business cycle, Discussion Paper 3108,
CEPR.
Artis, M.J., H.-M. Krolzig, and J. Toro (1999), The European business cycle, Discussion Paper 2242,
CEPR.
Artis, M.J. and W.Zhang (1995), International business cycles and the ERM: Is there a European
business cycle?, International Journal of Finance and Economics 2, 1-16.
Artis, M.J. and W.Zhang (1999), Further evidence on the international business cycle and the ERM: Is
there a European business cycle?, Oxford Economic Papers 51, 120-132.
European Central Bank (2002), Characteristics of the euro area business cycle in the 1990s, Monthly
bulletin, July, 39-49.
Forni, M. and L. Reichlin (2001), Federal policies and local economies; Europe and the US, European
Economic Review 45, 109-134.
Hamilton, J.D. (1989), A new approach to the economic analysis of nonstationary time series and the
business cycle, Econometrica 57, 357-384.
Kaufmann, S. (2003), The business cycle of European countries. Bayesian clustering of countryindividual IP growth series, Working Paper 83, Oesterreichische Nationalbank.
Frühwirth-Schnatter, S. and S. Kaufmann (2003), Model-based clustering of multiple time series,
mimeo, Johannes Kepler University Linz and Oesterreichische Nationalbank.
Mitchell, J. and K. Mouratidis (2002), Is there a common Euro-zone business cycle? National Institute
of Economic and Social Research.
16
Business Cycle Empirics in the Euro Area
Lucrezia Reichlin (Universite Libre de Bruxelles and CEPR)
The presentation addresses several problems of empirical business cycle analysis in the Euro area.
It started to ask whether it is possible to identify a euro area business cycle that has specific
characteristics, possibly different from the US cycle and the English speaking group of countries within
the G7.
Two observations lead to think that business cycles in the G7 (with the exception of Japan) have been
sharing common characteristics:

Since 1963, output volatility has decreased everywhere.

Cycles have become longer and less synchronized.
Moreover, overall cross-correlation of output per capita in the G7 countries, have remained stable in
the long-run (cycles eight years or longer).
These observations apply to business cycles since the early sixties to the end of the nineties. But what
about recent events, i.e. business cycles since the creation of the euro?
The presentation highlights two facts:

The US experienced a recent recession while this is not clear for the Euro area

Since 2000, output in the euro area has been smoother than in the US, although growth has
been lower.
Although one can detect (and the NBER indeed declared it) a short recession in the US in 2001, this is
clear for the Euro area where much uncertainty prevails.
This observation leads to the second part of the talk, i.e. the discussion of the difficulties of real time
assessment of business cycle conditions in the euro area and the illustration of Eurocoin, the CEPR
coincident business cycle indicator, as a tool of timely assessment of the euro area cyclical conditions.
The presentation recalls that real time assessment of economic conditions (cycle and potential) based
on GDP is generally problematic. The main reasons cited are:
17

GDP available at quarterly level only

GDP released with a delay of at least a quarter

GDP has large measurement error and it is subject to large revisions

Historical revisions in the euro area have been large
Moreover, real time assessment in the euro area, it is claimed, is more problematic than in the US.
These are some of the reasons given:

Measurement error is typically higher (generally, larger revisions than in the US).

Monthly data that could serve as measures of overall cyclical conditions, such as sales are not
very reliable

GDP is not clearly associated to key quarterly variables such as investment and employment
and monthly variables such as industrial production
The presentation illustrated some of these points comparing real time business cycle dating in Italy
and in the US.
Finally, the presentation illustrated the key features of Eurocoin. The index, which, unlike GDP is
monthly and made available before GDP statistics are released, is constructed using hundreds of time
series that, when aggregated, can help capturing the GDP cyclical signal. The idea is to use leading
variables to reconstruct the signal at the end of the sample when GDP data are not available and use
both coincident and leading variables to estimate the monthly GDP cyclical signal. The index is
updated as new information become available in a non-synchronous way.
The presentation focused on Eurocoin in the last two years and analyzed its stability. It showed that,
although second quarter GDP growth in the euro area has been negative, the CEPR indicator has
been positive and, in the past four months, showed increasing rate of growth.
18
One Europe, One Vote? The political economy of European
Union representation in international organizations
Jeffry A. Frieden (Harvard University)
Since the Treaty of Rome, many proponents of European integration have hoped that a single Europe
would speak with more authority – and more influence – in the international arena. Recent transAtlantic conflicts have served to heighten the view that one Europe would be more influential if it had
one foreign policy voice.
The same calculation has been applied to other areas of Europe’s international relations, including in
the economic realm. A common European representation in the International Monetary Fund/World
Bank system would have more votes than the United States – and thus a veto. At the Bank for
International Settlements, the Group of 8, the World Trade Organization, and other international
economic institutions, there is a widespread view that a pooled European presence would increase
Europe’s influence. This is particularly clear in international monetary affairs, where there already is a
common European institution, the European Central Bank, that is the natural representative of the
euro zone at least.5
But a common European international representation is more complicated than may appear at first
glance.
A collective European voice requires a collectively agreed upon policy and bargaining
position, which means that it requires compromise among EU members. In this sense, adopting a
common international EU policy is analogous to adopting a common internal EU policy: it requires that
member states weigh the potential benefits of a common policy against the potential costs of a policy
that is not to their liking. And just as with other EU policies – such as the general focus on subsidiarity
– there is a clear tradeoff between the advantages of scale and the disadvantages of overriding
heterogeneous preferences.
We can think about the issue in a more positive mode by asking what factors make it more or less
likely that the EU will pool its presence in international institutions. Already, EU representation varies
among international organizations. In some cases, this is for obvious reasons, such as in trade: it is
hard to imagine a single market with 15 different trade policies. 6
But, just as the principle of
subsidiarity has been applied in ways that lead some EU policies to be decided at the EU level, while
others remain national, so too is it the case – and is it likely to continue to be the case – that the EU’s
5
6
See, on this, McNamara and Meunier 2002.
See, on European trade representation, Meunier 2000 and Meunier and Nicolaidïs 1999.
19
international presence will vary among issue areas and institutions. The discussion here helps us
understand why.
To analyze the choices available to the European Union’s member states, I make use of simple spatial
intuition.7 I make a series of assumptions for the sake of clarity of presentation; all of them can be
abandoned or made more complex without altering the general points of the discussion.
assumption is that policy can be mapped onto a single dimension.
One
This helps simplify analysis
significantly, and in this instance it is close enough to reality to be defensible. I assume also a series
of voting rules. To start with, let us say that both the EU and the international institutions operate on
the principle of one country, one vote, with simple majority rule. Many international organizations do
operate on this principle, and it is a reasonable starting point. I further assume that EU members can
decide, by unanimity rule, to pool their votes to equal the number of member states; the new common
policy is binding and is made by majority rule. It is realistic to assume that creating a common
international representative would require the consent of all EU members; it is also reasonable that
once such a joint agent were in place, the EU’s position would be by something less than unanimity.
In any event, these assumptions are not essential, and I relax them later.
An example helps illustrate the analysis. For simplicity’s sake, think of an international organization
(IO) with five member countries: two non-EU countries on the center-right and right, and an EU of
three member countries arrayed from center to left. The policy in question could be any one of a
number, such as:
financial regulation, where movement leftward implies less support for stringent prudential
regulation and supervision of national banks
macroeconomic policy, where movement leftward implies more support for simulative fiscal
or monetary policies
IMF conditionality, where movement leftward implies less support for restrictive IMF
conditions
trade policy, where movement leftward implies more support for trade liberalization –
perhaps in a particular area, such as agriculture or services
For the purposes of this essay, in order to avoid too much concentration on the specifics of
any one economic policy arena, I use examples drawn from outside the economic realm.
The
examples all involve the Iraq war, mapped so that movement leftward implies less support for the
United States position, and based on the obvious fiction that some international institution was
essential to the conduct of military operations in Iraq. The example is useful in large part because the
positions of both EU and non-EU governments are well known and relatively easy to map onto a single
dimension.
7
The models are not presented here, but are available from the author upon request. Hug and König 2002, and
Rodden 2002, consider other aspects of EU politics with similar tools.
20
The equilibrium policy is that of the median country-voter. But if the three EU members pool their
votes, they first arrive at a common policy by majority rule – that of the median EU country-voter. By
virtue of the EU’s three pooled votes, this common EU policy then becomes the majority-rule outcome.
The result makes the EU better off – the outcome is closer to the ideal points of more EU member
states than is the median voter. So pooling has improved the outcome from the standpoint of the EU
as a whole.
However, a European country that occupies the position as the international median voter would not
agree to this pooling of representation. Pooling shifts the equilibrium outcome away from it, so it
makes this country worse off – in the Iraqi case, assuming that the UK is the EU country closest to the
US, pooling would have shifted the outcome away from Britain’s preferred position, towards a less
bellicose military stance. This simple example indicates an important point:
Observation 1. Member states whose preferences are farther from the EU median than
they are from the international median are more likely to oppose pooling
representation, while those whose preferences are closer to the EU median than to the
international median are more likely to support it.
The reason is that the closer a member state’s preference is to the EU’s expected collective
preference; the better off it will be with pooling. After all, the more similar are the policy views of a
government (or group or individual) to those of the EU, the more it would like the EU’s views to prevail.
This helps explain in general, for example, why France and Germany, whose foreign policy views tend
to be closer to the EU median, are more interested in a common foreign policy than the United
Kingdom, which tends to be an outlier. The observation is analogous to similar observations about
other EU-wide policies, in which preference outliers tend to oppose centralization while those toward
the center of the preference distribution support it, such as when Scandinavians worry that EU-level
social policies will be less generous than their own national policies. 8
But this example is an unrepresentative one. For one thing, EU members are a simple majority in the
example, which is rarely likely to be the case; for another, pooling does not affect the EU’s bargaining
power. So now let us consider an international institution with seven members, of which there are
three EU members (preference distribution and voting rules are as above). With all states voting
individually, the outcome is that of the median state-voter, which is not an EU country. Even if all three
EU votes are pooled, there is no impact on the outcome – all that a common representation does is
stack EU votes at a common point, and with majority rule there is no consequential effect. This simple
observation illustrates a broader point, that the degree to which joint EU representation affects
outcomes is sensitive to many features of the international environment: the number of countries in
question, the distribution of their preferences, and so on. One can imagine many instances in which
pooling would have an impact – such as if the common EU representative were pivotal – but
8
For a more general statement of this point, and others related to the treatment here, see Crémer and Palfrey
1999.
21
Observation 2. In and of itself, pooling representation does not necessarily increase
EU influence over bargained outcomes.
This highlights an aspect of the discussion of Europe’s international role that is often disregarded: the
expectation that a united Europe is greater than the sum of its parts. Most proponents and analysts
regard the advantage of pooling to be more than simply additive: that is, there must be something
more to pooling EU representation than putting together national votes, otherwise it could be done by
way of voting coalitions. And there are reasons to think that a more formal shared representation
could in fact have an added impact on EU bargaining power. For one thing, in many issue areas a
joint stance improves the EU’s outside option (that is, the alternative to a bargained outcome). In
negotiations over financial regulations, for example, a country’s bargaining power is a function of how
costly it would be for the country not to conform to the agreed-upon regulations. And in this context, it
is likely that an EU with a common European regulatory framework – which it could implement on its
own if negotiations with the US and Japan broke down – would be much more powerful than 15 EU
member states each with a different regulatory framework.
Whatever the reason, there is typically expected to be some value added from a joint representative
beyond that of the member states’ own votes. Analytically, we can think of some issue areas and IOs
in which pooling is more likely to increase EU bargaining power than in other issue areas and IOs.
Common EU positions on trade or financial regulation – where outside options, thus bargaining power,
are related to size – are likely to be greater than the sum of their parts, while a common EU position on
Amazonian biodiversity may not be.
This situation, in which pooled IO membership improves the EU’s bargaining power, can be
represented by assuming that pooling gives the EU the equivalent of two additional votes, for a total of
five.
By construction, then, the outcome will be determined by the EU’s collectively defined
preference. This indicates that increasing the EU’s bargaining power can have a substantial impact on
outcomes. Depending on the distribution of preferences within the EU, the outcome may even make
all EU member states better off (or at least leave them the same). Indeed, the closer EU member
state preferences are to one another, and the farther they are from those of other members of the IO,
the greater the benefits and the lower the costs of a common position. In this case, this distribution of
preferences would have led even the United Kingdom to support a common EU position on Iraq – the
outcome would have been at least as close to the UK’s views as that without a common EU stance.
It is also the case, however, that a different distribution of preferences within the EU can make the
outcome with a pooled representative less appealing to some EU members. The gain in the EU’s
international bargaining power, in this instance, is more than offset by the intra-EU conflict of interests.
This may be closer to the reality in the Iraq example: it is likely that Britain’s position was in fact closer
to that of the United States than it was to the median EU member – so that Britain would not have
supported a common stance.
22
This discussion illustrates another fundamental point about the way in which EU positions are
developed and expressed in the international arena:
Observation 3. There is a tradeoff between the added bargaining power of a common
EU representative, on the one hand, and the need to override heterogeneous
preferences, on the other.
This point is analogous to that in the literature on other government functions, such as on European
subsidiarity, currency unions, or the size of nations. There are scale economies to the aggregation of
government functions, such as diplomacy; but there are costs to forcing heterogenous actors to adopt
a common policy position.9
It might be useful to give some illustrative examples. Starting with the first observation, it stands to
reason that countries whose views are more like those of other nations than they are like those of their
fellow EU members would be unlikely to want to increase the EU’s international influence. A country
with views on macroeconomic policy coordination, or financial regulation, or foreign aid, or IMF policy
very similar to those of the United States and very different from those of other EU members, would be
foolish to do anything to increase the EU’s relative influence and diminish that of the United States.
The second observation is in a sense related to the first. Even an EU preference outlier would care
little about the choice of common or distinct EU voting if this were unlikely to have little impact on
international policy outcomes. And it is not hard to think of instances in which a joint and common
delegation would not necessarily increase the EU’s bargaining power – the General Assembly of the
United Nations, perhaps – and so the change in European representation would be of little import.
The third observation, about the tradeoff between increased bargaining power and the need for a
common EU position, is perhaps the most important and richest in implications and applications. Just
as similar propositions provide a way of thinking about the choices facing EU members considering
centralizing authority at the European level, it fixes ideas about the positive political economy of –
rather than the normative opinions or journalistic punditry relating to – European international
representation. Every individual country, this says, must weigh the impact of a greater international
role of the Europe of which they are a part against the compromises they will have to make to arrive at
a common European position. The comparative static properties of this observation are simple and
important and can be applied to policy areas, or international institutions, in which there is variation in
EU bargaining power and member state preferences. They can be applied to analyses of individual
member-state views on a common international representation, or on the overall likelihood of such a
common representation emerging. This discussion implies, for example, that all else equal, the more
pooling increases the EU’s international bargaining power, the more likely it is. All else equal, the
9
See, as important examples, Alesina and Spolaore 1997, Bolton and Roland 1997, and Casella 1992. A similar
tradeoff between risk-sharing and redistribution is identified in Persson and Tabellini 1996.
23
greater the divergence of views among EU member states, the less likely is the EU to agree on a
common international voice. All else equal, countries that anticipate very serious compromises – that
whose policy positions are far from those of their EU partners – are less likely to support pooling. And
all else equal, countries with positions similar to those of their European partners are likely to be most
favorable to common representatives.
Extensions and applications
Even this rudimentary discussion reveals some of the complexity of the political economy of a common
European international voice. If the discussion is made more realistic, the implications are more
complex still.
There are two areas in which our assumptions were particularly simplistic:
the
distribution of preferences, and the voting rules both inside the EU and internationally. In addition, we
ignored the domestic politics of these issues.
The distribution of preferences.
Virtually any outcome is possible if government preferences are
differently arranged. The relationship among member state views, and between member state views
and those of non-EU countries, defines the range of feasible coalitions, and the range of possible
outcomes. It is thus a fundamental determinant of constraints on EU decision-making. While this
makes actual analysis more complex, it does not reduce the usefulness of the exercise. Coming up
with a realistic sense of government preferences is central to understanding the likely outcome of intergovernmental bargaining.
Voting rules.
The impact of different voting rules on outcomes is also crucial, especially in the
international arena, where majority rule is quite rare. Indeed, it is also rare in the European Union
itself. So we need to consider how intra-EU voting rules, and the voting rules of the international
organizations of which the EU is a member, might affect outcomes.
We can start with an extreme case, which is actually quite common both internationally and in the EU:
unanimity, which gives each country a veto. The fact that with unanimity rule each EU member state
has a veto would normally provide no reason for any EU member to want to move toward a common
representation, as this implies giving up its veto. Moving from a situation in which EU member states
have 15 vetoes to one in which they have only one collective veto might make most EU members
better off, if the countries losing the veto were far from the EU median, but it would not be accepted by
countries that would lose out as a result.
And in many (if not most) reasonable preference
distributions, pooling in a unit-veto system would make the EU worse off.
The more general point is that voting rules change bargaining outcomes, in predictable if hardly simple
ways. And voting rules in international institutions can themselves be highly complex. The European
Union uses a combination of unanimity, simple majority, and qualified majority voting; and, depending
on the issue area, one would have also to include the roles of the European Commission, the
European Parliament, and the European Central Bank. International organizations also have a wide
24
variety of voting rules, including the IMF’s quota-based weighted-voting system, with weights that have
changed over time. Without going into detail on the matter, suffice it to say, once more, that an
analysis of the institutional causes and consequences of the EU’s international presence is sensitive to
intra-EU and international voting rules.
Domestic politics. One last observation is worth making, about the domestic politics of the EU’s
international role. Just as governments of member states formulate their views on a common EU
position based on how close it is likely to be to their own, so too can groups within countries. An
interest group closer to the EU median than to its national median will prefer that policy be made at the
European, rather than the national level, and vice versa. For example, if the domestic Left is closer to
the European median, it will prefer a common European representative (whose views will be closer to
its own than to those of its national government), while the interests of the domestic Right are
opposite. This relationship – which applies, of course, to European policies more generally, such as
when the Left in a left-wing country prefers national to European social policies – helps explain some
of the partisan characteristics of debates over Europe’s international role. Many on the British Left
prefer a European foreign policy to a more moderate independent British foreign policy; many on the
Swedish Left prefer an independent Swedish foreign policy to a more moderate European foreign
policy.
Concluding remarks
The discussion in this essay leads to few unambiguous conclusions. But that is one of its points: the
implications of a common European international position depend in very important ways on
circumstance. The distribution of preferences within and outside the EU, along with the voting rules
used by the EU and the international institutions have a profound effect on the ways in which Europe’s
member states would aggregate their views up into a common position, and on the impact of that
common position on international affairs.
But all is not lost to scholarly indecisiveness. There are several implications of general note. Most
generally, the aggregation of European Union representatives into one EU-wide joint and common
representative brings into play a fundamental tradeoff, between the benefits of increased bargaining
power and the costs of compromise among heterogeneous interests. This, in turn, has significant
implications. European governments with more extreme views (compared to the rest of the EU) will be
less likely to support a common international position.
The greater the added bargaining power
associated with a pooled representation, the more likely EU member states are to support it. Domestic
groups will support national or joint representation based on whether their own positions are closer to
the national view or to the EU-wide view. These considerations are hardly trivial, and they will play an
important role as Europe moves toward a more federal form of representation in international politics.
25
The Euro Area and World Interest Rates
Menzie Chinn* (University of Wisconsin, Madison and NBER)
and
Jeffrey Frankel** (Harvard University and NBER)
Acknowledgements: We thank the discussant, Pierre Siklos, as well as Robert Rasche, Ignazio
Angeloni, Jakob de Haan, Petra Geraats, and the other conference participants for useful comments.
Gabriele Galati and Frank Westermann provided useful suggestions, and Thomas Laubach provided
data on CBO forecasts. Chinn acknowledges the hospitality of the Chicago Fed, where he was a guest
scholar during the writing of this paper. The views contained herein are solely those of the authors,
and do not necessarily represent those of the institutions with which the authors are associated.
* Corresponding Author: LaFollette School of Public Affairs; and Department of Economics, University
of Wisconsin, 1180 Observatory Drive, Madison, WI 53706-1393. Email: [email protected]
** John F. Kennedy School of Government, 79 JFK Street, Cambridge, MA 02138. Email:
[email protected]
26
1. Introduction
How has European monetary integration affected the determination of world interest rates?
This
question comprises a number of empirical and policy debates. The first question is: Is there a “world
interest rate”? In other words, has the dismantling of the legal and regulatory impediments to capital
controls, accompanied by decreases in transactions costs, resulted in an essentially integrated pool of
financial capital? With the rapid evolution of financial markets in the Euro area, a re-examination is
warranted. The opportunity to revisit the theme of world interest rates is particularly welcome because
it has been over a decade since the last extensive discussion of the “world interest rate” and its
determinants (Blanchard and Summers, 1984; Barro and Sala-i-Martin, 1990). Another decade of data
is always useful, and another chance to see if the 30-year trend of increasing integration across
markets has continued. But there are two more topical issues to be addressed as well.
The next question has to do specifically with the achievement of European Economic and Monetary
Union in 1999. In the past, US interest rates have had a greater influence on rates in Europe than the
influence of European interest rates on the United States, even though the European economies in the
aggregate are roughly as large as the United States – larger, if one includes the non-euro members of
the European Union.10 One explanation is that the asymmetry arose from strategic interaction between
one central bank in the United States versus 15 central banks in Europe.
The US has had a first-
mover advantage (which game theory could model as a Stackleberg equilibrium), and the European
monetary authorities have been left with “take it or leave it.” A second, not inconsistent, explanation for
the asymmetry is that it arose from the fact that the European countries have been more sensitive to
their exchange rates, because they have been more open to international trade as a share of their
GDPs (which in turn is primarily because they are smaller, and secondarily because they are close
together, while the US has fewer natural trading partners). Each of these two explanations should
have disappeared now. That the ECB now speaks for all 12 euro countries should have obviated the
first explanation. That the euro area as a whole is no longer substantially more open to trade than the
United States should have obviated the second explanation. Thus the year 2003, when we have the
benefit of four years of experience with EMU, is a good time to see if the asymmetry of US dominance
remains, or if Europe is taking on a more central role in world financial markets.11
The other topical issue concerns the role of fiscal policy. One might have thought that the debate over
whether fiscal policy affects interest rates would have been settled by now. [The authors might have
thought that the debate had been settled in the direction of rejecting Ricardian equivalence as a
practical description of the real world, with all sorts of available theoretical explanations.] But the issue
Furthermore, US interest rates have had a greater influence on third countries – especially those in the Western
Hemisphere and East Asia – than have European interest rates. E.g., Chinn and Frankel (1994) and Frankel and
Wei (1995).
11
That the United States has gone deeply into net international debtor position is another reason to ask if its
dominance over international financial markets may have diminished over time, even while its geopolitical
dominance has increased.
10
27
has taken on renewed controversy in the light of a current domestic debate in the United States
regarding the 2001 and 2003 tax cuts and the associated budget deficits. Gale and Orszag (2003)
review the literature regarding effects of current and expected future budget deficits on interest rates, 12
and conclude:
“…studies that (properly) incorporate deficit expectations in addition to current deficits
tend to find economically and statistically significant connections between anticipated
deficits and current long-term interest rates.” (p. 20)
But others strongly disagree with many of these studies, and with the Gale and Orszag’s overall
characterization of the state of empirical evidence. Looking at additional European data may shed
additional light on this unsettled debate. Perhaps the major incremental contribution of this paper is to
make a first attempt at measuring the effect of expected future euro-area budget deficits on European
interest rates, in the manner that others have done for the United States. Furthermore, it is not just the
long-run fiscal outlook in the United States that is deteriorating; many European countries face even
larger future fiscal demands from the next generation of retirees. Meanwhile, the Stability and Growth
Pact that was supposed to limit European fiscal deficits appears to be coming unravelled. Thus the upto-date international evidence on interest rate determination should be equally useful on both sides of
the Atlantic.
These two issues – trans-Atlantic monetary transmission and the effects of deficits on interest rates –
are conveniently addressed in the same study. That is because interest rates are determined by
multiple factors. Perhaps this seemingly obvious fact needs to be stated explicitly, in that the official
response of the Bush White House to critiques of its fiscal policy was that “interest rates do not move
in lockstep with budget deficits (Hubbard, 2002).” This proposition is of course true: because interest
rates are influenced by a number of factors, including most plausibly the cyclical position of the
economy, monetary policy, and international influences, interest rates can often be observed to
change at times when fiscal policy has not, even under the hypothesis that government borrowing
causes interest rates to be higher than they otherwise would. Indeed, it will turn out in our regression
analysis that conditioning on a variety of factors, including expected future deficits, cyclical position,
and cross-Atlantic influences, is essential to uncovering the effects of domestic debt.
In each case -- the nature and extent of international macroeconomic spillovers and the transmission
mechanism for monetary and fiscal policy -- the ultimate motivation concerns effects on real economic
activity. Long-term interest rates are thought to matter for economic activity more than short rates.
Indeed, expected future budget deficits should in theory matter for long-term interest rates, not short-
12
To some observers, the tax cuts enacted by the Bush Administration seem unusually designed to lose tax
revenue in the long run, relative to the fiscal stimulus delivered in the short run. Thus the distinction between
current deficits and expected future deficits may be more relevant now than in the past.
28
term rates -- the focus of most statistical studies of international linkages. Consequently, in the last
section of the paper, we will focus on long-term inflation adjusted rates.
2. Interest Rates at the Short Horizon
At first glance, the proposition that there is, increasingly, a single world interest rate seems
questionable. The nominal money market interest rates for the US, UK, Germany and Japan do not
appear to be converging. Of course, there are good reasons to expect rates to diverge, even with
perfectly mobile and substitutable capital. 13 Expected inflation rates and exchange rate changes can
introduce wedges between observed interest rates. The corresponding real interest rates (nominal
rates adjusted by previous year’s inflation) have converged, especially since the 1970’s, when large
variations in inflation occurred. However, it is not apparent that these rates have converged further
since, say, the 1980’s when many of the developed country restrictions were dismantled.
In contrast, nominal interest rates have converged for the four largest Euro area economies: Germany,
France, Italy and Spain. So we have disparate trends – increasing convergence within the Euro area,
but persistent divergence outside. Indeed, the Euro area short-term interest rate has diverged
substantially from that in the U.S., in both directions at different times. This outcome has been the
inspiration for any number of critiques and defenses of the conduct of Euro area monetary policy. We
make no comment here on this interesting subject, except to note that convergence of the levels of
nominal variables denominated in different currencies need not carry implications for differences in the
real ease of monetary policy.
We undertake an analysis of whether, statistically speaking, US nominal money market rates drive
German, vice versa, or whether both drive each other. A series of vector error correction specifications
incorporating monthly data on nominal US and German (subsequently Euro area) rates are estimated,
imposing a long run unitary elasticity between US or European rates.
For the earlier period of 1973m03-1995m12, the US rate does not move to close statistical
disequilibria while the German rate does. In the later period of 1996m01-2003m04, this same pattern
persists. In a sense, this is not the most interesting comparison; the behavior of the euro rate is the
variable of attention. Over the shorter period beginning in 1999m01 period, we find that the euro area’s
money market rate also seems to respond more than does the US rate.
3. The Evidence at Longer Maturities
13
See Frankel (1991; 1992) for a discussion. Frankel and MacArthur (1988) present estimates of the degree of
capital mobility. Froot and Frankel (1989) and Chinn and Frankel (2002) present additional interpretations.
Relatedly, Siklos and Wohar (1997) argue that the convergence of inflation rates is a key reason for the
convergence of interest rates.
29
If our motivation for considering international interest rate linkages is transmission of monetary policy,
then we should ultimately be more interested in long-term real interest rates than short-term nominal
rates. Tests of the integration of markets for bonds of longer maturity using the covered interest parity
and uncovered interest parity conditions are much less common than those at the short horizon,
largely because of the dearth of appropriate data. In principle, one requires zero coupon yields for
constant maturities.
Moreover, when it comes to statistical analysis of expected returns (for tests of
uncovered interest parity, real interest parity, or the exchange risk premium) very few non-overlapping
observations on the realized change in the exchange rate or price level are available when the
maturity or horizon is long-term.
Several recent studies have examined how well markets for government bonds are integrated using
the covered and uncovered parity conditions, at 5 and 10 year horizons (Popper, 1993; Meredith and
Chinn, 1998). They have not necessarily supported the conventional wisdom that financial markets
are less highly integrated at these longer horizons than at short horizons. However, Meredith and
Chinn, as well as Fujii and Chinn (2001) obtained point estimates that often deviated from the values
consistent with uncovered interest and real interest parity. Perhaps more important is the fact that
even some large developed countries consistently fail the ex post RIP test – including Italy. Once one
leaves the set of G-7 countries, one finds even less evidence in favor of either uncovered interest
parity or real interest parity. In Madarassy and Chinn (2002), for instance, the Spanish peseta fails real
parity tests (using forecasted inflation). Hence, how nominal and real rates of different countries covary remains an open research question.
Here, we examine quarterly data on ten-year benchmark bonds, adjusted by lagged one-year
inflation.14 Once again operating in the vector error correction framework, we assess whether US rates
(rUS) respond to European rates (r*), and vice versa. The regressions are implemented over two subsamples: 1973q1-1995q4, and 1996q1-2003q2. For the early sample, the US rate in all cases fails to
respond to European long-term real rates. In contrast, European rates respond fairly strongly. French
and Italian rates respond most rapidly, closing a real interest rate gap (up to a constant) by over ten
percent per quarter. Ignoring the short run dynamics, this means that the half life of a deviation was a
little over one and a half years.
The results for the more recent period since the end of 1995 are somewhat ambiguous. In certain
cases, the US long-term real rate appears to move to close the gap between rates: for France, Italy,
and Spain. Interestingly, in the US-Germany pair, the German rate still moves to close the gap. What
is clear is that the US rate appears to close the gap at a more rapid pace than they did in the earlier
period. For an even more recent period, we can examine the aggregate behavior of the Euro area
bond rates. In this later sub-sample, US rates adjust at a rapid rate of 0.37; euro area rates – as
measured by a weighted average of ten year bond rates – adjust at a rate of 0.12. Strong conclusions,
however, must be tempered by an acknowledgement of the imprecision of estimates.
14
This procedure follows from the results in Sack (2000).
30
Each of the individual time series seems to have only limited information regarding the dynamics of
real interest rates. In order to obtain more precision, we impose some homogeneity, and estimate
panel fixed effects regressions, once again breaking the sample into two sub-samples. The results
confirm the impression that US rates in the early period were unresponsive to disequilibria, while Euro
area rates were. In the later period, both rates appear to revert to parity with statistical significance.
Interestingly, the US rate of reversion is now slower than the European rate; but it is estimated with
much greater precision, and so enters with higher significance. The basic symmetry disappears when
we re-estimate using SUR; then only the European rates revert, and at a fairly rapid pace. The half life
(once again ignoring short run dynamics) is about 5 quarters.
These results suggest that up to 1995, US real interest rates appeared to be driven by own-dynamics,
while the real interest rates of Germany, France, Italy and Spain seemed to move to close gaps with
the US. In the post 1995 period, the evidence is more ambiguous, with the US rate responding in
some country pairs, and not in others. If we treat the aggregated European country rates as a single
rate, then both US and Euro area interest rates respond to gaps between the two rates.
4. Government Bond Markets: Integrated or Not?
Thus far in our examination of interest rates on government bonds, the analysis has relied primarily
upon the pure time series evidence. However, examination of interest rates alone tells us little about
the determinants of those rates. Here, we tackle the issue of the medium-run determinants of real
interest rates. The central issue will be the relative importance of identifiable global versus national
factors.
The question of whether interest rates are determined in national or global markets has been a source
of debate over the past few years. On one side are those who view the capital market as a single pool
of funds for the OECD countries (Ford and Laxton, 1999). Sometimes, in fact, the complete integration
of the capital markets is taken as given, as in Barro and Sala-i-Martin (1990). On the other side are
those who aver that, while global factors are important, national factors retain a key importance
(Christiansen and Pigott, 1997; Breedon et al., 1999).
One factor that is omitted from all previous cross-country studies of the interest rate-debt nexus is the
role of expectations regarding deficits and debt. We cited the Gale-Orszag survey at the outset.
Among the many relevant studies, two examples are particularly recent and relevant. Canzoneri,
Cumby and Diba (2002) find that changes in the 5 year and 10 year ahead forecasted budget deficits
result in a statistically significant increase in the spread between short term and long term interest
rates (which they interpret in light of the fiscal theory of the price level). Laubach (2003) finds robust
evidence of a relationship between 5 year and 10 year ahead projected deficits and debts and the
level of long term real interest rates in the United States.
31
In this study, the regressions we estimate are of the form:
it   0   1 t   2 d t   3 Et (d t  2  d t )   4 y   5it,W  ut
(1)
where iℓ denotes a long term interest rate, π an inflation rate, y the output gap, and Et (.) a subjective
expectations operator. In the baseline specification, we use the lagged one year inflation rate as a
proxy measure for expected long term inflation.
The logic underlying this specification is straightforward. Expected inflation becomes built into the long
term nominal interest rate.
With risk aversion, it is possible that there is an additional premium as
well. Government debt-to-GDP, in the absence of complete Ricardian equivalence, has an impact to
the extent that government financing crowds out private spending. The same argument applies to
expected future debt. The output gap enters in as a summary measure of private sector demand for
savings. Finally, the “world” interest rate enters to capture international factors. For the non-US
economies, the US long term interest rate is used; for the US, the German interest rate is used.
We compile data on the Euro area economies of the Germany, France, Italy and Spain. Taken
together, these countries comprise 80% of Euro area GDP (in 2003q1). We also examine data for the
US, UK, and Japan. The variables of interest are current and expected (2 years ahead) net debt to
GDP ratio, long term interest and inflation rates and the output gap. The net debt measures are
obtained from the OECD’s semi-annual publication Economic Outlook.
In the basic set of time series regressions, it is clear that a simple regression including only domestic
variables – that is omitting US interest rates – yields dismal results, except for the United States, and
to a lesser extent, the United Kingdom. Many of the coefficients are wrong-signed. Regressions
augmented with world factors – in this case the US interest rate for non-US countries yield results
much more in accord with the maintained hypothesis that both global factors and domestic factors
matter. In all non-US cases but two, the coefficients are correctly signed.
The short span of our data sample argues that we should exploit the information in the cross section
as well as that in the time series dimension. Consequently, we estimate fixed effects panel regressions
using pooled data for France, Germany, Italy and Spain. The results are reported in Table 1. The
inflation coefficient is now closer to its posited value, while both the debt-to-GDP ratio and projected
change are significant. The US interest rate is strongly related to the national long term rate.
Current debt has a coefficient of 0.05, while the projected change in the debt ratio obtains a coefficient
estimate of 0.11 (column 4). All coefficients are statistically significant. We also test for whether the US
interest rate is merely proxying for G-7 debt, as suggested by Ford and Laxton (1999). In our sample
(which differs from that studied by Ford and Laxton and Breedon et al.), the G-7 debt ratio has a
32
perverse sign. The expected change in the debt ratio is no longer significant, but that is a
consequence of allowing inflation rates to enter in freely.
These findings are striking, especially when placed in the context of the existing literature. However, it
is important to observe that the evidence pertains to a long period of 15 years, of which only three
years of post-EMU data are encompassed. In the most recent period, these country-specific effects
must surely have shrunk for euro area countries. So while Hartmann et al. (2003) claims that “the
integration of government bond market has advanced less than is the case for money market”, the
yield spreads are now quite small relative to pre-EMU – on the order of 10 to 30 basis points, as
opposed to multiple percentage points in 1988.
This observation suggests that Euro area wide debt
might now be more important post-EMU. Unfortunately, this effect cannot yet be discerned empirically.
5. Conclusion
Events in government bond markets returned to the fore in 2003. We expect that attention will remain
fixed upon yields in these markets in the rest of this decade. The reasons are obvious. In the United
States, the policy-induced change in the cyclically adjusted budget balance from surplus to deficit is
likely to collide with additional financing demands from the private sector as the economy recovers. In
the Euro area, the impending increases in public expenditures associated with populations that are
aging even more rapidly than in the US will also put upward pressure on debt stocks (EEAG, 2003;
European Commission, 2001) and hence interest rates.
Our analysis indicates that over the past three decades, short and long-term interest rates have been
driven more from the US side than the European side. However, since European Monetary Union went
into effect, long-term real rates in both the United States and the Euro area have tended to move in
such a manner as to close any gaps that open up between them. This is suggestive of two-way
influences, although a structural economic model is necessary to make a stronger conclusion.
Conditioning on foreign interest rates enables us to discern more sharply the domestic influences as
well. One key contribution of our study is the finding of a role for actual levels and expected changes
in national stocks of government debt over the past 15 years, thereby extending to Europe a result that
others have found for the United States. The fact that global debt stocks do not explain particularly well
the evolution of country-by-country real interest rates indicates that long term government debt is not
perfectly substitutable. Unfortunately, we are unable to determine whether this characterization has
changed since monetary union. For example, aggregate Euro area debt might now better explain the
real interest rates for the long term government debt of Euro area governments. But, for now, we have
too few observations to address this conjecture.
33
References
Barro, Robert J. and Xavier Sala-i-Martin, 1990, “World real interest rates,” NBER Macroeconomics
Annual 1990 (Cambridge, MA: MIT Press): 15-59.
Blanchard, Olivier J., and Lawrence H. Summers, 1984, “Perspectives on high world real interest
rates,” Brookings Papers on Economic Activity 2:1984: 273-324.
Breedon, Francis, Brian Henry, and Geoffrey Williams, 1999, “Long-term Real Interest Rates:
Evidence on the Global Capital Market,” Oxford Review of Economic Policy 15(2): 128-142.
Canzoneri, Matthew B., Robert E. Cumby and Behzad Diba, 2002, “Should the European Central Bank
and the Federal Reserve be concerned about fiscal policy?” in Rethinking Stabilization Policy
(Kansas City: Federal Reserve Bank).
Chinn, Menzie and Jeffrey Frankel, 1994, "Financial Barriers in the Pacific Basin: 1982-1992," Journal
of Economic Integration 9(1): 62-80.
Chinn, Menzie and Jeffrey Frankel, 2002, “Survey data on exchange rate expectations: More
currencies, more horizons, more tests,” in W. Allen and D. Dickinson (editors), Monetary
Policy, Capital Flows and Financial Market Developments in the Era of Financial Globalisation:
Essays in Honour of Max Fry (London: Routledge): 145-167.
Christensen, Hans and Charles Pigott, 1997, “Long-term Interest Rates in Globalised Markets,”
Economics Department Working Papers No. 175 (Paris: OECD).
EEAG, 2003, Report on the European Economy. (Munich: European Economic Advisory Group at
CESifo).
European Commission, 2001, Budgetary Challenges Posed by Ageing Populations: The Impact on
Public Spending on Pensions, Health and Long-Term Care for the Elderly and Possible
Indicators of the Long-Term Sustainability of Public Finances (Brussels: ECFIN, October).
Ford, Robert and Douglas Laxton, 1999, “World Public Debt and Real Interest Rates,” Review of
Economic Policy 15(2): 77-94.
Frankel, Jeffrey, 1991, "Quantifying International Capital Mobility in the 1980's." In National Saving
and Economic Performance, D. Bernheim and J. Shoven, eds., University of Chicago Press:
Chicago, 227-260.
34
Frankel, Jeffrey, 1992, "Measuring International Capital Mobility: A Review," American Economic
Review 82(2): 197-202.
Reprinted in International Financial Integration, edited by
Sylvester Eijffinger and Jan Lemmen, Edward Elgar Publishing; forthcoming 2003.
Frankel, Jeffrey , and Alan MacArthur, 1988, "Political vs. Currency Premia in International Real
Interest Differentials:
A Study of Forward Rates for 24 Countries,"
European Economic
Review 32 (June), 1083-1121.
Frankel, Jeffrey, and Shang-Jin Wei, 1995, "European Integration and the Regionalization of World
Trade and Currencies: The Economics and the Politics" in Monetary and Fiscal Policy in an
Integrated Europe, edited by Barry Eichengreen, Jeffry Frieden, and Jurgen von Hagen,
Springer-Verlag Press, New York and Heidelberg.
Froot, Kenneth, and Jeffrey Frankel, 1989, "Forward discount bias: Is it an exchange risk Premium?"
Quarterly Journal of Economics 104(1) (February), 139-161.
Fujii, Eiji and Menzie Chinn, 20001, “Fin de Siècle real interest parity,” Journal of International
Financial Markets, Institutions and Money 11(3/4) 289-308.
Gale, William G. and Peter R. Orszag, 2003, “The economic effects of long-term fiscal discipline,”
Urban-Brookings Tax Policy Center Discussion Paper No. 8 (Washington: Brookings, April).
Hartmann, Philipp, Angela Maddaloni and Simone Manganelli, 2003, “The Euro Area financial system:
Structure, integration and policy initiatives,” ECB Working Paper No. 230 (Frankfurt: ECB,
May).
Hubbard, Glenn, 2002, “Remarks,” AEI, December 10, 2002; Tax Notes 30th Anniversary (p.8).
Laubach, Thomas, 2003, “New evidence on theinterest rate effects of budget deficits and debt,”
Finance and Economics Discussion Paper No. 2003-12 (Washington: Board of Governors of
the Federal Reserve System, `March).
Madarassy, Rita and Menzie Chinn, 2002, “Free to flow? New results on capital mobility amongst the
non-G7 economies,” Santa Cruz Center for International Economics Working Paper 02-20
(Santa Cruz: University of California, May).
Meredith, Guy and Menzie Chinn, 1998, “Long horizon uncovered interest parity,” NBER Working
Paper #6797.
35
Popper, Helen, 1993, “Long-term covered interest parity—evidence from currency swaps,” Journal of
International Money and Finance 12(4): 439-48.
Sack, Brian, 2000, “Deriving inflation expectations from nominal and inflation-indexed Treasury yields,”
Finance and Economics Discussion Papers No. 2000-33 (Washington, DC: Board of
Governors of the Federal Reserve, May).
Siklos, Pierre and Mark Wohar, 1997, “Convergence in Interest Rates and Inflation Across Countries
and
over
Time,”
Review
of
International
Economics
5(1):
129-141.
36
Table 1
Determinants of European long term interest rates: inflation, debt and G-7 debt, output gap, and
foreign interest rate
(1)
(2)
(3)
(4)
(5)
(6)
1.469***
1.097***
1.167***
1.00
0.992***
1.00
(0.251)
(0.218)
(0.172)
-0.029
0.072***
0.057**
0.048**
0.138***
0.064***
(0.027)
(0.022)
(0.023)
(0.020)
(0.031)
(0.020)
0.091
0.122*
0.088
0.112**
0.081
0.130**
in debt ratio
(0.089)
(0.070)
(0.054)
(0.049)
(0.052)
(0.053)
output gap
-0.025
0.142
0.218
(0.179)
(0.163)
(0.223)
Inflation
debt ratio
expected change
(0.196)
Foreign interest
rate
1.117***
1.093***
1.138***
0.923***
(0.167)
(0.155)
(0.142)
(0.208)
G-7 debt ratio
0.380***
(0.061)
2 yr ahead G-7
-0.061
debt ratio
(0.038)
N
Adj.R2
60
60
60
60
60
60
0.76
0.88
0.88
0.65
0.88
0.65
Notes: Fixed effects regression using annual data, in levels (Newey-West robust standard errors in
parentheses). Percentage variables defined in decimal form. N is the number of observations, and
Adj.R2 is the adjusted R-squared,. *(**)[***] denotes significance at the 10%(5%)[1%] level. Country
specific intercepts for Germany, France, Italy and Spain.
37
Monetary Policy Transmission in the Euro Area: Any
Changes after EMU?
By Ignazio Angeloni and Michael Ehrmann, European Central Bank *
The paper can be downloaded at: http://www.ecb.int/pub/wp/ecbwp240.pdf
1. Introduction
Monetary policy transmission in Europe has become a much more interesting research subject lately.
EMU is a major institutional reform, likely to change the European economy in many ways. The
monetary policy transmission process (MTP) – the set of links through which monetary policy affects
the economy – is a central element in this transformation. While these changes unfold, and before they
are even fully understood, the ECB needs the best available information on how transmission works in
the new arrangement in order to conduct its policy.
In this paper we attempt a first assessment of the nexus between EMU and the euro area transmission
process, using the most recent data. At least since Lucas’ celebrated “critique”, it has been recognized
that regime changes impact on the expectations formation mechanisms and hence on the economy’s
response to policy. It is hard to think of examples where this line of reasoning could apply more
forcefully than EMU. However, it is important to bear in mind that EMU is a process, not a one-time
event. The transition to a new currency and monetary policy was something economic agents had time
to prepare for, and adjust to, over a number of years. This complicates significantly the task of
identifying causal links. In any event, there are several related questions, only partly of causal nature,
that we are interested in: Has the euro area MTP changed lately? Has it changed in coincidence with
EMU? Has it changed because of EMU? What is the direction of change?
The last question links up with a specific issue of great policy relevance, namely, that of the
differentiation of the effects of monetary policy across euro area countries. Several authors have
argued that the euro area monetary transmission process is uneven across countries, in a way that
could complicate the conduct of the single monetary policy. Others have shown that such differences,
when detected, are not robust to changes in empirical methodology and data, such that the jury is still
out.
Unfortunately, studying the monetary policy transmission process in the euro area is difficult at this
stage. Four years of data are not enough to approach the issue in a systematic way. Adapting our
strategy to the nature of the problem, we focus on a selected number of variables at the intermediate
*
The views expressed in this paper are not necessarily shared by the ECB or the Eurosystem. Correspondence:
European Central Bank, Directorate General Research, Kaiserstrasse 29, 60311 Frankfurt am Main, Germany.
Email addresses: [email protected] and [email protected].
38
stages of the transmission process, regarding specifically banking and financial markets. These
sectors are of key importance among the transmission channels. In these sectors, EMU-related
changes are likely to occur rapidly, and a wealth of high frequency and cross-sectional information is
available. This means looking inside the “black box” of the transmission process, rather than
examining directly the effect of policy on final variables such as output and prices. We regard such
direct analysis as unfeasible today. We use both descriptive evidence and formal econometric
analyses, trying to make them complementary. We focus mainly on post-EMU data, reserving earlier
samples for comparison. We make extensive use of evidence from other countries (“controls”), to
identify EMU-specific changes.
2. Evidence on the Bank-Lending Channel
As we noted, banks are potentially very relevant for the transmission mechanism, given their
overwhelming role in financial intermediation in continental Europe. Banks are also likely to have
adjusted early, and quickly, to changes in market conditions following the introduction of the euro,
given their daily operational contact with central banks and with the money market, which was fully
integrated since day one.
We proceed in steps. We first survey some qualitative evidence on cross-border banking. Foreign
diversification of banking activity would be an element suggesting that monetary transmission is
becoming more homogeneous, because it would imply that households and firms have access to an
increasingly homogeneous supply of bank products (loans and deposits, primarily) throughout the
area. It should be noted in this regard that banking markets could be contested, and virtually integrate,
even without significant cross-border activity. The converse, however, would not be true: if sizeable
cross-border business is observed, this should be prima facie evidence of integration. Evidence of
cross-border “action” would then constitute a sufficient condition, but not a necessary one. In any case,
we will combine and crosscheck the balance sheet data with information on how interest rate changes
induced by monetary policy are transmitted to bank customers.
2.1 - Evidence on bank penetration
In the paper we show data on cross-border loans in the interbank market as well as to non-banks, and
on banks’ cross-border holdings of securities over time reveals whether these market segments are
increasingly integrated. The picture that emerges differs across segments, though. Loans to nonbanks increase somewhat, although not much. The increase seems to be part of a general rising
trend, since loans to outside the euro area increase in a similar way. Hence, there seems to be no
evidence of a decisive change after EMU so far. For cross-border interbank loans, the movements are
significantly stronger, showing an increase in lending within the euro area between 1997 and 2002
from around 25 to around 33 percent; interestingly, this is not part of a general trend: interbank loans
outside the area declined during this period. However, no sign of break in the trend is visible around
1999. Even clearer movements show up for banks’ securities holdings. For the euro area, these
holdings increase from 20 to 60 percent (for non-bank issues) and from 15 to 30 percent (bank issues)
39
in the 5-year span. For Germany and for the euro area, an acceleration of the upward trend is visible in
1999. No similar trend is apparent for securities issued in the rest of the EU and in the rest of the
world.
Direct contact with foreign customers is not the only way in which banks operate abroad. Opening of
branches or subsidiaries, stock acquisitions and mergers are alternative ways, each with different
organizational, regulatory and tax implications. In the paper we look first at evidence on branches and
subsidiaries. Foreign penetration in this form is fairly stable in the period under investigation (19972001) at what seems a relatively small level (8-9 percent). The presence of foreign banks is more
sizeable in the smaller countries, which one would expect to be more open to foreign banking.
Furthermore, the level of foreign penetration found for the euro area countries lies in a similar range
than the one in the control group of Sweden, the UK and Denmark (although the numbers stand higher
for the UK, reflecting London’s importance as financial center).
A somewhat different picture emerges if one looks at M&A activity. Some acceleration in cross-border
merger activity occurs after 1999. Though domestic mergers are still prevalent, cross-country ones
within the euro area are on the rise: the yearly average is 17.7 after 1999, as against 9 before. Most of
the activity in the euro area has taken place in countries where bank concentration is lowest,
suggesting that the starting asymmetries in concentration, albeit still high, are diminishing.
In sum, it is clear that bank integration in the euro area is not advancing fast. The introduction of the
euro does not seem to have triggered revolutionary changes so far. But there are significant changes
going on. Movements look more marked after 1999 in some key measures, such as interbank flows,
securities holding, and to a lesser extent M&As. However, it remains an open question whether the
current level of bank integration is sufficient for a homogeneous transmission of monetary policy. To
answer this question, the reaction of prices of banking products to monetary policy should provide a
more direct answer. We now turn to this.
2.2 - Transmission of monetary policy to bank interest rates
In the paper we analyze monthly data on lending and deposit interest rates, across a variety of
instruments and maturities, on a comparable basis for 5 euro area countries and for the euro area as a
whole. Our aim is to study how bank lending and deposit rates react to changes in the money market
rates. These results provide answers to the set of questions posed in the introduction in two ways.
First, changes in the pass through parameters over time can tell us something about whether the
transmission through bank rates has changed. Breakpoints in 1999 are obviously a focus of attention.
Moreover, also the direction of change – e.g., more powerful transmission; or a more rapid one – is of
interest. Second, the wealth of sectoral and country data allows a variety of tests of cross-country
homogeneity. Obviously, we are particularly interested in the question of whether post-1999 the passthrough of monetary policy to bank rates has become more homogeneous across countries.
40
We find clear evidence for a stronger immediate response of bank rates to monetary policy, as well as
a stronger maximum effect for most countries, a phenomenon which is completely absent in the
control group. Moreover, the dispersion of these two parameters across the euro area countries
declines sizeably under EMU. On the contrary, we find no evidence that the speed of the pass through
to bank rates is faster, either in the euro area or in the control group. Overall, the message is: a
stronger and more cohesive transmission mechanism after 1999, but not a speedier one.
With the help of rolling window regressions, we also check whether the development that we identify is
indeed located at or around the start of 1999. The results are very suggestive. The impact response of
bank rates generally starts rising once observations after 1999 enter the regression window and keeps
on rising while new observations from the new regime are added in, a pattern that cannot be found for
the countries outside the euro area.
To conclude, the overall message from the evidence from bank interest rates seems rather clear.
There has indeed been a structural change in the way euro area banks set their lending and deposit
rates. The change is in the direction of a stronger response of bank rates to central bank signals. The
starting point is around 1999. Since this effect is not found for countries outside the euro area, we
conclude that it is caused by monetary union. There is also evidence that the estimated responses of
banks to policy signals are increasingly homogeneous across countries.
3. Evidence on the Financial Market Channels
Our analysis in the paper concentrates, first, on something that approximates the interest rate channel
(IRC) in its purest form, looking at measures of cohesion among nominal and real interest rates on
riskless assets. Subsequently, we approach one angle of the “asset price” channel, by comparing the
impact of monetary policy on different national stock markets.
3.1 - Interest rate channel
Our basic idea here is that, if the IRC has become homogeneous in the euro area after 1999, then it
must be true that the structure of riskless rates in real terms follows the same law of motion in all
member countries. This is obvious for nominal rates, which after EMU are forced to coincide by
arbitrage given the absence of exchange-rate risk. It is not obvious for real rates, which include also
expected inflation. A homogeneous comparison between pre- and post-EMU evidence can be based
on measures of co-movements of real interest rates, at different maturities, across the two time
periods. We calculate measures of interest rate co-movement (using GDP weighted bilateral variances
of the interest differential), within the euro area countries and for three control groups: the euro area
vs. the rest of the EU; the euro area vs. the US and Japan; and four main US Census regions among
themselves.
As one would expect, the variance of the nominal interest rate differentials vanishes after 1999 in the
euro area, but not between the euro area and the control groups. More interestingly, the variance of
41
the real interest differentials (in level), short and long, dramatically declines too. In the control
countries, there is also some decline, but less strong. There is no decline among the US regions. Data
in first differences show a different pattern of convergence. The main reduction in the variance is
between the first period (1990-94) and the second (1995-98), not after 1999; there is a similar decline
between euro area and other EU, but not with the other control groups. Within the US the variance of
the differential does not decrease. The overall message is mixed: there is a sharp convergence, but it
does not take place unambiguously after EMU (it depends a lot on whether levels or changes are
considered), nor does it exclusively take place among euro area members. The other EU countries
(UK, Sweden, Denmark) converge too. The developments are instead quite different if one looks other
areas (euro area vs. US and Japan, or within US).
To investigate the co-movements at different frequencies, in the paper we show spectral densities of
real interest rate differentials. It emerges that the variance is concentrated in the lowest frequency
ranges (above 6-12 months), in all three periods. There is a sharp reduction in the variance at these
frequencies as one moves towards subsequent periods, particularly after 1999, signaling that the low
frequency components of the variance are being removed. In the charts on short term and first
difference data, a hump shape shows up in 1990-94 in the spectrum between 3 and 6 months. In the
later periods the spectral density of short real rate differentials flattens, suggesting that the stochastic
processes driving short real rates in different euro area countries may approach random noise. The
typical spectral shape of a white noise (flat) is reached only after 1999, however. Similarly for longterm rates, where a hump characterizes the spectrum in the first period, and vanishes in the
subsequent periods, particularly after 1999. All this confirms that the convergence in the stochastic
processes driving real rates (short and long) took place well before 1999, but that after 1999 there was
further progress.
3.2 – Stock market channel
The stock market is a key link of the transmission mechanism according to both monetarist and
Keynesian views. Tobin’s q theory assigns to stock prices a central role in transmitting policy shocks to
firms’ investment. At the same time, stock prices also affect the consumer, through wealth effects. In
Europe, stock ownership is limited but growing fast, hence exploring this channel is important.
Furthermore, the response of stock prices reveals the markets’ view of the effects of monetary policy.
We start by analyzing how the national stock market indices respond to a monetary policy surprise.
The focus of our interest is whether the effects on national markets are sufficiently homogeneous.
There are three countries whose estimated coefficients are somewhat far from the average: Germany
and Portugal with a higher response, and Ireland, where the response shows the wrong sign. A test of
full cross-country homogeneity is therefore rejected, but comfortably accepted if Portugal and Ireland
are excluded. The test of equality across the 5 largest countries is also accepted.
42
The residual differences among the national responses could in principle be explained by a different
sector composition of national stock markets. Controlling for this, we conclude that our estimated
sector differences do not seem able to explain much of the observed national differences. The latter,
however, as we have seen, are not very large or significant to begin with.
4. Conclusions
Our analysis of changes in the transmission process with EMU has shown that banks have indeed
changed something in their behavior. The key bank decision variables, the prices on the products they
offer, have started responding to monetary policy more strongly around EMU. Of course, different
explanations for this are possible. Banks could have behaved in such way because of increased
pressure from the euro-induced new competitive environment. The evidence on certain aspects of
bank international penetration hints in this direction. There are market segments (interbank lending,
securities taking) where the move towards integration is sizeable. But other segments, more important
from the viewpoint of the transmission mechanism (like direct lending to cross-border customers;
branching and mergers abroad) lag behind.
On the other hand, other factors could also explain this change in bank behavior. Financial market
interest rates have also begun to move differently. Market rates are the main drivers of bank rates.
There is a likely link between the two. Market rates had already started to comove in the pre-EMU
period 1995-98, but the comovement strengthened after 1999. This is supporting evidence that the
process of interest rate determination, in banking and in the financial markets, has moved in the
direction one would expect. It is also evidence that the interest rate channel is a contributing factor
leading to both changes in the transmission process, and more cross-country homogeneity.
If all this holds true, then one would expect to also see changes in the transmission of monetary policy
to asset prices post-1999. Present and expected future real rates are the discount factors that
translate future income flows into current asset prices. Other things being equal, stronger interest rate
co-movements across countries should tend to generate homogeneous stock price responses across
countries. Unfortunately, our data did not allow this comparison across time, but we observed that,
post-1999, the impact of monetary policy on stock prices is broadly similar.
43
The CEPR/ESI
Seventh Annual Conference on
“The Euro Area as an Economic Entity”
Professor Hermann Remsperger
Member of the Executive Board
of the Deutsche Bundesbank
“Inflation differentials in EMU
causes and implications”
in Eltville
on Saturday, 13 September 2003
44
Ladies and gentlemen
I am to speak here today as a central banker on the subject of “Inflation differentials in EMU – causes
and implications”. But is this subject really relevant to monetary policy? After all, there are regional and
sectoral differences in price developments in every currency zone.
In a market economy, price movements reflect principally the varying conditions of supply and
demand. Why, then, should one make the inflation differentials in the euro area into a problem in the
first place? What is more, inflation differentials cannot be affected by monetary policy directly, since
there cannot be any regionally oriented monetary policy in a currency union. Monetary policy in a
currency union is uniform.
However, in the short term or even in the medium term, the impact of monetary policy may be different
from one region or sector to another and thus inflation differentials could be of relevance to monetary
policy.
It is also of relevance, because we would not have expected still to be having remarkable differences
in price developments now that the euro is in its fifth year.
The key question to be asked is “Should something be done to counteract the inflation differentials?
And if so, what?”
I The diagnosis
However, before I go into these questions, I would first like to stress that I am rather satisfied with the
Eurosystem’s stability record.
True, the average inflation rate over the past three years from 2000 to 2002 has been somewhat
outside the definition of price stability. And it has also been a full percentage point higher than during
the qualification and start-up stage between 1997 and 1999.
But, whereas the conditions in that early period were extremely favorable for stabilization policy, what
followed were exchange rate and oil price shocks, food crises, a number of increases in indirect taxes
and other things besides.
Even so, it has been largely possible to keep the Eurosystem’s stability promise. That has to be put
down as a success.
Admittedly, that is not the case for all the euro-area countries. In four of the original member states,
the average annual price rise in the past three years has been more than 3%, with the figure in Ireland
45
being as much as 4½% per year. Given rates of increase on this scale, it is undoubtedly not
appropriate to speak of price stability.
In addition, it has to be noted that, in those countries which already had an above-average rate of
inflation, prices accelerated more rapidly following the qualification and start-up phase of monetary
union. For that reason, the dispersion of inflation rates has widened again.
In terms of their range, the inflation differential in the qualification and start-up stage was less than
1½ percentage points; later, it was more than 3 percentage points.
An increase in the spread is also apparent if alternative measures of dispersion are used. The
unweighted standard deviation, say, went up from about ½ percentage point to almost 1 percentage
point. This finding is therefore fairly robust, at least if the rates over three years are averaged.
In order to understand disappointments about the extent of inflation dispersion, I would like to call to
mind again the Maastricht inflation criterion: A country is ready for monetary union only if its inflation
rate is no more than 1.5 percentage points higher than the average rate of inflation of the three
countries with the best stability performance.
In the convergence and start-up phase, the average annual rate of inflation in Austria, France and
Germany was 0.8%. At that time, the critical 2.3% mark was not being overshot by any country.
In the following phase from 2000 to 2002, the three countries just mentioned again had the lowest
inflation rates in the euro area, although the average was 1.8%. Together with the margin allowed by
the inflation criterion, this produced a critical value of 3.3%.
The higher threshold was now being overshot by no less than three founding members of monetary
union, namely Ireland, the Netherlands and Portugal.
This raises the question as to the sustainability of the convergence achieved in the qualifying and
start-up stage.
At the qualification stage, there was not much argument about the inflation criterion. It is true that
some academics pointed out that considerable inflation differentials were to be expected in a larger
monetary union.15 Among the political decision-makers, the inflation criterion was regarded as not
much of a problem, however, especially when compared with the fiscal policy criteria.
15
For example, M B Canzonerie, B Diba and G Eudey, Trends in European productivity and real
exchange rates: implications for the Maastricht convergence criteria and for inflation targets after EMU, CEPR
Discussion Paper No 1417, June 1996, and J von Hagen and M J M Neumann, Real exchange rates within and
between currency areas: How far away is EMU? Review of Economics and Statistics Vol 76 (1994), p 236-244.
46
Hardly anyone anticipated that convergence, once achieved, would weaken again under a single
monetary policy. But this is precisely what has happened.
As you know, the example of the United States is very often used to put these inflation differentials into
perspective. Although the dispersion of inflation rates in the euro area is greater than it used to be
within the individual member states, it is similar to that in the USA. 16
However, it has to be taken into account that owner-occupied housing is included in the US CPI, but
not in the European HICP. Housing thus has a much greater weight in the US CPI than it does in the
HICP. This is important because there is hardly any other component where regionally divergent price
developments have such a major impact as they do in housing.
This can be seen from the rent component of the HICP. The interregional standard deviation of the
changes in rents is greater than that of services, which, in turn, is larger than that of industrial goods
(excluding energy).
Looking at house prices, one finds average annual growth rates of 10% per year or more in Ireland,
the Netherlands and Spain. It is only in Portugal that house price developments have remained
subdued.
These considerations lead me to draw the tentative conclusion that an appropriate coverage of owneroccupied housing would result in the measured divergence of inflation rates in the euro area being
larger and perhaps exceeding the dispersion of inflation rates in the USA.
II The causes
If you now ask how this renewed widening of euro-area inflation differentials has come about, I would
first like to emphasize that, in my view, it is impossible to identify a single dominant cause. I think it
may be helpful to distinguish between “disappearing” and “reappearing” causes of inflation
differentials.
The convergence-related
components
of
the inflation differentials
may be described
as
“disappearing”.17 Convergence in the level of economic variables necessarily involves temporary
16
European Central Bank, The dispersion of inflation across the euro area countries and the US
metropolitan areas, Monthly Bulletin, April 2003, p 22-24.
17
Empirical studies suggest, firstly, that a certain relationships exists between the inflation differentials in
the euro area and the output gaps. Secondly, it is found that prices rise more quickly in countries with a low price
level. It is thus likely that cyclical and convergence-related factors operate together. See European Central Bank,
Inflation differentials in a monetary union, Monthly Bulletin, October 1999, p 35-44, and later, P Honohan and
P R Lane, Divergent inflation rates in EMU, Papers presented at the 37th Economic Policy Panel Meeting,
11th/12th April 2003, Athens. Nevertheless, it is shown by J H Rogers, Monetary union, price level convergence,
47
differences in the rates of change in the relevant variables. Borrowing from the literature on growth,
one could speak here in a broader sense of beta convergence, which tends to result in always the
same countries showing either high or low rates of inflation.
In this context, the direct convergence of prices and interest rates through arbitrage across regions
should be mentioned first. Nominal convergence has undoubtedly been reinforced by the elimination of
the residual foreign exchange risk since the beginning of 1999.
The dwindling of the risk premia brought about a largely uniform interest rate level on capital markets.
In some countries, this is likely to have generated a substantial cyclical stimulus. An important factor
was the rapid rise in property prices encouraged by the convergence of interest rates.18
On the product markets, too, greater price transparency and the elimination of exchange rate risk have
assisted price convergence. Nevertheless, the euros contribution to the integration of the product
markets in Europe should not be overestimated either.
What was more important was the opening of the markets as a result of the Internal Market Program.
This had already led to a strong convergence of prices for traded goods in the first half of the 1990s. In
1999, according to Rogers, the dispersion of such prices had already ceased to be greater than in the
USA.19 Since then, there has been hardly any further narrowing in the dispersion of prices, 20 and the
dispersion of rates of price change remained approximately unchanged.
By contrast, the convergence of prices for non-traded goods – above all, services – seems to have
accelerated. By and large price increases were higher in countries with below average price levels.
The dispersion of inflation rates for services, measured by the (unweighted) standard deviation, almost
doubled between the qualification and start-up phase and the period of 2000-2002.
How has this stronger convergence of prices for non-traded goods come about? The traditional
approach to explaining such indirect convergence is the often-cited Balassa-Samuelson model.
However, this approach has had a somewhat mixed fate recently.
and inflation: How close is Europe to the United States? Board of Governors of the Federal Reserve System,
International Finance Discussion Paper No 740, October 2002, that, in elaborate econometric specifications, the
influence of the price gap is no longer statistically different from zero.
18
P Honohan and P R Lane, Divergent inflation rates ..., op cit.
19
However, according to the European Central Bank, Price level convergence and competition in the euro
area, Monthly Bulletin, August 2002, p 39-49, the cross-country dispersion of prices in the euro area is still
higher than the dispersion within individual countries.
20
J H Rogers, Monetary union ..., op cit.
48
Initially, a very prominent role was assigned to the Balassa-Samuelson mechanism in explaining the
inflation differentials in Europe.21
Later, it was pointed out that, since the second half of the 1990s, the relationship between the changes
in the productivity gap and the inflation differentials was no longer very strong. 22 Various modifications
of the basic model have been proposed, which can not only loosen but also strengthen the
relationships between productivity convergence and inflation divergence. 23
I cannot go into the details of that debate today. But I remain convinced that the Balassa-Samuelson
approach gives us important insights into the price dynamics of what is still a heterogeneous currency
area. At the same time, it is my view that the recent inflation differentials are too large to be explained
solely in terms of differences in productivity growth.
Certainly, besides the cyclical stimuli triggered by interest rate convergence, country-specific factors
and asymmetrical shocks have also had a major impact on inflation differentials.
This brings me to the reappearing component of the inflation differentials. Quite diverse factors can
lead at any time to new inflation differentials. They increase the inflation rates’ variance without
necessarily determining the country-ranking.
One major reason for continuing inflation differentials consists in price reactions to constantly recurring
regional supply and demand shocks.24
Especially in a monetary union such as the euro area, where the regional labor markets, owing to
different languages and social security systems, are so far not very closely interlinked, the price
21
For an overview of the various contributions, see M Kieler, The ECB's Inflation Objective, IMF
Working Paper WP/03/91, May 2003.
22
See, for example, P Honohan and P R Lane, Divergent inflation rates ..., op cit, or E Ortega, Persistent
inflation differentials in Europe, Banco de Espana, Documento de Trabajo No 0305, 2003.
23
See inter alia J Strauss, The influence of traded and nontraded wages on relative prices and real
exchange rates, Economics Letters Vol 55 (1997), p 391-395; R MacDonald and L Ricci, PPP and the Balassa
Samuelson Effect: The role of the distribution sector, IMF Working Paper WP/01/38, March 2001; C Fischer,
Real currency appreciation in accession countries: Balassa-Samuelson and investment demand, Economic
Research Centre of the Deutsche Bundesbank, Discussion Paper 19/02, July 2002; T Unayama, Product variety
and real exchange rates: The Balassa-Samuelson model reconsidered, Journal of Economics Vol 79 (2003), p.
41-60; P R Bergin, One Money one price? Pricing to market in a monetary union, mimeo, May 2001,
forthcoming in European Economic Review.
24
Such shocks may potentially explain quite large inflation differentials, as has been shown by M Duarte
and A L Wolman, Regional inflation in a currency union: fiscal policy versus fundamentals, European Central
Bank, Working Paper No 180, September 2002. Similar results are also arrived at by J Andrés, E Ortega and
J Vallier, Market structure and inflation differentials in the European Monetary Union, Banco de Espana,
Documento de Trabajo No 0301, 2003.
49
mechanism has a major part to play in balancing supply and demand.25 Short-term inflation
differentials can help to stabilize output and the labor market.
However, this mechanism can function smoothly only if wages are sufficiently flexible. If overall
inflation is low, this may call for wage cuts in some regions or sectors.
In the past, adjustments of this kind were not always observable on the scale that would have actually
been necessary.26 In the meantime, variable wage components have gained in importance. Even so,
the possibility cannot be ruled out that a certain asymmetry between the upward and downward
flexibility of wages will persist, thus limiting the effectiveness of the wage mechanism.
Besides country or sector-specific shocks, shocks which impact on the currency area as a whole may
also have varying price effects. For one thing, different regions may be affected in different ways owing
to their pattern of specialization. For another, they may react differently, say, owing to differences in
market structures27 or even in how expectations are formed.
Longer-term empirical studies show that expectation formation in the wage and price-setting process
in Germany tends to be forward-looking, whereas, in other countries, it is geared more to the past. 28
This can have a significant impact on price dynamics and give rise to inflation differentials. For
example, an oil price shock that impacts on wages in one of the countries because expectation
formation is oriented to the past, while this is not the case in other countries, will temporarily open up
inflation differentials.
Shock-like exchange rate movements are of major importance for short to medium-term price
dynamics in the euro area. Although there are no longer any nominal exchange rate movements within
the euro area, that is not the case vis-à-vis the dollar area.
The direct impact of exchange rate shocks on consumer prices is determined by the share of imported
goods in private consumption and the strength of the pass-through effect. That effect is powerful and
fast for oil products, but very much weaker and slower in the case of other goods. Owing to differences
25
This is stressed principally by the OECD, Inflation persistence in the Euro area, OECD Economic
Outlook 72, 2002, p 163-171. See also European Central Bank, Inflation differentials ..., op cit.
26
For Germany, see C Knoppik and T Beissinger, How rigid are nominal wages? Evidence and
implications for Germany, Institute for the Study of Labour (IZA) Discussion Paper No 357, September 2001,
forthcoming in Scandinavian Economic Journal, 2003, and C Knoppik and J Dittmar, A semi-parametric
analysis of downward nominal wage rigidity in the GSOEP 1984-2000, University of Regensburg Discussion
Paper No 374, September 2002.
27
J Andrés, E Ortega and J Vallier, Market structure and inflation differentials ..., op cit.
28
For example, P Benigno and J D Lopez-Salido, Inflation persistence and optimal monetary policy in the
euro area, Board of Governors of the Federal Reserve System, International Finance Discussion Paper No 749,
July 2002.
50
in the composition of private consumption, price developments may therefore differ in the short to
medium term even with uniform shocks.
Hüfner and Schröder, for example, point out that the short-term effect in the Netherlands, with its high
share of imports from non-European countries, is much greater than in Germany. 29
In addition to the direct impact of exchange rates on consumer prices, there may be a cyclical effect
via a change in the competitive position, which is also determined by an economy’s degree of
openness to countries outside the euro area. This is likely to have been a major factor for Ireland, in
particular, over the past few years.
Permit me to summarize my brief analysis of causes as follows. No single factor can be identified as
the cause of the inflation differentials. Instead, there are a number of different reasons that are not of
equal significance for every country in the euro area.
Principally, there are three factors, which I regard as important.
Firstly, the process of nominal convergence, especially in those countries, which had relatively high
nominal interest rates prior to European monetary union, created an expansionary stimulus.
Secondly, the real catching-up process of those euro-area countries with comparatively low
productivity also seems to have contributed to the inflation differentials.
Thirdly, country-specific factors have generated major stimuli to price developments. Chief among
those factors was the varying impact of the euros temporary weakness.30
The heterogeneous nature of economic policy measures, especially with regard to tax policy and the
regulation of network industries, should certainly also be included among the country-specific features.
With a view to country specific factors we have to note that in some countries wage moderation during
the convergence stage later gave way to strong wage increases, which intensified the divergence of
inflation rates.31
In my opinion there are good reasons for expecting that the inflation differentials will not simply
disappear. Perhaps they will be somewhat smaller in the future. That is particularly true for the inflation
differentials, which have been caused by the convergence process.
29
F P Hüfner and M. Schröder, Exchange rate pass-through to consumer prices: A European perspective,
ZEW Discussion Paper 02-20, 2002.
30
See, for example, P Honohan and P R Lane, Divergent inflation rates ..., op cit.
31
See P Honohan and P R Lane, Divergent inflation rates ..., op cit, for a discussion of the Irish
experience.
51
Experience in the United States, however, indicates that the decline may not be substantial. Empirical
studies show that the dispersion of inflation rates in the USA is relatively stable over time.
With these remarks I am already approaching the last part of my talk: What are the implications of
inflation differentials for business activity and economic policy?
III The implications
I shall deal with the question of their impact on economic activity first. Then, I shall address specific
areas of policy: fiscal policy, structural policy and, last but not least, monetary policy.
The first question is about the asymmetric regional impact of a single nominal interest rate owing to
inflation rate differentials across countries. With regard to Germany, in particular, it is often stated that
the real interest rate is too high from a German point of view and that, in view of the low inflation rate,
monetary conditions are too restrictive in this country. 32
I do not agree with this perspective. As I see it, changes in the international competitive position
stemming from inflation differentials are in the long run more important than the differentials in real
interest rates. Even so, viewed in the shorter term, differentials in real interest rates can indeed
generate considerable effects.
In a monetary union, money market rates in all the participating countries are more or less identical.
Inflation differentials thus create diverging short-term real interest rates: those countries with a
comparatively low rate of inflation have a higher real short-term interest rate and vice versa. This
reinforces the original disinflationary or inflationary tendencies.
The extent to which longer-term real interest rates adjust to inflation differentials hinges crucially on
the inflation expectations in the participating countries. High regional inflation rates are often reflected
in higher inflation expectations.
However, inflation expectations do not change in step with the current inflation rate. There is a
tendency to expect a return to the mean value or to the inflation rate aimed at by the central bank. The
real interest rate differentials are thus somewhat lower than the inflation differentials.
See, for example, M Feldstein, Britain must avoid Germany’s mistake, Financial Times of 21 April
2003 or Deutsches Institut für Wirtschaftsforschung (2003) Grundlinien der Wirtschaftsentwicklung 2003/2 in
DIW-Wochenbericht 1-2/03.
32
52
In August this year, for example, the inflation rate in Spain was 1.8 percentage points higher than in
Germany, while the long-term real interest rate based on inflation expectations was only 1 percentage
point above the German level.
Added to this is the fact that countries with marked inflation rates are often faced with high property
prices. If the resulting increase in the nominal wealth impacts on aggregate demand in the country
concerned, the inflation-accelerating effect is strengthened further.
These destabilizing processes are countered by a stabilizing mechanism, however. Inflation
differentials between economies in a monetary union lead to changes in their relative prices or rather
their real exchange rates.
Owing to a real “appreciation”, countries with higher-than-average inflation rates suffer a loss in price
competitiveness, while countries with relatively low inflation rates gain in price competitiveness. The
consequence is that export demand in the countries with higher inflation rates tends to decline, which
has a dampening impact on price developments in those countries. Conversely, demand tends to
increase in countries with lower inflation rates.
This effect also occurs in relationship to those countries which do not belong to the monetary union.
Since the euros (flexible) nominal exchange rate against the currencies of such countries is geared to
economic developments of the euro area as a whole, euro-area countries with an above-average
inflation rate suffer a deterioration in their competitive position vis-à-vis non-euro-area countries. Euroarea countries which have a below-average inflation rate gain in general.
The relationship between the destabilizing real interest rate effect and the stabilizing exchange rate
effect is ultimately an empirical question. Studies show that in the long-run the terms-of-trade effect
seems to be more important than the real interest rate effect. However, among the individual countries
there is a marked heterogeneity concerning the relative weight of the two channels.
While simulations for Germany, and those for the USA, suggest that the effect on competitiveness is
predominant, in other countries the real interest rate effect predominates over the competition effect
and thus – temporarily – works in a procyclical way.33
Additionally, we can not rely on the stabilizing effect of inflation differentials if one member state of the
monetary union finds itself in the process of a persistent decline in the general price level. In
particular, it is likely that the slump in aggregate demand, brought about by a protracted large-scale
slide in prices, will set off mechanisms that result in self-reinforcing deflationary developments. And
these may can even spread to other countries of the monetary union.
33
For the USA see I Arnold and C Kool, The role of inflation in regional adjustment: Evidence from the
United States, mimeo.
53
However, I want to stress explicitly, that a scenario of this kind is hardly likely to occur in the euro
area. Germany – which, in the recent debate, has been most under suspicion of slipping into a
deflation scenario – shows no signs of an impending deflation. 34
Recently, the HICP went up by 1 percent year over year. A declining price level is not expected in this
country for the future either. On the contrary, for the next years the economic agents are anticipating a
more or less constant inflation rate of 1½%.
Additionally, we should not overlook the fact that euro-area countries with higher inflation rates
also have to bear higher costs of inflation. Such costs include menu costs, increasing
distortions in tax systems based on the nominal principle and a sub-optimal cash holding.
Even with low inflation rates, these costs can be considerable. 35
It is definitely in the own interest of the countries in the euro area to keep their respective rates
of inflation at a low level, since they also have to bear the costs of inflation. 36 After all, this
includes preventing their own export industry from suffering the damage which may result from
a persistent loss of competitiveness and which may permanently reduce the country’s longerterm growth opportunities.
This applies just as much to the euro-area accession countries. Convergence of their inflation rates to
the euro-area level reduces their inflation costs. Furthermore, within the enlarged monetary union it
will ease potential tensions between countries with a low inflation rate and countries that have higher
rates of price increase. Thus, convergence of inflation rates at a low level before entry into EMU can
help to strengthen the stability of the monetary union.
If the destabilizing impact of inflation differentials threatens to predominate or if a country’s inflation
rate is felt to be too high, the question arises as to what economic policymakers could do to limit
inflation differentials.
In the world of the traditional Mundell-Fleming model of a single currency area, which is characterized
by recurring stochastic supply and demand shocks, the answer is clear: the sole appropriate
34
Regarding this assessment, see also The debate on deflationary risks in Germany in the June Monthly
Report of the Deutschen Bundesbank, p 15–28.
35
Lucas, R. E. (2000), Inflation and Welfare, in: Econometrica, Band 68, Nr. 2, S. 247-274.
36
This point is stressed by Feldstein, M. (2001), Economic Problems of Ireland in Europe, NBER
Working Paper 8264, May 2001.
54
stabilizing cyclical policy instrument in a monetary union is fiscal policy, which has the task of
smoothing fluctuations in output.
In the two-country model, the stabilization of output simultaneously results in smaller fluctuations in
the respective inflation rates. In this context, the price effects show up, so to speak, as a side-effect of
a fiscal policy which is stabilizing output.
However, there are two main arguments against the deployment of fiscal policy to limit inflation
differentials. Firstly, the empirical evidence for the price-stabilizing impact of a discretionary fiscal
policy tends to be weak.37 The general limitations of an anticyclical fiscal policy due to inside and
outside lags are well known.
Secondly, we have to take due account of the fact that at least part of the inflation differentials is not
stochastic, but rather convergence-related. The convergence process itself might be protracted if the
economies which are catching up with the advanced economies were to respond to this “structural”
component of the inflation differentials with a persistently restrictive fiscal policy.
At the same time, however, it is imperative to ensure that fiscal policy does not itself lead to a
widening of inflation differentials. For example, an accommodating fiscal policy has been given as a
reason for the relatively high inflation rates in some euro-area countries.
In the debate on the economic policy implications of inflation differentials the call for structural
reforms is often heard. Structural policy measures, such as the deregulation of major product
markets and the labor market, make it easier for an economy to respond to shocks, which are
among the causes of the dispersion of inflation rates in the euro area.
In addition, structural policy measures could lead to temporary price changes in the affected
product markets and these changes can have an impact on inflation differentials. One
example of this is provided by the market for telecommunications, in which increased
competition has resulted in several years of ongoing downward price pressure.
37
R Perotti (2003), Estimating the Effects of Fiscal Polities in OECD Countries, Journal of the European
Economic Association, forthcoming and M Duarte and A L Wolman (2002), Regional Inflation in a Currency
Union: Fiscal Policy vs. Fundamentals, ECB Working Paper No 180.
55
In the case of the countries which are catching up with the advanced economies, deregulation
might indeed lead to a dampening of inflationary developments, especially if they dampen
price rises in the services sector.
Admittedly, not all structural policy measures have the effect of reducing inflation. In certain
cases, they might even lead to higher prices. This could be the case if the still heavily
regulated housing markets in many of the participating countries were to be liberalized.
Using structural policy measures as an instrument for a reduction in inflation differentials is
therefore subject to certain qualifications, especially as their price-dampening impact may
even reinforce inflation differentials in countries with low inflation rates. When I say this I do
not want to question the general importance of deregulation policy measures. The opposite is
true.
Wage and labor market policies can help to reduce inflation differentials caused by demand or supply
shocks. A higher degree of labor force mobility between countries as well as sectors can ease the
adjustment pressure on regional wages. Making the underlying conditions on the labor market more
flexible, particularly with regard to the wage formation process, should therefore be an essential item
on the economic policy reform agenda.
To my mind, wage policy has a particular responsibility in terms of responding to economic shocks
which impact directly on prices. A temporary increase in the general price level – say, as a result of
higher oil prices or higher indirect taxes – should not lead to second-round effects. Forward-looking,
stability-oriented wage agreements will help to prevent the inflation differentials from becoming
entrenched.
In contrast to wage policy the possibilities of monetary policy to counter inflation differentials are very
limited. As a policy for the entire currency area monetary policy is not in a position to distribute its
effects among the participating countries. Hence, it is unable to exert an influence on the inflation
differentials.
Even so, there are two ways in which the ECB takes account of the inflation differentials and the
prevention of deflationary tendencies. First, the Governing Council has created monetary conditions
which allow adequate scope for real economic growth.
56
Second, the Governing Council, in its recent clarification of its monetary policy strategy, spelled out its
stability policy objective in terms of a medium-term euro-area inflation rate of “just under 2%”. Given
the current inflation differentials, that definition, in my view, provides an adequately large safety
margin against deflation for all the member countries. And with this definition the ECB explicitly took
account of the implications of inflation differentials within the euro area. 38
Now, please allow me to summarize my talk in five points. First, I would like to state that
regional inflation differentials in a currency area are a widespread phenomenon. In the euro
area the underlying driving forces will change over time. The convergence-related causes will
peter out. Thus, recurrent supply or demand-side shocks will gain in relative importance.
Empirical findings from other countries speak against the expectation that the inflation
differentials will simply disappear.
Second, I want to stress the point, that as a general rule, the real economic effects of
divergent inflation rates with a single monetary policy are self-correcting in the long-run.
However, in the short-run a widening of the heterogeneity between the participating countries
can occur. The inflation differentials in the qualification and start-up stage of the euro area
were smaller than in the last three years.
Third, it is imperative not to overlook the fact that high inflation rates give rise to
macroeconomic costs in the individual countries concerned. Moreover, persistent inflation
differentials may lead to tensions in economic policy in the euro area.
Fourth, while the single interest rate policy of the ECB cannot take account of regional
developments, national economic policies can increase the capability of national economies to
cope with demand and supply shocks. Wage policies in particular have a key responsibility for
a smooth adjustment to shocks. Forward looking and productivity oriented wage agreements
would help to prevent inflation differentials from becoming entrenched.
“To maintain inflation rates close to 2% over the medium term ... underlines the ECB’s commitment to
provide a sufficient safety margin to guard against the risks of deflation. It also addresses the issue of the
possible presence of a measurement bias in the HICP and the implications of inflation differentials within the
euro area.” (ECB press release „The ECB’s monetary policy strategy‘, 8 May 2003)
38
57
Fifth, and with regard to EMU-enlargement it is crucial for the acceding countries to have
achieved a sufficient degree of convergence before joining the single currency area. 39
39
The Bundesbank has already made repeated reference to this (see, for example, Deutsche Bundesbank
(2001), Monetary aspects of the enlargement of the EU, Monthly Report, October 2001, p 15-30).
58