Eurozone cycles: An analysis of phase synchronization

• Author: Sana Hussain,Brigitte Granville
• DOI: http://doi.org/10.1002/ijfe.1576
• Date: 01 April 2017
• Published date: 01 April 2017

RESEARCH ARTICLE

Eurozone cycles: An analysis of phase synchronization

Brigitte Granville | Sana Hussain

Queen Mary University of London, Mile End

Road, London E1 4NS, UK

Correspondence

Brigitte Granville, Queen Mary University of

London, Mile End Road, London E1 4NS,

UK.

Email: b.granville@qmul.ac.uk

JEL Classification: C14; E32; E44

Abstract

This paper analyses synchronization, both across and between business and

financial cycles (growth and classical) in a subset of 10 countries representative of

the Economic and Monetary Union. Employing an extended data set from 1960 to

2013, we find evidence of synchronization across financial cycles. In case of

business cycles, we find contrasting results: There is significant synchronization

across growth cycles but no evidence of a common classical cycle. This confirms,

first, that economic and financial variables in the Economic and Monetary Union

behave differently and, second, that synchronization in business cycles arises from

synchronized deviations from the trend, but the underlying macroeconomic

fundamentals are not in synch. Furthermore, we adopt a novel approach to break

down our full sample period into smaller subperiods to follow the evolution of

synchronization over time. Our results highlight the role played by the monetary

union in further increasing macroeconomic divergences.

KEYWORDS

business cycles,concordance, Eurozone, financialcycles, macroeconomic divergences/monetary policy,

time‐frequencyanalysis

1|INTRODUCTION

This paper evaluates business and financial cycle synchroni-

zation in a subset of 10 countries representative of Europe's

Economic and Monetary Union (EMU) for the period

1960–2013. Employing an extended dataset covering periods

both before and after the introduction of the euro, we exam-

ine both classical and growth cycles in order to analyse syn-

chronization in recessions and expansions, as well as

synchronization, in high‐and low‐growth periods. In contrast

to the existing literature, this allows us to further probe

whether the observed patterns of synchronization arise from

a comovement of output level or output gap, that is, whether

it is the classical cycle that exhibits synchronization or the

growth cycle. Our paper also augments the existing studies

by examining synchronization in smaller subperiods to fol-

low its evolution more closely. We adopt a novel approach

in breaking down the full sample time series into smaller sub-

periods based on graphical analysis of the mean and standard

deviation of the relevant time series.

Our aim is to assess whether with the introduction of the

euro, EMU members spend more time in the same cyclical

phase as was posited by some authors at the time of its incep-

tion such as Artis and Zhang (1997) and Frankel and Rose

(1998). Others such as De Haan et al., (2008: 265) have

argued that the monetary union can result in less business

cycle synchronization as the exchange rate can no longer

act as a shock absorbing mechanism, meaning that all adjust-

ments must be borne by the real economy, with Eurozone

(EZ) members relying on either export (Germany) or domes-

tic demand (Spain and France) as drivers of economic recov-

ery (McCarthy, 2006).

This question is of importance in a monetary union as it is

linked to the effectiveness of implementing common counter-

cyclical economic policies in response to the euro crisis to

revive economic growth. If business cycles are not synchro-

nized, a one‐size‐fits‐all monetary policy may not be optimal

as some countries will be in the contraction phase while others

will be in the recovery phase of theircycles. The existenceof a

common business cycle is under debate (partly because of the

use of different data and methods), but the consensus is that

there is no common business cycle in the EZ and that the eco-

nomic trend is one of divergence rather than convergence

(Gayer, 2007; Hallett & Richter, 2008). De Haan et al.

Received: 4 June 2015 Revised: 26 November 2016 Accepted: 12 February 2017

DOI: 10.1002/ijfe.1576

Int J Fin Econ. 2017;22:83–114. Copyright © 2017 John Wiley & Sons, Ltd.wileyonlinelibrary.com/journal/ijfe 83

(2008: 266) determine in their business cycle synchronization

survey that business cycles in the EZ are “substantially out of

synch”and that there is no movement towards the “emergence

of a ‘European’business cycle.”Similarly, Beine, Candelon,

and Hecq (2000) find no common cyclical features in the

EZ, and Kose, Otrok, and Whitman (2003) find no evidence

of a European cycle. Camacho, Perez‐Quiros, and Saiz

(2006) also conclude that the establishment of the EMU has

not increased the levels of comovement across these econo-

mies. Artis, Marcellino, and Proietti (2005) observe low signs

of synchronization across the EZ.

Any such divergence would affect the sustainability of the

monetary union especially if the dynamics of financial cycles

are considered (Borio, 2014). Therefore, drawing on

Claessens, Kose, and Terrones (2012), we also explore the

interaction between business and financial cycles. The syn-

chronization of business and financial cycles within each

country can magnify fluctuations via a feedback mechanism

where financial variables affect real variables and vice versa

(Claessens et al., 2012). Transmission of this impact via the

common monetary mechanism can then influence the EZ as

a whole. For instance, if economic recessions are accompa-

nied by financial downturns, the adverse impact may be more

severe and prolonged (Claessens et al., 2012). It is, therefore,

of prime importance to further deepen our understanding of

the link between business and financial cycles (Borio,

2014; Claessens et al., 2012; Egert & Sutherland, 2014).

For this reason, our paper also analyses the interaction

between business and financial cycles, which has not been

done exclusively for the EZ before.

Following this introduction, Section 2 presents our data

and methodology. Concordance indices are used to investi-

gate the extent of comovement in economic cycles.

Section 3 examines the synchronization of the different phases

(recessions and expansions) of classical business cycles

(measured by industrial production) across our sample of

EMU members, the different phases (downturns and upturns)

of classical financial cycles (measured by equity prices) across

those same countries participating in EMU, the synchroniza-

tion of the different phases (high rate and low rate) of the

growth version of business and financial cycles, and the con-

cordance between business and financial cycles. We test

whether this comovement or concordance is statistically sig-

nificant and whether it has intensified or diminished over

time. Section 4 provides concluding comments.

2|MEASURING SYNCHRONIZA-

TION IN CYCLES

Observed patterns of synchronization depend on the choice

of measurement methods. The choices involved here con-

cern: the cycle (classical vs. growth and cycles in growth

rates), the concordance (correlatio n vs. concordance index),

and the detrending technique (linear, band pass, or high pass;

Hallett & Richter, 2008: 73). We follow Hodrickand Prescott

(1997: 2) in assuming “that no one approach dominates all

the others and that it is best to examine the data from a num-

ber of different perspectives.”Our methods consist of, first,

detecting cycles by identifying turning points and, second,

determining synchronization by calculating concordance

indices. This methodology was developed for studying busi-

ness cycles and applied to financial cycles by Claessens

et al. (2012), Drehmann, Borio, and Tsatsaronis (2012) and

Pagan and Sossounov (2003).

For detecting cycles, we first focus on classical cycles,

which consider the level of the underlying time series. Clas-

sical cycles have been defined by Burns and Mitchell

(1946: 3) as the sequential patter n of expansions (the time

period from a trough to a peak) and recessions (the time

period from a peak to a trough) in the level of economic

activity, with the rider that “this sequence of change is recur-

rent but not periodic.”The Burns and Mitchell rules guided

the National Bureau of Economic Research procedure for

producing the reference dates of the business cycle for the

United States. For classical financial cycles, the expansion

phase is termed an upturn and the contraction phase a down-

turn (Claessens et al., 2012: 180).

Second, we focus on growth cycles, which involve

removing the permanent component (trend) from the under-

lying time series. Growth cycles are defined by Kydland

and Prescott (1990) as the deviation of the variable of interest

from its long‐term trend. Although classical and growth

cycles are related, the growth cycle measures the upward

and downward deviation of economic or financial activity

from its long‐term trend rather than the level. Therefore, con-

trary to classical cycles, the trend and cyclical components

have to be separated for identifying growth cycles. This

requires identifying “the factors determining long run

economic growth from those determining cyclical fluctua-

tions”(Stock & Watson, 1999: 9). However, breaking down

the relevant time series into trends and cycles is not easy as

both the trend and cycle influence each other and an appro-

priate filtering technique is required.

2.1 |Data

The data were obtained from the OECD statistics database for

10 countries selected as representative of the present‐day EZ:

Austria, Belgium, France, Germany, Greece, Ireland, Italy,

the Netherlands, Portugal, and Spain.

For business cycles, we use 10 seasonally adjusted

monthly time series of the industrial production (IP) index

from 1960:1 through 2013:12. The data measure volume

changes over time as indices, seasonally adjusted with 2010

as the base year. IP refers to the volume of output generated

84 GRANVILLE AND HUSSAIN

by production units grouped into industrial sectors (such as

mining, manufacturing, and electricity gas and water) in line

with the International Standard Industrial Classification of all

economic activities. We reject the view that IP cannot be used

as a proxy for total output because “manufacturing activity

represents less than 20 per cent of aggregate output in the

Eurozone”(De Haan et al., 2008: 236) given that most of

the cyclical variation in the EZ economy is explained by

the industrial sector (Gayer, 2007: 2) and a “historically

strong correlation between IP and GDP data”Gayer (2007:

7) has been observed. Moreover, most of the related analyses

are based on IP data, for example, Artis and Zhang (1997),

Artis, Krolzig, and Toro (2004), Camacho et al. (2006), Har-

ding and Pagan (2006), Inklaar and De Haan (2001), and

Massmann and Mitchell (2004), not only because IP data is

available at a monthly frequency from 1960 but also because

it displays more cyclical sensitivity than gross domestic prod-

uct estimates. This allows greater precision in measuring

business cycles by capturing more of the high‐frequency

fluctuations.

In contrast to business cycles, there is no obvious mea-

sure for financial cycles (Borio, 2014). Related literature

identifies financial cycles in three distinct but interdependent

market segments, namely, credit, residential real estate, and

equity prices (Claessens et al., 2012). We follow Pagan and

Sossounov (2003) in using equity prices as they exhibit

greater volatility featuring more upturns and downturns.

These are also available at a monthly frequency for a longer

time period and therefore facilitate greater precision in iden-

tifying cycles. For financial cycles, we consider 10 monthly

time series of the share price index from 1957:1 through

2013:12. The OECD database defines the share price index

as the prices of companies traded on national or foreign stock

exchanges. The share price index is an indicator of fluctua-

tions in the equity market and can be viewed as a proxy for

fluctuations in the overall financial markets. Monthly data

are simple arithmetic averages of the closing daily values

with 2010 as the base year.

2.2 |Identifying turning points

2.2.1 |Classical cycles

For identifying classical cycles, we employ the business

cycle dating algorithm developed by Harding and Pagan

(2002) using the insights of Bry and Boschan (1971) set

out in Table 1. We apply this algorithm to the natural loga-

rithm y

t

of the monthly index of IP Y

t

over the years

1960–2013 and the natural logarithm p

t

of the monthly index

of share prices P

t

over the years 1957–2013. Our censoring

rules follow the National Bureau of Economic Research def-

initions: cycles must have a minimum length of 15 months,

phases must have a minimum length of 6 months, and the

window over which local maxima (peaks) and minima

(troughs) are computed is 5 months.

A classical business cycle peak is said to occur at time tif

yt>yt−5;yt−4…yt−1

ðÞ

and yt>ytþ1;ytþ2…ytþ5



and a trough occurs at time tif

yt<yt−5;yt−4…yt−1

ðÞand yt<ytþ1;ytþ2…ytþ5



:

Similarly, a classical financial cycle peak occurs at time t

if

pt>pt−5;pt−4…pt−1

ðÞand pt>ptþ1;ptþ2…ptþ5



and a trough occurs at time tif

pt<pt−5;pt−4…pt−1

ðÞand pt<ptþ1;ptþ2…ptþ5



:

2.2.2 |Growth cycles

For identifying growth cycles, we first filter y

t

and p

t

before

applying a modified version of the Bry–Boschan algorithm to

TABLE 1 Bry Boshan (BB) procedure for programmed determina-

tion of turning points

Step Procedure

1. Determination of extremes and substitution of values.

2. Determination of cycles in 12‐month moving average (extremes

replaced).

A. Identification of points higher (or lower) than 5 months on

either side.

B. Enforcement of alternation of turns by selecting highest of

multiple peaks (or lowest of multiple troughs).

3. Determination of corresponding turns in Spencer curve

(extremes replaced).

A. Identification of highest (or lowest) value within 5 months

of selected turn in 12‐month moving average.

B. Enforcement of minimum cycle duration of 15 months by

eliminating lower peaks and higher troughs of shorter cycles.

4. Determination of corresponding turns in short‐ter m moving

average of 3 to 6 months, depending on MCD (months of

cyclical dominance).

A. Identification of highest (or lowest) value within 5 months

of selected turn in Spencer curve.

5. Determination of turning points in unsmoothed series.

A. Identification of highest (or lowest) value within 4 months,

or MCD term, whichever is larger, of selected turn in short‐term

moving average.

B. Elimination of turns within 6 months of beginning and end of

series.

C. Elimination of peaks (or troughs) at both ends ofseries which

are lower (or higher) than values closer to end.

D. Elimination of cycles whose duration is less than 15 months.

E. Elimination of phases whose duration is less than 5 months.

6. Statement of final turning points.

Source: Bry and Boschan (1971, p.21; table 1)

GRANVILLE AND HUSSAIN 85