Economic Systems 35 (2011) 574–585

Contents lists available at ScienceDirect

Economic Systems
jo urnal home page : www.elsevier.com/loc ate/ecos ys

Determinants of European stock market integration

David Büttner, Bernd Hayo *
Faculty of Business Administration and Economics, Philipps-University Marburg, Universitätsstr. 24, 35032 Marburg, Germany

A R T I C L E I N F O A B S T R A C T

Article history: We analyse the determinants of stock market integration among EU

Received 10 February 2010 member states for the period 1999–2007. First, we apply bivariate
Received in revised form 14 September 2010 DCC-MGARCH models to extract dynamic conditional correlations
Accepted 14 October 2010
between European stock markets, which are then explained by
Available online 12 June 2011
interest rate spreads, exchange rate risk, market capitalisation, and
business cycle synchronisation in a pooled OLS model. By grouping
JEL classification: the countries into euro area countries, ‘‘old’’ EU member states
E44
outside the euro area, and new EU member states, we also evaluate
F3
the impact of euro introduction and the European unification
F36
G15 process on stock market integration. We find a significant trend
toward more stock market integration, which is enhanced by the
size of relative and absolute market capitalisation and hindered by
Keywords:
Stock market integration foreign exchange risk between old member states and the euro area.
European unification Interest rate spreads and business cycle synchronisation are also
DCC-MGARCH model significant factors in explaining equity market integration.
ß 2011 Elsevier B.V. All rights reserved.

1. Introduction

The integration of financial markets is important to both market participants and policymakers. In
integrated markets, capital flows freely to where it will generate the highest return. Integrated
financial markets have easier access to foreign capital, but are also more vulnerable to financial crises
occurring in other areas of the world. Moreover, any increase in the degree of global financial market
integration decreases the opportunity for diversification. It is thus essential to achieve a better
understanding of the factors driving financial market integration. In this study, we analyse
determinants of stock market integration from 1999 to 2007 using data from European Union (EU)
member states. To assess the impact of political factors on financial market integration, we group the

* Corresponding author. Tel.: +49 0 64 21 28 23 091; fax: +49 0 64 21 28 23 088.

E-mail address: hayo@wiwi.uni-marburg.de (B. Hayo).
URL: http://www.uni-marburg.de/fb02/makro

0939-3625/$ – see front matter ß 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecosys.2010.10.004
 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585 575

countries based on their European integration status into euro area countries (EMU), old EU member
states without the euro (OMS), and new EU member states (NMS).1
Although there are many studies analysing the evolution of stock market integration over time, few
authors attempt to disentangle the driving factors behind this process. Recently, the impact of
European political and economic integration on European stock market integration has been studied
more intensively.
Kim et al. (2005) analyse the impact of EMU on stock market integration by estimating time-
varying conditional correlations between the stock market returns of member states and euro area
weighted aggregate returns.2 They use a bivariate EGARCH framework over the period January 1989 to
May 2003. The authors find that EMU unidirectionally causes a shift to higher European stock market
integration. Moreover, they search for further causes of stock market integration, e.g. foreign exchange
rate volatility, output correlations and seasonal effects. However, only financial market sophistication
measures yield statistically significant results.
Kenourgios et al. (2009) examine the integration in European (euro area, central Europe and the
Balkans) equity and bond markets from January 1997 to October 2006. They use a modified version of
the asymmetric generalized dynamic conditional correlation (AG-DCC) and discover an increase in
integration during the period of the dotcom collapse in 2000, the start of negotiations on EU
membership with the Balkan countries in 2000, the entry of the euro into circulation in 2002, and the
entry of CE countries into the European Union in 2004. Additionally, they find an increase in
integration of the euro area, central European and Balkan countries over time.
Covering the period from January 1986 to June 2000, Fratzscher (2002) examines the impact of
EMU on financial market integration. Within a GARCH framework he finds that European equity
markets have become highly integrated only since 1996 and that the elimination of exchange rate risk
associated with the creation of EMU can explain a large part of financial market integration.
Erb et al. (1994) investigate G-7 cross-country stock exchange correlations from 1970 to 1993 and
discover that correlations are lower when the business cycles of two countries are out of phase and
when they are simultaneously experiencing a growth period. In contrast, during recessions
correlations are higher.3
Using extreme value theory and monthly index return data from 1959 till 1997 for the US, UK,
Germany, France, and Japan, Longin and Solnik (2001) find that the correlation increases in bear
markets but not in bull markets and is unrelated to market volatility.
Pretorius (2002) finds the inverse of industrial production growth differentials and trade linkages
to have a positive impact on bilateral correlations of 10 emerging stock markets over the period of
1995 until 2000. While other factors, such as inflation differentials, interest rates, stock market size,
market volatility, and a time trend do not appear to play a major role, there is a positive impact on
correlations for countries belonging to the same region.
Summing up, the empirical evidence for the impact of European political integration on financial
market integration is stronger than the evidence for the influence of macroeconomic factors.
Our study adds various new elements to the literature: First, it differentiates countries according to
their level of European integration. Second, it includes new member states which joined the EU in
2004. Third, in contrast to other studies (such as Kim et al., 2005), we estimate bilateral dynamic

1
Members of the groups are: (1) Euro area: Austria, Belgium, Finland, France, Germany, Ireland, Italy, The Netherlands,
Portugal, and Spain. Greece is not included, as it introduced the euro only in 2001, nor is Luxemburg, as its markets are small. (2)
Old EU member states outside the euro area: Denmark, Sweden, and The United Kingdom. (3) New EU member states: the Czech
Republic, Hungary, and Poland. These countries were chosen because they have not yet introduced the euro, but have
sufficiently large financial markets.
2
Using a regime-switching model, Baele (2005) shows that European integration had a positive effect on European stock
market volatility spillovers in the 1980s and 1990s. Bartram et al. (2007) and Cappiello et al. (2006) find a positive effect of EMU
on stock market interdependence. In their study of the correlations between spot and futures stock markets in France, Germany
and the UK, Antoniou et al. (2003) show that correlations among French and German markets are higher than with their UK
counterparts. Savva et al. (2009) confirm the positive impact of Euro introduction on stock market correlations among France
and Germany.
3
Croci (2004) also looks at the impact of European macroeconomic convergence on integration of German, French, Italian and
Spanish financial markets over the period of January 1994 through June 2004, yielding mixed results. She uses three different
measures, inter alia DCC-MGARCH.
576 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585

correlations instead of using correlations versus a European stock market index. Using DCC-MGARCH
(dynamic conditional correlation multivariate generalized autoregressive conditional heteroskedas-
ticity) models to retrieve bilateral conditional correlations, we improve on the existing approaches put
forward to explain stock market integration. We test rigorously for the joint significance of variables
across country groups as well as for symmetry in the size of the effects. Finally, we explore potential
differentials between the use of real time and revised data in the analysis.

2. Data description and econometric approach

Figs. 1–4 and Table A1 in Appendix give an overview of the development of the daily stock market
returns.4 The observations are adjusted for public holidays to ensure comparability of the data. It is
noteworthy that the mean returns of the NMS countries are higher than those of EMU and OMS. This
reflects the catching-up process of the transition economies, which is frequently accompanied by an
inflow of foreign capital additionally driving up stock market returns. With regard to market volatility, as
measured by the standard deviations, all markets seem to be similar.5 Excess kurtosis is higher in EMU
and especially in smaller economies.6 Generally, all stock market series display excess kurtosis and a few
show pronounced skewness, indications of non-normal distributions and the presence of ARCH.
Our empirical indicator of the degree of integration of European equity markets is the (conditional)
correlation of returns between these markets. Taking into account that financial integration is a
dynamic process, we allow the estimated correlations to vary over time.7 We use bivariate DCC-
MGARCH models (Engle, 2002) to extract the conditional correlations, as this model class combines a
parsimonious model specification with rich dynamics.8
Following Engle (2002) and Engle and Sheppard (2001), we use the residuals of ARMA (4,4)
processes of n stock market returns. The conditional covariance matrix of these residuals ut jF t1 
Nð0; Ht Þ is characterised as follows:

Ht  Dt Rt Dt ; (1)

where
qffiffiffiffiffiffiffi
Dt ¼ diagf hi;t g; and Rt ¼ diagfQt1 gQ t diagfQt1 g; (2)

Q t ¼ ð1  a  bÞQ̄ þ aðet1 e0t1 Þ þ bQ t1 ; (3)

pffiffiffiffiffiffi
uit ¼ hit eit ; (4)

where Ft1 captures all information up to t1, and Q is the matrix of unconditional correlations. The
conditional variances are based on a FIGARCH (1,d,1) (fractionally integrated GARCH) model (Baillie
et al., 1996), which allows for long memory behaviour and a slow rate of decay after a volatility shock:

1 1
hit ¼ vi ½1  bi L þ f1  ½1  bi ðLÞg ai ðLÞð1  LÞd u2i;t (5)

4
We use the respective ‘‘blue chip’’ indices as these usually contain the stock with the highest liquidity and market
capitalisation.
5
An exception is the Finnish stock market, which is characterised by a somewhat higher standard deviation.
6
Indeed, running a regression of excess kurtosis on average market capitalisation over the observation period reveals that
with an increase of 1 trillion s in market capitalisation excess kurtosis falls by approximately 1. This phenomenon was also
observed by Bekaert et al. (1998) for emerging market returns.
7
Using Engle–Sheppard tests (Engle and Sheppard, 2001) in a preliminary CCC-MGARCH model, we do indeed find that the
hypothesis of constant correlations can be rejected for most of our country pairs (see Table A2 in the Annex). The main results
have been confirmed by using a different lag length (10 lags).
8
For an overview of dynamic correlation models, see Silvennoinen and Teräsvirta (2008).
 200
250

0
50
100
150
0
50
100
150
200
250
300
350
0
50
100
150
200
250
300
350
400
450

0
50
100
150
200
250
300
350
400
450
500
Jan. 99 Jan. 99 Jan. 99 Jan. 99

Jan. 00 Jan. 00 Jan. 00 Jan. 00

DK
IE

CZ
PT
AT

Jan. 01 Jan. 01 Jan. 01 Jan. 01

SE
Jan. 02 Jan. 02
IT

Jan. 02 Jan. 02
BE

HU
Jan. 03 Jan. 03 Jan. 03 Jan. 03

UK

PL
Jan. 04 Jan. 04 Jan. 04 Jan. 04
FI
NL

Jan. 05 Jan. 05 Jan. 05 Jan. 05

D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585

Figs. 1–4. European stock indices 1999–2007 (1999 = 100).

Jan. 06 Jan. 06 Jan. 06 Jan. 06
ES
FR

Jan. 07 Jan. 07 Jan. 07 Jan. 07

DE
577
578 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585

We are interested in the elements of the conditional correlation matrix Rt, i.e., the dynamic
conditional correlations between stock market returns:

q
i j;t
ri j;t ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi (6)
qi j;t qi j;t

with i,j = 1,2 and qij,t as elements of Qt.

For the 16 countries in our sample, we estimate 120 bilateral DCC-MGARCH models. Table A3 in
Appendix shows the means of the estimated dynamic correlations. Values over 0.5 are highlighted.
Correlations among NMS are lower than those among OMS and EMU countries, suggesting a positive
effect of European political integration on financial market interdependence.
The estimated dynamic correlations of European stock markets are explained by variables proxying
for the maturity of financial markets, exchange rate risk, the degree of business cycle synchronisation,
and seasonal as well as trading day effects.
A number of authors (e.g., Fratzscher, 2002; Baele et al., 2004; Kim et al., 2005) suggest that the
introduction of the euro promotes financial market integration by eliminating exchange rate risks and
short-term interest rate spreads. We model exchange rate risk by extracting the conditional
volatilities of exchange rate returns using a GARCH model.
Table A4 in Appendix shows descriptive statistics of bilateral exchange rates. Among the results,
the (unconditional) standard deviation of the Danish crown/euro exchange rate stands out as being
remarkably low, as Denmark is keeping the crown in a very narrow band around a central parity versus
the euro. At the same time, the exchange rates versus the Hungarian forint and the Polish zloty display
high standard deviations and high excess kurtosis. Again, these observations reflect exchange rate
policy. Poland has a free float now and until February 2008, Hungary was operating a (sometimes
shaky) hybrid ERM II style exchange rate regime versus the euro. In contrast, the Czech crown is less
volatile, as the Czech National Bank uses a managed float to keep volatility low.
Daily short-term interest rate spreads are computed using the absolute three-month interest rate
differentials of the respective countries.9
To account for the state of financial market sophistication, we include two measures of the depth
and size of these markets in our model: the sums of absolute and relative (in terms of GDP) market
capitalisation of the two markets under consideration (see Kim et al., 2005).10 Fig. A1a–d shows the
development of relative stock market capitalisation for the observation period. The most interesting
pattern can be observed for Finland: after a peak in 2000, market capitalisation collapsed to one third
due to the end of the dotcom bubble. However, market capitalisation increased substantially after
2002 in Finland and other euro area countries, particularly Austria. In the OMS, the recovery of market
capitalisation started at about the same time. Interestingly, Sweden surpassed the United Kingdom in
terms of market capitalisation toward the end of the observation period. Also, the stock market
capitalisation in the NMS started to increase at the end of 2002, but from markedly lower initial values.
While still below the average value of the EMU and OMS countries the catching-up process led Poland
to reach levels similar to some OMS countries.
Erb et al. (1994) find that the degree of business cycle synchronisation has a significantly positive
effect on stock market integration. We evaluate the impact of goods market integration on financial
market integration by adding two (monthly) lags of an indicator variable for business cycle
synchronisation, which takes the value 1 if the output gaps (extracted from monthly industrial
production) of two countries show the same sign and 1 if they have opposing signs.11 The cumulative
effect of business cycle synchronisation over (almost) one year is captured by a variable that consists
of the sum of the business cycle synchronisation indicators from lag 3 to lag 12, thereby allowing for
possible lags in the processing of information on output developments.12 We also examine whether

9
The data were retrieved from Eurostat and national statistical offices/central banks.
10
Data sources: ECB and Eurostat.
11
The data are from the OECD webpage (MEI original release data and revisions database). To extract the output gaps, we
derive the trend using a Hodrick–Prescott filter with l = 14,400 (Hodrick and Prescott, 1997).
12
At a preliminary stage we also vary the specification with regard to the underlying sizes of the respective output gaps.
However, no substantial changes in estimation results occurred.
 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585 579

the phase of the business cycle matters by including dummy variables for booms and recessions.13
Controlling for omitted trending factors, we include a deterministic time trend. Finally, dummy
variables capture the financial crisis following the terrorist attacks on 11 September 2001, as well as
deterministic seasonal and trading day effects.
As shown by Orphanides and van Norden (2002), using revised output data to estimate reactions of
economic agents can lead to severely biased estimates. Thus, the authors advise employing real-time
data that reflect the actual state of information at the time of decision making. To investigate the
importance of this distinction in the construction of the data set, we estimate the model using both
kinds of data for industrial production.14
Using these factors, we explain the conditional dynamic correlations between equity markets in
the various groups of countries. Estimation takes place within one large stacked OLS model with
group-individual regressors and fixed effects (see Eq. (7)). The advantages of this model set-up include
an increase in estimation efficiency due to the large number of observations and the possibility of
testing for asymmetries across country groups using powerful standard tests; the disadvantages
include the assumptions of equal coefficients within the groups and a common error structure. We
estimate Eq. (7):

Corri;t ¼ Consti þ Corri;t1 ai þ Controli;t g i þ Exchange Rate Riski;t1 bi

þ Interest Rate Spreadi;t1 ni þ Absolute Market Capitalisationi;m1 t i
X
2
þ Relative Market Capitalizationi;m1 ji þ Boomi;mr mi;mr
r¼0

X
2 X
2
þ Recessioni;mr $i;mr þ Business Cycle Syncronisationi;mr hi;mr
r¼0 r¼0

þ Cumulated Business Cycle Syncronisationi;t ki þ ut i

¼ EMU=EMU; EMU=OMS; EMU=NMS; OMS=OMS; OMS=NMS; NMS=NMS; (7)

where Corri,t is the bivariate dynamic stock market correlation between two countries belonging to
the groups EMU, OMS, or NMS at time t (see Eq. (6)), Control includes a deterministic trend as well as
dummies for January, Friday, 11 September 2001, and 12 September 2001. m  1, m  2, . . . refers to the
previous 1, 2, . . . month. All other variables are self-explanatory.

3. Estimation results

Table 1 provides an overview of our results for output gap estimates based on real-time data for
industrial production. Tests indicate that there is no statistical difference from estimators obtained with
revised output data. Thus, contrary to Orphanides and van Norden (2002), we do not find the data
distinction to be important. Note that the estimation takes place within the framework of one large model
and the columns in Table 1 refer to the coefficient associated with the observations for each particular
country group within this specification. Correlations are normalised to 100 for ease of interpretation.15
As the sample is very large (230,520 observations), making the statistical tests highly sensitive to
violations of the null hypotheses, we test at a 0.5% level following Leamer (1983), who recommends
adjusting the significance level inversely with the sample size. Since the estimates exhibit evidence of
autocorrelation and heteroscedasticity, we apply robust standard errors based on Andrews (1991).
We discover a significant deterministic trend toward greater stock market integration. This trend
in integration between old and new member states, as well as between euro area participants and new

13
We use a threshold of one standard deviation of the output gap to define a boom/recession, which implies that small output
gaps yield a value of zero. Choosing different thresholds does not change our conclusions.
14
The correlation coefficients between real-time and revised output gap estimates are 0.68, 0.62, and 0.75 for EMU, OMS, and
NMS, respectively.
15
Please note that, by construction, correlation coefficients are bounded between 1 and 1. Therefore, in principle, OLS
estimation is inappropriate. However, comparing our coefficients with those based on Tobit estimates shows that they are
almost exactly equal (up to four digits after the decimal point).
580 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585

Table 1
Explaining dynamic conditional correlations of European stock market returns.

i= EMU/EMU EMU/OMS EMU/NMS OMS/OMS OMS/NMS NMS/NMS

Consti 0.534* 0.646* 1.014* 0.921* 1.443* 0.173

Corri,t1 98.443* 98.697* 95.862* 97.554* 94.213* 99.375*
Trendi  100 0.013* 0.008* 0.030* 0.023* 0.042* 0.003
11/09/2001i 0.322 0.360 2.064 1.021 2.064 0.584*
12/09/2001i 3.476 4.993* 7.512* 8.472* 9.177* 1.395*
Januaryi 0.033 0.025 0.002 0.034 0.085 0.018
Fridayi 0.009 0.059 0.077* 0.077 0.140 0.001
Exchange rate riski,t1 4.184* 0.650 6.574 0.875 0.300
Interest rate spreadi,t1 0.037 0.009 0.014 0.002 0.002
Market Capi,m1 (% of GDP) 0.132* 0.128* 0.536* 0.204 0.847* 0.645
Market Capi,m1 (Trillion s) 0.237* 0.076* 0.250* 0.001 0.023 0.245
Boomi 0.069 0.040 0.021 0.068 0.083 0.042
Boomi,m1 0.058 0.035 0.049 0.173 0.027 0.031
Boomi,m2 0.081 0.029 0.039 0.006 0.105 0.110
Recessioni 0.071 0.022 0.230* 0.029 0.181 0.322*
Recessioni,m1 0.030 0.022 0.023 0.293 0.016 0.119
Recessioni,m2 0.020 0.107 0.099 0.145 0.226 0.021
Business Cycle Synchronisationi 0.004 0.014 0.009 0.037 0.039 0.002
Business Cycle Synchronisationi,m1 0.009 0.008 0.014 0.042 0.005 0.007
Business Cycle Synchronisationi,m2 0.009 0.001 0.024 0.035 0.025 0.025
Cumulated Business Cycle 0.002 0.003 0.001 0.006 0.002 0.001
Synchronisationi,m3 to m12
No. of observations 230,520
Log-Likelihood 532,352
Normality test Chi2(2) = 1,105,300*
Hetero test 15.26*
AR 1–2 test: 59.92*
R2 0.98

Note: Stock market correlations are normalised to 100. m  1, m  2. . . refer to values of the previous 1,2, . . . month.
*
Significance at a 0.5% level using robust standard errors based on Andrews (1991).

members, is stronger than the correlation between the new EU member states themselves, which is
not significant.
Foreign exchange risk depresses stock market integration among old EU member states and
participants in the euro area. The absence of foreign exchange rate risk in the euro area leads to higher
equity market integration. These results are in accordance with previous findings in the literature and
suggest that adoption of the euro by those EU member countries that are still outside the euro area will
foster financial market integration. In contrast, interest rate differentials do not play a role.
We approximate the maturity and depth of equity markets by including relative and absolute
market capitalisation. As expected, these indicators exert a positive impact on market integration,
implying that the deepening of financial markets (especially in the new member states) will result in
enhanced market correlations in the future.
The point estimates show that correlations tend to be higher on Fridays and during January but, in
contrast to the findings of Kim et al. (2005), these seasonal effects are not significant (with one
exception). The downturn of financial markets following the terrorist attacks on 11 September 2001
tightens the correlation of all stock market returns except within the euro area.
Finally, we find that when euro area and new EU member states are in a recession, their mutual
financial market correlation decreases, whereas it increases when the new member states are jointly
in a recession. However, at a first glance, our individual parameter estimates suggest that business
cycle synchronisation does not play a major role in explaining stock market integration. A more
thorough analysis reveals that this outcome is, due to collinearity between the variables, understating
the statistic significance of our results. Table 2 shows that when testing them jointly, all variable
groups have a significant impact on correlations.
 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585 581

Table 2
Testing determinants: group effects and asymmetry.

Exclusion tests Asymmetry tests

Consti F(6, 230,396) = 745.9* Chi2(15) = 102.6*

Corri,t1 F(6, 230,396) = 4,820,900* Chi2 (15) = 450.0*
Trendi F(6, 230,396) = 293.9* Chi2 (15) = 81.3*
Januaryi F(6, 230,396) = 5.0* Chi2 (15) = 5.1
Fridayi F(6, 230,396) = 24.3* Chi2 (15) = 10.5
11/09/2001 F(6, 230,396) = 76.8* Chi2 (15) = 12.5
12/09/2001 F(6, 230,396) = 1573.8* Chi2 (15) = 588.9*
Exchange rate riski,t1 F(5, 230,396) = 39.5* Chi2 (10) = 26.0*
Interest rate spreadi,t1 F(5, 230,396) = 18.4* Chi2 (10) = 6.9
Market capi,m1 (% of GDP) F(6, 230,396) = 148.3* Chi2 (15) = 65.9*
Market capi,m1 (Billion s) F(6, 230,396) = 337.4* Chi2 (15) = 124.4*
Boomi F(6, 230,396) = 3.1* Chi2 (15) = 2.9
Boomi,m1 F(6, 230,396) = 10.6* Chi2 (15) = 10.2
Boomi,m2 F(6, 230,396) = 10.6* Chi2 (15) = 8.9
Boomi, Boomi,m1, Boomi,m2 F(18, 230,396) = 28.5* Chi2 (45) = 25.8
Recessioni F(6, 230,396) = 16.2* Chi2 (15) = 16.2
Recessioni,m1 F(6, 230,396) = 11.3* Chi2 (15) = 11.2
Recessioni,m2 F(6, 230,396) = 5.6* Chi2 (15) = 5.4
Recessioni, Recessioni,m1, Recessioni,m2 F(18, 230,396) = 36.5* Chi2 (45) = 17.0
Business Cycle Synchronisationi F(6, 230,396) = 13.8* Chi2 (15) = 8.3
Business Cycle Synchronisationi,m1 F(6, 230,396) = 19.8* Chi2 (15) = 4.1
Business Cycle Synchronisationi,m2 F(6, 230,396) = 9.3* Chi2 (15) = 8.8
Cumulated Business Cycle Synchronisationi,m3 to m12 F(6, 230,396) = 8.5* Chi2 (15) = 8.4
Business Cycle Synchronisationi, Business Cycle F(24, 230,396) = 49.2* Chi2 (60) = 28.0
Synchronisationi,m1, Business Cycle
Synchronisationi,m2, Cumulated Business Cycle
Synchronisationi,m3 to m12
Note: m  1, m  2. . . refer to values of the previous 1,2, . . . month.
*
Significance at a 0.5% level using robust standard errors based on Andrews (1991).

It is interesting to see whether our variables have the same effect on the correlation between
different country groups.16 Table 2 contains the outcome of tests for symmetric influences. We find
that constants, lagged effects of correlations, time trends, the 12 September 2001 dummy, exchange
rate risk, and both market capitalisation variables have asymmetric effects over the country groups.
For all other variables, particularly the business cycle indicators, we cannot reject symmetric effects
across the different country groups.

4. Conclusion

We analyse the determinants of stock market integration between EU member states using dynamic
conditional correlations estimated by DCC-MGARCH models. Our indicator of financial market
integration is then analysed by means of a pooled OLS model that groups EU member countries into three
categories: euro area members, old EU member countries not participating in EMU, and new member
states. We conduct our analysis for both real-time and revised output data and discover no noteworthy
differences. Regarding the general development of European stock market integration, we find that for
almost all groups of countries there is a significant trend toward more integration. However, foreign
exchange risk and interest rate spreads depress integration among old EU member states and for
participants of the euro area. Therefore, if non-euro area countries adopt the euro, an increase in stock
market correlations vis-à-vis the euro area can be expected. Moreover, insofar as a common European
monetary policy leads to more synchronised business cycles, we can expect ‘second-round effects’ for
financial market integration. In our sample, we find that when euro area and new EU member states are in
a recession, their mutual financial market correlation decreases, whereas it increases when the new
member states are jointly in a recession. The size of relative and absolute market capitalisation also

16
Therefore, we test whether the group coefficients are jointly equal to each other.
582 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585

promotes equity market integration. If markets deepen, higher correlations can be expected in the future,
especially for the new member states, which are in the process of catching up. We do not find significant
evidence of seasonal or trading day effects. As we conduct our tests at a probability level of 99.5% there
exists only a very small probability of making a type I error of inference, i.e. concluding that there is a
significant effect although in fact there is not. Systematic testing reveals that there is very little
asymmetry in the parameter estimates across groups. Of our main variables of interest, only exchange
rate risk and indicators of market capitalisation show significantly different coefficients. Overall, we find
that nominal determinants of stock market integration seem to be more important to this process than
real determinants.

Appendix A

See Tables A1–A4 and Fig. A1.

Table A1
Descriptive statistics for European stock markets.

Mean Std. Devn. Skewness Exc. Kurt. Min Max

AT 0.0007 0.011 0.60 4.02 0.08 0.05

BE 0.0001 0.013 0.37 8.26 0.08 0.12
FI 0.0004 0.023 0.52 6.85 0.17 0.15
FR 0.0002 0.015 0.05 2.66 0.08 0.07
DE 0.0002 0.017 0.05 2.67 0.09 0.08
IE 0.0002 0.012 0.47 3.4 0.06 0.06
IT 0.0002 0.011 0.57 4.38 0.08 0.06
NL 0.0000 0.015 0.43 8.29 0.14 0.11
PT 0.0003 0.010 0.09 5.8 0.06 0.07
ES 0.0003 0.013 0.16 2.37 0.07 0.06
DK 0.0004 0.012 0.19 2.87 0.07 0.07
SE 0.0002 0.017 0.04 2.11 0.09 0.09
UK 0.0000 0.012 0.18 2.53 0.06 0.06
CZ 0.0008 0.014 0.42 3.2 0.09 0.07
HU 0.0007 0.016 0.03 3.59 0.10 0.09
PL 0.0008 0.015 0.12 4.34 0.09 0.12
Source: Datastream.
Note: The number of observations is 1938 for all stock markets.

Table A2
Results Engle–Sheppard (5) tests of constant correlations.

AT BE FI FR DE IE IT NL PT ES DK SE UK CZ HU PL

AT 36.8** 57.5** 24.4** 21.6** 35.4** 41.2** 53.4** 34.5** 27.8** 18.6** 57.2** 24.0** 26.8** 35.0** 25.9**
BE 36.8** 26.4** 72.3** 66.7** 12.6 35.4** 119.0** 22.6** 17.5** 49.2** 61.2** 65.7** 17.0** 27.2** 29.2**
FI 57.5** 26.4** 16.1* 28.8** 10.3 22.2** 46.1** 3.0 14.2* 7.1 21.4** 14.7* 24.6** 26.0** 13.0*
FR 24.4** 72.3** 16.1* 38.4** 13.4* 12.7* 291.8** 34.2** 44.5** 15.4* 28.2** 91.4** 18.0** 30.5** 19.1**
DE 21.6** 66.7** 28.8** 38.4** 13.7* 16.8* 134.8** 29.3** 21.8** 5.4 36.7** 27.1** 17.1** 15.6* 12.4
IE 35.4** 12.6 10.3 13.4* 13.7* 18.9** 54.7** 8.0 21.4** 12.8* 20.5** 12.3 20.9** 8.8 9.2
IT 41.2** 35.4** 22.2** 12.7* 16.8* 18.9** 132.3** 23.9** 40.4** 17.8** 11.5 18.2** 40.8** 21.6** 27.6**
NL 53.4** 119.0** 46.1** 291.8** 134.8** 54.7** 132.3** 58.9** 123.5** 93.3** 111.0** 280.4** 24.8** 41.3** 8.3
PT 34.5** 22.6** 3.0 34.2** 29.3** 8.0 23.9** 58.9** 8.4 17.5** 29.6** 5.8 6.6 10.1 17.6**
ES 27.8** 17.5** 14.2* 44.5** 21.8** 21.4** 40.4** 123.5** 8.4 17.7** 25.1** 31.4** 8.5 18.7** 11.1
DK 18.6** 49.2** 7.1 15.4* 5.4 12.8* 17.8** 93.3** 17.5 **
17.7** 11.4 13.3* 23.5** 15.9* 16.5*
SE 57.2** 61.2** 21.4** 28.2** 36.7** 20.5** 11.5 111.0** 29.6** 25.1** 11.4 52.3** 19.7** 26.3** 20.5**
UK 24.0** **
65.7 14.7 *
91.4 **
27.1** 12.3 18.2** 280.4** 5.8 31.4** 13.3* 52.3** 16.0* 18.3** 10.8
CZ 26.8** 17.0** 24.6** 18.0** 17.1** 20.9** 40.8** 24.8** 6.6 8.5 23.5** 19.7** 16.0* 22.0** 20.6**
HU 35.0** 27.2** 26.0** 30.5** 15.6* 8.8 21.6** 41.3** 10.1 18.7** 15.9* 26.3** 18.3** 22.0** 12.0
PL 25.9** 29.2** 13.0* 19.1** 12.4 9.2 27.6** 8.3 17.6** 11.1 16.5* 20.5** 10.8 20.6** 12.0

Note: The table reports the test statistics of the Engle–Sheppard test (Engle and Sheppard, 2001), with H0 stating that the
correlations between two stock markets are constant.
*
Significance at 5% level.
**
Significance at 1% level.
 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585 583

Table A3
Means of dynamic correlations.

AT BE FI FR DE IE IT NL PT ES DK SE UK CZ HU PL

AT 0.47 0.43 0.51 0.48 0.39 0.48 0.52 0.37 0.48 0.42 0.46 0.48 0.39 0.39 0.34
BE 0.47 0.56 0.74 0.70 0.51 0.63 0.75 0.50 0.69 0.51 0.63 0.68 0.37 0.36 0.31
FI 0.43 0.56 0.74 0.69 0.51 0.62 0.73 0.50 0.67 0.53 0.75 0.68 0.42 0.43 0.39
FR 0.51 0.74 0.74 0.89 0.58 0.76 0.88 0.60 0.85 0.61 0.80 0.84 0.44 0.45 0.42
DE 0.48 0.70 0.69 0.89 0.55 0.71 0.84 0.56 0.80 0.56 0.78 0.77 0.41 0.43 0.38
IE 0.39 0.51 0.51 0.58 0.55 0.54 0.59 0.40 0.53 0.46 0.54 0.60 0.30 0.35 0.34
IT 0.48 0.63 0.62 0.76 0.71 0.54 0.73 0.52 0.71 0.55 0.67 0.66 0.40 0.41 0.40
NL 0.52 0.75 0.73 0.88 0.84 0.59 0.73 0.57 0.79 0.60 0.77 0.82 0.44 0.46 0.42
PT 0.37 0.50 0.50 0.60 0.56 0.40 0.52 0.57 0.60 0.44 0.53 0.54 0.34 0.32 0.35
ES 0.48 0.69 0.67 0.85 0.80 0.53 0.71 0.79 0.60 0.55 0.74 0.76 0.42 0.42 0.37
DK 0.42 0.51 0.53 0.61 0.56 0.46 0.55 0.60 0.44 0.55 0.58 0.57 0.38 0.38 0.34
SE 0.46 0.63 0.75 0.80 0.78 0.54 0.67 0.77 0.53 0.74 0.58 0.74 0.43 0.44 0.43
UK 0.48 0.68 0.68 0.84 0.77 0.60 0.66 0.82 0.54 0.76 0.57 0.74 0.42 0.45 0.41
CZ 0.39 0.37 0.42 0.44 0.41 0.30 0.40 0.44 0.34 0.42 0.38 0.43 0.42 0.48 0.43
HU 0.39 0.36 0.43 0.45 0.43 0.35 0.41 0.46 0.32 0.42 0.38 0.44 0.45 0.48 0.51
PL 0.34 0.31 0.39 0.42 0.38 0.34 0.40 0.42 0.35 0.37 0.34 0.43 0.41 0.43 0.51

Table A4
Descriptive statistics of foreign exchange rate returns.

Mean Std. Devn. Skewness Exc. Kurt. Min Max

CZK/s 0.000133 0.0037 0.03 4.32 0.02 0.02

DKK/s 0.000001 0.0002 0.38 3.59 0.00 0.00
HUF/s 0.000020 0.0049 2.09 20.21 0.02 0.05
PLN/s 0.000032 0.0069 1.50 15.69 0.03 0.09
SEK/s 0.000022 0.0036 0.05 2.12 0.02 0.02
GBP/s 0.000032 0.0047 0.34 3.13 0.03 0.03
DKK/CZK 0.000134 0.0037 0.01 4.23 0.02 0.02
HUF/CZK 0.000134 0.0053 1.07 9.43 0.05 0.02
PLN/CZK 0.000061 0.0067 0.98 9.59 0.07 0.04
SEK/CZK 0.000144 0.0048 0.05 2.40 0.03 0.02
GBP/CZK 0.000145 0.0057 0.32 2.36 0.04 0.02
HUF/DKK 0.000005 0.0049 1.83 17.67 0.05 0.02
PLN/DKK 0.000081 0.0069 1.21 12.57 0.08 0.03
SEK/DKK 0.000007 0.0036 0.02 2.16 0.02 0.02
GBP/DKK 0.000009 0.0047 0.27 3.04 0.03 0.03
PLN/HUF 0.000084 0.0064 0.14 4.05 0.03 0.04
SEK/HUF 0.000009 0.0057 0.85 8.85 0.03 0.05
GBP/HUF 0.000009 0.0066 0.74 7.46 0.04 0.06
SEK/PLN 0.000046 0.0071 0.92 8.75 0.03 0.08
GBP/PLN 0.000052 0.0071 1.23 14.70 0.03 0.09
GBP/SEK 0.000008 0.0054 0.13 1.19 0.03 0.02
Source: Eurostat.
Note: Exchange rates in price notation. The number of observations is 1938 for all currency pairs. Czech crown: CZK, Hungarian
forint: HUF, Polish zloty: PLN, Danish crown: DKK, Swedish crown: SEK, British pound: GBP.
584 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585

a 350
FI FR DE
300
NL ES
250

200

150

100

50

0
Jul-99

Jul-00

Jul-01

Jul-02

Jul-03

Jul-04

Jul-05

Jul-06

Jul-07
Jan-99

Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Jan-06

Jan-07
90
b
80
70
60
50
40
30
20 AT BE
10 IE IT
PT
0
Jul-99

Jul-00

Jul-01

Jul-02

Jul-03

Jul-04

Jul-05

Jul-06

Jul-07
Jan-99

Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Jan-06

Jan-07
c 250
DK SE UK
200

150

100

50

0
Jul-99

Jul-00

Jul-01

Jul-02

Jul-03

Jul-04

Jul-05

Jul-06

Jul-07
Jan-99

Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Jan-06

Jan-07

60
d CZ HU PL
50

40

30

20

10

0
Jul-99

Jul-00

Jul-01

Jul-02

Jul-03

Jul-04

Jul-05

Jul-06

Jul-07
Jan-99

Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Jan-06

Jan-07

Sources: ECB, Eurostat, National Stock Exchanges.

Fig. A1. (a) Stock market capitalisation in EMU countries (in % of GDP), (b) stock market capitalisation in EMU countries (in % of
GDP), (c) stock market capitalisation in OMS countries (in % of GDP) and (d) stock market capitalisation in NMS countries (in % of
GDP).
Sources: ECB, Eurostat, National Stock Exchanges.
 D. Büttner, B. Hayo / Economic Systems 35 (2011) 574–585 585

References

Andrews, D.W.K., 1991. Heteroskedasticity and autocorrelation consistent covariance matrix estimation. Econometrica 59,
817–858.
Antoniou, A., Pescetto, G., Violaris, A., 2003. Modelling international price relationships and interdependencies between the
stock index and stock index futures markets of three EU countries: a multivariate analysis. Journal of Business, Finance &
Accounting 30, 645–667.
Bekaert, G., Erb, C.B., Harvey, C., Viskanta, T., 1998. Distributional characteristics of emerging market returns and asset
allocation. Journal of Portfolio Management 24, 102–116.
Baele, L., Ferrando, A., Hördahl, P., Krylova, E., Monnet, C., 2004. Measuring financial integration in the euro area. ECB Occasional
Paper Series 14, Frankfurt/Main.
Baele, L., 2005. Volatility spillover effects in European equity markets. Journal of Financial and Quantitative Analysis 40, 373–
401.
Baillie, R.T., Bollerslev, T., Mikkelsen, H.O., 1996. Fractionally integrated generalized autoregressive conditional heteroskedas-
ticity. Journal of Econometrics 74, 3–30.
Bartram, S.M., Taylor, S.J., Wang, Y.-H., 2007. The Euro and European financial market dependence. Journal of Banking & Finance
31, 1461–1481.
Cappiello, L., Engle, E.F., Sheppard, K., 2006. Asymmetric dynamics in the correlations of global equity and bond returns. Journal
of Financial Econometrics 4, 537–572.
Croci, M., 2004. Integration of the euro equity markets: An empirical analysis. Working Paper, Università Politecnica delle
Marche - Department of Economics, Ancona.
Engle, R.F., 2002. Dynamic conditional correlation—a simple class of multivariate GARCH models. Journal of Business and
Economic Statistics 20, 339–350.
Engle, R.F., Sheppard, K., 2001. Theoretical and Empirical Properties of Dynamic Conditional Correlation Multivariate GARCH.
Mimeo, UCSD, La Jolla, CA.
Erb, C., Harvey, C.R., Viskanta, T.E., 1994. Forecasting international equity correlations. Financial Analysts Journal 50, 32–45.
Fratzscher, M., 2002. Financial market integration in Europe: on the effects of EMU on stock markets. International Journal of
Finance and Economics 7, 165–193.
Hodrick, R.J., Prescott, E.C., 1997. Postwar U.S. business cycles: an empirical investigation. Journal of Money, Credit, and Banking
29, 1–16.
Kenourgios, D., Samitas, A., Paltalidis, N., 2009. Financial market dynamics in an enlarged European Union. Journal of Economic
Integration 24, 197–221.
Kim, S.J., Moshirian, F., Wu, E., 2005. Dynamic stock market integration driven by the European Monetary Union: an empirical
analysis. Journal of Banking & Finance 29, 2475–2502.
Leamer, E.E., 1983. Model choice and specification analysis. In: Griliches, Z., Intriligator, M.D., Schmidt, P. (Eds.), Handbook of
Econometrics, vol. 1. North-Holland, Amsterdam, pp. 285–330.
Longin, F., Solnik, B., 2001. Extreme correlation of international equity markets. The Journal of Finance 56, 649–676.
Orphanides, A., van Norden, S., 2002. The unreliability of output-gap estimates in real time. Review of Economics and Statistics
84, 569–583.
Pretorius, E., 2002. Economic determinants of emerging stock market interdependence. Emerging Markets Review 3, 84–105.
Savva, C.S., Osborn, D.R., Gill, L., 2009. Spillovers and correlations between US and major European stock markets: the role of the
euro. Applied Financial Economics 19, 1595–1604.
Silvennoinen, A., Teräsvirta, T., 2008. Multivariate GARCH Models. SSE/EFI Working Paper Series in Economics and Finance 669,
Stockholm.