The Eurozone and cross-country equity return correlations : divergence with the rest of the EU?

(1)

The Eurozone and cross-country equity

return correlations: divergence with the

rest of the EU?

Stef Konijn (11051264)

University of Amsterdam

June 2018

BSc Economics and Business

Track: Economics and Finance

Supervised by: dr. Esther Eiling

Word count: 6801

(2)

Stef Konijn – University of Amsterdam

ABSTRACT ... 3

I. INTRODUCTION ... 3

II. LITERATURE REVIEW ... 5

III. METHODOLOGY ... 7

IV. DATA ... 9

V. RESULTS AND ANALYSIS ... 10

VI. ROBUSTNESS CHECK ... 14

VII. CONCLUSION ... 18

REFERENCE LIST ... 20

APPENDIX ... 22

Statement of originality

This document is written by Stef Konijn who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in

creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents

(3)

Abstract

This paper examines quarterly cross-country equity market return correlations for the Eurozone and other EU members between 1995 and 2015. We find that the average

correlation between other EU members is unexpectedly higher but shows no significant upward trend whilst the correlation of the Eurozone does, indicating a

convergence between the two. Furthermore, macroeconomic factors that could affect correlations are examined and we find that trade openness could explain movements in Eurozone correlations, but the precise effect of the euro itself remains

unclear.

I. Introduction

Since the introduction of the euro on January 1st_{, 1999, a total of nineteen European}

countries have officially joined the Eurozone. The single currency has issued a large debate among policy makers, researchers, and market participants about its effects on financial markets (Cappiello, Kadareja, and Manganelli, 2010). One of the main areas of interest has been about the cross-country correlations of equity markets. Practitioners, such as investors, are for example interested in the optimal diversified portfolio for which correlations and exchange rate risk are a major concern. The euro makes it possible for identical financial assets to be perfect substitutes even when they are traded in different Eurozone countries (Jappelliand and Pagano, 2008).

In 1989, Eun and Shim (1989) found that the United States’ equity market movements can have significant effect on other national stock markets. Prior research from the 1960s and 1970s, however, found little correlations among

markets and showed that diversification across countries had significant benefits for investors. In de decade before Eun and Shim’s research, market correlations between developed nations rose from a low 0.2 in the 1970s to 0.5 in 1989 and thereafter a continuous increase to 0.8 by the end of the 2000s (Quinn and Voth, 2008). After decades of debates among academics, some studies have even noted that industry factors are now more important factors in driving international equity returns than country factors (see, for instance, Eiling, Gerard, Hillion, and Roon, 2012; Moerman, 2008). This has largely been attributed to global liberalization of financial markets in the world. One example of such liberalization efforts is within Europe where the European Union (EU) plays a key role. Membership of the Economic and Monetary Union (EMU) of the EU has even been used in a study to account for capital control openness of European countries due to the conditions imposed for membership (Quinn and Voth, 2008). Example of such trade liberalization within the Eurozone are

(4)

the removal of controls on capital flows and the harmonization of regulations (Jappelliand and Pagano, 2008).

Integration may happen in the euro area, but not all members of the EMU have joined the euro and the same counts for other European countries that are a member of the EU. This begs the question how the introduction of the euro has influenced the whole of the EU, including countries that did not adopt the single currency. Therefore, the research question of this paper is as follows: Have EU members that have not adopted the euro diverged from the Eurozone in terms of cross-country equity market correlations since the introduction of the euro?

The goal of this study is to investigate how Eurozone correlations and those of non-Eurozone countries have moved over time. Moreover, we examine if the

correlations between the two regions have converged or diverged since the

introduction of the euro. This is studied over a period of time from four years before the introduction of the euro, in 1995, towards the very end of the financial crisis, in 2015. Also, the economic factors that might play a role in influencing correlations are investigated. This paper uses techniques for estimating quarterly correlations similar to Eiling and Gerard (2015), but the focus is on the European equity markets only.

This paper adds to the existing literature as it focuses specifically on Europe and uses the whole of the EU, including recent members, in the study. Where most research focuses on the first years after the introduction of the euro, we examine the effects of the euro for fifteen years after its introduction in 1999.

Surprisingly, we find that the average correlation of the EU members that have not adopted the euro is higher than the average of the Eurozone. This result is

however not robust for a smaller sample period or if we use the developed Eurozone markets only. Furthermore, a convergence between the two correlations over time is found, which is robust for all tests except if use developed markets only. The

economic channels which may affect comovements differ across regions. In the Eurozone the trade openness and the number of observations seem to play a more important role. For the non-euro area, it remains unclear which factors exactly

explain correlation movements over time, but the number of observations does seem to have some effect.

The remainder of this paper is structured as follows. First, in section two, the existing literature on global and European equity return correlations will be

discussed. Also, factors used in previous studies that possibly affect comovements are examined. Next, in the third section, the methodology of this study will be

presented, followed by an explanation of the data in section four. In section five, the results and analysis will be discussed, followed by our robustness checks in section

(5)

six. Finally, the seventh section concludes. Additional tables and graphs are included in the Appendixes.

II. Literature review

As noted before, equity market correlations have been of major interest to academics for many years. Research as old as Grubel’s (1968) investigate the role of correlations for diversified portfolios and find that, at the time, for most Western countries there is little evidence of significant correlations or at least low ones with the United States. According to Quinn and Voth (2008), who study a century of global equity market correlations, capital account openness was greatly reduced during both World Wars and the following capital controls such as the Bretton Woods system had a significant effect on market comovements till the second half of the 20th_{century. Financial}

liberalization, however, has greatly increased stock market correlations over the past decades. Quinn and Voth’s (2008) main finding is that there is strong evidence that countries with little capital controls show greater equity market correlations.

In Europe, a similar increase in comovements has been found. Cappiello, Kadareja, and Manganelli (2010) report that the degree of equity return

comovements has increased since the introduction of the euro and find that it is robust for global trends. Specifically, financial, industrial, and consumer service sectors are the most important factors in driving the increase in correlations. Moreover, Brooks and Del Negro (2002) even argue that in their research Europe is the only region where the balance between country and industry effects has

significantly changed in the 1990s following the fiscal and monetary integration of the Maastricht Treaty. McAllister and Lizieri (2006), furthermore, remark that it is difficult to examine the precise effect of monetary integration in Europe due to the many other forms of global and regional integration taking place. They find that the

European factor has a significant effect in explaining equity returns, but that its effect is similar for non-Eurozone and non-EU countries as for countries that have adopted the euro. For real estate equities, the increase in correlations is even greater for the former two groups of countries than for the Eurozone. One possible explanation given is that broader economic integration may play a larger role than the monetary integration associated with the euro. Bekaert, Harvey, Lundblad, and Siegel (2013) find similar results that EU membership, but not the adaptation of the euro, has increased financial integration in Europe. Ferreira and Ferreira (2006) show that industry factors have become of growing importance in Europe in determining equity returns, and that in the post-euro period (1999-2001) they have become of similar importance as country factors. They, however, find similar results for their sample of five European non-EMU countries as for the eleven EMU countries examined, which

(6)

indicates that the first two years of the euro had no significant effect. Flavin (2004) finds the same results for a sample between 1995 and 2002 of the original eleven Eurozone countries and four other European countries. Bekaert, Hodrick, and Zhang (2009) also find an upward trend in correlations in the euro area but find that this is mostly driven by global trends. Nevertheless, the upward trend seems to coincide with the liberalization of financial markets in major European markets, such as France and Italy. Additionally, Savva and Aslanidis (2009) examine the comovements of Central and Eastern European countries with the Eurozone. They find that Czech, Slovenian, and Polish stock markets have had a significant increase in correlations with the euro area between 1997 and 2008 which is mainly driven by EU related developments. The Hungarian and Slovak market comovements, however, remain unchanged.

The question remains what specific factors of the euro have contributed to the increase of comovements. As noted before, Quinn and Voth (2008) find that financial liberalization results in an increase of correlations. This could be due to a decline of the home bias for portfolio holdings by European investors following the reduction of capital controls (Brooks and Del Negro, 2002; De Santis and Gerard, 2006). This has also been shown by Coeurdacier and Martin (2009) who find the euro and the association financial liberalization has reduced transaction costs for equity holdings by a total of approximately 27% which resulted in investors being more inclined to purchase equity within the Eurozone. Furthermore, higher correlations can be

explained by a growing trend towards multinational companies so that they are more exposed to international business cycles (Brooks and Del Negro, 2002; Adjaouté and Danthine, 2004). This could also have been fueled by the reduction of international capital controls. Moreover, a study has found that the harmonization of interest rates is one of the causes of further integration of European stock markets (Ferreira and Ferreira, 2006).

As for other comovements, after the introduction of the euro, Gonçalves, Rodrigues, and Soares (2009) find that bilateral correlation of business cycles has increased more among EMU-members than among other OECD countries. Moreover, Abad, Chulià, and Gómez-Puig (2009) show that euro bond markets have become less vulnerable to global risk factors and more vulnerable to EMU risk factors, indicating a higher degree of integration. Further integration of these markets could also affect the cross-country equity correlations.

However, Adjaouté and Danthine (2004), also note that according to Ricardian trade theory countries stick to their comparative advantages if international trade controls are abolished. This would lead to geographical specialization which should in turn lead to lower cross-country correlations. Also, Francis, Hunter, and Hasan (2002)

(7)

find that high levels of currency volatility can lead to higher equity correlations. This could indicate that correlations in the euro area actually decrease after the

introduction of the single currency.

In this paper, three macroeconomic factors that potentially explain cross-country equity correlations are examined. Proxies for these factors are given in sections five and six. First, trade openness is included. Chen and Zhang (1997) find that countries with high bilateral trade have higher comovements in financial markets. In other words, trade within a region, which measures economic

integration, is positively correlated with financial integration. We thus expect trade openness to positively affect equity correlations within a region. Second, we consider economic growth. According to Chambet and Gibson (2008), higher income per capita could lead to investors being more inclined to invest abroad due to lower risk aversion and lower information costs. This in turn increases cross-country equity correlations, so we expect a positive relationship. Third, the development of stock markets is included. As equity markets develop, foreign investors might be more willing to invest in the local stock market (Eiling and Gerard, 2015). Thus, we expect the development of equity markets to have a positive relationship with the

correlations.

All in all, existing literature has found an increase in comovements for the Eurozone. This increase can however also be found for other EU members and European countries. It remains unclear how the euro has exactly influenced equity market correlations, but we expect that the equity correlations of the Eurozone have diverged from the rest of the EU.

III. Methodology

In this paper, a similar technique is used as the one used by Eiling and Gerard (2015). We are not interested in the correlations between two countries, but in the average correlations of daily excess returns within a specific region 𝑎 over time. First, daily excess returns are found by subtracting the daily risk-free rate from the daily total equity market returns for each individual country 𝑖. We define a period 𝑡 as a quarter and take the average of the daily cross-sectional dispersion to further improve our estimation result accuracy. The same data-generating process for excess returns is used as by Bekaert, Hodrick, and Zhang (2009) which, in this study, is as follows:

𝑟̃_𝑖,𝑡 = 𝐸(𝑟̃_𝑖,𝑡) + 𝛽_𝑖,𝑡𝑔𝑙𝑜𝑏𝑎𝑙∙ 𝐹̃_𝑡𝑔𝑙𝑜𝑏𝑎𝑙+ 𝛽_𝑖,𝑡𝑙𝑜𝑐𝑎𝑙 ∙ 𝐹̃_𝑡𝑙𝑜𝑐𝑎𝑙+ 𝜀̃_𝑖,𝑡 (1) where 𝑟̃_𝑖,𝑡 is the excess return of country 𝑖, 𝐸(𝑟̃_𝑖,𝑡) is its expectation, 𝐹̃_𝑡𝑔𝑙𝑜𝑏𝑎𝑙 and 𝐹̃_𝑡𝑙𝑜𝑐𝑎𝑙 are global and local factors, respectively, and 𝜀̃_𝑖,𝑡 is the country-specific

(8)

idiosyncratic return. The factors 𝐹̃_𝑡𝑔𝑙𝑜𝑏𝑎𝑙 and 𝐹̃_𝑡𝑙𝑜𝑐𝑎𝑙 drive the excess return with exposures 𝛽_𝑖,𝑡𝑔𝑙𝑜𝑏𝑎𝑙 and 𝛽_𝑖,𝑡𝑙𝑜𝑐𝑎𝑙, where the local factors only play a role within region 𝑎.

Correlations are affected by the role the different global, local and

idiosyncratic factors play within a region over time. To find the relative importance of these factors, we have to make certain assumptions due to the fact that we have a relatively high number of data points. Because of this, it becomes more difficult to precisely estimate the individual exposures to global and local factors and country residual risk using daily returns within a period.

The same assumptions are made as those made by Eiling and Gerard (2015). First, it is assumed that all countries within region 𝑎 have the same exposures to global and regional factors and that the idiosyncratic country-level variance is the same for all countries within the region. In symbols, that is as follows: 𝛽_𝑖,𝑡𝑔𝑙𝑜𝑏𝑎𝑙 = 𝛽_𝑎,𝑡𝑔𝑙𝑜𝑏𝑎𝑙 and 𝛽_𝑖,𝑡𝑙𝑜𝑐𝑎𝑙 = 𝛽_𝑎,𝑡𝑙𝑜𝑐𝑎𝑙, and 𝜎_{𝜀𝑖,𝑡}2 = 𝜎_{𝜀𝑎,𝑡}2 , for all countries 𝑖 within region 𝑎. The factors are however allowed to differ across regions and over time. What follows is that all correlations within a region are the same for all pairs of countries. This is known as dynamic equicorrelation and it provides consistent parameter estimates (Engle and Kelly, 2008). Secondly, similar to Eiling and Gerard (2015), we assume the factors and the country-specific idiosyncratic return are normally distributed, such that:

𝐹̃_𝑡𝑔𝑙𝑜𝑏𝑎𝑙~ (0, 𝜎_𝑊,𝑡2 ₎

𝐹̃_𝑎,𝑡𝑙𝑜𝑐𝑎𝑙~ (0, 𝜎_{𝐿𝑎,𝑡}2 ) 𝜀̃_𝑖,𝑡 ~ (0, 𝜎_{𝜀𝑎,𝑡}2 ₎

where 𝐹̃_𝑎,𝑡𝑙𝑜𝑐𝑎𝑙 is the local factor within region 𝑎. This results in the following return variance for each country 𝑖 within region 𝑎:

𝜎_𝑎,𝑡2 = 𝛽_𝑎,𝑡𝑔𝑙𝑜𝑏𝑎𝑙 2∙ 𝜎_{𝑔𝑙𝑜𝑏𝑎𝑙,𝑡}2 + 𝛽_𝑎,𝑡𝑙𝑜𝑐𝑎𝑙 2∙ 𝜎_{𝑙𝑜𝑐𝑎𝑙 𝑎,𝑡}2 + 𝜎_{𝜀𝑎,𝑡}2 (2) so that the variance can be split up into global, local, and idiosyncratic risk factors, of which the global and local factors constitute to systematic risk. According to Eiling and Gerard (2015), the results of this setup is that all pairwise correlations for each set of countries is equal. Within both regions cross-country correlations for each quarter 𝑡 are then equal to the proportion of idiosyncratic variance to total variance, or: 𝜌_𝑎,𝑡 = 𝛽𝑎,𝑡 𝑔𝑙𝑜𝑏𝑎𝑙2 ∙𝜎_{𝑔𝑙𝑜𝑏𝑎𝑙,𝑡}2 +𝛽𝑎,𝑡𝑙𝑜𝑐𝑎𝑙2∙𝜎𝑙𝑜𝑐𝑎𝑙 𝑎,𝑡2 𝛽_𝑎,𝑡𝑔𝑙𝑜𝑏𝑎𝑙2∙𝜎_{𝑔𝑙𝑜𝑏𝑎𝑙,𝑡}2 +𝛽_𝑎,𝑡𝑙𝑜𝑐𝑎𝑙2∙𝜎_{𝑙𝑜𝑐𝑎𝑙 𝑎,𝑡}2 +𝜎_{𝜀𝑎,𝑡}2 = 1 − 𝜎𝜀𝑎,𝑡 2 𝜎_𝑎,𝑡2 (3)

(9)

for all countries within region 𝑎. This correlation is interpreted as the average cross-country equity return correlation within region 𝑎 in period 𝑡 (Eiling and Gerard, 2015).

Like Ferreira and Ferreira (2006), daily excess returns are assumed to be serially uncorrelated due to the fact that European markets open and close at approximately the same time each day. We measure the idiosyncratic country-level variance for region 𝑎 with 𝑁 countries in period 𝑡, which is again defined as a quarter, as follows: σ ̂2 εa,t = 𝑣2𝑎,𝑡 = 1 𝑁−1∑ 𝑣 2 𝑖,𝑡 𝑁 i=1 (4) where 𝑣2_𝑖,𝑡 = ∑𝐷𝑡 (𝑟_𝑖,𝑑− 𝑟_{𝐸𝑊 𝑎,𝑑} d=1 )2

where 𝐷_𝑡 is defined as the number of trading days during quarter 𝑡. Similar to Eiling and Gerard (2015), total variance during quarter 𝑡 for region 𝑎 is found by taking the cross-sectional average of the total variance over all countries in that region:

σ̂2_a,t = 𝑉2_𝑎,𝑡 = 1 𝑁∑ 𝑉 2 𝑖,𝑡 𝑁 i=1 (5) where 𝑉2_𝑖,𝑡 = ∑𝐷𝑡 𝑟2_𝑖,𝑑 𝑑=1

Finally, Eiling and Gerard (2015) propose the following estimator for the average level of the cross-country correlation for region 𝑎 in quarter 𝑡:

𝜌̂_𝑎,𝑡 = 1 − 𝑣2𝑎,𝑡

𝑉2_𝑎,𝑡 (6)

IV. Data

Our dataset consists of twenty-eight countries that are currently a member of the European Union. The data is subdivided into two groups: the Eurozone and other EU members that are not part of the Eurozone. Similar to Flavin (2004), countries that have recently joined the monetary union, in this case Latvia and Lithuania which joined the euro in 2014 and 2015 respectively, are omitted from our Eurozone sample. We expect it to be improbable that they had any significant effect on correlations within the Eurozone due to their recent adaptation. They are both included in the group of other EU members not part of the Eurozone which totals eleven countries. This group also includes the United Kingdom. The Eurozone now includes a total of seventeen countries, excluding Latvia and Lithuania.

Data is used from Q1 1995 up until Q4 2015 as soon as a total return index, and possibly an exchange rate to the euro, was available. We find 5479 daily

observations for each index available from beginning to end, totaling to eighty-four quarters over twenty-one years. Due to the limited availability of exchange rates

(10)

between local currencies and the euro and to have as many data points as possible, index returns of non-Eurozone countries were first, when needed, converted to the British pound sterling and then into euros. For the Eurozone twelve countries and for the non-Eurozone six countries already have data in 1995 such that it is sufficient to calculate correlations. The German government 3-month interest rate is used as the risk-free rate and found by dividing the annual rate by the number of trading days within the year. Table 𝐴1 in the Appendix shows the summary statistics of the daily total returns for both the Eurozone and the other EU countries as well as for the risk-free rate. We use Datastream to find the daily total return indexes, daily exchange rates, and interest rates. Daily returns are adjusted for holidays and market-closures for longer than one day by taking the return following the closure and dividing it by the number of days that is was closed for plus one. This average is then used for all days that the stock market was closed for as well as the day following the closure.

V. Results and Analysis

This section gives several results where all tests use the whole quarterly time period from 1995 up until 2015 with our full sample of twenty-eight countries. Other time periods and restricted samples are presented in the robustness exercises in the next section. Table 𝐼 reports the summary statistics of the cross-country equity return correlations for members of the Eurozone and for the other EU members for the full sample period. Also, the statistics for a new set of correlations, the difference

between the two, is shown, which is found by subtracting the correlations of the non-Eurozone from the correlations corresponding to the non-Eurozone. Surprisingly, over the whole time period, the average correlation for the group of non-euro countries (0.3236) is higher than for the countries that have adopted the euro (0.2831), also

Table 𝑰. Summary statistics of quarterly correlations for the Eurozone and other EU-members and the difference between the two for the whole sample

This table reports the summary statistics of the quarterly correlations for both regions, and the difference between the two, from Q1 1995 up until Q4 2015, totaling to eighty-four periods. The difference between the two correlations is found by subtracting the non-Eurozone from the non-Eurozone correlation.

Mean Median Stdev Min Max Skew Kurt

Correlations

Eurozone 0.2831 0.2722 0.1248 0.04 0.60 0.409 0.175

Non-euro 0.3236 0.3134 0.1129 0.10 0.62 0.397 -0.290

(11)

Stef Konijn – University of Amsterdam Co rrela tio n s

Figure 𝑰. Movement of the quarterly correlation for the Eurozone and other EU-members for the whole sample

indicated by the negative average value of the difference (-0.0405). Figure 𝐼 shows the movement of the quarterly correlations over the twenty-one years used in our sample. Furthermore, in Figure 𝐴1 in the Appendix, the movement over time of the difference between the two correlations is shown.

Next, to see if there are certain time trends for the correlations over time, we use a simple linear time trend regression, similar to Eiling and Gerard (2015), that looks as follows:

𝑦_𝑡 = 𝛼 + 𝛽𝑡𝑖𝑚𝑒∙ 𝑡 + 𝜀_𝑡 (7)

where 𝑦_𝑡is the series of correlations corresponding to our group of interest and 𝑡 is the linear trend over time. The obtained results are shown in table 𝐼𝐼. The Eurozone shows an upward trend significant at the 1%-level of an increase of 0.003 per quarter. For the EU members that do not participate in the euro, there is no evidence of a significant trend. The test for the difference between the two correlations, which is again defined as the non-Eurozone correlation subtracted from the Eurozone

correlation, is done separately. For the difference in comovements, we find a 1%-significant upward trend of 0.002 per quarter indicating that the correlations of the Eurozone and other EU members have converged between 1995 and 2015. This seems in accordance with figure 𝐼 which shows that the cross-country correlation of the euro area has increased towards the correlation of the non-euro area and at some points even surpassed it. So, in contrast to our expectations, correlations between the two regions have actually converged over time. We will come back to this in our robustness checks in the next section.

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 Euro Non-euro

(12)

Subsequently, tests are done to examine which macroeconomic factors may influence the movement of correlations over time. We look at macroeconomic variables as well as measures for the amount of relevant observations. For this, we will again use a time series regression separate for each region, but this time with multiple variables. The regression is as follows:

𝑌_𝑖,𝑡 = 𝑐_𝑖+ 𝛾_𝑖′∙ 𝑋_𝑖,𝑡 + 𝜀_𝑖,𝑡 (8)

where 𝑌_𝑖,𝑡 is the estimated correlation for one region in quarter 𝑡, and 𝑋_𝑖,𝑡 are the factors which may influence correlations. For the factors, we will use the following estimates. First, similar to Bekaert and Harvey (1997), Ng (2000), and Baele and Inghelbrecht (2009), to measure trade openness, the ratio of imports plus exports over nominal GDP is used. Data is retrieved from Datastream and the proportion, expressed as a percentage, is found by summing up the imports, exports and the GDP for all countries within a region whenever data is available. The second factor is economic growth for which the same proxy as Eiling and Gerard (2015) is used, namely growth in real GDP per capita for both regions. Data is again retrieved from Datastream and is found by first calculating the nominal GDP growth for the whole region and then adjusting it for the equally weighted inflation rate. For the non-Eurozone, the inflation rate is corrected for relatively high inflation rates by excluding post-soviet emerging markets prior to 1999, as defined by the Morgan Stanley Capital International (MSCI; see Table 𝐴1 in the Appendix). Third, the average number of countries for which equity return data was available per quarter is included to account for countries that are later included in our sample. For the Eurozone this number varies between twelve and seventeen and for the other EU members it varies between almost six and eleven.1_{Finally, for the euro area, the number of countries}

that have adopted the euro and for which data was available per quarter is

1_{For two countries, the Czech Republic and Hungary, data starts one day later, on January 3}rd_{1995, which}

explains why there are almost six countries in the beginning.

Table 𝑰𝑰. Testing for trends in quarterly equity market correlations for the Eurozone and other EU-members and the difference between the two for the whole sample

This table reports the summary statistics of the quarterly correlations for both regions, and the difference between the two, from Q1 1995 up until Q4 2015, totaling to eighty-four periods. The difference between the two correlations is found by subtracting the non-Eurozone from the Eurozone correlation. ‘DW’ gives the Durbin-Watson statistic to measure the first-order autocorrelation.

***, **, and * denote 1%, 5%, and 10% significance levels.

𝛽𝑡𝑖𝑚𝑒 _𝑅2 _𝑡 _𝑝 _𝐷𝑊

Eurozone 0.003∗∗∗ 0.259 5.349 0.000 0.914

Non-euro 0.000 0.005 0.622 0.536 1.066

(13)

included as a variable. This number is zero prior to 1999 and increases from eleven to seventeen afterwards.

Our findings for both regions are presented in table 𝐼𝐼𝐼. For the euro area, we find that for none of the regressions the real GDP per capita growth has any

significant effect on correlations. In the first regression, trade openness has a 1%-significant positive effect of an additional 0.008 in correlations for every percentage increase in exports and imports over GDP. However, if we add the number of

observations available, the effect of trade openness becomes significantly negative, and if we add the number of countries using the euro, it has no significant effect at all. The former negative effect is not in line with the discussed literature and our expectations. According to Bekaert, Hodrick, and Zhang (2009), this negative effect of trade openness could be explained by an increase in competitiveness and industrial specialization due to higher economic integration which in turn lower cash flow correlations. Furthermore, the number of observations available and the number of countries that have adopted the euro both have a positive effect significant at the 1%-level of 0.085 and 0.011, respectively, for every additional country added to the sample. For the other EU members that have not adopted the euro, we find no

Table 𝑰𝑰𝑰. Testing for factors in quarterly equity market correlations for the Eurozone and other EU members for the whole sample

This table reports the OLS regression results for both regions from Q1 1995 up until Q4 2015. The quarterly correlation estimates for each region are regressed on a series of macroeconomic variables as well as measures for the number of observations.

Macroeconomic variables include trade over GDP and real GDP per capita growth, and the observation measures include the number of observations available as well as the number of observations using the euro. The two-sided 𝑡-statistics are shown in parenthesis below the estimated coefficients.

regr. trade/GDP GDP growth Observations

available Observations using euro 𝑅2 Eurozone 1) 0.008∗∗∗ (4.927) -0.000 (-1.847) 0.232 2) −0.006∗∗ (-2.175) -0.001 (-0.067) 0.085∗∗∗ (5.436) 0.439 3) 0.001 (0.272) 0.002 (0.142) 0.011∗∗∗ (2.809) 0.301 Non-euro 1) -0.000 (-0.181) 0.002 (0.521) 0.004 2) 0.004 (1.395) 0.001 (0.362) −0.019∗ (-1.913) 0.048

(14)

significant effect for the trade openness, nor for economic growth. For the number of observations available, however, we find, in contrast to the Eurozone region, a

negative effect significant at the 10% level of -0.019 per country added. This could be explained by the fact that we add smaller emerging markets to our sample which at first only consisted of relatively developed markets with higher cross-country

correlations. The fit of our regression, indicated by 𝑅2, is rather low for the non-euro area when compared to the Eurozone. This is further examined in the next section.

VI. Robustness Check

This section presents a series of additional robustness tests using the same methods of examination presented in the previous section. First, tests are done for a shorter sample period starting in 2001. Next, annual tests are performed in order to add more macroeconomic variables. Finally, a different set of countries is used to focus on developed markets only.

In Figure 𝐼, a clear distinction can be seen between the first six years, where the non-euro correlations are significantly higher than the Eurozone correlations, and the rest of the sample, where the correlations of both regions move more or less in the same direction. Therefore, a separate sample using only data after 2001, where the two correlations are equal for the first time, is used. Correlation summary statistics, time trend tests, and the factor regressions are presented in Tables 𝐴2, 𝐴3, and 𝐴4 in de the Appendix, respectively.

In contrast to the sample using the whole time period, the average correlation of the Eurozone (0.3311) is now higher than the average correlation of the group of other EU members (0.3091), which corresponds to our expectations and the existing literature. The difference before 2001 couldbe explained by the lack of certain

(emerging) markets for which no data was yet available. This is especially the case for the non-euro area for which five eastern European emerging markets have no data available yet, such that the comovements are overestimated prior to 2001.

Therefore, our new sample can be seen as a sample where almost all countries have data available. Only two countries in the Eurozone and two countries in the group of other EU members have no data available as of 2001.2_{This also influences our time}

trend tests, shown in Table 𝐴3. Now, there is no significant trend for the euro area, but there is an upward trend of 0.003 per quarter significant at the 1%-level for the other EU members. Furthermore, the differences seem to get smaller with 0.003 per

2_{For the Eurozone, data starts on September 6}th_{, 2004, for Cyprus and on April 3}rd_{, 2006, for Slovenia. For the}

(15)

quarter indicating that correlations are converging. As for the regressions with factors that may influence comovements, shown in Table 𝐴4, we find that, similar to the tests for the whole sample period, for none of the regressions there is a significant effect of growth in real GDP per capita on correlations. Observations available again has a rather strong significant effect for both the Eurozone (0.080) as well as for other EU members (0.095), while it was previously significantly negative for the latter region (-0.019). It seems unlikely that the addition of only two countries, namely Croatia and Romania, has such a profound effect, but if we add a time variable which is also strictly increasing, we find that observations available still has a positive effect of 0.070 per additional country significant at the 5%-level. For the regressions

without the amount of observations, trade openness shows no significant effect for the Eurozone, whilst it previously did, and for the other EU members, a 5%-significant upward trend of 0.006 per percentage increase of imports and exports over GDP, whilst it previously showed no significant effect. The fit of our regressions for the non-euro area is also better compared to the regressions done using the whole sample period and the 𝑅2 for the other EU members is now even higher than the 𝑅2 for the Eurozone, for which it is lower than in section five.

Next, in order to measure the effect of the development of equity markets, we include market capitalization over GDP as a macroeconomic variable in our

regression. Bekaert and Harvey (1997), Carrieri, Errunza, and Hogan (2007), and Eiling and Gerard (2015) use the same proxy. Due to the limited data availability of this variable, we now define 𝑡 as a year, and use annual correlations which are found by simply taking the average over four quarters for each year. The same is done for exports and imports over GDP. Data on market capitalization over GDP is retrieved from the World Bank and is for each region found by taking the equally weighted average over all countries. Annual data for the real GDP per capita growth is retrieved from the IMF. For the Eurozone, it is assumed that Latvia and Lithuania do not

significantly affect the total Eurozone inflation. Inflation rates for the group of other EU members are again adjusted by excluding the inflation rate of emerging markets prior to 1999. The movement of annual correlations over time is shown in Figure 𝐴2 in the Appendix. Correlation summary statistics, time trend tests, and the factor regressions are shown in Tables 𝐴5, 𝐴6, and 𝐴7, respectively.

The mean of the correlations and the difference between the two is of course the same as the ones found in section five. The time trend tests also show similar findings, but on a different scale, namely a significant upward trend in correlations of 0.010 per year for the Eurozone and 0.009 per year for the differences between the two, again indicating a convergence over time. The group of other EU members

(16)

shows no significant trend. Furthermore, in none of the factor regressions does the development of stock markets show a significant effect on comovements. The additional variable does however influence the significance of other factors. For the first time, economic growth shows a significant effect. For the Eurozone, we find an increase of 0.004 in correlations for every additional percentage point in GDP per capita growth. Also, the effect of the number of countries that have adopted the euro in our sample becomes insignificant with the addition of the factor for equity market development. For other EU members, nothing changes except for the number of observations which now has an insignificant effect. For the Eurozone, the fit of our model has increased and for the non-euro area also, but not as much as for our limited sample period shown earlier. All in all, the inclusion of equity market

development does not seem to affect our regressions that much. This could be due to the fact that European markets are already relatively developed compared to the emerging markets examined in the literature discussed before.

Finally, we make a separate sample for developed markets only. This is done because developed European markets tend to have a significantly higher average correlation compared to emerging Eastern European markets (Eiling and Gerard, 2015). We focus on developed markets as that encompasses the majority of the Eurozone and the EU. Similar to Bekaert, Hodrick, and Zhang (2009), the market classifications by Morgan Stanley Capital International (MSCI) as of 2015 are used and can be found in Table 𝐴1 in the Appendix. Greece is considered a developed market as it has been so for the majority of time between 1995 and 2015. For the Eurozone, thirteen out of

seventeen countries are left, and for other EU members, only three out of eleven countries remain part of our sample. This also largely omits the problem of some smaller emerging markets having only limited data available as now only two countries for the Eurozone and none for the other EU members do not have data from the beginning.3_{However, it also introduces a new problem, namely that there}

are only three countries left for the non-euro area which leads to greater

comovement estimates. This can be seen in Figure 𝐼𝐼which shows the movement of correlations for both regions over time. The correlations for the non-euro area are strictly higher than those of the euro area, which contradicts existing literature. We thus expect that our findings cannot be extrapolated for the whole non-euro area. Summary statistics and the results of time trend tests are shown in Tables 𝐴8 and 𝐴9, respectively. For the factor regressions, new variables are estimated excluding the emerging markets and the results can be found in Table 𝐴10.

(17)

Figure 𝑰𝑰. Movement of the quarterly correlation for the Eurozone and other EU-members for developed markets only

The average of the Eurozone correlations has increased, but not as much as for the other EU members which has more than doubled compared to the full sample used in section five. We do however find that the average correlation for the thirteen remaining euro countries (0.4526) has surpassed the average of the full sample of eleven non-euro countries (0.3236) found in the previous section. As for time trends, we still find a significant upward trend for the Eurozone, albeit smaller, of 0.001 per quarter, and now also one for the non-euro area of 0.001 per quarter. No

convergence or divergence over time can be found between the two. The factor regressions show that the effect of trade openness for the Eurozone has declined or become insignificant for the first and second regression. This may indicate that

bilateral trade probably plays a more profound role in emerging markets in explaining equity market comovements. Also, the effect of the number of observations using the euro now has an insignificant effect. This could be explained by the fact that the introduction of the euro had a bigger impact on emerging equity markets than on developed markets. For the other EU members, nothing changes and the test using the number of observations cannot be performed as it is constant over time. All in all, there seems to be a real difference between developed and emerging European markets as has also been shown in previous studies.

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Correl at ion s Eurozone Non-euro

(18)

VII. Conclusion

In this paper, cross-country equity market return correlations for the Eurozone and for the group of other EU members are examined over time, covering the pre-euro period from 1995 and the post-euro period up until 2015. We use the average of the daily cross-sectional dispersion to find the quarterly correlations and furthermore examine time trends and look into the factors that affect comovements.

First, we find that the average correlation of the non-euro area is surprisingly higher than the average of the Eurozone, which is robust for all our tests except for the shorter time period which starts in 2001. We do however find that the average correlation of the thirteen developed Eurozone markets is higher than the average of the full sample of eleven non-euro countries. In most of our tests we find an upward trend for the correlations within the Eurozone, while a significant upward trend for the other EU members is found only for the limited sample period. This could indicate that the euro has indeed increased equity correlations since its introduction. Also, a convergence between correlations is found in all tests except for developed markets only. This again goes against our expectations that the Eurozone correlation would diverge from the correlation of the non-euro area.

Second, the factor regressions show that the (macroeconomic) channels which may affect comovements differ for both regions. Economic growth and stock market development do however not seem to have a significant effect for either region. For the Eurozone, the effect of trade openness is the same for all our robustness

exercises, but the direction and significance differ if we include different variables. At first, we find a positive significant result, but if we include observations available the effect becomes negative and if we include observations available using the euro, the channel becomes insignificant at all. Observations available itself has a relatively constant significant positive effect in all our tests. The effect of observations available using the euro is positive, but its significance is not robust for our checks, such that it is unclear if the euro had indeed a positive effect on correlations. For the non-euro area, trade openness does not seem to affect comovements except for one single regression. The effect of the amount of observations available is also unclear as we find significantly positive, negative as well as insignificant results. The fit of our

regression is the best for our annual tests including stock market development for the Eurozone and for a shorter sample period for the non-Eurozone.

The implications of our results for market participants are as follows. Since 1995, diversification across markets in the EU has become less effective for investors. For developed European markets, the effectiveness is even smaller, but it has not changed as radically. Moreover, this study shows the macroeconomic factors that

(19)

connect national stock markets and how they differ between the Eurozone and other EU members. The results might also be of interest for the examination of possible future euro adapters such as Bulgaria, Croatia, and Romania.

One of the limitations of this study is that equity return correlation estimates are not adjusted for serial correlation. Also, this paper examines the Eurozone and the rest of the EU as a whole, while the financial integration with the rest of Europe for individual countries might also be of interest, even more so for investors.

Additionally, Figure 𝐼 seems to show rather sudden shifts between upward and downward trends in the movement of correlations over time which affect both regions. Future research could examine these potential structural breaks and its causes.

In conclusion, the research question is answered as the correlations for the Eurozone and for EU members that have not adopted the euro have converged over time which is robust for most of our checks. The channels through which

comovements are affected and the precise effect of the euro, however, remain largely unclear so the results should be addressed carefully.

(20)

Reference List

Abad, P., Chuliá, H., and Gómez-Puig, M. (2010). EMU and European government bond market integration. Journal of Banking and Finance, 34(12), 2851-2860.

Adjaoute, K., and Danthine, J. (2004). Equity returns and integration: is Europe changing? Oxford Review of Economic Policy, 20(4), 555-570.

Baele, L., and Inghelbrecht, K. (2009). Time-varying Integration and International diversification strategies. Journal of Empirical Finance, 16(3), 368-387.

Bekaert, G., and Harvey., C.R. (1997). Emerging equity market volatility. Journal of Financial

Economics, 43(1), 29-77.

Bekaert, G., Harvey, C.R., Lundblad, C.T., and Siegel, S. (2013). The European Union, the Euro, and equity market integration. Journal of Financial Economics, 109(3), 583-603. Bekaert, G., Hodrick, R., and Zhang, X. (2009). International Stock Return

Comovements. Journal of Finance, 64(6), 2591-2626.

Brooks, R., and Del Negro, M. (2002). The Rise in Comovement Across National Stock

Markets: Market Integration or Global Bubble? (IMF Working Paper). International

Monetary Fund.

Cappiello, L., Kadareja, A., and Manganelli, S. (2010). The Impact of the Euro on Equity Markets. Journal of Financial and Quantitative Analysis, 45(2), 473-502.

Carrieri, F., Errunza, V., and Hogan, K. (2007). Characterizing World Market Integration through Time. Journal of Financial and Quantitative Analysis, 42(4), 915-940. Chambet, A., and Gibson, R. (2008). Financial integration, economic instability and trade

structure in emerging markets. Journal of International Money and Finance, 27(4), 654-675.

Chen, N., and Zhang, F. (1997). Correlations, trades and stock returns of the Pacific-Basin markets. Pacific-Basin Finance Journal, 5(5), 559-577.

Coeurdacier, N., and Martin, P. (2009). The geography of asset trade and the euro: Insiders and outsiders. Journal of The Japanese and International Economies, 23(2), 90-113. De Santis, R., and Gerard, B. (2006). Financial integration, international portfolio choice and

the European monetary union (Working Paper No. 626). European Central Bank,

Eiling, E., Gerard, B., Hillion, P., and de Roon, F.A. (2012). International portfolio diversification: Currency, industry and country effects revisited. Journal of

International Money and Finance, 31(5), 1249-1278.

Eiling, E., and Gerard, B. (2015). Emerging Equity Market Comovements: Trends and Macroeconomic Fundamentals. Review of Finance, 19(4), 1543-1585.

Engle, R., and Kelly, B. (2012). Dynamic Equicorrelation. Journal of Business & Economic

Statistics, 30(2), 212-228.

Eun, C., and Shim, S. (1989). International Transmission of Stock Market Movements. Journal

of Financial and Quantitative Analysis, 24(2), 241-256.

Ferreira, M. Almeida, and Ferreira, M. Ângelo. (2006). The Importance of Industry and Country Effects in the EMU Equity Markets. European Financial Management, 12(3), 341-373.

Flavin, T. (2004). The effect of the Euro on country versus industry portfolio

diversification. Journal of International Money and Finance, 23(7), 1137-1158. Francis, B.B., Hunter, D.M., and Hasan, I. (2002). Return-Volatility Linkages in the

International Equity and Currency Markets (Discussion Paper No. 2002/09). Bank of

(21)

Gonçalves, C.E.S., Rodrigues, M., and Soares, T. (2009). Correlation of business cycles in the euro zone. Economics Letters, 102(1), 56-58.

Grubel, H. (1968). Internationally Diversified Portfolios: Welfare Gains and Capital Flows. The

American Economic Review, 58(5), 1299-1314.

Jappelli, T., and Pagano, M. (2008). Financial market integration under EMU (CFS Working Paper No. 2008/33). Goethe University, Frankfurt.

McAllister, P., and Lizieri, C. (2006). Monetary Integration and Real Estate Markets: The Impact of Euro on European Real Estate Equities. Journal of Property Research, 23(4), 281-303.

Moerman, G. (2008). Diversification in euro area stock markets: Country versus industry. Journal of International Money and Finance, 27(7), 1122-1134.

MSCI (2017, October). MSCI Global Investable Market Indexes Methodology. Retrieved from https://www.msci.com/eqb/methodology/meth_docs/MSCI_October2017_GIMIMeth odology.pdf

MSCI (n.d.). Market Classification. Retrieved from https://www.msci.com/market-classification

Ng, A. (2000). Volatility spillover effects from Japan and the US to the Pacific–Basin. Journal

of International Money and Finance, 19(2), 207-233.

Quinn, D., and Voth, H. (2008). A Century of Global Equity Market Correlations. American

Economic Review, 98(2), 535-540.

Savva, C., and Aslanidis, S. (2010). Stock market integration between new EU member states and the Euro-zone. Empirical Economics, 39(2), 337-351.

(22)

Appendix

𝑻𝒂𝒃𝒍𝒆 𝑨𝟏. Descriptive statistics equity returns and risk-free rate

This table shows the summary statistics of the equity returns of all Eurozone and other EU countries as well as of the risk-free rate between 1995 and 2015. ‘Market classification’ denotes if a country is considered a developed or an emerging market by Morgan Stanley Capital International (MSCI) as of 2015 (MSCI, 2017, n.d.). Greece is considered a developed market as it has been so for the majority of time between 1995 and 2015. For the sake of simplicity, frontier and standalone markets are also considered an emerging market. ‘Starting date’ indicates the first day when data was available and ‘data points’ denotes the total amount of daily observations available for that country. The mean and standard deviation are given annually, and the minimum and maximum equity return are given per day.

market classification starting date data points mean (% p.a.) stdev (% p.a.) min (% p.d.) max (% p.d.) Eurozone

Austria Developed 02-Jan-95 5479 1.35 23.10 -10.57 13.60

Belgium Developed 02-Jan-95 5479 11.48 19.34 -7.99 9.77

Cyprus Developed 06-Sep-04 2954 -20.32 41.56 -14.38 18.48

Estonia Emerging 04-Jun-96 5108 18.50 24.26 -19.42 13.72

Finland Developed 02-Jan-95 5479 13.69 28.49 -15.99 15.66

France Developed 02-Jan-95 5479 7.71 22.29 -8.90 10.91

Germany Developed 02-Jan-95 5479 12.33 23.69 -8.51 11.39

Greece Developed 02-Jan-95 5479 -7.37 34.00 -16.59 17.58

Ireland Developed 02-Jan-95 5479 8.91 20.79 -13.04 10.20

Italy Developed 02-Jan-95 5479 2.97 23.48 -8.28 11.60

Luxembourg Developed 05-Jan-99 4433 3.04 20.50 -10.92 9.51

Malta Emerging 28-Dec-95 5221 8.20 11.89 -7.32 10.04

Netherlands Developed 02-Jan-95 5479 7.81 21.83 -8.85 10.23

Portugal Developed 02-Jan-95 5479 1.30 18.84 -9.87 10.73

Slovakia Emerging 02-Jan-95 5479 9.91 23.85 -9.00 12.61

Slovenia Emerging 03-Apr-06 2544 2.09 19.22 -33.20 49.89

Spain Developed 02-Jan-95 5479 -1.93 26.68 -9.62 15.63

Non-euro

Bulgaria Emerging 23-Oct-00 3964 17.24 24.90 -18.88 23.40

Croatia Emerging 03-Jun-02 3544 -1.20 26.49 -10.39 14.44

Czech Republic Emerging 03-Jan-95 5478 8.93 24.86 -14.53 18.23

Denmark Developed 02-Jan-95 5479 20.61 26.49 -10.75 14.06

Hungary Emerging 03-Jan-95 5478 24.78 34.18 -18.37 19.86

Latvia Emerging 04-Apr-00 4173 17.74 23.10 -13.69 12.30

Lithuania Emerging 04-Apr-00 4173 14.23 17.39 -12.65 12.59

Poland Emerging 02-Jan-95 5479 10.83 31.48 -11.10 11.11

(23)

𝑭𝒊𝒈𝒖𝒓𝒆 𝑨𝟏. Movement of the difference between quarterly correlations of the Eurozone

and other EU members over time

This figure shows the movement of the difference between quarterly cross-country equity correlations in the Eurozone and other EU members over time. The difference is found by subtracting the correlation of the non-Eurozone from the correlation of the Eurozone. 𝑻𝒂𝒃𝒍𝒆 𝑨𝟏. (Continued) market classification starting date data points mean (% p.a.) stdev (% p.a.) min (% p.d.) max (% p.d.) Non-euro

Romania Emerging 01-Dec-05 2631 5.41 33.26 -25.86 12.35

Sweden Developed 02-Jan-95 5479 17.09 30.39 -10.53 12.11

UK Developed 02-Jan-95 5479 4.72 20.22 -8.96 9.97 Risk-free rate German 3M 𝑟𝑓 02-Jan-95 5479 3.23 0.10 -0.002 0.020 -0,4 -0,3 -0,2 -0,1 0 0,1 0,2 0,3 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Dif fe re n ce in co rre lat ion s

(24)

Table 𝑨𝟐. Summary statistics of quarterly correlations for the Eurozone and other EU-members and the difference between the two for a limited sample period

This table reports the summary statistics of the quarterly correlations for both regions, and the difference between the two, from Q1 2001 up until Q4 2015, totaling to sixty periods. The difference between the two correlations is found by subtracting the non-Eurozone from the Eurozone correlation.

Correlations

Eurozone 0.3311 0.3157 0.1064 0.15 0.60 0.762 0.329

Non-euro 0.3091 0.3046 0.1092 0.10 0.62 0.392 -0.042

Difference 0.0219 0.0229 0.0882 -0.17 0.23 -0.013 -0.118

Table 𝑨𝟑. Testing for trends in quarterly equity market correlations for the Eurozone and other EU-members and the difference between the two for a limited sample period

This table reports the summary statistics of the quarterly correlations for both regions, and the difference between the two, from Q1 2001 up until Q4 2015, totaling to eighty-four periods. The difference between the two correlations is found by subtracting the non-Eurozone from the non-Eurozone correlation. ‘DW’ gives the Durbin-Watson statistic to

measure the first-order autocorrelation.

𝛽𝑡𝑖𝑚𝑒 𝑅2 _𝑡 _𝑝 _𝐷𝑊

Eurozone 0.000 0.000 0.022 0.983 1.044

Non-euro 0.003∗∗∗ 0.187 3.657 0.001 1.288

(25)

Table 𝑨𝟒. Testing for factors in quarterly equity market correlations for the Eurozone and other EU members for a limited sample period

This table reports the OLS regression results for both regions from Q1 2001 up until Q4 2015. The quarterly correlation estimates for each region are regressed on a series of macroeconomic variables as well as measures for the number of observations. Macroeconomic variables include trade over GDP and real GDP per capita growth, and the observation measures include the number of observations available as well as the number of observations using the euro. The two-sided 𝑡-statistics are shown in parenthesis below the estimated coefficients.

available Observations using euro 𝑅2 Eurozone 1) -0.001 (-0.295) 0.001 (0.968) 0.002 2) −0.010∗∗∗ (-2.870) 0.001 (0.966) 0.080∗∗∗ (3.559) 0.186 3) -0.003 (-0.767) 0.002 (0.113) 0.007 (0.794) 0.013 Non-euro 1) 0.006∗∗ (2.569) -0.002 (0.579) 0.104 2) -0.002 (-0.599) -0.003 (-0.869) 0.095∗∗∗ (3.200) 0.243

Figure 𝑨𝟐. Movement of the annual average correlation for the Eurozone and other EU-members over time for the whole sample

0 0,1 0,2 0,3 0,4 0,5 0,6 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Correl at ion s Euro Non-euro

(26)

Table 𝑨𝟓. Summary statistics of annual correlations for the Eurozone and other EU-members and the difference between the two for the whole sample

This table reports the summary statistics of the annual average correlations for both regions, and the difference between the two, from 1995 up until 2001, totaling to twenty-one

periods. The difference between the two correlations is found by subtracting the non-Eurozone from the non-Eurozone correlation.

Correlations

Eurozone 0.2831 0.2783 0.1088 0.07 0.53 0.269 0.467

Non-euro 0.3236 0.3018 0.0880 0.18 0.47 0.108 -0.942

Difference -0.0405 0.0099 0.1193 -0.27 0.12 -0.563 -0.926

Table 𝑨𝟔. Testing for trends in annual equity market correlations for the Eurozone and other EU-members and the difference between the two for the whole sample

This table reports the summary statistics of the annual correlations for both regions, and the difference between the two, from 1995 up until 2015, totaling to twenty-one periods. The difference between the two correlations is found by subtracting the non-Eurozone from the Eurozone correlation. ‘DW’ gives the Durbin-Watson statistic to measure the first-order autocorrelation.

Eurozone 0.010∗∗∗ 0.347 3.177 0.005 0.586

Non-euro 0.001 0.007 0.354 0.728 0.890

(27)

Table 𝑨𝟕. Testing for factors in annual equity market correlations for the Eurozone and other EU members for whole sample

This table reports the OLS regression results for both regions from 1995 up until 2015. The annual correlation estimates for each region are regressed on a series of macroeconomic variables as well as measures for the number of observations. Macroeconomic variables include trade over GDP, real GDP per capita growth, and market capitalization over GDP, and the observation measures include the number of observations available as well as the number of observations using the euro. The two-sided 𝑡-statistics are shown in parenthesis below the estimated coefficients.

***, **, and * denote 1%, 5%, and 10% significance levels. regr. trade/ GDP GDP growth MCap/ GDP Observations available Observations using euro 𝑅2 Eurozone 1) 0.009∗∗∗ (3.130) 0.003 (1.078) 0.000 (0.892) 0.379 2) −0.009∗ (-1.819) 0.004∗∗ (2.287) 0.000 (-0.466) 0.100∗∗∗ (4.034) 0.692 3) 0.002 (0.372) 0.002 (0.288) 0.000 (0.884) 0.010 (1.477) 0.453 Non-euro 1) -0.001 (-0.352) -0.002 (-0.827) -0.001 (-0.444) 0.059 2) 0.003 (0.564) -0.002 (0.495) -0.001 (0.697) -0.018 (-1.080) 0.123

Table 𝑨𝟖. Summary statistics of quarterly correlations for the Eurozone and other EU-members and the difference between the two for developed markets only

This table reports the summary statistics of the quarterly correlations for both regions, and the difference between the two, from Q1 1995 up until Q4 2015, totaling to eighty-four periods. The difference between the two correlations is found by subtracting the non-Eurozone from the non-Eurozone correlation. Only developed markets as defined by MSCI are used, which can be found in Table 𝐴1.

Correlations

Eurozone 0.4526 0.4543 0.1472 0.16 0.75 0.083 -0.537

Non-euro 0.6918 0.6930 0.1025 0.39 0.89 -0.277 0.042

(28)

Table 𝑨𝟏𝟎. Testing for factors in quarterly equity market correlations for the Eurozone and other EU members for developed markets only

This table reports the OLS regression results for both regions from Q1 1995 up until Q4 2015 for developed markets only, as defined by MSCI, which can be found in Table 𝐴1. The quarterly correlation estimates for each region are regressed on a series of macroeconomic variables as well as measures for the number of observations. Macroeconomic variables include trade over GDP and real GDP per capita growth, and the observation measures include the number of observations available as well as the number of observations using the euro. The two-sided 𝑡-statistics are shown in parenthesis below the estimated coefficients.

available Observations using euro 𝑅2 Eurozone 1) 0.004∗ (1.894) 0.004 (0.262) 0.042 2) -0.003 (-0.779) 0.005 (0.352) 0.082∗∗ (1.998) 0.088 3) 0.000 (0.137) 0.005 (0.344) 0.007 (1.406) 0.066 Non-euro 1) 0.001 (0.161) -0.003 (-0.916) 0.011

Table 𝑨𝟗. Testing for trends in quarterly equity market correlations for the Eurozone and other EU-members and the difference between the two for developed markets only

This table reports the summary statistics of the quarterly correlations for both regions, and the difference between the two, from Q1 1995 up until Q4 2015, totaling to eighty-four periods. The difference between the two correlations is found by subtracting the non-Eurozone from the non-Eurozone correlation. Only developed markets as defined by MSCI are used, which can be found in Table 𝐴1. ‘DW’ gives the Durbin-Watson statistic to measure the first-order autocorrelation.

Eurozone 0.001∗∗ 0.057 2.222 0.029 0.842

Non-euro 0.001∗∗∗ 0.090 2.846 0.006 1.445

The Eurozone and cross-country equity return correlations : divergence with the rest of the EU?