UK’s stock market integration with the European Union

(1)

i

UK’s stock market integration with the European Union

Master Thesis in

FINANCE (Track: Asset Management)

Johannes Mahr (11391294)

07/2016

Supervised by:

dhr. dr. J.E. (Jeroen) Ligterink

University of Amsterdam Amsterdam Business School Faculty for Economics and Business Plantage Muidergracht 12, 1018 TV Amsterdam

johannesmahr@gmx.at + 43 676 949 5482

(2)

ii

Statement of Originality. This document is written by Johannes Mahr 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)

iii

Abstract. EU stock markets have exhibited a strong trend towards financial integration since

the mid-nineties. After having recovered from massive economic shocks, as the financial- or

the sovereign debt crisis, the EU and its efforts promoting financial integration across member

states faced another striking event in June 2016, viz. the British vote in favor of leaving the

EU. Thus, this paper takes the perspective from the UK and explores a long-term relationship

with EU equity markets as well as the impact of the Brexit on equity market integration. Using

industry valuation differentials, a sample with 26 European countries unveils that UK has

un-dergone a period of integration with EU members between 1990-2016. A sub-sample further

reveals time-varying properties of the financial integration measure and shows that the UK’s

trend towards integration has partially reversed in recent years.

JEL classification: G11, F36

(4)

iv

The establishment of the European Union (EU) along with the introduction of a common currency and monetary policy has become the most ambitious effort in mankind to unify different countries on a cultural, social as well as economic dimension. The latter is specifically related to an integration of capital markets, which depicts a phenomenon of increasingly correlated security movements across equity and debt markets between member countries in the EU (regional integration) as well as the EU as a trading block and rest of the world (global integration). As set out in the annual reports of the European Central Bank (ECB), capital market - or financial integration is one of the most important goals as it ensures a functioning and efficient implementation of monetary policies (European Central Bank (2017)). However, it is not only the regulatory bodies having a deep interest in strong financial linkages of EU member countries. Multinational companies (MNCs) as well as investment firms like-wise require an understanding of the underlying mechanisms and time-varying developments for a sound decision making process. MNCs, for instance, require knowledge of the determinants of the cost of capital for asset pricing or financial forecasting purposes. One of these determinants can be found in the level of financial integration, which predicts a negative relationship with the cost of capital accord-ing to prevalent theories (e.g. Stulz (1999)). Investment firms, on the other side, are interested in the price of risk and international diversification benefits when it comes to an investment decision. Both factors tend to decrease with proceeding financial integration (Kearney and Lucey (2004)).

Stock market integration in the EU has been subject to numerous studies. In line with the goal of regu-latory EU bodies, contemporary literature suggests that EU equity market integration strengthened in-deed during the past two decades: Bekaert, Harvey, Lundblad and Siegel (2013) for instance, provide evidence that EU membership decreased industry valuation differentials among EU country pairs and that the implementation of the Euro had minor effects on integration in the same direction. Syllignakis and Kouretas (2010) show in their analysis that also Eastern European countries have become more financially connected after having joined the EU. However, they also stress that there was a significant slowdown of this process observable in the years after the financial crisis (2008-2009). Furthermore, a recent research article from Bartram and Wang (2015) finds that the European sovereign debt crisis

(6)

2

following the financial crisis also contributed to a mitigation of the European financial market integra-tion process.

In view of the British vote in June 2016 in favor of leaving the European Union (hereinafter referred to as “Brexit”), the threat of a disintegration of the European financial markets appears to be likely. As revealed above, European capital market’s integration proved to be sensitive to such striking events. It follows that a tendency towards an integrated European capital market may have further weakened due to events like these. Therefore, this paper aims at shedding light on stock market integration of the United Kingdom (UK) with the EU over time, with the ultimate goal of yielding an understanding of UK’s contribution to capital market integration in the EU as well as the Brexit’s implications to it.

Most of the studies in this field use long-term sample periods (e.g. Aggraval, Lucey et al. (2013) or Panayotis and Anna (2013)) because of low-frequency data unavailability of economic variables. More-over, there has been a tendency amongst researchers to investigate correlations on debt markets as a well-established methodology proxy for capital market integration (e.g. Cipollini and Coakley (2015) Christiansen (2014)). By contrast, there seems to be little contemporary literature exploring equity mar-ket integration, which may be attributed a greater extent of disagreement on the issue of which ap-proaches can be considered as useful techniques or measurements. However, I adopt a relatively new methodology as proposed by Bekaert et al. (2011), which is rooted in the convergence of industry val-uation differentials across an integrated market, rules out most of the concerns raised by researchers regarding stock market integration measures. In this sense, this paper investigates in a first analysis a long-term equity market integration between the UK and EU member states as well as possible conse-quences of the time after the financial crisis, including the sovereign debt crisis, by investigating the evolution of UK’s stock market integration in the period from 1990 to 2016. To my knowledge, the most recent study about UK’s equity market integration with European countries was conducted by Antonios et al. (2007) in which they observed that UK equity markets are highly integrated with Europe. They also found that correlations between cross-country equities are higher during bear markets and lower in recovery periods. The first important question in this regard is if the UK has undergone a long-term period of integration with EU member states as compared to non-EU countries. I provide an

(7)

3

answer to this by unveiling that UK equity markets exhibit significantly higher integration with EU markets in the period of 1990-2016. In particular, the simplest specification of OLS estimates shows that UK stock markets are by 182 basis points more integrated with EU members than with non-EU countries. This EU-effect also survives the inclusion of time-varying- and time invariant controls as well as time- and country-pair fixed effects, which yields strong evidence that the UK stock markets have strongly integrated with EU equity markets over time. Furthermore, I show that a benchmark economy, i.e. Germany, displays slightly weaker integration levels in terms of an economical and sta-tistical perspective.

However, a long-term effect might not necessarily hold for a mid-term investigation capturing recent years and the time around the British vote for leaving the EU. Therefore, in a second part of my research I use a sample period from 2011 to 2016 along with low frequency observations and ultimately split it the sample to understand the EU equity markets’ reactions before and after the threat of Britain leaving the EU had become imminent. A consistent sample of 26 country pairs unveils striking results: Whereas the EU-effect for UK equity markets essentially remains intact for the period of 2011 to 2013, it disap-pears in the consecutive years completely. This suggests that UK’s long-term trend towards stock mar-ket integration with the EU has reversed in recent years.

This paper contributes to stock market integration literature in two ways: First, I employ a novel ap-proach initially proposed by Bekaert et al. (2011) and derive an equity market integration measure that captures the stock market integration from a UK perspective. The underlying mechanism involves the notion that industry valuations of equities converge with the magnitude of integration in a given stock market. In this regard, firm-level earnings yields of all publicly listed companies as reported by Datastream are used to form portfolios on the industry level, which not only reduces noise but also ensures that firms with similar growth or risk profiles are aggregated within one industry. Eventually, these differentials serve as a proxy for equity market segmentation and are explained by whether a country has a EU membership or not in an Ordinary Least Squares (OLS) regression framework. To my knowledge, this is the first paper to examine long-term equity market integration of the UK with the

(8)

4

EU by employing an analysis solely relying on country-pairs’ valuation differentials. Also, by conduct-ing a comparable analysis for Germany, I show that UK’s equity markets exhibit a relatively strong connection with the EU over time. Secondly and not less important, I reveal that a trend towards more integration has not survived the tumultuous times around the Brexit vote. After the threat of the Brexit had become concrete in 2014 (the party propagating the Brexit (UKIP) won the elections for the Euro-pean Parliament), the UK-EU integrational effect disappeared during the period from 2014 to 2016 whereas the previous years exhibit significant tendencies towards equity market integration. Speaking from a statistical view, from 2014 onwards UK’s EU membership cannot explain a variation UK-EU industry valuation differentials anymore. This appears to be a striking reversal in view of the strong results yielding evidence towards equity market integration between the UK and the EU for the long-term period OLS estimates.

The remainder of this paper is divided into five sections: After the abstract and introduction of Section 1, Section 2 provides a review of the main theories and applications in capital market integration liter-ature. Section 3 introduces the adopted methodology, the measure and its control variables. Section 4 presents the data and descriptive statistics. Section 5 and 6 show the main results, the limitations and robustness check. Section 6 concludes and provides an anchor for future research in this field.

2. Literature review 2.1. Financial integration

During the last two decades, the research conducted with respect to capital market integration has gained much importance (Sharma and Seth (2012). This trend can most likely be ascribed to increasing glob-alization, which involves international market integration as a phenomenon of decreasing barriers in the movement of labor, capital as well as goods and services. From an economic point of view, the rise of trading blocs and economic unions, the internationalization of markets for securities and significant advances in technology fostered the development of financial global links. In simple terms, financial integration means that all market participants have virtually no restrictions in the decision of portfolio allocation (Albulescu et al. (2015)). However, an optimal diversification strategy may be limited by the tendency to overinvest in domestic assets rather than following the principles of portfolio allocation

(9)

5

theory (Baele et al. (2007)). This phenomenon has been named the home bias puzzle. Literature pro-vides evidence that home bias indeed exists across developed markets, as Balta and Delgado (2009) show for EU markets or Dongmin and Lilian (2009) for US investors. Cooper et al. (2013) provide a literature survey on the equity home bias puzzle. They identify the main factors contributing to a home-biased investment decision to be higher costs of information, transaction and legal execution. The puz-zling element refers to the fact that global and regional trends towards financial integration should mit-igate the barriers to the optimal inclusion of foreign assets to an investor’s portfolio. However, they stress that the levels of home bias still depict a prominent role in investors’ minds today.

Increasingly interrelated markets and associated change of investor’s decision can directly or indirectly affect a country’s macroeconomic environment through national income, employment or exchange rates (Kearney and Lucey (2004)), which may eventually result in a set of positive or negative side-effects for the economy. The notion that financial integration across different markets can be seen as beneficial enjoys popularity among researchers. Baele et al. (2004) describe risk sharing as one out of three major advantages. Under full international risk sharing as defined by Obstfeld and Rogoff (1996) and the efficient market hypothesis, countries can completely protect themselves against domestic fluctuations in consumption and income through engaging in international diversification. Their theoretical model involves the notion that full risk sharing opportunities imply perfect correlation of consumption flows across two countries. However, at the same time the co-moving patterns do not exist with respect to economic shocks, what ultimately decouples idiosyncratic risk from fluctuations in consumption and allows a full insurance against domestic output shocks. With increasing cross-border financial asset holdings and declining investor’s home bias, the benefits of hedging portfolios against domestic eco-nomic shocks are even more pronounced. Bracke and Schmitz (2011) showed that net capital gains and dividend income serve as the main channels through which international risk sharing activities are ex-ecuted. They further show that the countercyclical properties of the proposed channels of risk sharing have gained more importance over time. The risk-sharing argument implies for an UK or EU investor that the long-term trend of increasing financial integration yields better protection opportunities against economic shocks. However, against the expectation that risk-sharing is closely linked with financial

(10)

6

integration in the Euro area, Ballabriga, and Villegas-Sánchez (2017) reveal in a recently published article that there is no evidence of decreased income fluctuation risk in response to stronger connected financial markets.

As suggested by Baele et al. (2004), a second and third potential benefit stemming from financial inte-gration are efficient capital allocation and economic growth. They define financial inteinte-gration by using three characteristics a fully integrated market typically has, which include an equal set of rules and regulations governing this market, unrestricted access to the market as well as all participant facing equal treatment by the market. This definition naturally implies that investors face less restrictions in integrated markets when it comes to the investment decision process. Therefore, they can invest their funds according to their expertise and beliefs about securities movements, which introduces improved capital allocation and brings financial markets closer to efficiency (Levine (2001)). The economic mechanism that positively links economic growth with financial integration seems intuitive. If the flow of capital is facilitated there should be positive effects on less developed countries through a productive reallocation of funds coming from more developed countries. This leads to an enhanced domestic output and increased competition in emerging markets, which eventually drives economic growth (Acemoglu and Zilibotti (1997)). In theory, the shift of capital may not only bring the necessary funds where they are needed, but also advances in technology and actively managed foreign direct investments, which are supposed to ensure a long-term economic growth (Rusek (2004)). However, the global evidence on those benefits associated with financial integration are somewhat restrictive. Obstfeld (2009), for ex-ample, caution that it is hard to find unambiguous evidence on the relation between financial openness and economic growth. The reason for this is the endogeneity inherent to the decision of financially liberalizing an economy. However, a majority of developing countries continue to seek economic growth through opening their financial system to the world. The author concludes that this may have a positive impact on economic growth under certain circumstances, including the implementation of pol-icies fostering stability or a strong trade balance and capital reserves. On the other side, a lack of ful-filling these criteria may lead to negative side-effects for economic growth. In a setting of developed countries, Rusek (2004) reached the conclusion that financial integration was the key for the relatively

(11)

7

high growth rates in the U.S compared to Europe or Japan around the turn of the millennium. However, the author identifies broadness, depth and liquidity of capital markets to be the main prerequisites for financial integration and an associated economic growth. A very recent paper of Epaulard and Pom-meret (2016) attempted to overcome the empirical issues and calibrated a stochastic growth model which unveiled that financial integration brings on average 0.8 to 1.7 percentage points of economic growth per year across emerging markets. Regarding the EU setting, there is consensus among re-searchers that financial integration led through an enhancement of financial systems to a rise in eco-nomic growth (e.g. Friedrich et al. (2013) or Dudian and Popa (2013)).

It shall be noted at this point that financial integration also has negative aspects to consider. In this respect, the financial crisis depicts an illustrative example by unveiling the strong connection or finan-cial links between global capital markets which enabled the devastating effects of the crash of the U.S. housing market to spread over the world.

2.2. Measuring capital market integration

A steadily increasing interconnected world has urged policy makers, institutional and private investors as well as researchers to gain a better understanding of the mechanisms driving the cross-country finan-cial linkages. After having pointed out potential advantages and drawbacks from finanfinan-cial integration as well as the underlying empirical evidence above, it is necessary to dive into the most important contributions and econometric techniques in this field.

Kearney and Lucey (2004) provide an overview on different approaches of measuring the interdepend-ence of financial markets. They state that all kinds of approaches aiming at measuring financial market integration are rooted in two applications: direct and indirect measurements. Among direct measures the most prominent representative is the equalization of interest rates or yield spreads. Embedded in the theory of the law of one price, this approach assumes that in a perfectly integrated financial market the rates of return should be equal across countries when converted to a mutual currency. The methodology used for the analysis in the following sections and proposed by Bekaert et al. (2013) can be assigned to this category. They proposed a novel measure of financial integration that is derived from industry

(12)

8

earnings yields differentials across countries. The expectation is to find lower differentials among EU-member country-pairs than among their non-EU counterparts. Indeed, by employing a sample of 528 distinct country-pairs they find a significantly higher magnitude of financial integration among EU member states during the period from 1990-2007. This yields evidence that EU membership has posi-tively affected equity market integration of member states. However, they could not find a channel, such as trade integration or financial regulation, through which this EU effect is driven. Moreover, a main contribution to equity market integration with respect to the European Monetary Union (EMU) depicts the finding that not the Euro but EU membership itself has fostered financial integration.

The group of indirect measurements, on the other hand, involves the notion of international market completeness as an underlying assumption. In this sense and as outlined in the risk sharing argument above, financial integration is perfect if a country can fully reduce its idiosyncratic risk to a systematic one by diversifying to an international portfolio. This mechanism serves as an insurance to fluctuations in national income or production flows (Obstfeld and Rogoff (1996)). An example of indirect measures involves the study conducted by Bekaert et al. (2003), in which they examine structural breaks in capital flows as a channel of equity market integration.

After having outlined the underlying theory and definition of financial integration, Kearney and Lucey (2004) propose three major techniques to measure financial integration. The first major category of empirical models includes international asset pricing models. The simplest example of these models depicts a world CAPM (WCAPM) model. The assumptions made are similar to those of the traditional capital asset pricing model (CAPM) as introduced by Sharpe (1964) and Lintner (1965), including most importantly and among others capital market efficiency and free lending or borrowing at the risk-free rate. Instead of employing the correlation between a risky asset and a domestic index as the beta, the price of risk is rather determined by a global return factor in the WCAPM. Karolyi and Stulz (2002) point out that the use of this model naturally involves a problem of multicollinearity in that indices from large economies, such as the EU or the US, account for a major part of the movements of a given world index. Moreover, the WCAPM requires an assumption of fully integrated financial markets so that the price of risk is entirely determined by global market risk. Karolyi and Stulz (2002) further provide a

(13)

9

review of related surveys and stress that one of the key weaknesses of a single-factor approach is that the security market line is too flat in an empirical setting and therefore fails to explain risky asset returns. This potentially relates to the fact that asset returns in the cross-section vary in response to other factors than global risk. With respect to the fully integrated markets assumption, the counterpart to this model is an approach that adopts the standard CAPM while assuming a country is fully segmented from the rest of the world. However, Kearney and Lucey (2004) caution that neither of these approaches has yielded convincing results.

Therefore, literature has come up with more sophisticated multi-beta and time-varying international asset pricing models, which have been more successful. For instance, Apergis et al. (2011) extended the single-factor WCAPM with the proven factor foreign exchange risk (De Santis and Gerard (1998)), which they found to be reflected in German stock returns. With respect to intertemporal international asset pricing models, Kerney and Lucey (2004) emphasize the importance of an approach that allows for a dynamic evolution of financial integration. This is in line with Finance literature, which suggests that risk premia in equity markets are generally time-varying rather than constant (Cochrane (2005)). Recently, there has been some effort to account for the time-variant component of the international CAPM, which provided convincing results regarding the pricing error and the slope of the security market line (e.g. Rossi and Timmermann (2015) or Baillie and Cho (2016)). However, concerning the empirical part of this paper, the necessity of allowing for time-varying integration becomes redundant. This is because financial integration is entirely reflected in the convergence of earnings yields, which is in turn only determined by discount rates and capitalized growth opportunities (Bekaert et al. (2011)).

The second main category as proposed by Kearney and Lucey (2004) includes models adopting corre-lations or cointegration approaches. The underlying idea is that assets are more likely priced on a global level what implies securities returns follow commoving trends (Karolyi and Stulz (2002)). Studies in this field usually employ conditional correlations or volatility spillovers. The former refers to the as-sumption that integrated equity markets exhibit a common stochastic trend. Following this logic, more integrated markets indicate a more pronounced common trend or an increased covariance respectively. Strongly related to this concept is the notion of volatility spillovers in integrated markets, which depicts

(14)

10

a contagion risk in the sense that economic shocks can be transmitted more easily through integrated markets (Gatfaoui (2013)). A recent study conducted by Choudhry and Jayasekera (2014) investigated the effects of the financial crisis on volatility spillover effects. They found a significant increase in volatility spillovers in a sample period capturing the period before, during and after the financial crisis. The results are particularly strong for the UK and Germany, which contributes to my expectation of the UK having undergone a long-term period of integration with EU countries.

Despite the bulk of studies revealing a trend towards increased integration over time -as Kearney and Lucey (2004) conclude-, measuring financial integration through cross-country correlations has several drawbacks. They stress that results may be driven by mutual economic shocks that affect cross-country security co-movements that would have otherwise not exhibited a common trend. Furthermore, Karolyi and Stulz (2002) identify the problem of testing two hypotheses jointly. The first one indicates that integrated markets have common co-movements or volatility spillovers and that changes in correlations are reflected in changes of these common factors, respectively. The second hypothesis implies that time-varying correlations are the result of overacting of uninformed investors. What makes this joint hypoth-esis problematic is that there is little consensus about empirically disentangling these two hypotheses when adopting an approach using conditional correlations or spillovers.

2.3. Integration of debt and equity markets

In a review of extant literature, Kearney and Lucey (2004) point out the structure of integration. Capital or financial market integration is just a subset of the highest goal of the EU regulatory bodies, viz. economic integration. While economic integration clearly refers to the EU’s commitment to a single market along with the freedom of movement of goods, services, labor and capital, capital market inte-gration has a somewhat narrower definition. Baele et al. (2004) further divide European financial mar-kets into government bond-, corporate bond-, euro money-, bank credit- and equity marmar-kets. For the sake of relevance with respect to contemporary literature, the following analysis focuses on bonds and equity markets.

(15)

11

The links between government bond markets are usually measured by sovereign yield spreads or the correlations thereof and have been explored in several research articles. For instance, Laopodis (2008) found that there is an increase in linkages among the government Euro bond markets observable in the period after the introduction of the Euro. Employing the conditional correlations approach by using a GARCH model, they unveiled especially strong results for the UK, Germany and the US. While adopt-ing an international CAPM approach, Abad et al. (2010) explored government bond returns as well and found similar results for European Monetary Union (EMU) countries. However, both surveys did not include the period during and after the financial crisis into account, which may have a significant impact on the results presented in those papers. A more recent study conducted by Cipollini and Coakley (2015) did so and explored the effect on debt market integration of the sovereign debt crisis. They found strong evidence that after 2010 sovereign bond markets have experienced a trend towards segmentation again. Christiansen (2014) found comparable results when studying government bond markets: Financial in-tegration is generally stronger for EMU countries and for older EU member states while the financial crisis essentially weakened the magnitude of bond market integration in the EU.

Baele et al. (2004) showed that yield curves converged across all capital markets in the years after the introduction of the Euro, which provides evidence that also equity markets were affected by the EMU in terms of higher financial linkages. One of the aims of the EMU was to decrease fragmentation on the European capital market in order to provide a unified access to capital within the Euro-area. Conse-quently, capital market financing has gained importance and the cost of capital has decreased for Euro-pean companies and governments, which also points in a direction of increased integration (Bris et al. (2009)). Naturally, investment flows and portfolio allocation were also affected by these developments due to decreased market and foreign exchange risk (Bartram and Karolyi (2006)) and the question of accurate portfolio reformation arose, trying to answer how to achieve the greatest benefits from inter-national diversification. Especially the importance of industry relatively to country factors has increased in response to the rise of capital market integration (Eiling et al. (2012)). In addition to the trend towards forming portfolios based on industrial sectors, the EU’s integration in the world market should be con-sidered as well. Kim, Moshirian et al. (2005) assessed the effect of the EMU on stock market integration

(16)

12

with the US and Japan. They found that on a global level the establishment of the EMU along with price stability has contributed to a strengthened integration of EU capital markets. The vast majority of the literature shows similar findings when it comes to the investigation of European equity markets. Pa-nayotis and Anna (2013) for example, analyzed spillover effects on European stock markets and docu-mented strong as well as increasing conditional correlations across EU member states. Whereas the authors excluded the period around the financial crisis, Bekaert et al. (2013) found that even after ac-counting for the financial as well as the sovereign debt crisis, the strong interconnection between EU equity markets survived. To sum, the majority of contemporary literature tend to agree upon the fact that EU capital markets have integrated over time and that this effect prevails even after economic shocks. So far, researchers have focused on EU countries, EMU countries, North-, South or Eastern European countries on an aggregate level. However, in view of the Brexit the role of the UK has become much more important. To gain an understanding about the possible consequences of Britain leaving the EU, it is inevitable to explore how UK have contributed to the integrational effects the EU could achieve among its member states. The last researchers to investigate this field were Antonios et al. (2007). Their study unveiled that UK equity markets exhibit a relatively high level of integration with Europe. Yet, there is more than a decade ever since, which was coined by a series of economic as well as political shocks, in between their sample period and the Brexit vote of June 2016. This gap in information about the UK-EU equity market relation will be closed within the scope of the empirical analysis.

As mentioned above, most researchers make use of measures constructed by the evolution of equity return correlation (e.g. Adjaoute and Danthine (2004)) or volatility spillover (e.g. Fratzscher (2002)) to measure the degree of equity market integration. However, due to the difficulty of accurate measure-ments, the validity of these models is often doubted (Karolyi and Stulz (2002)). That is why this paper will make use of a very recent and simple approach to measure equity market segmentation, which was initially introduced by Bekaert et al. (2011). As contrasted with standard international finance literature, their methodology refrains from using asset pricing models requiring econometric estimations. This allows for avoiding the difficulties of those standard models, which were already discussed above. By adopting this measure, this paper aims at revealing the role of the UK in the EU’s process of equity

(17)

13

market integration. The approach builds on the assumption of yield convergence with industry valuation differentials as a measure of segmentation. Valuation in this sense means earnings-yields, the inverse of the price-earnings ratio, which captures capitalized growth opportunities as well as systematic risk through discount rates. Earnings-yields have a linear relationship to discount rates and growth opportu-nities. These are the two main variables determining a stock’s price according to the Gordon’s growth and the dividend discount model. In this setting, discount rates capture financial market integration and capitalized growth opportunities reflect economic integration. Thus, the assumption that discount rates and growth opportunities are industry rather than country specific leads to the conclusion that differ-ences in industry valuations should converge amongst different countries in an integrated market. Therefore, in a fully integrated market, industry earnings yields should converge to zero. Since this idea forms the basis of the empirical analysis of this paper, the detailed mechanics are further discussed in the following section. For now, it is noteworthy that they found very strong results yielding evidence towards decreasing segmentation in developed countries whereas the levels in emerging markets stayed constant. Due to the success of the measure, they extended the analysis to a EU setting an investigated the equity market integration of EU amongst EU member states over time (Bekaert et al. 2013). Again, they could show very strong results indicating that the EU member states have significantly integrated with each other in the period from 1990-2007. However, the investigation in that field does not unveil how each of the EU countries drive this effect. Since they only report a measure of integration for the EU as whole, the contribution of Germany and most importantly UK remains unknown. Hence, one of the aims of this paper is to fill the gap in UK’s contribution of the EU-effect by isolating the magnitude and direction of UK’s integration with the EU while presenting a benchmark country of comparable size and economic strength, i.e. Germany. Although from historical research we know that there are some indicators that the UK financial markets exhibit relatively higher correlation with their US coun-terpart (Cheng (1998) or Fraser and Oyefeso (2005)) as compared to EU members, the vast amount of literature suggests the EU has been successful in implementing their long-term plan of a unified capital market (Bartram and Wang (2015), Bekaert et al. (2013) or Hardouvelis et al. (2006)). In this regard, my expectation about UK integration is in line with the findings about financial integration within the

(18)

14

EU and I can derive the following hypothesis: The UK has undergone a long-term period of increasing stock market integration with EU member states from 1990-2016.

However, in view of recent events and the gap in stock market integration related literature thereof, this research will not only yield a new insight of the UK’s long-term stock market integration with Europe, but also contributes in evaluating possible impacts of the Brexit in future years by employing a shorter sub-sample period with monthly observations which is split in a period before and after the threat of the Brexit has become imminent. The underlying hypothesis is that the UK’s trend towards stock market integration with EU markets has reversed since the beginning of 2014. As the hypothesis proves to be true, it can be expected that the impact of the Brexit on European stock market integration will be less severe because equity markets have already (at least partially) incorporated the risk of fragmentation. Moreover, after two decades of EU financial integration, the reversal will make the attractiveness of portfolio diversification across EU member states increase.

3. The model

I employ differences in bilateral stock market valuation across industries to measure the degree of UK’s equity market integration with the EU. The underlying rationale has been discussed in the previous section. Building on the mechanism of converging equity earnings-yields in an integrated market, a measure of segmentation can be derived by computing the absolute difference of industry-portfolio yields, weighted by the relative importance of the industry. The weights correspond to the market cap-italization of a given industry for a country-pair. The model adopted for the purposes of the following analysis is essentially equal to the approach suggested by Bekaert et al. (2013). However, instead of measuring the degree of segmentation for all EU countries I focus on the relationship between UK-EU and UK-non-EU countries. This allows to isolate the magnitude of UK equity market integration with the EU relatively to non-EU countries. To formalize the idea of the segmentation measure, the following equation has been derived:

(19)

15

This measure gauges the degree of segmentation by capturing the difference in equity valuations, IV, of the UK and country j for industry k at time t. NUK,j,t denotes the number of industries and MCUK,j,k,t

weighs the differential according to the relative market capitalization of industry k in the UKand country j at time t.

SEG will serve as the dependent variable in the following OLS-regression model:

𝑆𝐸𝐺 , , = 𝛼 + 𝛽 ∗ 𝐸𝑈 , , , + 𝛽 ∗ 𝑋 , , + 𝑐 , + 𝑑 + ɛ , , (2)

The subscripts UK and j denote the two countries which were compared in terms of stock valuation differentials at time t. While the former subscript obviously always indicates the UK, subscript j repre-sents all EU and non-EU countries accordingly. The variable EU is a dummy-variable which is one if the UK is compared to a EU country and zero for a non-EU country. Regarding the results section, this indicator variable shows the coefficient of interest. It measures the average segmentation, as proxied for by the differences in earnings-yields, between UK-country-pairs and UK-non-EU country-pairs. Since I expect segmentation to be lower (or likewise integration to be higher) among UK-EU country-pairs, the corresponding coefficient should be negative.

The controls are chosen in line with the caveats of the model as suggested by Bekaert et al. (2011). 𝑋_{, ,} represents those controls, which may affect valuation differentials for other reasons than being a EU member. These include measure induced -, time-series - as well as time-invariant controls, which will be discussed below. The most important assumption of the model is that all firms within one industry exhibit the same systematic risk. The implication is that industry differentials should converge to zero in integrated markets. This makes the model essentially independent from beta estimations and yields:

𝛽 , = 𝛽 (3)

Assuming that betas or systematic risk are industry- rather than firm specific also implies that leverage induced financial risk is equal across countries. This may not hold for all country pairs and could cause upward bias in the 𝛽 coefficient of equation (2) in that smaller differences in leverage could lead to a

(20)

16

decrease in the segmentation measure between a given country-pair. For example, the progress in tech-nologies and the developments in the financial sector have made easier for companies to achieve a targeted leverage ratio. Therefore, differentials in leverage ratios have narrowed over time. Because debt levels are reflected in earnings-yields through the price, they may also converge in response to narrowing leverage ratios, which ultimately could introduce the bias to the EU indicator variable. In this sense, the model is extended by the difference in absolute leverage between the UK and another country.

Next, as part of measure induced controls differences in earnings growth - and price volatilities are added to the model. The underlying rationale relates to the linear relationship between valuation ratios and discount rates as well as capitalized growth opportunities. The mathematical derivation can be found in Bekaert et al. (2011). The basic intuition is, however, that the declining volatilities of macro-economic variables during the “Great Moderation” (Mullineux et al. (2011)) may have caused volatility differentials to converge. In response, this may affect valuation ratios in a narrowing manner, leading to an upward bias of the 𝛽 estimator since volatilities are reflected in a stock’s valuation. To control for both, discount rates and growth opportunities as incorporated in earnings yields, earnings growth - and price volatility are added to the model, respectively.

Last, the model uses the natural logarithm of the number of firms per country-pair. A high number of firms implies a large economy which is typical for highly developed and industrialized countries. Thus, the segmentation measure might be upward biased because it could be driven by the size of the local capital market, which is proxied for by using the summed number of firms of a given country-pair. Furthermore, a high number of firms also leads to more accuracy in the estimator of interest, which is due to a lower influence of outliers or extreme variability in the dataset. Because the segmentation measure uses absolute values, a higher accuracy leads to lower differentials what may likewise intro-duce upward bias to measure.

The next set of control variables are time-invariant and time-series controls. The former group includes the difference in GDP as of 1990, which proxies for the potential difference in economic development between a country-pair. For instance, through high per capita GDP a country had relatively better

(21)

17

chances become a member of the EU because the convergence criteria of the Treaty of Maastricht (Eu-ropean Union (1992)) set out the minimum requirements for GDP to debt ratio. Therefore, differences in the GDP of 1990 influenced which countries would become members of the EU and thereby may introduce bias to the estimator. Furthermore, it is well documented that home bias plays a major role in investments (see e.g. Balta and Delgado (2009) for a EU or Dongmin and Lilian (2009) for a US setting). The main implication is that investors tend to be biased towards domestic equities when making an investment decision. This is mostly due to potential difficulties stemming from the legal environment or transaction costs. In this respect, the model also includes the distance between a country-pair to address issues regarding investor’s home bias. Last, I added an Eastern European indicator variable, which aims at taking into consideration the fact that Eastern European countries had lower chances to join the EU compared to their Central- and Western European due to their deficits in economic and political environment.

The following set of a time-series controls is included to factor in effects, which are constant in the cross-section but vary over time. In this regard and based on the suggestions of Bekaert et al. (2013), I aggregated the earnings yields of all founder states, which are Belgium, France, Germany, Italy, Lux-emburg and the Netherlands, to derive the variable Founder States Earnings Yield and added the aver-age thereof to the model. To introduce a measure of global integration, I further included the variable Difference in Earnings Yields between Founder States and NASDAQ firms, which allows to understand the integration between Europe, as proxied by founder states, and the US, as proxied by all active com-panies listed with the NASDAQ.

To take additional unobserved factors into account, 𝑐, and 𝑑 extend the regression and denote coun-try-pair and time fixed effects, respectively.

In line with the findings of stock market integration in the EU as discussed in the previous section, the expectation arises that UK exhibits a trend towards increasing integration over a long-term sample pe-riod. However, I also suspect that recent years show a reversal of this trend. On the one hand, this is due to the recovery period after the sovereign debt crisis, which is typically associated with a tendency towards disintegration (Antonios et al. (2007)). Moreover, this assumption is also fueled by the political

(22)

18

environment. When UK’s prime minister David Cameron announced a referendum about an EU mem-bership in case of his re-election in January 2013 (BBC News Services (2013)) and the Brexit promoting UKIP party even won the EU Parliament elections in 2014 (BBC News Services (2014)), efficient cap-ital markets should have incorporated the risk of a Brexit already at this stage. Thus, I expect the mag-nitude of the segmentation measure to have weakened during the past couple of years as compared to a long-term measure capturing the time of 1990-2016.

4. Data

For the long-term sample from 1990-2016, the segmentation measure requires yearly firm-level earn-ings yields for European firms, which were retrieved in USD from Datastream. American as well as global depository receipts, preference shares, exchange traded funds and its derivatives are excluded. I use Datastream’s industry classification benchmark (ICBSN) to assign a sector for each static observa-tion. A list of the 38 industries employed can be found in Appendix 1. Furthermore, I exclude companies whose country of issuance (GGISN) does not match with the country of residence. For instance, com-pany “a” is headquartered in country “A” but decided to issue shares in country “B”, which leads to an exclusion of the company. Due to data unavailability after applying the aforementioned filter criteria, Albania, Bosnia, Bulgaria, Belarus, Kazakhstan, Lithuania, Latvia, Moldova, Montenegro, Macedonia, Serbia, Slovenia, Slovakia and Ukraine drop out of the sample. The following step involves establishing the segmentation measure as specified in equation (1). While Bekaert et al. (2011, 2013) do not mention explicitly the item they use, I decided to employ the annual average of high and low earnings yields (WC09206) because this item takes yearly fluctuations better into account and enables smoother move-ments. After dividing the earnings yields by 100, setting negative values to zero and values larger than one to missing, I deleted observations having an error message when retrieving earnings yields data. This cleaning process left a sample of 77,299 observations for the period of 1990-2016. Next, I averaged the data to an industry level, leaving one observation per industry, per year and per country, yielding a sample of 14,335 observations. I deleted another 54 observations, which did not have data for any in-dustry in any given year (this concerned only Romania and Croatia). Aggregating the observations on the country level brings a total of 658 observations. A list of all countries along with their average

(23)

19

segmentation for the long-term sample period can be found in Appendix A2. Ultimately, I weighed the earnings yield differentials with the relative market capitalization (WC08001). This implies summing the firm-level market capitalization per industry for each country-pair and establishing the weights ac-cording to the total market capitalization across all industries. Having a short glimpse in the results section’s Table two, specification one, unveils that a linear regression of the SEG measure against the UK-EU indicator yields a statistically coefficient of -0.0192, implying that -on average- the UK was less segmented or stronger integrated with the EU than with non-EU countries by an absolute 192 basis points. Of course, this result is not meaningful since no controls are used in this model. However, this result gives a first insight on UK stock market’s strong connection with the EU.

The SEG-measure induced control variables include the difference in absolute leverage between the UK and country j. The model employs the variable Difference in Debt to Capital (abs.) to account for a potential upward bias. The item (WC08221), which was used on this behalf, measures yearly records of the total debt to total capital ratio. Negative values were set to zero. The original paper by Bekaert et al. (2013) relied on data from Bureau van Dijk's OSIRIS database in order establish more precision by using different leverage ratios for industrial and financial firms. However, due the relative small sample (658 observations for the annual analysis) and the lower coverage of financial information for this spe-cific set of firms in the OSIRIS database, I decided to employ Datastream’s leverage item. The trade-off is losing some precision for the sake for more statistical power. The derivation of the Difference in Net Income Growth vol. (abs.) variable involved computing log growth rates for recorded net income (WC01706). Next, I calculated volatilities for 3 year moving window and required at least three non-missing observations. Again, Bekaert et al. (2013) established higher precision to this variable by re-quiring eight consecutive observations with non-missing data. For my relatively small dataset, this would have caused too many observations to drop and hence posed a threat to statistical validity. Price volatility is available for a yearly basis at Datastream (WC08806) and measures the movement to a high and low from a mean price. The data was then divided by 100 and negative values were set to zero. As compared to Bekaert et al. (2013) and their more restrictive criteria for data selection of this variable, I once again employed this specific Datastream item for purposes of greater data coverage. SEG and all

(24)

20

measure induced controls were winsorized at the lowest and highest percentile and -for the sake of consistency- were weighted according to relative market capitalization.

Panel A of Table 2 shows the summary statistics for the long-term sample of 1990-2016 with yearly observations. When the data is aggregated from the firm- to the country level there remain 658 obser-vations. This involves the UK being matched with 26 EU or non-EU countries (see Appendix A2 for a list of countries). The proportion of UK-EU country pairs amounts to 66.9%. The average segmentation conditional on the UK-EU indicator for UK-EU country-pairs is 4.01% whereas UK-non-EU pairs ex-hibit a higher segmentation with 5.93%. The unconditional standard deviation is 4.39%, which shows that there is a relatively high variation in the data when comparing it to the summary statistics as re-ported by Bekaert et al. (2013). However, this is not surprising considering that their unbalanced sample has 5,665 observations, which is due to the cross-matching of all EU and non-EU countries. Despite the difference in sample size, the comparison of the average segmentation conditional on EU membership shows that the UK is by 192 basis points more integrated with the EU than with non-EU countries whereas Bekaert et al. (2013) report a 222 basis points higher integration among all EU member states. This already gives a first insight on the magnitude of UK’s integration with the EU insofar as the UK exhibits comparable integration levels as the EU as a whole.

The examination of possible Brexit-effects requires a shorter sample period and reduced frequency ob-servations. Therefore, I use a sample period from 2011 to 2016, which will naturally serve as a robust-ness test for the result of section 4 in that it unveils if a UK-EU membership effect is also observable in the mid-term and most importantly allows to investigate the stock market reactions from the uncertain times of 2014 onwards. That is why I eventually split the sample in a period from 2011-2013 and 2014-2016, respectively. It was at the beginning of 2014 when major banks such as JPMorgan or Citi (Brinded (2014)) addressed serious concerns about negative consequences of UK leaving the EU. Also, the year of 2014 was marked by a major trend setting event: In May 2014, Nigel Farage and his Brexit promoting UK independence party (UKIP) won the EU Parliament election with an increase of over ten percentage points as compared to the previous election (BBC News Services (2014)). This was a first major

(25)

21

Table 1. Summary Statistics. The Conditional Mean refers to whether a country-pair consists of a UK and EU member or not.

Conditional Mean VARIABLES

Panel A, yearly from 1990-2016 N Mean SD Min Max UK-EU

UK-non-EU

UK-EU indicator 658 0.6690 0.4710 0.0000 1.0000 1.0000 0.0000

SEG 658 0.0464 0.0439 0.0002 0.7590 0.0401 0.0593

Number of Firms 658 3.5780 0.4290 2.1500 4.4890 3.5950 3.5450

Difference in debt to capital (abs.) 658 0.1760 0.0935 0.0000 0.7600 0.1810 0.1640

Difference in net income growth vol. (abs.) 658 0.0102 0.0143 0.0000 0.0987 0.0111 0.0082

Difference in price vol. (abs.) 658 0.0830 0.0519 0.0000 0.3540 0.0825 0.0841

Distance between country-pair (airline, in 1,000 km) 658 1.3720 0.7230 0.3220 3.2220 1.2410 1.6370

Eastern European indicator 658 0.1630 0.3690 0.0000 1.0000 0.1160 0.2570

Difference in GDP as of 1990 (abs., in 1,000 USD) 658 10.220 6.8750 1.6150 30.360 7.6700 15.370

Difference in earnings yields between founder states and NASDAQ 658 0.0497 0.0109 0.0334 0.0763 0.0493 0.0505

Founder states earnings yield 658 0.0619 0.0010 0.0464 0.0836 0.0624 0.0608

Conditional Mean VARIABLES

Panel B, monthly from 2011-2016 N Mean SD Min Max UK-EU

UK-non-EU

UK-EU indicator 1,930 0.7930 0.4050 0.0000 1.0000 1.0000 0.0000

SEG 1,930 0.0447 0.03990 0.0000 0.523 0.0433 0.0504

Number of Firms 1,930 3.5440 0.4460 2.0790 4.408 3.5230 3.6350

Difference in debt to capital (abs.) 1,930 0.1110 0.0778 0.0000 0.508 0.1130 0.1040

Difference in net income growth vol. (abs.) 1,930 0.0093 0.0117 0.0000 0.152 0.0094 0.0088

Difference in price vol. (abs.) 1,930 0.3540 0.5680 0.0000 5.407 0.3310 0.4440

Distance between country-pair (airline, in 1,000 km) 1,930 1.4270 0.7150 0.3220 3.2220 1.3170 1.6130

Eastern European indicator 1,930 0.2410 0.4280 0.0000 1.0000 0.2610 0.1670

Difference in GDP as of 1990 (abs., in 1,000 USD) 1,930 10.490 6.8680 1.5820 30.36 9.0090 16.190

Difference in earnings yields between founder states and NASDAQ 1,930 0.0369 0.0531 0.0000 0.654 0.0692 0.0752

(26)

22

indicator for UK’s skeptical attitude towards the EU membership. In this sense, I expect the magnitude and/or the statistical significance of the EU indicator in model (2) to decrease.

When adopting a shorter sample period, it is inevitable to use monthly observations on this behalf in order to get meaningful statistical results. A major problem that occurs naturally is data availability. In particular, measure-induced controls are reported on a quarterly basis but are required as monthly ob-servations for an OLS regression analysis. Researchers in economics and finance have tackled this problem by using the cubic spline interpolation methodology. For instance, Luo, Cheng et al. (2015) adopted this approach to reduce quarterly to monthly observations on the firm-level. The advantage of this methodology as compared to the use of Datastream’s pre-calculated industry indices as suggested by Bekaert et al. (2013) is that it overcomes the problem of firm identification. Datastream does not provide a time-series composition of the reported indices. That is why I cannot determine the exact number of firms per country-pair at a given point in time. Thus, the potentially important control vari-able Number of Firms per country pair would become meaningless. Furthermore, avoiding Datastream’s pre-calculated indices and relying und the cubic spline approach allows for a better comparability as well as consistency in relation to the analysis of the long-term sample. The drawback of using cubic spline interpolation to fill missing data is that it violates the standard assumption OLS linear regressions by implementing a potential source of autocorrelation. I deal with this problem in two steps: First, after retrieving quarterly data for Difference in debt to capital (abs.) and Difference in net income growth vol. (abs.) by using the same items as for the long-term sample, I calculated for both variables the Bayes information criterion (BIC) as well as the Akaike information criterion (AIC). These two criteria will form the basis of choosing the optimal lag-length for Newey-West standard errors, which are not only robust against heteroskedacity but also against autocorrelation, which provides a solution to the problem of the cubic spine interpolation as outlined above.

Regarding the segmentation measure as defined in equation (1), I employed Datastream’s twelve-month trailing price earnings ratios (PE) and took the invert of this measure to derive the earnings yields. By the same token, I used a twelve-month trailing market capitalization (MC) data to weigh the earnings yield differentials. The observations for the control variable Difference in price vol. (abs.) are available

(27)

23

through a trailing 12-month basis on the monthly level. Again, the segmentation measure as well as its induced controls were winsorized at the 1st_{and the 99}th_{percentile and were weighted by their relative}

market capitalization.

5. Results

To gain a first understanding of how industry valuation differentials between the UK and the EU mem-bers as well as non-memmem-bers have evolved over time, Fig. 1 shows the average segmentation from 1990-2016. While UK-EU differentials are moving in a 2.5% to 4.5% corridor, the UK-non-EU country-pairs exhibit a far more volatile range of 3.5% to 7.0%. More importantly, during the whole sample period, average UK-EU integration is below UK-non-EU integration, which serves an indicator of UK’s strong connection to Europe in terms of capital market integration. This may be the result to the EU’s strong efforts to establish a single market across all member states, which includes the facilitation of the

Figure 1 shows the average difference in industry valuations between UK-EU and UK-non-EU country-pairs in the period of 1990-2016

(28)

24

movement of goods, capital, services and labor. It shall be noted, however, that the interpretation of Fig.1 does not imply a causality between EU membership and UK’s lower segmentation or higher in-tegration with member states. This could likewise be the result of closer links in terms of economic development, a general trend of stronger global integration between more developed countries or even geographical proximity, to name a few. Therefore, the following regression analysis aims at examining economical as well as statistical importance of the EU-effect for UK equity markets while accounting for unobserved factors, which may have influenced the magnitude of stock market integration. Table 2 exhibits the results for the segmentation measure and its controls while employing yearly observations during 1990-2016. All the following analyses have the focus on the UK-EU coefficient. This indicator variable measures the average differences in earnings-yields between EU country-pairs and UK-non-EU country-pairs. A negative sign implies that average segmentation was lower (or likewise aver-age integration was higher) among UK-EU country-pairs and is hereinafter referred to as an EU-effect. For example, a univariate specification, which involves a linear regression of SEG on the UK-EU-indi-cator variable, shows that UK stock markets were on average and in absolute terms by 194 basis points stronger integrated with EU equity market than with non-EU European markets. The second specifica-tion adds measure induced control variables. Not surprisingly, the EU-effect weakens as compared to the absolute 194 basis points from specification one but the economic magnitude of this absolute de-crease is low with 13 basis points. None of the measure induced controls are statistically significant and consequently cannot explain any of the variation in the valuation differentials. The third regression specification further includes time-invariant controls. Again, the EU-indicator decreases (by 38 basis points) in absolute terms while it stays significant at a 99% level of confidence. The distance between two countries as well as the Eastern European indicator variable have the expected sign and are eco-nomically as well as statistically important. An additional 1,000 km to London can be related to an increase of 150 basis points in valuation differentials. The fact if a country is Eastern European or not is associated with a 218 basis points higher SEG. These findings are in line with my expectation and contemporary literature. For instance, Balta and Delgado (2009) show that EU policies have not been successful in diminishing the home bias issue within the common market. Furthermore, Syllignakis and Kouretas (2010) show that some Eastern European countries have financially connected with EU capital

(29)

25

Table 2. reports the results of linear ordinary least squares regression (OLS) regression models (1-3) and fixed effect models (4-6) with the segmentation measure (SEG) as the dependent variable and the UK-EU indicator as the independent variable of interest. There are observations for the sample period from 1990 to 2007. Standard errors are reported in parentheses and are robust to heteroskedacity.

UK-EU integration from 1990-2016

Specification 1 2 3 4 5 6 7

Rel. MCap weighted SEG

UK-EU indicator -0.0194*** -0.0181*** -0.0143*** -0.0139*** -0.0167*** -0.0211*** -0.0145**

(0.0048) (0.0040) (0.0040) (0.0038) (0.0036) (0.0061) (0.0067)

Number of Firms 0.0007 -0.0099 -0.0094 0.0007 -0.0174 -0.0270

(0.0080) (0.0101) (0.0101) (0.0080) (0.0148) (0.0216)

Difference in debt to capital (abs.) -0.0483 -0.0714 -0.0712* -0.0396 -0.0808*** -0.0714***

(0.0408) (0.0439) (0.0428) (0.0378) (0.0186) (0.0191)

Difference in net income growth vol. (abs.) -0.0343 0.0401 0.0360 0.1590 0.0977*** 0.1020***

(0.0851) (0.0847) (0.0904) (0.1190) (0.0343) (0.0364)

Difference in price vol. (abs.) 0.1530 0.1610 0.1640 -0.0167*** -0.0211*** -0.0145**

(0.1180) (0.1130) (0.1140) (0.0036) (0.0061) (0.0067)

Difference in GDP as of 1990 (abs.) -0.0008** -0.0008**

(0.0003) (0.0003)

Distance between country-pair (airline) 0.0150*** 0.0149***

(0.0039) (0.0039)

Eastern European indicator 0.0218*** 0.0219***

(0.0051) (0.0051)

Difference in earnings yields between founder states and

NASDAQ 0.3180*

(0.1620)

Founder states earnings yield 0.0385

(0.1860)

Constant 0.0593*** 0.0523* 0.0741** 0.0535

(0.0047) (0.0286) (0.0356) (0.0392)

Observations 658 658 658 658 658 658 658

R-squared 0.043 0.079 0.157 0.163 0.582 0.658 0.674

Time Fixed Effects No No No No Yes No Yes

Entity Fixed Effects No No No No No Yes Yes

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(30)

26

markets after their accession, but by far not all of them, such as the Baltic states, Romania or Bulgaria. The fourth specification further adds time-variant controls. The UK-EU coefficient becomes slightly weaker from an economic perspective, remains significant at a 99% confidence level, though. The dif-ference in earnings yields between EU founder states and NASDAQ firms shall reflect global trend of integration. Although the coefficient exhibits only statistical significance at a 90% confidence level, it shows that a 1% increase in global integration was associated with a 0.32% increase in valuation dif-ferentials between UK-EU and UK-non-EU country-pairs. Specifications 5-7 include time-fixed effects, country-pair fixed effects and both, respectively. Naturally, a fixed effect model excludes time-varying and time-invariant factors as well as a constant due to multicollinearity. Moreover, the model serves as robustness check in that it takes additional unobserved factors into account, which may have influence on industry valuation differentials. All fixed effect regressions show a negative and significant UK-EU coefficient, which strengthens the hypothesis that UK has undergone a long-term period of integration with EU-member states’ equity markets. The magnitude of the EU-effect of specification 7 is compa-rable to the one of specification 4, exhibiting an absolute difference of only 6 basis points. This indicates that the model works properly and has been applied consistently. A comparison of the results of Bekaert et al. (2013), who tested an equal model while investigating an overall measure of EU-integration, un-veils that the UK shows no surprises or irregularities in terms of stock market integration with the EU. Throughout the sample period, the coefficient of interest is between roughly one and two percent across all specifications, which is the same corridor found by the mentioned authors. This leads to the conclu-sion that the UK was subject to similar mechanism being responsible for a general trend towards equity market integration in the EU. Also, the results from Table 2 support the hypothesis that the UK has undergone a long-term period of stock market integration with the EU.

Isolating a UK-EU effect provides some understanding about the integrational properties of UK stock markets, but the question remains how these results compare to other member states. Factoring in the size of capital markets and economic importance, the most logical benchmark is to be located in Ger-many. Germany is not only one of the founders of the EU, it has also been historically a strong promoter of the European single market and political decisions have been made rather from a European perspec-tive rather than a German one (Lemke (2010)).

(31)

27

Table 3 reports the results of the German stock market integration with the EU. For the sake of con-sistency, the methodology adopted as well as the constitution of the sample essentially remained equal within the scope this analysis. The univariate specification shows that German equity markets were stronger integrated with EU stock markets by an absolute value of 205 basis points as compared to non-EU stock markets. This represents a slightly stronger equity market connection than for the equivalent UK-EU indicator of Table 2. Again, specification two adds the measure-induced control variables, which lets the DE-EU indicator decrease by 8 basis points in absolute terms. The OLS regressions in specification three and four further include time-variant and invariant control variables. Both, the dis-tance between Berlin and country j as well as the Eastern European indicator, are positive and significant and are in a comparable range as compared to the results of Table 2. What is noteworthy about specifi-cation four is that the DE-EU coefficient was almost divided by half in comparison to specifispecifi-cation one. This eventually yields a lower magnitude of integration for DE-EU country-pairs than for UK-EU coun-try-pairs by an absolute 134 basis points, which is surprising considering Germany is one of the driving forces behind the promotion of the EU and its efforts to harmonize capital markets as opposed to the UK and its somewhat more complicated relationship to EU and its institutions (Oliver, 2015).

Taken together though, the results for German stock market integration with the EU reveal a similar picture when looking at the range of the EU dummy coefficient across the OLS specifications: As in Table 2 or in the results provided by Bekaert et al. (2013), the EU-indicator is negative, significant and decreases from roughly two to one percentage-point when adding the three sets of controls. Unexpect-edly, however, the EU-effect does not survive the time-fixed - and entity-fixed effect specification in the German setting as opposed to the strong results from a UK perspective. In particular, one can ob-serve that the DE-EU indicator loses explanatory power when including country-pair fixed effects in specification 6, with the coefficient decreasing from 184 to 112 basis points in absolute terms and the confidence interval declining from a 99% to a 90% level. This implies that there may be some unob-served factors, that vary across country-pairs but are fixed over time, causing bias to the standard OLS estimator of specification 4. One can possibly think of cultural differences across countries which may eventually affect the decision of becoming an EU member or not through elected political

(32)

28

Table 3. reports the results of linear ordinary least squares regression (OLS) regression models (1-3) and fixed effect models (4-6) with the segmentation measure (SEG) as the dependent variable and a DE-EU indicator as the independent variable of interest. There are yearly observations for the sample period from 1990 to 2007. Standard errors are reported in parentheses and are robust to heteroskedacity.

DE-EU integration from 1990-2016

Specification 1 2 3 4 5 6 7 SEG DE-EU indicator -0.0205*** -0.0197*** -0.0110** -0.0105** -0.0184*** -0.0112* 0.0012 (0.0050) (0.0053) (0.0048) (0.00474) (0.0040) (0.0067) (0.0077) Number of Firms 0.0010 0.0112** 0.0118** 0.0020 0.0041 0.0374 (0.0086) (0.0050) (0.00501) (0.0048) (0.0222) (0.0255)

Difference in debt to capital (abs.) 0.0513** 0.0377** 0.0379** 0.0570*** -0.0020 -0.0104

(0.0253) (0.0165) (0.0165) (0.0162) (0.0181) (0.0184)

Difference in net income growth vol. (abs.) -0.0346 0.0302 0.0324 -0.0003 -0.0007** -0.0007**

(0.1010) (0.1140) (0.115) (0.0003) (0.0003) (0.0003)

Difference in price vol. (abs.) -0.0184 -0.0206 -0.0185

(0.0333) (0.0301) (0.0301)

Difference in GDP as of 1990 (abs.) 0.0097 0.0120

(0.0352) (0.0353)

Distance between country-pair (airline) 0.0175*** 0.0176***

(0.0026) (0.0026)

Eastern European indicator 0.0255*** 0.0257***

(0.0053) (0.00531)

Difference in earnings yields between founder states and NASDAQ 0.262

(0.163)

Founder states earnings yield 0.0146

(0.179)

Constant 0.0610*** 0.0508* -0.0031 -0.0194

(0.0048) (0.0282) (0.0168) (0.0212)

Observations 659 659 659 659 659 659 659

R-squared 0.041 0.057 0.129 0.132 0.551 0.637 0.666

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Time Fixed Effects No No No No Yes No Yes

(33)

29

representatives. When comparing the statistical relevance to the findings of Bekaert et al. (2013), the reader shall bear in mind that the authors explored cross-country integration to capture an effect for the whole EU. This paper, by contrast, isolates the effect for UK or Germany and hence has an almost ten-fold smaller sample.

In the light of the tumultuous past decade, the investigation of the period after the sovereign debt crisis is of particular interest. Literature has shown that economic shocks have the potential of slowing down the process of integration in the EU (e.g. Kučerová (2013) or Bartram and Wang (2015)). Figure 2 shows the subperiod from 2011-2016 with monthly observations in order to take into account the time-varying feature of financial integration (Baele and Inghelbrecht (2010)) on the one hand, and to capture the effects before and after capital markets realized that the Brexit is a realistic threat. Up until the end of 2014, we can see a similar picture compared to the long-term analysis: The degree of UK’s equity market integration with the EU is relatively constant with one to two percentage points lower valuation differentials as compared to non-EU countries in the period from 2011-2014. However, the two lines in

Figure 2. depicts weighted monthly industry valuation differentials for UK and EU as well as UK and non-EU country-pairs from 2011-2016

UK’s stock market integration with the European Union