Equity market (des-)integration after the Eurozone crisis
A style regression analysis approach
Master Thesis July 2018
Institution: University of Amsterdam – Amsterdam Business School
Programme: MSc Finance
Track: Quantitative Finance
Name: Koen van Leeuwen
Student number: 10548793
Abstract
In the last decade, Europe has witnessed an important sequence of events that might have affected the level of equity market integration. By using data from 2003-2017 and employing a style regression analysis, it is examined if the level of equity market integration in the EU has decreased after both the financial global crisis and the sovereign debt crisis. The results show a significant increase in the relative importance of country effects. Therefore, the level of equity market integration has decreased after the two crises. Furthermore, the dynamics of the process of equity market integration show that the largest decrease occurred in the transition from the ante-crisis period to the crisis period and that the level of equity market integration has been varying continuously.
Keywords: equity market integration, EU, EMU, crisis, style analysis, country vs. industry effects
Statement of Originality:
This document is written by Koen van Leeuwen, 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 mentioned in the text and its references have been used creating it. The UvA is responsible solely for the supervision of completion of work, not for contents.
Content
1. Introduction ... 1
2. Literature review ... 3
2.1 Measuring equity market integration ... 3
2.2 Europe ... 4 2.3 Crisis ... 5 3. Methodology ... 7 3.1 Methodology ... 7 3.2 Hypotheses ... 10 4. Data ... 12 4.1 Data collection ... 11 4.2 Descriptive statistics ... 11 5. Results... 15
5.1 Full sample regression ... 15
5.2 Average specific variances ... 16
5.3.1 Variance ratios ... 18
5.3.2 Dynamics ... 19
5.4 Robustness checks ... 20
6. Conclusion ... 23
7. Bibliography ... 25
8. Appendix 1 – Conditional volatility ... 29
9. Appendix 2 – National industry sectors ... 30
10. Appendix 3a – Style regression country ... 32
11. Appendix 3b – Style regression industry ... 33
12. Appendix 4 – Average weights ... 34
13. Appendix 5 – Code for standard errors ... 35
1 1. Introduction
Equity market integration plays a critical role in the functioning of markets. The level of equity market integration has implications for the benefits arising from international diversification strategies (Goetzmann et al., 2005), for the cost of capital (Hardouveli et al., 2007), for the robustness of national equity markets (Kearny & Lucey, 2004), for growth opportunities across industries (Bekaert et al., 2013) and for the level of economic development (Bekaert et al., 2001, Colacito & Croce, 2010 and Lee & Hsiesh, 2014). For these reasons, equity market integration has been studied extensively. More specifically, the level of equity market integration in Europe after the foundation of the European Union (hereafter: EU) and of the European Monetary Union (hereafter: EMU) has been subject of study. In the academic literature it is acknowledged that the process of equity market integration is a dynamic and time-varying process (Bekaert & Harvey, 1995). Over the last decade, Europe has witnessed an important sequence of events that might have affected the level of equity market integration. In September 2008, the Lehman Brothers filed for bankruptcy, which marks the start of the global financial crisis (FT reporters, 2008). In Europe, the global financial crisis was succeeded by the sovereign debt crisis, which specifically affected the so-called GIIPS1 countries (Bekaert et al. 2013). These crises showed that
reconsidering leaving the Euro is a real possibility (Bartram & Wang, 2015). In the aftermath of these crises, right wing populists advocating for leaving the European Union have gained popularity (Stavrakakis et al., 2017). Ultimately, the majority of the UK electorate voted for leaving the EU in a referendum held in June 2016 (Walker, 2017).
Given the fact that the process of equity market integration is dynamic, these events might have affected the level of equity market integration in Europe. However, the effect of these events has not been examined yet. This is remarkable, since it is known that major economic events can change the fundamentals of the relationships between markets (Huyghebeart & Wang, 2010). As far as known, datasets used in recent empirical analysis
include observations until 2011 (Chen et al., 2014), therefore the period from 2011-2017 remains unexamined. Studying this period is, in the light of the discussed events, relevant because these events are indicators of the vulnerability of the EMU. Therefore, the research question of this master thesis is the following: “Has the level of European equity market integration decreased in the post-crisis period?” Furthermore, the dynamics of equity market integration are examined by
2 studying the level of equity market integration in ante-crisis period, in the crisis period and in the post-crisis period.
The research question is answered by employing style regressions as developed by Sharpe (1992) and used by Eiling et al. (2012) for examining the relative importance of country effects versus industry effects. A dataset containing returns from November 2003 – December 2017 is used, thereby covering the up till now unexamined post-crisis period. By comparing the level of equity market integration in the ante-crisis period with its level in the post-crisis period, it is examined if the level of this dynamic integration process has significantly been affected by both crises. Furthermore, it is assessed if the observed increase in equity market integration after the advent of the Euro is a sustainable long-term relationship or if this relationship has been affected or even reverted by countries preventing the European banking system to collapse and trying to enhance their individual creditworthiness. An in-depth analysis of the dynamics of equity market integration is provided by testing the difference between the level of equity market integration in the ante-crisis period and in the crisis period and by testing the difference between the level of equity market integration in the crisis period and in the post-crisis period. By
focusing on the relative importance of country effects in the ante-crisis period, the crisis period and the post-crisis period, another chapter to the ongoing discussion on this relative importance is added.
In this thesis, a significant increase in the relative importance of country effects in the post-crisis period relative to the ante-crisis period is reported. Therefore, the level of equity market integration in the EU is lower after the crisis than before. The process of equity market integration, which was increasing after the advent of the Euro has been reverted by both the global financial crisis and the sovereign debt crisis. The largest decrease in the level of equity market integration is found in the transition from the ante-crisis period to the crisis period. The decrease from the crisis period to the ante-crisis period is much smaller, but significant.
Therefore, the dynamics of the results show that during the period 2003-2017, the level of equity market integration within the EU has continuously been varying.
The decrease in the level of equity market integration within the EU implies the following: an increase in the diversification benefits within the EU, an increase in the cost of capital and a decrease in the robustness of national equity markets. Furthermore, the decrease in equity market integration is an indicator of the vulnerability of the EMU and of the sentiment of
3 individual countries towards a unified Europe.
The remainder of this thesis is structured as follows: Section 2 elaborates on the literature review, Section 3 discusses the methodology, Section 4 provides the descriptive statistics and Section 5 discusses the results. In Section 6, a conclusion is drawn and recommendations for further researches are done.
2. Literature review
To gain a better understanding of equity market integration, several measures including the one used in this thesis, are discussed. An analysis of past empirical findings on equity market
integration within Europe is provided as well. Specifically emphasizing the empirical findings of a few authors regarding equity market integration in crises highlights the importance of
examining the process of equity market integration in the EU during and after the crises.
2.1 Measuring equity market integration
Equity market integration is a dynamic, time-varying process and therefore is challenging to measure. For this reason, it is not remarkable that in the vast academic literature on equity market integration numerous measures and methodologies have been introduced. Griffin (2002) employs alternative specifications of the Fama and French model to identify a global factor driving national equity returns. Other authors focus on changes in correlations between national equity returns to determine the level of equity market integration (Panton et al., 1976). A more sophisticated approach than examining cross-country correlations is the employment of a
cointegration model, which provides an intuitive measure for equity market integration (Chan et al., 1997). Bekaert et al. (2011) introduce a model-free measure of market integration.
Volosovych (2011) employs a Principal Component Analysis (PCA) to measure time-varying equity market integration. Bekaert et al. (2013) use industry valuation differentials to assess equity market integration. Bartram & Wang (2015) asses equity market integration by employing a conditional copula dependence model.
In this thesis, equity market integration is defined as the relative importance of country effects to industry effects, as utilized by Eiling et al. (2012). The method of measuring this relative importance is elaborated upon in Section 3. By focusing on both country and industry returns, one is able to gain proper understanding of the factors that might drive Eurozone equity
4 returns, whereas the majority of the mentioned alternative measures of equity market integration measure and quantify the process, but do not necessarily directly identify its drivers.
The academic debate on the relative importance of country effects to industry effects has currently been ongoing for more than 40 years. In his seminal paper, Lessard (1974) focused on both industry and country returns. By regressing a global factor and a national index factor on individual firm returns, he found that industry effects are less important than country effects. Heston & Rouwenhorst (1994) confirmed these findings by employing a multi-factor model with country and industry dummies. Adjouaté & Danthine (2004) extend the approach of
Rouwenhorst (1994) by adding a more flexible test to assess the relative importance of country and regional effects. Their results suggest a reversion of the relative importance: industry effects dominate. Eiling et. al (2012) introduce the concept of style analysis to assess the relative
importance of country returns and also identify a reversion in the relative importance. This paper adds another chapter to the discussion on this relative importance by shedding light on the effects of both the global financial crisis and the sovereign debt crisis on the process of equity market integration.
2.2 Europe
In the last decennia, Europe has been subject to financial, economic and monetary integration. Therefore, equity market integration within Europe and the effect of the advent of the euro and the formation of the EMU has been examined extensively. Adjouaté & Danthine (2004) report an increase of correlations of returns within the EU. Furthermore, they contribute to the debate on the relative importance of country effects to industry effects by initially reporting a superiority of country effects, which has reverted in 1999. However, they conclude that this reversal itself is time-varying, since their latest observations in 2004 indicate an increase in the importance of country effects. Baele (2005) focusses on equity market integration in Western Europe and finds an increase in the importance of common factors to explain country returns. Bekaert et al. (2013) report similar results and discuss both the effect of membership of the EMU and the EU on the level of equity market integration. They find that EU membership itself has increased equity market integration, but that membership of the EMU has not, since irrespective of having implemented the euro or not, there seems to be an increase in the level of equity market
5 increasing importance of industry factors but disagree by reporting a decrease in the cost of capital for countries that have joined the EMU. This decrease is insignificant for countries that have joined the EU. Kim & Moshirian (2005) also state that the formation of the EMU has been the driving force for the increasing process of equity market integration within Europe. The findings of Fratscher (2002) confirm this by stating that particularly for countries that have adopted the euro, the European equity markets have replaced the US equity market as the main component for explaining equity returns. Aggarwal et al. (2010) focus on the effect of political events on the integration of equity markets in Europe and observe increasing interdependencies between markets, but state that political events have a negligible effect on the direction of the process of integration. Eiling et al. (2013) report a reversion of the relative importance of country effects to industry effects after 1999.
The preponderance of the academic literature reports an increasing level of equity market integration within Europe since the start of this age. The main conclusion is that the unique process of financial, economic and monetary integration within Europe has significantly
increased equity market integration. However, the long-term effects of the global financial crisis, the sovereign debt crisis and its aftermath has not been examined in these studies. Bekaert et al. (2013) do extend their dataset to July 2012 to check the robustness of their results to the effects of the global financial crisis, but do not examine the long-term effect of the crisis itself.
2.3 Crisis
This thesis contributes to the existing literature by focusing on the impact of the recent crises on the level of equity market integration in the EU. The effect of crises and, the global financial crisis in particular, has been examined on both a global and regional scale. Huyghebaert & Wang (2010) state that major economics events can affect the relationship between stock markets. Both the global financial crisis and the sovereign debt crisis that hit Europe are major economic events. Chen et al. (2014) find empirical results that support the assertion of Huyghebaert & Wang (2010) by concluding that the global financial crisis has influenced the determinants of international equity market integration. With respect to crises in general, Forbes & Rigobon (2002) state that there is a high level of co-movement between equity markets during crises. They report empirical findings for the 1997-1998 stock crash in Asia, the 1994 Mexican
6 highest volatility, which are commonly crisis periods, display a higher level of integration. In bear markets, the integration is higher. However, in such crisis periods, the benefits of
diversification are most needed.
The effects of the financial crisis on a global scale have been subject of study. Zhang et al. (2013) find that the global financial crisis has caused a change in the relationship between developed stock markets and the BRICS2 markets. They acknowledge themselves that the crisis is recent history, such that the results are not reporting a long-term effect. However, they expect that their findings will hold on the long run. Lekhonen (2015) specifically studies the period 2007-2009. He finds that equity market integration has increased for emerging markets during the crisis, but actually decreased for developed markets. Furthermore, he states that the findings regarding the effects of the global financial crisis depend on the definition of the crisis, whereas he finds that the liquidity crisis of 2007 actually increased and that the crisis, starting after the bankruptcy of the Lehman Brothers, decreased equity market integration. Bekaert et al. (2011) show that integration in the US has decreased towards the end of 2008 but increased to its ante-crisis level in 2009.
The Asian stock crisis of 1997-1998 has been examined extensively. In et al. (2001) find a higher level of equity market integration during the crisis, since domestic markets tend to react to news in foreign markets as well. Click & Plummer (2005) focus on a post-crisis period of two years, where they find that the ASEAN-5 stock markets3 are cointegrated after the crisis.
However, they notice that these markets are not fully integrated. Yang et al. (2003) report similar findings for a broader set of Asian countries by reporting both short-term and long-term
cointegration relationships between stock markets after the stock market crisis.
As mentioned before, Bekaert et al. (2013) extend their dataset to July 2012 and find that their conclusion, an increase in equity market integration in Europe, still holds. Gebka &
Karoglou (2013) include observations up until 2010 and remark that the financial crisis has a positive effect on the process of equity market integration within Europe. Furthermore, they predict that this effect will hold on the long-run. Bartram & Wang (2015) report increasing market dependence after the bankruptcy of the Lehman Brothers, but find lower dependence during the sovereign debt crisis. They suggest that country effects might increase in importance
2 Brazil, Russia, India, China and South-Africa
7 again.
For the different crises, various results have been reported. Therefore, making general statements about the effect of crises on equity market integration is undesirable. Lekhonen (2015) points this out by stating that no crisis is the same and therefore should be examined individually. By focusing on Europe, a region that has been integrating over the last twenty years and has been affected by two crises and other shocks, the long-term effects of these events are assessed. For Europe, and specifically the EMU, the academic literature has not yet elaborated upon this matter. This thesis therefore adds value by comparing the level of equity market integration in the post-crisis period with its level in the ante-crisis period. By focusing on long-term effects, rather than on short-long-term effects (as done by Lekhonen, 2015, Bekaert et al., 2013, Gebka & Karoglou, 2013 and Bartram & Wang, 2015) conclusions regarding the current level of equity market integration can be drawn. Over more, it can be assessed if the observed increase in equity market integration after the advent of the euro is a sustainable long-term relationship or if this relationship has been affected by countries preventing the European banking system to collapse and trying to enhance their individual creditworthiness.
3.1 Methodology
Sharpe (1992) introduced the concept of style analysis, a technique that aims to minimize the variance of the difference between a replicating portfolio and a certain return. Eiling et al. (2012) utilized this style analysis approach to measure equity market integration by comparing country and industry effects. Following this approach, style analysis is used to measure equity market integration and to examine if the process of equity market integration has decreased in the post-crisis period. The methodology consists of seven steps.
The first step is calculating the returns. Returns on national industry indices are obtained by the following calculation:
𝑟𝑘𝑖𝑛𝑑 = 100 ∗ (ln(𝑅𝐼𝑡) − ln(𝑅𝐼𝑡−1)), (1)
where RI is the mnemonic for Total Return Index.
The second step is constructing the replicating portfolios. By adding up national industry returns within a country, considering their respective weight in terms of market value, country return-based portfolios are constructed. This country return replicating portfolio is constructed by the following calculations:
8 𝑟𝑙,𝑡𝑐𝑡𝑟= 𝑤1,𝑡∗ 𝑟1,𝑙,𝑡𝑖𝑛𝑑+ 𝑤2,𝑡∗ 𝑟2,𝑙,𝑡𝑖𝑛𝑑+ ⋯ + 𝑤𝑘,𝑡∗ 𝑟𝑘,𝑙,𝑡𝑖𝑛𝑑, (2) where 𝑤𝑖,𝑡 =
𝑀𝑎𝑟𝑘𝑒𝑡𝑉𝑎𝑙𝑢𝑒(𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑘),𝑡
∑𝐾𝑘=1MarketValue,𝑡 (3)
Analogously, the industry return-based portfolios are constructed by adding up each national industry return across countries, considering their respective weight in terms of market value. Therefore, the constituents of one industry return-based portfolio (e.g. industry A) are the value-weighted returns of industry A across all nine countries.
The third step is fitting a Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model to the data. Using this GARCH model, a control for time-varying return volatility is implemented (Engle, 1982). The sample includes two major crises, therefore controlling for this volatility is necessary. By dividing the returns by the estimated conditional volatility (𝜎𝑡) the standardized returns are obtained. 𝜎𝑡 is obtained by fitting the GJR-GARCH model to the value-weighted portfolio of country returns, such that the sample-wide volatility is incorporated. Appendix 1 provides a plot of the conditional volatility estimated by the GJR-GARCH model.
The fourth step is employing the style analysis regressions, which can be expressed in two different empirical specifications. According to Sharpe (1992) the style regression is subject to two restrictions: short positions are not allowed (i.e., β ≥ 0) and coefficients need to sum up to one. This last restriction implies that the portfolio is the best replicating portfolio for the given return.
The regression equations are the following:
𝑟𝑙,𝑡𝑐𝑡𝑟= 𝛼𝑙+ ∑𝐿𝑙=1β𝑙,𝑘𝑟𝑘,𝑡𝑖𝑛𝑑+ 𝜀𝑙,𝑡𝑐𝑡𝑟 (4)
𝑟𝑘,𝑡𝑖𝑛𝑑 = 𝛼𝑘+ ∑ β𝑘,𝑙𝑟𝑙,𝑡𝑐𝑡𝑟 𝐾
𝑘=1 + 𝜀𝑘,𝑡
𝑖𝑛𝑑, (5)
where 𝑟𝑙,𝑡𝑐𝑡𝑟 denotes the return of the replicating portfolio of country 𝑙 and 𝑟𝑘,𝑡𝑖𝑛𝑑 denotes the return of the replicating portfolio of industry 𝑘. The simple regressions as denoted in equations (4) and (5) might contain endogeneity problems, since the country return on which the industry returns are regressed (4) is constructed by national industry returns, which are also constituents for the industry returns as used in equation (4). Therefore, a total of 190 extra replicating portfolios are constructed, which are labelled as the ’filtered’ portfolios. For example, when employing the style regression for Germany, the independent variables are the ten Eurozone wide industry
9 returns as constructed in the second step. However, when employing the filtered regression for Germany, the German national industry returns are filtered out of the ten Eurozone wide industry returns. The Eurozone wide industry returns will be value-weighted portfolios of the national industry returns of the remaining eight countries, excluding Germany. The same reasoning holds for employing the style regression where an industry returns is the dependent variable, as
expressed in equation (5). For example, when employing the style regression for Basic Materials, the independent variables are the nine country returns as constructed in the second step.
However, when employing the filtered regression for Basic Materials, the national industry returns on Basic Materials will be filtered out of the country returns. The country returns will be value-weighted portfolios of national industry returns of the remaining nine industries, excluding Basic Materials.
The fifth step is defining the country and industry specific variances. If the 𝑅𝑐𝑡𝑟2 of the first regression (4) would be equal to one, then a country return is perfectly replicated by a portfolio of industry returns. Then 𝜀𝑙,𝑡𝑐𝑡𝑟 would be equal to zero. This implies that 𝜀𝑙,𝑡𝑐𝑡𝑟, the
residual variance, measures the variance in the country return that is not captured by the portfolio of industry returns and therefore represents the country-specific variance (Eiling et al., 2012). Return variances between country may differ, therefore using (1 − 𝑅𝑐𝑡𝑟,𝑙2 ) as measure for country-specific variance is the most reliable. Analogous reasoning holds for calculating the industry-specific variance.
The sixth step is averaging both the country-specific variances and the industry-specific variances. The averaged country-specific variance is:
𝐴𝐶𝑆𝑉 = ∑𝐿𝑙=1𝑤𝑙∗ (1 − 𝑅𝑐𝑡𝑟,𝑙2 ). (6)
Then the averaged industry-specific variance is:
𝐴𝐼𝑆𝑉 = ∑ 𝑤𝑘∗ (1 − 𝑅𝑖𝑛𝑑,𝑘2
𝐾
𝑘=1 ) (7)
, where both the equally weighted (EW) and the value-weighted (VW) measures for 𝑤𝑙 and 𝑤𝑘 are used.
The seventh and last step is constructing the variance ratio. The goal of this research is to compare the relative importance of industry and country effects. By dividing the averaged country-specific variance by the averaged industry-specific variance, a measure for comparing the relative importance is constructed. Therefore, the variance ratio is:
10
𝑉𝑅 =𝐴𝐶𝑆𝑉
𝐴𝐼𝑆𝑉 (8)
If the variance ratio is equal to one, country effects and industry effects are considered to be equally important. If the variance ratio is smaller than one, industry effects are dominating country effects.
3.2 Hypotheses
The hypothesis is as follows: “The level of equity market integration has decreased after the two crises in the Eurozone. ”
The main measure for comparing the level of equity market integration is the variance ratio, as described in equation (8). Therefore, the null hypothesis is stated as follows:
𝐻0: 𝑉𝑅𝑎𝑛𝑡𝑒 > 𝑉𝑅𝑝𝑜𝑠𝑡.
From this follows that the alternative hypothesis is formulated as follows: 𝐻1: 𝑉𝑅𝑎𝑛𝑡𝑒 < 𝑉𝑅𝑝𝑜𝑠𝑡.
The main arguments for this hypothesis are summarized here. Firstly, the process of equity market integration is dynamic, which explains the constant academic interest in equity market integration. Secondly, the process is impacted by major economic events, such as crises (Huyghebaert & Wang, 2010 and Chen et al., 2014). Thirdly, the sovereign debt crisis and the events in its aftermath showed that leaving the euro is a real possibility (Bartram & Wang, 2015). Therefore, process of equity market integration could have been decreased as the confidence in the EMU might have decreased. Fourthly, the results of Lekhonen (2010) and Bartram & Wang (2010) indicate that a decrease the level of equity market integration in Europe is a realistic expectation. These arguments have extensively been discussed in Section 2.
11 4. Data
4.1 Data collection
The sample consists of nine countries that have implemented the euro at its introduction. Austria and Ireland are excluded because of the unavailability of returns of major national industries. Luxembourg is excluded because of its unique capital in- and outflow. The sample contains data from November 2003 – December 2017. The first subsample contains data from November 2003 – August 2008 and is referred to as the ante-crisis period. The second subsample involving the two crises contains data from September 2008 – June 2013. The bankruptcy of the Lehman Brothers in September 2008 marked the start of the global financial crisis (FT reporters, 2008). In Europe, this crisis swiftly reverted into the sovereign debt crisis. The third subsample contains data from July 2013 – December 2017. Graph 2 supports the proposition that the European economy started to recover after July 2013, therefore this month is pinpointed as the end of the crisis period in this research. The data for the sample of returns that is used for the style analysis regressions is composed of returns for ten Eurozone-wide industry sectors and returns for nine countries. These returns are constructed by using the weekly national industry returns of each country. Both the ten weekly national industry indicesand the market value of the constituents of these indices are obtained from Datastream. An overview of the national industries, its coverage of the market and constituents is provided in Appendix 2.
4.2.1 Descriptive statistics – Country returns
In Table 1, the statistics of the country returns for ante-crisis period, the period involving the crises and the post-crisis period are reported. In the period before the crisis, all countries have a positive average weekly return and the volatility is on average at a level around 2%. Greece displays the highest return and Finland displays the highest volatility. Italy has the lowest average return with a value of 0.1452% and Spain has the lowest volatility with a value of 1.7757%.
In the period involving the two crises, Finland, Greece, Italy, Portugal and Spain have on average negative weekly returns. Greece displays with an average return of -0.4139 the most negative return. It also has the highest volatility with a value of 4.8486 %, which is
approximately two times the volatility of Greece in the ante-crisis period. For all countries, the volatility has increased above a level of 3%. Therefore, it can be concluded that volatility is high, and that the mean is relatively small for the crisis period.
12
Table 1 – Country returns
Table 1 summarizes the statistics of the country returns for the three periods. The country returns are calculated by composing replicating portfolios of the national industry returns within a country. The ante-crisis period (July 2003 – August 2008) is denoted by ‘’Ante’’, the period involving the two crises (September 2008 – June 2013) is denoted by ‘’Crisis’’ and the post-crisis period (July 2013 – December 2017) is denoted by ‘’Post’’. Furthermore, the mean and the standard deviation (SD) are reported. N is the number of observations aggregated over the three periods.
Country returns
Ante Crisis Post
Mean SD Mean SD Mean SD N
Belgium 0.2085 1.9342 0.1434 3.0089 0.2359 2.0740 741 Finland 0.2342 2.6711 -0.0282 3.5680 0.3028 2.2854 741 France 0.2084 1.8369 0.0688 3.1327 0.2640 2.1490 741 Germany 0.2296 1.7973 0.1077 3.1496 0.2377 2.2524 741 Greece 0.2531 2.4341 -0.4139 4.8486 -0.2293 5.8796 741 Italy 0.1452 1.7879 -0.0703 3.6311 0.2552 2.7610 741 Portugal 0.1879 1.7942 -0.0120 3.0453 0.1264 2.7931 741 Spain 0.2422 1.7757 -0.0169 3.5899 0.2522 2.5600 741 Netherlands 0.2134 1.9472 0.0122 3.3657 0.2815 2.1373 741 In the post-crisis period, the average of the weekly returns has increased for all countries relative to the crisis period. For Finland, Italy, Portugal and Spain, the return has increased from a negative to a positive level. The return for Greece has increased as well but remained negative at -0.2293%. Furthermore, for all countries except Greece the volatility has decreased. Finland shows, with a decrease of 1.23 percentage points, the largest decrease in volatility. Greece contrasts with this decrease in volatility by showing an increase of 1.03 percentage points. Overall, it can be concluded that in the post-crisis period, relative to the crisis period, the mean return has increased, and the volatility has decreased. Besides, all countries excluding Greece and Portugal show higher average returns in the post-crisis period than in the ante-crisis period. This implies that recovery from the crises has been rather strong.
Greece, Italy, Portugal and Spain (hereafter: GIPS countries) display, on average over the three periods, the lowest volatility adjusted return. Graph 1 displays the difference in returns between the GIPS countries and the value-weighted portfolio of Belgium, Finland, France, Germany and the Netherlands (hereafter: core countries).
13
Graph 1 – Plotted returns
Graph 1 displays both the return of the value weighted portfolio of the core countries and the return of the value weighted portfolio of the GIPS countries.
Graph 1 shows that the global financial crisis generated a large shock in both the core countries and the GIPS countries. Graph 2 displays the market value for respectively the full sample, the core countries and the GIPS countries. These market values are used for constructing the value weighted Eurozone wide country portfolio, which is used for estimating the conditional
volatility. The total market value experienced sharp declines in both the global financial crisis and the sovereign debt crisis. The market value of the core countries moves along the total market value but displays less sharp declines. These sharp declines are therefore caused by the sharp decline in the market value of the GIPS countries.
14
4.2.2 Descriptive statistics – Industry returns
In Table 2, the statistics for the industry returns for the ante-crisis period, the period involving the crises and the post-crisis period are reported.
Table 2 – Industry Returns
Table 2 summarizes the descriptive statistics of the industry returns for the three periods. The industry returns are calculated by constructing replicating portfolios of the national industry returns across countries. The mean and standard deviation (SD) are reported. The average weight of an industry in the given time period within the value-weighted portfolio of industry returns is denoted by ‘’Weight’’. These weights will be used for constructing the averaged variances as stated in equation (7). N is the number of observations aggregated over the three periods. ‘’Basic Mat’’ denotes the industry Basic Materials, ‘’Cons Goods’’ denotes the industry Consumer Goods and ‘’Cons Svs’’ denotes the industry Consumer Services.
Industry Returns
Ante Crisis Post
Mean SD Weight Mean SD Weight Mean SD Weight N
Basic Mat 0.4180 2.1382 6.8628 0.0535 3.8797 7.5542 0.2219 2.4653 7.6092 741 Cons Goods 0.20405 2.0058 9.4599 0.2427 3.2584 17.6523 0.2586 2.2193 21.3717 741 Cons Svs 0.1254 1.7414 8.2878 0.1352 2.5690 7.5340 0.2452 2.0314 8.0295 741 Financials 0.1513 2.0681 30.1380 -0.1080 4.4542 21.6566 0.2405 2.9157 21.0967 741 Healthcare 0.1341 1.6372 6.3344 0.2406 2.4099 4.3582 0.1992 2.1495 5.4143 741 Industrials 0.2185 2.3187 9.7108 0.0911 3.3181 14.2067 0.2912 2.1556 15.6902 741 Oil & Gas 0.2641 2.2500 7.1550 -0.029 3.6352 6.7189 0.1351 2.9823 4.4524 741 Technology 0.0434 2.7553 4.8967 0.0540 3.2197 4.1318 0.3221 2.1716 5.3599 741 Telecom 0.1417 2.0357 7.4003 -0.0540 2.7336 5.6382 0.2544 2.5783 4.4440 741 Utilities 0.4031 1.8738 9.7546 -0.1926 3.2507 10.5487 0.2206 2.2814 6.53195 741
In the ante-crisis period, the industry Basic Materials has the largest average weekly return with a value of 0.4180%. All other industry returns are positive as well and display low volatility, where the industry Technology has the highest standard deviation with a value of 2.75%.
During the crisis period, industries are highly volatile with a minimum standard deviation of 2.41%, which is in the range of the highest volatility in the ante-crisis period. Financials displays the highest volatility with a value of 4.45%, which is more than twice the volatility of this industry in the ante-crisis period. Furthermore, Financials, Oil & Gas, Telecom and Utilities have negative average weekly returns. The industry Financials remains the largest industry in the value weighted portfolio but loses 28% of its weight which is equivalent to a decrease of 8.48 percentage points.
15 In the post-crisis period, no industry reports a negative average weekly return. The
Financials industry shows greatest recovery, improving to an average weekly return of 0.24%. Furthermore, overall volatility is decreased where no industry reports a volatility higher than 3%.
5. Results
The style regressions are employed for four different samples: the full sample, the ante-crisis sample, the crisis sample and the post-crisis sample. The structure of this Section is as follows. First, the coefficients and specific variances for the full sample style regressions are discussed. In this discussion, the interpretation of the results will be elaborated upon. Second, both the country specific and industry specific variances for the ante-crisis period, the crisis period and the post-crisis period are reported and discussed. Third, the variance ratios are presented and discussed, after which a conclusion regarding equity market integration in the post-crisis period is drawn. The dynamics of equity market integration are examined by testing the difference between the variance ratio in the ante-crisis period and the variance ratio in the crisis period, and by testing the difference between the variance ratio in the crisis period and the variance ratio in the post-crisis period. Fourth, alternate specifications of the methodology are employed in two robustness checks.
5.1 Full sample regression
The full sample style regressions are employed for both regressions with country returns as independent variable and regressions with industry returns as independent variable. The results of these regressions are presented in respectively Appendix 3a and Appendix 3b. In both
appendices, the results for the full and the filtered regressions are reported. The (1-R2) are higher for all the filtered regressions, which means that the model explains the variance of the
dependent variable less. This is explained by the fact that the overlapping elements are filtered out. The (1-R2) for the filtered regression for Finland is 9.6 percentage points higher than the (1-R2) for the full regression, which is the largest difference in the whole sample. The coefficients for this regression also display differences. This is explained by the fact that Finland has a large weight in the Eurozone wide technology sector (as shown in Appendix 3b). Since the national industry returns of Finland have been taken out of the technology sector in the filtered
16 In Appendix 3a, it is reported that Greece has a country specific variance of 72.2% for the filtered style regression. Therefore, the portfolio of industry returns only explains 27.8% of the variance of the return of Greece. This country specific variance is time-varying: for Greece, it decreases during the crisis period and increases in the post-crisis period.4 With a value of 8.7% France has the lowest country specific variance. Therefore, the portfolio of industry returns explains the variance in the return of France relatively well.
In Appendix 3b, it is reported that the coefficients for both France and Germany in the portfolio estimating the return of the ten industries are relatively large. This is explained by the fact that France and Germany are the largest economies in the sample, as shown by the average weights in the value-weighted portfolio of countries reported in Appendix 4. The industry specific variance is the largest for the healthcare industry with a value of 59.1%.
All reported country and industry specific variances, measured by (1-R2), are
significantly different from zero. This result holds for the style analyses in respectively the ante-crisis period, the ante-crisis-period and the post-ante-crisis period. Therefore, in neither the full sample and its subsamples country specific variances or industry specific variances are absent. The early research of Rouwenhorst (1999) shows that between 1978 and 1998 both industry and country variances were already present in the European equity market.5
5.2 Average specific variances
Using the (1-R2) resulting from the regressions for the ante-crisis period, the crisis period and the post-crisis period, the average specific variances are constructed. These are reported in Table 3. For both the full and the filtered returns, the value weighted variances are lower than the equally weighted variances. This is caused by assigning an equal weight to Greece and Germany,
whereas Greece has a large country specific variance and small market weight and Germany has a small country specific variance and a large market weight. Furthermore, the average variances resulting from the full regressions are lower than those resulting from the filtered regressions. For instance, the average country variance is, aggregated over the three periods, 61.58% higher for the filtered regression relative to the full regression. The elimination of the overlapping
4 The (1-R2) for all models in the 3 remaining sample periods are available on request. 5 Rouwenhorst speaks of country effects and industry effects.
17 returns explains the lower R2 and thus results in a higher (1-R2) for both the country style and industry style regressions.
Table 3 – Average specific variances
In this table, the average country specific variances (ASCV) and industry specific variances
(AISV) are reported for the three periods. Both the variances computed by the equally weighted (EW) and the value weighted (VW) method are reported. The VW method uses the weight of the
country or industry in a Eurozone wide portfolio for the timespan of the regression. Under ‘’Full’’, the average variances for the full style regressions are reported. Under ‘’Filtered’’, the average variances for the filtered style regressions are reported.
Average variances
Equally weighted Full Value weighted
Ante Crisis Post Ante Crisis Post
ACSV 22.52% 20.46% 23.02% 11.28% 9.39% 8.95%
AISV 32.84% 23.97% 24.71% 27.15% 18.82% 17.75%
Filtered
ACSV 28.29% 25.27% 27.74% 16.75% 15.70% 15.13%
AISV 42.48% 32.11% 29.02% 37.20% 28.76% 24.98%
The equally weighted country specific variances for the full regression report a decreased average country variance during the crisis period and show an increase in the post-crisis period. However, the value weighted variance reports a decreasing pattern for all periods. The average variance during the crisis (9.39%) has, relative to the ante-crisis variance (11.28%) decreased with 16.75%. The decrease in the post-crisis period, relative to the crisis period is lower with a value of 4.69%.
The industry specific variances for the full regression show a similar pattern. The equally weighted industry variances show a decrease in the crisis-period and a slight increase in the post-crisis period. The value weighted variances, again, show a decreasing pattern where the largest difference is observed between the ante-crisis period and the crisis period. The average variance during the crisis (18.82%) has, relative to the ante-crisis variance (27.15%) decreased with 30.68 %. The decrease in the post-crisis period, relative to the crisis period is lower with a value of 5.69%.
The results found for the full regressions are confirmed by the results for the filtered regression. Both the value weighted country specific and industry specific measures show a decreasing trend. Furthermore, the decrease is the largest in the transition from the ante-crisis
18 period to the crisis period.
The difference in trend between the value-weighted and equally-weighted measures is
explained by considering the country specific variance of Greece in the post-crisis period. This is nearly 80 % and is therefore increasing the post-crisis equally-weighted country variance. The value-weighted post-crisis country variance assigns Greece its weight according to its market value and therefore decreases.
The average variances for both countries and industries show a decreasing trend. This is contrasting with the findings of Eiling et al. (2012), who find a decreasing average country-specific variance and an increasing average industry country-specific variance after the advent of the euro. Over more, the results in Table 3 report the largest decreases for the industry variances with values of 30.68% and 5.69%, whereas the decreases for the country variances are lower with values of 16.75% and 4.69%.
5.3.1 Variance ratios
In Table 4, the variance ratios are reported. For both the filtered and full regressions and two alternative measures of the ratios, an increase is reported. This indicates that country effects have increased in importance. For the value-weighted measure resulting from the filtered regression, an increase from 0.4504 to 0.6056 over the period before the crisis and the period after the crisis is reported. This is an increase of 34.45%. The difference between the variance ratio in the ante-crisis period and the post-ante-crisis period is significant at the 1% level, as shown by the p-value. This significance level holds for both specifications of the regressions and both definitions of the measures. Therefore, the relative importance of country effects has increased and the level of equity market integration in Europe has decreased.
This result is in line with the concept of time-varying equity market integration.
Furthermore, the results confirm that major economic events influence the relationship between stock markets, as Huyghebaert & Wang (2010) stated in their research. The reported increase in the relative importance of the country effects is in line with the expectations of Bartram & Wang (2015) and Lekhonen (2015). The reported decrease in the level of equity market integration within Europe conflicts with the expectations of several authors (Bekeart et al., 2013, Gebka & Karoglou, 2013 and Eiling et al., 2012). Forbes & Rigobon (2002) and Pukthuatong & Roll (2009) report an increase in equity market integration during crises, but do not focus on the long run. This study does not focus on the short run, because the period containing the two crises is
19 defined in such a way that the short-run effect cannot be captured. The long-run effect of the global financial crisis and the sovereign debt crisis is a reversion of the process of equity market integration and a decrease of its level. This change might reflect the decreasing confidence in the EMU and in a unified Europe.
Table 4 – Variance ratios
This table reports the variance ratios (VR) for the ante-crisis, crisis and post-crisis periods. This variance ratio measures the relative importance of the averaged country specific variances versus the averaged industry specific variances. If the variance ratio is equal to one, the two variances are considered to be equally important. However, if the variance ratio is smaller than one, the industry variances are considered to be more important. Under ‘’Full’’, the variance ratios resulting from the full style regressions are reported. Under ‘’Filtered’’, the variance ratios resulting from the filtered style regressions are reported. Standard errors are between brackets. The p-value is the p-value of testing the null hypothesis that the variance ratio in the post-crisis period is smaller than the variance ratio in the ante-crisis period. This p-value is the p-value of testing the hypothesis that is formulated in Section 3.2.
Variance ratios
Equally weighted Full Value weighted
Ante Crisis Post p-val Ante Crisis Post p-val
VR 0.6857 0.8543 0.9316 0.0000 0.4153 0.4988 0.5043 0.0000 (0.03) (0.02) (0.02) (0.02) (0.01) (0.01) Filtered VR 0.6661 0.7869 0.9557 0.0000 0.4504 0.5457 0.6056 0.0000 (0.04) (0.03) (0.02) (0.03) (0.02) (0.02) 5.3.2 Dynamics
The process of equity market integration is more elaborately studied by testing the difference of the variance ratios between the several subsamples. Table 4 reports a significance difference for this measure between the ante-crisis period and the post-crisis period; however, by examining the difference of this measure between all periods, the dynamics of the decrease in equity market integration are identified. Table 5 reports the p-values of the difference in variance ratios between all periods.
Table 5- P-values
In this table, the p-values for two null hypotheses (denoted by H0) are reported. It is tested if the variance ratio in
the crisis-period is equal to the variance ratio in the ante-crisis period and if the variance ratio in the crisis-period is equal to the variance ratio in the post-crisis period.
p-values
Equally weighted Full Value weighted
H0 VRante=VRcrisis VRcrisis=VRpost VRante=VRcrisis VRcrisis=VRpost
p-value 0.0000 0.0026 0.0002 0.7325
Filtered
20 For all specifications, the null hypothesis that the variance ratio in the ante-crisis period is equal to the variance ratio in the crisis-period is rejected at a 2% significance level. The null hypothesis that the variance ratio in the crisis equals the variance ratio in the post-crisis period is not
rejected for the value-weighted ratio resulting from the full regression. However, it is rejected for both the equally weighted and the value weighted measure resulting from the filtered regression, respectively at 1% and 5% significance level. Summarizing, for the value-weighted measure of the filtered regression, both hypotheses are rejected. Therefore, the process of equity market integration has been continuously changing.
5.4 Robustness checks
Two alternate specifications of the methodology are performed to test the robustness of the discussed results. Results from both tests are compared with the results from the main methodology. Firstly, the unstandardized returns are used for the style analysis regressions. Therefore, the application of the GARCH model to adjust for high volatility periods is abandoned.
Secondly, the sector Financials is eliminated from the sample. The results reported in Appendix 3a show that Financials is a main determinant for the country returns, since it has a positive coefficient for every country, Finland excluded. Furthermore, the industry weights reported in table 2 show that the aggregated average weight over the full sample for the sector Financials is equal to 24.29%.6 Lee et al. (2013) find similar results and conclude that the sector Financials has significant power for explaining the movement in market returns for Asian
countries. After eliminating Financials, the country returns are composed of nine industry returns (instead of ten), which are value-weighted across and within countries, excluding both the return and the market value of the industry Financials. Therefore, it is examined if the main results regarding equity market integration hold if the main explanatory industry is taken out of the sample.
21
5.4.1 Unstandardized returns
Table 6 reports the variance ratios of the unstandardized returns. The unstandardized returns are not scaled by their volatility.
Table 6 – Variance ratios unstandardized returns
This table reports the variance ratios (VR) for the ante-crisis, crisis and post-crisis periods. These variance ratios are constructed using the (1-R2) resulting from the regressions of the unstandardized returns.
Variance ratios
Equally weighted Full Value weighted
Ante Crisis Post p-val Ante Crisis Post p-val
VR 0.6604 0.2237 1.0214 0.0000 0.3990 0.1510 0.5130 0.0000
(0.03) (0.03) (0.01) (0.02) (0.02) (0.01)
Filtered
VR 0.6604 0.7264 0.9471 0.0000 0.4303 0.4820 0.5828 0.0000
(0.04) (0.03) (0.01) (0.04) (0.03) (0.01)
For the ratios resulting from the full regression, extremely low values for the crisis period are reported, which contrasts with the findings for the standardized returns. This is correlated to the high volatility during the crisis period. Therefore, using a GARCH model to take account for this volatility is proven to be essential. The main result, a rejection of the null hypothesis that the variance ratio in the post-crisis period is lower than the variance ratio in the ante-crisis period, holds at all significance levels. In comparison, the reported increase of the variance ratio from the ante-crisis period to the post-crisis period for the unstandardized returns is equal to 35.44 %, whereas this is 34.45% for the standardized returns.7 Both empirical specifications therefore report a significant increase of the relative importance of country effects.
5.4.2 No Financials
In Table 7, the variance ratios based on data excluding Financials are reported. These variance ratios show a similar decreasing pattern as the ratios based on standardized returns. The reported p-values show that the main result, a rejection of the null hypothesis that the variance ratio in the post-crisis period is lower than the variance ratio in the ante-crisis period, holds at all
significance levels. The variance ratio however, is higher than the ratio for both the standardized and unstandardized returns with a value of 0.7488, whereas for the standardized and
unstandardized returns the ratio has a value of respectively 0.6056 and 0.5828. The reported
22 increase from the ante-crisis period to the post-crisis period is equal to 57.74%, which is 23.29 percentage points higher than the reported increase for the standardized returns. The exclusion of Financials therefore caused a higher increase in the relative importance of country effects.
Table 7 – Variance ratios excluding Financials
This table reports the variance ratios (VR) for the ante-crisis, crisis and post-crisis periods. These variance ratios are constructed using the (1-R2) resulting from the regressions, excluding Financials as constituent.
Variance ratios
Equally weighted Full Value weighted
Ante Crisis Post p-val Ante Crisis Post p-val
VR 0.8871 1.1500 1.2018 0.0000 0.4317 0.6121 0.6402 0.0000 (0.04) (0.06) (0.06) (0.02) (0.03) (0.03) Filtered VR 0.7698 1.0051 1.0707 0.0000 0.4747 0.6735 0.7488 0.0000 (0.04) (0.02) (0.02) (0.04) (0.02) (0.01) 5.4.3 Dynamics
Appendix 6 shows that the robustness checks report various dynamics of the level of equity market integration. For the unstandardized returns, the p-values indicate that the difference of the variance ratios in the crisis period and the post-crisis period are significant at all significance levels for all specifications of the measure, whereas the ratios resulting from the filtered regression report p-values of difference with a value of 0.1868 and 0.2390 for the difference between the ante-crisis period and the crisis period. Table 6 reports a rather low variance ratio during the crisis and a major increase during the post-crisis period. This is coherent with the p-values of difference for the crisis period and the post-crisis period.
For the robustness check excluding Financials, the p-values indicate that the difference of the variance ratios in the ante-crisis period and the crisis period are significant at all significance levels for all specifications of the measure, whereas the p-values for the difference between the crisis period and the post-crisis period of the filtered regression report a significant difference. These results are similar to the p-values of difference for the standardized returns.
The results are robust to two robustness checks. After using unstandardized returns and using standardized returns, but excluding Financials, the main result holds: a significant increase in the level of equity market integration is reported when comparing the ratios of the ante-crisis and the post-crisis period. The robustness checks report different dynamics of the level of equity market integration between the different periods. This difference is caused by a misspecification
23 of the variance ratios based on the unstandardized returns. The p-values based on the portfolios excluding Financials confirm the dynamics described by the p-values resulting from the
standardized returns.
6. Conclusion
This thesis adds another chapter to the debate on the relative importance of country and industry effects by assessing the level of equity market integration after a period of turmoil. Weekly national industry returns between 2003 and 2017 are used to construct country returns and Eurozone-wide industry returns. By employing a style regression analysis, it is examined if the level of equity market integration within the EU has been affected by the two crises and its aftermath. Furthermore, the dynamics of equity market integration within the EU are assessed. The results show a decrease in the level of equity market integration after the crisis. A significant increase in the relative importance of country effects between the ante-crisis and the post-crisis period is reported. Over more, the process of equity market integration has been changing
continuously, as a significant difference in the relative importance of country effects between the ante-crisis and crisis period and between the crisis period and post-crisis period has been
reported. The decrease in the level of equity market integration is the largest between the ante-crisis and ante-crisis period. Therefore, the results confirm that the process of equity market
integration is a dynamic and time-varying process.
The preponderance of academic papers reports an increasing level of equity market integration within Europe after the foundation of the EMU. The results of this thesis show that this process of increasing dominance of industry effects has been reverted by the effects of the crises within Europe. This is in line with the assertion of Huyghebaert & Wang (2010), who state that the fundamentals of relationships between equity markets are affected by major economic events. Furthermore, the dynamics of the process show that the level of equity market integration has continuously been decreasing. This highlights afresh the importance of re-examining the level of equity market integration within the EU, a geographical area that was considered to be highly integrated.
Due to the unavailability of data, both Austria and Ireland are not included in the sample. However, these two countries did implement the Euro at its introduction. By repeating this study at a later point in time, these two countries can be incorporated since the unavailability of data is
24 merely present in the ante-crisis period. Then the sample would consist of 11 countries and broader statements on the level of equity market integration within the EU could be made. Furthermore, statistical inferences are drawn based on the assumption that the returns are normally distributed. Weekly returns are used, because longer-term horizon returns display a normal distribution, in contrary to daily returns. However, by constructing standard errors based on an alternative test, for instance a t-test, the results of this study would be even more reliable. A decrease in the level of equity market integration has strong implications for both investors and policy makers. The decrease implicates that diversification benefits within Europe has increased, that the cost of capital has decreased, that the growth opportunities between industries differ and that the level of economic development has decreased. Furthermore, the decrease of equity market integration is an indicator of the vulnerability of the EMU and reflects a decreasing confidence in the EMU and a unified Europe. Therefore, European policy makers need to undertake action to increase the confidence in both the EMU and the EU. Furthermore, one needs to consider that countries leaving the Euro might be a real possibility.
To further examine the long-term effects of the two crises on equity market integration, future research is needed. This thesis focuses on nine EU-countries that have implemented the Euro at its introduction. However, after the advent of the euro, more countries have implemented the currency. After 2002, Bulgaria, Croatia, Cyprus, The Czech Republic, Estonia, Latvia, Lithuania, Malta, Slovakia and Slovenia have adopted the Euro. Examining the level of equity market integration and the long-term effect of the crisis on this level including (some of) these countries is a valuable addition to the research done in this thesis. Since the level of equity market integration for nine core countries of the EU has decreased after the crisis, it is interesting to examine if this result is exacerbated for countries that have adopted the Euro at a later stage.
25 7. Bibliography
Adjouaté, K, Danthine, J. (2004). Equity returns and integration: Is Europe changing? Oxford Review of Economic Policy 20, 555-570.
Aggarwal, R., Lucy, B., & Muckley, C. (2010). Dynamics of Equity Market Integration in Europe: Impact of Political Economy Events, Journal of Common Market Studies 48, 641-660.
Baele, L. (2005). Volatility Spillover Effects in European Equity Markets, Journal of Financial and Quantitative Analysis 40, 373-401.
Bartram, S.M., Wang, Y. (2015). European financial market dependence: An industry analysis. Journal of Banking & Finance 59, 146-163.
Bekaert, G, Harvey, C.R. (1995). Time-Varying World Market Integration, The Journal of Finance 50, 403-444.
Bekaert, G., Harvey, C.R., Lundblad, C.T., & Siegel, S. (2011). What Segments Equity Markets?, The Review of Financial Studies 24, 3841-3890.
Bekaert, G., Harvey, C.R., Lundblad, C.T., & Siegel, S. (2013). The European Union, the Euro and equity market integration, Journal of Financial Economics 109, 583-603.
Chan, K.C., Gup, B.E., & Pan, M. (1997). International stock market integration: A study of eighteen nations, Journal of Business, Finance and Accounting 24, 803-813.
Chen, M., Chen, P., & Lee, C. (2014). Frontier stock market integration and the global financial crisis, North American Journal of Economics and Finance 29, 84-103.
Christoffersen, P.F. (2012). Elements of Financial Risk Management. Oxford: Elsevier.
Click, R.W., Plummer, M.G. (2005). Stock market integration in ASEAN after the Asian financial crisis, Journal of Asian Economics 16, 5-28.
Colacito, R., Croce, M., M. (2010). The Short and Long Run Benefits of Financial Integration, American Economic Review 100, 527-321.
26 Eiling, E., Gerard, B., & De Roon, F. (2012). Euro-Zone Equity Returns: Country versus
Industry Effects, Review of Finance 16, 755-798.
Engle, R.F. (1982). Autoregressive Conditional Heteroskedasticity with Estimates of the Variance of United Kingdom Inflation, Econometrica 50, 987-1007.
Forbes, K.J., Rigobon, R. (2002). No Contagion, Only Interdependence: Measuring Stock Market Comovements, The Journal of Finance 57, 2223-2261.
FT reporters (2008). Lehman Brothers files for bankruptcy, Financial Times, 16 September 2008.
Gebka, B., Karoglou, M. (2013). Have the GIPSI settled down? Breaks and multivariate stochastic volatility models for, and not against, the European financial integration, Journal of Banking & Finance 37, 3639-3653.
Goetzmann, W.N., Li, L., & Rouwenhorst, K.G. (2005). Long-Term Global Market Correlations, The Journal of Business 78, 1-38.
Griffin, J.M. (2002). Are the Fama and French Factors Global or Country Specific?, The Review of Financial Studies 15, 783-803.
Hardouvelis, G.A., Malliaropulos, D., & Priestly, R. (2007). The impact of EMU on the equity cost of capital, Journal of International Money and Finance 26, 305-327.
Heston, S.L, Rouwenhorst, K.G. (1994). Does industrial structure explain the benefits of international diversification? Journal of Financial Economics 36, 3-27.
Huyghebaert, N., Wang, L. (2010). The co-movement of stock markets in East Asia. Did the 1997-1998 Asian financial crisis really strengthen stock market integration? China Economic Review 21, 98-112.
In, F., Kim, S., Yoon, J.H., & Viney, C. (2001). Dynamic interdependence and volatility of Asian stock markets Evidence from the Asian crisis, International Review of Financial Analysis 10, 87-96.
27 Kearny, C, Lucy, B.M. (2004). International equity market integration: Theory, evidence and implications, International Review of Financial Analysis 13, 571-583.
Kim, S.J, Moshirian, F. (2005). Dynamic stock market integration driven by the European Monetary Union : An empirical analysis, Journal of Banking & Finance 29, 2475-2502.
Lee, C., Chen, M., & Chang, C. (2013). Dynamic relationships between industry returns and stock market returns, The North American Journal of Economics and Finance 26, 119-144.
Lekhonen, H. (2015). Stock Market Integration and the Global Financial Crisis, Review of Finance 19, 2039-2094.
Lessard, D.R. (1974). World, National, and Industry Factors in Equity Returns, The Journal of Finance 29, 379-391.
Panton, D.B., Lessig, V.P., & Joy, O.M. (1976). Comovement of International Equity Markets: A Taxonomic Approach, The Journal of Financial and Quantitative Analysis 11, 415-432.
Pukthuanthong, K, Roll, R. (2009). Global market integration: An alternative measure and its application, Journal of Financial Economics 94, 214-232.
Rouwenhorst, K.G. (1999). European Equity Markets and the EMU, Financial Analysts Journal 55, 57-64.
Sharpe, W.F. (1992). Asset Allocation: Management Style and Performance Measurement, Journal of Portfolio Management, 7-19.
Stavrakakis, Y., Katsambekis, G., Nikisianis, N., Kioupkiolis, A., & Siomos, T. (2017). Extreme right-wing populism in Europe: revisiting a reified association, Critical Discourse Studies 14, 420-439.
Volosovych, V. (2011). Measuring financial market integration over the long-run: Is there a U-shape? Journal of International Money and Finance 30, 1535-1561.
Walker, N. (2017). Brexit timeline: events leading to the UK’s exit from the European Union. House of Commons Library, no. 07960.
28 Yang, J., Kolari, J.W., & Min, I. (2003). Stock market integration and financial crisis: the case of Asia, Applied Financial Economics 13, 477-486.
Zhang, B., Li, X., & Yu, H. (2013). Has recent financial crisis changed permanently the correlations between BRICS and developed stock markets? North American Journal of Economics and Finance 26, 725-738.
29 Appendix 1 – Conditional volatility
Equity indices and returns display the leverage effect: negative returns tend to push up volatility more than positive returns do (Christoffersen, 2012, p.11). The GJR-GARCH model which is an extension of the general GARCH model controls for this effect. The conditional variance is then given by:
𝜎𝑡+12 = 𝜔 + 𝛼𝑅𝑡2 + 𝛾𝐼𝑡𝑅𝑡2+ 𝛽𝜎𝑡2, where 𝐼𝑡 = 1 if 𝑅𝑡 < 0.
The plotted conditional volatility for the value-weighted portfolio of country returns, estimated by this model is displayed below.
As indicated by the descriptive statistics in section 2 and in graph 1, the conditional volatility is the highest in the both crises periods with a maximum of almost 9% at the end of 2008.
30 Appendix 2 – National industry sectors
In this table, the constituents of the national industry returns and therefore the coverage of the market are displayed. The reported constituents are the total of all countries. The constituents per country might differ and therefore one country might have no tobacco sector as constituent for the industry Consumer Goods but does have other sectors within this industry. The titles of the columns report the ten industries.
Industry constituents
Basic Materials Cons Goods Cons Services Financials Healthcare Chemicals Automobiles Food & Drug
Retailers
Banks Healthcare & Equipment Services
Forestry & Paper Beverages General Retailers Financial Services Pharmaceuticals & Biotechnology Industrial Metals Food Producers Media Equity Investment
Instruments Industrial Mining Household Goods
& Home Construction
Travel & Leisure Life Insurance
Mining Leisure Goods Non-life Insurance
Personal Goods Real Estate
Investment & Services
Tobacco Real Estate
31 Appendix 2 – National industry sectors (continued)
Industry constituents (continued)
Industrials Oil & Gas Technology Telecom Utilities Aerospace & Defense Alternative Energy Software & Computer Services Fixed Line Telecommunication Electricity Construction & Materials
Oil Equipment & Services Technology Hardware & Equipment Mobile Telecommunication
Gas, Water & Multiutilities General
Industrials
Oil & Gas Producers Industrial Engineering Industrial Transportation Support Services
