Market Impact on financial market integration: Cross-quantilogram analysis of the global impact of the euro

Market Impact on financial market integration: Cross-quantilogram analysis of the global impact of the euro

Journal Pre-proof Market Impact on financial market integration: Cross-quantilogram analysis of the global impact of the euro Sebastian Lindman, Tom T...

6MB Sizes 0 Downloads 55 Views

Journal Pre-proof Market Impact on financial market integration: Cross-quantilogram analysis of the global impact of the euro Sebastian Lindman, Tom Tuvhag, Ranadeva Jayasekera, Gazi Salah Uddin, Victor Troster

PII: DOI: Reference:

S0927-5398(19)30086-6 https://doi.org/10.1016/j.jempfin.2019.10.005 EMPFIN 1145

To appear in:

Journal of Empirical Finance

Received date : 15 December 2018 Revised date : 8 October 2019 Accepted date : 11 October 2019 Please cite this article as: S. Lindman, T. Tuvhag, R. Jayasekera et al., Market Impact on financial market integration: Cross-quantilogram analysis of the global impact of the euro. Journal of Empirical Finance (2019), doi: https://doi.org/10.1016/j.jempfin.2019.10.005. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier B.V.

*Separate Title Page

Journal Pre-proof Market Impact on Financial Market Integration: Cross-quantilogram Analysis of the Global Impact of the Euro

by

of

Sebastian Lindman Department of Management and Engineering, Linköping University, Sweden E-mail: [email protected]

pro

Tom Tuvhag Trafikverket, 28 Luntgatan 602 19 Norrköping, Sweden E-mail: [email protected] +46 72-084 84 54

re-

Ranadeva Jayasekera Trinity Business School, Trinity College; Dublin, Ireland E-mail: [email protected]

lP

Gazi Salah Uddin Department of Management and Engineering, Linköping University, Sweden E-mail: [email protected]

Jo

urn a

Victor Troster Departament d'Economia Aplicada, Universitat de les Illes Balears E-mail: [email protected]

*Highlights (for review)

Journal Pre-proof Highlights First study to use a cross-quantilogram method to examine financial market integration The degree of dependence is relatively stronger in a common currency group There is a strong heterogeneous dependence structure for the UK and Germany

of

Dynamics in Cross-quantile correlation show discontinuities in the dependence structure

Jo

urn a

lP

re-

pro

Results can be used to develop effective policy instruments for the ECB and EMU

*Blinded Manuscript Click here to view linked References

Journal Pre-proof Market Impact on Financial Market Integration: Cross-quantilogram Analysis of the Global Impact of the Euro

Abstract

urn a

JEL codes: C1, C5, G1

lP

re-

pro

of

We contribute to the literature by providing a more comprehensive understanding of the impact the euro has had on financial market integration with economies of different characteristics outside and within the European market via inclusion of market conditions influence on the level of financial integration. Our paper employs the recently developed cross-quantilogram (Han et al., 2016) approach to examine quantile dependence between the conditional stock return distributions of Germany and the UK with that of three common currency groups within EMU (Finland, France, and Italy), two global leading markets (the US and Japan), and two of the most promising emerging markets (China and India). We find three key results. First, both the EU membership and the common currency union affect the degree of financial market integration. Nevertheless, disentangling the effects of EU membership from the common currency shows that the common currency group has an additional impact on financial integration, as the degree of dependence is stronger in the common currency group than in the sovereign currency group and other groups. Second, there is a heterogeneous dependence structure, which is strongly observed for the UK and German stock returns with that of developed (the US and Japan) and emerging markets (India and China). Third, cross-quantile correlations change over time, especially in low and high quantiles, indicating that they are prone to jumps and discontinuities in the dependence structure. As far as we are aware, this is the first study in this field employing a cross-quantilogram method to examine the impact of different market conditions on the correlations, making our study a pioneer in the field of stock market integration.

Jo

Keywords: Financial market integration; Cross-quantilogram; Euro; EMU

1

Journal Pre-proof 1

Introduction

Is the membership in a common currency union more beneficial for financial market integration than using a sovereign currency? By entering into a common currency area, the traditional requirements for most functions performed by a sovereign central bank disappear. A common central bank and its monetary policy can influence the integration through synchronized business cycles. The common currency union can also reduce the impact of adverse shocks

of

through risk sharing (Mundell, 1973). As illustrated by the global financial crisis in 2008, the countries adopting the euro increased their business cycle synchronization, which provided an

pro

effective stabilization instrument (Bekiros et al., 2015). An increased financial market integration within the Eurozone and the European Union (EU) supported countries like Greece but simultaneously raised questions of the willingness to participate in the union. In this paper, we employ a cross-quantilogram approach (Han et al., 2016) to analyze quantile dependence between the conditional distributions of stock returns of Germany and the UK with that of three

re-

common-currency groups within the EMU (Finland, France, and Italy), two global leading markets (the US and Japan), and the two most promising emerging markets (China and India). We apply a full-distributional approach that delivers a quantile-based dynamic correlation at various quantiles of the stock returns distribution. To our knowledge, this is the first paper that

lP

provides a complete characterization of the dependence between stock markets over the entire distribution of stock returns. We show that both the EU membership and the common currency union affect the degree of financial market integration. Nevertheless, disentangling the effects of EU membership from the common currency indicates that the common currency union has

urn a

an additional impact on financial integration, as the common currency group has stronger correlations than the sovereign currency group and other groups. This is clearly highlighted in the comparison between Sweden and Norway, both with sovereign currencies, but different EU membership statuses. We also find a heterogeneous dependence structure, which is strongly observed for the UK and German stock returns with that of developed (the US and Japan) and emerging markets (India and China). Finally, cross-quantile correlations change over time,

Jo

especially in low and high quantiles, indicating that financial market integration increases during economic and financial turbulence periods. This suggests that common policy instruments (such as active monetary policy by the European Central Bank, ECB, and migration policy by the Economic and Monetary Union of EU, EMU) could provide an effective stabilization instrument for the entire Euro area.

2

Journal Pre-proof The countries within Europe have developed closer economic ties since the 1950s that today are expressed in a single internal market. The creation of the common market aims to ensure the free movement of capital, goods, labor, and services (European Commission, 2018). Because of globalization, we live in a highly connected world characterized as a global common market nowadays. The EU benefits from globalization as it negotiates trade agreements with countries and regions around the world for its members, e.g., the Transatlantic Trade and

of

Investment Partnership between the EU and the United States of America, US. These sorts of trade agreements are negotiated to eliminate customs duties, remove or reduce customs tariffs,

pro

or provide a general framework for bilateral economic relationships (European Commission, 2018). Therefore, these trade agreements enhance trade and integration, connecting the EU to other countries and regions. The continued global integration hinges on the ability of countries to adjust, adapt, and encourage globalization. A cost of global integration arises when countries give up the autonomy in the form of monetary policy since increased global integration

re-

demands parties to agree to more equivalent game rules (Ku and Yoo, 2013). As previously pointed out, there are several advantages and disadvantages of high financial market integration. To understand better how countries, economies, and markets are affected in

lP

different prospects such as recessions, booms, fluctuations, and economic uncertainties, it is essential to first examine and understand the level of co-movement of business cycles and dependence in economic and financial market integration. The concept of economic integration is comprehensive with several definitions in previous studies and theoretical works; for

urn a

instance, Balassa (1962) defines economic integration as the transformation in welfare occurring in an area when one or more areas reduce their regulations against each other. The transformation, and thus higher integration, can be achieved by increasing trade in goods or by increasing flows of ideas (Rivera-Batiz and Romer, 1991). Nevertheless, Phylaktis and Ravazzolo (2002) suggest that economic integration is highly connected with financial market integration. Accordingly, if two countries share a co-movement in their Gross Domestic

Jo

Products (GDPs), they will face a co-movement in their cash flows and in their financial markets. Further, Schwert (1990), Roll (1992), and Canova and De Nicolo (1995), among others, provide evidence of a positive relation between economic and financial market integration in the US and in Europe. Therefore, if two countries have a high level of integration in the financial markets, we interpret the countries to have synchronized business cycles, thus making financial market integration and business cycle synchronization cooperative concepts.

3

Journal Pre-proof Previous studies in the field of business cycle synchronization have generally focused on three areas: (1) the short- and long-term relationship with the Vector Autoregressive (VAR), Vector Error Correction (VECM), and Autoregressive Distributed Lag (ADL) models; (2) Generalized Autoregressive Conditional Heteroskedasticity (GARCH) Dynamic Conditional Correlation (DCC) models examining the correlation in business cycle synchronization; and (3) determining the dependence, directionality, or causality in a time- or frequency scale.

of

Furthermore, previous studies focused on investigating the level of financial market integration within a common currency area. A frequently studied currency area is the Economic and

pro

Monetary Union of EU (EMU), which have had a significant impact towards an increased business cycle synchronization of the member countries.

The theoretical framework often concludes based on optimum currency area theory, which declares different characteristics for a currency area to be optimum (Mundell, 1961). Among the benefits of being part of a common currency area is the increased risk sharing within the

re-

area (Mundell, 1973). Due to unexpected changes and increased uncertainty followed by globalization and today’s economic context, recessions such as the global financial crisis in 2008 will likely occur again. On the other hand, a country being part of a common currency

lP

area loses its autonomy to regulate monetary policy. The outcome of an optimum common currency area is increased business cycles co-movement within the area (Mundell, 1961). However, how the countries as individuals and groups integrate with countries of different area theory.

urn a

characteristics outside the currency area has no obvious conclusion from the optimum currency

Therefore, the purpose of this study is to investigate the impact a common currency union has on the integration of financial markets. We compare countries participating in the union and countries not participating with major economies outside and within the market of the common currency union. Since the literature already has studied the impact within a common currency group, we aim to shed light towards the integration with major global and emerging economies.

Jo

Further, we aim to study the impact of the common currency union on the financial market integration during different market regimes. We investigate how the relationship of stock market integration affects our countries and groups depending on them experiencing adverse, normal, or good market conditions, namely how the business cycle varies. The impact of market conditions is essential to analyze this to globalization and changing uncertainty in different market regimes, with the global financial crisis as an evident example. Therefore, we apply cross-quantilogram (CQ) analysis to identify whether there are differences in the effects 4

Journal Pre-proof depending on market regimes. The CQ analysis is a relatively new method, as it provides innovative features of dependence methodology originating from 2016 (Han et al., 2016), and it studies the directional predictability, in particular for financial time series; hence, it captures the volatility. This approach uncovers the dependence structure and directional predictability via correlation properties at different market conditions, such as lower-normal-higher quantiles of the distribution, and it provides the time-varying characteristics at different market

of

conditions. In addition, the CQ approach considers the asymmetry spillovers between the two investigated variables and captures shocks as well.

pro

Our starting point is Europe where we have selected three countries within the EMU: Finland, France, and Italy that make up the common currency group. The sovereign currencies group consists of Hungary, Norway, and Sweden. Our major economies include three groups, the European drivers’ group that consists of the economic center of Europe, Germany, and the financial center of Europe, the UK. The Global group consists of the US and Japan. The last

re-

group in this study, the Emerging market group, consists of China and India. Considering our diverse sample, the limitation in previous studies, and our choice of methodology, we focus on the following research questions:

lP

1. What impact has the common currency union had on the financial market integration for countries adopting the Euro and for countries using a sovereign currency in Europe? 2. Are there different integration developments towards the three groups: European drivers, global, and emerging countries?

urn a

3. How do different market conditions affect the level of financial market integration? The United Kingdom, UK, composes the greatest sceptic, declaring referendum for leaving the EU with the result of Brexit. The UK’s exit might jeopardize the economy’s access to one of its most significant markets, currently accounting for 44 percent of its exports and 53 percent of imports (House of Commons Parliament, 2016). Therefore, one of EU’s fundamental

Jo

keystones, the idea of a single market with free movement of goods, services, capital, and labor for approximately 500 million consumers (European Union, 2018) is threatened. Following the news of the UK leaving the EU, it is interesting to examine whether there was a shift in financial integration for the UK, altering from the European market towards the global and emerging market, thereby differencing the UK’s development from Germany’s. Although UK is part of the EU the monitory policy adopted by the Bank of England is different to that of the European Central Bank. Whilst the ECB has been strictly targeting price stability through low inflation with the hope of fostering sustainable long-term economic growth, the Bank of England has 5

Journal Pre-proof shown greater flexibility and willingness to consider other objectives such as promoting full employment, preventing recession, and pursuing quantitative easing. Another fundamental difference between the ECB and the Bank of England is the willingness of the latter to act as a lender of last resort, in contrast to the monetary policy adopted by the ECB, which has led to the fall of the UK bond yields, whereas bond yields are rising in the Eurozone. These results could be used to infer some of the implications of Brexit on the UK. Nevertheless, since our consider Brexit as a secondary, but important, analysis.

of

primary focus is the impact of a common currency area on financial market integration, we

pro

Deriving from the previous studies and our theoretical framework, the following hypotheses apply. A high degree of financial integration should exist within the EU due to its homogeneity with a common market and regulations. The countries that adopted the euro should display a higher degree of integration since they share even more similarities. The connection between Europe and the US is inevitable due to their long history, with Transatlantic Trade and

re-

Investment Partnership for example, and they should present a stronger integration than for Europe with any other country or group. As for the UK’s decision to leave the EU, we study the following interesting phenomenon. One of the arguments put forth in support of Brexit was

lP

that the post-Brexit UK would be in a better position to establish profitable trading relationships with countries outside the EU. Hence, we investigate whether there is increased financial market integration with countries outside the EU in support of this argument. We apply time series analysis to investigate the correlation, which we interpret as the level of

urn a

financial market integration, between the stock markets of twelve countries. Exponential Generalized Autoregressive Conditional Heteroscedasticity (EGARCH), Dynamic Conditional Correlations (DCC), and Cross-Quantilogram are used as econometric models. Our approach focuses on comparing the results from these models to analyze the correlations on (1) an aggregated level, (2) with a time-dependent condition, and (3) under different market regimes, to make conclusions regarding financial market integration. This approach evaluates the

Jo

correlation on both static and dynamic levels. By breaking down the countries correlations into different market regimes, a more complete picture can be generated. Monthly data are obtained from the Thompson Reuters DataStream, World Bank database, and the OECD database, covering the period from January 1993 to September 2017. The results indicate that higher correlations for the common currency group than for the sovereign currency group also with the European drivers and the global countries do exist. We do not find such a pattern with the

6

Journal Pre-proof emerging countries. Furthermore, market regimes influence financial market integration; in particular, adverse market conditions increase stock market integration. Our contribution to the literature is to improve the knowledge of the impact the euro has had on financial integration with economies of different characteristics outside and within the European market. Further, our paper contributes to the literature via inclusion of market conditions influence on the level of financial market integration. To the best of our knowledge,

of

this is the first study in this field conducting a cross-quantilogram method to examine the impact of different market conditions on the correlations, making our study a pioneer in the field of

pro

financial market integration. Therefore, we intend to fill a gap in the existing literature by combining the aggregated and time-dependent approaches including the market regimes. This paper proceeds with a presentation of the related literature in Section 2 that outlines the theoretical framework including financial integration and optimum currency area theory.

re-

Section 3 describes the methodology that consists of methods used in this paper such as EGARCH, DCC, and cross-quantilogram. Section 4 discloses the data and primary analysis, where summary statistics and break and stationarity tests are presented. In Section 5, the implications in Section 6. 2

Related literature

lP

empirical results are presented and a discussion is held, followed by conclusions and policy

The literature of economic and financial integration has historically focused on Europe and on

urn a

the global developed countries such as the US and Japan. A highly researched question is the impact of common currency areas on business cycle synchronization. The question proceeds from Mundell’s (1961) work of optimum currency area theory. In Europe, this has been studied concerning the implementation of the euro as the common currency for the EMU. By using different time series approaches, several authors (e.g., Yang et al., 2003a; Lafuente and Ordóñez, 2009; Rua, 2010) have found that the EMU significantly increased financial

Jo

integration between the member countries. During the last decades, a large number of studies have examined financial integration in emerging markets, as for Asia (e.g., Johnson and Soenen, 2002; Moneta and Rüffer, 2009; Komatsubara et al., 2017). The fundamental issues of business cycle are sensitive to specific geographical or common currency or trade union. The measures of the business cycle activity are fundamentally based on the three different forms, such as economic activity (i.e., GDP and Inflation), financial integration (i.e., equity and currency markets), and trade integration (bi vs multilateral trade connectivity). To understand the 7

Journal Pre-proof business cycles properties, GARCH-type dynamic correlation, VAR-VECM based directionality, cyclizing adjustment process, and multiscale modelling are widely used. Focusing on the financial integration, Yang et al. (2003a) applied a VAR model to conclude that the stock markets of EMU member economies have become more integrated and interdependent. Another interesting finding is the results regarding the stock market of the UK. Since the foundation of EMU, the level of financial integration has declined with the UK (Yang

of

et al., 2003a). Lafuente and Ordóñez (2009) go further as they state that long-run financial integration only appears among the Eurozone members and not with the UK. This long-run

pro

relationship between European stock markets, however, occurred before the implementation of the euro and it was examined through the dynamic conditional correlation multivariate GARCH model (DCC-MV-GARCH). Overall, Rose and Engel (2002) verify the findings of higher synchronized business cycles for the member countries of a currency union than for countries with sovereign currencies. By employing a sample not only referring to EMU, they find that

re-

the countries inside a currency union experience more trade and less volatile real exchange rates. The usage of the exchange rate as a proxy variable for business cycle synchronization is in contrast rare, though Moneta and Rüffer (2009) have found the exchange rate of Japan yen

lP

and US dollar to be influential for synchronization. The multiscale approach distinguishes changes in the pattern of business cycle synchronization, through the inclusion of time and frequency domain (Bekiros et al., 2015).

The European market has grown, and it has become a dominant financial market according to

urn a

Fratzscher (2002). Fratzscher (2002) analyzes European equity markets between 1986 and 2000, using a GARCH model, with three key results: (1) European equity markets have become highly integrated since 1996; (2) the euro area market has taken over from the US as the dominant market in Europe; and (3) the high integration within European equity markets is mainly explained by the development of EMU and the elimination of exchange rate volatility. A greater impact for European financial markets and a shift from the US towards Europe is

Jo

further examined by Diebold and Yilmaz (2014), using VAR and a developed connectedness framework. Focusing on the integration between the US and Europe, Pérez-Rodríguez (2006) finds a high correlation of the European currencies, the euro and British pound, with the US dollar by employing a DCC-GARCH model. Further, a strong connection is shown to exist between the ECB reference rate and the US traded spot rates, according to this study. A high degree of connectedness could partially be explained by Wang and Wen (2007), who analyze

8

Journal Pre-proof inflation co-movements. Their results are an argument to connect inflation and GDP as synchronization measures. In contrast to the economies in Europe, which can mostly be characterized as developed countries, the Asian market consists of a higher share of developing countries. Several studies have shown high financial market integration for the developing economies within Asia (e.g., Johansson, 2011; Komatsubara et al., 2017). This integration has increased significantly since

of

2007, and the global financial crisis in 2008 could potentially explain this (Komatsubara et al., 2017). This argument is supported by a second study from Yang et al. (2003b), showing

pro

strengthened stock market integration during the Asian financial crisis in 1997. Sethapramote (2015) used the Association of Southeast Asian Nations (ASEAN) countries and measured static and dynamic correlations in several macroeconomic and policy variables. A main finding was the evidence of synchronization in the key variables. Further, this study showed that the influence of trade integration on business cycle synchronization was the greatest one within

re-

ASEAN. The financial integration was found most important between ASEAN and the US. Eickmeier and Breitung (2006), Park and Shin (2009), Quah and Crowely (2012), and Sethapramote (2015) are examples of studies that include several variables, conducting

lP

different types of integration. Nevertheless, only Park and Shin (2009) studied a sample consisting of countries from Europe, Asia, and America. Their research question was to analyze East Asia’s business cycle since 1990 and to examine whether the cycle has strengthened within the group, decoupling from the EU and the US. This makes that study one of the few in the field

urn a

analyzing cross-regional integration.

In sum, the previous literature has focused on synchronization within a specific geographical area or a common currency area, with Europe and the EMU as relevant examples. Variables like GDP, industrial production, and inflation have been used along with financial variables such as equity and stock indices. Among the time series methods, VAR, GARCH, and Wavelet analysis are the most commonly used ones in previous studies. Overall, a significant impact of

Jo

EMU towards an increased business cycle synchronization has been shown, with higher financial integration for the euro countries than for countries with sovereign currencies. Characteristics such as being a core country also positively affect the integration among countries in the EU. For the Asian market, financial crises tend to strengthen integration. Few studies have examined cross-regional integration. Therefore, we see a research gap in the previous literature.

9

Journal Pre-proof 3

Data and preliminary analysis

The employed dataset consists of monthly stock market indices. The investigated period spans January 1993 to September 2017, and it comprises observations from the last day of each month. The starting point synchronizes with the implementation of the Maastricht Treaty by the EU, which was essential for the establishment of EMU in 1999 (Bekiros et al., 2015). The selection of countries is based on our purpose, and it contains twelve countries divided into five

of

sub-groups based on their perspective: common currency, sovereign currency, European drivers, and global and emerging markets. The common currency group consists of Finland,

pro

France, and Italy; this selection is done because these countries belong to the EMU as well as that the selection is diversely seen to geographic distribution. Finland is located in Northern Europe, while France and Italy are located in the western and southern parts of Europe. Further, France and Italy can be viewed as two core countries, while Finland has the characteristic of a peripheral country in Europe, which may have implications for the level of financial market

re-

integration (Aguiar-Conraria and Soares, 2011). For the sovereign currency group, we have selected Hungary, Norway, and Sweden. These countries have sovereign currencies inside Europe. Norway and Sweden belong to Northern Europe, while Hungary is a country in Eastern Europe. As a robustness check, we also analyze two major Eastern European economies that

lP

joined the EU during our sample period, Poland and the Czech Republic, which have a sovereign currency. This helps disentangle EU membership from common currency effects. Overall, the diversifications allow for different conditions for the candidates.

urn a

Germany and the UK compose the European drivers. Our approach assumes that Germany is the primary driver of the European economic activity (Bekiros, et al., 2015) and the UK is the driver of the financial activity in Europe (Lane and Milesi-Ferretti, 2008). In the global group, we have selected the US and Japan due to their characteristics as financial centers. The US is a world actor and the leading financial driver in the world, whereas Japan is the leading financial driver on the Asian continent, and the Japanese yen is considered the third most traded currency

Jo

in the world (Kim et al., 2013). Both countries are often included as anchor countries in previous studies examining financial integration (e.g., Hartmann et al., 2003; Sethapramote, 2015). Among the emerging countries, we have selected China and India since they are the most promising ones with a high average growth rate (Herd and Dougherty, 2007). According to Berdiev and Chang (2015), China has increasingly influenced the economic structure of the world in the recent decades.

10

Journal Pre-proof Globalization has led to a significant increase in financial transactions among countries. Capital flows freely over a global market, where the highest return moderates the direction (Büttner and Hayo, 2011). Business cycle synchronization can contribute to financial stability internationally and influence the degree of capital market integration across countries over time (Lafuente and Ordóñez, 2009). Coincidentally, Büttner and Hayo (2011) explain financial integration to be vulnerable to financial crises, thus making the examination of financial

of

markets important. Financial integration tends to affect the relationship between output growth and volatility, besides increasing the international correlation in consumption and GDP

pro

(Lafuente and Ordóñez, 2009).

All stock market indices are extracted in their local currency and transformed into a common currency. We have selected US dollars as the common currency due to its view as a world currency. Since our sample focuses on the European market, with EMU as our common currency area, one could argue that the transformation is a better fit to the euro. Nevertheless,

re-

the euro as a currency was first implemented in 1999, while our sample period starts in 1993, and hence the US dollar is chosen. The transformation to US dollar was conducted via Thompson Reuters DataStream exchange-rate conversion function. The detailed data sources

lP

and variables definitions are presented in the Panel A of Table 1. This transformation implies the influence of exchange rate fluctuations on the indices. Nevertheless, to compare each country’s development, each stock market index needs to be expressed in a mutual currency. We find it better to use US dollars than the euro since the choice of euro would have given the

urn a

sovereign currency countries a disadvantage against the common currency countries towards Germany. Hence, the exchange rate fluctuation would only affect the sovereign currency countries.

Panel B of Table 1 presents the summary statistics of the first differences of stock market indices for the individual countries, during the period from January 1993 to September 2017. The first difference denotes stock market return. In comparison with the entire sample, the stock

Jo

markets of China and Hungary are more volatile, as measured by the standard deviation of the returns. Skewness values are negative for all series, indicating an asymmetry at the left tail of the returns distribution. The kurtosis values suggest non-normal distributions for almost all series, with the exceptions of Italy and Japan. These two countries have kurtosis values close to that of the normal distribution. A negative skewness and a high kurtosis indicate the presence of cluster volatility in the series, strengthening the argument of using the GARCH approach.

11

Journal Pre-proof The Jarque-Bera (J-B) test results confirm that most returns are not normally distributed; only the stock market returns of Japan exhibit a J-B statistic smaller than the critical value, and they are hence normally distributed. Non-normal distribution does not cause a problem for the estimation itself, but the estimations tend to be less efficient. The ARCH(12) test results indicate that the series are affected by heteroscedasticity and cluster volatility that further strengthens the argument to use the GARCH model for the estimations. Nonetheless, Hungary stock returns

of

exhibit insignificant ARCH(12) statistics, implying no heteroscedasticity. Only Finland and Japan suffer from autocorrelation due to their significant Q(12) Ljung-Box statistics.

pro

Insert Table 1

Figure 1 plots the log-level of the series. The stock market displays a positive trend. For the sample, we can identify a common pattern during the global financial crisis and the Dot-com bubble with peaks before the crises followed by a significant drop. The Zivot and Andrews

re-

(1992)’s structural break unit root test (Panel B of Table 2) reveals structural breaks for all countries during the Dot-com bubble.

To investigate further the presence of structural breaks, we perform an extended structural break

lP

test by Bai and Perron (2003); the results are illustrated in the Panel C of Table 2. The BaiPerron multiple structural break test is conducted with sequential estimations of multiple breaks. This sequential method is based on its robustness to misspecification in the number of breaks (Bai, 1997). We specify a model with a constant and a linear deterministic trend within

urn a

each segment defined by break dates, which minimized the Bayesian Information Criterion (BIC). The Bai and Perron multiple structural breaks display a common pattern during the global financial crisis in 2008, as we find a significant break date during the global financial crisis for all stock return series. These findings confirm the visual analysis of the stock market. From Figure 1, all series seem to be non-stationary in log-level, and they are therefore transformed to first differences, which are displayed in Figure 2. As illustrated in the first-

Jo

difference series, cluster volatility seems to appear. Cluster volatility means that a major change is followed by additional major changes, while a minor change is followed by additional minor changes, forming a cluster pattern in the series. This pattern further motivates the GARCH approach since it handles changes in volatility. Insert Figure 1 Insert Figure 2 12

Journal Pre-proof

We find evident outliers for all countries when plotting the standard residuals. From an economic perspective, it is significant to retain all outliers since they do describe market anomalies. The primary source of outliers appears during business cycle shocks, as the global financial crisis. Economic crises influence financial integration (Komatsubara et al., 2017). We employ the Augmented Dickey-Fuller (1979) and Phillips-Perron (1988) unit root tests to test

of

for stationarity. The testing procedure follows the Pantula principle. Panel A of Table 2 displays the ADF and PP for first differences; there is no indication of unit roots, and all first-differenced

pro

series are considered stationary since we reject the null hypothesis at the 5% significance level. For most series, the null hypothesis cannot be rejected, when combining the ADF and PP, indicating a unit root in log-level. For those series, where an indication of stationarity in the log-level is identified, another test for stationarity, the KPSS test, is applied1. Combining the stationarity tests with the visual trends of the series, we conclude that all series are stationary

re-

in first differences.

Insert Table 2

lP

The final preliminary analysis performed is the unconditional correlation matrix; it illustrates the correlation between the countries of the sample on Table 3. Inside the common currency group, French and Italian stock markets display a relatively high correlation between themselves. In comparison, Finland has lower correlations with the other two members,

urn a

although still displaying a relatively medium to high level towards the whole sample. Looking outside the common currency group, it displays mutually strong relationships with the European countries, and the US features for the group. Further, for the common currency group, relatively high correlations are found with geographically close countries. Within the sovereign currency group, most of the countries illustrate a similar pattern as the common currency group correlations with the European countries and the US. Hungary and Sweden have higher

Jo

correlations with European countries than with the economies outside Europe. In particular, Sweden has high correlations with the European countries. Norway displays a high correlation with the UK, followed by Sweden and France for the stock market. Hungary distinguishes themselves from the rest of the European countries with low correlations. Overall, the common currency countries as a group have higher correlations than the sovereign currency group.

1

The results are available upon request to the corresponding authors.

13

Journal Pre-proof Insert Table 3 4

Methodology

In this paper, we apply a DCC-EGARCH model as a fundamental benchmark to investigate the dynamic conditional correlation, as it is the most applicable method used to study financial integration, as seen in the previous literature. The EGARCH model developed by Nelson

of

(1991)2 is a development from the univariate GARCH model that allows for asymmetric effects of negative and positive shocks. The EGARCH estimations generate values for positive or adverse shocks, referred to as Theta-values in this paper. Therefore, the EGARCH, unlike other

pro

GARCH models, analyzes the predictions of how the series are affected by economic shocks. Our specifications of dynamic conditional correlations are based on the Engle (2002)’s model, where the conditional correlations matrix is time-dependent. These ARMA(p, q)-EGARCH(m, n) model specifications capture the asymmetries in the conditional volatility that may occur in

re-

the time-paths of return series.

We implement the cross-quantilogram model that was introduced by Linton and Whang (2007) to measure the predictability on different quantiles of the distribution in a stationary time series, based on the correlogram of “quantile hits” (Han et al., 2016). Linton and Whang (2007) applied

lP

a quantilogram to test for directional predictability and to examine the null hypothesis that a given time series has no directional predictability. The test for predictability is conducted by comparing the quantilogram with a pointwise confidence interval. Han et al. (2016) highlight several advantages for directional predictability of quantilogram compared with other tests.

urn a

Among the advantages, this method is based on “quantile hits” that do not require moment conditions like the ordinary correlogram, and it is applicable for series with strong tail dependence. Han et al. (2016) developed the univariate quantilogram framework by Linton and Whang (2007) to a multivariate setting to measure the quantile dependence between two stationary time series. The cross-quantilogram applies conditional quantiles to measure the directional dependence between two different time series after parsimoniously controlling for

Jo

the information at the prediction. Moreover, the test statistic for directional predictability is asymptotically valid uniformly over a range of quantiles. The cross-quantilogram captures the serial dependence between two series at different conditional quantiles, as expressed in Equation (1). Let {( 𝑦%& , 𝑦(& )* : 𝑡 ∈ ℤ } be two different 2

This methodology is not detailed here for brevity, but it can be found in details in ARCH (Engle,1982), GARCH (Bollerslev, 1986), EGARCH (Nelson, 1991), DCC (Engle, 2002), and subsequent contributions.

14

Journal Pre-proof series of stock market returns. We denote the conditional distribution of 𝑦0& given 𝑦1& as 𝐹34 (⋅ |𝑦1& ), for (𝑖, 𝑗) ∈ {1, 2} and 𝑖 ≠ 𝑗. Let 𝑞0,& (𝜏0 ) = inf@𝑢: 𝐹34 B𝑢C𝑦1& D ≥ 𝜏0 F be the 𝜏0 -conditional quantile of 𝐹34 (⋅ |𝑦1& ), with 𝜏0 ∈ (0, 1), for (𝑖, 𝑗) ∈ {1, 2} and 𝑖 ≠ 𝑗. The cross-quantilogram is a measure of the serial dependence between two events {𝑦%& ≤ 𝑞%,& (𝜏% )} and {𝑦(,&HI ≤ 𝑞(,&HI (𝜏( )} for an integer k and any pair of 𝜏 = (𝜏% , 𝜏( )* ∈ ℑ, with ℑ ≡ ℑ% × ℑ( and ℑ0 ∈

of

(0, 1), where ℑ denotes the range of quantiles we are interested in analyzing the directional predictability. The indicator function is denoted by 1[∙], and @1P𝑦0& ≤ 𝑞0,& (∙)QF is the “quantilehit” process for i = 1, 2. Let 𝜓S (𝑢) ≡ 1[𝑢 < 0] − 𝜏, then the cross-quantilogram is defined as 𝑞(,&HI (𝜏( )}Q for 𝑘 = 0, ±1, ±2, …, as follows:

𝐸 \𝜓S] ^𝑦%& − 𝑞%,& (𝜏% )_ 𝜓S` ^𝑦(,&HI − 𝑞(,&HI (𝜏( )_a b𝐸 \𝜓S(]

^𝑦%& − 𝑞%,& (𝜏% )_a b𝐸

\𝜓S(`

^𝑦(,&HI − 𝑞(,&HI (𝜏( )_a

.

(1)

re-

𝜌S (𝑘) =

pro

the cross-correlation of the “quantile-hit” processes 1P{𝑦%& ≤ 𝑞%,& (𝜏% )}Q and 1P{𝑦(,&HI ≤

The cross-quantilogram of Equation (1) captures serial dependence between two series at

lP

different quantile levels. Taking 𝜏 ≡ (𝜏% , 𝜏( )* = (𝜏def , 𝜏ghi )* as an example, 𝜌S (1) measures the cross-correlation between the Finnish market being below quantile 𝑞def (𝜏def ) at time 𝑡 and the German market being below quantile 𝑞ghi (𝜏ghi ) at time 𝑡 − 1. Therefore, 𝜌S (1) = 0 implies that whether the German market is above or below quantile 𝑞ghi (𝜏ghi ) at time 𝑡 does not help

urn a

predict whether the Finnish market will be above or below the quantile 𝑞def (𝜏def ) on the next period at time 𝑡 + 1. In contrast, if 𝜌S (1) ≠ 0 there exists directional predictability from Germany to Finland at 𝜏 = (𝜏def , 𝜏ghi )* .

Han et al. (2016) consider the quantile regression proposed by Koenker and Bassett (1978), * n based on the parametric conditional quantile functions 𝑞k%,& (𝜏% ) = 𝑥%& 𝛽% (𝜏% ) and 𝑞k(,& (𝜏( ) =

Jo

* n 𝑥(& 𝛽( (𝜏( ) as follows:

y

* 𝛽n% (𝜏% ) = argmin w 𝜚S] (𝑦%& − 𝑥%& 𝛽% ), s] t ℝv

&z%

(2) y

* 𝛽n( (𝜏( ) = argmin w 𝜚S` (𝑦(& − 𝑥(& 𝛽( ), s` t ℝv

&z%

15

Journal Pre-proof *

where 𝜚S4 (𝑢) ≡ 𝑢(𝜏0 − 1[𝑢 < 0]), 𝑥%& = B1, 𝑦%,&HI , 𝑦(& D ,

and

*

𝑥(& = B1, 𝑦%& , 𝑦(,&HI D ,

for

observations {(𝑦%& , 𝑦(& )}y&z% . Then, we estimate the cross-quantilogram of Equation (1) by calculating its sample analog, based on the estimated conditional quantile functions 𝑞k%,& (𝜏% ) = * n * n 𝑥%& 𝛽% (𝜏% ) and 𝑞k(,& (𝜏( ) = 𝑥(& 𝛽( (𝜏( ) obtained from Equation (2).

The cross-quantilogram method results in two outputs, quantile cross-correlation heatmaps and

of

rolling windows for different market regimes. In this paper, the results are specified with a lag length of one, a number of bootstrap iterations of 500, and a significance level of 0.05. A lag

pro

length of one indicates that we measure the correlations between two months. The selection of bootstrap iterations is evaluated with replications ranging from 100 to 1000. Based on robustness, where the results are consistent over time, 500 bootstrap replications are selected. The significance level of 0.05 is selected based on standard econometric arguments to prevent

re-

from rejecting a correct null hypothesis.

The quantile cross-correlations heatmap is based on calculating the average quantile crosscorrelations between countries. The interpretation of the quantiles is that lower quantiles, e.g. quantile 0.05, display adverse market conditions, quantiles in the middle represent normal

lP

market conditions, e.g. quantile 0.50, and higher quantiles indicate good market conditions, e.g. quantile 0.95. The interpretation of these results is based on a scale from dark red via white to dark blue, where dark red illustrates high positive correlation and dark blue represents high negative correlation. Interpreting market conditions influence is essential due to the

urn a

development of globalization and changing uncertainty in different market regimes. The rolling-window estimation is based on the average quantile cross-correlation during the window and the time-dependent change from one window to another. Hence, the rollingwindow-cross-quantilogram model examines the development of integration over time when controlling for the impact of market conditions. The selected rolling-window size is 48 months; the argument for this selection is based on political term periods that often extends 48 months.

Jo

The rolling-window figures also include confidence intervals; the interpretation of the results is that when the blue line is inside the confidence intervals the change is statistically significant. When the blue line is outside the confidence interval, the result is not statistically significant.

16

Journal Pre-proof 5.

Results and discussion 5.1 Exponential Generalized Autoregressive Conditional Heteroskedasticity

We apply the DCC-EGARCH3 approach to investigate the financial market integration. The primary advantage of the EGARCH, in comparison with other GARCH-type models, are the predictions of how the series are affected by economic shocks. This impact is measured by the Theta-values presented in Table 4. Theta 1 (θ1) indicates how one country’s volatility is affected

of

by a positive shock, whereas Theta 2 (θ2) represents how the same country’s volatility is affected by an adverse shock. Both values are expressed in absolute values and denote changes

pro

in volatility to a shock at time 𝑡 − 1, 𝜀&H% . If θ2 is greater than θ1, the country is more volatile during an adverse shock and thus the stock market returns, as displayed in Table 4, are more affected after an adverse shock than they would be if a positive shock occurred. Table 4 reports the results of the DCC-EGARCH estimations for the stock market. Panel A

re-

displays the mean equation for the univariate EGARCH, where the selected ARMA-order coefficients and significance levels are presented. The variance equation for the univariate EGARCH model is displayed in Panel B, where the ARCH and GARCH terms, as well as the Theta-values, are disclosed. Panel C shows the multivariate EGARCH Rho-value, which is

lP

interpreted as the aggregated correlation over the period. Finally, diagnostic tests for the univariate EGARCH and the multivariate DCC-EGARCH estimations are shown in Panel D. The diagnostic test interpretation is described in the sections above. For Italy, Hungary, Sweden, and the UK, the GARCH (1, 1) is the best-fitted model, whereas for Finland, France,

urn a

and Germany the GARCH (2, 2) is the best-fitted model. The GARCH (2, 1) is the best-fitted model for Norway.

France, Norway, Germany, and the UK display α + β values above one, indicating a non-mean reverting result for the univariate estimation. These results do not correspond to the multivariate estimation, where all α + β are below one, and the outcome may, therefore, depend on the criteria for the model specification. On Panel D, the Jarque-Bera test statistic indicates that all

Jo

stock market returns, except the UK, are affected by non-normality. In comparison with the normal distribution test results presented in the Panel B of Table 1, both the values of excess kurtosis and skewness demonstrate improvements. This implies that the EGARCH with a tStudent distribution is a better model than the standard Ordinary Least Squares (OLS). The

3

To find feasible ARMA processes, all orders up to ARMA(p, q) and GARCH(m, n) are performed and in those cases several ARMA orders are fitted; the best-fitted model is selected by minimizing the Akaike’s criterion (AIC) value.

17

Journal Pre-proof Ljung-Box test, Q2 (12), is rejected for all variables, suggesting that the univariate model is well adjusted to the data, and there is no serial correlation on the residuals. Insert Table 4 All θ1 values in Table 4 are negative, indicating a decrease in volatility after a positive shock for all stock market returns. However, some θ1 values are statistically non-significant, making

of

the interpretations uncertain. Finland stands out from the sample with the lowest absolute value for θ1, illustrating that this country is less affected by a positive shock. This conclusion cannot though be drawn due to a lack of significance. Overall, no clear pattern is found within each

pro

group, and thereby no group seems to be more affected by a positive shock than the other groups. The analysis of the θ2 values reports positive values for all countries, showing that the volatility increases after an adverse shock. As for the θ1 values, a lack of significance exists for additional countries so that no clear pattern is found within each group. Nevertheless,

re-

comparing the Theta-values of the European drivers, they seem to be more affected by an adverse shock than by a positive shock, since their θ2 values are higher than their θ1 values, according to absolute values. This pattern also occurs in Finland, Hungary, Italy, and Sweden. For France and Norway, the situation is the opposite; their values of θ1 are higher than that of

lP

θ2, indicating that these countries are more affected by a positive shock than by a negative one. Overall, these results indicate that an adverse shock is more influential than a positive one. Yet, a lack of significance makes the interpretations uncertain.

urn a

To conclude the theoretical framework, it would be reasonable for the common currency group to have similar Theta-values due to the criteria of homogeneity within the common currency group (Mundell, 1961). The characteristics of homogeneity can be expressed in several ways and thus do not need to express similar Theta-values. Instances where the common currency group shows lower Theta-values than the sovereign currency group are possible due to the possibility of risk sharing within the currency area, which should affect and decrease volatility

Jo

in adverse shocks (Mundell, 1973). Especially, as the stock market returns are referred to as financial variables, low Theta-values should occur for the common currency group according to Brada and Mendez (1988) and Hosny (2013). They argue that increased financial integration led to reduced risk and higher investments in the financial market. Previous studies, e.g., Yang et al. (2003a), Lafuente and Ordóñez (2009), Afonso and Sequeira (2010), and Rua (2010), find that euro countries have a higher level of financial integration within Europe, thus this could imply less volatility and lower Theta-values.

18

Journal Pre-proof 5.2 Dynamic Conditional Correlation The DCC-EGARCH estimations result in aggregated correlation values (Rho), Table 4, and time-dependent graphs that display the Rho-value over time (Figure 3). When interpreting the Rho-values, we have considered values in the 70th percentile and above as high correlations, in the 50th percentile and below as low correlations. These margins in percentile are deliberative, as we focus on a general impression when analyzing the results and determining high and low

of

correlations.

Table 4 shows the DCC-EGARCH results illustrated in the unconditional correlation matrix,

pro

indicating a high correlation with the European drivers for the common currency group. Further, the correlations towards Germany are higher than the correlations with the UK, for each one of the countries of the common currency group (Table 4, Panel C). Unreported t-tests of difference of means show that these differences between correlations with Germany and the UK are statistically significant for France and Italy. These results are also interpreted over time in

re-

Figure 3. Mundell (1961) partly explains this high degree of correlation. To begin with, we need to determine the function of the European drivers. Germany as an EU country is a member of the common currency area. According to Mundell (1961), an optimum currency area can be

lP

split into interregional and international areas. For the interregional area, a common currency is applied, which is the case of our common currency group and Germany. For the international area, the countries have sovereign currencies but do cooperate due to factor mobility within the area.

urn a

Because of the EU’s fundamental keystone, the idea of a single market with free movement of goods, services, capital, and labor, all EU countries can freely use each other’s services. According to Lipsey (1957), the consumption effects within a union such as the EU increase financial integration (Hosny, 2013; Marinov, 2014). This occurs in the common currency group but also in the UK, Hungary, and Sweden since they are part of the EU. An essential distinction is Mundell’s (1961) requirement of a fixed exchange rate within the optimum currency area.

Jo

Nevertheless, we argue that a high correlation towards the European drivers is applicable for the common currency group (Mundell, 1961 and 1973), especially towards Germany (Afonso and Furceri, 2009; Rua, 2010). However, our results depart from Yang et al. (2003a) and Lafuente and Ordóñez (2009), in the sense of showing high correlation towards the UK for the common currency group.

19

Journal Pre-proof 5.2.1 Common currency group – European drivers As previously stated, we find a general high level of correlation between the common currency countries and the European drivers, from which the integration is slightly higher towards the German stock market. Further, France and Italy display a higher correlation level than Finland. France reports the highest Rho-values, around 0.856, followed by Italy with 0.735, while Finland has values lower than 0.674. This diversity follows the analysis of Aguiar-Conraria and

of

Soares (2011), who show that core countries have higher correlation compared with peripheral countries.

pro

In comparison with France, Italy, and Germany, Finland did not join the EU during its establishment but rather decades later. Camacho et al. (2006) find higher integration in business cycles for the founder members of a currency area than for new members. Hence, the level of financial integration is lower for Finland in line with the concept of accounting for EU as a possible optimum currency area with interregional characteristics (Mundell, 1961). Conversely,

re-

all referred EMU countries did implement the euro as their currency during the same period, weakening the argument for Finland having a lower correlation. Nevertheless, in line with previous research, we consider the type of each economy, as a core or a peripheral country, to

lP

be more important for the level of correlation than the time within the currency area. In AguiarConraria and Soares (2011)’s paper, France is the common currency country with the highest characteristics of a core country, which partly can explain the constant trend in Figure 3 and the result of France overall expressing the highest Rho-values. While Finland and Italy have much

urn a

more fluctuations with a positive trend over time, France is stable between 0.85 and 0.9 with Germany and between 0.8 and 0.85 with the UK. Finland displays the highest correlations of the common currency countries during the implementation of the euro as a common currency during 2002, when interpreting the timedependent Rho-values and their trends in Figure 3. Right after 2000 to 2001, Finland had a low correlation lower than 0.6, but it experienced an increasing correlation with the European

Jo

drivers during and after the euro implementation. Again, this could depend on France and Italy being more homogeneous with Germany, as core countries, than on Finland (Aguiar-Conraria and Soares, 2011) as well as on the time of entry into the EU (Mundell, 1961; Camacho et al., 2006). An interesting development that occurs in Figure 3 for the common currency group is that the post-financial crisis period is characterized by a peak in correlations, after which the correlations progress in a similar path to that of Germany and the UK. Considering the global financial crisis in 2008 as an adverse shock, the common currency countries have the advantage 20

Journal Pre-proof of risk sharing (Mundell, 1973), thus affecting the volatility and correlations similarly as a natural process. Insert Figure 3 5.2.2 Sovereign currency group – European drivers The sovereign currency group presents a similar pattern of high correlations towards both

of

Germany and the UK as the common currency group did. Lower Rho-values, both aggregated and time-dependent, are though observed for the sovereign currency group in comparison with the common currency group, as the mean values with Germany and the UK are around 0.635

pro

and 0.696, respectively. The lower correlation towards Germany (a member of the common currency group) is in line with Mundell (1961), who advocates the sovereign currency countries’ role as international and not interregional players in the EU. A lower level of correlation with Germany also indicates similar results to those presented in Rose and Engel

re-

(2002).

Sweden distinguishes themselves with high correlation towards the European drivers, and it displays a higher level of correlation, around 0.756, than both Finland and Italy as common

lP

currency countries. Sweden cannot be characterized as a core country; instead, a general high level of homogeneity between Sweden and the European drivers with an extent of trade may be an argument for high Rho-values. On the other hand, Hungary reports a medium-low correlation with the European driver, around 0.560. With Mundell (1961) as a benchmark, Norway is more

urn a

likely to possess lower level of correlations than Hungary and Sweden. Norway is not part of EU, and hence it cannot be viewed as an international country in an optimum currency area. Instead, Norway is somewhere between Hungary and Sweden with Rho-values around 0.682. When interpreting the time-dependent Rho-values in Figure 3, a similar pattern arises within the sovereign currency group, which peaks in correlation around the boom of the global financial crisis in 2008 followed by a steady decrease.

Jo

5.2.3 Correlations with the global countries The correlations with the global countries show that Japan and the US display higher correlations with the common currency group and the European drivers than with the sovereign currency group. This could be due to the group’s different roles in the global economy. The euro is one of the leading currencies in the world that could influence the common currency group; further, Germany’s correlation with the global countries, who also have strong currencies (Kim et al., 2013), could be another contributing factor. In addition, Pérez-Rodríguez 21

Journal Pre-proof (2006) finds that the euro and the British pound exchange rates are highly correlated with the US dollar. Trade alliances such as the Transatlantic Trade and Investment Partnership are possible explanations for a higher level of correlation of all groups with the US (Baldwin, 2008), since all European countries (except Norway) are part of the EU. As the correlations with the US can be characterized as rather high, for the European drivers around 0.743, for the common currency

of

group around 0.636, and for the sovereign currency group around 0.615, the correlations with Japan are considered as low ones. This suggests that the role of the US, as the global financial

pro

leader, is more influential on the stock market than Japan. The mean Rho-value of the European drivers with Japan is around 0.472, while the common currency group and the sovereign currency group display mean values around 0.457 and 0.406, respectively. An increasing correlation in the aftermath of the global financial crisis is associated with the

re-

global countries, when interpreting the values of time-dependent Rho in Figure 3. The financial crisis in 2008 burst out in the US and then became global through Europe. Since we live in a highly connected market with great mobility and integration (Dreher, 2006), events like the global financial crisis are likely to affect the global economy in multiple directions. We can

lP

also see an increase in the correlation for all groups towards the emerging countries after the financial crisis. This increase in correlation does not continue, reaching a peak for both global and emerging markets during the post-crisis period. Besides, the argument of higher correlations after the global financial crisis can be strengthened by higher θ2 values. Higher

urn a

values of θ2 occur consequently for the European drivers, as previously explained, implying that the volatility of Germany and the UK are more affected by a negative shock since the start of the financial crisis.

5.2.4 Correlations with the emerging countries In comparison with the correlations with the global markets, the correlations with the emerging

Jo

markets are lower. All groups have a mean Rho-value of 0.29, indicating that the level of correlation is independent of a country being part of the EMU towards the emerging market. Even China shows considerably lower correlations, with no country expressing a Rho-value greater than 0.150, as presented in the Panel C of Table 4, whereas the correlations are on a similar level for India as for Japan. Historically, China and India have not been as developed as the other countries in our sample, and they are not homogeneous towards the rest of the sample, which is an important criterion for the level of economic and financial integration (Mundell,

22

Journal Pre-proof 1961). Developing countries have further not experienced risk sharing through globalization according to Kose et al. (2009). As evident in our cross-quantilogram estimations below, financial market integration increases in general during adverse market conditions. It is also during adverse market conditions (such as after an adverse shock) that the incentives of risk sharing are at their highest level. Thus, lower correlations with the emerging countries, in contrast to the other developed countries, are

of

reasonable. According to studies such as Johansson (2011), He and Liao (2012), and Komatsubara et al. (2017), the emerging countries are more characterized as developing

pro

countries. This implies that the changing economic conditions in the emerging market are not yet visible, even if we look at the time-dependent Rho and its development (Figure 3). Nevertheless, Figure 3 shows that the trend of correlation of the emerging countries with other countries is reminiscent with the trend with global markets. Therefore, for a global market where countries are connected and influenced by business cycles, the level of connectedness

re-

differs as evident by the results of lower correlations for the emerging countries and the highest level of correlation with the US and the European drivers. 5.2.5 Summary of the DCC-EGARCH analysis

lP

To summarize the DCC results, we find the highest correlation levels with the European drivers followed by the US as a global country. As for the European drivers, we find a higher level of integration for the common currency group than for the sovereign currency group. In addition, towards the global countries, the common currency group together with the European drivers for all groups.

urn a

display the highest correlations. The lowest correlations appear towards the emerging countries

The degree of correlation depends on characteristics such as the level of homogeneity (Mundell, 1961), risk sharing (Mundell, 1973), and factor mobility (Mundell, 1961), thus, making globalization important for financial market integration. In addition, the geographical proximity

Jo

seems to influence the level of financial integration. Our results are analogous to AguiarConraria and Soares (2011), who find that geographically close countries have more synchronized business cycles. A final reason as to why the financial market integration within the EMU is significantly higher is the elimination of exchange rate volatility (Fratzscher, 2002). This result does not indicate an alteration in the correlation of the European drivers with the global- and emerging markets, implying no signs of preparation of the UK for leaving the EU.

23

Journal Pre-proof 5.3 Cross-quantilogram Based on the cross-quantilogram estimations, we can interpret an aggregated and timedependent correlation depending on different market regimes. The results of the crossquantilogram estimations are displayed in heatmaps (Figure 4) and in rolling-window estimations (Figure 5). 5.3.1

Common currency group – European drivers

of

The common currency countries overall display a more distinct pattern with higher correlations with Germany than with the UK, according to the DCC-EGARCH estimations. When

pro

examining the cross-quantilogram estimations, we interpret that the bad market condition is influential for increasing correlations with the common currency group. The common currency group alone experiencing a bad market condition, below 0.1, tends to increase the financial market integration, both with Germany and with the UK. The rolling-window estimations overall demonstrate an increasing influence of normal market conditions on the increasing

re-

integration, with an average around 0.15, with Germany and the UK.

Along with the results displayed in Table 4, the cross-quantilogram and rolling-window estimations indicate that the increased volatility of the θ2 values corresponds with increasing

lP

financial market integration during bad market conditions. Hence, an adverse shock, and experiencing a bad market condition, induces the EMU (common currency) countries to have a higher level of financial market integration within the currency area. We find a change in correlation during bad market conditions with Germany that depends on Germany’s

urn a

membership in EMU. Further, this effect is stronger for an adverse shock and under a bad market regime than for a positive shock and under a good market regime. Therefore, it could be interesting to examine whether one of the greatest benefits of joining a currency area is the

5.3.2

Jo

higher financial market integration, which can be viewed as an insurance against recessions.

Insert Figure 4

Sovereign currency group – European drivers

The correlations for the sovereign currency group with the European drivers are not as uniform as they are for the common currency group. This difference may not be surprising according to Mundell (1961), since the common currency group should be more homogeneous as a group than the sovereign currency countries. The sovereign currency countries do increase the correlation with Germany, in general, when Germany experiences good market conditions and 24

Journal Pre-proof the sovereign currency countries experience bad market conditions. The rolling-window estimations show a general increase in financial market integration with Germany during a bad market condition for the sovereign currency group. Similarly, the correlation with the UK increases depending on the sovereign currency group experiencing a bad market condition. The results for Sweden, an EU country (with a sovereign currency), differ from those for Norway, which is not an EU country (but it has a sovereign currency). The DCC-EGARCH

of

results suggest that the correlation between Norway and Germany are positive and stable over time (Panel E of Figure 3). Nevertheless, Panel E of Figure 4 shows that this correlation is

pro

negative when Germany faces a bad market condition and Norway is under a good market condition. Further, there is increasing financial market integration between Norway and Germany, when Norway is under a bad market condition, especially after the great financial crisis in 2008 (Panel E of Figure 5). We observe a similar pattern of financial integration

re-

between Norway and the UK.

On the other hand, the correlation between Sweden and Germany is positive and stable after the implementation of the euro in 2002, regardless of Sweden experiencing bad or moderate market conditions (Panel F of Figure 5). The correlation between Sweden and the UK displays

lP

a similar pattern. Therefore, the CQ analysis demonstrates that being part of the EU provides a stable financial market integration with the European drivers as an EU country (Sweden) presents a higher and more stable correlation than a non-EU country (Norway).

urn a

When the UK experiences a good market condition, it is reasonable to assume that they pull Europe’s economy as a financial driver, increasing the integration through the common market. The heatmaps show a general increase in correlation when the countries experience bad market conditions, emphasizing the UK’s role as a financial driver; nevertheless, this increase is not comprehensive and is displayed at different market conditions for the UK with different countries.

Correlations with the global countries

Jo

5.3.3

The correlation with the US increases when the common currency group, the sovereign currency group, and the European drivers experience bad market conditions (Figure 4). These findings are confirmed by the rolling-window estimations (Figure 5). For the common currency group, notably with the exception of France and Italy uniform performers, we find an average increase in correlation of 0.1. Homogeneity in geography or the fact that France and Italy are more core countries than Finland (Aguiar-Conraria and Soares, 2011) could be the reason 25

Journal Pre-proof behind this, as discussed for the DCC-EGARCH results. Especially, when these core countries face good market conditions, signs of decreasing financial market integration arise for the stock market when the US is below its normal market condition. With Germany and the UK being core countries as European drivers, signs of decreasing financial integration were evident. Nevertheless, the correlations increase when the drivers experience bad market conditions and the US experiences normal or good market conditions.

of

As in the results of the DCC-EGARCH analysis, all groups report lower correlations with Japan than with the US. According to the cross-quantilogram estimations, the overall financial market

pro

integration is weak with Japan and other groups, regardless of the market regimes. France is an exception, showing an increasing correlation with Japan during normal market conditions. Nevertheless, the rolling-window estimations between France and Japan do not indicate an increase in correlation during normal market conditions.

re-

Interpreting the heatmaps for the European drivers, we find weak signs of change in integration regardless of the market conditions of Japan. Only Germany displays market-dependent integration with Japan. This change in integration occurs in normal market conditions for Germany and in bad market regimes for Japan. A weak increase in correlation (from 0.1 to 0.2)

lP

appears for the post-euro period and onwards, when interpreting the rolling-window estimations of normal market conditions between Germany and Japan. The significant difference in the correlation with the global countries can partly be explained through their different roles in the global market. The US could be viewed as a global market leader, which is highly integrated

urn a

with the European market through globalization (Dreher, 2006); conversely, Japan may be more integrated with the Asian market than with the European one, according to previous studies such as Berdiev and Chang (2015). Another explanation could be the higher level of similar characteristics (Mundell, 1961) of the US and the European countries as they all are part of the Western world and hence share similar political context and institutions. Correlations with the emerging countries

Jo

5.3.4

The DCC-EGARCH estimations indicate low levels of integration for all groups and countries with the emerging countries. From the cross-quantilogram analysis, we cannot interpret influence or consensus regarding market regimes for the correlations between or within the groups. Combining the results of the DCC-EGARCH and the cross-quantilogram estimations, we find our groups to be less financially integrated with the emerging countries than with the rest of the countries. Different results in the cross-quantilogram and the rolling-window estimations indicate uncertain interpretations without a clear linkage of financial integration. 26

Journal Pre-proof The correlations with China reveal a strong increase when the European drivers and China are facing normal to good market conditions. The rolling-window estimations do not yield the same clear picture, as they show an increase in correlations over time when China experiences normal market conditions. The correlations with India increase when the market conditions are good for the group and normal for India, while the correlations with China are higher when the European drivers are in the upper-tail quantiles and China in the lower quantiles. The rolling-

of

window estimations support these results for India, where after the global financial crisis there is evidence of an increasing correlation in normal and good market conditions. For China, this

pro

increase is absent in the rolling-window estimations. Insert Figure 5 5.3.5

Summary of the cross-quantilogram analysis

The cross-quantilogram results reveal that different market regimes affect the correlation

re-

development differently. The DCC-EGARCH results in Table 4 and Figure 3 display high correlations between France and Italy with Germany, which are stable over time. Nevertheless, the CQ results of Figures 4-5 show that the financial market integration between the common currency group and Germany increases during bad market conditions. This relationship is also

lP

shown for the UK. The higher influence of bad market conditions may depend on Germany’s and the UK’s function as drivers of the European economy. In particular, when bad market conditions appear, for example, during a crisis, the common currency countries resort to the drivers to get economic pace. This analysis also helps explain why the sovereign currency

urn a

countries increase their financial market integration with the UK, when experiencing bad market conditions.

In addition, the CQ analysis illustrates that the financial integration between the sovereign group and the European drivers depends on market conditions. The correlations between the sovereign group and the European drivers are stable and positive in the DCC-EGARCH

Jo

estimations presented in Table 4 and Figure 3. Conversely, the CQ analysis of Figures 4-5 shows an increase in financial market integration with Germany during a bad market condition for the sovereign currency group. Similarly, the correlations with the UK increase depending on the sovereign currency group experiencing a bad market condition. Further, the CQ results highlight that being part of the EU provides a stable financial market integration with the European drivers as an EU country (Sweden) presents a higher and more stable correlation than a non-EU country (Norway).

27

Journal Pre-proof The CQ results demonstrate that bad market conditions for the groups increase the stock market integration with the US. The DCC-EGARCH rolling-window estimations indicate a stable and strong correlation between the US and almost all countries, except for Hungary and the UK. Nevertheless, the CQ analysis suggests that the correlations between the US and each one of the countries may become negative when the US faces a bad market condition (Figure 4). Further, normal market conditions seem to increase the integration with Japan and some

of

countries.

The impact of market regimes on financial integration for the emerging markets display a

pro

different picture. The emerging market is less affected overall by market conditions compared with the European drivers and the Global group. Some correlations show that bad market conditions decrease the level of integration, while this is an absent characteristic in the other correlations. Our findings indicate that countries with more homogeneous characteristics that also have a higher degree of financial market integration, namely the European countries with

re-

the US, are more affected in bad conditions. Conversely, the emerging countries, which did not show high correlations in the DCC-EGARCH analysis, display a diverse impact of market regimes. These results provide evidence of the effects of globalization (Kose et al., 2009) on

lP

market regimes and on financial market integration.

5.4 Robustness check: Poland and the Czech Republic As a robustness check, we analyze two economies that joined the EU during our sample period with a sovereign currency. We aim at drawing some suggesting evidence to disentangle EU

urn a

membership from common currency. We examine the stock market indices of Poland and the Czech Republic, which are major Eastern European economies that joined the EU in May 2004. Both countries have a sovereign currency inside the EU. We use the Warsaw Stock Exchange WIG Total Return Index (WIG) and the Prague Stock Exchange Index (PE) for the stock market indices of Poland and the Czech Republic, respectively. The sample period of Poland spans March 1993 to September 2017, and the sample period of the Czech Republic spans April 1994

Jo

to September 2017. We employ fewer observations due to data availability on the stock markets of these two countries.

We also gather the stock market indices of Poland and the Czech Republic in their local currency and convert them into US dollars. Unreported unit root tests on the log-level of both series find that they are nonstationary, whereas the first-differenced indices (returns) are stationary at the 5% level. Thus, we analyze the returns of the stock market indices of Poland and the Czech Republic. We first fit a DCC-EGARCH model to both stock market returns. The 28

Journal Pre-proof GARCH(1,1) is the best-fitted model for Poland and the Czech Republic, as it is for Hungary. Unreported Ljung-Box Q2(12) tests show that there is no serial correlation on the residuals of the selected EGARCH models. Figure A.1 in the Appendix reports the time-dependent Rho-value of the DCC-EGARCH estimations for Poland and the Czech Republic. In line with the results for the sovereign group presented in Figure 3, both countries present high correlations towards Germany and the UK,

of

although their Rho-values are lower than the ones of the common currency group with the European drivers. Poland and the Czech Republic have mean Rho-values with Germany of

pro

0.500 and 0.402, respectively. Similar to Hungary (with a mean Rho-value of 0.560), these two countries display medium-low correlations with the European driver.

In addition, towards the global countries, the common currency group together with the European drivers report higher correlations than the ones of Poland and the Czech Republic, in

re-

consonance with the results for the sovereign group displayed in Figure 3. The Rho-values of Poland and the Czech Republic also have a similar pattern to the ones of the sovereign currency group over time, peaking in correlation around the boom of the global financial crisis in 2008

lP

followed by a steady decrease.

Figure A.2 in the Appendix presents the CQ heatmaps, and Figure A.3 shows the CQ rollingwindow estimations for Poland and the Czech Republic. Consistent with the results presented in Figure 5, the correlations of Poland and the Czech Republic with the European drivers are

urn a

not as uniform as they are for the common currency group. Poland and the Czech Republic do increase the correlation with Germany, when Germany experiences good market conditions and these two sovereign countries face bad market conditions. In addition, the correlations between the US and each one of these two countries are not significant when the US is under bad market conditions, similarly to the correlations between Norway and the US. The rolling-window estimations highlight a general increase in financial market integration

Jo

with Germany during a bad market condition for the Czech Republic. Similarly, the correlation between the Czech Republic and the UK increases depending on the Czech Republic experiencing a bad market condition. Nevertheless, Poland displays a stable pattern of low rolling-window correlations with Germany and the UK, which is robust to different market conditions.

29

Journal Pre-proof 6

Conclusion

Our paper employs the recently developed cross-quantilogram (CQ; see Han et al., 2016) approach to examine quantile dependence between the conditional distributions of stock returns of Germany and the UK with that of three common-currency groups within EMU (Finland, France, and Italy), two global leading markets (the US and Japan), and the two most promising emerging markets (China and India). As opposed to previous studies that are mostly based on

of

linear correlation or causality modelling, we apply CQ measures to detect and identify the nature of dependence effects amongst Eurozone and the broad Euro area business cycles. This

pro

approach utilizes a model-free measure of correlation between two variables across the quantiles of each distribution, and it examines the dependence under specific market circumstances, for example in bearish (lower-quantile), bullish (upper-quantile), and normal (intermediate-quantile) markets that other models (e.g., switching model, DCC-GARCH-type model, and quantile regression) are unable to capture.

re-

We find three key results. First, both the EU membership and the common currency union affect the degree of financial market integration. Nevertheless, disentangling the effects of EU membership from the common currency indicates that the common currency group has an

lP

additional impact on financial integration, as the degree of dependence is stronger in the common currency group than in the sovereign currency group and other groups. Besides, there is a heterogeneous dependence structure, which is strongly observed for the UK and German stock returns with that of developed (the US and Japan) and emerging markets (India and

urn a

China). This is mainly due to country-specific differences and asymmetries of trade and fiscal policies. Finally, cross-quantile correlations change over time, especially in low and high quantiles, indicating that they are prone to jumps and discontinuities in the dependence structure. Thus, the financial market integration increases during economic and financial turbulence periods. This implies that common policy instruments (such as active monetary policy by the ECB and migration policy by EMU) could provide an effective stabilization

Jo

instrument for the entire Euro area.

We contribute to the literature by providing a more comprehensive understanding of the impact the euro has had on financial integration with economies of different characteristics outside and within the European market, by considering the influence of market conditions on the level of financial market integration. To the best of our knowledge, this is the first study in this field employing a cross-quantilogram method to examine the impact of different market conditions on the correlations, making our study a pioneer in the field of financial market integration. 30

Journal Pre-proof Therefore, we fill a gap in the existing literature by combining the aggregated and timedependent approaches that include the market regimes. The empirical results illustrating the implication of an “active monetary policy” by the ECB and “migration policy” by EMU due to increased dependence via market connectedness under the different economic conditions (normal vs crisis) could provide important inferences to develop effective policy instruments for the ECB and EMU. As the dependence structure

of

mechanism varies across the different market and economic conditions (stronger turbulence times), it might be effective to consider an optimal monetary policy and a common migration

pro

policy to accommodate for the time-varying asymmetric nature of business cycles. References

Afonso, A. & Furceri, D., 2009. Sectoral business cycle synchronization in the European Union. Economics Bulletin, 29(4), pp. 2996-3014.

re-

Afonso, A. & Sequeira, A., 2010. Revisiting business cycle synchronisation in the European Union, s.l.: Working Paper 22/2010/DE/UECE. Aguiar-Conraria, L. & Soares, M. J., 2011. Business cycle synchronization and the Euro: A wavelet analysis. Journal of Macroeconomics, 33(3), pp. 477-489.

lP

Bai, J., 1997. Estimating multiple breaks one at a time. Econometric Theory, 13(3), pp. 315352. Bai, J. & Perron, P., 2003. Computation and analysis of multiple structural change models. Journal of Applied Econometrics, 18(1), pp. 1-22.

urn a

Balassa, B. A., 1962. The Theory of Economic Integration. London: Richard D. Irwin. Baldwin, R., 2008. Big-think regionalism: A critical survey, s.l.: NBER working paper series: Working Paper 14056. Bekiros, S., Nguyen, D. K., Uddin, G. S. & Sjö, B., 2015. Business cycle (de)synchronization in the aftermath of the global financial crisis: Implications for the Euro area. Studies in Nonlinear Dynamics & Econometrics, 19(5), pp. 609-624. Berdiev, A. N. & Chang, C.-P., 2015. Business cycle synchronization in Asia-Pacific: New evidence from wavelet analysis. Journal of Asian Economics, Volume 37, pp. 20-33.

Jo

Bollerslev, T., 1986. Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31(3), pp. 307-327. Brada, J. C. & Mendez, J. A., 1988. An estimate of the dynamic effects of economic integration. The Review of Economics and Statistics, 70(1), pp. 163-168. Büttner, D. & Hayo, B., 2011. Determinants of European stock market integration. Economic Systems, 35(4), pp. 574-585. Camacho, M., Perez-Quiros, G. & Saiz, L., 2006. Are European business cycles close enough to be just one? Journal of Economic Dynamics & Control, 30(9-10), pp. 1687-1706. 31

Journal Pre-proof Canova, F., & De Nicolo, G., 1995. Stock returns and real activity: A structural approach. European Economic Review, 39(5), pp. 981-1015. Dickey, D. A. & Fuller, W. A., 1979. Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association, 74(366a), pp. 427431.

of

Diebold, F. X. & Yilmaz, K., 2014. On the network topology of variance decompositions: Measuring the connectedness of financial firms. Journal of Econometrics, 182(1), pp. 119-134. Dreher, A., 2006. Does globalization affect growth? Evidence from a new index of globalization. Applied Economics, 38(10), pp. 1091-1110.

pro

Eickmeier, S. & Breitung, J., 2006. How synchronized are new EU member states with the euro area? Evidence from a structural factor model. Journal of Comparative Economics, 34(3), pp. 538-563. Engle, R., 2002. Dynamic conditional correlation: A simple class of multivariate generalized autoregressive conditional heteroskedasticity models. Journal of Business & Economic Statistics, 20(3), pp. 339-350.

re-

Engle, R. F., 1982. Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica, 50(4), pp. 987-1007.

lP

European Commission, 2018. Trade agreements. [Online] Available at: https://ec.europa.eu/info/business-economy-euro/trade-non-eucountries/trade-agreements_en [Accessed 22 April 2018]. European Union, 2018. One market without borders: European Union. [Online] Available at: https://europa.eu/european-union/topics/single-market_en [Accessed 22 April 2018].

urn a

Fratzscher, M., 2002. Financial market integration in Europe: On the effects of EMU on stock markets. International Journal of Finance & Economics, 7(3), pp. 165-193. Han, H., Linton, O., Oka, T. & Whang, Y.-J., 2016. The cross-quantilogram: Measuring quantile dependence and testing directional predictability between time series. Journal of Econometrics, 193(1), pp. 251-270. Hartmann, P., Maddaloni, A. & Manganelli, S., 2003. The Euro-area financial system: Structure, integration, and policy initiatives. Oxford Review of Economic Policy, 19(1), pp. 180-213.

Jo

He, D. & Liao, W., 2012. Asian business cycle synchronization. Pacific Economic Review, 17(1), pp. 106-135. Herd, R. & Dougherty, S., 2007. Growth prospects in China and India compared. The European Journal of Comparative Economics, 4(1), pp. 65-89. Hosny, A. S., 2013. Theories of economic integration: A survey of the economic and political literature. International Journal of Economy, Management and Social Sciences, 2(5), pp. 133-155. House of Commons Parliament, 2016. Brexit: impact across policy areas, s.l.: House of Commons Library: Briefing Paper. 32

Journal Pre-proof Johansson, A. C., 2011. Financial markets in East Asia and Europe during the global financial crisis. The World Economy, 34(7), pp. 1088-1105. Johnson, R. & Soenen, L., 2002. Asian economic integration and stock market comovement. The Journal of Financial Research, 25(1), pp. 141-157. Kim, B.-H., Kim, H. & Min, H.-G., 2013. Reassessing the link between the Japanese yen and emerging Asian currencies. Journal of International Money and Finance, Volume 33, pp. 306-326.

of

Koenker, R. & Bassett, Jr., G., 1978. Regression quantiles. Econometrica, 46(1), pp. 33-50.

pro

Komatsubara, T., Okimoto, T. & Tatsumi, K.-i., 2017. Dynamics of integration in East Asian equity markets. Journal of the Japanese and International Economies, Volume 45(C), pp. 37-50. Kose, M. A., Prasad, E. S. & Terrones, M. E., 2009. Does financial globalization promote risk sharing? Journal of Development Economics, 89(2), pp. 258-270. Ku, J. & Yoo, J., 2013. Globalization and sovereignty. Berkeley Journal of International Law, 31(1), pp. 210-234.

re-

Lafuente, J. A. & Ordóñez, J., 2009. The effect of the EMU on short and long-run stock market dynamics: New evidence on financial integration. International Journal of Financial Markets and Derivatives, 1(1), pp. 75-95. Lane, P. R. & Milesi-Ferretti, G. M., 2008. The drivers of financial globalization. American Economic Review, 98(2), pp. 327-332.

lP

Linton, O. & Whang, Y.-J., 2007. The quantilogram: With an application to evaluating directional predictability. Journal of Econometrics, 141(1), pp. 250-282. Lipsey, R. G., 1957. The theory of customs unions: Trade diversion and welfare. Economica, New Series, 24(93), pp. 40-46.

urn a

Marinov, E., 2014. Economic Integration Theories and the Developing Countries, s.l.: MPRA Paper No. 63310. Moneta, F. & Rüffer, R., 2009. Business cycle synchronization in East Asia. Journal of Asian Economics, 20(1), pp. 1-12. Mundell, R. A., 1961. A theory of optimum currency areas. The American Economic Review, 51(4), pp. 657-665. Mundell, R. A., 1973. Uncommon arguments for common currencies. In: H. G. Johnson & S. A. K., eds. The Economics of Common Currencies. Oxon: Routledge, pp. 114-132.

Jo

Nelson, D. B., 1991. Conditional heteroskedasticity in asset returns: A new approach. Econometrica, 59(2), pp. 347-370. Park, Y. C. & Shin, K., 2009. Economic integration and changes in the business cycle in East Asia: Is the region decoupling from the rest of the world? Asian Economic Papers, 8(1), pp. 107-140. Pérez-Rodríguez, J. V., 2006. The euro and other major currencies floating against the U.S. dollar. Atlantic Economic Journal, 34(4), pp. 367-384.

33

Journal Pre-proof Phillips, P. C. B. & Perron, P., 1988. Testing for a unit root in times series regression. Biometrika, 75(2), pp. 335-346. Phylaktis, K., & Ravazzolo, F., 2002. Measuring financial and economic integration with equity prices in emerging markets. Journal of International Money and Finance, 21(6), pp. 879903. Quah, C.-H. & Crowley, P. M., 2012. Which country should be the monetary anchor for East Asia: the US, Japan or China? Journal of the Asia Pacific Economy, 17(1), pp. 94-112.

of

Rivera-Batiz, L. A. & Romer, P. M., 1991. Economic integration and endogenous growth. The Quarterly Journal of Economics, 106(2), pp. 531-555.

pro

Roll, R., 1992. Industrial structure and the comparative behavior of international stock market indices. The Journal of Finance, 47(1), pp. 3-41. Rose, A. K. & Engel, C., 2002. Currency unions and international integration. Journal of Money, Credit and Banking, 34(4), pp. 1067-1089. Rua, A., 2010. Measuring comovement in the time-frequency space. Journal of Macroeconomics, 32(2), pp. 685-691.

re-

Schwert, G. W., 1990. Stock returns and real activity: A century of evidence. The Journal of Finance, 45(4), pp. 1237-1257. Sethapramote, Y., 2015. Synchronization of business cycles and economic policy linkages in ASEAN. Journal of Asian Economics, 39, pp. 126-136.

lP

Wang, P. & Wen, Y., 2007. Inflation dynamics: A cross-country investigation. Journal of Monetary Economics, 54(7), pp. 2004-2031. Yang, J., Kolari, J. W. & Min, I., 2003b. Stock market integration and financial crises: The case of Asia. Applied Financial Economics, 13(7), pp. 477-486.

urn a

Yang, J., Min, I. & Li, Q., 2003a. European stock market integration: Does EMU matter? Journal of Business Finance & Accounting, 30(9-10), pp. 1253-1276.

Jo

Zivot, E., Andrews, D.W.K., (1992). Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis. Journal of Business & Economic Statistics, 10(3), 251–270.

34

Journal Pre-proof Table 1. Stock market indices and summary statistics Panel A: Stock market indices Indices

Mnemonics

Period

United Kingdom

UK-DS Market $ - Tot Price Index

TOTMUK$(PI)

1993M01-2017M09

Germany

DAX 30 Performance - Tot Price Index

DAXINDX(PI)~U$

1993M01-2017M09

Japan

TOPIX - Tot Price Index

TOKYOSE(PI)~U$

1993M01-2017M09

United States

S&P 500 Composite - Tot Price Index

S&PCOMP (PI)~U$

1993M01-2017M09

China

Shenzhen Se Composite - Price Index

CHZCOMP(PI)~U$

1993M01-2017M09

India

S&P BSE (100) National - Price Index

IBOMBSE(PI)~U$

1993M01-2017M09

Common

France

France-DS Market - Tot Price Index

TOTMKFR(PI)~U$

1993M01-2017M09

Currency Union

Finland

OMX Helsinki - Tot Price Index

HEXINDX(PI)~U$

1993M01-2017M09

(CCU)

Italy

Italy-DS Market $ - Tot Price Index

TOTMIT$(PI)

1993M01-2017M09

Non-common

Sweden

OMX Stockholm - Price Index

SWSEALI(PI)~U$

1993M01-2017M09

Currency Union

Norway

Norway-DS Market $ - Tot Price Index

TOTMNW$(PI)

1993M01-2017M09

Hungary

Budapest Price Index

BUXINDX(PI) ~U$

1993M01-2017M09

Emerging

(NCCU)

pro

Global

re-

Europe

of

Country

Notes: Country specific stock indices, index identification mnemonics, and sample period.

Median

Std. Dev.

CHN

0.0062

0.0080

FIN

0.0085

0.0107

FRA

0.0051

0.0083

GER

0.0070

0.0109

HUN

0.0093

IND

0.0063

ITA

0.0028

JPN

0.0012

NOR

0.0065

SWE

0.0078

UK

0.0035

Kurtosis

J-B

ARCH(12)

Q(12)

0.1083

-0.131

7.455

245.573***

8.227***

1.116

0.0794

-0.350

4.726

42.759***

3.415***

1.838**

0.0569

-0.695

4.417

48.567***

3.543***

1.222

0.0660

-0.832

5.397

104.954***

2.806***

0.815

0.0157

0.1001

-0.754

7.755

306.903***

0.779

0.719

0.0120

0.0866

-0.328

4.203

23.160***

2.869***

0.528

0.0036

0.0683

-0.267

3.675

9.128***

2.226***

1.195

0.0035

0.0531

-0.032

3.422

2.249

2.162**

2.403***

0.0111

0.0733

-1.259

7.830

365.887***

4.121***

1.353

0.0109

0.0669

-0.611

5.067

71.110***

2.976***

1.479

0.0052

0.0463

-0.725

5.423

98.304***

5.012***

0.839

0.0110

0.0416

-0.878

4.961

85.390***

3.127***

0.931

Jo

US

Skewness

urn a

Mean

lP

Panel B: Summary statistics

0.0059

Notes: All series used in Table 1 are in first differences. The notation *, **, and *** in the Jarque-Bera (J-B) test, ARCH(12), and Q(12) indicates rejection of the null hypothesis at the 10%, 5%, and 1% significance level, respectively. CHN denotes China, FIN denotes Finland, FRA denotes France, GER denotes Germany, HUN denotes Hungary, IND denotes India, ITA denotes Italy, JPN denotes Japan, NOR denotes Norway, SWE denotes Sweden, UK denotes the United Kingdom, and US denotes the United States of America. Sample period: January 1993 – September 2017. Source: Thompson Reuters DataStream.

35

Journal Pre-proof Table 2. Unit root tests Panel A: ADF and PP tests Phillips–Perron (PP) test

ADF(𝛿) -16.155 (0) *** -8.905 (2) *** -9.076 (2) *** -16.558 (0) *** -15.634 (0) *** -15.587 (0) *** -7.165 (3) *** -8.955 (2) *** -15.276 (0) *** -8.230 (2) *** -15.116 (0) *** -15.914 (0) ***

PP(𝛿) -16.319 (6) *** -14.216 (3) *** -15.759 (5) *** -16.564 (5) *** -15.605 (2) *** -15.614 (3) *** -16.433 (5) *** -14.352 (5) *** -15.312 (5) *** -15.818 (7) *** -15.355 (6) *** -16.007 (7) ***

PP(𝜑) -16.331 (6) *** -14.404 (1) *** -15.741 (5) *** -16.542 (5) *** -15.601 (2) *** -15.594 (3) *** -16.434 (5) *** -14.329 (5) *** -15.329 (5) *** -15.825 (7) *** -15.374 (6) *** -15.986 (7) ***

of

ADF(𝜑) -16.179 (0) *** -8.983 (2) *** -9.069 (2) *** -16.537 (0) *** -15.632(0) *** -15.567 (0) *** -7.166 (3) *** -8.977(2) *** -15.300 (0) *** -8.245 (2) *** -15.151 (0) *** -15.894 (0) ***

pro

CHN FIN FRA GER HUN IND ITA JPN NOR SWE UK US

Augmented Dickey-Fuller (ADF) test

lP

re-

Notes: All test statistics are in first differences. ADF (𝛿) and PP (𝛿) are the models with intercept. ADF (𝜑) and PP (𝜑) represent a model with intercept and deterministic trend. The notation *, **, and *** indicates rejection of the null hypothesis at the 10%, 5%, and 1% significance level, respectively. Figures in ADF parenthesis indicate the selected lag length, and figures in PP parenthesis indicate the bandwidth used. We used a maximum lag length of 12 and the minimum AIC as a decision criterion to determine unit root processes. Source: Thompson Reuters DataStream.

Panel B: Zivot-Andrews structural break unit root test

GER HUN IND ITA JPN NOR SWE UK US

ZA (𝜑) -4.338 (4) * -3.868 (3) *

Time Break 2001M07 1998M01

-3.992 (3) ***

2008M06

-3.056 (4)

1998M04

-3.464 (0) ***

2001M02

-3.614 (0) **

2000M04

-3.697 (1) **

2011M05

-3.877 (1) ***

2003M12

-3.768 (1) ***

2004M09

-4.229 (1) ***

2003M05

-4.201 (4) ***

2008M06

-4.181 (4) ***

2008M06

-3.288 (3)

1997M07

-3.548 (3) *

2003M06

-4.145 (1) ***

2003M10

-4.326 (1) ***

2004M09

-3.556 (3)

2008M06

-4.045 (3) **

2000M09

-3.967 (4) **

2008M06

-3.836 (4) **

2008M06

-2.668 (3)

2007M11

-3.234 (3) **

2009M09

urn a

FRA

Time break 2002M07 1988M01

Jo

CHN FIN

ZA (𝛿) -4.155 (4) ** -3.708 (3) *

Notes: This table reports the Zivot-Andrews Structural Break test results. ZA (𝛿) represents the model with intercept, and ZA (𝜑) denotes the model with intercept and trend. This test is performed with a maximum number of lags of four. The notation *, **, and *** indicates rejection of the null hypothesis at the 10%, 5%, and 1% significance level, respectively. Figures in parentheses indicate the selected lag length.

36

Journal Pre-proof Panel C. Bai-Perron multi-structural break test t-stat

Period

t-stat

Period

t-stat

Period

t-stat

Period

1993M01-

15.379 ***

1996M09-

21.839

2000M052004M09

1.895 *

2012M02-

***

-3.620 ***

2008M06-

2000M04

38.255 ***

2004M10-

1996M08 1993M01-

94.250

1999M11-

29.589

2003M07-

131.555

2008M09-

***

2003M06

***

2008M08

***

2013M08

19.635 ***

2013M09-

1999M10 1993M01-

109.387 ***

1999M11-

53.589 ***

2003M07-

75.912 ***

2008M09-

19.177 ***

2013M07-

313.939 ***

1999M11-

55.327 ***

2003M07-

56.929 ***

2008M09-

22.641 ***

2013M09-

45.668 ***

1997M01-

23.084 ***

2001M03-

3.212 ***

2004M11-

8.406 ***

2008M09-

22.375 ***

1999M08-

18.142 ***

2003M05-

-2.706 ***

2008M06-

2.114 **

2012M02-

142.006 ***

1998M01-

35.239 ***

2001M09-

21.415 ***

2013M09-

31.254 ***

1999M07-

28.201 ***

2003M08-

145.350 ***

1998M08-

37.901 ***

2003M10-

59.862 ***

1999M11-

20.954 ***

2003M07-

163.625 ***

1999M10-

137.43 3 ***

2003M06-

83.095 ***

1999M11-

GER

1993M011999M10

HUN

1993M011996M12

IND

1993M011999M07

ITA

1993M011997M12

JPN

1993M011999M06

NOR

1993M011998M07

SWE

1993M011999M10

UK

1993M011999M09

US

1993M011999M10

2003M06 2003M06 2001M02 2003M04 2001M08 2003M07 2003M09 2003M06

2003M05

2003M06

2008M08 2008M08 2004M10 2008M05 2008M06 2008M08

66.419 *** 8.537 ***

2013M06 2013M08 2008M08

2012M01 2017M09 2017M09

of

1999M10

2008M05

2017M09 2014M01

pro

FRA

Period

2012M01

2008M072013M08

2008M092017M09

2017M09

2012M04

16.174 ***

2008M09-

0.140

2012M05-

33.033 ***

2008M09-

2008M10-

2008M08 2008M08

2008M08

2003M072008M09

58.198 ***

2012M04

2013M10

2013M09

t-stat

2017M09

2.708 ***

2014M02-

0.438

21.524 *** 21.248 *** 18.887 *** 5.326 ***

2017M09

14.427 *** 17.578 ***

17.426 ***

2012M05-

61.673 ***

2008M09-

2017M09

5.647 ***

10.450 ***

re-

FIN

t-stat

lP

CHN

Period

2017M09 2017M09

19.425 ***

2013M11-

25.593 ***

2013M10-

2017M09

2017M09

19.931 *** 8.094 *** 28.576 *** 29.348 ***

Jo

urn a

Notes: We perform the Bai-Perron multi-structural break test with sequential tests for all subsets break specification. We specify a model with a constant and a linear deterministic trend within each segment defined by break dates, which minimized the BIC. The adopted trimming percentage is 15, with maximum levels of 5 and significance level of 0.05. We employed a HAC (NeweyWest) coefficient’s covariance matrix with degrees-of-freedom adjustment and allowance of error distribution to differ across breaks. The notation *, **, and *** indicates rejection of the null hypothesis at the 10%, 5%, and 1% significance level, respectively. Source: Thompson Reuters DataStream.

37

Journal Pre-proof Table 3. Unconditional correlation matrix CHN

FIN

FRA

GER

HUN

IND

ITA

JPN

NOR

SWE

UK

1.000

FIN

0.101

1.000

FRA

0.149

0.731

1.000

GER

0.170

0.687

0.898

1.000

HUN

0.068

0.566

0.655

0.604

1.000

IND

0.173

0.419

0.492

0.471

0.497

1.000

ITA

0.115

0.649

0.818

0.757

0.580

0.480

1.000

JPN

0.049

0.483

0.479

0.417

0.359

0.358

0.422

1.000

NOR

0.149

0.639

0.755

0.689

0.666

0.492

0.660

0.478

1.000

SWE

0.105

0.793

0.837

0.817

0.618

0.516

0.735

0.501

0.760

1.000

UK

0.155

0.707

0.864

0.806

0.627

0.470

0.717

0.520

0.807

0.807

1.000

US

0.159

0.681

0.776

0.790

0.593

0.445

0.619

0.478

0.661

0.756

0.798

pro

of

CHN

US

1.000

Jo

urn a

lP

re-

Notes: All series used in Table 3 are in first differences. The unconditional correlation matrix is calculated by applying a Pearson correlation coefficient. Source: Thompson Reuters DataStream.

38

Journal Pre-proof Table 4. Estimation results of the DCC-EGARCH model for the stock market

0.008 -0.828*** -0.022 0.927***

0.889*** -0.153 0.204** 274.52c

UK

0.007

0.009** 0.194 -0.022 -0.021

0.005*

0.039 -0.003 0.149** 0.007 0.045

-0.200

0.062 0.022 -0.000

-55.602b*** 0.197

-62.921b*** 0.704

-0.254 0.106 383.079d

-0.205* 0.247 400.489c

-55.638b*** 1.530 0.476 -0.152 0.834*** -0.095 0.187 409.863d

0.614*** 0.749*** 0.441*** 0.609*** 0.141** 0.501*** 0.028* 0.339

0.752*** 0.760*** 0.462*** 0.739*** 0.108 0.478*** 0.017*** 0.943***

0.441*** 0.726*** 0.137* 0.446*** 0.014** 0.958***

0.502*** 0.759*** 0.123 0.454*** 0.009* 0.961***

-22.113 -0.331** 0.416 7.540** 13.119

-22.344 -0.060 1.875*** 43.553*** 7.276

-14.431 -0.324** 1.287*** 25.593*** 15.258

-15.424 -0.200 0.164 2.297 5.011

-5.414*** 1.176 1.145 -0.268

re-

0.540*** 0.579*** 0.315*** 0.498*** 0.071 0.445*** 0.012** 0.950***

GER

pro

-4.736*** -0.173

0.007* 0.368 -0.003 -0.073 -0.086 -0.372

EU SWE

of

NCCU NOR

HUN

-21.013 -0.102 2.433*** 73.512*** 6.739

lP

CCU FIN FRA ITA Panel A: Mean equation – Univariate EGARCH Const (m) 0.008* 0.006* 0.003 AR(1) 0.443** -0.608* AR(2) 0.461** -0.023 AR(3) 0.028 AR(4) 0.127** MA(1) -0.334 0.085 0.654* MA(2) -0.640*** MA(3) 0.037 MA(4) MA(5) Panel B: Variance equation – Univariate EGARCH Const (v) 13.202 -59.367b*** -5.485*** α1 -0.546 0.834*** 0.990*** α2 -0.070 -0.300 β1 1.400*** -0.096*** -0.704*** β2 -0.400 0.877*** θ1 -0.004 -0.243*** -0.137* θ2 0.465*** 0.169 0.377*** c d Log-L 362.005 457.942 394.756c Panel C: DCC – Multivariate EGARCH 0.673*** 0.882*** 0.744*** 𝜌GER 0.663*** 0.830*** 0.727*** 𝜌UK 0.462*** 0.475*** 0.435*** 𝜌JPN 𝜌US 0.650*** 0.682*** 0.577*** 𝜌CHN 0.150** 0.146** 0.075 𝜌IND 0.421*** 0.466*** 0.478*** α 0.024** 0.055** 0.019*** β 0.903*** 0.102 0.937*** Panel D: Diagnostic tests AIC -21.551 -23.860 -21.981 Skewness -0.154 -0.034 0.232 Excess kurtosis 1.120*** 1.008*** 1.481*** Jarque-Bera 16.630*** 12.591*** 29.704*** 2 Q (12) 7.472 10.795 17.267

0.652

0.324 -0.192** 0.267** 509.08d

Jo

urn a

Notes: Estimation results of the DCC-EGARCH model for the stock market. Univariate and multivariate model specifications are based on the best-fitted model with convergence in both estimations and insignificance in Q2 (12) at the 10%, 5%, and 1% significance levels for the diagnostic test. If several specifications are fitted with insignificant Q2(12), the one with the lowest AIC is chosen. Univariate and multivariate model specifications are first tested on a GARCH(1, 1) process; if no model is fitted with insignificant Q2(12), we further test for GARCH(2, 1), (1, 2), and (2, 2), and we choose a specification according to the lowest AIC. The notation *, **, and *** indicates rejection of the null hypothesis at the 10%, 5%, and 1% significance level, respectively. The notation b indicates the constant multiplied with 107. The notation c indicates α + β < 1 and d indicates α + β > 1. In Panel D, the skewness, excess kurtosis, Jarque-Bera, and Q2(12) are calculated on the selected univariate GARCH model.

39

Journal Pre-proof Figure 1. Stock market indices in log-level China

Finland

7 6

Germany

France

10

9

5

8.4

10.0

8.0

9.5

7.6

9.0

7.2

8.5

6.8

8.0

8 4 7

2

6 94

96

98

00

02

04

06

08

10

12

14

16

7.5

6.4 94

96

98

00

Hungary

02

04

06

08

10

12

14

16

94

96

98

00

02

04

India

6

4.5

2.5 06

08

10

12

14

16

96

98

00

Norway

02

04

06

08

10

12

14

16

4.5 4.0

7.0

3.5

6.5

3.0

6.0

2.5

5.5

2.0

5.0

1.5

8.4

7.6

7.2

96

98

00

02

04

06

08

10

12

14

16

6.8 94

96

98

00

02

04

06

08

96

98

00

02

04

06

08

10

12

14

16

Unted Kingdom

8.0

94

04

06

08

10

12

14

16

10

12

14

16

1.8

94

Sweden

7.5

02

2.0

5.6 94

8.0

00

Japan

pro

04

98

2.2

6.0

3.0

02

96

2.4

6.4

3.5

00

94

16

2.6

6.8

3

98

14

2.8

4.0

96

12

3.0

7.2

5

94

10

7.6

5.0

2

08

Italy

5.5

4

06

of

3

10

12

14

16

94

96

98

00

02

04

06

08

10

12

14

94

96

98

00

02

04

06

08

United States 8.0

7.5

7.0

6.5

6.0 94

16

96

98

00

02

04

06

08

10

12

14

16

re-

Notes: The period covered is 1993M01-2017M09. Source: DataStream International.

Finland

China

.6

lP

Figure 2. Stock market indices in first differences .4

France

.2

.4

.2

.1

.2 .2

.1

.0

.0

.0

.0

-.1

-.2 -.2

-.6

-.4 96

98

00

02

04

06

08

10

Hungary

12

14

16

-.2

-.3

94

96

98

00

02

04

06

08

10

12

urn a

94

-.1

-.2

-.4

14

16

-.3 94

96

98

00

02

India

.6

.0

06

08

10

12

14

16

-.6 96

98

00

02

04

06

08

10

Norway

12

14

16

94

96

98

00

02

04

06

08

10

12

-.3 14

16

96

98

00

02

04

06

08

10

12

14

16

-.4 96

98

00

02

04

06

08

10

12

14

16

98

00

02

04

06

08

10

12

14

16

96

98

00

02

04

06

08

10

12

14

16

12

14

16

.1

.0 .0 -.1 -.1

-.3 96

12

.2

.1

-.2

94

10

United States

.2

-.4

94

94

United Kingdom

-.2

-.3

08

-.2 94

.0

-.2

Jo

-.1

06

-.1

-.2

.2

.0

04

.0

.4

.1

02

.1

Sweden

.2

00

-.1

-.4

94

98

.0

-.2

-.4

96

.2

.1

.0

-.2

94

Japan

.2

.2

.2

04

Italia

.4

.4

Germany

.3

14

16

-.2 94

96

98

00

02

04

06

08

10

12

14

16

Notes: The period covered is 1993M02-2017M09. Source: DataStream International.

40

94

96

98

00

02

04

06

08

10

Journal Pre-proof Figure 3. Estimation results of the DCC-EGARCH model for the stock market European market (EU)

Global markets (GM)

Emerging markets (EM)

pro

of

Panel A: Finland vs EU, GM, and EM

lP

re-

Panel B: France vs EU, GM, and EM

urn a

Panel C: Italy vs EU, GM, and EM

Jo

Panel D: Hungary vs EU, GM, and EM

41

Journal Pre-proof European market (EU)

Global markets (GM)

Emerging markets (EM)

pro

of

Panel E: Norway vs EU, GM, and EM

urn a

lP

Panel G: Germany vs EU, GM, and EM

re-

Panel F: Sweden vs EU, GM, and EM

Jo

Panel H: United Kingdom vs EU, GM, and EM

42

Journal Pre-proof Figure 4. Cross-quantilogram heatmaps for the stock market European market (EU)

Global markets (GM)

Emerging markets (EM)

CHN

JPN

IND

FIN

pro

FIN

re-

FRA

ITA

ITA

US

IND

urn a

ITA

CHN

ITA

JPN

Jo

GER

ITA Panel D: Hungary vs EU, GM, and EM

HUN

HUN

US

IND

HUN

UK

FRA

CHN

lP

FRA

JPN

GER UK

FRA

US

UK

FRA

ITA

FIN

CHN

JPN

GER

Panel B: France vs EU, GM, and EM

FRA Panel C: Italy vs EU, GM, and EM

FIN

of

FIN

US

UK

FIN

IND

GER

Panel A: Finland vs EU, GM, and EM

HUN

HUN

43

HUN

Journal Pre-proof European market (EU)

Global markets (GM)

Emerging markets (EM)

CHN

JPN US

of

SWE

re-

SWE

SWE

SWE

SWE

JPN

GER

GER

US

IND

urn a

SWE

CHN

lP

Panel G: Germany vs GM and EM

NOR

CHN

pro

NOR

JPN

GER

NOR Panel F: Sweden vs EU, GM, and EM

UK

NOR

IND

NOR

US

UK

NOR

IND

GER

Panel E: Norway vs EU, GM, and EM

GER

GER

UK

IND

UK

US

Jo

JPN

CHN

Panel H: United Kingdom vs GM and EM

UK

UK

Notes: The cross-quantilogram correlation is estimated using Eq. (1), and statistical significance is measured using the Ljung-Box test.

44

Journal Pre-proof Figure 5. Cross-quantilogram rolling-window estimations for the stock market Panel A: Finland vs EU, GM, and EM under different market conditions Q = 0.50

JPN

FIN

US

FIN

FIN

CHN

CHN

FIN

FIN

FIN

IND

IND

FIN

Jo

FIN

UK

urn a

US

FIN

CHN

FIN

US

lP

re-

FIN

JPN

JPN

FIN

IND

of

pro

FIN

UK

UK

FIN

FIN

Q = 0.95

GER

GER

GER

Q = 0.05

FIN

FIN

45

FIN

Journal Pre-proof Panel B: France vs EU, GM, and EM under different market conditions Q = 0.50

FRA

UK

pro

UK

re-

JPN

lP

FRA

US

US

CHN

FRA

urn a

CHN

FRA

US

FRA

FRA

JPN

FRA

FRA

FRA

Jo

IND

IND

FRA

FRA

CHN

UK JPN

FRA

IND

FRA

of

FRA

FRA

Q = 0.95

GER

GER

GER

Q = 0.05

FRA

FRA

46

FRA

Journal Pre-proof Panel C: Italy vs EU, GM, and EM under different market conditions

UK

pro

UK

ITA

JPN

lP

US

ITA

CHN

ITA

urn a

CHN

ITA

US

ITA

re-

JPN

JPN

ITA

US

UK

ITA

ITA

ITA

of

ITA

ITA

ITA

Jo

IND

IND

ITA

ITA

CHN

GER

ITA

IND

Q = 0.95

GER

Q = 0.50

GER

Q = 0.05

ITA

ITA

47

ITA

Journal Pre-proof Panel D: Hungary vs EU, GM, and EM under different market conditions Q = 0.50

GER HUN

UK

pro

UK

re-

JPN

lP

HUN

US

US

CHN

HUN

urn a

CHN

HUN

US

HUN

HUN

JPN

HUN

HUN

HUN

Jo

IND

IND

HUN

HUN

CHN

UK JPN

HUN

HUN

HUN

of

HUN

IND

Q = 0.95

GER

GER

Q = 0.05

HUN

HUN

48

HUN

Journal Pre-proof Panel E: Norway vs EU, GM, and EM under different market conditions Q = 0.50

GER UK

pro

UK

UK

NOR

re-

JPN

JPN

lP CHN

CHN

NOR

NOR

NOR

IND

IND

NOR

Jo

NOR

NOR

urn a

CHN

NOR

US

US

NOR

NOR

JPN

NOR

NOR

NOR

of

NOR

US

GER

NOR

IND

Q = 0.95

GER

Q = 0.05

NOR

NOR

49

NOR

Journal Pre-proof Panel F: Sweden vs EU, GM, and EM under different market conditions Q = 0.50

GER UK

pro

UK

re-

JPN

lP

US

SWE

CHN

SWE

urn a

CHN

SWE

US

SWE

SWE

JPN

SWE

US

UK JPN

SWE

SWE

SWE

of

SWE

SWE

SWE

Jo

IND

IND

SWE

SWE

CHN

GER

SWE

IND

Q = 0.95

GER

Q = 0.05

SWE

SWE

50

SWE

Journal Pre-proof Panel G: Germany vs GM and EM under different market conditions Q = 0.95

JPN

Q = 0.5

JPN

US

pro

GER

reGER

lP

IND

IND

GER

GER

CHN

CHN

CHN

GER

GER

Jo

urn a

GER

GER

of

GER

US

US

GER

51

GER

IND

JPN

Q = 0.05

GER

Journal Pre-proof Panel H: United Kingdom vs GM and EM under different market conditions Q = 0.95

JPN

Q = 0.50

JPN

US

pro

UK

reUK

lP

IND

IND

UK

UK

CHN

CHN

CHN

UK

UK

UK

of

UK

US

US

UK

UK

UK

IND

JPN

Q = 0.05

UK

Jo

urn a

Notes: CQ figures are estimated based on Eq. (1) via recursive-rolling samples. The first, second, and third columns show stock return distributions at the 5%, 50%, and 95% quantiles, respectively. The horizontal axis represents the starting year of the rolling window. The blue lines are rolling CQs, whereas the red lines indicate a 95% confidence interval for the null hypothesis of no-predictability. The confidence interval is calculated by a bootstrap procedure, where 500 bootstrap replications are used.

52

Journal Pre-proof Appendix

Figure A.1. DCC-EGARCH estimation results: Poland and the Czech Republic European market (EU)

Global markets (GM)

Emerging markets (EM)

re-

pro

of

Panel A: Poland vs EU, GM, and EM

Jo

urn a

lP

Panel B: Czech Republic vs EU, GM, and EM

53

Journal Pre-proof Figure A.2. Cross-quantilogram heatmaps: Poland and the Czech Republic European market (EU)

Global markets (GM)

Emerging markets (EM)

pro

US

POL

lP

CZE

CZE

IND

CZE

US

UK

CZE

re-

JPN

GER

Panel B: Czech Republic vs EU, GM, and EM

POL

CHN

POL

CZE

POL

IND

POL

UK

POL

of

CHN

JPN

GER

Panel A: Poland vs EU, GM, and EM

CZE

Jo

urn a

Notes: The cross-quantilogram correlation is estimated using Eq. (1), and statistical significance is measured using the Ljung-Box test.

54

Journal Pre-proof Figure A.3. Cross-quantilogram rolling-window estimations: Poland and the Czech Republic Panel A: Poland vs EU, GM, and EM under different market conditions

lP

urn a

of JPN

POL

POL

POL

CHN

CHN

Jo

CHN

POL

POL

US

POL

US

US

re-

POL

JPN

JPN

POL

POL

POL

UK

pro

POL

UK

UK

POL

POL

POL

IND

IND

POL

IND

Q = 0.95

GER

Q = 0.50

GER

GER

Q = 0.05

POL

POL

55

POL

Journal Pre-proof Panel B: Czech Republic vs EU, GM, and EM under different market conditions Q = 0.50

GER UK

pro

UK

CZE

lP CHN

CHN

CZE

CZE

CZE

IND

IND

CZE

Jo

CZE

CZE

urn a

CHN

CZE

US

US

CZE

CZE

JPN

re-

JPN

JPN

CZE

CZE

CZE

of

CZE

US

GER UK

CZE

IND

Q = 0.95

GER

Q = 0.05

CZE

CZE

CZE

Notes: CQ figures are estimated based on Eq. (1) via recursive-rolling samples. The first, second, and third columns show stock return distributions at the 5%, 50%, and 95% quantiles, respectively. The horizontal axis represents the starting year of the rolling window. The blue lines are rolling CQs, whereas the red lines indicate a 95% confidence interval for the null hypothesis of no-predictability. The confidence interval is calculated by a bootstrap procedure, where 500 bootstrap replications are used.

56