1
Integration, Contagion, and Crises in the Eurozone
By STEF J. DE JONG
ABSTRACT
In this paper I analyse the transmission of equity market shocks on synthetic equity market portfolios in the Eurozone. I investigate the occurrence of contagion for Eurozone countries from the perspective of a three-factor model. Furthermore, I analyse the role of financial and economic integration on the transmission of equity markets shocks and on contagion. I find evidence of contagion in both crisis periods. Contagion mainly occurred from domestic and European equity markets to individual domestic portfolios. Countries with a larger exposure to the European equity market experienced less contagion overall. Financial and economic integration influences the transmission of financial shocks and influenced the degree of contagion in the crises periods. (JEL
E44, E61, F45, G15, G18)
Supervisor: prof. dr. K. F. Roszbach Co-assessor: prof. dr. L. H. Hoogduin Course code: EBM000A20
Institution: University of Groningen, Faculty of Economics and Business
There is a substantial body of empirical literature (which I describe in detail in section I) which concludes that a further increase in financial and economic integration is necessary in order to improve the financial stability in the Eurozone. This literature argues that the movement towards a single European market improves risk-sharing (e.g. Baele et al. 2004), spurs economic growth (e.g. Chari et al. 2012), and can work as a market-based insurance mechanism (e.g. De Grauwe 2016). Both policy-makers and market participants have been actively advocating and pursuing policies for the movement towards a single European financial market. However, increased linkages between countries can be a source of distress and contagion in times of crisis (e.g. Imai and Takarabe 2011; Bacchetta and Wincoop 2016).
The discussion regarding the issues of interconnectedness and contagion intensified during the financial crisis of 2007 to 2009. This crisis originated in a small segment of the U.S. lending market. When the subprime home mortgage market bubble burst, U.S. financial institutions ran into trouble. Then, the crisis spread towards the economies of other countries, where it affected economies and equity markets worldwide. Some equity markets experienced a steeper decline in value than that of the originating country, the U.S. equity market. In Europe, a second crash occurred as some countries were no longer able to refinance their government budgets. Falling government bonds prices in certain countries led to a domestic banking crisis as banks experienced significant losses on their balance sheets. Also, banks experienced funding problems when domestic liquidity dried up. Financial markets calmed with the ECB announcement of free unlimited support under some restrictions. These two crisis periods, both with different origins, provide an ideal laboratory to investigate, debate and compare the presence of contagion, and the role of integration in European equity markets.
Motivated by the contradictions on financial and economic integration in the literature and by the impact of the crises in the past decade, I study the transmission of global, European, and domestic shocks to equity markets in the Eurozone. Furthermore, I investigate the role of financial and economic integration on the transmission of equity market shocks and on equity market contagion. The goal of this paper is to uncover whether contagion occurred in the Eurozone during two crisis periods, and to study what effect financial and economic integration has on the transmission of financial shocks and on equity market contagion.
contagion. Returns across stocks unrelated to the three factors may still be correlated during the crisis, which is captured by the fourth type of contagion. Finally, I analyse channels of contagion to shed light on how financial and economic integration within the Eurozone affects the transmission of financial shocks and contagion.
To briefly summarize my findings, I find evidence for financial and economic integration to influence the transmission of financial shocks and contagion. Additionally, I find significant evidence of contagion during the Great Financial Crisis (GFC). First, contagion occurred for the global, European and domestic factors. The evidence for domestic and European contagion is strongest, while global contagion had a relatively large effect, as global exposure increased relatively the most. Second, countries with high equity market integration, measured as the exposure towards the European risk factor, experienced the smallest degrees of contagion. Third, especially industries related to imports and exports experienced contagion from the global risk factor. Fourth, I find significant evidence that a greater degree of financial integration amplified contagion from the global risk factor during the GFC. Last, I find strong evidence for contagion unrelated to the three risk factors.
For the Sovereign Debt Crisis (SDC), I find strong evidence of contagion from the European risk factor, especially for countries in Southern Europe and Eastern Europe. Again, countries with a higher European exposure seem to be the least affected by the crises in terms of contagion. Financial integration is negatively related to European and domestic contagion during this crisis. For this crisis I also find strong evidence of contagion unrelated to the three risk factors in the model.
This work contributes to the literature at two different points. First, it adds to the literature of international market integration and shock transmission. What is added is the detailed approach with respect to the discussion of integration and its mechanism in influencing the transmission of shocks. The variables used to measure integration are specifically focused on European integration. This is done for an inclusive set of countries for the Eurozone and provides details on specific regions and industries within the Eurozone. Additionally, it adds to the literature regarding crises and contagion. The data period contains both the GFC and the SDC and tries to provide an in-depth explanation on the role of integration and the possibilities of contagion.
I. Literature
In the last two decades, the landscape of the European financial and economic markets changed dramatically. The introduction of the Euro has led to the movement of more integrated financial markets in the Eurozone. Since then, both European policy makers and market participants have been actively pursuing policies to increase financial integration. Abolishment of national currencies led to the market being able to price the inherent risk of two bonds or equities to be based on pure corporate risks, rather than on the risk-return composition being dominated by the risk of exchange rate fluctuations. However, De Grauwe (2013) and Schoenmaker (2011) argue that idiosyncratic movements have been left unconstrained by the introduction of the fully centralized Eurozone money and monetary policy. Macroeconomic policies remained in the hands of national governments. This has left the monetary union with very little to be able to converge economic boom and bust cycles as they originate at the national level, without becoming a common economic boom-and-bust cycle dynamic at the Eurozone level. In order to allow countries to converge their economic cycles, De Grauwe (2016) suggests that further integration of financial markets is necessary to enhance financial stability. Markets can act as an insurance mechanism when they are fully integrated. Integrated markets allow for risk-sharing between different countries, as asymmetric shocks can be carried by all countries. In the absence of European fiscal unification, the main risk-sharing mechanism must come from integrated financial markets. The integration of financial markets thereby provides an insurance mechanism and facilitates a smooth functioning of the monetary union.
stock returns and show that the probability of joining the EMU increases the relative importance of European risk factors.
The presence of a ‘home bias’ is common in public equity markets. Domestic investors hold mostly domestic equities and thereby do not effectively diversify their risks, even if the countries allow free movement of capital. Gârleanu et al. (2015) show that participants in incomplete markets suffer from growing informational frictions when they progressively participate in more unfamiliar markets. The home bias has been reduced significantly in Europe since the introduction of the Euro. Balli et al. (2010) show that the introduction of the Euro contributed to the reduction in this home bias. Bosch and Schoenmaker (2008) also find this reduction in home bias for European equity and bond markets.
Increased integration in the Eurozone may also have its dark side. Bley (2009) argues that the monetary policy convergence of the Eurozone has facilitated a divergence of economic variables. This divergence has led to changing returns behaviours between countries and hence diverging equity markets. Furthermore, increased integration may lead to increased probabilities of contagion. Imai and Takarabe (2011) investigate the relation between banking integration and the vulnerability of economies to financial shocks in Japan. They argue that increased integration may harm the economy as large financial institutions can aggressively transmit shocks and increase economic volatility. The effect may be contributed by home bias and poor lending portfolios. Bacchetta and Wincoop (2016) find that business cycle panic is synchronized across countries as long as there is a minimum level of economic integration. Only when the economies are sufficiently integrated are they subject to coordinated panics. They argue that a panic, when it occurs, can be perfectly synchronized across countries even when economic integration is limited.
strength of local fundamentals. Investors may even stop discriminating between these fundamentals as a result of herding behaviour.
Summarizing, the literature clearly introduces the potential gains from increased integration in both financial markets and trade linkages. Improvements in European integration may enhance further risk-sharing, economic growth, and it can help facilitate a smooth functioning of the centralized monetary policy. However, further integration may influence the transmission of financial shocks and it may lead to increased probabilities of contagion during crisis periods. Therefore, it is important to assess how the two crisis periods influenced the Eurozone and to analyse what role European integration has on the transmission of financial shocks and contagion.
II. Empirical Framework
In this section, I first explain the baseline model. Then, I add dummy variables for the crisis periods in order to investigate contagion. Thereafter, I introduce the full model in order to assess the role of financial and economic integration on the transmission of shocks and contagion. Finally, I explain the construction and estimation procedure of the three factors.
The empirical model
To investigate equity market exposures, I use the three-factor model proposed by Bekaert et al. (2014). This model allows me to distinguish the different exposures of country-industry equity portfolios. Moreover, by the addition of time dummy variables for two crises periods I am able to investigate contagion that occurred in the Eurozone, and it allows me to investigate the originations of contagion. Finally, by adding variables that proxy for economic and financial integration, I can assess the role of these variables on the exposures and contagion related to the three factors.
would imply that the world equity market is perfectly integrated. The interdependence model takes the following form:
𝑅𝑖,𝑡= 𝐸𝑡−1[𝑅𝑖,𝑡] + 𝛽𝑖,0′𝐹𝑡 + 𝑒𝑖,𝑡 (1)
where 𝑅𝑖,𝑡 is the excess return of the industry-portfolio i in week t (the weekly return of the portfolio minus the three-month German T-bill rate in weekly units). 𝐸𝑡−1[ 𝑅𝑖,𝑡] is the expectation of the excess return on the portfolio, measured as a linear function of the lagged excess return. 𝐹𝑡 is a vector of the global, European and domestic risk factors, and 𝑒𝑖,𝑡is the error term. I use the interdependence model
to capture the risk exposures of the country-industry portfolios related to the three factors. These exposures are captured by the betas for the three factors. This approach allows me to provide insights on the structure of the Eurozone equity markets.
In the past decade, countries in the Eurozone suffered from two crises, the Great Financial Crisis (GFC) and the Sovereign Debt Crisis (SDC). In order to assess whether and how contagion occurred, I first formulate the following two hypotheses:
Hypothesis (A): 𝐻0: The Eurozone experienced no contagion during the Great Financial Crisis. 𝐻1: The Eurozone experienced contagion during the Great Financial Crisis, especially contagion related to the global factor.
Hypothesis (B): 𝐻0: The Eurozone experienced no contagion during the Sovereign Debt Crisis. 𝐻1: The Eurozone experienced contagion during the Sovereign Debt Crisis, especially contagion related to the European factor.
beta coefficients to change for each of the three factors during the crisis period. To test for contagion, I estimate the following model:
𝑅𝑖,𝑡 = 𝐸𝑡−1[𝑅𝑖,𝑡] + 𝛽𝑖,𝑡′𝐹𝑡+ 𝜂𝑖𝐺𝐹𝐶𝑡+ 𝜍𝑖𝑆𝐷𝐶𝑡+ 𝑒𝑖,𝑡 (2)
𝛽𝑖,𝑡 = 𝛽𝑖,0+ 𝛾𝑖,0𝐺𝐹𝐶𝑡+ 𝜗𝑖,0𝑆𝐷𝐶𝑡 (3)
The coefficients of these dummies and their interactions, 𝛾𝑖,0, 𝜗𝑖,0, 𝜂𝑖 and 𝜍𝑖 in (2) and (3) capture the different sources of contagion. 𝛾𝑖,0 and 𝜗𝑖,0 are included to capture the contagion related to the domestic, European and global factors. They capture the change in risk exposure of the country-industry equity portfolios to the three factors during the two crisis periods. If these coefficients are positive and significant, then I define this as factor related contagion. I refer to the 𝛽𝑖,0 coefficients as the interdependence coefficients.
I include 𝜂𝑖 and 𝜍𝑖 in equation (2) to capture the effect of contagion unrelated to the factor loadings. A substantial negative coefficient indicates excess comovement for the portfolios during the crisis period, which I define as factor unrelated contagion. These two coefficients can capture multiple effects. First, behaviour shown by investors in cases of panic is non-fundamental and is thus not reflected in the exposures to the three factors. This non-fundamental contagion can be the result of irrational herding behaviour. Kaminsky and Reinhart (2000) suggest that herding behaviour is associated with contagion. In case of herding behaviour, investors are no longer discriminating across firms and countries based on their economic fundamentals and start to act the same way as the majority around them. A rational explanation for non-fundamental excess comovement is that during a crisis all investors may face the same liquidity constraints or margin calls at the same time. This can result in increased downward pressure on all portfolios and thus it can lead to increases in the excess comovement. Furthermore, the coefficients can capture the effects of economic fundamentals not captured by the factor loadings. I investigate contagion for each region and industry. This may provide a better understanding of how contagion occurred during the two crisis periods.
Finally, I investigate the effects of financial and economic integration on the exposures and contagion related to the three factors. First, I formulate the following hypothesis:
𝐻1: Financial and economic integration in the Eurozone influences the transmission of shocks related to the three factors and influenced contagion related to the three factors during the two crisis periods.
To test for the effects of financial and economic integration in the Eurozone on the exposures and contagion, I include 𝑍𝑖,𝑡−𝑘 into the model. 𝑍𝑖,𝑡−𝑘 consists of three variables to proxy financial and economic integration for country i and at time t – k, where k is the number of lags. I display the variables in Table I in Section II. I provide a further discussion on the variables in Section II. I include 𝑍𝑖,𝑡−𝑘 in (5) to assess its effect on the exposures of the three factors. Furthermore, I include 𝑍𝑖,𝑡−𝑘 in both (6) and (7) to assess the effects of financial and economic integration in crisis periods on contagion.The full model takes the following form:
𝑅𝑖,𝑡 = 𝐸𝑡−1[𝑅𝑖,𝑡] + 𝛽𝑖,𝑡′𝐹𝑡+ 𝜂𝑖𝐺𝐹𝐶𝑡+ 𝜍𝑖𝑆𝐷𝐶𝑡+ 𝑒𝑖,𝑡 (4) 𝛽𝑖,𝑡= 𝛽𝑖,0+ 𝛽1𝑍𝑖,𝑡−𝑘+ 𝛾𝑖,𝑡𝐺𝐹𝐶𝑡+ 𝜗𝑖,𝑡𝑆𝐷𝐶𝑡 (5)
𝛾𝑖,𝑡= 𝛾𝑖,0+ 𝛾1𝑍𝑖,𝑡−𝑘 (6)
𝜗𝑖,𝑡 = 𝜗𝑖,0+ 𝜗1𝑍𝑖,𝑡−𝑘 (7)
where 𝛽1 in (5) captures the effect of the variables in 𝑍𝑖,𝑡−𝑘 on the exposures related to the three factors. The total factor exposure of a portfolio outside crisis periods is now dependent on 𝛽𝑖,0 and 𝛽1𝑍𝑖,𝑡−𝑘. The effect of the variables in 𝑍𝑖,𝑡−𝑘 during crisis periods, with respect to the three factors, is measured by the coefficients 𝛾1 and 𝜗1 for the GFC in (6) and SDC in (7), respectively. These coefficients provide insights in the transmission channel of the global, European and domestic factors with respect to European integration. The variables are specific to the country in which the industry-portfolio is located and are lagged by two quarters. The model is estimated for all portfolios jointly, using pooled OLS. I account for cross sectional dependence between countries by clustering the standard errors of the portfolios within countries.
Estimation procedure for the factor loadings
time-varying factor loadings and changes in structural expectations, I conduct the regression for every consecutive period of 26 weeks. I use this procedure as it allows the factor structure to sufficiently change every six months. First, I estimate the global factor by regressing the MSCI world index returns on the European returns:
𝑟𝑡𝐺= 𝛼 + 𝛽𝑟𝑡𝐸𝑈+ 𝜀𝑡 (8)
where 𝑟𝑡𝐺 and 𝑟𝑡𝐸𝑈 are the MSCI world index and the Europeans return at time t, respectively. 𝜀𝑡 is the error term which I use as the global factor. I measure the European factor as the value weighted return of all European countries, excluding the country it is measured for. The domestic factor is extracted in a similar fashion as the global factor. First, I regress the domestic market return on the global and European returns, and again I use the residuals of the regression as part of the domestic factor:
𝑟𝑖,𝑡𝐷 = 𝛼 + 𝛽1𝑟𝑡𝐺+ 𝛽2𝑟𝑡𝐸𝑈+ 𝜀𝑖,𝑡 (9)
where 𝑟𝑖,𝑡𝐷 is the equity portfolio of industry i in country D. Similar to the European factor, the domestic factor 𝑅𝑡𝐷 is value-weighted for all country-industry portfolios located in the same country as portfolio i, except for portfolio i itself. By construction, the global and domestic factor have a zero-mean. I use
this approach as it avoids spurious correlations and adding-up constraints. The more correct approach in notation would be to define the European and domestic factors as 𝑅𝑡𝐸𝑈/𝐷 and 𝑅𝑡𝐷/𝑖, but I use 𝑅𝑡𝐸𝑈 and 𝑅𝑡𝐷 for convenience.
III. Data
industry j. Then, I calculate the portfolio return by taking the weighted sum of company returns, for each industry: 𝑅𝑗,𝑐,𝑡 = ∑ ( 𝑀𝑉𝑖,𝑗,𝑐,𝑡 𝑀𝑉𝑗,𝑐,𝑡) ∗ 𝑟𝑖,𝑗,𝑐,𝑡 𝑁𝑖,𝑗,𝑐,𝑡 𝑖=1 (10)
where 𝑁𝑖,𝑗,𝑐,𝑡 is the number of companies in industry j in country c at time t. 𝑅𝑗,𝑐,𝑡 is the return on the synthetic industry portfolios. Table A in the Appendix provides the number of companies I use in the construction of the portfolio and their corresponding countries. The ten industries correspond to the industry classification benchmarks, found in the note under Table A. Furthermore, I report the countries and their corresponding regions in Table B in the Appendix. Not all countries have companies that are traded sufficiently in all sectors on their national stock exchange, hence not all countries have a matching number of industry-portfolios. In total, the sample contains 177 equity portfolios, for 24 European countries. The sample period runs from January 2001 until March 2018. I obtain all data on stock returns, market values and industry classification benchmarks from Datastream.
Proxies for integration in Europe
To proxy for integration in Europe, I use cross-border monetary financial institutional (MFI) loans, cross-border European trade and misalignment of industries. The first measure, cross-border MFI loans, proxies for financial integration, as suggested by Bekaert et al. (2014). The second measure, cross-border trade, is commonly used in the literature as a proxy for trade integration (e.g. Inklaar, Jong-A-Pin and de Haan (2008), Bekaert et al. (2005), and Baele (2005)). Finally, I use industry misalignment as the last measure, as proposed by Baele and Inghelbrecht (2009), to proxy for similarities in the industry composition of a country compared to the Eurozone. The measures for cross-border trade and cross-border MFI loans are both constructed in a similar fashion:
Table I
Variables for measuring integration
This table reports the variables in 𝑍𝑖,𝑡−𝑘 used in (5) to (7), and their sources. All the statistics shown are calculated
across the 177 country-industry portfolios for the entire time sample.
where I scale both measures by the GDP of each country. I measure industry misalignment as:
𝑀𝑖𝑠𝑎𝑙𝑖𝑔𝑛𝑚𝑒𝑛𝑡𝑐,𝑡= √∑ ( 𝑤𝑖,𝑡𝑐 − 𝑤𝑗,𝑡𝐸𝑈)² 𝑁𝑐
𝑗 = 1 (13)
where 𝑁𝑐 is the number of industries within country c, 𝑤𝑗,𝑡𝑐 is the weight of each industry j in country c at time t and 𝑤𝑖,𝑡𝐸𝑈 is the total weight of the whole industry of all countries in Europe1. The weights are
computed as the market capitalization of an industry within a country divided by the total market capitalization of all industries within the country. I report the summary statistics and data sources of each variable in Table I.
1Again, the weights of the industries for 𝑤
𝑖,𝑡𝐸𝑈 are measured for country c, while excluding country c from the European
industry sample, to avoid endogeneity issues. Strictly speaking, 𝑤𝑖,𝑡𝐸𝑈/𝑐 should be denoted but I use the former for notational ease.
Variables Units Frequency Definition
Unit of
I use the amount of cross-border MFI (excluding ESCB2) loans, scaled by GDP, as the proxy for
financial integration. I include this measure to capture the degree of risk sharing and financial openness in the Eurozone. More financial openness and risk sharing should be positively related to the amount of cross-border MFI loans. By holding cross-border positions in their loan portfolio, financial institutions are able to diversify idiosyncratic country risks. More financially open countries are likely to attract more foreign investment due to lower entry barriers. However, more banking integration also comes with downsides. Schoenmaker (2011) argues that the increased linkages between countries increases the risk financial instability as long as the individual countries have national financial autonomy, as increased financial integration reduces the effect of national policies. This may cause countries with more banking integration to experience larger European exposures. De Grauwe (2016) also argues that an incomplete banking union leads to financial instability. Increased financial linkages between countries may increase exposures and contagion. Japelli and Pagano (2008) argue for the importance of cross-border measures for financial integration as the integration into a single market decreases the effectiveness of national size-based measures. Therefore, I focus in this paper on cross-border positions.
To proxy for the effects of economic integration and trade openness in the Eurozone, I use European imports plus exports, scaled by GDP. Inklaar, Jong-A-Pin and de Haan (2008) find that trade intensity affects the synchronization of the business cycles. Synchronized business cycles increase the interdependence of equity markets between countries. Bekaert et al. (2005) and Baele (2005) document a positive relation between the degree of trade integration for a country and the exposure of the country’s returns towards regional and global equity markets. Kaminsky and Reinhart (2000) find a positive relation for the trade linkages between two countries and the transmission of idiosyncratic shocks and probabilities of contagion. Trade linkages are thus important in assessing the propagation of shocks. Also, they argue that trade often takes place within a region, and thus tends to be more intra- than extra-regional in nature. Therefore, the measure I use for trade integration is the intra-European imports plus exports.
The transmission of shocks in a country is likely to be dependent on the structure of the industry composition within the country. If the structure of a country’s industrial composition becomes more aligned to that of another country or region, then their returns should become similar. Baele and Inghelbrecht (2009) document a strong negative relation between a region becoming increasingly different in its industry structure relative to the region and the degree of which shocks are transmitted. They find that as an industry is more evenly spread out over a region, it becomes more exposed to global shocks. I investigate the misalignment of industries within countries relative to the industries within the Eurozone.
If the goal of the European Commission, financial stability, can indeed be achieved by further integration of the Eurozone, then the coefficients for the trade and financial integration variables of 𝑍𝑖,𝑡−𝑘 should be negative during normal times for the domestic and global factors and positive for the European factor. During the crisis periods, these increased linkages may be a source of contagion, if this is the case, then their coefficients will be positive. The variable misalignment is included to determine whether the movement towards one single market increases the possibilities of contagion, and whether a country with a more (less) misaligned industry portfolio is better (worse) protected from non-domestic shocks.
I lag all three variables in 𝑍𝑖,𝑡−𝑘 in (5) to (7) by two quarters. By using this lag, I prevent the possibility of a spurious relationship between the returns of the equity market portfolios and the variables within 𝑍𝑖,𝑡−𝑘, as this prevents an unobserved factor being able to influence both simultaneously.
IV. Empirical Results
In this section, I discuss the results of the models introduced in Section II. I divide this section into two subsections. In the first subsection, I discuss the results of the interdependence model in (1). Then, I discuss the results of the interdependence model with time dummies to capture the effects of contagion, as presented in (2) and (3), to test for Hypotheses (A) and (B). In this subsection, I will also explore the variation of the coefficients from these two models across portfolios, by aggregating over European regions and industries. Then, in the second subsection, I report and discuss the estimates of the full model, (4) to (7). The full model contains the financial and economic integration proxies in order to test Hypothesis (C). Finally, I provide two robustness tests for the full model.
Interdependence and contagion
Table II
Interdependence The table reports the estimates of the interdependence model:
𝑅𝑖,𝑡= 𝐸𝑡−1[𝑅𝑖,𝑡] + 𝛽𝑖,0′𝐹𝑡 + 𝑒𝑖,𝑡 (1)
The table depicts the estimates of the unweighted average degree of interdependence of all portfolios in the sample using pooled OLS. The model is corrected for heteroskedasticity by using country-clustered standard errors. For the subscripts, D denotes the domestic factor, EU the European factor and G the global factor. ***, **. and * indicate the statistical significance of the estimates at the 1%, 5%, and 10% levels, respectively.
Coeff. St. Err. Interdependence 𝐸𝑡−1[𝑅𝑖,𝑡] -0.006 0.005 𝛽𝐷 0.353 *** 0.056 𝛽𝐸𝑈 0.689*** 0.060 𝛽𝐺 0.108*** 0.027 Constant -0.063*** 0.022 Observations 158,920 R² 0.205
First, I estimate the interdependence model and discuss the results on the structure of Eurozone equity markets and the averaged exposures of the equity portfolios. By excluding the crisis dummies, the model collapses to a three-factor model, where a coefficient for a factor equal to one implies full integration. In Table II, I present the estimated parameters of the interdependence model as in (1).
Each of the reported βs of equation (1) in Table II reflect the average degree of exposure across all the 177 country-industry portfolios. The exposures to all three factors, measured as the beta coefficients, are highly significant. The relatively high exposure to the European factor of 0.689, almost twice the size of the exposure to domestic factor, suggests that the portfolios on average indicate a rather well integrated equity market within the Eurozone. Next, the domestic factor dominates the global factor. These first results already provide some insights for the average European country and its equity market. European equity markets are an important factor for a country’s equity markets, while the global equity market seems to be of a much smaller influence.
In Table III, I report the estimated beta coefficients of equation (1) of Table II across portfolios, aggregated over European regional groups and industries. When exploring the variation per region in panel A, I find that the exposure to the European factor generally dominates the exposure to the domestic and global factors. The exception is Eastern Europe, for which the domestic factor exposure is higher than the European factor. Both Eastern Europe and Southern Europe denote a relatively high coefficient for the domestic factor of 0.420 and 0.496, respectively. When comparing the 𝛽𝐸𝑈 for the
Table III
Interdependence across Regions and Industries
The table reports the estimates of the interdependence model of Table II, aggregated over regions and industries: 𝑅𝑖,𝑡= 𝐸𝑡−1[𝑅𝑖,𝑡] + 𝛽𝑖,0′𝐹𝑡 + 𝑒𝑖,𝑡 (1)
The table depicts the estimates and standard errors using pooled OLS of Table II, but on a regional (Panel A) and industry (Panel B) basis. Thus, the number of observations and the R² are the same. For the subscripts, D denotes the domestic factor, EU the European factor and G the global factor. ***, **. and * indicate the statistical significance of the estimates at the 1%, 5%, and 10% levels, respectively.
Panel A: By region
Interdependence
Region 𝛽𝐷 St. Err. 𝛽𝐸𝑈 St. Err. 𝛽𝐺 St. Err.
Eastern Europe 0.420*** 0.061 0.356** 0.127 0.152** 0.050 Northern Europe 0.313*** 0.069 0.630*** 0.101 0.128*** 0.034 Southern Europe 0.496*** 0.075 0.793*** 0.079 0.117** 0.047 Western Europe 0.156* 0.064 0.883*** 0.079 0.047 0.065 Panel B: By industry Interdependence
Industry 𝛽𝐷 St. Err. 𝛽𝐸𝑈 St. Err. 𝛽𝐺 St. Err.
Basic materials 0.342*** 0.116 0.800*** 0.102 0.246*** 0.074 Consumer goods 0.257*** 0.050 0.564*** 0.067 0.011 0.044 Consumer services 0.359*** 0.111 0.650*** 0.086 0.131** 0.060 Financials 0.342*** 0.067 0.890*** 0.093 0.047 0.042 Health care 0.367*** 0.063 0.638*** 0.089 0.038 0.049 Industrials 0.423*** 0.086 0.733*** 0.076 0.189*** 0.038
Oil & gas 0.425*** 0.083 0.668*** 0.095 0.171*** 0.044
Technology 0.274* 0.130 0.641*** 0.147 0.197*** 0.065
Telecommunications 0.376*** 0.063 0.618*** 0.076 -0.040 0.038
Utilities 0.332*** 0.071 0.612*** 0.099 0.148* 0.083
The results of the European factor could be due to the use of the weighted average of all countries for the European factor. If financially developed countries are clustered within one region, and financial development is reflected in relatively high weights for their country equity market portfolios in the construction of the EU factor, then the region could be of large influence on the European factor.
Table IV
The Interdependence Model with Contagion
The table reports the estimates of the interdependence model with the inclusion of contagion parameters: 𝑅𝑖,𝑡= 𝐸𝑡−1[𝑅𝑖,𝑡] + 𝛽𝑖,𝑡′𝐹𝑡+ 𝜂𝑖𝐺𝐹𝐶𝑡+ 𝜍𝑖𝑆𝐷𝐶𝑡+ 𝑒𝑖,𝑡 (2)
𝛽𝑖,𝑡= 𝛽𝑖,0+ 𝛾𝑖,0𝐺𝐹𝐶𝑡+ 𝜗𝑖,0𝑆𝐷𝐶𝑡 (3)
The table depicts the estimates of the unweighted average degree of interdependence and contagion of all portfolios in the sample using pooled OLS. The model is corrected for heteroskedasticity by using country-clustered standard errors. For the subscripts, D denotes the domestic factor, EU the European factor and G the global factor. ***, **. and * indicate the statistical significance of the estimates at the 1%, 5%, and 10% levels, respectively. Coeff. St. Err. Interdependence 𝐸𝑡−1[𝑅𝑖,𝑡] -0.008 0.005 𝛽𝐷 0.324*** 0.055 𝛽𝐸𝑈 0.626*** 0.072 𝛽𝐺 0.080*** 0.028 Contagion 𝛾𝐷 0.126* 0.066 𝛾𝐸𝑈 0.138*** 0.044 𝛾𝐺 0.098* 0.053 𝜗𝐷 0.066 0.077 𝜗𝐸𝑈 0.112*** 0.033 𝜗𝐺 0.085 0.052 𝜂 -0.555*** 0.097 𝜍 -0.133*** 0.045 Constant 0.034 0.037 Observations 158920 R² 0.208
technology. This intuitively makes sense, as these sectors are relatively more dependent on global factors.
In order to test whether contagion occurred during the two crisis periods, I estimate equations (2) and (3). The addition of the time dummies for the two crisis periods in (3) allows the beta coefficients to take different values during the crisis periods. I measure contagion related to the three factors as changes in the degree of exposures towards the three factors during the crisis periods. This effect is captured by 𝛾𝑖,0and 𝜗𝑖,0. Contagion unrelated to the three factors is captured by 𝜂𝑖and 𝜍𝑖. I report the
estimates of the interdependence model with contagion dummies, denoted by equations (2) and (3), in Table IV. I report the average 𝛽𝑖,0 for the interdependence coefficients of the three factors, and 𝛾𝑖,0 and
𝜗𝑖,0 as the contagion coefficients for the three factors during the GFC and the SDC. The superscripts
denote the origin of the three factors. 𝜂𝑖and 𝜍𝑖are the coefficients for the crisis time dummies without
improvement of the model, and to the addition of new variables. I will return to this and address the issue in Table VI later.
The estimated coefficients from equations (2) and (3) in Table IV reveal several interesting patterns. First,the coefficients for the factor related exposures, reflected by the betas in Table IV, have decreased slightly compared to the interdependence model from equation (1) in Table II. This can be the result of the inclusion of the crisis dummy parameters as they can ‘dummy out’ the crisis periods leading to an overall reduction of comovements between portfolios before and after the crisis. Second, I find strong evidence for the rejection of the 𝐻0 of Hypothesis (A). I find strong evidence of factor-related contagion during the GFC, captured by 𝛾 parameters in Table IV. The importance of financial shocks from all three factors rose during the GFC. This is reflected in the positive and significant 𝛾 parameters in Table IV. The exposure to the global factor increased relatively the most during this crisis period, as a result of global contagion. Global exposure increased from 0.080 to 0.178 during the GFC, an increase of over 100%. Contagion from domestic sources increased slightly over 35%, while the contagion from Europe increased slightly over 20%. This offers a key insight that during a global crisis these exposures may increase significantly, although in normal times exposures towards certain factors may be small. Third, I also find strong evidence for the rejection of the 𝐻0 of Hypothesis (B). The positive and significant 𝜗 for the European factor in Table IV is evidence for European contagion during the SDC. The coefficient for 𝜗 reflects a rise in importance of European shocks during the SDC. Fourth, coefficients for the time dummies for the GFC and SDC, 𝜂𝑖 and 𝜍𝑖, are both negative with values of -0.555 and -0.133 and highly
significant, with the largest coefficient for the GFC. Especially during the GFC, contagion seems to be partially driven by factors unrelated to the three factors. This could be the result of herding behaviour of investors, or that a large group of investors face the same margin calls or liquidity constraints at the same time. This seems less the case during the SDC, where the effect is still significant but also relatively small when compared to the GFC. A reason for this can be that the first crisis affected the whole of Europe, while the second crisis mainly affected the southern European countries and Ireland. In all, the positive and significant estimates of the interdependence model with crisis time dummies are evidence for the occurrence of contagion during both crises periods. However, the changes in exposures and their severity were different for both crisis periods. This is likely due to the different origins of the two crises.
The breakdown of the coefficients from (2) and (3) per European region and industry in Table V reveals some interesting patterns.I will first discuss the coefficients displayed in Panel A of Table V. First, for Eastern Europe the exposure for the European factor, captured by 𝛽𝐸𝑈, now becomes
insignificant with the inclusion of the contagion parameters. This insignificant value confirms the earlier suggestion of a low degree of equity market integration for Eastern Europe with the Eurozone. During the two crises periods the importance of European shocks increased greatly for Eastern Europe, reflected in the large and significant values for 𝛾 and 𝜗 related to the European factor. This indicates that Eastern Europe experienced relatively more European contagion. Second, the coefficients for 𝛾, which capture contagion effects for the GFC, are significant for only two regions, Northern Europe and Eastern Europe, and only for the domestic and European factor. These two regions, with Eastern Europe in particular, have the lowest values for their 𝛽s for the European factor. This can be an indication that countries with low exposure to the European factor may be more susceptible to contagion from a crisis with its origination outside of the Eurozone. Third, all four regions experienced significant contagion unrelated to the three factors during the GFC, reflected in the negative and significant coefficients for 𝜂. During the SDC only Eastern Europe and Southern Europe experienced factor-unrelated contagion, as reflected by the coefficients for 𝜍. This form of contagion could be driven by the economic fundamentals of a country, as suggested by Bekaert et al. (2014).
I report the breakdown of the coefficients of (2) and (3) per industry in Panel B in Table V. The exposures of all industries to the domestic and European factors are significant, reflected by the 𝛽 coefficients related to these two factors. Not all industries experienced contagion related to the three factors during the two crisis periods. The coefficients for the crisis parameters related to the three factors, 𝛾 and 𝜗, remain insignificant for consumer services, health care and technology for all three factors. Other industries did experience a large rise in the importance of domestic shocks during the GFC. This is reflected by the highly significant and large coefficients for 𝛾𝐷, for consumer goods, industrials, and to a lesser extent oil and gas. These differences in contagion may be driven by the degree to how much an industry is affected by a shock in demand. What stands out is the importance of the European factor for the financials sector in both normal times, reflected in the 𝛽𝐸𝑈, and during crisis periods, reflected in 𝛾𝐸𝑈 and 𝜗𝐸𝑈. 𝜂 is significant and negative for all industries, which suggests that contagion unrelated to the three factors affected all industries during the GFC. During the SDC, 𝜍 is highly significant only for the financial sector, and significant to a lesser extent for industrials and utilities. Contagion unrelated to the three factors clearly influenced industries very differently over the two crisis periods. Contagion seems to mostly flow through industries that intuitively have the most linkages towards other countries, such as the import and export related industries and financials.
Table V
Contagion across Regions and Industries
The table reports the estimates of the interdependence model with crisis time dummies of Table IV, aggregated over regions and industries:
𝑅𝑖,𝑡= 𝐸𝑡−1[𝑅𝑖,𝑡] + 𝛽𝑖,𝑡′𝐹𝑡+ 𝜂𝑖𝐺𝐹𝐶𝑡+ 𝜍𝑖𝑆𝐷𝐶𝑡+ 𝑒𝑖,𝑡 (2)
𝛽𝑖,𝑡= 𝛽𝑖,0+ 𝛾𝑖,0𝐺𝐹𝐶𝑡+ 𝜗𝑖,0𝑆𝐷𝐶𝑡 (3)
The table depicts the estimates and standard errors using pooled OLS of Table IV, but on a regional (Panel A) and industry (Panel B) basis. Thus, the number of observations and the R² are the same. For the subscripts, D denotes the domestic factor, EU the European factor and G the global factor. ***, **. and * indicate the statistical significance of the estimates at the 1%, 5%, and 10% levels, respectively.
Panel A: By region Interdependence Contagion Region 𝛽𝐷 𝛽𝐸𝑈 𝛽𝐺 𝛾𝐷 𝛾𝐸𝑈 𝛾𝐺 𝜗𝐷 𝜗𝐸𝑈 𝜗𝐺 𝜂 𝜍 Eastern Europe 0.346*** 0.213 0.092** 0.322* 0.345** 0.161 0.141 0.166** 0.218* -0.841*** -0.350** Northern Europe 0.211*** 0.556*** 0.098 0.324** 0.163** 0.068 0.285*** 0.119* 0.100 -0.517** 0.039 Southern Europe 0.456*** 0.728*** 0.09* 0.101 0.087 0.127 0.139 0.198*** 0.021 -0.447*** -0.311*** Western Europe 0.166** 0.853*** 0.060 -0.015 0.070 -0.074 -0.028 0.039 0.001 -0.197** 0.063 Panel B: By industry Interdependence Contagion Industry 𝛽𝐷 𝛽𝐸𝑈 𝛽𝐺 𝛾𝐷 𝛾𝐸𝑈 𝛾𝐺 𝜗𝐷 𝜗𝐸𝑈 𝜗𝐺 𝜂 𝜍 Basic materials 0.317** 0.703*** 0.163* 0.101 0.174** 0.261* 0.044 0.223*** 0.234** -0.538*** -0.073 Consumer goods 0.192*** 0.496*** -0.093 0.327*** 0.176** 0.326** 0.043 0.055 0.310** -0.602*** -0.011 Consumer services 0.304*** 0.606*** 0.103** 0.184 0.066 0.163 0.095 0.121 -0.022 -0.514*** -0.032 Financials 0.279*** 0.700*** 0.053 0.234* 0.406*** -0.057 0.303*** 0.333*** 0.044 -0.421** -0.314*** Health care 0.354*** 0.610*** 0.038 0.022 0.082 -0.056 0.094 0.003 0.070 -0.591*** -0.096 Industrials 0.309*** 0.616*** 0.142*** 0.394*** 0.257*** 0.028 0.298** 0.188*** 0.300*** -0.624*** -0.120** Oil & gas 0.435*** 0.590*** 0.135** 0.109** 0.185** 0.062 -0.135 0.102 0.174** -0.421*** -0.153
Technology 0.231* 0.675*** 0.251*** 0.257 -0.079 -0.123 -0.014 -0.054 -0.221 -0.325* 0.174
Table VI
Actual and Fitted Returns
The table reports the actual equity market returns per country, over both crisis periods (from 2007Q3 to 2009Q3 and from 2010Q2 to 2012Q3). The actual returns are compared to the fitted returns of the models presented in Tables II and IV. The portfolio returns displayed in the table are unweighted averages within countries. The countries are ranked on their actual equity returns during the crises. The model parameters shown are from the contagion model.
Interdependence
Model Contagion model
Actual returns Fitted returns Fitted returns Model parameters
Country Returns Rank Returns Rank Returns Rank 𝛽𝐷 𝛽𝐸𝑈 𝛽𝐺 𝛾𝐷 𝛾𝐸𝑈 𝛾𝐺 𝜗𝐷 𝜗𝐸𝑈 𝜗𝐺 𝜂 𝜍
sample period, but the fitted and actual returns are displayed for the crisis periods only. The fitted returns of the models can be compared to each other and to the actual returns during the crisis period in Table VI. I do the computation of the fitted returns by estimating equation (1) for the interdependence model and equations (2) and (3) for the interdependence model with crisis time dummies. After the estimation, I obtain the 𝑅̂𝑖,𝑡 for each portfolio i at time t. The total returns, 𝑅̂𝑐, are then obtained for each country from the unweighted average returns of each industry within the country during the crisis periods. Then, the estimated 𝑅̂𝑐 can be compared to 𝑅𝑐, the actual returns per country. This allows me to compare the performance of the interdependence and contagion models during crisis periods.
I report the actual and fitted returns of equations (1) to (3) in Table VI. I also report the estimates of the interdependence model with crisis time dummies, equations (2) and (3), per country to explore further variance in the country-industry portfolios. In the first set of columns of the table, the countries are ranked on their actual performance during the crisis periods from worst to best. Thus, countries with larger decreases in their returns during the crises periods rank higher. The second and third sets of columns show the predicted returns from the interdependence model of (1), the model with crisis time dummies (2) and (3), and their rankings according to their fitted returns. The countries affected most by the crises, reflected by the largest declines in actual returns, are those situated in the Eastern and Southern Europe. Both models seem to be insufficient in predicting the returns during the crisis. The rankings based on the predictions of both models are not in line with the ranking based on the actual values. The interdependence model with crisis time dummies provides fitted returns closer to the actual returns. This justifies the addition of these dummies to the interdependence model.
European Integration
In the previous section I focussed on whether contagion occurred in the Eurozone during the two crisis periods. Furthermore, I provided insights on the exposures of different European regions and industries, and how they were affected during the two crises. In this subsection, I turn to the role of financial and economic integration and how it influences the transmission of financial shocks as reflected by the three factors, and I test for Hypothesis (C) described in Section II. I test whether country-specific financial and economic integration within the Eurozone affects the exposures related to the three factors. Furthermore, I test whether financial and economic integration influenced contagion related to the three factors. To test for this, I estimate the full model as presented in equations (4) to (7) in Section II.
The estimated coefficients of equations (4) to (7) are reported in Table VII. I estimate the system of equations in quarterly frequency instead of weekly frequency, due to the quarterly frequency of the variables in 𝑍𝑖,𝑡−𝑘. I discuss the variables in 𝑍𝑖,𝑡−𝑘 in detail in Section II. I adjust the variable for the excess returns for this change in frequency, by taking the total return over the last quarter.
I report the coefficients for interdependence, reflected by 𝛽𝑖,0, in Table VII. The significance of the
coefficients changes little after the inclusion of 𝑍𝑖,𝑡−𝑘, when these are compared to the 𝛽s of equation (2) and (3) in Table IV. The coefficients for interdependence for all three factors remain significant, but the coefficient for the global factor loses some of its significance. The size of the coefficients for interdependence in Table VII does change, as the variables in 𝑍𝑖,𝑡−𝑘 now also influence the total exposure to the factors. Also, using a quarterly rather than weekly frequency in the data is likely to have an influence on the results.
The coefficients for the cross-border MFI loans in Table VII are highly significant for the domestic and European factor. The coefficient for the domestic factor is negative, which indicates that countries with high MFI loans have lower domestic exposures. The reverse is true for the exposure of the European factor, countries with more MFI loans have a higher exposure to the European factor. This potentially validates the theory that international lending allows MFIs to better diversify away their idiosyncratic risks from countries and to become less reliable on the domestic factors. The coefficients for the interaction of cross-border Eurozone trade with the three factors are insignificant for the European and global factors. Hence, they are not likely to have an important influence on the exposure to these two factors. Baele and Inghelbrecht (2009) find that increased integration through trade has a positive effect on European betas. However, they do not include country-specific exposures in their research, but only regional and global exposures. They argue that this result is likely due to increased cross-border participation of international firms, or due to the convergence in cash flow shocks. Bekaert et al (2013) use the same argumentation. The coefficient for cross-border trade interacted with the domestic factor is negative and significant. Countries that are more (less) engaged in cross-border trade experience a lower (higher) exposure to their domestic factor. A country’s economy becomes less dependent on domestic factors if total trade as a percentage of GDP increases. The coefficient for the interaction of misalignment regarding the European factor has a negative coefficient and is highly significant. Coefficients for the other two factors remain insignificant. Countries with an industry composition substantially different from that of the rest of the Eurozone, have a lower exposure to European shocks. I use the three variables in 𝑍𝑖,𝑡−𝑘 in (5) as proxies for financial and economic integration, in order to answer Hypothesis (C). Clearly, as discussed above, these three variables influence the transmission of domestic and European shocks.
Now, I will discuss the results of the interaction of the variables in 𝑍𝑖,𝑡−𝑘 interacted with the three factors and the crisis time dummies in equations (6) and (7). I present these results in Table VII. This allows me to answer the second part of Hypothesis (C), whether financial and economic integration influenced the degree to which countries experienced contagion. First, I will discuss the coefficients of 𝑍𝑖,𝑡−𝑘 in (6), for the Great Financial Crisis (GFC). Then, I discuss the results of equation (7) for the Sovereign Debt Crisis (SDC).
Table VII
Interdependence, Contagion and European Integration
The table reports the estimates of the full model by pooled OLS, as denoted in (4) to (7). The table depicts the estimates of the unweighted average degree of interdependence and contagion of all portfolios in the sample. The model is corrected for heteroskedasticity by using country-clustered standard errors. For the subscripts, D denotes the domestic factor, EU the European factor and G the global factor. ***, **. and * indicate the statistical significance of the estimates at the 1%, 5%, and 10% levels, respectively.
Domestic Factor European Factor Global Factor
Coeff. St. Err. Coeff. St. Err. Coeff. St. Err. Coeff. St. Err.
Constant 0.336** 0.157 𝐸𝑡−1[𝑅𝑖,𝑡] 0.891*** 0.029 𝛽𝑖,0 0.610*** 0.144 1.450*** 0.181 0.460* 0.251 𝛽1∗ MFI Loans -0.134*** 0.021 0.169*** 0.028 0.027 0.053 𝛽1∗ Trade -0.438** 0.199 -0.354 0.295 -0.287 0.320 𝛽1∗ Misalignment 0.538 0.456 -1.398*** 0.474 0.119 0.811
Great Financial Crisis -2.357*** 0.466 0.515 0.434 0.024 0.184 1.812*** 0.479
𝛾1∗ MFI Loans 0.128 0.135 -0.016 0.045 0.363*** 0.116
𝛾1∗ Trade 0.134 0.176 -0.006 0.179 -0.036 0.338
𝛾1∗ Misalignment -1.515 1.037 -0.306 0.699 -3.081* 1.574
Sovereign Debt Crisis -1.456*** 0.144 -0.384 0.232 -0.311 0.216 -1.260** 0.592
𝜗1∗ MFI Loans -0.182*** 0.055 -0.130*** 0.043 -0.036 0.158
𝜗1∗ Trade 0.143 0.412 -0.047 0.264 1.907** 0.895
𝜗1∗ Misalignment 1.181 1.081 0.820 0.786 -0.315 1.906
Observations 10.098
MFI loans are positively related to the importance of the global factor during the crisis and had an amplifying effect on the transmission of global shocks. Rather than through Eurozone financial integration, this result could be due to increased financial globalization. Cross-border MFI loans could proxy for the total amount of loans or for the loans with the US, as intra-Eurozone lending and lending to the US or global market are likely to be positively related. Misalignment of industries has a negative coefficient for the interaction with the GFC. This provides some evidence that a misaligned industry indeed allows the country to experience a decreased transmission of shocks and contagion from outside of the country. If further integration of European markets causes industries to become more similar to each other, then this may increase the susceptibility of the countries equity market with respect to European shocks.
The only significant coefficient for the interaction of the SDC with the three factors is for the global factor. This coefficient is negative, which reflects that during the SDC the importance of the global factor actually decreased in this period. Only countries with high trade as a percentage of GDP experienced increased exposure to the global factor, as reflected in the positive and significant coefficient for the interaction term for cross-border trade and the SDC. The amount of Intra-Eurozone trade and total cross-border trade for a country are likely to be positively related. Countries with higher Eurozone trade as a percentage of GDP are also likely to trade more with countries outside the Eurozone. Therefore, these countries also experience a higher exposure to the global factor. The coefficients for the cross-border MFI loans are negative and significant for domestic and European factors during the SDC. The negative coefficients show that domestic and European exposures decreased for countries with larger Eurozone loan portfolios. The negative relation between MFI loans and the domestic and European exposures can be due to the better possibilities to diversify idiosyncratic risks for countries with higher MFI loans.
Robustness
In the Appendix I report two robustness tests for the full model, in order to test for the robustness of the results. For the first robustness test, presented in Table C1 in the Appendix, I exclude the interaction of the crisis time dummies 𝐺𝐹𝐶𝑡 and 𝑆𝐷𝐶𝑡 with the three factors of the model. Hence, the model only contains interactions which include 𝑍𝑖,𝑡−𝑘. This means that the model only captures contagion related to the variables in 𝑍𝑖,𝑡−𝑘. Furthermore, I lag the variables in 𝑍𝑖,𝑡−𝑘 by only one period. The magnitude of the coefficients for most of the coefficients change, but the signs remain the same. Also, the significance of most factors stays the same. The exception is the interaction of 𝑍𝑖,𝑡−𝑘 with the global factor during the GFC. Cross-border trade now also becomes weakly significant, as reflected by its coefficient. Misalignment loses its significance, but its sign doesn’t change. I present the second robustness test in Table C2 in the Appendix. For this test, I exclude contagion from the global factor during the SDC. I exclude this form of contagion as the origin of this crisis lies within Europe, thus contagion from the global factor may be counterintuitive. Again, the coefficients change in magnitude, but the signs for the coefficients remain the same. Significance of the coefficients only changes for variables in the two crisis periods. First, cross-border trade now becomes weakly significant when interacted with the global factor during the GFC. Second, the coefficient for the European factor interacted with the SDC time dummy becomes negative and significant. This suggests that the European factor actually became less important during the SDC. In all, the coefficients for the variables without interaction with the crisis time dummies do not change sign or significance when I make small changes to the model, or to the number of lags for 𝑍𝑖,𝑡−𝑘. For the two crisis periods, the significance and signs of some of the coefficients does change. This is likely, as the changes to the model mostly impact the variables related to the crisis periods. The coefficients for MFI loans do not change sign or significance, while the coefficients for Misalignment and cross-border trade change in significance.
In this section I have discussed the results of the interdependence model in equation (1) and the contagion model, as in (2) and (3) in order to assess Hypotheses (A) and (B). Also, I explored the variance in the coefficients of these two models, based on European regions and industries. Then, I provided the performance of the two models, and included the coefficients of the contagion model per country. Then, I discussed the effects of integration in the Eurozone on the factor exposures and on contagion during the two crisis periods in order to discuss Hypothesis (C). Finally, I presented two robustness tests for the full model.
crisis periods. For both hypotheses I rejected the 𝐻0, and concluded that countries in the Eurozone experienced contagion during both crisis periods. The GFC affected all regions and industries through factor unrelated contagion, while the SDC mostly affected Southern and Eastern Europe. I found that countries with relatively low equity market integration towards the European factor experienced the greatest degrees of contagion during the GFC. This was confirmed by the results of the analysis per country in Table VI, where overall countries with low European exposures experienced highest degrees of both factor related and non-factor related contagion. It seems that greater integration into a single equity market, reflected by the European exposure, leads to decreases in contagion. When I compared the performance of the two models, the model of equation (2) and (3) estimated returns that were more in line with reality. Then, I introduced the full model of equations (4) to (7) in order to test Hypothesis (C). In the full model I investigated the role of Eurozone integration on the exposures and contagion of the country-industry equity portfolios. I rejected the 𝐻0 and concluded that financial and economic integration, proxied by three variables, does play a role in the transmission of financial shocks. Furthermore, financial and economic integration influenced the degree to which contagion occurred during the two crisis periods. Financial integration, proxied by cross-border Eurozone MFI loans, amplified the transmission of global shocks during the GFC. During the SDC financial integration reduced the importance of domestic and European shocks. The results for financial integration are robust to changes in the model. Trade integration reduces exposure to the domestic factor, while during the SDC it amplified the transmission of global shocks. Countries with more misaligned industries experience lower exposures to the European factor, while during the GFC these countries experienced decreased contagion from global shocks. In all, I find that financial and economic integration influences the exposures of a country, and the transmission of shocks in crisis periods.
V. Conclusion
Since the construction of the Euro, European policymakers have been pursuing and advocating policies regarding the integration of markets of European countries into a single market. The discussion on financial and economic integration intensified with the occurrence of the Great Financial Crisis, as some blamed increased international financial and economic linkages for the transmission of contagion. In this paper I use a dataset with the returns of synthetic return portfolios in 24 Eurozone countries from 2001 to 2018 to study whether and where contagion occurred in the Eurozone. Furthermore, I study the role of financial and economic integration on the transmission of shocks and contagion.
markets and the European equity market to the domestic equity portfolios. Furthermore, I find strong evidence of contagion from European equity markets during the Sovereign Debt Crisis. Second, I find that countries with lower exposures to the European factor experience greater instances of contagion during the crisis periods. The smallest changes in exposures during crises periods occurred predominantly for countries with higher European factor exposures. Third, I find evidence that financial and economic integration plays an important role in the transmission of equity market shocks. The proxies for integration capture either a positive relation with the European factor, or a negative relation with the domestic factor. Financial integration also played a role in the transmission of global contagion during the Great Financial Crisis, where it had an amplifying effect on global equity market shocks, while during the Sovereign Debt Crisis it reduced the transmission of contagion from domestic and European equity markets. Finally, contagion unrelated to the three factors played an important role in both crisis periods.
REFERENCES
Bacchetta, Phillipe, and Eric van Wincoop, 2016. “The Great Recession: A Self-Fulfilling Global Panic.” American Economic Journal: Macroeconomics, 8(4): 177-198.
Baele, Lieven, 2005. “Volatility Spillover Effects in European Equity Markets.” Journal of Financial and Quantitative Analysis, 40(2): 373-401.
Baele, Lieven, and Koen Inghelbrecht, 2009. “Time-varying Integration and International Diversification Strategies.” Journal of Empirical Finance, 16(1): 368-387.
Baele, Lieven, Annalisa Ferrando, Peter Hördahl, Elizaveta Krylova, and Cyril Monnet, 2004. “Measuring Financial Integration in the Euro Area.” European Central Bank Occasional Paper Series 14.
Balli, Faruk, Syed Abul Basher, and Hatice Ozer-Balli. 2010. “From home bias to Euro bias: Disentangling the effects of monetary union on the European financial markets.” Journal of Economics and Business, 62(1): 347-366.
Bekaert, Geert, Campbell R. Harvey, and Angela Ng, 2005. “Market integration and Contagion.” Journal of Business, 78(1) 39-69.
Bekaert, Geert, Campbell R. Harvey, Christian T. Lundblad, and Stephan Siegel, 2013. “The
European Union, the Euro and equity market integration.” Journal of Financial Economics, 109(1): 583-603.
Bekaert, Geert, Michael Ehrmann, Marcel Fratzscher, and Arnaud Mehl, 2014. “The Global Crises and Equity Market Contagion.” The Journal of Finance, 69(6): 2597-2649.
Bekaert, Geert, Robert J. Hodrick, and Xiaoyan Zhang, 2009. “International Stock Return Comovements.” The Journal of Finance, 64(6): 2591-2626.
Bley, Jorg, 2009. “European Stock Market Integration: Fact or Fiction?” Journal of International Financial Markets, Institutions & Money, 19(1): 759-776.
Bosch, Thijs, and Dirk Schoenmaker, 2008. “Is the home bias in equities and bonds declining in Europe?” Investment Management and Financial Innovations, 5(4): 90-102.
Chari, Anusha, Peter Blair Henry, and Diego Sasson, 2012. “Capital Market Integration and Wages.” American Economic Journal: Macroeconomics, 4(2): 102-132.
De Grauwe, Paul, 2013. “Design Failures in the Eurozone: Can they be fixed?” European Economy Economic Papers 491.
De Grauwe, Paul, 2016. “Economics of Monetary Union.” Oxford University Press, Eleventh Edition. Eichengreen, Barry, Andrew K. Rose, and Charles Wyplosz, 1996. “Contagious Currency Crises.”
National Bureau of Economic Research (NBER) Working Paper 5681.
Garcia-Herrero, Alicia, and Philip Wooldridge, 2007. “Global and regional financial integration: progress in emerging markets.” BIS Quarterly Review, September: 57-70.
Hardouvelis, Gikas A., Dimitrios Malliaropulos, and Richard Priestley. 2006. “EMU and European Stock Market Integration.” Journal of Business, 79(1): 365-392.
Imai, Masami, and Seitaro Takarabe. 2011. “Bank Integration and Transmission of Financial Shocks: Evidence from Japan.” American Economic Journal: Macroeconomics, 3(1): 155-183.
Inklaar, Robert, Richard Jong-A-Pin, and Jakob de Haan, 2008. “Trade and Business Cycle Synchronization in OECD Countries – A re-examination.” European Economic Review, 52(1): 646-666.
Japelli, Tullio, and Marco Pagano, 2008. “Financial Market Integration under EMU” European Economy Economic Papers 312.
Kaminsky, Graciela L., and Carmen M. Reinhart, 2000. “On Crises, Contagion and Confusion.” Journal of International Economics, 51(1): 145-168.
APPENDIX
Table A
Stock indices per Country
This table shows the number of firms used in the construction of the country-industry equity portfolio, and the index on which the firms are traded.
Country Name of stock index No. of listed firms Country Name of stock index No. of listed firms
Austria ATX 20 Ireland ISEQ 45
Belgium BEL20 20 Italy MIB 30
Bulgaria SOFIX 15 Latvia OMX 35
Croatia CROBEX 23 Lithuania OMX 30
Czech Republic PSE 12 Luxembourg LuxX 9
Denmark OMX20 20 Netherlands AEX 25
Estonia OMX 10 Portugal PSI 20 18
Finland OMX25 25 Romania BET 15
France CAC 40 40 Slovenia SBI 9
Germany DAX 30 Spain IBEX35 35
Greece Athex 20 25 Sweden OMX30 30
Hungary BUX 16 UK FTSE100 100
Source: Datastream. Note: The broad sectors that are used for the industry classifications in the value weighted country-industry equity portfolios are the following: (i). Basic Materials, (ii). Consumer Goods (iii). Consumer Services (iv). Financials (v). Health Care (vi). Industrials (vii). Oil & Gas (viii). Technology (ix). Telecommunications (x). Utilities.
Table B
Regions and their corresponding countries
This table shows the regions, and their corresponding countries.
Region Countries
Eastern Europe Bulgaria, Czech Republic, Hungay, Romania, Slovenia
Northern Europe Denmark, England, Estonia, Finland, Latvia, Lithuania, Sweden Southern Europe Croatia, Greece, Italy, Portugal, Spain
Table C.1
Robustness Test excluding Crisis Interactions with the factors
The table reports the estimates of the full model by pooled OLS, as denoted in (4) to (7) but excludes the interaction of the GFC and SDC time dummies with the three factors. The table depicts the estimates of the unweighted average degree of interdependence and contagion of all portfolios in the sample. The model is corrected for heteroskedasticity by using country-clustered standard errors. For the subscripts, D denotes the domestic factor, EU the European factor and G the global factor. ***, **. and * indicate the statistical significance of the estimates at the 1%, 5%, and 10% levels, respectively.
Domestic European Global
Coeff. St.Err. Coeff. St.Err. Coeff. St.Err. Coeff. St.Err.
Constant 0.348** 0.158 𝐸𝑡−1[𝑅𝑖,𝑡] 0.889*** 0.029 𝛽𝑖,0 0.629*** 0.106 1.282*** 0.133 0.404** 0.165 𝛽1∗ MFI Loans -0.135*** 0.021 0.167*** 0.033 0.028 0.054 𝛽1∗ Trade -0.444** 0.207 -0.264 0.29 -0.263 0.289 𝛽1∗ Misalignment 0.503 0.389 -1.095** 0.429 0.240 0.775
Great Financial Crisis -2.262*** 0.468
𝛾1∗ MFI Loans 0.119 0.161 -0.042 0.041 0.216** 0.103
𝛾1∗ Trade 0.540* 0.300 0.033 0.226 1.205* 0.605
𝛾1∗ Misalignment -0.868* 0.480 -0.218 0.544 -0.213 1.581
Sovereign Debt Crisis -1.498*** 0.150
𝜗1∗ MFI Loans -0.196** 0.078 -0.127** 0.050 0.041 0.190
𝜗1∗ Trade 0.148 0.413 -0.227 0.258 1.326 0.826
𝜗1∗ Misalignment 0.192 0.881 0.295 0.724 -2.765 1.803
Observations 10.098
