Trade,Industrial Dissimilarity,FDI and Business Cycle Co-movements: EC3SLS Evidence from Eurozone Economies

Abstract

This article uses simultaneous equations error component three-stage least squares (EC3SLS) panel data technique to find the direct and indirect impacts of trade, industrial dissimilarity and FDI on the business cycle synchronisation of Eurozone economies. The period of analysis is 1990–2009. These are the major findings: (a) trade, industrial dissimilarity and FDI have both direct and indirect effects on the business cycle synchronisation of sample economies; (b) EC3SLS estimates show that closer trade ties among these Eurozone countries have led to more synchronised business cycle co-movements, because common disturbances are more prevalent and intra-industry trade dominates; (c) the bilateral FDI flows have served as a source of destabilisation rather than a source of synchronisation; (d) industry-specific shocks have almost lost their importance both in terms of generating more trade and in terms of raising the output correlation of sample countries; (e) trade intensity and FDI flows are positively and significantly correlated, thereby suggesting that more FDI encourages more trade and vice-versa; (f) trade shows a negative relationship with industrial dissimilarity, which implies that bilateral trade in these countries promotes similar industrial structures; and finally, (g) these economies characterise intra-industry trade patterns, as expected in the case of these highly integrated Eurozone economies. Although the results show the presence of intra-industry trade types, these economies have diversified their production processes at the higher income levels.

JEL: E32, F02, F10, F11, F15, C33

Keywords

Business Cycle Co-movements Trade Intensity Industrial Dissimilarity FDI Eurozone Economies EC3SLS

1. Introduction

The theory of optimum currency areas (OCAs) proposes a broad set of prerequisites that a geographic region needs to satisfy to form a currency union.¹ In this pursuit, many treaties and convergence policies (starting with the Marshall Plan of 1948 and going up to the Delors Plan of 1989 and the Maastricht Treaty of 1991) were initiated, prior to the establishment of the euro as a common currency in Europe. The only guideline for prospective entrants (of the Eurozone) was: a potential entrant would find it more beneficial to abandon its independent monetary policy and join a monetary union, only if it is subject to symmetric economic shocks (Bayoumi & Eichengreen, 1993, 1997; Masson & Taylor, 1993). However, the argument did not last long and was soon replaced with endogenous OCA criteria (Frankel & Rose, 1997, 1998). The endogenous OCA theory argues that international trade patterns and business cycle correlation are endogenous. But international trade is not the only channel of business cycle transmission. For instance, shock transmission can take place through industrial structures, similar fiscal policies, foreign direct investment (FDI), currency unions, monetary and financial integration, exchange rate volatility, and economic integration. The theoretical interactions among these determinants of international business cycle synchronisation are very complex. It is these interactions that are explored in this study. More specifically, this article analyses the impact of trade, industrial dissimilarity and FDI on the business cycle synchronisation of Eurozone economies using a panel data set spanning the period 1990–2009. Disentangling the relative contributions of trade, FDI and industrial dissimilarity is crucial from the point of view of business cycles research. It is also a relevant policy question, in the sense that the international correlation of business cycles is an important metric used to measure the desirability and feasibility of a potential entrant joining a currency union.² More generally, the channels this study propose to determine are relevant to policymakers asking if, and why, they should be concerned with foreign developments affecting domestic fluctuations.

The rest of the article is organised as follows. In the second section, an extensive review of the related literature is carried out. The detailed review helps locate the major research gaps in the existing literature. In the third section, the econometric model, the formulation of variables along with their data sources and the empirical strategy used in this article are discussed. In the fourth section, the estimation results of Models (1) and (2) are analysed. Finally, in the last section some concluding remarks are offered.

2. Extensive Literature Review and Major Research Gaps

Theoretically, the relation between trade integration and business cycle synchronisation is ambiguous, because greater trade integration can transmit demand shocks occurring in one country to another country, thereby increasing synchronisation. It can also lead to specialisation in production (Dornbusch, Fisher, & Samuelson, 1977). Specialisation will promote industry-specific shocks, thereby reducing synchronisation (Krugman, 1993; Kose & Yi, 2002). The same applies to the relationship between financial integration or FDI and business cycle synchronisation. Given the ambiguity of economic theory on these issues, a vast empirical literature has emerged to study the effect of these macro-economic variables on the business cycle synchronisation of economies. Overall, these studies tend to provide evidence of a positive relation between economic integration and synchronisation, especially for highly developed economies (see, e.g., Baxter & Kouparitsas, 2005; Clark & van Wincoop, 2001; Frankel & Rose, 1998; Gächter & Riedl, 2014; Gonçalves, Rodrigues, & Soares, 2009; Hanus & Vacha, 2015; Hsu, Wu, & Yau, 2011; Imbs, 2004; Inklaar, Jong-A-Pin & de Haan, 2008; Kolasa, 2013). A detailed chronological review of the empirical literature based on business cycle synchronisation is carried out in Table 1. Tables 1–7 report the results. The Table 1 provides information on the study objective, sample data set, sample period, estimation technique used to answer the respective objective and the major findings of the study. An appraisal of the empirical literature (given in Table 1) reveals that although theoretical as well as empirical studies have produced somewhat mixed results, some patterns in the existing literature on business cycle synchronisation are easily observable. First, the overall effect of trade on business cycle synchronisation is strong and positive. Second, similarity in industrial structures show mixed impact on business cycle synchronisation. Finally, business cycles in regions at higher levels of financial integration are significantly more synchronised. The studies mentioned in Table 1 have enriched the literature on business cycle synchronisation. Yet the indirect effects of major macro-variables on the business cycle synchronisation of Eurozone economies have been largely ignored; most studies analyse the co-movements of macro-economic variables in the European region using the structural vector autoregression (SVAR) technique. The SVAR technique was developed as an alternative to traditional macro-econometric modelling in 1980. The earlier SVAR methodology was atheoretical in nature and the procedure was almost mechanical. Although the later developments in SVAR technique widely adopted identifying assumptions, these orthogonal conditions/assumptions related to underlying shocks in SVAR technique are fairly restrictive. The need is to explore this issue using panel econometric techniques; the studies of business cycle synchronisation clearly indicate that the issue of synchronisation still requires improvement not only in terms of econometric analysis but also in the way the research has been carried out so far.

Table 1
Empirical Studies on Business Cycle Synchronisation

Year Author(s) Objective of the Study Sample Data set Estimation Technique Major Findings

1989 Blanchard and Quah Supply disturbances and demand disturbances are used to interpret fluctuations in output and unemployment. US real GNP growth and unemployment rate data; sample period (1950:Q2–1987:Q4). VAR system; variance decomposition Demand shocks create hump-shaped effect on output and unemployment, which disappears approximately after 2–3 years. The supply shocks have an effect on output that cumulates over time and reaches its peak after 5 years. The study drew huge attention and laid the foundation for future research in this area. Assumption: Supply disturbances have permanent effect on output while demand disturbances do not have. Note: GNP data are quarterly. Monthly unemployment data are averaged to provide quarterly observations.

1993 Bayoumi and Eichengreen To investigate the future path of European Commission (EC) as a monetary union and hence to extract the exact information on underlying aggregate demand and supply shocks 12 EC members and 11 OECD countries (annual sample data from 1960–1988); US regions data (1963–1986) Vector autoregression (VAR) decomposition technique The findings suggest that the underlying shocks are significantly more idiosyncratic across EC countries than across US states, thereby making it difficult to operate as a monetary union.

1998 Frankel and Rose To study an empirical relationship between the extent of trade and the similarity of the shocks and business cycles Panel of 30 years data for 21 industrialised countries Cross-country correlations; HP-filtered and differencing detrending techniques The results—along with the robustness checks—provide a strong positive relationship between the degree of bilateral trade intensity and the cross-country bilateral correlation of business cycle activity. That means, higher trade integration leads to more synchronised business cycles. This endogenous nature of the relationship between the OCA criteria serves the authors as a straightforward application of the Lucas critique. That is, a potential member is more likely to satisfy the OCA criteria for entry into a currency union ex post than ex ante.

2004 Horvath and Rátfai To analyse the relative incidence of demand and supply shocks in both the participating and the accession EMU countries 8 EMU accession economies and reference group (Germany, France, Italy); sample period (1993:Q1–2000:Q3), except Hungary (1995:Q1–2000:Q3) Bivariate vector auto regression models While the results show significant degree of symmetry in business cycle shocks among EMU accession economies, the idiosyncratic shocks between current and would-be member countries dominate. More specifically, the shocks among EMU accession and reference group economies, in particular, Germany, tend to be uncorrelated. These results have an important policy bearing: it could be costly for the participants to proceed with the union.

2004 Imbs To access empirically the relative impact of trade intensity, financial integration and sectoral specialisation on business cycle correlations of economies 18 developed countries group; quarterly data (1983:Q1–1998:Q3). Simultaneous equations model; 3SLS; FEM and REM techniques; EC3SLS The results reveal a strong intra-industry trade link with business cycles correlations, and a weak link between inter-industry trade and cycles correlations. However, the overall effect of trade on business cycle synchronisation is strong and positive. Same is the case of specialisation on business cycles correlation. Also, business cycles in regions at higher levels of financial integration are significantly more synchronised. It should be noted that one of the motivations for such a study was that international correlations of business cycles serves an important parameter to determine the desirability and feasibility of a currency union.

2005 Baxter and Kouparitsas To find out the determinants of business cycle co-movement between countries Data set of over 100 countries (both developing and developed); major sample period 1970–1995 ‘Extreme bounds analysis’ applied to band-pass filtered data The study finds that bilateral trade between two countries is highly correlated with the business cycles of economies, while similarity in their industrial structures does not lead to business cycle synchronisation. Same is the case of currency union variable. Among the ‘gravity’ variables, distance is negatively related to business cycle correlation, as one can expect. By taking a large number of variables, the different variants of indexes, and a large number of robust estimations, the study serves benchmark guidance for future research in this area.

2007 Calderón, Chong, and Stein To see the impact of trade intensity on business cycle synchronisation among developing and industrial countries Annual data for 147 countries spanning the period 1960–1999 Cross-sectional and panel data techniques imposed on a number of indexes The study finds that the impact of trade intensity on business cycle synchronisation of developing economies is significantly positive, but smaller than that of industrialised countries. This difference arises because of the differences in their patterns of specialisation and bilateral trade.

2008 Darvas and Szapáry To study the prevailing degree of synchronisation between the countries from Central and Eastern European Countries (CEECs), and also check the OCA endogeneity of the euro currency union (CU) members Hodrick-Prescott (HP) and the Band-Pass (BP) filters; dynamic factor models using maximum likelihood (ML) estimation and Kalman-filtering of the state-space representation 8 CEECs, 10 EMU members and various other countries used as control group; quarterly data (1983–2002) grouped in 4 non-overlapping 5-year periods (1983–1987, 1988–1992, 1993–1997, 1998–2002) The major findings of this study are (a) Hungary, Poland and Slovenia are most synchronised, (b) Czech Republic and Slovakia are least synchronised, (c) the Baltic states (Estonia, Latvia, Lithuania) are not synchronised at all during the entire period of study, and (d) the core EMU countries (Italy, Germany, Belgium, France, Austria, the Netherland) exhibit higher synchronisation in their business cycles, while the periphery countries (Finland, Ireland, Portugal, Spain) show lower co-movements. However, the periphery countries are moving towards the higher synchronisation level of the core EMU countries.

2008 Bussière, Fidrmuc, and Schnatz To check the integration of Central and South Eastern European countries (CSEECs) with the Euro area (EA) Annual data set of 61 countries spanning from 1980–2003 Gravity models using panel fixed and random effects technique The gravity model results suggest: rapid trade integration of the CSEECs and the EA; and a gradual convergence to ‘normal’ trade levels.

2008 Inklaar, Jong-A-Pin, and de Haan To re-investigate the relationship between intensity of trade and business cycle synchronisation 21 OECD economies; 1970–2003 sample—further divided into 3 sub-samples (1970–1981, 1981–1992, 1992–2003) Multivariate model; extreme bounds analysis; OLS and instrumental variables techniques The study confirms a positive impact of trade intensity, specialisation and convergence in fiscal and monetary policies on business cycle synchronisation. The merit (of this study) lies in its incorporation of sample heterogeneity and outliers in the analytical framework.

2009 Baltagi, Demetriades, and Law To empirically examine the relationship of trade and financial openness with the financial development of economies Using annual panel data from developing and industrialised countries Dynamic panel estimation techniques The estimated results support the hypothesis that trade and financial openness affect positively and significantly the banking sector development. The results also suggest that the marginal effects of trade openness are negatively related to the degree of financial openness and vice-versa. Thus, this research provides a partial support to the Rajan-Zingales hypothesis (this hypothesis states that both the trade and financial openness are necessary for financial development). The study has an important policy implication: relatively closed low-income countries can benefit most from opening up their trade and/or capital accounts.

2009 Gonçalves, Rodrigues, and Soares To compare post-euro adoption bilateral business cycle correlations of EMU members with the OECD economies 22 OECD and Eurozone countries; major data set (1980–2007) Differences in-differences technique The results confirm that bilateral correlations of the business cycle have increased more amongst the EMU members as compared to the OECD economies after the adoption of the euro.

2009 Cerqueira and Martins To find out the determinants of business cycle synchronisation (BCS) using panel approach 20 OECD countries; Major data set (1970–2002) GDP cross-correlation index for panel data that captures time-variability Major findings: (a) bilateral trade and EU membership have a positive effect on BCS, (b) industrial dissimilarity is non-significant, and (c) financial openness has negative effect on BCS.

2010 Hsu To answer the question: is it feasible for East Asia to form a common currency area? Or Does a common symmetric shock dominate East Asian economies? A post-Asian crisis data set (1997–2007); one group of economies belong to East Asia while the second group belongs to the EA The study uses a 3-variable SVAR model to identify three kinds of shocks: global, regional and domestic The results suggest (a) a sharp decline of asymmetric domestic shocks in the post-crisis period, (b) most East Asian countries respond to regional and global shocks in a symmetric way, and (c) feasibility of common CU in East Asia in the near future. However, the study provides a limited picture of the influence of global shocks on the East Asia. It would have been of much more interest had the study considered the impact of shocks on East Asia originating from major economies of the world, such as the US and the UK.

2010 Savva, Neanidis, and Osborn To examine business cycle correlations between the EA and the currently negotiating and recently added EU states 9 enlargement countries, 5 core and 4 periphery EA countries, Turkey, Croatia; Monthly data (1980:M1–2006:M6) Bivariate VAR-GARCH The findings reveal that all the negotiating and newly added EU countries have increased their business cycle synchronisation with the EA more than two times, or at least have changed from negative to positive correlations, since the early 1990s. The study adds to the existing literature in two broad ways as follows. First, while the existing literature only considers either constant correlation coefficients or sub-samples divided using assumed exogenous break dates; this study determines the timing and the length of structural shifts endogenously. Second, this study considers the possibility of multiple transition paths, thus allowing for more complex correlation regimes.

2011 Hsu, Wu, and Yau Add-on the simultaneous equation model of Imbs (2004) by including FDI as one more possible channels of shock transmission Cross-sectional as well as panel-data from 1998 to 2002 on G7 countries and 8 major OECD countries The study estimates single equation as well as error-components three stage least squares (EC3SLS) simultaneous equation model The main findings of the study are as follows. First, EC3SLS results are superior to the OLS single equation estimates and 3SLS estimates of cross-sectional regressions. Second, both trade and FDI serve important channels of shock transmission across borders. Last, specialisation in production or industrial dissimilarity has an indirect impact on the business cycle synchronisation of economies via trade and FDI.

2011 Artis and Okubo To re-examine the correlation between trade and business cycle synchronisation 19 developed countries; real GDP data (1870–2000), trade(1870–2006), CU and other dummies Modified GDP index of Cerqueira and Martins (2009) Just like previous studies, this study also finds a positive effect of trade on business cycle synchronisation even in the current times of globalisation. The impact of economic integration and CUs also comes out to be significantly positive.

2011 Peersman To estimate the effects of aggregate supply, aggregate demand, monetary policy and exchange rate shocks on the USA, the UK and EA Estimations based on post Bretton Woods period, that is, 1974–2008 period. Two-country SVAR model The main findings of the study are (a) symmetric shocks with the EA play a very important role in explaining fluctuations in the UK output, (b) the relative importance of asymmetric supply and demand shocks is around 20% in the long run and, hence, cannot be ignored, (c) exchange rate serves as an independent source of shock, and (d) the UK business cycle seems to be more synchronised with the US business cycle.

2012 Cover and Mallick How rationale Bank of England is to lower its bank rate in response to contractionary shocks and increase its bank rate in response to expansionary shocks UK economy; the data covers the period 1985:Q1 to 2011:Q1 SVAR technique The estimated results imply that two shocks (technology shock and IS shock) are relatively more dominating and explain nearly all the variation in the real GDP (level of output) and the unemployment rate. The study contributes hugely towards making a broader use of SVAR technique.

2013 Kolasa To investigate the differences between economic fluctuations in CEECs and the EA 5 CEECs and the EA; Quarterly data (1995:Q1–2011:Q4) Business cycle accounting framework; VAR system; Dynamic stochastic general equilibrium business cycle model The results (of this study) indicate the business cycles of CEECs and EA countries differ from each other, although substantial convergence has been achieved in the last few years. The study contributes the existing literature by showing a divergence from the mostly used popular methods of filtering techniques and SVAR.

2014 Gächter and Riedl To analyse the impact of ‘euro’ on business cycle synchronisation of EMU countries 27 EMU economies; annual data (1993–2011) System GMM; time-varying correlation index Major finding: EMU membership has not only led to symmetric business cycles across member states but has also decreased the spread in output gaps among them. Thus, these results substantially strengthen the idea that a country is more likely to satisfy the criteria for entry into a CU ex post than ex ante, as proposed by Frankel and Rose (1998).

2014 Magkonis and Tsopanakis To investigate the effects of both aggregate fiscal and financial stress indexes on key macroeconomic variables G7 countries; quarterly data (1980:Q1–2009:Q2) SVAR model The empirical results show that financial and fiscal shocks affect negatively the key macroeconomic variables. Also, there is a week feedback from a fiscal shock to a financial sector and vice-versa.

2015 Enea, Palaăca, and Tigănăs To evaluate business cycle synchronisation within the Group of Seven (G7) countries G7 countries; annual data (1970–2011) Bifactorial analysis; Hodrick-Prescott method; Pearson approach The study finds (a) trade as an important channel of shock transmission, (b) high degree of synchronisation exists during recessions, (c) during economic expansion, there are 2 well-defined macro cycles corresponding to each continent—Europe, North America and the emergent Asian cycle.

2015 Hanus and Vacha To analyse the business cycle synchronisation of the Visegrad Group (V4) with the EU 16 EU countries; monthly data (1990:M1–2014:M12) Wavelet methodology —cohesion measure with time-varying weights High business cycle correlation of the V4 with the EU

2016 Miles and Vijverberg To find out whether the euro has increased or decreased business cycle synchronisation among its member countries Eurozone and non-euro countries; quarterly data (1983:Q1–2009:Q4) Markov switching model Out of 8 Eurozone economies, only one—the Netherlands has strongly synchronised its business cycle since euro introduction

2016 Di Giorgio To find out the business cycle synchronisation of Central and Eastern European Countries (CEECs) with the EA CEECs and EA aggregate; quarterly data (1993:Q1–2014:Q1) Markov regime switching model; synchronisation tests Sample economies show (a) strong business cycle synchronisation during recessionary times, (b) weak cycle correlations economic expansion and high growth times, except Hungary and Poland (c) CEEC cycles are dependent on EA cycles.

Year	Author(s)	Objective of the Study	Sample Data set	Estimation Technique	Major Findings
1989	Blanchard and Quah	Supply disturbances and demand disturbances are used to interpret fluctuations in output and unemployment.	US real GNP growth and unemployment rate data; sample period (1950:Q2–1987:Q4).	VAR system; variance decomposition	Demand shocks create hump-shaped effect on output and unemployment, which disappears approximately after 2–3 years. The supply shocks have an effect on output that cumulates over time and reaches its peak after 5 years. The study drew huge attention and laid the foundation for future research in this area. Assumption: Supply disturbances have permanent effect on output while demand disturbances do not have. Note: GNP data are quarterly. Monthly unemployment data are averaged to provide quarterly observations.
1993	Bayoumi and Eichengreen	To investigate the future path of European Commission (EC) as a monetary union and hence to extract the exact information on underlying aggregate demand and supply shocks	12 EC members and 11 OECD countries (annual sample data from 1960–1988); US regions data (1963–1986)	Vector autoregression (VAR) decomposition technique	The findings suggest that the underlying shocks are significantly more idiosyncratic across EC countries than across US states, thereby making it difficult to operate as a monetary union.
1998	Frankel and Rose	To study an empirical relationship between the extent of trade and the similarity of the shocks and business cycles	Panel of 30 years data for 21 industrialised countries	Cross-country correlations; HP-filtered and differencing detrending techniques	The results—along with the robustness checks—provide a strong positive relationship between the degree of bilateral trade intensity and the cross-country bilateral correlation of business cycle activity. That means, higher trade integration leads to more synchronised business cycles. This endogenous nature of the relationship between the OCA criteria serves the authors as a straightforward application of the Lucas critique. That is, a potential member is more likely to satisfy the OCA criteria for entry into a currency union ex post than ex ante.
2004	Horvath and Rátfai	To analyse the relative incidence of demand and supply shocks in both the participating and the accession EMU countries	8 EMU accession economies and reference group (Germany, France, Italy); sample period (1993:Q1–2000:Q3), except Hungary (1995:Q1–2000:Q3)	Bivariate vector auto regression models	While the results show significant degree of symmetry in business cycle shocks among EMU accession economies, the idiosyncratic shocks between current and would-be member countries dominate. More specifically, the shocks among EMU accession and reference group economies, in particular, Germany, tend to be uncorrelated. These results have an important policy bearing: it could be costly for the participants to proceed with the union.
2004	Imbs	To access empirically the relative impact of trade intensity, financial integration and sectoral specialisation on business cycle correlations of economies	18 developed countries group; quarterly data (1983:Q1–1998:Q3).	Simultaneous equations model; 3SLS; FEM and REM techniques; EC3SLS	The results reveal a strong intra-industry trade link with business cycles correlations, and a weak link between inter-industry trade and cycles correlations. However, the overall effect of trade on business cycle synchronisation is strong and positive. Same is the case of specialisation on business cycles correlation. Also, business cycles in regions at higher levels of financial integration are significantly more synchronised. It should be noted that one of the motivations for such a study was that international correlations of business cycles serves an important parameter to determine the desirability and feasibility of a currency union.
2005	Baxter and Kouparitsas	To find out the determinants of business cycle co-movement between countries	Data set of over 100 countries (both developing and developed); major sample period 1970–1995	‘Extreme bounds analysis’ applied to band-pass filtered data	The study finds that bilateral trade between two countries is highly correlated with the business cycles of economies, while similarity in their industrial structures does not lead to business cycle synchronisation. Same is the case of currency union variable. Among the ‘gravity’ variables, distance is negatively related to business cycle correlation, as one can expect. By taking a large number of variables, the different variants of indexes, and a large number of robust estimations, the study serves benchmark guidance for future research in this area.
2007	Calderón, Chong, and Stein	To see the impact of trade intensity on business cycle synchronisation among developing and industrial countries	Annual data for 147 countries spanning the period 1960–1999	Cross-sectional and panel data techniques imposed on a number of indexes	The study finds that the impact of trade intensity on business cycle synchronisation of developing economies is significantly positive, but smaller than that of industrialised countries. This difference arises because of the differences in their patterns of specialisation and bilateral trade.
2008	Darvas and Szapáry	To study the prevailing degree of synchronisation between the countries from Central and Eastern European Countries (CEECs), and also check the OCA endogeneity of the euro currency union (CU) members	Hodrick-Prescott (HP) and the Band-Pass (BP) filters; dynamic factor models using maximum likelihood (ML) estimation and Kalman-filtering of the state-space representation	8 CEECs, 10 EMU members and various other countries used as control group; quarterly data (1983–2002) grouped in 4 non-overlapping 5-year periods (1983–1987, 1988–1992, 1993–1997, 1998–2002)	The major findings of this study are (a) Hungary, Poland and Slovenia are most synchronised, (b) Czech Republic and Slovakia are least synchronised, (c) the Baltic states (Estonia, Latvia, Lithuania) are not synchronised at all during the entire period of study, and (d) the core EMU countries (Italy, Germany, Belgium, France, Austria, the Netherland) exhibit higher synchronisation in their business cycles, while the periphery countries (Finland, Ireland, Portugal, Spain) show lower co-movements. However, the periphery countries are moving towards the higher synchronisation level of the core EMU countries.
2008	Bussière, Fidrmuc, and Schnatz	To check the integration of Central and South Eastern European countries (CSEECs) with the Euro area (EA)	Annual data set of 61 countries spanning from 1980–2003	Gravity models using panel fixed and random effects technique	The gravity model results suggest: rapid trade integration of the CSEECs and the EA; and a gradual convergence to ‘normal’ trade levels.
2008	Inklaar, Jong-A-Pin, and de Haan	To re-investigate the relationship between intensity of trade and business cycle synchronisation	21 OECD economies; 1970–2003 sample—further divided into 3 sub-samples (1970–1981, 1981–1992, 1992–2003)	Multivariate model; extreme bounds analysis; OLS and instrumental variables techniques	The study confirms a positive impact of trade intensity, specialisation and convergence in fiscal and monetary policies on business cycle synchronisation. The merit (of this study) lies in its incorporation of sample heterogeneity and outliers in the analytical framework.
2009	Baltagi, Demetriades, and Law	To empirically examine the relationship of trade and financial openness with the financial development of economies	Using annual panel data from developing and industrialised countries	Dynamic panel estimation techniques	The estimated results support the hypothesis that trade and financial openness affect positively and significantly the banking sector development. The results also suggest that the marginal effects of trade openness are negatively related to the degree of financial openness and vice-versa. Thus, this research provides a partial support to the Rajan-Zingales hypothesis (this hypothesis states that both the trade and financial openness are necessary for financial development). The study has an important policy implication: relatively closed low-income countries can benefit most from opening up their trade and/or capital accounts.
2009	Gonçalves, Rodrigues, and Soares	To compare post-euro adoption bilateral business cycle correlations of EMU members with the OECD economies	22 OECD and Eurozone countries; major data set (1980–2007)	Differences in-differences technique	The results confirm that bilateral correlations of the business cycle have increased more amongst the EMU members as compared to the OECD economies after the adoption of the euro.
2009	Cerqueira and Martins	To find out the determinants of business cycle synchronisation (BCS) using panel approach	20 OECD countries; Major data set (1970–2002)	GDP cross-correlation index for panel data that captures time-variability	Major findings: (a) bilateral trade and EU membership have a positive effect on BCS, (b) industrial dissimilarity is non-significant, and (c) financial openness has negative effect on BCS.
2010	Hsu	To answer the question: is it feasible for East Asia to form a common currency area? Or Does a common symmetric shock dominate East Asian economies?	A post-Asian crisis data set (1997–2007); one group of economies belong to East Asia while the second group belongs to the EA	The study uses a 3-variable SVAR model to identify three kinds of shocks: global, regional and domestic	The results suggest (a) a sharp decline of asymmetric domestic shocks in the post-crisis period, (b) most East Asian countries respond to regional and global shocks in a symmetric way, and (c) feasibility of common CU in East Asia in the near future. However, the study provides a limited picture of the influence of global shocks on the East Asia. It would have been of much more interest had the study considered the impact of shocks on East Asia originating from major economies of the world, such as the US and the UK.
2010	Savva, Neanidis, and Osborn	To examine business cycle correlations between the EA and the currently negotiating and recently added EU states	9 enlargement countries, 5 core and 4 periphery EA countries, Turkey, Croatia; Monthly data (1980:M1–2006:M6)	Bivariate VAR-GARCH	The findings reveal that all the negotiating and newly added EU countries have increased their business cycle synchronisation with the EA more than two times, or at least have changed from negative to positive correlations, since the early 1990s. The study adds to the existing literature in two broad ways as follows. First, while the existing literature only considers either constant correlation coefficients or sub-samples divided using assumed exogenous break dates; this study determines the timing and the length of structural shifts endogenously. Second, this study considers the possibility of multiple transition paths, thus allowing for more complex correlation regimes.
2011	Hsu, Wu, and Yau	Add-on the simultaneous equation model of Imbs (2004) by including FDI as one more possible channels of shock transmission	Cross-sectional as well as panel-data from 1998 to 2002 on G7 countries and 8 major OECD countries	The study estimates single equation as well as error-components three stage least squares (EC3SLS) simultaneous equation model	The main findings of the study are as follows. First, EC3SLS results are superior to the OLS single equation estimates and 3SLS estimates of cross-sectional regressions. Second, both trade and FDI serve important channels of shock transmission across borders. Last, specialisation in production or industrial dissimilarity has an indirect impact on the business cycle synchronisation of economies via trade and FDI.
2011	Artis and Okubo	To re-examine the correlation between trade and business cycle synchronisation	19 developed countries; real GDP data (1870–2000), trade(1870–2006), CU and other dummies	Modified GDP index of Cerqueira and Martins (2009)	Just like previous studies, this study also finds a positive effect of trade on business cycle synchronisation even in the current times of globalisation. The impact of economic integration and CUs also comes out to be significantly positive.
2011	Peersman	To estimate the effects of aggregate supply, aggregate demand, monetary policy and exchange rate shocks on the USA, the UK and EA	Estimations based on post Bretton Woods period, that is, 1974–2008 period.	Two-country SVAR model	The main findings of the study are (a) symmetric shocks with the EA play a very important role in explaining fluctuations in the UK output, (b) the relative importance of asymmetric supply and demand shocks is around 20% in the long run and, hence, cannot be ignored, (c) exchange rate serves as an independent source of shock, and (d) the UK business cycle seems to be more synchronised with the US business cycle.
2012	Cover and Mallick	How rationale Bank of England is to lower its bank rate in response to contractionary shocks and increase its bank rate in response to expansionary shocks	UK economy; the data covers the period 1985:Q1 to 2011:Q1	SVAR technique	The estimated results imply that two shocks (technology shock and IS shock) are relatively more dominating and explain nearly all the variation in the real GDP (level of output) and the unemployment rate. The study contributes hugely towards making a broader use of SVAR technique.
2013	Kolasa	To investigate the differences between economic fluctuations in CEECs and the EA	5 CEECs and the EA; Quarterly data (1995:Q1–2011:Q4)	Business cycle accounting framework; VAR system; Dynamic stochastic general equilibrium business cycle model	The results (of this study) indicate the business cycles of CEECs and EA countries differ from each other, although substantial convergence has been achieved in the last few years. The study contributes the existing literature by showing a divergence from the mostly used popular methods of filtering techniques and SVAR.
2014	Gächter and Riedl	To analyse the impact of ‘euro’ on business cycle synchronisation of EMU countries	27 EMU economies; annual data (1993–2011)	System GMM; time-varying correlation index	Major finding: EMU membership has not only led to symmetric business cycles across member states but has also decreased the spread in output gaps among them. Thus, these results substantially strengthen the idea that a country is more likely to satisfy the criteria for entry into a CU ex post than ex ante, as proposed by Frankel and Rose (1998).
2014	Magkonis and Tsopanakis	To investigate the effects of both aggregate fiscal and financial stress indexes on key macroeconomic variables	G7 countries; quarterly data (1980:Q1–2009:Q2)	SVAR model	The empirical results show that financial and fiscal shocks affect negatively the key macroeconomic variables. Also, there is a week feedback from a fiscal shock to a financial sector and vice-versa.
2015	Enea, Palaăca, and Tigănăs	To evaluate business cycle synchronisation within the Group of Seven (G7) countries	G7 countries; annual data (1970–2011)	Bifactorial analysis; Hodrick-Prescott method; Pearson approach	The study finds (a) trade as an important channel of shock transmission, (b) high degree of synchronisation exists during recessions, (c) during economic expansion, there are 2 well-defined macro cycles corresponding to each continent—Europe, North America and the emergent Asian cycle.
2015	Hanus and Vacha	To analyse the business cycle synchronisation of the Visegrad Group (V4) with the EU	16 EU countries; monthly data (1990:M1–2014:M12)	Wavelet methodology —cohesion measure with time-varying weights	High business cycle correlation of the V4 with the EU
2016	Miles and Vijverberg	To find out whether the euro has increased or decreased business cycle synchronisation among its member countries	Eurozone and non-euro countries; quarterly data (1983:Q1–2009:Q4)	Markov switching model	Out of 8 Eurozone economies, only one—the Netherlands has strongly synchronised its business cycle since euro introduction
2016	Di Giorgio	To find out the business cycle synchronisation of Central and Eastern European Countries (CEECs) with the EA	CEECs and EA aggregate; quarterly data (1993:Q1–2014:Q1)	Markov regime switching model; synchronisation tests	Sample economies show (a) strong business cycle synchronisation during recessionary times, (b) weak cycle correlations economic expansion and high growth times, except Hungary and Poland (c) CEEC cycles are dependent on EA cycles.

Source: The author.

This study makes an attempt to fill these gaps by investigating both the direct as well as indirect effects of trade, industrial dissimilarity and FDI on the business cycle synchronisation of Eurozone economies. In other words, the study explores the interactions between trade, specialisation patterns, and FDI and their linkages with cyclical co-movements of Eurozone countries. In addition to these variables, the use of other variables (per capita GDP difference variable, similar monetary policy and legal origin variables, and gravity variables in general) help trace the impact of those variables on the endogenous variables (trade, industrial dissimilarity and FDI, in our case) and through endogenous variables on the business cycle synchronisation of the Eurozone economies. While these other variables are not of direct interest, one needs to model their effects so as to be able to discern the total impact (both direct and indirect) of major macro-variables on the business cycle synchronisation of economies. And this becomes one of the reasons that this study adopts the error component three-stage least squares (EC3SLS) technique proposed by Baltagi (1981), which has the characteristics of simultaneous equations. Simultaneity, implicit in most theories, is also revealing empirically. Our main variable of interest is output correlation. Even though this variable is endogenous in this analysis, the EC3SLS provides consistent estimates. Besides, this technique adopts the panel data approach, which means it captures both cross-sectional and time-series information. Another advantage of the EC3SLS approach is that it can tackle the problem of endogeneity and the indirect effects of each variable are also measured. Thus, the entire strategy of this study is to address two major issues raised in the debate about output synchronisation in the euro area:

Have business cycles of Eurozone economies become more similar?

What are the major factors of business cycle synchronisation among these economies?

3. The Econometric Model

3.1. Model Specification and Data

To find out the direct and indirect effects of trade intensity (T) and industrial dissimilarity (ID) on business cycle correlation (ρ), the study uses the following three-equation model:

\begin{array}{l} [\begin{matrix} ρ \\ T \\ I D \end{matrix}] = [\begin{matrix} α_{0} & α_{1} & α_{2} & α_{3} & 0 & 0 & 0 & 0 & 0 & 0 \\ β_{0} & 0 & β_{1} & 0 & β_{2} & β_{3} & β_{4} & 0 & β_{5} & β_{6} \end{matrix}] \\ [\begin{matrix} γ_{0} & γ_{1} & 0 & 0 & 0 & 0 & 0 & γ_{2} & γ_{3} & 0 \end{matrix}] \\ [\begin{matrix} 1 \\ T \\ I D \\ M P \\ L a n g u a g e \\ D i s \tan c e \\ A d j a c e n c y \\ G D P_{G} \\ \begin{matrix} G D P_{P} \\ P C G D P_{D} \end{matrix} \end{matrix}] + [\begin{matrix} ϕ_{1, i, j} \\ ϕ_{2, i, j} \\ ϕ_{3, i, j} \end{matrix}] + [\begin{matrix} ε_{1, i, j, t} \\ ε_{2, i, j, t} \\ ε_{3, i, j, t} \end{matrix}] ​ \end{array}

(1)

Where,

i, j, t are the values of index country pairs (i, j) in year t; and ϕ is the time-invariant country-pair-specific term used to control for individual heterogeneity. This heterogeneity can arise because of any reason—be it political, ethical, cultural, geographic, regional, religious, administrative, historical or any other factor. The incorporation of these country-pair effects controls the probable presence of heterogeneity bias in our estimates (Cheng & Wall, 2004). Thus, these country-pair effects are specific to the country pairs but common to all time periods—say, for example, for Germany and Italy country-pair from the year 1990–2009. The only restriction is that they are symmetric in nature, that is, ϕ_ij = ϕ_ji. In addition, time effects, which are common to the country-pairs but specific to the time period t, are also used in some specifications of the model. These time-specific effects control the probable presence of heterogeneity across time in our estimates. ϵ is an idiosyncratic random error.

For some sample countries, FDI data is not available (see Note 9), therefore, Model (1) is modified to incorporate the FDI variable as an additional endogenous variable. Accordingly, the resultant model consists of four equations as given below:

\begin{array}{l} [\begin{matrix} ρ \\ T \\ ID \\ FDI \end{matrix}] = [\begin{matrix} α_{0} & α_{1} & α_{2} & α_{3} & α_{4} & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ β_{0} & 0 & β_{1} & β_{2} & 0 & β_{3} & β_{4} & β_{5} & 0 & β_{6} & β_{7} & 0 \\ γ_{0} & γ_{1} & 0 & γ_{2} & 0 & 0 & 0 & 0 & γ_{3} & γ_{4} & 0 & 0 \\ λ_{0} & λ_{1} & λ_{2} & 0 & λ_{3} & 0 & 0 & 0 & 0 & 0 & 0 & λ_{4} \end{matrix}] [\begin{matrix} 1 \\ T \\ ID \\ FDI \\ MP \end{matrix}] \\ [\begin{matrix} Language \\ Distance \\ Adjacency \\ {GDP}_{G} \\ \begin{matrix} {GDP}_{P} \\ {PCGDP}_{D} \end{matrix} \\ LO \end{matrix}] + [\begin{matrix} f_{1,i,j} \\ f_{2,i,j} \\ f_{3,i,j} \\ f_{4,i,j} \end{matrix}] + [\begin{matrix} ε_{1,i,j,t} \\ ε_{2,i,j,t} \\ ε_{3,i,j,t} \\ ε_{4,i,j,t} \end{matrix}] ​ \end{array}

(2)

Where,

ρ represents business cycle correlation variable. In this analysis, output correlation variable is constructed on the basis of two different measures of output. These are: (i) HP-filtered output (ρHP); and (ii) first-differenced output (ρFD). The former is obtained as a natural logarithm of real GDP, detrended with a Hodrick-Prescott (Hodrick & Prescott, 1997) filter. The latter is obtained as an annual growth rate of real GDP. Since the data used is of yearly frequency, the smoothing parameter used in this case is 6.25 (Ravn & Uhlig, 2002). The two measures of output serve to check whether the results are sensitive to a particular type of output measure or not.³,⁴ The annual data for real GDP (1990–2009) is taken from the World Bank’s World Development Indicators and is measured in 2005 constant US dollars.

Bilateral trade intensity index (T_i,j,t) used in this analysis is defined as:

T_{i, j, t} = \frac{x_{i, j, t} + m_{i, j, t} + x_{j, i, t} + m_{j, i, t}}{X_{i, t} + M_{i, t} + X_{j, t} + M_{j, t}} ​

(3)

Where, x_i,j,t (x_j,i,t) is the value of exports from country i (country j) to country j (country i) in year t; m_i,j,t (m_j,i,t) is the value of the imports of country i (country j) from country j (country i) in year t; X_i,t and X_j,t (M_i,t and M_j,t) are the values of country i’s exports (imports) and country j’s exports (imports) to all countries in year t, respectively. This index has become the standard trade intensity index in recent years because it saves us from understating actual trade flows, a concern acknowledged by Frankel and Rose (1998). A higher value of T_i,j,t indicates greater trade intensity between countries i and j. The annual export and import data is extracted from Direction of Trade Statistics (DOTS) CD_ROM of International Monetary Fund (IMF) and is measured in current US dollars.

Krugman (1991) suggested a measure of industrial dissimilarity. Subsequently, the measure has been used by many economists in their studies (e.g., Hsu et al., 2011). We also use the same index as follows:

I D_{i, j, t} = \sum_{k}^{K} | S_{k, i, t} - S_{k, j, t} | ​

(4)

Where, S_k,i,t is the GDP share of industry k in country i in year t; and S_k,j,t—the GDP share of industry k in country j in year t. A larger value of ID_i,j,t indicates a greater degree of industry dissimilarity in industrial structure. This indicator uses manufacturing sector value-added shares relative to total economy taken from the OECD (2012) STAN Indicators Rev. 4 Database.

There is no standard measure of bilateral FDI intensity; so this index is measured in the same way as the index of bilateral trade intensity:

F D I_{i, j, t} = \frac{I n w a r d f d i_{i, j, t} + O u t w a r d f d i_{i, j, t} + I n w a r d f d i_{j, i, t} + O u t w a r d f d i_{j, i, t}}{I n w a r d F D I_{i, t} + O u t w a r d F D I_{i, t} + I n w a r d F D I_{j, t} + O u t w a r d F D I_{j, t}} ​

(5)

Generally, FDI represents capital flows across national borders. This study considers bilateral FDI flows as a proxy for financial integration. The data on bilateral inward and outward FDI is collected from the OECD (2014) International Direct Investment Statistics.

Our sample data set consists of Eurozone countries. Many steps were taken to integrate these Eurozone economies in the past; therefore, we expect a positive relationship between trade intensity and business cycle co-movements, that is, α₁ > 0, which is confirmed in most of the estimated results. As far as the sign of α₂ is concerned, we expect a negative relationship between industry dissimilarity and output correlation (α₂ < 0) if business cycles are driven by industry-specific shocks. In other words, countries with greater similarities in their industrial structure tend to move together. Regarding individual bilateral trade and industrial dissimilarity equations, classical Ricardian theory predicts a positive linkage between trade and specialisation, which means that industrial dissimilarities generate more trade (Balassa, 1986; Dornbusch et al., 1977). This implies that β₁ and/or γ₁ need to be greater than zero. Also, FDI can transmit shocks across economies through various channels such as technology diffusion, or its connection with international financial markets through the process of reallocation of capital, which implies that FDI can serve as important carrier of disturbances across national boundaries in the current times of economic globalisation. Hence, it is a case of α₃ > 0. Also, β₂ > 0 because we believe that common characteristics specific to the country pair are likely to govern bilateral trade and FDI flows. Such a relationship is supported by the EC3SLS estimates. The sign of λ₂ is governed by the same principle as that of γ₁. In general, the sign of λ₂ is ambiguous. The absence of any rigorous theory related to the relationship between FDI and specialisation deepens the ambiguity.

A set of exogenous variables is used to achieve the identification of the model (Equations (1) and (2)).⁵ As far as these variables are concerned, we expect α₃ > 0 in Equation (1) or α₄ > 0 in Equation 2, because Eurozone economies in our sample data set are broadly similar in economic structure, use the common euro currency and face similar monetary policy behaviour. Monetary policy closeness between a country-pair is measured as the correlation of short-term interest rates between the two countries.⁶ Also, two countries which are highly integrated through FDI and which share similar monetary policy objectives might transmit the idiosyncratic shock occurring in one country to the other country’s real activities via the FDI channel. This implies that λ₅ > 0. In addition, gravity variables are included—as they play an important role in explaining bilateral trade flows. These gravity variables include dummies for common official language and adjacency, geographic distance, and log of the product of two countries GDP (GDP_P). Except for GDP_P, the data on these gravity variables is collected from the database of the Centre d’Etudes Prospectives et d’Informations Internationales (CEPII). The huge literature available on the gravity model of international trade governs the signs of these variables; we expect β₃ > 0, β₄ < 0, and β₅ > 0. Besides, La Porta, Lopez, & Shleifer (2008) emphasise the important role that the common legal origin of two countries can play in transmitting financial shocks across national borders. We have λ₄ > 0.

Also, following Imbs (2004), two exogenous determinants of specialisation—the logs of the ratios of two countries GDP (GDP_G) and the logs of the products of two countries GDP (GDP_P)—are used in Models (1) and (2). These two variables are expected to affect the patterns of specialisation. In the economics paradigm, it is believed that economies initially diversify by spreading their economic activities more equally across sectors, but begin specialising once they reach higher levels of economic development. Since our entire sample data comprises Eurozone countries, we expect a positive relationship between specialisation and these two variables. In other words, pairs of countries at different stages of development, as measured by the gap between their GDPs, tend to display different economic structures. To sum it up, it can be said that a negative relationship should exist between industry dissimilarity and GDP_G and/or GDP_P in our case; a positive relationship will mean that these sample economies have continuously diversified their production processes at higher and higher stages of development. This relationship can be elaborated further by looking at the sign and significance of the PCGDP_D variable (the absolute difference of per capita GDP between a pair of countries) because it helps us to identify the type of trade that takes place among these euro club countries. A positive sign for this variable suggests that trade in these countries is based on comparative advantage, which lends support to the inter-industry trade type; while a negative sign shows that these economies feature intra-industry trade patterns (Kunroo, Sofi, & Azad, 2016; Rault, Sova, & Sova, 2009). This, in turn, invokes the classical Ricardian theory—countries with different factor endowments and thus, in comparative advantages, would exchange more goods of an inter-industry nature.⁷ More elaborately, if labour is the only factor of production, then comparative advantage will be calculated on a per worker basis. However, when both the capital and labour are considered to be the two factors of production, then inter-industry trade between two countries will be such that the country with more abundance in capital will produce and export capital-intensive goods, while a country that is abundant in labour will produce and export labour-intensive goods. On the contrary, countries indulging in intra-industry trade will export and import those kinds of goods that are produced along the same lines of production. Such a type of trade has risen significantly since the 1980s in most of the OECD and other highly open economies (Blaster, 2015). For instance, China imports capital-cum-technology-intensive computer components and uses its abundant cheap labour to assemble these components in to a final product, that is, finished computer. This finished computer is then exported back to the USA and other developed European countries with a small marginal value addition, in terms of components assembling costs and small tariff costs. Hence, intra-industry trade is created. Also, Google and other software application developers for mobile and other computing devices are highly concentrated in the USA. Today, hardware and other components are mostly made outside of the USA, in China, Japan and Korea, for example. Hence, the USA and the rest of the world do trade in the same industry (i.e., again intra-industry trade). The data related to these variables (GDP_G, GDP_P, and PCGDP_D) is collected from World Development Indicators, World Bank.

3.2. Econometric Methodology and Descriptive Statistics

The focus of this study is the estimation of simultaneous equation models represented by Equations (1) and (2), using the error component EC3SLS estimation technique proposed by Baltagi (1981). The use of this procedure allows efficient estimation in models with panel data.⁸ Moreover, the simultaneous equation EC3SLS econometric technique adopts the panel data approach, which incorporates both cross-sectional and time-series information. In fact, the adoption of the simultaneous equation EC3SLS estimation method is even supported by the unconditional correlations of all the exogenous and endogenous variables used in the study. These correlations are reported in Table 2. From the table, we find that both measures of cycle synchronisation, that is, ρHP and ρFD, are positively and significantly (at the 5% level) correlated with trade and monetary policy, and negatively correlated with industry dissimilarity. However, FDI does not show any significant correlation. Also, the correlation of cycle synchronisation with trade is greater than that with FDI and/or industry dissimilarity, suggesting that the shock transmission mechanism via trade across these Eurozone economies is more influential than that from the channels of FDI and/or production specialisation over the sample period 1990–2009. Regarding the other equations (trade, industry dissimilarity or specialisation in production, and FDI), FDI shows a significantly positive correlation with both trade and industry dissimilarity, while industry dissimilarity shows a significantly negative (though small) correlation with trade. The gravity variables show significant association with trade; industry dissimilarity shows a significantly positive correlation with both GDP_G and GDP_P; FDI does not show any correlation with legal origin; PCGDP_D shows a significantly negative correlation with trade; and monetary policy shows a negative but significant correlation with FDI. Generally, the same results are confirmed by the EC3SLS estimates as reported in the next section. The pairwise sample correlations clearly highlight the need for using a simultaneous equation model panel data technique that takes care of both direct and indirect effects, which is what has been done in this study.

Table 2
Pairwise Sample Correlations among the Variables

ρHP ρFD T Industry dissimilarity FDI MP LO ADJ LANG DIST GDP_G GDP_P PCGDP_D

ρHP 1.000 – – – – – – – – – – – –

ρFD 0.760* 1.000 – – – – – – – – – – –

T 0.418* 0.359* 1.000 – – – – – – – – – –

Industry dissimilarity –0.233* –0.165* –0.084* 1.000 – – – – – – – – –

FDI 0.082 –0.075 0.351* 0.153* 1.000 – – – – – – – –

MP 0.253* 0.241* 0.262* 0.108* –0.185* 1.000 – – – – – – –

LO 0.067* 0.250* 0.271* –0.236* 0.124 0.256* 1.000 – – – – – –

ADJ 0.360* 0.287* 0.584* –0.081* 0.252* 0.200* 0.201* 1.000 – – – – –

LANG 0.293* 0.285* 0.182* –0.019 0.054 0.179* 0.183* 0.533* 1.000 – – – –

DIST –0.486* –0.264* –0.371* 0.021 –0.302* –0.209* –0.112* –0.525* –0.394* 1.000 – – –

GDP_G –0.075* –0.193* –0.049 0.100* –0.016 0.244* 0.034 0.082* 0.081* –0.242* 1.000 – –

GDP_P 0.108* 0.121* 0.250* 0.293* 0.281* 0.348* 0.152* 0.103* 0.101* –0.032 0.023 1.000 –

PCGDP_D –0.323* –0.324* –0.383* 0.024 –0.113 –0.263* –0.184* –0.211* –0.162* 0.058* 0.202* –0.191* 1.000

	ρHP	ρFD	T	Industry dissimilarity	FDI	MP	LO	ADJ	LANG	DIST	GDP_G	GDP_P	PCGDP_D
ρHP	1.000	–	–	–	–	–	–	–	–	–	–	–	–
ρFD	0.760*	1.000	–	–	–	–	–	–	–	–	–	–	–
T	0.418*	0.359*	1.000	–	–	–	–	–	–	–	–	–	–
Industry dissimilarity	–0.233*	–0.165*	–0.084*	1.000	–	–	–	–	–	–	–	–	–
FDI	0.082	–0.075	0.351*	0.153*	1.000	–	–	–	–	–	–	–	–
MP	0.253*	0.241*	0.262*	0.108*	–0.185*	1.000	–	–	–	–	–	–	–
LO	0.067*	0.250*	0.271*	–0.236*	0.124	0.256*	1.000	–	–	–	–	–	–
ADJ	0.360*	0.287*	0.584*	–0.081*	0.252*	0.200*	0.201*	1.000	–	–	–	–	–
LANG	0.293*	0.285*	0.182*	–0.019	0.054	0.179*	0.183*	0.533*	1.000	–	–	–	–
DIST	–0.486*	–0.264*	–0.371*	0.021	–0.302*	–0.209*	–0.112*	–0.525*	–0.394*	1.000	–	–	–
GDP_G	–0.075*	–0.193*	–0.049	0.100*	–0.016	0.244*	0.034	0.082*	0.081*	–0.242*	1.000	–	–
GDP_P	0.108*	0.121*	0.250*	0.293*	0.281*	0.348*	0.152*	0.103*	0.101*	–0.032	0.023	1.000	–
PCGDP_D	–0.323*	–0.324*	–0.383*	0.024	–0.113	–0.263*	–0.184*	–0.211*	–0.162*	0.058*	0.202*	–0.191*	1.000

Source: The author.

Notes: * shows significance at 0.05; ρHP = Hodrick-Prescott filtered output measured from the natural logarithm of real GDP; ρFD = first-differenced output measured as the annual growth rate of real GDP; T = bilateral trade; industry dissimilarity = industrial dissimilarity; FDI = bilateral foreign direct investment; MP = common monetary policy; LO = legal origin; ADJ = adjacency; LANG = common language; DIST = bilateral distance; GDP_G = GDP gap; GDP_P = GDP product; PCGDP_D = per capita GDP difference.

For a start, the bilateral correlations of the business cycles are computed on the basis of the cyclical component of annual real GDP data isolated using the Hodrick-Prescott (Hodrick & Prescott, 1997) filter. Another method used is the first-differenced output measure. The purpose is to estimate the business cycle component of the real GDP macro variable. The descriptive statistics for all the panel variables used in this analysis are given in Table 3. The total number of panels for the data set is 105, but this reduces to 21 for the smaller FDI data set corresponding to Model (2). The first column gives the overall mean value of the variable in the entire panel. The standard deviation reports three measures: (a) the overall standard deviation (around the grand mean $\overset{─}{x} = \frac{1}{N T \sum_{i} \sum_{t} x_{i t}} $ ) of a variable in the entire panel is $\sqrt{\frac{1}{N T - 1} \sum_{i} \sum_{t} {(x_{i t} - \overset{─}{x})}^{2}}$ , (b) between standard deviation of the variable across individuals, that is, $\sqrt{\frac{1}{N - 1} \sum_{i} {(\overset{─}{x_{i}} - \overset{─}{x})}^{2}}$ , and (c) within standard deviation of a variable over time for each individual (around individual mean ${\overset{─}{x}}_{i} = \frac{1}{T \sum_{t} x_{i t}} $ ) is $\sqrt{\frac{1}{N T - 1} \sum_{i} \sum_{t} {(x_{i t} - \overset{─}{x_{i}} + \overset{─}{x})}^{2}} $ , where $(x_{i t} - \overset{─}{x_{i}} + \overset{─}{x}) $ is centred on $\overset{─}{x}$ . Time-invariant regressors such as language, adjacency and legal origin have 0 within the variation/standard deviation. The Min and Max columns represent the minimums and maximums of say, x_it for overall, ${\bar{x}}_{i}$ for between, and $(x_{i t} - {\bar{x}}_{i} + \bar{x}) $ for within. The last column gives the number of observations and time period of every variable.

Table 3

Panel Summary Statistics

Variable	Mean	Standard Deviation	Min	Max	Observations
Industry dissimilarity	–	–	1	105	N = 2100
Year	–	–	1990	2009	N = 2100
HP-filtered Correlation
Overall	0.692	0.166	0.255	0.935	N = 2100
Between	–	0.167	0.255	0.935	n = 105
Within	–	1.95E–16	0.692	0.692	T = 20
First-differenced Correlation
Overall	0.674	0.151	0.231	0.911	N = 2100
Between	–	0.151	0.231	0.911	n = 105
Within	–	1.80E–16	0.674	0.674	T = 20
Trade
Overall	0.018	0.03	0	0.146	N = 2100
Between	–	0.03	0.001	0.131	n = 105
Within	–	0.011	–0.059	0.074	T = 20
Industrial Dissimilarity
Overall	5.31	4.349	0	23.481	N = 2100
Between	–	3.516	0.467	16.691	n = 105
Within	–	2.581	–5.537	14.317	T = 20
Foreign Direct Investment
Overall	0.041	0.066	–0.481	0.672	N = 420
Between	–	0.032	–0.001	0.102	n = 21
Within	–	0.058	–0.467	0.61	T = 20
Monetary Policy
Overall	0.799	0.251	0.148	0.997	N = 2100
Between	–	0.251	0.148	0.997	n = 105
Within	–	1.85E–16	0.799	0.799	T = 20
GDP Product
Overall	48.518	14.039	0	57.231	N = 2100
Between	–	6.679	35.504	56.924	n = 105
Within	–	1.22E + 01	0.534	62.101	T = 20
GDP Gap
Overall	0.161	2.176	–5.864	4.767	N = 2100
Between	–	2.051	–4.141	4.522	n = 105
Within	–	0.753	–3.401	4.302	T = 20
Per Capita GDP Difference
Overall	9.155	1.172	2.268	11.455	N = 2100
Between	–	1.021	6.716	10.874	n = 105
Within	–	0.583	4.453	11.004	T = 20
Distance
Overall	1394.06	711.62	149.926	3383.273	N = 2100
Between	–	714.861	149.926	3383.273	n = 105
Within	–	1.681	1319.052	1398.008	T = 20
Language
Overall	0.086	0.281	0	1	N = 2100
Between	–	0.281	0	1	n = 105
Within	–	0	0.086	0.086	T = 20
Adjacency
Overall	0.152	0.361	0	1	N = 2100
Between		0.361	0	1	n = 105
Within	–	0	0.152	0.152	T = 20
Legal Origin
Overall	0.286	0.453	0	1	N = 14616
Between	–	0.453	0	1	n = 812
Within	–	0	0.286	0.286	T = 18

Source: The author.

4. Empirical Results

Out of 15 sample countries, bilateral FDI data is available only for seven Eurozone economies. Therefore, we examine the results under a three-equation model (Equation (1) EC3SLS estimates reported in Table 4) and a four-equation model (Equation (2) EC3SLS estimates reported in Table 5).⁹ Due to data availability, we collect annual observations spanning the period 1990–2009.¹⁰,¹¹ The results (Table 4) show that both bilateral trade and a common monetary policy have a positive and significant relationship with output synchronisation, while industry dissimilarity has a significantly negative relationship with business cycle synchronisation. This provides evidence that high bilateral trade intensity leads to a synchronised business cycle, industrial structures strongly influence business cycles between pairs of countries, and that having the same monetary policy helps to explain business cycle correlation. On the other hand, the relatively small coefficient of industry dissimilarity in the trade equation weakly challenges the well-perceived idea that industrial differences between two countries generate more trade between them (Balassa, 1986). The negative and significant coefficient of the PCGDP_D variable in the trade equation (Panel B) is also in line with our expectation. This suggests that trade exists along similar lines of production, as one would expect in the case of these highly integrated Eurozone economies. In terms of the implication for trade theory, the result is in line with the arguments for intra-industry trade, that is, these sample Eurozone economies trade in goods that are similar in quality and technology. Such a conclusion is even solidified by the significantly negative coefficient of trade in the industry dissimilarity equation ( Table 4, Panel C). This result contradicts the classical Ricardian theory that there is a positive linkage between trade and specialisation. The emphasis is on the negative but significant coefficient of trade in Panel C (industry dissimilarity equation) of Table 4. Such findings are consistently further supported by the positive and significant coefficients of GDP_G and GDP_P—these variables determine the development stages of industrial specialisation—in Panel C, thereby suggesting that these Eurozone economies have diversified even at higher levels of income. The empirical performance of the so-named gravity variables (which account for trade flows) is in line with the gravity literature. Clearly, these variables show a high predictive power on trade flows. More elaborately, the coefficients of GDP product, distance and adjacency in Table 4 suggest that economic size increases trade flows between two countries, bilateral distance between two countries reduces the flow of trade between them, while two countries that share borders with each other will trade more. However, a common language exhibits a significantly negative impact on bilateral trade flows.

Table 4
Simultaneous Equations EC3SLS Panel Eurozone Data Estimates of Model (1)

Output measure HP-filtered
First-differenced

Time effects No Yes No Yes

Panel A: Correlation

Trade 2.622 (0.116)* 2.677 (0.115)* 1.906 (0.111)* 1.933 (0.109)*

Industrial dissimilarity –0.013 (0.001)* –0.012 (0.001)* –0.009 (0.001)* –0.008 (0.001)*

Monetary policy 0.115 (0.013)* 0.112 (0.013)* 0.107 (0.012)* 0.105 (0.012)*

Panel B: Trade

Industrial dissimilarity –0.002 (0.001)* –0.002 (0.000)* –0.002 (0.001)* –0.002 (0.000)*

PCGDP difference –0.006 (0.001)* –0.007 (0.001)* –0.006 (0.001)* –0.007 (0.001)*

GDP product 0.001 (0.001)* 0.001 (0.000)* 0.001 (0.001)* 0.001 (0.000)*

Language –0.021 (0.002)* –0.021 (0.002)* –0.021 (0.002)* –0.022 (0.002)*

Distance –8.53e–06 (7.57e–07)* –8.58e–06 (7.63e–07)* –6.61e–06 (7.66e–07)* –6.65e–06 (7.72e–07)*

Adjacency 0.039 (0.002)* 0.040 (0.002)* 0.041 (0.002)* 0.042 (0.002)*

Panel C: Industrial Dissimilarity

Trade –48.859 (3.211)* –42.184 (3.214)* –49.083 (3.212)* –42.321 (3.215)*

GDP gap 0.194 (0.041)* 0.194 (0.040)* 0.203 (0.041)* 0.203 (0.040)*

GDP product 0.115 (0.007)* 0.112 (0.007)* 0.115 (0.007)* 0.112 (0.007)*

Output measure	HP-filtered	First-differenced
Panel A: Correlation
Trade	2.622 (0.116)***	2.677 (0.115)***	1.906 (0.111)***	1.933 (0.109)***
Industrial dissimilarity	–0.013 (0.001)***	–0.012 (0.001)***	–0.009 (0.001)***	–0.008 (0.001)***
Monetary policy	0.115 (0.013)***	0.112 (0.013)***	0.107 (0.012)***	0.105 (0.012)***
Panel B: Trade
Industrial dissimilarity	–0.002 (0.001)***	–0.002 (0.000)***	–0.002 (0.001)***	–0.002 (0.000)***
PCGDP difference	–0.006 (0.001)***	–0.007 (0.001)***	–0.006 (0.001)***	–0.007 (0.001)***
GDP product	0.001 (0.001)***	0.001 (0.000)***	0.001 (0.001)***	0.001 (0.000)***
Language	–0.021 (0.002)***	–0.021 (0.002)***	–0.021 (0.002)***	–0.022 (0.002)***
Distance	–8.53e–06 (7.57e–07)***	–8.58e–06 (7.63e–07)***	–6.61e–06 (7.66e–07)***	–6.65e–06 (7.72e–07)***
Adjacency	0.039 (0.002)***	0.040 (0.002)***	0.041 (0.002)***	0.042 (0.002)***
Panel C: Industrial Dissimilarity
Trade	–48.859 (3.211)***	–42.184 (3.214)***	–49.083 (3.212)***	–42.321 (3.215)***
GDP gap	0.194 (0.041)***	0.194 (0.040)***	0.203 (0.041)***	0.203 (0.040)***
GDP product	0.115 (0.007)***	0.112 (0.007)***	0.115 (0.007)***	0.112 (0.007)***

Source: The author.

Notes: Standard errors are in parenthesis. Constant estimates are not reported. ***, ** and * significant at the 1%, 5% and 10% levels, respectively.

Table 5

Simultaneous Equations EC3SLS panel Eurozone Data Estimates of Model (2)

Output measure	HP-filtered		First-differenced
Time effects	No	Yes	No	Yes
Panel A: Correlation
Trade	1.691 (0.140)***	1.531 (0.133)***	1.698 (0.138)***	1.438 (0.131)***
Industrial dissimilarity	0.002 (0.002)	0.003 (0.002)	–0.004 (0.002)*	–0.002 (0.002)
FDI	–1.111 (0.156)***	–0.927 (0.141)***	–1.511 (0.154)***	–1.207 (0.139)***
Monetary policy	0.233 (0.127)*	0.233 (0.126)*	0.774 (0.124)***	0.789 (0.121)***
Panel B: Trade
Industrial dissimilarity	–0.001 (0.001)*	–0.002 (0.001)***	–0.001 (0.001)	–0.002 (0.001)***
FDI	0.494 (0.037)***	0.421 (0.034)***	0.483 (0.037)***	0.414 (0.033)***
PCGDP difference	–0.003 (0.001)***	–0.003 (0.001)***	–0.003 (0.001)***	–0.003 (0.001)***
GDP product	0.011 (0.001)***	0.012 (0.001)***	0.010 (0.001)***	0.012 (0.001)**
Language	0.002 (0.004)	0.002 (0.004)	0.005 (0.004)	0.006 (0.004)
Distance	–9.05e–07 (2.98e–06)	–2.06e–06 (2.94e–06)	–3.32e–06 (2.87e–06)	–4.17e–06 (2.84e–06)
Adjacency	0.024 (0.002)***	0.027 (0.002)***	0.024 (0.002)***	0.028 (0.002)***
Panel C: Industrial Dissimilarity
Trade	–5.242 (4.876)	–7.672 (4.684)	–4.491 (4.873)	–7.320 (4.683)
FDI	10.074 (4.807)**	16.285 (4.266)***	10.064 (4.807)**	16.374 (4.266)***
GDP gap	0.091 (0.076)	0.103 (0.076)	0.131 (0.076)*	0.131 (0.075)*
GDP product	0.826 (0.130)***	0.742 (0.129)***	0.805 (0.130)***	0.732 (0.129)***
Panel D: FDI
Trade	1.016 (0.083)***	0.912 (0.086)***	1.016 (0.082)***	0.917 (0.086)***
Industrial dissimilarity	0.001 (0.002)	0.004 (0.002)***	0.000 (0.001)	0.004 (0.002)**
Monetary policy	–0.025 (0.086)	–0.002 (0.090)	–0.005 (0.005)	0.018 (0.091)
Legal origin	0.001 (0.005)	0.003 (0.005)	–0.002 (0.005)	–0.001 (0.005)

Source: The author.

Notes: Standard errors are in parenthesis. Constant estimates are not reported. ***, ** and * significant at the 1%, 5% and 10% levels, respectively.

Given the above evidence that bilateral trade is positively linked and industry dissimilarity is negatively linked with the business cycle correlation of the sample countries, the other major concern raised in this article is to check whether FDI serves as another crucial channel in transmitting economic disturbances across countries. FDI is included in the model as it is expected to be an important source of shock transmission across borders in the current period of globalisation. Besides, FDI capital flows serve as a better proxy of financial integration.

Table 5 reports the EC3SLS estimates of the model including the FDI variable. These results show that FDI has a significant, sizeable, but negative impact on the output correlation of Eurozone economies. This is in contrast to our intuition that intensive FDI activities could contribute to output co-movements in the same way as trade activities would. In fact, FDI serves more a source of disturbance than a source of stabilisation/synchronisation. No doubt, even this disturbance is part of business cycles. This is even evident from: (a) now industry dissimilarity has insignificant impact on output correlation (Table 5, Panel A); (b) trade does not show any significant impact on industry dissimilarity (Panel C); and, (c) industry dissimilarity does not show any significant impact on both trade and FDI either at the 1 per cent or at 5 per cent levels when time effects are not considered (Panel B and Panel D). Nonetheless, (b) and (c) show a significant impact when the crisis period is not included, as reported in the next section (Table 7). But FDI does have some interesting influences in the sense that: (a) it has positive, significant and sizeable impact on trade flows and vice-versa (Panel D and Panel B), indicating that more FDI encourages more trade and higher trade between two countries increases FDI flows between them; (b) the roles of language, distance and GDP gap do not matter statistically and significantly in the current period of globalisation (Panel B and Panel C); (c) and that all other variables retain their signs and significance as exhibited in Table 4. However, legal origin or a common monetary policy does not have a significant effect on FDI flows between countries. All this clearly shows that FDI has an important influence on various macro-variables in highly open Eurozone Customs Union economies.

4.1. Robustness Check (Excluding the Crisis Period)

This section presents the estimated results of Equations (1) and (2) based on a sample data set restricted to the period between 1990 and 2006. The intention is to exclude the period of the US financial crisis and the Eurozone sovereign debt crisis. This alteration has nearly no effect on the sign and statistical significance of the coefficient estimates, except that some coefficients change marginally in terms of their magnitude. The only deviation from the earlier results is that now the impacts of bilateral trade on industry dissimilarity (Table 7, Panel C) and of industry dissimilarity on FDI (Table 7, Panel D) become statistically significant. The results corresponding to Equations (1) and (2) are reported, respectively, in Tables 6 and 7.

Table 6
Simultaneous Equations EC3SLS Panel Eurozone Data Estimates of Model (1), Excluding Crisis Period

Output measure HP-filtered
First-differenced

Time effects No Yes No Yes

Panel A: Correlation

Trade 2.581 (0.127)* 2.650 (0.125)* 1.896 (0.119)* 1.930 (0.118)*

Industrial dissimilarity –0.013 (0.001)* –0.013 (0.001)* –0.008 (0.001)* –0.008 (0.001)*

Monetary policy 0.121 (0.014)* 0.118 (0.014)* 0.110 (0.014)* 0.109 (0.014)*

Panel B: Trade

Industrial dissimilarity –0.002 (0.001)* –0.002 (0.000)* –0.002 (0.001)* –0.002 (0.000)*

PCGDP difference –0.006 (0.001)* –0.007 (0.001)* –0.006 (0.001)* –0.007 (0.001)*

GDP product 0.001 (0.001)* 0.001 (0.000)* 0.001 (0.001)* 0.001 (0.000)*

Language –0.024 (0.002)* –0.024 (0.002)* –0.024 (0.002)* –0.025 (0.002)*

Distance –7.98e–06 (8.22e–07)* –8.10e–06 (8.31e–07)* –6.13e–06 (8.32e–07)* –6.24e–06 (8.40e–07)*

Adjacency 0.038 (0.002)* 0.040 (0.002)* 0.040 (0.002)* 0.042 (0.002)*

Panel C: Industrial Dissimilarity

Trade –51.968 (3.475)* –43.858 (3.479)* –52.214 (3.476)* –44.004 (3.481)*

GDP gap 0.195 (0.043)* 0.197 (0.044)* 0.203 (0.043)* 0.204 (0.044)*

GDP product 0.118 (0.007)* 0.114 (0.007)* 0.119 (0.007)* 0.114 (0.007)*

Output measure	HP-filtered	First-differenced
Panel A: Correlation
Trade	2.581 (0.127)***	2.650 (0.125)***	1.896 (0.119)***	1.930 (0.118)***
Industrial dissimilarity	–0.013 (0.001)***	–0.013 (0.001)***	–0.008 (0.001)***	–0.008 (0.001)***
Monetary policy	0.121 (0.014)***	0.118 (0.014)***	0.110 (0.014)***	0.109 (0.014)***
Panel B: Trade
Industrial dissimilarity	–0.002 (0.001)***	–0.002 (0.000)***	–0.002 (0.001)***	–0.002 (0.000)***
PCGDP difference	–0.006 (0.001)***	–0.007 (0.001)***	–0.006 (0.001)***	–0.007 (0.001)***
GDP product	0.001 (0.001)***	0.001 (0.000)***	0.001 (0.001)***	0.001 (0.000)***
Language	–0.024 (0.002)***	–0.024 (0.002)***	–0.024 (0.002)***	–0.025 (0.002)***
Distance	–7.98e–06 (8.22e–07)***	–8.10e–06 (8.31e–07)***	–6.13e–06 (8.32e–07)***	–6.24e–06 (8.40e–07)***
Adjacency	0.038 (0.002)***	0.040 (0.002)***	0.040 (0.002)***	0.042 (0.002)***
Panel C: Industrial Dissimilarity
Trade	–51.968 (3.475)***	–43.858 (3.479)***	–52.214 (3.476)***	–44.004 (3.481)***
GDP gap	0.195 (0.043)***	0.197 (0.044)***	0.203 (0.043)***	0.204 (0.044)***
GDP product	0.118 (0.007)***	0.114 (0.007)***	0.119 (0.007)***	0.114 (0.007)***

Source: The author.

Notes: Standard errors are in parenthesis. Constant estimates are not reported. ***, ** and * significant at the 1%, 5% and 10% levels, respectively.

Table 7

Simultaneous Equations EC3SLS Panel Eurozone Data Estimates of Model (2), Excluding Crisis Period

Output measure	HP-filtered		First-differenced
Time effects	No	Yes	No	Yes
Panel A: Correlation
Trade	1.859 (0.151)***	1.671 (0.145)***	1.644 (0.149)***	1.435 (0.141)***
Industrial dissimilarity	–0.002 (0.003)	–0.001 (0.002)	–0.000 (0.003)	–0.001 (0.002)
FDI	–1.268 (0.161)***	–1.050 (0.149)***	–1.485 (0.161)***	–1.236 (0.148)***
Monetary policy	0.051 (0.146)	0.075 (0.143)	0.768 (0.142)***	0.787 (0.139)***
Panel B: Trade
Industrial dissimilarity	–0.004 (0.001)***	–0.004 (0.001)***	–0.003 (0.001)***	–0.004 (0.001)***
FDI	0.491 (0.038)***	0.430 (0.035)***	0.495 (0.037)***	0.432 (0.035)***
PCGDP difference	–0.003 (0.001)***	–0.003 (0.001)***	–0.003 (0.001)***	–0.003 (0.001)***
GDP product	0.013 (0.001)***	0.014 (0.001)***	0.012 (0.001)***	0.013 (0.001)**
Language	–0.001 (0.005)	0.001 (0.004)	0.004 (0.004)	0.006 (0.004)
Distance	–2.30e–07 (3.35e–06)	–8.95e–07 (3.34e–06)	–1.10e–06 (3.25e–06)	–1.89e–06 (3.25e–06)
Adjacency	0.026 (0.003)***	0.028 (0.003)***	0.026 (0.002)***	0.029 (0.002)***
Panel C: Industrial Dissimilarity
Trade	–11.536 (5.069)**	–13.456 (4.892)***	–15.217 (5.081)***	–16.737 (4.905)***
FDI	25.350 (4.466)***	28.874 (4.067)***	26.719 (4.467)***	30.012 (4.068)***
GDP gap	0.341 (0.075)***	0.314 (0.074)***	0.236 (0.075)***	0.215 (0.074)***
GDP product	0.462 (0.131)***	0.423 (0.129)***	0.571 (0.131)***	0.521 (0.129)***
Panel D: FDI
Trade	0.851 (0.087)***	0.789 (0.088)***	0.887 (0.086)***	0.823 (0.088)***
Industrial dissimilarity	0.008 (0.002)***	0.011 (0.002)***	0.008 (0.002)***	0.011 (0.002)***
Monetary policy	–0.064 (0.092)	–0.034 (0.094)	–0.006 (0.091)	0.014 (0.094)
Legal origin	0.003 (0.005)	0.004 (0.005)	–0.004 (0.005)	–0.001 (0.005)

Source: The author.

Notes: Standard errors are in parenthesis. Constant estimates are not reported. ***, ** and * significant at the 1%, 5% and 10% levels, respectively.

5. Discussion and Conclusions

The contribution this article aims to make is to find out the total impact of trade, industrial dissimilarity, and FDI on the business cycle synchronisation of Eurozone economies by using a panel data set covering the period 1990–2009. The findings of this study reveal that bilateral trade between two countries is positively and significantly correlated with business-cycle correlation between them,¹² implying that closer trade ties among these Eurozone countries result in more synchronised business cycle co-movements, because common disturbances are more prevalent and intra-industry trade dominates. Further, trade shows indirect effects on cycle synchronisation via FDI and ID (industry dissimilarity). Apart from trade, the study suggests that the channels of FDI, production specialisation and monetary policy closeness are almost equally important in explaining business cycle correlations.

Similarly, industry dissimilarity has both direct and indirect impacts on the output correlation of economies. However, though significant, industry dissimilarity’s direct impact on the business cycle synchronisation of the sample Eurozone economies is very small, indicating that the industry-specific shocks have almost lost their importance in raising output correlation of countries. In fact, these industry-specific shocks become insignificant when FDI flows are taken into consideration. Similar conclusions are drawn from the indirect effects of industry dissimilarity on the business cycle synchronisation of Eurozone economies via trade and FDI.

Nonetheless, the impact of bilateral trade flows on the business cycle synchronisation of Eurozone economies via industry dissimilarity is negative, significant and of a large magnitude (Table 4, Panel C). Even in the presence of FDI, this impact is retained when the crisis period is excluded (Table 7, Panel C). This goes against the positive linkage predicted between trade and specialisation by classical Ricardian theory, which means that industrial dissimilarities do not generate more trade in our case. From the viewpoint of international trade theories, developing countries or two sets of heterogeneous economies that are at different levels of economic development will have inter-industry trade. This trade is driven by differences in relative factor endowments, which means that goods that are produced with abundant factors will be exported and those that require scarce resources for their production will be imported. Such a hypothesis is tested by a positive coefficient of absolute differences in the per capita GDP variable. A negative sign on this coefficient is associated with intra-industry trade structure, which assumes the existence of simultaneous exports and imports of goods in the same sector but at the same or different stages of processing, that is, similar products of the same quality and technology. Thus, intra-industry trade targets the variety of products (e.g., the trade in cars in the automobile industry). Since this article considers highly integrated Eurozone economies, it is believed that intra-industry trade type exists among these economies. Such a result is confirmed by the negative coefficient of the per capita GDP difference variable in our analysis. This implies that intra-industry trade is taking place among these sample economies. This intra-industry trade is driven by economies of scale enjoyed by these Eurozone economies in varying degrees. However, contrary to our expectation, the two proxy variables—the GDP gap and GDP product—determining the developmental stages of industrial specialisation are significantly positive, thereby suggesting that these Eurozone economies have diversified even at higher levels of income. The gravity variables that supply the DNA of the geography of trade have the expected signs. For example, greater geographic distance between two countries reduces their bilateral trade because of high transportation costs, and sharing a common border boosts trade. Common monetary policy exhibits only direct effects and no indirect effects on the business cycle synchronisation of Eurozone economies.

Like trade and industry dissimilarity, FDI has both direct and indirect impacts on the output correlation of economies. FDI shows a negative but statistically significant impact on the output correlation of Eurozone countries, which implies that bilateral FDI serves more as a source of instability than of business cycle synchronisation. This might imply that these countries outsource part of their production processes in the form of foreign investments to other countries. It also implies that these countries enhance and replace their own technology through technology transfers and new investments. The bottom line is that it highlights the important role that foreign investment can play in generating different phases of business cycles in the current times.

As far as the indirect effects are concerned, FDI influences business cycle synchronisation of the sample economies via trade and similarities in industrial structure. The sizeable and significantly positive impact of FDI on industrial dissimilarity shows that FDI creates dissimilar industrial structures across these economies, suggestive of the fact that vertical investments co-exist in these economies. Moreover, trade intensity and FDI flows are positively and significantly related to each other. This means that more FDI encourages more trade and vice-versa, so trade and FDI complement each other. The incorporation of FDI in our model also reverses some of the interesting relations. For instance, industry dissimilarity has an insignificant impact on output correlation; monetary policy loses its robustness on the output correlation of the sample Eurozone economies; language and geographic distance lose their importance in determining the size of trade flows between countries; and common legal origins and/or common monetary policy do not determine FDI flows at all. In sum, besides trade and industrial dissimilarity, the present study shows the important influence that FDI has on the output correlation of Eurozone economies, and their impact on various other macro-variables, thereby suggesting the presence of indirect effects in these sample euro currency union countries. Furthermore, the study raises a question on the proper utilisation of FDI flows, at least in the case of these sample Eurozone economies.

Footnotes

Acknowledgements

This work was carried out when the author was a Junior Sir Ratan Tata Fellow (at the Institute of Economic Growth, New Delhi) under the Sir Ratan Tata Trust Fund for Young Social Scientists. An earlier version of this article was presented at a seminar at the Institute of Economic Growth and subsequently published in the form of a working paper (IEG Working Paper No. 371/2018). The author is very grateful to Prof. Pravakar Sahoo, Dr Sushil Sen, Sheeraz Ahmad Kunroo, Mudabera Gulzar and the anonymous referee(s) for their support and helpful comments on earlier drafts of this article.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no financial support for the research, authorship and/or publication of this article.

Notes

References

Artis

, & Okubo

(2011). Does international trade really lead to business cycle synchronization? A panel data approach. The Manchester School, 79(2), 318–332.

Balassa

(1986). The determinants of intra-industry specialization in United States trade. Oxford Economic Papers, 38(2), 220–233.

Baltagi

B. H.

(1981). Simultaneous equations with error components. Journal of Econometrics, 17(2), 189–200.

Baltagi

(2008). Econometric analysis of panel data. West Sussex: John Wiley & Sons.

Baltagi

B. H.

, Demetriades

, & Law

S. H.

(2009). Financial development and openness: Evidence from panel data. Journal of Development Economics, 89(2), 285–296.

Baxter

, & Kouparitsas

M. A.

(2005). Determinants of business cycle comovement: A robust analysis. Journal of Monetary Economics, 52(1), 113–157.

Bayoumi

, & Eichengreen

(1993). Shocking aspects of European monetary integration. Adjustment and growth in the European Monetary Union. In Torres

& Giavazzi

(Eds.), Adjustment and growth in the European Monetary Union (pp. 193–229). Cambridge: Cambridge University Press.

Bayoumi

, & Eichengreen

(1997). Ever closer to heaven? An optimum-currency-area index for European countries. European Economic Review, 41(3), 761–770.

Blanchard

O. J.

, & Quah

(1989). The dynamic effects of aggregate demand and supply disturbances. The American Economic Review, 79(4), 655–673.

10.

Blaster

(2015). Globalisation—prismatically unfolded: The economics of trade and growth underpinning globalisation. Retrieved from https://balaster.weebly.com/downloads.html

11.

Bussière

, Fidrmuc

, & Schnatz

(2008). EU enlargement and trade integration: Lessons from a gravity model. Review of Development Economics, 12(3), 562–576.

12.

Calderón

, Chong

, & Stein

(2007). Trade intensity and business cycle synchronization: Are developing countries any different? Journal of International Economics, 71(1), 2–21.

13.

Cerqueira

P. A.

, & Martins

(2009). Measuring the determinants of business cycle synchronization using a panel approach. Economics Letters, 102(2), 106–108.

14.

Cheng

I-H.

, & Wall

H. J.

(2004). Controlling for heterogeneity in gravity models of trade and integration (Working Paper No. 1999–010E). Federal Reserve Bank of St. Louis. Retrieved from http://research.stlouisfed.org/wp/1999/1999–010.pdf

15.

Clark

T. E.

, & van Wincoop

(2001). Borders and business cycles. Journal of International Economics, 55(1), 59–85.

16.

Cover

, & Mallick

(2012). Identifying sources of macroeconomic and exchange rate fluctuations in the UK. Journal of International Money and Finance, 31(2), 1627–1648.

17.

Darvas

, & Szapary

(2004). Business cycle synchronization in the enlarged EU: Comovements in the new and old members. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=508564.

18.

Dornbusch

, Fischer

, & Samuelson

P. A.

(1977). Comparative advantage, trade, and payments in a Ricardian model with a continuum of goods. American Economic Review, 67(5), 823–839.

19.

Enea

S- F.

, Palaşcă

, & Ţigănaş

(2015). G7 Countries—advocates of the global business cycle. Procedia Economics and Finance, 20, 193–200.

20.

Fidrmuc

(2004). The endogeneity of the optimum currency area criteria, intra-industry trade, and EMU enlargement. Contemporary Economic Policy, 22(1), 1–12.

21.

Frankel

J. A.

, & Rose

A. K.

(1997). Is EMU more justifiable ex post than ex ante? European Economic Review, 41(3), 753–760.

22.

Frankel

J. A.

, & Rose

A. K.

(1998). The endogenity of the optimum currency area criteria. The Economic Journal, 108(449), 1009–1025.

23.

Gächter

, & Riedl

(2014). One money, one cycle? The EMU experience. Journal of Macroeconomics, 42, 141–155.

24.

Di Giorgio

(2016). Business cycle synchronization of CEECs with the Euro area: A regime switching approach. Journal of Common Market Studies, 54(2), 284–300.

25.

Gonçalves

C. E. S.

, Rodrigues

, & Soares

(2009). Correlation of business cycles in the euro zone. Economics Letters, 102(1), 56–58.

26.

Hanus

, & Vacha

(2015). Business cycle synchronization of the Visegrad Four and the European Union (IES Working Paper No. 19/2015). IES FSV, Charles University. Retrieved from http://ies.fsv.cuni.cz

27.

Hodrick

R. J.

, & Prescott

E. C.

(1997). Postwar US business cycles: An empirical investigation. Journal of Money, Credit and Banking, 29(1), 1–16.

28.

Horvath

, & Rátfai

(2004). Supply and demand shocks in accession countries to the Economic and Monetary Union. Journal of Comparative Economics, 32(2), 202–211.

29.

Hsu

C-C.

, Wu

J-Yi.

, & Yau

(2011). Foreign direct investment and business cycle co-movements: The panel data evidence. Journal of Macroeconomics, 33(4), 770–783.

30.

Hsu

H-F.

(2010). Is a common currency area feasible for East Asia? A multivariate structural vector autoregression approach. Asian Economic Journal, 24(4), 391–411.

31.

Imbs

(2004). Trade, finance, specialization, and synchronization. Review of Economics and Statistics, 86(3), 723–734.

32.

Inklaar

, Jong-A-Pin

, & de Haan

(2008). Trade and business cycle synchronization in OECD countries—A re-examination. European Economic Review, 52(4), 646–666.

33.

Kolasa

(2013). Business cycles in EU new member states: How and why are they different? Journal of Macroeconomics, 38(B), 487–496.

34.

Kose

, & Yi

K-M.

(2002). The trade comovement problem in international macroeconomics (FRB of New York Staff Report No. 155). doi: 10.2139/ssrn.368201

35.

Krugman

(1991). Geography and trade. Cambridge, Massachusetts and London, England: MIT press.

36.

Krugman

(1993). What do we need to know about the international monetary system? Princeton, NJ: International Economics Section, Department of Economics Princeton University.

37.

Kunroo

M. H.

(2015). Theory of optimum currency areas: A literature survey. Review of Market Integration, 7(2), 87–116. doi: 10.1177/0974929216631381

38.

Kunroo

M. H.

, Sofi

I. A.

, & Azad

N. A.

(2016). Trade implications of the euro in EMU countries: A panel gravity analysis. Empirica, 43(2), 391–413.

39.

Kunroo

M. H.

, Sofi

I. A.

, Khurana

, & Mogha

S. K.

(2017). Re-visiting the impact of the euro on trade flows: New evidence using gravity equation with Poisson count-data technique. In Deep

, Bansal

J. C.

, Das

K. N.

, Lal

A. K.

, Garg

, Nagar

A. K.

& Pant

(Eds.) Proceedings of the sixth international conference on soft computing for problem solving. Advances in intelligent systems and computing (vol. 547; pp. 229–239). Singapore: Springer.

40.

La Porta

, Lopez

, & Shleifer

(2008). The economic consequences of legal origins. Journal of Economic Literature, 4(2), 285–332.

41.

Magkonis

, & Tsopanakis

(2014). Exploring the effects of financial and fiscal vulnerabilities on G7 economies: Evidence from SVAR analysis. Journal of International Financial Markets, Institutions and Money, 32, 343–367.

42.

Masson

P. R.

, & Taylor

M. P.

(1993). Policy issues in the operation of currency unions. Cambridge: Cambridge University Press.

43.

Miles

, & Vijverberg

C-P. C.

(2016). Did the euro common currency increase or decrease business cycle synchronization for its member countries? Economica. Retrieved from https://onlinelibrary-wiley-com-443.web.bisu.edu.cn/doi/abs/10.1111/ecca.12201

44.

OECD (2012). STAN indicators rev. 4, STAN: OECD structural analysis statistics (database). doi: 10.1787/data-00651-en

45.

OECD OECD international direct investment statistics, 2014. Paris: OECD Publishing. doi: 10.1787/idis-2014-en

46.

Otto

, Voss

G. M.

, & Willard

(2001). Understanding OECD output correlations. Sydney: Reserve Bank of Australia.

47.

Peersman

(2011). The relative importance of symmetric and asymmetric shocks: The case of United Kingdom and the euro area. Oxford Bulletin of Economics and Statistics, 73(1), 104–118.

48.

Rault

, Sova

, & Sova

A. M.

(2009). Modelling international trade flows between CEEC and OECD countries. Applied Economics Letters, 16(15), 1547–1554.

49.

Ravn

M. O.

, & Uhlig

(2002). On adjusting the Hodrick-Prescott filter for the frequency of observations. The Review of Economics and Statistics, 84(2), 371–380.

50.

Savva

, Neanidis

, & Osborn

(2010). Business cycle synchronization of the euro area with new and negotiating member countries. International Journal of Finance & Economics, 15(3), 288–306.