Macroeconomic and financial implications of multi-dimensional interdependencies between OECD countries

3.1 Trade

One mechanism through which a country-specific shock can have wider consequences in the international economy is trade. Being at the core of the OECD's interests since its inception, the exchange of goods and services between national economies is an uncontroversial element of exposure to the variability of a foreign economy both in terms of volumes (which are in turn, associated with productive trends) and prices (where additional factors such as financial or even fiscal variables have a considerable impact).

Recent studies such as Beetsma and Giuliodori (2011), and Inklaar, Jong-A-Pin, and De Haan (2008) focus on analysing the importance of the trade channels and assessing their role in the transmission of spillovers from national policies. They were able to confirm the decisive influence of trade exchanges in the transmission of shocks between the OECD economies, suggesting that trade integration strengthens business cycle synchronization. Besides trade dependency, the international trade literature also emphasizes the importance of commodity and geographical trade clusters in relation to the escalation of spillover effects (Cheewatrakoolpong & Manprasert, 2014). Kali and Reyes (2010) also conclude that countries that become centres of trade are highly prone to suffer from serious detrimental effects during crises. In turn, a number of recent works on international spillovers have discussed in depth how crucial trade linkages are in terms of a country's vulnerability to a crisis as well as the effect of increased trade linkages that originates from a process of integrated activity in core areas that are deeply connected to the economic cycle.

OECD countries are strongly interconnected by trade linkages. The intra-industry trade in goods between OECD member states was 60–70% larger than the flow of exports leaving the OECD to non-member countries (OECD, 2010). Cross-border intermediate and final goods interrelations can have substantial implications for propagation mechanisms and the linkages among demand, trade, and production across OECD countries. Indeed, the literature demonstrates that a large fall in demand of goods and services played an important role in driving the trade collapse during the global recession of 2008–2009 (Alessandria, Joseph, & Virgiliu, 2010; Eaton, Kortum, Neiman, & Romalis, 2016). Similarly, Bems, Johnson, and Yi (2010) show that demand forces can account for roughly 70% of the trade collapse and that trade linkages are substantially important in terms of each country's vulnerability to recent crises.

Bems et al. (2010), an authoritative study of its kind, suggest that trade could transmit crises internationally through a combined integration of three channels: (a) a competitiveness effect, (b) an income effect and (c) a cheap-import effect. A competitiveness effect occurs when devaluation of one country's currency drives up the relative competitiveness of its exports and diminishes the competitiveness of exports from other countries. An income effect occurs when a crisis affects income and growth within the crisis country, by reducing purchases of imports from abroad. A cheap-import effect occurs when devaluation of one country's currency down the relative price of its exports and thereby reducing prices in countries that import these goods.

Within the OECD context, Eichengreen and Rose (1999), utilizing a trade-weighted matrix of macroeconomic variables, show that trade linkages instead of macroeconomic similarities have acted as a crucial channel for the contagion mechanisms of transmission between economies. Similarly, using a global input–output framework, Bems et al. (2010) show that the large magnitude of trade spillovers originating from and within many OECD countries played a crucial role in explaining both the collapse of trade and transmission of the global recession. Several studies such as Alessandria et al. (2010) and Eaton et al. (2016) conclude that trade linkages were large and significant in the context of international transmission of a crisis. Therefore, the first part of the multi-dimensional interdependency matrix must focus on the role of OECD's deeply connected trade linkages and quantify the role of trade in the exposure to shocks across OECD economies.

Identifying the relative degree of vulnerability towards a specific economy has been done through the use of an indicator of bi-lateral trade-openness by using total trade in terms of exports and imports to reflect importance of country i for country j⁵ as outlined by Pesaran et al. (2004) as:

(3.1)

with i ≠ j and where _i,j and _i,j represent, respectively, the mean of exports and imports between country i and country j during the studied time horizon.

3.2 Finance

Another important thread in the links between economies is found in the exchange of financial resources. The financial channel of transmission has been studied in the context of macroeconomic policies by Faini (2006) with special attention to interest rates and fiscal spillovers as well as by Corsetti and Müller (2011) who conclude on the dominant effect of financial factors in an international transmission mechanism that operates on domestic and foreign private spending through expectations, interest rates and international asset markets, in that sequence, extending the study on the role and impact on private spending of the sovereign risk channel as analysed in Corsetti, Pericoli, and Sbracia (2011) in an international, interdependent context.

As far as financial interdependencies are concerned, OECD (2018) suggests that the degree of financial interlinkages between OECD economies has increased significantly over the last two decades. The importance of direct investment, in particular, has grown notably over time. The escalation of direct investment flows is partly explained by increased global cross-border mergers and acquisitions to a certain extent, following a decreasing pattern after the global financial crisis. Global mergers and acquisitions have not only resumed again recently, but also its regional composition changed (OECD, 2018). The EU countries position has changed from being net outward investors to becoming the world's largest net recipient of cross border merger and acquisitions. In addition to foreign direct investment exchanges, the outstanding stock of gross financial derivatives has also increased as financial products became more complex as a result of financial innovation over the last decade (OECD, 2018). OECD statistics suggest that in 2005–2007, the total volume of international mergers and acquisitions formed up to 80% of global foreign direct investment flows which then followed a decreasing pattern to under 60% during 2008 and 2014 (OECD, 2014). Since then it has been following an increasing pattern with substantial individual cross-border deals taking place (UNCTAD, 2017).

In most OECD economies, the total volume of capital investments by foreign-owned firms is around 2% of domestic GDP, and this ratio is significantly higher in smaller sized economies (OECD, 2015; OECD, 2018). Recent OECD figures suggest that, in the majority of advanced and emerging OECD economies, the stock of foreign direct investment (FDI) assets and investment income from FDI as a percentage of GDP have doubled or tripled over the past two decades, although the share is smaller in emerging than in advanced economies (OECD, 2018). This is noted particularly in the United Kingdom with 4.5% of its GDP is formed by FDI investment income, then in Canada with 3.5% of its domestic GDP followed by France, Germany and the United States (in a descending order). As far as the stock of FDI assets are concerned, this ratio is significantly higher in the United Kingdom with 80% of its domestic GDP, followed by France (75%), Germany (65%) and Italy (45%).

Increased gross asset and liability positions as well as growing bilateral direct investment exchanges translate into international exposures, turning financial centres as increasingly vulnerable to shocks which makes the exact identification of exposures a complex task. Other consequences, resulting from increased financial interconnectedness, are mainly base erosion, profit shifting, exposure to the risk of illicit financial flows – such as money-laundering, tax evasion and international bribery – all at the forefront of the OECD agenda (OECD, 2018). The degree of financial interlinkages among Western European countries and those in Central, Eastern, and South-eastern Europe (CESE) has grown notably as well, with a consequential increase in foreign ownership of the CESE banking systems. Arvai, Driessen, and Otker (2009) show that the financial interlinkages in Europe are economically significant. Also, it should be noted that intra-EU FDI flows rose by 40% in 2015 and reached 365 billion (data from Eurostat).

A recent OECD study (OECD, 2018) underlines the importance of financial interdependencies for the propagation of economic shocks across borders. They demonstrate that geographical diversification of increased cross border flows enables financial markets to grow and diversify risks, assisting companies to expand and diversify the risks by investing where the rate of return is highest while mitigating negative domestic shocks. On the other hand, such connectedness increases the risks of exposure to negative spillovers across borders. In turn, changes in economic conditions abroad become more influential than national economic policies, particularly on domestic investment and employment decisions as well as corporate earnings. Such connectedness and increased geo-diversification explain the significant role played by global factors in equity prices (OECD, 2018). There is a great deal of literature on the effects of financial interconnectedness on the growth in volatility that supports OECD's view (see, e.g., Kose, Prasad, Rogoff, & Wei, 2009; Prasad, Rogoff, Wei, & Kose, 2007). Moreover, increased financial flows, particularly FDI and portfolio flows, have also been shown to raise income inequality in both advanced and emerging market economies (Basu & Guariglia, 2007). In addition to these papers, other studies of scholars like Kazi, Mehanaoui, and Akbar (2014) and Guesmi, Kaabia, and Irfan (2013) have documented the substantial importance of financial exposures across OECD countries.

The complexity and variety of the interactions discussed above, in turn, highlight the necessity to concentrate on specific aspects as FDI positions given the role they play for the transmission of economic shocks across borders⁶ which, as explained above, is intended to reflect a more structural element of financial exposure as opposed to short-term (and possibly speculative) financial flows on the basis of the former having a stronger relationship with the real sector of recipient countries and, therefore, with macroeconomic aggregates such as consumption, investment, and output.

For the reasons outlined above, the second part of the multi-dimensional interdependency matrix focuses on the increased role of external financing as a source of funding and aims to represent the extent of possible channels of contagion through financial exchanges. This component represents the role of FDI as a factor of international financial exposure, specifically calculated as:

(3.2)

with i ≠ j and where and stand, respectively, for the means of total outflows and inflows of direct investments between country i and country j.

3.3 Bilateral migration flows

An under-explored element in previous literature is migration, which is typically associated with allocative, distributive and external effects.⁷ Migration of various kinds within OECD countries has become increasingly common as a way for individuals to diversify household income and reduce the divide they perceive between their own circumstances and those in more advanced countries. A recent OECD study (OECD, 2017a) shows how increasingly important migration exchanges have become in member economies. Overall, the migrant stock made up for 13% of the population in OECD countries in 2015, which was around 10% in 2000 (OECD, 2017a). In addition, around 5 million people migrated permanently to OECD countries in 2016, well above the pre-crisis previous peak in 2007. There was also a surge in asylum applications in 2015–2016. Global international tourist arrivals rose by over 4% per annum on average in the two decades of 2016 (OECD, 2018).

Migration linkages between economies are associated to economic and financial impacts which have not been captured by conventional measurements of international interaction. Trade or financial statistics, for example, fail to properly reflect the economic exposure to the effects of migration on productivity, public finance (both on income and expenditure, see Lisenkova et al., 2014), wages (Damette & Fromentin, 2013) and aggregate demand.⁸ Current migration links, between not necessarily neighbouring countries, form intricate networks displaying distinctive, typically asymmetric, intensities and therefore cannot be represented by geographical distances either.

Indirect improving effects of migration exchanges operate through three main channels: (a) labour markets, (b) remittances and (c) international students and tourist traffic. Within each bi-lateral relationship, the composition of migration exchanges, similarly, brings about consequential effects at both ends of these links. The movement of skilled migrants, for example, whether temporary (cyclical) or in the longer term, constitutes a significant factor for the performance of increasingly international business models with accentuated use in advanced OECD economies.⁹ Similarly, at the lower end of the skills range, shortages of human resources in source countries for example, have implications on productivity and household income. In this paper, we do not differentiate by skills in migration exchanges which means all the described relationships are equally contained in our weighting scheme.

In 2015, total remittance flow from migrants in OECD countries to their home countries was around 500 billion USD. In the 21st century, migrants are seen not only as a source of foreign currency, but also as bearers of skills, knowledge and social remittances (Faist & Fauser, 2011; Levitt, 1998). This view often contains the underlying assumption that migration has significant potential for tackling inequality, either through the cash remittances that migrant workers are able to send back, or through the skills, technology transfer and knowledge they acquire abroad and then take back to their country of origin. Social remittance flows are often larger than either bilateral aid or foreign direct investment (Ratha & Shaw, 2007). Kapur and McHale (2009) argue that emigration from poorer to richer countries increases the income of the migrants, and also that of their relatives who remain in the country of origin. In fact, migration holds greater potential for reducing inequality when the migrants are from poorer households. In addition to long term migration's potential impacts on labour market and remittances, increasing flow of short-term migration such as international students (OECD, 2017a) and growing international tourist traffic facilitate new digital technologies usage. Thus, increased flow of short-term migration has a knock-on effect on creating economic opportunities, enabling firms to access new markets, and offering consumers a wider choice of goods and services (OECD, 2017b).

As discussed above, migration exchanges in the OECD can have significant impacts on economic performance, primarily through the pressure they put on labour markets. Our inclusion of migration weights is ultimately aimed at reflecting the intensity and asymmetric nature of these bi-lateral relationships between economies in the studied sample (not to be confused by the reader with any attempt to simulate shocks from variations in migration). Heterogeneities between countries can be marked as they assume distinct roles such as origin or destinations with different intensities in the exchanges for which a differentiating factor is often the stance of the immigration policies they apply (see Pedersen, Pytlikova, & Smith, 2008). Although deeply constrained by data limitations, analysing the bi-directional linkages is a crucial task for a more complete understanding of international exposures. However, no attempt has been made in the GVAR literature to measure these cross-border factors. Thus, to account for the role played by migration as transmission mechanisms, total bilateral stocks of migration are used for formulating an index of the relative position between countries. Data from the World Bank has been used for these purposes (see the Data section below). Two pre-2008 financial crisis periods, 1990 and 2000, were selected in an attempt to depict more structural, long-term relationships in terms of migration. Then, given the availability of data, a higher weight (70% in each bi-lateral set) was assigned to the most recent series, 2000, which sits approximately in the middle of the typical timespan of other macroeconomic series in the study. That weight is a choice largely made to emphasize the higher relevance of the most recent available series in the context of the overall time-coverage of our analysis.

Each bi-lateral relationship is, this way, calculated as:

(3.3)

where, typically, Migr_i,j,y ≠ Migr_j,i,y.

3.4 Geographical proximities

Tobler's (1979) first law of geography, ‘everything is related to everything else, but near things are more related than distant things’ is appropriate here as an indicator of geographical proximities between economies can absorb other channels of interaction not captured by the other three channels outlined above. On the other hand, while geographic closeness does not strictly precondition the intensity-based ordering of the first three considered relationships, it is still a factor of interaction in other aspects with final expression in economic and financial performance. Among those, we can mention institutional and legal similarities which benefit contractual arrangements, technology transfers, educational and scientific cooperation, common development projects and their funding, regional (including inter-city) agreements, etc. This theme surfaces in Krugman's (1991) work emphasizing the importance of locational decisions by various centripetal forces and linkages that are caused by large local markets. They have demonstrated that economic integration in geographic regions has been central in terms of concentration of economic activities.

Studies analysing the effect of economic geography generally fall into two categories, with one strand consisting of the studies of Brakman, Garretsen, and Schramm (2004) for Germany, Hanson (2005) for the United States and Mion (2004) for Italy, who all assume labour to be perfectly mobile and real wages to be equalized. The second strand meanwhile concentrates on the same effect at international level, and is represented by the work of Redding and Venables (2004), in which real wage levels are influenced by intermediate factors of production. Studies at both national and international level have shown geography to have a significant impact on access to markets, which then shapes income levels. Krugman (1991) and Venables (1994) also document the economic relevance of such proximities in examining regional integration. More recent studies such as Redding and Venables (2003) produced an authoritative study of its kind, showing that the geography of access to markets and sources of supply is statistically significant and is essential in explaining cross-country variation in income per capita. Around 29% of cross regional variation in wage levels can be explained by a region's proximity to customer markets (Lopez Rodriguez & Faina, 2007).

Taking into account the context discussed above, we propose to add a proxy to measure those additional linkages based on the country's location within the OECD space. The inclusion of geographic proximity (1/dist_i,j, where dist_i,j is the geographic distance between countries i and j) in the weighting scheme, thus, serves the purpose of representing other channels of interaction between economies not captured by the previous three measurements. An index of inverse distances that assigns greater weight to closer neighbourhoods is calculated using data from the World Bank's API¹⁰ (geo-localization of capital cities). Inverse-distance weights are commonly used in configurations of spatial econometrics involving units that are geographically dispersed (LeSage & Pace, 2009).

Given the presence of dominant hubs of regional activity, the current context of economic, financial and migratory interactions (all three included in a hypothetical matrix X) between pairings of countries, i, j and i, k, shows that violations to the rule:

are frequent, especially when country i is one of such hubs.¹¹ This way, geographic distances are considered to be true complements of the other three channels.

In this study, using geographic data from the World Bank's Indicators API, an index with inverse distances between capital cities, W_i,j, has been calculated as:

(3.4)

3.5 Composite weights based on key linkages

While proposing an innovative scheme of weights in GVAR models, we are well aware that such a complex picture of cross-country interdependencies will require a great deal of acceptance from scholars who have predominantly concentrated on trade linkages.

Further justification for the weighting scheme draws from the suggestions in three well-known papers. First, Eickmeier and Ng (2015) consider a range of different connectivity matrices and assess differences through a forecasting exercise. Second, Crespo Cuaresma, Feldkircher and Huber (2016) analyse weight schemes that allow for a variety of weights for different foreign variables in the system, specifically evaluating alternative weight schemes to suit the likelihoods of the GVAR model. Third, Gross (2013) proposes estimating weights rather than choosing them exogenously for the model.

To avoid a potential problem of endogeneity in the formulation of our composite weights, the study utilizes averages, formulated in a way that their outcomes are as disengaged from quarterly policy variations as possible. By using either temporarily long or very punctual averages in time according to data availability with the prime objective of measuring how important country i has been for country j in terms of the considered exchanges and features. As described above, trade and financial weights, for example, calculate averages over the entire timespan of the series so that quarterly policies are completely diluted in these calculations (macroeconomic policies would require a large, multi-annual persistent stance to generate a significant endogenous association with these metrics, such a scenario is implausible). At the other extreme, migration weights rely on only two dates 1990 and 2000. The short-term dynamics of variations in immigration policies, for example, would have little impact on the snapshots of the flows taken in those dates.¹²

The final composite matrix, Z (see Appendix C and Figure C1), is calculated from the corresponding entries in each of the previous matrices as:

(3.5)

with Z_i,i = 0 being a result of the null diagonals in all of the constituent matrices. For the purposes of estimations these weights have been column-normalized.¹³

Based on historical trend as well as volume of trade and financial exchanges (see Appendix C, Figures C3 and C4) between OECD economies, trade and financial channels are given preponderant roles in the depiction of exposures to foreign variables (adding up to 70% of the total weights).

In turn, as discussed above, bilateral migration exchanges are also considered consequential factors of exposure. Given the relative importance of such linkages in the OECD context (see Appendix C, Figure C5), migration holds the role of a secondary factor of exposure (25% of the total weights, which is lower that the weights assigned to trade and finance). Lastly, geographical proximities are also taken into consideration in order to include other aspects of spatial interdependencies among OECD economies (see Appendix C and Figure C2). Their residual, although complementary, role is reflected in a lower participation in the final composite weights (5% of the weights).

It is worth noting that, although the weights assigned to each contributing factor of exposure are the product of the authors' choice, this is not arbitrary or random as it takes into account observed features of international interaction. Most importantly, while showing consistency with the prevailing descriptions of foreign exposure found in previous literature (where trade and finance are the key, predominant drivers), they allow for the inclusion of secondary and residual factors of interaction acknowledging, at the same time, their complementary roles in a scheme that ensures comparability across the sample (as opposed to an alternative where the weightings are the product of estimations).

In addition, the reader should recall that, in the context of our study, weights are intended to provide a measurement of the intensity of relevant relationships between country pairings. The variables representing those weights are, therefore, not used as direct factors of shock-simulation. Their participation in the modelled system is restricted to the measurement of multi-dimensional exposures to foreign shocks. Key qualities of the final weights, Z, are (a) their reflection of multiple factors intervening in economic and financial interactions/exposures with intensities specific to each country pair in a standardized, coherent scheme, (b) their depiction of asymmetries in relation to the included exposures (except only for the case of geographic distances) and (c) their flexibility to materialize only in the values strictly supported by the underlying data (if two countries have very low migration exchanges, then Migr_i,j → 0, for example, without affecting any of the depicted exposures to other channels which would still vary within their own respective ranges).

Additionally, regional aggregation weights w_{i, ℓ} required by the GVAR framework are defined following.

Dées, Mauro, Pesaran, and Smith (2007) as the averages of GDP-purchasing power parity (PPP) individual series between 1995 and 2013 (annual data), :

(3.6)

with N_ℓ representing the number of countries in the region ℓ. The data used for the calculation of these weights and the result of its application to the variables in the model are presented in Appendix D.

3.6 Data

This study's sample consists of 24 selected OECD countries'¹⁴ quarterly data between 1989Q4 and 2013Q3 (N = 24, T = 96) comprising of geographic data, macroeconomic aggregates, financial indicators as well as migration and key open-economy variables.

4.1 Contemporaneous effects of foreign variables on their domestic counterparts

The cross-country impact elasticities from changes in foreign variables provide a crucial source of information in relation to the interactions and potential spillovers between economies. The setting chosen in the study allows us to examine the elasticities with respect to variations in the global environment as seen from the perspective of each economy. Those elasticities, computed from the cointegrating VECMX models²⁰ can be understood as the contemporaneous change of a domestic variable as a result of a 1% variation in its foreign equivalent.

The complete table of elasticities is included in Appendix E jointly with their corresponding Newey-West t-ratios (based on heteroskedasticity and autocorrelation corrected standard errors). From them, the following substantial findings should be highlighted: Sweden, Portugal, Germany, Norway, Japan, and Italy (in that order) present the largest elasticities to foreign variations in GDP. Although this may not be surprising for the smaller economies in this list, it is particularly relevant in the case of Germany and Japan, given the relative size of their economies, since it reflects a high degree of exposure to global levels of activity. Contrastingly, the elasticity in the case of the U.S. economy is among the lowest. This indicates that even with the considerable size of the international exchanges it has with the world, the U.S. domestic economy displays a high degree of resilience against external variations in economic activity.

Large elasticities (most of them larger than one) to global variations in lending rates are shown by Austria, Canada, Netherlands and Finland revealing their high exposure to the patterns followed by foreign financial variables. The relative depth of their involvement in foreign financial exchanges as compared to the size of their own economies (notably in the case of the Netherlands) appears to be an important factor of this sensitivity, particularly when compared to cases like Germany with a larger domestic economy and, then, a much lower elasticity to global variations in this respect.

Similarly important in this evaluation of the global circumstances surrounding credit markets are the results showing that national corporate borrowing rates are markedly susceptible to global disturbances. Sweden, Spain, France, Italy, Canada, Germany, the United Kingdom, Ireland and Netherlands all show statistically significant elasticities ranging from 0.78 to 1.38, which indicates that the foreign component of this variable has consequential implications for the operation of credit markets including cases of large financial hubs like the United Kingdom and Germany. A consequential finding in terms of elasticities to global variations is the one that shows a high degree of responsiveness of share prices. Ten of the economies in the study sample display elasticities close to one or even considerably larger, no other variable displays the same levels of sensitivity to foreign developments. This feature is shared indistinctively among the economies with diverse sizes and global roles as Greece, Ireland, Finland, Italy, Germany, Japan, France, Portugal, Spain, and Sweden; all at the top of this scale. This exhibits not only the considerable degree of their financial markets integration, but also the fact that some economies would be in better positions to face large variations while others, in this list, for example, would find such events highly detrimental.

4.2 Dynamic analysis

This section addresses the analysis of the model's dynamics through a number of shocks intended to reveal two main features: (a) the heterogeneous response to comparable disturbances across regions and (b) the impact and persistence of spillovers between regions. The study uses generalized impulse-response functions (GIRFs),²¹ taking advantage of their independence of the ordering of the variables and, crucially, of their ability to express the features of dissimilar responses to the modelled disturbances²²in a comparable way.²³ The graphs in this section display two main components: (a) the own-region effect of the modelled shock and (b) its inter-regional spillovers. For comparative purposes, each panel corresponds to an independent scenario as described in its caption.