International business cycle synchronization: what is the role of financial linkages?

Ambrogio Cesa-Bianchi, Jean Imbs and Jumana Saleheen.

It is a well-known fact that financial integration has increased dramatically over the past few decades.  Has this rise led to higher or lower business cycle synchronization? The answer depends crucially on the source of the shock.  In response to common shocks, financial integration tends to lower business cycle synchronization.  But in response to a country-specific shock, business cycles are more synchronised between countries that are more financially integrated.

Why are business cycles synchronized across countries?

The recent global financial crisis has identified financial market integration as the major culprit for the  large and fast diffusion of shocks across borders.  In an influential paper, Rey (2013) describes the presence of a Global Financial Cycle, where an “international credit channel” enables financial conditions to propagate across borders.  While cycles are unambiguously synchronised between trade partners, there is less agreement about the role played by  greater financial integration.

Theory is an imperfect guide.   In a two-country real business cycles model, a country-specific productivity shock – that creates a disparity between the return to investment across countries – makes capital flow to wherever returns are higher  (or to ‘make hay where the sun shines’).  So, greater financial integration lowers the international synchronization of business cycles (Backus, Kehoe, and Kydland, 1992).  But in a similar two-country model, augmented with credit or collateral constraints, a country-specific shock that makes the constraint bind at home is contagious abroad, as domestic agents recall foreign assets to meet their constraint (Devereux and Yetman, 2010; Kalemli-Ozcan, Papaioannou, Perri, 2013; Allen and Gale, 2000). So, greater financial linkages can increase the international synchronization of business cycles.

One common feature of these models is that they analyze the consequences of purely “idiosyncratic” (i.e., country-specific) shocks.  But they are silent about what happens to business cycle synchronisation in the face of “common shocks”.  In a recent paper, we show that accounting for common shocks – and their heterogeneous effects across countries – has first-order consequences on the estimated impact of financial integration on business cycle synchronization.

On the one hand, we show that capital tends to flow between the same pairs of countries in the face of common shocks.  In response to positive shocks, capital tends to flow from countries that are relatively unresponsive (or “inelastic”) to common shocks to countries that are relatively responsive (or “elastic”) to common shocks.  The opposite happens in response to negative common shocks.  This generates a divergence in their business cycles, implying that financial flows do not help ‘make hay where the sun shines’.  On the other hand, we document a positive association between business cycle synchronization and financial linkages in response to idiosyncratic shocks.

Measuring business cycle synchronization: the role of common and idiosyncratic shocks

A standard and simple measure of cross-country business cycle synchronisation is the opposite of the pairwise difference in annual GDP growth rates (in absolute value).   This variable, labelled S_{ijt} , is intuitive: it takes negative values, increasing with the degree of synchronization between a pair of countries, and reaching an upper bound at zero for perfectly synchronized countries.

The measure S_{ijt}  presents two key advantages: first it is readily observable at yearly frequency. Second, unlike the Pearson correlation coefficient, it is invariant to the volatility of the underlying shock (see Forbes and Rigobon, 2001; Corsetti, Pericoli, and Sbracia, 2002). Figure 1 reports the behaviour of S_{ijt}  in the cross-section of 18 advanced economies from 1980 to 2012.

Figure 1.  A simple measure of cross-country business cycle synchronisation.S

Note.  The solid line plots the evolution over time of the average value of S_{ijt}  for the 1980-2012 period.  The average is computed across 153 country pairs (our sample spans 18 countries) for each year.

Figure 1 reveals a somewhat counterintuitive fact: the standard measure of synchronisation, S_{ijt} , falls in 1991, when there was a recession in a number of advanced economies.  And it falls sharply again during the Global Financial Crisis.

This is due to the fact that the simple measure S_{ijt}  does not take into account (i) that GDP growth rates respond to both country-specific shocks as well as common shocks; and (ii) that common shocks have heterogeneous effects across countries.  In other words, even though GDP growth rates were falling in most countries during Global Financial Crisis, the differential pace of contraction across countries shows up in S_{ijt}  as a fall in synchronisation. For example, US annual GDP growth went from about -0.3 in 2008 to about -2.8 percent in 2009, whereas UK GDP growth went from -0.3 to -4.3 percent.  As a result, the synchronization measure S_{ijt}  fell from 0 to -1.5.

To address this, we decompose country-specific GDP growth rates into a common and an idiosyncratic component.  To identify common shocks and their heterogeneous effects across countries, we use principal components techniques that allow country-specific loadings on the factors.

This decomposition also allows us to decompose the traditional measure of business cycle synchronization in two elements: a synchronisation measure that is based on the pairwise difference of GDP’s response to idiosyncratic shocks (S^{\epsilon}_{ijt} ); and a synchronisation measure that is based on the pairwise difference of GDP’s response to common shocks (S^{F}_{ijt} ).  We report this decomposition, together with the original measure S_{ijt} , in Figure 2.

Figure 2.  The conventional measure of cross-country business cycle synchronisation and its decomposition in idiosyncratic and common components.


Note.  The solid line plots the evolution over time of the average value of  for the 1980-2012 period.  The chart also reports the cross-sectional averages of the idiosyncratic component (S^{\epsilon}_{ijt} , dashed line) and the common component (S^{F}_{ijt} , dotted line) of S_{ijt} . The average is computed across 153 country pairs (our sample spans 18 countries) for each year.

Figure 2 confirms our above conjecture, showing that the decline in S_{ijt}  observed during the 1991 or 2008 recessions is clearly associated with common shocks, as S^{F}_{ijt}  drops substantially in both cases.  In contrast, S^{\epsilon}_{ijt}  increases in 2008, and does not fall in 1991.  This suggests the measure S_{ijt}  conflates two mechanisms: the international propagation of idiosyncratic shocks, S^{\epsilon}_{ijt} , and the international equilibrium response to common shocks, S^{F}_{ijt} .  The former is a measure of co-movement across countries; the latter is a measure of the dispersion in GDP growth rates in response to common shocks.  The presence of these two elements needs to be taken into account by empirical work that tries to estimate the effect of finance on synchronization.

Reinterpreting the evidence

Armed with the above decomposition, we revisit the evidence of the impact of finance on synchronisation.   Our approach builds on the conventional in the literature (see Kalemli-Ozcan, Papaioannou, and Peydro, 2013): we regress business cycle synchronisation on the intensity of bilateral bank linkages (using  BIS Locational Banking Statistics).

We run regressions using the three measures of synchronisation described above, namely S_{ijt} , S^{F}_{ijt} , and S^{\epsilon}_{ijt} .   The standard measure of synchronisation, S_{ijt} , implies the well-known result that increased financial linkages lower business cycle synchronisation, a finding that jars with what was observed during the global financial crisis.

We then show that this same result holds when using S^{F}_{ijt} , i.e.  when synchronization is conditioned on common shocks only.  The result comes from permanent differences across countries in terms of how growth and capital flows respond to common shocks.

What happens when we use S^{\epsilon}_{ijt} ? This synchronization measure should capture the equilibrium response of GDP to country-specific shocks, i.e.  the object of interest of most international business cycle models.  We find that, in response to idiosyncratic shocks, increased financial integration increases business cycle synchronization.


Do more financially integrated economies have more synchronised business cycles?  The answer crucially depends on the source of the shock.  Using a sample of advanced economies, we find that the answer is yes if we use a synchronization measure that strips out the heterogeneous responses of GDP to common shocks.  Differently in response to common shocks, financial flows seem to respond to “safe haven” motives, flowing systematically between the same economies and lowering their business cycle synchronization.

Our result points to the crucial role played by financial frictions in the international transmission of shocks, even in “tranquil” times.  While in a frictionless world financial integration allows perfect risk sharing, when financial frictions are at work a strategic trade-off between openness and stability may emerge.

Ambrogio Cesa-Bianchi works in the Bank’s Global Interconnections and Spillovers Division, Jumana Saleheen is Technical Head of Division in the Bank’s Financial Stability Directorate and Jean Imbs works at the Paris School of Economics and a Research Director at France’s CNRS.

If you want to get in touch, please email us at You are also welcome to leave a comment below. Comments are moderated and will not appear until they have been approved.

Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

A version of this blogpost has previously been published on VoxEU