Financial deglobalisation in banking?

BIS Working Papers

No 650

 

Financial deglobalisation in banking?

by Robert N McCauley, Agustín S Bénétrix, Patrick M McGuire and Goetz von Peter

 

Monetary and Economic Department

June 2017

 

JEL classification: F36, F4, G21

Keywords: Financial globalisation, international banking; consolidation; ownership

 

BIS Working Papers are written by members of the Monetary and Economic Department of the Bank for International Settlements, and from time to time by other economists, and are published by the Bank. The papers are on subjects of topical interest and are technical in character. The views expressed in them are those of their authors and not necessarily the views of the BIS.

This publication is available on the BIS website (www.bis.org).

 

© Bank for International Settlements 2017. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated.

ISSN 1020-0959 (print)

ISSN 1682-7678 (online)

Financial deglobalisation in banking?

Robert N McCauley, Agustín S Bénétrix, Patrick M McGuire and Goetz von Peter

 

 Abstract

This paper argues that the decline in cross-border banking since 2007 does not amount to a broad-based retreat in international lending (“financial deglobalisation”). We show that BIS international banking data organised by the nationality of ownership (“consolidated view”) provide a clearer picture of international financial integration than the traditional balance-of-payments measure. On the consolidated view, what appears to be a global shrinkage of international banking is confined to European banks, which uniquely responded to credit losses after 2007 by shedding assets abroad – in particular, reducing lending – to restore capital ratios. Other banking systems’ global footprint, notably those of Japanese, Canadian and even US banks, has expanded since 2007. Using a global dataset of banks’ affiliates (branches and subsidiaries), we demonstrate that the who (nationality) accounts for more of the peak-to-trough shrinkage of foreign claims than does the where (locational factors). These findings suggest that the contraction in global lending can be interpreted as cyclical deleveraging of European banks’ large overseas operations, rather than broad-based financial deglobalisation.

 

Keywords: Financial globalisation, international banking; consolidation; ownership

JEL classification: F36, F4, G21.


 Bank for International Settlements (BIS), Trinity College Dublin, BIS and BIS, respectively. The authors thank Claudio Borio, Jaime Caruana, Ingo Fender, Philip Lane, Swapan Pradhan, Hyun Song Shin, Philip Wooldridge, Charles Wyplosz, and participants in a BIS seminar for comments. We thank Jakub Demski, Pablo Garcia-Luna and Jose Maria Pastor Vidal for research assistance. The views expressed are those of the authors and not necessarily those of the Bank for International Settlements or Trinity College Dublin.

 

1. Introduction

"Peak trade" denotes the hypothesis that international trade is no longer growing faster than global GDP, which may preclude the strategy of trade-led economic growth.[1] A parallel thesis, which Caruana (2017b) dubs "peak finance", asserts that global finance has passed its high point and that financial deglobalisation has begun.[2] In particular, observers have interpreted the decline since 2007 in cross-border bank lending as evidence of financial deglobalisation.[3] In this paper, we show that this decline is driven by changes in only certain banks' balance sheets, and is thus not evidence of a broad-based deglobalisation trend.

Taken at face value, the BIS international banking statistics make a stronger case for peak finance in banking than global trade data make for peak trade (Graph 1). After falling relative to economic activity in the Great Financial Crisis (GFC) of 2007­09, global trade has declined only marginally in the latest observation. By contrast, the amounts of cross-border claims of BIS reporting banks in relation to global GDP rose from the early 1990s, peaked during the GFC, and fell until recent quarters. These data are compiled on a balance of payments basis (ie on a locational basis), according to the location of the creditor bank. They show a decline of cross-border bank assets from a peak near 60% of global GDP in 2007 to less than 40% in recent quarters. Lane and Milesi-Ferretti (2001, 2007, 2017) have used such stocks, along with other external assets, to measure international financial integration.[4]

1 Whether trade deglobalisation is cyclical or secular remains to be established. See Tomb and Trivedi (2017); Timmer et al (2016); Saravelos and Winkler (2016); Hoekman (ed) (2015); Constantinescu et al; Boz et al (2014); Shin (2016); BIS (2017).

2An early use of "financial deglobalisation" was by Broda et al (2009).

3Forbes (2014), Forbes et al (2016, 2017); but see Eichengreen (2016), Hooper et al (2016) and Verma.

4Their work has given rise to a very productive line of research that has taken on countries' net foreign asset positions, aggregate debt-equity ratios, and currency valuation effects (Gourinchas et al (2010), Lane and Shambaugh (2010a,b), Gourinchas and Rey (2014); and Benetrix et al (2015)).

 
 

Such "locational" measures of external assets suffer from two limitations: double­counting some positions and ignoring other relevant ones. As a result, they do not best serve to assess globalisation trends in banking. To be sure, giving priority to where the banking business is conducted (not by whom), as the BIS locational statistics do, can be useful in analysing macroeconomic aggregates, such as employment and value added.6 But cross-border claims double-count positions in which a bank's headquarters funds its branch in a financial centre like London, before lending abroad. At the same time, a bank's local positions booked in an affiliate abroad (eg a UK bank's loan to a French resident through the bank's Paris branch) are not captured in the external positions of either the bank's home country or the affiliate's host country. On a consolidated view, these are foreign positions - the bank has lent to a borrower outside the home country, even if it is booked and funded locally. And bank positions by location rather than by ownership cannot be related to the financial strength of particular banks or to the policies of the banks' home government or supervisors. As a result, this perspective has little to say on the drivers of (de)globalisation.

In this paper, we take multinational banks aggregated by the country of headquarters as the unit of analysis. By analogy to work on multinational firms, we "draw borders around groups of firms classified by nationality rather than around geographical entities" (Baldwin et al (1998), Avdjiev et al (2016)). We examine the foreign claims of 29 banking systems that report the BIS consolidated banking statistics (CBS) for evidence of deglobalisation.

We contribute to a growing body of work that puts the emphasis on banks' consolidated balance sheets rather than on locational (cross-border) bank positions. Much of this literature focuses on the effect of capital constraints on banks' foreign lending. In particular, McCauley and Yeaple (1994) and Peek and Rosengren (1997, 2000) analysed their effect on Japanese banks' lending abroad. McGuire and Tarashev (2008) analysed their effect on lending to emerging markets. Avdjiev et al (2012) analysed their effect on European banks' lending in central and eastern Europe. Forbes et al (2017) found that both higher required capital and monetary policy incentives for domestic lending led UK banks to cut loans abroad.

In recent papers close to ours, Cerutti and Claessens (2017) found that equity market indicators of bank vulnerability drove the banks' retrenchment after the crisis, though more through a contraction in cross-border than in local lending. In related work, Cerrutti and Zhou (2017) show that, in spite of the shrinkage of cross-border credit from core lenders, the network continued to proliferate linkages. They describe their findings as consistent with the speech that anticipated this paper (Caruana (2017b)), which regards financial deglobalisation as a regional (European) phenomenon. Bouvatier and Delatte (2015) use a traditional gravity model with a non-linear time trend to assess integration, and highlight the contraction of consolidated assets of banks headquartered in the euro area.

Our paper contributes to this body of work in three ways. First, in Section 2 we analyse step-by-step the differences in locational and consolidated measures of integration. European banks uniquely restored their capital ratios through asset shrinkage, and foreign claims bore the brunt of it, pointing to home bias. A retreat to the home market when a bank has suffered losses can reflect lower expected returns abroad or increased risk aversion (Giannetti and Laeven (2012)). But it can also reflect policy choices in the context of government support for banks (Rose and Wieladek (2014) and unconventional monetary policy that targets domestic lending (Forbes et al (2017)). Though consolidated European banks have retrenched, they have slashed assets disproportionately outside of their home offices.

Finally, in Section 5 we run a formal horserace between bank nationality and the effects of local economic conditions and policy measures in accounting for global bank retrenchment. This uses data that simultaneously capture bank nationality and bank location (eg German banks in Germany as separate from German banks in the United Kingdom). Nationality effects - again highlighting European banks - account for more of the decline in total foreign assets and in cross-border lending than do locational factors. Moreover, those global consolidated banking systems that suffered greater credit losses during and following the GFC tended to shrink their cross-border credit by more, in many locations around the world. European banks' problems thus reverberated globally.

2. The analytics and empirics of consolidation

Three reasons motivate building our discussion of deglobalisation on globally consolidated data. The first is that consolidated data include one dimension entirely absent from the residence-based (locational) perspective, namely banks' local positions in host countries around the world. Cross-border ownership is as much an expression of globalisation as cross-border positions. Second, consolidation removes intragroup positions, and thus avoids double-counting foreign claims that banks fund this way. And third, banking data organised by nationality of ownership provide a clearer picture of financial integration: the assignment of (arm's length) cross-border positions to bank nationalities leaves unchanged the global aggregate, but this step is essential for understanding how these positions respond to bank characteristics and the policies of the home government or supervisory authorities.

To illustrate these points, start from the locational perspective. A country's external bank assets include too much and too little from a nationality perspective. For the United States, when resident non-US banks' branches raise dollars from US money market funds and lend them abroad (Baba et al (2009)), then the corresponding US external asset (a claim entering the US international investment position) is not held by a US-owned bank. From a nationality perspective, this is too much - the non-US bank branch should be consolidated with its foreign parent. Similarly, when a US bank books an intragroup claim on its London branch, it does not have a claim on a foreign entity, although US external assets rise. This, too, is too much from a nationality perspective. However, when a US-owned bank affiliate in London accepts a deposit from a central bank (ie of foreign exchange reserves) and lends it to an emerging market firm, this is not recorded as a US external asset. From a nationality perspective, this is too little.

Clearly, consolidation has two distinct aspects. The first is to boil down a complex array of balance sheets into a single balance sheet that nets out intragroup balances but also adds in positions booked in offices abroad. The second is the assignment of this unified balance sheet to a particular home country. We use the location of a bank's headquarters. One can imagine other choices - for some purposes one might wish to drill down to the underlying equity holders, for instance.[5]

The reduction process nets out intragroup positions, the fluff of international finance, but adds in offshore positions against unaffiliated parties, the stuff of international finance. If a bank in A routes loans to B through C, perhaps an offshore financial centre, consolidation strikes A's claim on C and C's claim on B and reports just A's claim on B (Table 1). Likewise, with B's claims on A routed through C. From the perspective of risk exposure, the importance of consolidation is immediately apparent if one imagines something going wrong in B, leaving bankers and policymakers in A wondering what their exposure to the adverse event might be. Similarly, consolidation is useful to those in B, making clear that the ultimate lender and decision-maker is in A, not C.

5 Particular cases can challenge this assignment by headquarters — think of BCCI, incorporated in Luxembourg but headquartered in the United Kingdom (Goodhart (2011); Borio and Filosa (1994).

 

Consolidation of positions in offshore centres not only redistributes positions, but can also reduce them in aggregate. For example, the global Lane-Milesi-Ferretti (LMF) ratio - total external positions divided by global GDP - in Table 1 is (200+200+400)/(100+100+10) = 800/210, or almost 400%. But much of this is fluff and the consolidated aggregate openness is 400/210, or just under 200%.

Intragroup positions are big money. Of the $27 trillion in cross-border claims in the fourth quarter of 2016, about $9 trillion was intragroup claims (http://www.bis.org/statistics/a1_1.pdf). In other words, a third of external bank claims (ie those compiled on a balance of payments basis) are consolidated away.

If a bank uses an offshore centre to raise deposits and lend to unaffiliated parties, then consolidation in effect relocates that activity to the home country. Recall the case of the US bank affiliate in London accepting a deposit of foreign exchange reserves and lending it abroad. In this case, consolidated accounts would show a smaller UK balance sheet and a larger US balance sheet, with no effect on the aggregate LMF ratio.

Finally, consolidation brings into view purely local positions that are not captured at all in external assets. On a consolidated view, these are foreign positions - the bank headquartered abroad has liabilities to a host depositor, and credits to host country borrowers. The simplest possible example is a two-country world of like-sized (GDP = 100) economies with similar external positions (Table 2). External assets of 100 plus external liabilities of 100 put gross external positions at 200% of GDP for each country and the world (first memo item, the LMF ratio).

 

In addition, each country's banks have branches in the other country that attract local deposits and extend credit to the extent of 50. Thus, on a consolidated view, A's banks have 50 more in foreign assets and 50 more in foreign liabilities booked at their branches in B, and B's banks have 50 more in foreign liabilities and 50 more in foreign assets booked at their branches in A. A's banks have foreign assets plus foreign liabilities equal to 300; similarly, B's banks, 300. Consolidated openness is 300% of GDP for each country and the world.

Thus, in principle, consolidation has an ambivalent effect on measured international financial integration. If the multiplication of intragroup positions through offshore financial centres dominates, the locational measure could overstate international financial integration.If the effect of adding local positions dominates, then international financial integration is understated by external positions. In practice, these two effects of consolidation largely offset each other in aggregate, leaving the broad impression of deglobalisation in banking. The left-hand panel of Graph 2 shows the locational total from Graph 1 broken down into its constituent parts. Its juxtaposition to the consolidated data in the centre panel shows relatively subtle differences. The elimination of intergroup claims (the blue area in the left-hand panel), is roughly balanced by the addition of local claims (the green area in the centre panel). The subtle difference is that the consolidated claims only declined by about 20% of world GDP between 2007 and 2016, while locational claims declined by somewhat more than 20% of global GDP. Even without correcting for exchange rate valuation effects, it is evident that the local claims are more stable than cross-border positions (McGuire and von Peter (2016)).

 
 
It is worth emphasising that it is coincidence that the two effects of consolidation offset each other in aggregate. If there is a trend over time towards the multinational business model, with its reliance on locally funded local claims (McCauley et al (2012)), then the consolidated measure will indicate over time a higher level of international financial integration than will the locational measure. In other words, the ascendency of multinational banking over international banking means more risk-sharing in local positions that are not captured by balance-of-payments data.

Exclusion of such centres is not a satisfactory solution, particularly for countries that are a mix of offshore financial centre and real economy, like the United Kingdom.

Similarly, more jurisdictions reporting consolidated banking statistics to the BIS adds locally funded positions and raises the measure in the centre panel of Graph 2 relative to that in the left-hand panel.

In contrast to the globally aggregated positions (Graph 2), differences in the locational and consolidated bank stocks can be stark for individual countries, as Graph 3 shows. Differences in the outstanding stocks depend on both the overall banking activity in the country and on the structure of the global operations of the banks headquartered there. For example, at the far left, the United Kingdom's (GB) reading of -6.5% of combined GDP indicates the UK's locational bank aggregate is larger than the UK's consolidated aggregate. This reflects London's status as an international banking centre, host to some of the largest foreign affiliates of major banking systems. These foreign affiliates raise (mainly) cross-border deposit and wholesale funding which they in turn lend cross border to borrowers elsewhere. As a result, these foreign affiliates' cross-border positions in the UK's locational banking aggregate well exceed the positions of UK headquartered banks' affiliates abroad.

 

By contrast, the penultimate bar indicates that Swiss banks' consolidated claims exceed the locational claims in Switzerland by 2.4% of combined GDP. Neither domestic banks nor foreign bank affiliates in Switzerland have large cross-border claims positions. But, on a consolidated basis, Swiss banks have outsized global operations. McGuire and von Peter (2012) estimate that Swiss banks' foreign claims amounted to more than eight times Switzerland's GDP in 2007. Some 80% of these foreign claims were booked in offices located outside Switzerland.Japan (Graph 3, sixth bar from the left) is an example where the consolidated and locational bank measures are very close. Foreign banks' affiliates in Japan have mainly local claims on residents of Japan, with relatively small cross-border positions. For their part, Japanese banks' foreign claims are amongst the largest of all national banking systems; but they book the bulk of these at home.The differences across countries evident from Graph 3 make clear that a locational view for any particular country provides an unreliable guide to the evolution of any one banking system's balance sheet. Since internationally active banks have operations in multiple countries, banks headquartered in a given jurisdiction can respond to stresses by adjusting assets in locations abroad (see

3. The view by bank nationality

In this section we demonstrate that the apparent deglobalisation in banking is more regional that global. Graph 4 builds on the country breakdown in Graph 2 (right-hand panel), but with a more disaggregated view of the evolution of banks' foreign claims (unadjusted). The left-hand panel shows that the post-crisis contraction in European banks' foreign claims was shared: most European banking systems reported lower foreign claims in 2017 than in 2007, with Spanish banks being the notable exception.

 

Banks from the rest of the world, however, continued down the path of international financial integration (Graph 4, right-hand panel). Japanese, US, Canadian, Australian and emerging market banks expanded their foreign claims around the world, partly substituting for the retreat of European banks. In that sense, the trend towards deglobalisation is not broad-based.

Graph 5 contrasts the locational and consolidated views of deglobalisation. Here, banks are grouped into European (euro area, UK and Swiss) banks, US banks, other advanced banks and emerging market banks. The left-hand panel shows changes in trillions of US dollars and the right-hand panel percentage changes since 2007. The regional nature of the decline in bank claims is evident when the data are grouped by ownership (blue bars) rather than location (red bars). In particular, three observations stand out.

 

Banks' external claims by location and consolidated foreign claims by nationality therefore is not included in the consolidated data shown in Graphs 2-4 (Goh and Pradhan (2016)). Banks in China reported $778 billion in cross-border claims in June 2016. Moreover, Chinese banks' affiliates outside the mainland were estimated to have $1 trillion in assets offshore at the end of 2014 (McCauley and Ma (2015)). The latter in particular has grown rapidly (McGuire and van Rixtel (2012)). Chinese banks are very much integrating into global banking markets/

4. European bank deleveraging with home bias

We interpret the decline of European banks' foreign claims as a consequence of cyclical deleveraging of unsustainably risky bank balance sheets, rather than a secular deglobalisation trend. European banks uniquely shrank assets in order to strengthen their balance sheets in 2007-16. And, Spanish banks apart, the deleveraging of European banks displayed a home bias - which may reflect to some extent policy - further contributing to the shrinking of foreign claims on the rest of the world. The implication is that the proximate cause of the decline in European banks' foreign claims was a management decision on how to deleverage.

European banks adopted a unique deleveraging strategy: shrinking assets. Big banks headquartered in many countries raised the ratio of capital to risk-weighted assets after the 2007-09 crisis (Graph 6, left-hand panel, black dots). However, a decomposition of these increases (updating Cohen and Scatigna (2016)) reveals that European banks alone increased their capital strength in part through reducing total assets (a positive violet bar). By contrast, non-European big banks raised enough equity through retained earnings and equity issuance to strengthen their capital while expanding total assets. Put differently, European banks did not raise enough capital to achieve the 5-percentage point improvement in their weighted capital ratio without shedding assets.

However, the asset shedding by European banks was not a blind process, but one that protected the home market at the expense of foreign claims. Our evidence for this home bias is unfortunately limited to recent years. In the wake of the GFC, the countries reporting consolidated statistics to the BIS agreed to report the domestic claims of the same reporting banks, giving an advantage of consistency in reporting population. However, the reporting only began in the fourth quarter of 2013, yielding a late sample for testing for home bias. In Graph 7, right-hand panel, observations to the right of the 45 degree line can be interpreted as showing a trajectory of domestic and foreign claims that implies a home bias. UK and Swiss banks' foreign claims fell even as their domestic claims grew; this is a strong form of home bias. German banks' foreign claims contracted more than their domestic claims, while French and US banks' domestic claims grew faster than their foreign claims; all three cases illustrate a weaker form of home bias. By contrast, Spanish and Dutch banks' foreign claims expanded even as their domestic claims contracted, showing a foreign bias. Canadian, Italian and Japanese banks showed balance in their domestic and foreign growth strategies.

 

In sum, the financial deglobalisation that seems to characterise international banking looks different in a consolidated perspective. It seems more a regional phenomenon, limited to European banks. And it can be seen as reversing an earlier banking glut that afflicted Europe and that was based on very high leverage. In recent years, European banks shed foreign assets as part of their efforts to strengthen their capitalisation. Given their extensive overseas operations, this retrenchment was felt around the globe.

5. Deglobalisation and the nationality effect

In this section, we run a horserace between nationality effects and locational effects as drivers of the contraction in international banking in the wake of the crisis. This allows us to draw concrete inferences about how shedding of claims by particular consolidated banking systems affected the locational deglobalisation measures for host countries in which they operate. In the next subsection, we first describe a more granular version of the BIS international banking statistics that track how consolidated banks organise their global activity across host country locations. We then show the extent to which the contraction in European banks’ balance sheets shrank external bank assets outside their home countries. In the following sub-section, we use these granular data in panel regressions to quantify the extent to which bank nationality accounts for the contraction in foreign claims booked across all locations. We also relate our bank nationality effects to bank losses.

5.1 The interplay between bank nationality and location

A common image of international banking is that of a bank operating out of its headquarters at home, extending credit to domestic borrowers, and cross-border credit to borrowers abroad. On this view, problems in banks from a particular country would register primarily as a contraction in local credit in, and cross-border credit from, the home country. In that case, the locational view and the consolidated view would tell the same story about (de)globalisation. A more accurate picture of international banking is of multinational banking groups with offices in many countries, each of which extends credit to borrowers in the country in which it is located and cross-border credit to borrowers elsewhere (McCauley et al (2012)). Problems in banks of a particular nationality could result in coordinated contraction of cross-border credit from, and local credit in, many locations. The locational view would pick up the cross-border part in the many locations, but only the consolidated view would identify the common cause.

So far, we have documented that it was primarily (consolidated) European banks' foreign claims that contracted in the aftermath of the crisis. Yet, the consequences were felt around the world. Was this because European banks scaled back their activity in a single (or select few) locations in which they operate? This could arise, for example, if all European banks had large local positions in a single large country experiencing macroeconomic difficulty (eg the United States), or because all European banks have large holdings of cross-border exposures to a similar asset class booked in a single financial centre (eg the United Kingdom). In either case, the locational perspective may give an adequate picture of deglobalisation.

But if bank nationality played an independent role, in the sense that European banks scaled back activity in all (or many) locations in a coordinated way, then deglobalisation is neither broad-based nor a trend, necessarily. A coordinated change in claims for a particular bank nationality across all locations in which it operates - ie a bank nationality effect - points to balance sheet strains across the banks' entire consolidated structure, for example crisis-related losses and funding strains at the level of the parent bank.

The BIS international banking statistics have the granularity to show the assets of banks of a particular nationality in a particular location (eg German banks in the United Kingdom).16 For banks' home offices (eg German banks in Germany), we see banks' cross-border positions only. This disaggregation of a consolidated banking system's global positions reveals the extent to which that banking system sheds assets at different rates in the different host countries in which it operates.

We use these data in panel regressions where the dependent variable is the change in outstanding foreign claims booked by each bank-location pair. We include in each regression full sets of bank nationality and bank location dummies, as well as other controls, to test which set of dummies accounts for more of the overall cross­sectional variation. Note that, in a world where all banks have similar organisational structures (eg large offices in the United Kingdom) with similar portfolios, it would not be possible to identify separate location and nationality effects. In the real world, banking systems' structures and the countries in which they operate differ, according to the nature of the exposures on their balance sheets. This heterogeneity across locations and nationalities allows us to identify separately the importance of each effect.

As a prelude to our regression results, Graph 7 provides an indication of the extent to which the retrenchment of European banks affected external assets in many countries, not just in these banks' home countries. Cross-border claims of European banks' offices outside the home country shrank faster after the crisis than those booked in their home offices, driving up the share of their globally consolidated cross-border claims booked at home (Graph 7, left-hand panel). As we demonstrate below, the contractions in these banks' positions were clearly related to problems on these banks' global balance sheets (ie a nationality effect), but the effect of these problems was most pronounced in locations outside the home country.

 

And, as shown in the right-hand panel, understanding the changes in many countries’ external bank assets requires understanding the problems in European banks. Their foreign affiliates’ cross-border claims (blue line) contracted by roughly $5 trillion between end-2007 and end-2016. Banks of other nationalities increased their cross-border claims by about $2 trillion over the same period (gold line). What this reveals is that changes in these European banks’ foreign affiliates’ balance sheets had a larger effect on the external bank asset positions in other countries than did changes in the balance sheets of all the other banks located in these countries (including banks’ home offices). This makes it clear that balance sheet stresses that affect banks’ globally consolidated balance sheets have a significant impact on the locational measures of financial integration in not only the home country, but in all the other countries in which they have operations. We explore this issue more formally in the empirical analysis.

5.2 Identifying the nationality effect

The full granular data set captures quarterly positions, from 1999 onwards, of the home offices and foreign affiliates of 26 bank nationalities operating in more than 40 host countries. From an initial sample of 650 bank nationality-location pairs, we filter out observations with data shortcomings to create a sample of 280 affiliate groups spanning 18 parent countries and 38 different host jurisdictions, including major offshore centres. The combined peak assets among affiliates in the sample equals $36 trillion at end-2007, or 93% of total reported affiliate assets and total foreign claims too. The sample is unbalanced in the sense that not all bank nationalities have operations in all host countries.

We use a peak-to-trough approach to measure the contraction in the balance sheets of banks' foreign affiliates in the wake of the financial crisis. For each group of affiliates, the percentage contraction is calculated by comparing the peak value of total assets between Q1 2006 and Q1 2009 with the minimum asset value from Q2 2009 to Q1 2015. This approach takes into account that banks' assets peaked in different quarters, with most peaks occurring in Q4 2008 or Q1 2009; banks also shed assets at different rates and in different periods, eg due to the uneven effect across bank nationalities of the 2010-12 European sovereign debt crisis or other forces. The sample average of peak-to-trough contractions is 36% (weighted average), 39% (median) and 42% (average), respectively, with slightly larger contractions for the peak-to-trough decline in cross-border assets.

Our cross-sectional regression relates the rate of contraction to bank nationality, the location of operations, and to characteristics of each affiliate. The dependent variable, -AAin/Ain, is the percentage contraction of total assets (both cross-border and local) booked by bank affiliates of nationality n in host country i, indexed by in. The minus sign means that positive values are contractions (negative values are expansions). These growth rates are regressed on three groups of explanatory variables:

 
 

Location and nationality are captured through fixed effects at the country level: on captures factors common to banks of a specific nationality n, such as their group- wide business model or home country regulation. Likewise, is a set of dummy variables that absorb common factors affecting all affiliates operating in host country i, including local demand conditions and sovereign risk. The fixed effects and on are of interest in their own right, but in most empirical work they are merely included for consistency.21 We take location fixed effects as the baseline, and express nationality fixed effects as deviations that add or subtract from the contraction characteristic of any particular location.

We would like to obtain nationality and location effects that are robust to outliers, such as small offices or affiliates that relied on a vulnerable funding mix on the eve of the crisis. To this end, we include various controls (in square brackets in the equation) at the level of affiliates. They include balance sheet size of affiliates, measured as the logarithm of total assets (Ain), the (log of) geographical distance to the affiliates' home country (din), as well as pre-crisis ratios describing the structure of affiliates' assets and liabilities, Rrin.22 The coefficients a,5 and thus measure separately the extent to which a particular control variable accelerated (or mitigated if negative) the shedding of assets by the affiliates of nationality n in host country i. In related work, we focused on banks' funding mix and their reliance on non-core funding (McGuire and von Peter (2016)), and the results lent some support to the notion that these forms of funding were more fragile (Hahm et al (2013)).

Table 3 shows the results from various regression models based on equation (1). A consistent result is that there are systematic differences across nationalities and locations in the way bank affiliates contracted their positions since the crisis. A regression with no affiliate-specific variables (Model (1)) shows that the fixed effects jointly account for as much as 53% of the variance in balance sheet contractions. In this simple specification, the predicted contraction of bank affiliates in country i of nationality n simply equals the sum of the two respective fixed effects, + an.

 

The location (host country) effect explains 19% of variation. It estimates the decline in international banking activity in a particular country while taking out the influence of the nationality mix of the banks operating there. The magnitudes show that the average contraction specific to a given location was larger in troubled economies and some offshore centres serving as conduits (Graph 8). Affiliates in Greece, Spain, Korea and in several offshore centres (Panama, Cayman Islands, the Bahamas, Jersey and Guernsey) shed more than 40% of their assets. Claims booked by offices in the United Kingdom contracted at a rate close to the world average (42%). Many banks held structured finance products that lost value in the crisis in their affiliates located in London and in other financial centres.

For its part, the nationality effect alone accounts for 33% of variation in the credit contractions across affiliates. This captures the commonality among affiliates with the same parent country across all host countries. For example, Fortis and ABN AMRO were broken up during the crisis, and so the aggregate balance sheets of Belgian and Dutch banks' affiliates contracted in many host locations. The nationality effect implies that Austrian, Belgian and Dutch banks on average shed, respectively, 31% and 27% and 24% more of their assets than banks of other nationalities (after controlling for where their affiliates operate). Similarly, being an affiliate of a German bank added 18 percentage points to the average contraction in any given host country. By contrast, having a Japanese or Australian parent mitigated the contraction characteristic of a particular location by some 20 percentage points since these banking groups emerged from the crisis largely unscathed.

Note that there is no contradiction in finding two opposing fixed effects for the same country: they relate to distinct concepts. Spain, as an economy in crisis, saw a strong average contraction among affiliates located there, be they Spanish or foreign banks (Graph 8, left-hand panel). But Spanish banks, with a formidable presence across Latin America, saw an offsetting nationality effect pointing to continued expansion in spite of the difficulties at home (Graph 8, right-hand panel).

A weighted regression gives qualitatively similar results (Table 3, Model (2)). Giving larger affiliates more weight proportional to their total assets leads to an outsize regression fit, as the fixed effects are now determined to closely match the contractions at larger banking systems and host countries, respectively. Although the estimated fixed effects differ, the ranking of locations and nationalities remains similar. Again, the shaded rows in Table 3 underscore the significance of fixed effects and how much they contribute to explaining the contraction of bank affiliates.

The two sets of fixed effects absorb all factors common to host countries or to bank nationalities. This includes the extent to which funding structures were systematically more fragile among bank affiliates of the same nationality, or among bank offices in certain locations (such as financial centres). That said, to prevent outliers from influencing the fixed effects, we add basic controls at the affiliate level, measured at the pre-crisis peak. Many statistical commands calculate fixed effects before running the regression on demeaned variables; doing so delivers more significant coefficients on the affiliate-level regressors than the results shown here. In Table 3, we estimated both sets of fixed effects along with their standard errors, which allows for more sample variability (see McCaffrey et al (2012) for a review of methods).

Separating location and nationality factors in global bank retrenchment minimum as a share of their balance sheet. LIM explains 3% of the variance in balance sheet contractions across affiliates, given the fixed effects. The estimated coefficient (-0.3) suggests that banks doing 10% more business locally shrank by 3 percentage points less on average. The pairing of local assets with local funding apparently provided resilience, presumably because it is indicative of retail business.

To close the circle, the final column in Table 3 runs the peak-to-trough analysis on affiliates' cross-border assets with which our discussion of deglobalisation began. Results mirror those for total assets (Model (3)), with slightly lower explanatory power and sample size. As before, there are systematic patterns in the way banks shed assets depending on their nationality and location. Locational factors, however, are less significant now, presumably because affiliates' lending goes exclusively to borrowers in other countries. Even so, the contraction of bank affiliates moves the external positions of countries they operate in - going well beyond their home countries. The nationality effect remains potent in explaining credit contractions, pointing to a strong commonality among banks from the same home country, regardless of where they operate. Clearly, European banks shrank their operations in many locations, responding to financial distress at the group level.

Returning to the theme of deleveraging, we explore how the nationality effect relates to the credit losses banks suffered during the crisis. Credit losses in the financial crisis are known to have decimated the capital base of many internationally active banks. That this in turn led banks to cut credit can be gleaned from Graph 9: banking systems recording greater credit losses tended to shrink their balance sheets by more - and do so in many locations around the world.

These results underscore the importance of bank health at the group level. Banks with larger credit losses spread credit contractions across many host countries. That this effect plays an outsize role for European banks provides a perspective on the decomposition above of the change in bank capitalisation by nationality of bank. European banks' credit losses in 2007-16 left them hard choices to strengthen their capital. The "lessons from the Nordics" of the benefits of prompt resolution of bank credit problems were not fully taken on board in Europe (Borio et al (2010)). Unlike in the United States, there was generally no broad programme of government equity purchase, as under the US TARP, nor checks on dividends and other distributions (as under the stress tests). Shedding of foreign claims allowed management not to issue new equity when market prices languished below book values.

 

6. Conclusions

In sum, the financial deglobalisation that appears to characterise global banking looks different from a consolidated perspective. It is better interpreted as a regional phenomenon that reflects an earlier banking glut that afflicted Europe. European banks responded to losses by shedding assets around the world.

If it is accepted that the decline of cross-border and foreign banking claims is a symptom of a particular form of European bank deleveraging, what are the policy implications? One implication is that the authorities are well advised to insist on the strengthening of banks' balance sheets, rather than let losses fester.

Another is that there is a risk of mistaking bank deleveraging for a deglobalisation trend in banking. Doing so could influence the response to policies that tend to further banking fragmentation. If observers judge that the global banking market is already headed in the direction of deglobalisation, they may underestimate the cost of such policies. In other words, a misapprehension regarding the source of the contraction in global banking as secular rather than cyclical could actually prove self-fulfilling.

 

References

Avdjiev, S, Z Kuti and E Takats (2012): "The euro area crisis and cross-border bank lending to emerging markets", BIS Quarterly Review, December, pp 37-47.

Avdjiev, S, R McCauley and H S Shin (2016): "Breaking free of the triple coincidence in international finance," Economic Policy, vol 31, issue 87, July, pp 409-51.

Baba, N, R McCauley and S Ramaswamy (2009): "US dollar money market funds and non-US banks", BIS Quarterly Review, March, pp 65-81.

Baldwin, R, R Lipsey and D Richardson (1998): "Introduction", in R Baldwin, R Lipsey and D Richardson (eds), Geography and ownership as bases for economic accounting, University of Chicago Press for the National Bureau of Economic Research, pp 1-7.

Bank for International Settlements (2017): 87th Annual Report, Chapter VI, June.

Benetrix, A, P Lane and J Shambaugh (2015): "International currency exposures, valuation effects and the global financial crisis," Journal of International Economics, vol 96 (January), pp 98-109.

Borio, C (2013): "The Great Financial Crisis: setting priorities for new statistics", BIS Working Papers, no 408, April.

Borio, C and R Filosa (1994): the changing borders of international banking, BIS Economic Papers no 43, December.

Borio, C, R McCauley, P McGuire and V Sushko (2016): "Covered interest parity lost: understanding the cross-currency basis", BIS Quarterly Review, September, pp 45-64.

Borio, C, B Vale and G von Peter (2010): "Resolving the financial crisis: are we heeding the lessons from the Nordics?", BIS Working Papers, no 311, June.

Bouvatier, V and A-L Delatte (2015): "Waves of international banking integration: a tale of regional differences", European Economic Review, vol 80, pp 354-373.

Boz, E, M Bussiere and C Marsilli (2014): "Recent slowdown in global trade: cyclical or structural", VoxEU, 12 November.

Broda, C, P Ghezzi and E Levy Yeyati (2009): "The new global balance: financial de­globalisation, savings drain, and the US dollar", VoxEU, 22 May.

Caruana, J (2017a): "International financial crises: new understandings, new data", speech at the National Bank of Belgium, Brussels, 6 February. (2017b): "Have we passed 'peak finance'?", lecture to the International Center for Monetary and Banking Studies, Geneva, 28 February.

Cerutti, E and S Claessens (2017): "The great cross-border bank deleveraging: supply constraints and intra-group frictions", Review of Finance, pp 201-236.

Cerutti, E and H Zhou (2017): "The global banking network in the aftermath of the crisis: is there evidence of de-globalization", IMF Working Paper, forthcoming.

Cetorelli, N and L Goldberg (2011): "Global banks and international shock transmission: evidence from the crisis", IMF Economic Review, vol 59(1), pp 41-46.

Cohen, B and M Scatigna (2016): "Banks and capital requirements: channels of adjustment", Journal of Banking and Finance, vol 69, issue s1, pp S56-S69.

Constantinescu, C, A Mattoo and M Ruta (2015): "Explaining the global trade slowdown", VoxEU, 18 January.

Eichengreen, B (2016): "Globalization's last gasp", Project Syndicate, 17 November.

Fender, I and P McGuire (2010): "Bank structure, funding risk and the transmission of shocks across countries: concepts and measurement", BIS Quarterly Review, September, pp 63-79.

Forbes, K (2014): "Financial 'deglobalization'? Capital flows, banks and the Beatles", speech to Queen Mary University, London, 18 November.

Forbes, K, D Reinhardt and T Wieladek (2016): "How the interactions of monetary and regulatory policies may have been ahead of the anti-globalisation backlash", VoxEU, 23 December.  (2017): "The spillovers, interactions, and (un)intended consequences of monetary and regulatory policies", Journal of Monetary Economics, vol 85, January, pp 1-22.

Galstyan, V and P Lane (2013): "Bilateral portfolio dynamics during the Global Financial Crisis", European Economic Review, vol 57 (C), pp 63-74.

Giannetti, M and L Laeven (2012): "The flight home effect: evidence from the syndicated loan market during financial crises", Journal of Financial Economics, vol 104, no 1, pp 23-43.

Goh, K and S-K Pradhan (2016): China and Russia join the BIS locational statistics", BIS Quarterly Review, December, pp 25-26.

Goodhart, C (2011): The Basel Committee on Banking Supervision, Cambridge: Cambridge University Press.

Gourinchas, P-O and H Rey (2014): "External adjustment, global imbalances, valuation effects", in G Gopinath, E Helpman and K Rogoff (eds), Handbook of International Economics, vol 4, North Holland, pp 585-646.

Gourinchas, P-O, H Rey and N Govillot (2010): "Exorbitant privilege and exorbitant duty", Bank of Japan, IMES Discussion Paper, no 2010-E-20.

Hahm, J-H, H S Shin and K Shin (2013): "Non-core bank liabilities and financial vulnerability", Journal of Money, Credit and Banking, vol 45, issue S1.

Hoekman, B (ed) (2015): The global trade slowdown: A new normal?, CEPR Press.

Hooper, P, M Spencer, T Slok and M Luzzetti (2016): "Global financial integration and disintegration", Deutsche Bank Economics Global Economic Perspectives, 29 November.

Kreicher, L, R McCauley and P McGuire (2014): "The 2011 FDIC assessment on bank managed liabilities: interest-rate and balance-sheet responses", in R de Mooij and G Nicodeme (eds), Taxation of the financial sector, Cambridge: MIT Press.

Lane, P and J M Milesi-Ferretti (2001): "The external wealth of nations: measures of foreign assets and liabilities for industrial and developing countries", Journal of international Economics, vol 55, pp 263-294.  (2007): "The external wealth of nations mark II: Revised and extended estimates of foreign assets and liabilities, 1970-2004", Journal of International Economics vol 73, November, pp 223-250.  (2017): "International financial integration in the wake of the global financial crisis", IMF Working Papers, no WP/17/115, May.

Lane, P and J Shambaugh (2010a): "Financial exchange rates and international currency exposures", American Economic Review, vol 100, no 1. (2010b): "The long or short of it: determinants of foreign currency exposure in external balance sheets", Journal of International Economics, vol 80, issue 1, January, pp 33-44.

McCaffrey, D, J Lockwood and K Mihaly (2012): "A review of Stata commands for fixed- effects estimation in normal linear models", Stata Journal, vol 12, issue 3, pp 406-432.

McCauley, R and G Ma (2015): "Transforming central bank liabilities into government debt: the case of China", China & World Economy, vol 23, issue 4, July-August, pp 1-18.

McCauley, R, and P McGuire (2014): "Non-US banks' claims on the Federal Reserve", BIS Quarterly Review, March, pp 89-97.

McCauley, R, P McGuire and G von Peter (2012): "After the global financial crisis: from international to multinational banking?", Journal of Economics and Business, no 64, pp 7-23.

McCauley, R and S Yeaple (1994): "How lower Japanese asset prices affect Pacific financial markets", Federal Reserve Bank of New York Quarterly Review, vol 19, no 1, spring, pp 15-33.

McGuire, P and N Tarashev (2008): "Global monitoring with the BIS international banking statistics", BIS Working Papers, no 244, February.

McGuire, P and G von Peter (2009): "The US dollar shortage in global banking and the international policy response", BIS Working Papers, no 291, October. (2012): "The US dollar shortage in global banking and the international policy response", International Finance, vol 15, issue 2, Summer, pp 155-178.  (2016): "The resilience of banks' international operations", BIS Quarterly Review, March, pp 17-18.

McGuire, P and A van Rixtel (2012): "Shifting credit patterns in emerging Asia", BIS Quarterly Review, December, pp 65-78.

Okawa, Y and E van Wincoop (2012): "Gravity in international finance", Journal of International Economics, no 87, pp 205-15.

Peek, J and E Rosengren (1997): "The international transmission of financial shocks: the case of Japan", American Economic Review, vol 87, no 4, pp 485-505. (2000): "Collateral damage: effects of the Japanese bank crisis on real activity in the United States", American Economic Review, vol 90, no 1, March, pp 30-45.

Rose, A, and T Wieladek (2014): "Financial protectionism? First evidence" Journal of Finance, vol 69, no 5, pp 2127-2149.

Saravelos, G and R Winkler (2016): "Deglobalization is here to stay: what it means for global macro", Deutsche Bank Special Report, 16 November.

Shin, H S (2012): "Global banking glut and loan risk premium", IMF Economic Review, vol 60, issue 2, pp 155-192. (2016): "The bank/capital markets nexus goes global", speech at the London

School of Economics and Political Science, 15 November.

Timmer, M, B Los, R Stehrer and G de Vries (2016): "An anatomy of the global trade slowdown based on the WIOD 2016 release", University of Groningen, Groningen Growth and Development Centre Research Memorandum, no 162, December.

Tissot, B (2016): "Globalisation and financial stability risks: is the residency-based approach of the national accounts old-fashioned?", BIS Working Papers, no 587, October.

Tomb, I and K Trivedi (2017): "'Peak trade' is premature", VoxEU, 6 January.

Verma, S (2016): "A cheat sheet on the deglobalization of the financial world", Bloomberg, 6 December.