Appetite for sovereign debt
The link between bank capitalization and government bond holdings in the euro area
Master Thesis
Author: Toep van Dijk Student number: 10674985
1 Statement of Originality
This document is written by Student Toep van Dijk who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2 Abstract
The current regulatory environment of the EU does not force banks to hold any capital buffers against government debt exposures. As a result, banks within the euro area are very vulnerable to sovereign credit risk. From 2007 onwards euro area banks increased their domestic sovereign exposures drastically. Because of the inherent systemic nature of government debt – and the absence of counterbalancing capital when risks materializes – the amount of government debt that is currently held by euro area banks could have severe implications for financial stability when risk premiums on sovereign debt increase. I show in this paper a negative relationship between the level of bank capitalization and their domestic government bond holdings. The current regulatory treatment of sovereign debt incentivizes low capitalized banks to increase their government bond holdings, especially in the wake of increasing capital requirements due to the more stringent Basel III framework. Banks can meet these more binding capital requirements by shifting their portfolio from assets with a non-zero risk weight (i.a. loans to the private sector) to zero-risk-weighted assets (i.a. loans to governments). The crowding out of allocation of credit could subdue investment and consequentially hamper economic growth.
3 Table of Contents
Introduction 4
1. Context – policy measures and the sovereign-bank nexus 6
2. Patterns in sovereign exposure of the euro area banking sector 10
3. Literature 14
4. Data and methodology 19
5. Results 23
6. Policy Recommendations 27
Conclusion 30
Reference list 31
4 Introduction
Negative feedback loops between banks and sovereigns are crucial in order to understand the dynamics behind the recent European debt crisis. Bank bailouts across the euro area drastically worsened government finances. Vice versa, sovereign risk spread rapidly through the banking system, deteriorating balance sheets of several banks across the euro area. The European debt crisis drastically revealed that a ‘deadly embrace’ of banks and sovereigns could give birth to a downward spiral of shocks.
The newly formed euro area banking union, consisting of a Single Rulebook, a Single Supervisory Mechanism (SSM) and a Single Resolution Mechanism (SRM), is supposed to limit the interconnectedness between banks and member states. The SRM (more specifically: bail-in instead of bailout and, in extremis, use of the resolution fund) indeed reduces the likelihood that
sovereigns are ‘dragged into insolvency trough tax-funded bank bail outs’ (Buiter, Rahbari & Montilla, 2014). However, the sovereign-to-bank contagion channel is still in place. As a result, banks within the euro area are still very vulnerable to sovereign credit risk.
According to the Basel II and III accords sovereign risk should be counterbalanced by an amount of capital that corresponds to the underlying credit risk. However, following the less strict EU Capital Requirement Directive (CRD) interpretation of the Basel accords, EU banks apply a zero risk weight to all EU-currency denominated government bonds. Consequently, EU banks are not obliged to hold any capital against their EU sovereign exposures.
From mid-2007 onwards euro area banks increased their sovereign exposures with more than 45%, (ECB Statistical Data Warehouse 2015).1 Because of the inherent systemic nature of
government debt – and the absence of counterbalancing capital when risks materializes – the amount of government debt that is currently held by euro area banks could potentially again trigger a downward spiral of shocks. For this reason, both policy makers and academics increasingly question the rationale behind the current CRD prudential regulatory treatment of sovereign debt.
In this paper I show a negative relationship between banks’ capital position and the amount of domestic sovereign debt holdings in the euro area between 2007 and 2015. In other words, between 2007 and 2015 banks with relatively low capital levels tended to invest more in
government securities. The zero risk weight of sovereign exposures incentives financial institutions to reduce their allocation of credit to the real economy (that is financing of households or non-financial corporations) in favor of lending to governments. The surge in government bond holdings seems to be enhanced by the fact that banks are subject to increasingly binding capital
requirements, as a result of the Basel III accord. In order to meet these new capital standards, banks are implicitly encouraged to substitute assets with a high risk weight into assets with a low risk weight, such as government securities.
This paper is structured as follows. The first section gives some background about the European sovereign debt crisis, the bank-sovereign nexus and the current policy stance. The second section
1 The ECB defines MFIs as credit and financial institutions whose business is to receive deposits (or close
substitutes) from entities other than MFIs and – for their own account – grant credits and/or make investments in securities. In the rest of this paper banks and MFIs are used interchangeably.
5
provides an overview of the patterns and tendencies in sovereign exposures of euro area banks. The third section gives an overview of the existing literature about the link between banks and sovereigns in general and – more specifically – about the determinants of euro area banks’ sovereign bond holdings. The fourth section provides a brief outline of the data and methodology that are used and the fourth section summarizes the results. Finally, the fifth section lists some policy considerations.
6
1. Context – policy response and the sovereign-bank nexus
The financial positions of banks and their sovereigns are often closely intertwined. The situation in which weak banks and sovereigns push each other into a bad equilibrium is known as the vicious circle (also called ‘deadly embrace’ or ‘doom loop’) between banks and sovereigns. Although the European sovereign debt crisis was ultimately caused by excessive macroeconomic and financial imbalances, the sovereign-bank nexus is crucial in order to understand the dynamics during the crisis. The financial meltdown forced several sovereigns across the continent to bailout systemic banks in order to avoid a cascade of bank defaults and to minimize costs for the real economy. In the absence of discretionary revenue measures, the bailout costs ran into budget deficits, causing a rise in government debt in almost all euro area member states.
From 2010 onwards, financial markets started to penalize weak macroeconomic
fundamentals of periphery euro area member states (i.e. Cyprus, Greece, Ireland, Italy, Spain and Portugal). In contrast to the early years of the euro, investors now took the creditworthiness of the sovereign, including the level of government debt, into account when formulating risk premiums on sovereign debt. The surge in bond yields of periphery countries, in turn, weakened balance sheets of banks with substantial exposures to periphery governments. Several banks faced a drop in the market value of their assets. As a consequence, the funding costs of these banks rose: their perceived riskiness increased and the value of government bonds they used as collateral for secured funding declined. As the balance sheets of these banks weakened more and more, governments were once more urged to intervene and recapitalize these banks. As a result, government finances deteriorated further.
In response to the euro area sovereign debt crisis, euro area policy makers implemented several measures to specifically break the link between sovereigns and banks and to strengthen the financial resilience of the monetary union in general. The main institutional changes that relate to the sovereign-bank nexus are the bank recapitalization programs of the European Stability Mechanism (ESM) and the European banking union (for more information, see box 1). Box 1 – Institutional changes and the link between banks and sovereigns
European Stability Mechanism (ESM)
The ESM is a permanent intergovernmental crisis resolution mechanism for euro area countries. In September 2013 the ESM took over the tasks of the temporary European Financial Stability Fund (EFSF) and the European Financial Stabilization Mechanism (EFSM). The ESM has a maximum lending capacity of € 500 billion, paid by
contributions of the Member States according to their size. Because sovereign risk is often related to problems in the banking sector, the ESM has two specific bank recapitalization programs: an indirect and a direct recapitalization scheme.
Initially, the ESM was only allowed to recapitalize banks indirectly. Under the indirect recapitalization program the ESM would first lend the money to the sovereign and the sovereign would then transfer the money to the ailing financial institution. This implied that the vicious link between sovereigns and failing banks still existed, because a bank bailout by the ESM would still impact the sovereign’s debt ratio and credit
7 rating.
In order to break the vicious link between member states and troubled banks, the member states decided to give the ESM the possibility to recapitalize banks directly. The direct recapitalization framework was established in December 2014. Within this framework the ESM directly transfers capital to a troubled bank. As a consequence, the bank bailout would not be reflected in the debt ratio of the concerned sovereign.
However, it is questionable to what extend the direct recapitalization framework really breaks the bank-to-sovereign contagion channel. After all, because the ESM is funded by contributions of all participating member states, it is still the government that de facto bears the costs of bank failures. Instead of really breaking the vicious link, the direct recapitalization instrument rather mutualizes the bank-to-sovereign
contagion channel by shifting the link between banks and sovereigns from the national to the (pooled) European level.
The banking union
At the euro area summit in June 2012, the heads of state decided that supranational supervision on euro area banks would be be a precondition for introduction of the ESM instrument for direct bank recapitalization. This declaration of intent led the first brick for the establishment of the euro area banking union. The banking union is made out of three building blocks: a single rulebook, the Single Supervisory Mechanism (SSM) and the Single Resolution Mechanism (SRM). The single rulebook is the backbone of the banking union and consists of legislative texts that all banks in the EU have to comply with in order to create an equal level playing field. The main legislative texts of the single rulebook are the Bank Recovery and Resolution Directive (BRRD), the Capital Requirements Directive (CRD) and Capital Requirements Regulation (CRR) and the Deposit Guarantee Scheme Directive (DGSD).
The SSM, led by the ECB, is responsible for bank supervision and monitors the compliance of participating banks with the rules that are set in the single rulebook. As a result, all banks are subject to the same supervisory treatment.
The SRM is a framework for recovery and resolution of failing banks and consists of national resolution authorities (in most member states the national central bank), a European resolution authority (the so-called Single Resolution Board (SRB)) and a Single Resolution Fund (SRF). The SRB is the central-decision making body of the SRM and must ensure that banks that face difficulties are resolved efficiently. In order to minimize the costs for the taxpayer, the SRB applies the framework of the BRRD.
The BRRD requires that failing banks will be subject to the principle of bail-in. In a bail-in, bank’s creditors’ are forced to bear some of the burden of the costs of a bank failure. According to the BRRD, first a minimum amount of 8% of all liabilities must be bailed in (along the lines of a strict predefined hierarchy), before the SRF can be used. The SRF is funded by contributions of the banking sector and will only be used when other resolution options (including the bail-in tool) are exhausted. The
8
The newly introduced bail-in mechanism, as set out in the BRRD, indeed mitigates the transmission of shocks from sovereigns to banks. Within the new resolution framework, bank failures do not necessarily lead to an increase in government debt, as creditors of the bank, and ultimately the banking sector itself, bear the costs of a bank’s recapitalization. First, a substantial amount of all liabilities must be bailed in – and thereafter the SRF will be used – before the respective
government will be invoked to support an ailing bank.
But the bank-to-sovereign transmission channel is only one part of the ‘doom loop’. The other part of the nexus – the sovereign-to-bank transmission channel – is still in place. So far, policy makers in the euro area have not implemented measures that reduce the risks that are associated with large sovereign debt exposures of banks. In fact, the current regulatory
environment of the EU does not force banks to hold any capital buffers against government debt exposures.
The prudential treatment of sovereign debt on bank balance sheets in the EU follows from the ‘Capital Requirement Regulation’ (CRR). The CRR – in turn – is based on the framework set out by the Basel Committee of Banking Supervision (BCBS), located at the Bank of International Settlements (BIS). Yet, certain aspects of the transposition of the Basel framework in the CRR differ from the actual doctrine of Basel. Both ‘Basel II and III call for minimum capital requirements that commensurate with the underlying credit risk’. Jurisdictions may adopt either the
‘standardized approach’ (based on standardized credit ratings) or the ‘Internal Ratings-Based (IRB) approach’, or both. The IBR approach applies especially to large and internationally operating banks. Under the IRB Approach risk weights are determined according to a banks’ own assessment of the probability of default (PD) and loss-given default (LGD). For the 201 largest banks the average PD for sovereign exposures was around 0.1%, which is commensurate with a risk weight of 30% (Bank of International Settlements (BIS), 2013).
However, within the Basel framework, sovereigns are also allowed to ‘apply a lower risk weight to banks’ exposures to their sovereign of incorporation denominated in domestic currency and funded in that currency’ (BIS, 2006). As a result, risk weights of sovereign exposures vary across jurisdictions. In the EU, the CRR allows banks that generally follow the IRB Approach to stay permanently under the Standardized Approach for their sovereign exposures. Moreover, in applying the Standardized Approach, EU authorities ‘have set a zero risk weight not just to sovereign
exposures denominated and funded in the currency of the member state, but also to sovereign exposures denominated and funded in the currency of any other member state’ (BIS, 2013). As a consequence, the amount of risk-weighted assets is substantially lower for EU banks than for banks outside the EU.
In addition, the EU legislative framework does not set concentration limits for sovereign exposure either. Generally, the CRR limits the amount of exposures of banks to a single
counterpart to 25% of their adjusted own funds (that is: the sum of core tier 1 and core tier 2 capital) (European Commission, 2015). The exposure limit prevents banks from building up large exposures to a single counterparty and forces banks to reduce risk through diversification. However, the CRR exempts sovereign exposures from this basic rule when sovereign debt is zero risk weighted according to the Basel’s standardized approach (Ibid.). Because this applies to all sovereign debt denominated in EU currency, the exposure of EU banks to EU sovereign debt is by no means legally capped.
9
The current regulatory treatment of sovereign debt in the EU is based on the notion that sovereign debt is virtually risk free. Evidently, the fact that several sovereign defaults occurred throughout modern history (i.a. Argentina, Russia, Thailand and Mexico) shows the opposite. Moreover, the recent European sovereign debt crisis underlined again that even for advanced countries
government debt is not entirely risk free. The fact that risk premiums on government debt vary from country to country underscores that there is in fact a perceived risk of default.
In times of economic distress, sovereign bonds on bank balance sheets function as a transmission channel trough which government finances cause systemic risk (Ibid.). In the absence of capital buffers, an increase in bond yields could have severe implications for financial stability. Regarding the uneven distribution of sovereign exposures across euro area member states (for data, see next section), the current regulatory framework could potentially undermine the solidarity mechanism on which the banking union relies (such as use of the common Single Resolution Fund).
Moreover, the current regulatory framework provides incentives for banks to prefer lending to European sovereigns over lending to the private sector. This is enhanced by the fact that capital requirements are increasingly stringent in the wake of the implementation of new banking
regulations. Between 2010 and 2011 the members of the Basel Committee of Banking Supervision formulated new capital requirements for banks. The worldwide financial crisis underscored the urge for higher mandatory capital buffers of banks and de facto revision of the Basel II rulebook. In November 2010 the G20 countries agreed upon adopting the new standards that were formulated by the Basel Committee. The newly formed Basel III regulatory framework calls for a minimum tier 1 capital buffer of 6% (as a percentage of risk weighted assets), compared to a 4% tier 1 buffer under Basel II. Started in 2014, the new rules are gradually implemented until the Basel III regulatory framework is fully in effect in 2019. Despite new capital standards will not be fully implemented until 2019, most banks already started the adjustment process from the moment of the announcement in 2010.
There are several channels through which banks can improve their capital position and adjust to these higher regulatory standards: (1) by accumulating retained earnings, (2) by issuing new equity, (3) by implementing changes in the asset side of their balance sheet (e.g. sell assets and pay down debt) or (4) by reducing the amount of risk weighted assets by replacing high risk weighted assets by low risk weighted assets (Cohen & Scatigna, 2013). The last channel is particularly important when assessing the effect of the current regulatory stance regarding sovereign debt on the real economy. In contrast to government bonds, retail loans and loans to corporations (i.e. ‘the real economy’) have a non-zero risk weight.2 As a result, banks with
relatively low capital levels are implicitly stimulated to replace these loans by government debt in order to meet the more stringent capital requirements. The subsequent crowding out of allocation of credit to the real economy, in turn, leads (ceteris paribus) to lower investment and lower economic growth.
2 The risk weight of assets under the IBR approach depends on a bank’s own assessment based on the
perceived probability of default (also taken into account historical defaults within their portfolio and possible correlations). The standardized approach risk for instance applies the following risk weights on different asset classes: 35% risk weight for exposures secured by residential real estate, 100% risk weight for exposures secured by commercial real estate, 20% risk weight for AAA rated corporate bonds, 100% risk weight for unrated corporate bonds and 75% risk weight for retail exposures (BIS, 2015).
10
2. Patterns in sovereign exposures of the euro area banking sector3
Euro area banks hold a large share of their assets in domestic government securities. Data shows that domestic bond holdings by euro area banks were already markedly higher at the start of the 2007 financial meltdown than in both the UK and the US (Merler & Pisani-Ferry, 2012). At the beginning of 2007, before the financial meltdown, total government securities holdings of euro area banks accounted for almost 5% of total bank assets, whereas in the UK, for instance, domestic government securities holdings of the financial sector accounted only for 2% of total assets (ECB SDW, 2015). From 2007 onwards banks in the euro continued to increase their exposure to sovereign debt. Moreover, as euro area sovereign bond yield spreads widened between 2010 and mid-2012, several banks continued to increase their exposure to financially distressed
governments (EBA, 2010-2014).
Figure 1 gives an overview of the evolution of domestic government bond holdings of euro area banking sectors as a share of total assets for core (i.e. Austria, Belgium, Finland, France, Germany and the Netherlands) and periphery countries (i.e. Cyprus, Greece, Ireland, Italy, Portugal and Spain) between 2007 and 2015.
Figure 1 – Financial institutions’ holdings of domestic government securities (% of total assets)
A. Core countries
B. Periphery countries
The figure highlights that in almost all euro area countries, banks increased their sovereign exposures between 2007 and 2015. In addition, the figure also shows the huge differences in domestic sovereign debt holdings across euro area member states. Banks in Italy, Portugal and
3 In this section I used both county and bank aggregated data. The country aggregated data is
subtracted from the Statistical Data Warehouse (SDW) of the ECB and publicly available at
www.sdw.ecb.europa.eu (see the data and methodology section and/or the appendix for more details). The bank aggregated data is extracted from seven subsequent data collecting moments of the European Banking Authority between 2010 and 2013 (which includes the stress-test, transparency exercise and capital exercise). The data is also publicly available at http://www.eba.europa.eu/risk-analysis-and-data, but is highly
11
Spain have huge domestic sovereign exposures, while exposure of banks in Finland, Ireland and the Netherlands are relatively modest. In Italy, domestic sovereign exposures account for more than 10% of total bank assets. Given that the average capital holdings of Italian banks are also only slightly above the 10% of total (non-risk weighted) assets, this evidently could have implications for financial stability when volatility on financial markets causes risk premiums to increase (ECB SDW 2015).
Looking at bank specific data of the European Banking Authority (EBA), it becomes apparent that the amount of government bond holdings of varies greatly from bank to bank. It appears to be the case that, although all banks hold a substantial amount of their assets in sovereign debt, middle-sized banks are most exposed to sovereign debt. (European Banking Authority, 2010-2015).4 The dynamics that explain the patterns in domestic government bond
holdings of banks between 2007 and 2015 also varies from country for country. The surge in government debt holdings of Irish banks, for instance, mainly reflects the reduction of the size of the Irish banking sector.5 The large drop in government debt holdings of Greek banks, on the other
hand, seems to reflect two respective bailouts agreements of May 2010 and February 2012. The second bailout also included a haircut of more than 50% for private holders of Greek government debt. This led to a substantial drop in the total value of Greek government bonds holdings by Greek banks. The same dynamics might also explain the reduction of exposure to domestic government debt of Cypriot banks.
Regarding figure 1, it is important to note that the increase of total assets between 2007 and 2014 causes the patterns in figure 1 to understate the actual increase of domestic government bond holdings of euro area banks. The increase of domestic government debt exposure was on average much larger for periphery banks than for core banks, but also Austrian, Dutch and Finnish banks increased their sovereign exposures substantially. In absolute terms, the amount of
government bond holdings of banking sectors in Austria, Italy, Finland, Ireland, the Netherlands, Portugal and Spain more than doubled between 2007 and 2015. Portuguese banks show the largest increase in government bond holdings: between 2007 and 2014 their government exposures more than quintupled (ECB SDW, 2015).
Another pattern that arises from the data is the large degree of home bias of the euro area banking sector (see figure 2). Home bias can be defined as a bank’s domestic government securities
holdings divided by its total sovereign exposure holdings.6
4 Across time the smallest quartile of banks hold on average 0.072% of their assets in euro area sovereign
bonds. For the second and third smallest quartile this is on average 0.104% and 0.082% respectively. The largest quartile of banks hold on average 0.048% of their assets in euro area sovereign debt. Note that the EBA sample consists predominantly of large banks.
5 In the aftermath of the global financial meltdown the size of the Irish banking sector decreased. Given
that most banks tent to hold a vast share of their assets in liquid government bonds for liquidity reasons, the sharp increase and decrease of government exposure of Irish banks de facto reflects the overall size of the Irish banking sector.
6 Note that total government securities do not include extra euro area government bonds as the ECB
(unfortunately) does not report data on the bond holdings of governments outside the euro area. However, when looked at the 2014 stress test of the EBA (2014), the exposure of euro area banks to euro area government accounts for more than 80% of total sovereign exposure. Given that especially large banks are operating on a intercontinental scale and consequently hold more bonds outside the euro area (and the fact that the EBA imposes the stress test mainly on large banks), the real share can considered to be even higher.
12
Figure 2 – Financial institutions’ home bias towards government bonds
A. Core countries
B. Periphery countries
Figure 2 clearly shows that between 2007 and 2015 – against the background of the worldwide financial meltdown and subsequent European sovereign debt crisis – in almost all euro area countries (except for Finland), banks reduced their exposures to other euro area governments relative to their domestic government bond holdings. After the introduction of the EMU in 2002 the degree of home bias initially decreased gradually. Apparently, the absence of currency risk led banks to increase their exposures to other euro area governments. From 2007 onwards home bias increased again. As of 2015, the degree of home bias is roughly on the same level as before the introduction of the euro.
The figure also indicates that the level of home bias varies greatly from country to country. In general, home bias is remarkably higher in periphery countries than in core countries. The figures of Italy and Greece are particularly striking. Domestic sovereign exposures of Italian and Greek banks account on average for more than 95% of the total amount of sovereign exposures of their respective banking sectors. This seems to point out that Greek and Italian banks hold highly under-diversified sovereign portfolios.
When looking at the bank level data of the 2014 stress test of the European Banking Authority (EBA), it appears that the level of home bias is relatively large for small banks (EBA, 2014).7 Because large banks are usually multinational corporations with subsidiaries in several
countries, their portfolios are likely to be diversified across countries. Small banks, in contrast to multinational banks, operate on a smaller scale, within a country’s borders. As a result, small banks appear to be predominantly exposed to their own sovereign. Next to home bias, euro area banks also tend to be regionally biased. Most EU banks have substantial exposures to government debt of neighboring countries (EBA, 2010-2014). The Dutch ING Group for instance hold a
substantial amount of both Belgian (around 1.4% of total assets) and German (around 1.8% of
7 It should be noted that the EBA sample itself predominantly exists of large, systemic relevant banks,
13
total assets) government bonds (as of December 2013) (Ibid.). Regional bias could stem from expansion decisions. When banks decide to operate across borders they tend to expand their business first to geographically and culturally related countries. In addition, mergers and
acquisitions are more obvious between banks from neighboring countries. Information asymmetry might also play a role. Information asymmetry might be less prevalent and decisive for neighboring countries compared to other European countries. Bank asset managers are simply better aware of market tendencies in neighboring countries than in more distant countries.
Yet another tendency that emerges from the data the increase in government bond holdings of banks relative to their allocation of credit to the private sector. Figure 3 indicates that between the start of the European sovereign debt crisis in 2009 and the beginning of 2015 the ratio of domestic government bond holdings and lending to the real economy (loans to households and non-financial corporations) increased drastically. This is consistent with my hypothesis that euro banks
decreased their total amount of loans (higher risk weight) in favor of an increase of exposure to domestic government securities (lower risk weight).
In relative terms, the portfolio shift from loans to government securities was more clear-cut for banks in periphery countries. The fact that in Cyprus and Greece the ratio decreased between 2009 and 2015 can be attributed to the bail-out agreements, which implied – among other things – that private investors had to write-down (part of) their Cypriot and Greek government debt
holdings.
Figure 3 – Financial institutions’ loans-to-government debt ratio
A. Core countries
14 3. Literature
The European sovereign debt crisis and the ‘vicious link’ renewed the interest of policy makers and academics in the regulatory relationship between banks and sovereigns. In addition, there is an ongoing debate among academics on what actually drives euro area banks’ behavior towards sovereign debt and – more specific – why euro area banks increased their domestic sovereign exposures in the wake of increasing risk premiums. The first part of this section focuses on the general relationship between banks, sovereign debt and default. The second part focuses on recent academic literature on the driving forces of sovereign bond holdings by financial institutions, euro area banks in particular.
The relationship between the banking system and sovereign crisis is recently covered by Reinhart and Rogoff (2011). For a broad sample of seventy countries, Reinhart and Rogoff (2011) present empirical evidence for the existence of a two-way relationship between bank crises and sovereign debt crises throughout modern economic history (1800-2009). According to Reinhart and Rogoff (2011), ‘banking and debt crisis often occur simultaneously (or in proximity to one another) and, more often than not, banking crises precede debt crises’. The recent European sovereign debt crisis once again made clear that financial crisis and sovereign crisis often go hand in hand.
According to Gennaioli, Martin and Rossi (2013), there is a positive relationship between the level of development of the financial sector and the vulnerability of banks to sovereign defaults. When financial markets are more developed, banks are likely to be more affected by sovereign default. The reason for this is that financial institutions are generally more leveraged in developed countries. In addition, Gennaioli et al. (2013) argue that the ‘decline in private credit is larger when banks hold more government bonds’. More recently, Gennaioli, Martin and Rossi (2014) also find a reversed positive relationship between a sovereign default and the amount of sovereign bond holdings by banks. Based on an extensive data analysis (more than 18,000 banks), they argue that ‘during sovereign defaults banks increase their exposure to public bonds, especially large banks and when expected bond returns are high’. The notion that banks increase their investment in government bonds in the wake of increasing risk premiums on governments bonds increase is called the ‘carry trade hypothesis’. Gennaioli et al. (2014) conclude that it is of great importance not to ignore bond holdings of banks during normal times, because sovereign debt holdings explain a large part of the dynamics during economic distress.
In their DNB Working Paper Popov and Van der Hoorn (2013) distinguish two effects through which an increase in sovereign risk premiums could negatively impact banks. First, a devaluation of the stock of sovereign bond holdings deteriorates banks’ balance sheets (direct effect). Second, concerns about counterparty risk increases the costs of finance (indirect effect). Moreover, Popov and Van der Hoorn (2013) conclude that exposure to sovereign risk had a negative impact on ‘lending by banks in the syndicated loan market’, which is consistent the empirical findings I present.
Concerns about the sovereign-bank nexus are – among others (see for instance Buiter, Rahbari and Montilla (2014) and Korte and Steffen (2014)) – also shared by former Deputy General Manager of the BIS Hervé Hannoun (2011). Hannoun (2011) was one of the first that questioned the risk-free status of sovereign debt. According to Hannoun (2011), ‘sovereign risk
15
pricing in financial markets follows a well-known pattern: long periods of complacency during which risk premiums and perceptions are low, while risks are building up. These periods of complacency are followed by sudden changes or reversals in market sentiments, which are both too abrupt and too late’. Hannoun underlines the interconnectedness of banks and sovereigns by graphically showing a similar pattern between bank CDS spreads and sovereign CDS spreads across the euro area. In Hannoun’s view, it is crucial for regulators to clarify that, although sovereign bonds are a low risk asset class, sovereign debt must be subject to a capital charge that reflects the underlying risk.
Most academics distinguish the preferential (1) the preferential regulatory treatment, (2) carry trade behavior, (3) the macroeconomic environment (including the deficit absorption hypothesis) and (4) and home bias as forces behind the recent surge in sovereign exposures. It should be noted that these determinants are not mutually exclusive.
1. Preferential regulatory treatment
The view that the preferential regulatory treatment implies that sovereign bonds are an attractive investment for EU banks and – as a result – an important catalyst of sovereign debt holdings euro area banks is broadly shared among academics (see for instance Bonner, 2014; Acharya and Steffen, 2013; Korte & Steffen, 2014; Buch, Koetter and Ohls, 2013; Lang and Schröder, 2015; Popov and Van Horen, 2013 and Battistini, Pagano and Simonelli, 2014).
According to Korte and Steffen (2014) current treatment of sovereign debt in the EU CRD/CRR is an important channel through which domestic banks’ non-domestic sovereign
exposures amplify contagion among European sovereigns’. This was the case in Cyprus, when large exposures of the banking sector to Greek government debt drastically worsened government finances. Korte and Steffen (2014) illustrate the contagion channel by showing that Greek sovereign debt ratings and the Cypriot CDS spread followed a similar pattern. In addition, Korte and Steffen (2014) come up with a measure (called the ‘sovereign subsidy’) to quantify to what extend banks are undercapitalized as a result of the zero risk weight. By assigning appropriate risk weights to the sovereign exposures according to the IRB approach, they conclude that the ‘subsidy’ accumulates to approximately €750 billion as of June 2013. Moreover, they argue that the
‘sovereign subsidy’ nearly doubled over the last four years. Finally, Korte and Steffen (2014) state that sovereign bond exposures account on average for more than 200% of core tier 1 capital holdings of banks under the authority of the EBA.
Acharya and Steffen (2013) argue, based on 2010-2012 EBA data, that between March and December 2010 some banks with lower Tier 1 capital ratios, higher risk-weighted-assets and larger loan-to-asset ratios increased their holdings of Spanish and Italian sovereign debt. According to Acharya and Steffen (2013) banks that were short on equity had an incentive to shift their portfolios to the highest-yielding (and hence riskier) assets without having to ‘issue economic capital’. In addition, they argue that governments themselves could have had incentives to
preserve the regulatory arbitrage towards sovereign risk, in order to be able to continue borrowing in times of economic distress. Acharya and Steffen’s (2013) main point is that carry trade behavior of banks in crucial in order to understand the behavior of euro area banks (see ‘carry trade’
16
Based on bilateral daily transaction data (MiFID) of Dutch banks collected by the Dutch Authority of Financial Markets (AFM), Bonner (2014) states that ‘liquidity and capital requirements cause banks’ demand for government bonds to increase beyond their internal risk management targets’.
According to the MiFID-data, bond holdings differ substantially ‘between banks and over time’. Bonner (2014) uses Seemingly Unrelated Regression (SUR), in order to compare bank demand for government bonds to demand for other types of bonds. More specifically, he shows that at the mandatory capital threshold banks’ demand for government bonds increases, while banks’ demand for other types of bonds decreases. Results imply that a lower capital ratio in the previous month causes banks to buy considerably more government bonds and to moderately sell other types of bonds. In addition, banks that are close to the legal liquidity threshold also increase their demand for government bonds. The empirical findings of Bonner for Dutch banks are similar to the results I found on a country aggregated level.
In their ‘Bundesbank working paper’ Buch et al. (2013) show similar results. Based on panel information of all German banks they conclude that ‘larger and less well-capitalized banks and banks with a smaller deposit base’ tend to hold more sovereign bonds. According to Buch et al. (2013) German banks ‘hold more bonds from large euro-area, low-inflation and high-yield
countries’. Furthermore, they argue that German banks only since the fall of Lehman Brothers differentiated between OECD countries and de facto reacted upon macroeconomic fundamentals.
2. Carry trade behavior
As mentioned, Acharya and Steffen (2013) underline the importance of carry trade in
understanding the dynamics behind banks’ peripheral bond holdings during the 2009-2012 period. According to Acharya and Steffen (2013), euro area banks borrowed money at a low interest rate and invested it in high-yield (low quality) sovereign bonds. Undercapitalized banks betted at convergence of spreads between euro area countries and de facto speculated on commitment of political leaders to the euro. By doing so, undercapitalized banks could search for high yield and take on risk, while still being able to meet mandatory capital requirements.
Additionally, Acharya and Steffen’s (2013) estimate a multifactor model with pooled OLS in order to evaluate the relationship between a bank’s daily stock return and both the daily return on ten year periphery (Greece, Ireland, Portugal, Spain and Ireland) and ten year German bonds. As a result, they show that banks’ stock market returns are positively related to returns on GIPSI bonds and negatively related to German bonds. Put differently, banks were having long positions on high-yielding GIPSI bonds and short positions on low-high-yielding German bonds, suggesting carry trade behavior.
Furthermore, Acharya and Steffen (2013) state that the increase of exposure of Spanish and Italian bank to domestic sovereign debt was partly facilitated by the ECB’s ‘Long Term Refinancing Operations’ (LTRO’s) – which in December 2011 and March 2012 allowed banks to borrow against a three year average of the refinance rate. Through the LTRO-facility those banks had access to cheap money. Based on EBA data, Acharya and Steffen (2013) argue that after the LTRO tenders, Italian and Spanish banks increased their domestic sovereign exposures drastically. This view is also shared by ECB president Mario Draghi, who suggested in July 2014 that LTRO money was used to ‘trade sovereign debt and reap big profits’ (Black & Stearns, 2014).
17
Battistini et al. (2013) also find evidence in line with the ‘carry trade hypothesis’. Their findings indicate that banks in periphery countries reacted to an increase of domestic sovereign risk by increasing their domestic sovereign exposures.
3. Macroeconomic environment
Castro and Mencia (2014), on the other hand, undermine the ‘carry trade hypothesis’. They use a vector autoregressive model (VAR), in order to estimate the effect of carry trades (approximated by sovereign yields) on sovereign bond holdings, when corrected for alternative investment
opportunities (approximated by macroeconomic conditions or, more precisely: industrial production and unemployment). Castro and Mencia (2014) conclude – in contrast to Acharya and Steffen (2013) – that ‘macroeconomic conditions play a more relevant role than carry trade incentives to explain the increase in banks’ holdings of domestic sovereign debt in times of stress’.
The European Systemic Risk Board (ESRB) (2015) likewise argues that the evolution of sovereign debt exposure can be attributed to macroeconomic fundamentals. According to the ESRB, the increase of sovereign exposure was partly a result of an increase in government deficits in the years following the 2007 financial meltdown. The increase in public debt issuances was then absorbed by domestic banks as residual buyers. However, the ‘deficit hypothesis’ seems unable to explain the behavior of banks in some core countries, where the increase of government bond holdings was much higher than the actual increase in government debt. German banks for instance also increased their sovereign bond holdings drastically, whereas debt-to-GDP in Germany
increased only slightly between 2007 and 2014 (Eurostat, 2015).
4. Home bias
Many authors argue that home bias also explains some of the dynamics behind sovereign bond holdings of banks. As showed in the previous secton, most euro area banks are predominantly exposed to their domestic sovereign. Home bias of domestic banks can be a result of (1) financial repression (moral suasion) by the government, (2) implicit or explicit bail-out guarantees, (3) hedging against redenomination risk and/or (4) risk indifference or (5) cognitive behavior due to custom behavior or market knowledge.
Becker and Ivashina (2014), Ongena, Popov & Van Horen (2015), Acharya and Steffen (2013) and Battistini et al. (2013) and – less recently – Livschits and Schoors (2009) all argue that home bias can be rooted in financial repression by the government. Governments exceed pressure on domestic banks to buy domestic government bonds when the risk of default increases and finance conditions deteriorate. Becker and Ivashina (2014) argue that ‘financial repression of the banking sector was actively used by European countries during the sovereign debt crisis’. According to Becker and Ivashina (2014) the regulatory environment and carry trade behavior cannot explain why – for example – Spanish banks especially bought Spanish instead of Italian government bonds. They argue that the association of an increase in domestic government holdings with a contraction in corporate loans was more pronounced for banks that were influenced by the government through either direct ownership or board positions.
Financial repression of governments on the banking can also be connected to bailout guarantees. Bailout guarantees (be it implicit or explicit) follow from the fact that banks can be of systemic importance. Banks that load up on government debt commit sovereigns to their survival
18
as they are rather perceived to be ‘too big to fail’. Both Gennaioli et al. (2013) and Becker and Ivashina (2014) shortly mention the bailout guarantee mechanism as a determinant of home bias. Nonetheless, Becker and Ivashina (2014) do not find a relationship between bailouts and domestic bond holdings.
Battistini et al. (2013) argue that the increase of home bias in bank portfolios can also be attributed to an increase in euro area break-up risk. In case of a break-up of the euro area, both liabilities and domestic bond holdings of a financial institution would again be redenominated into a newly formed national currency. Consequently, Spanish banks are – for example – better hedged against redenomination risk of Spanish government bonds than Italian banks. Battistini et al. (2013) find a positive relationship between euro area systemic risks and bank’s sovereign
portfolios. They conclude that home bias and the perceived risk of a country’s disorderly exit from the EMU are positively correlated. This mechanism seems also able to explain the decreasing tendency in the home bias of euro area banks (see home bias chart in the previous section)
In addition, banks could also be indifferent to the risks associated with large domestic sovereign exposures, because banks are already likely to be affected by sovereign stress anyway through several other channels. The amount of NPLs might for instance increase as a result of the deteriorated macroeconomic conditions. In other words, banks do not self-impose concentration limits for domestic sovereign exposures, because the likelihood that a banking system will survive a sovereign default is already very small, regardless of the amount of domestic government bond holdings.
Finally, home bias can also be rooted in irrational behavior of investors. Although behavior is by nature impossible to quantify, home bias could be a result of habits or nationalist motives. In other words, banks prefer domestic government bonds because of tradition or cultural bias. Another explanation might be that foreign banks face asymmetrical information in foreign bond markets and therefore prefer to invest in domestic government securities. The information dynamics could also explain why banks – next to domestic government bonds – also tend to have large exposures to government debt of neighboring or culturally related countries (see previous section).
19 4. Data and methodology
Most data used in this analysis is subtracted from the ECB Statistical Data Warehouse (SDW). The ECB reports country aggregated data of balance sheet of Monetary Financial Institutions (MFIs). MFIs are defined as credit and financial institutions whose business is to receive deposits (or close substitutes) from entities other than MFIs and – for their own account – grant credits and/or make investments in securities. The ECB classifies money market funds also as MFIs. As of January 2015 more than 6,500 MFIs operate in the euro area, of which more than 90% are banks (ECB, 2015). All banks are legally obliged to report their balance sheet items to central banks in the countries in which they operate. The ECB, in turn, collects these data from central banks. For this analysis I used data on the following balance sheet items: (1) debt securities issued by the domestic general government (i.e. the banking sector’s domestic sovereign exposure), (2) capital and reserves (i.e. a banking sector’s level of capitalization), (3) total loans total loans to non-financial corporations, (4) total loans to households and non-profit institutions serving households and (5) total assets. All observations within these datasets refer to outstanding amounts at the end of the period.
With respect to use of these data, a couple of remarks should be made. First, it should be noted that general government debt securities, next to national government debt securities, also include debt securities of local and state governments. Second, the ECB distinguishes lending of financial institutions to non-financial corporations; households and non-profit organizations supporting households; financial corporations; and insurance companies and pension funds. I considered the sum of loans to non-financial corporations, households and non-profit
organizations supporting households to be the equivalent of bank lending to the real economy. Third, the ECB’s definition of capital and reserves include equity capital, non-distributed benefits or funds and specific and general provisions against loans, securities and other types of assets. It is important to note that – although the ECB definition includes reserves that banks hold at the ECB – variance in capital and reserve figures are mainly due to variances in banks’ capital. The amount of reserves that banks are obliged to hold at the ECB is more or less constant over the investigated period and similar for all countries. Hence, fluctuations of capital and reserves for greater part reflect fluctuations of the level of capitalization of the banking sector.
In addition to the ECB data, I used data from Eurostat on (6) sovereign bond yields and (7) gross general consolidated government debt of euro area countries. Sovereign bond yields refer to ‘monthly averages of central government bond yields on the secondary market, gross of tax, with a maturity of around 10 years’ (Eurostat, 2015).8 Following convention, spreads are calculated as the
difference between the German ‘bund’ and the prevailing secondary market long-term (10 year) bond yield of the country concerned. Data on a country’s indebtedness (as share of GDP)
compromises the subsectors: central government, state government, local government and social security funds (Eurostat, 2015). The data corresponds to quarterly end of period government debt. Monthly figures are received through linear interpolation of these quarterly figures. Box 7 in the appendix summarizes the data characteristics.
Because data on domestic government bond holdings is aggregated on a country level, I used a panel data model with time fixed effects. Time fixed effects regression controls for variables that
20
are constant across countries but evolve over time (for instance, the state of the world economy), which implies that the model has a different intercept for every monthly observation (Stock & Watson, 2012). The data analysis covers the following countries: Austria, Belgium, Finland, France, Germany, the Netherlands, Cyprus, Greece, Italy, Ireland, Portugal and Spain. The Baltic States, as well as Slovakia and Slovenia are excluded from the analysis, notably due to limited data coverage as a result of their late entrance to the EMU. Luxemburg and Malta are excluded from the analysis because of their size.
Box 2 summarizes the baseline model, which estimates the relationship between banks’ level of capitalization, sovereign yield spread and the sovereign’s indebtedness and the amount of government bond holdings. The model is estimated for the period January 2007 until January 2015. This period includes the worldwide financial meltdown and the European sovereign debt crisis as well as the G20 agreement upon adopting the new Basel III capital requirements.
Box 2 – Baseline model
(1) (2) (3) (4)
The first regression specification estimates the effect of the level of capitalization, the yield on domestic government bonds and a country’s indebtedness on the amount of domestic government bond holdings by banks as a share of total assets. The banking sector’s capitalization is defined as the amount of capital and reserves divided by the amount of total assets. It should be noted that supervisors generally look at capital as a share of risk-weighted assets when they assess a bank’s financial viability. Unfortunately, data on the level of capitalization of banks relative to their risk weighted assets is not issued on a country aggregated level, neither by the ECB nor by any other data collecting institution. The level of capitalization used in this equation, which is in fact close to the banking sector’s leverage ratio, can nonetheless be used as a proxy for the banking sector’s capitalization. A negative relationship between the level of capitalization and the amount of
21
government bond holdings indicates that low capitalized banks hold more government debt, suggesting that the zero risk-weight might provide incentives for banks to lend to their sovereign.
The return on domestic government bonds (vis-à-vis the Germand bund) is included because banks evidently take returns on their investment into account when they make investment decisions. Moreover, inclusion of yield spreads controls for carry trade behavior of euro area banks, as described by Acharya and Steffen (2013), and assesses whether this also took place on a systemic level and for a more prolonged period. A positive relationship between spreads and domestic government exposure could indicate such carry trade behavior. Government debt (as a percentage of GDP) enters the equation to incorporate the so-called deficit absorption hypothesis, according to which banks acted as residual buyers and mainly responded to consecutive years of government deficits and a subsequent surge of government debt, whether or not under the influence of moral suasion or financial repression by their government (see literature section). A positive relationship between sovereign debt and sovereign exposure might point in this direction.
The second specification is somewhat similar to the first, but estimates instead the
association between a bank’s level of capitalization, the yield on domestic government bonds and a country’s indebtedness with the amount of government bond holdings by banks relative to their lending to the real economy, which for clarity reasons is referred to as their portfolio composition. The portfolio composition can be seen as a proxy for banks shifting their portfolio away from lending to the real economy (i.e. households and non-financial corporations) to lending to their own sovereign. This captures again the notion that badly capitalized banks have an incentive to invest in sovereign debt in the wake of more demanding capital requirements. Moreover, this specification, in contrast to the first specification, also corrects for the fact that total assets of European banks also increased over the investigated period. The third and fourth specifications transform the dependent variable and relevant independent variables (sovereign debt and sovereign spread) of the first and second specification, respectively, into first differences.
Box 3 summarizes the extended model, which is also estimated for the period 2007 until 2015. Box 3 – Extended model
(5) (6) (7) (8)
22 (9) (10)
The fifth and sixth specifications are essentially the same as the first and second specifications augmented with a dummy variable for periphery countries and a lag for the level of capitalization. The latter is included because capitalization is assumed to influence bank behavior regarding sovereign debt in the next period. The dummy takes account of different characteristics of core and periphery countries, such as the differences in economic fundamentals, which eventually led to contrasting treatment by financial markets and capital flows from the periphery to the core.
Specifications 7 and 8 include interactions between the periphery dummy and both the level of the banking sector’s capitalization and the size of domestic sovereign debt. These
interactions are included to test whether the interaction between the level of capitalization and the amount of government debt with domestic government bond holdings (both as a share of total assets and as a share of lending to the real economy) is systematically different for core and periphery banks. It could, for instance, be the case that the deficit absorption hypothesis applies more to periphery countries, whereas the behavior of banks in core countries can be better explained by the level of capitalization. Finally, specifications 9 and 10, address the interaction between the periphery dummy and sovereign spread, to test whether periphery and core banks reacted differently to returns on sovereign debt. The patterns in bond yield spreads during the European debt crisis were obviously different for core and periphery countries, which could have resulted in different bank behavior towards domestic sovereign debt in these countries as well.
Box 4 summarizes the descriptive statistics of all the variables used in the analysis. Box 4 – Summary statistics
Variable Observations µ σ min max
country 1152 - - 1 12
date 1152 - - 1 96
domestic government exposure portfolio composition 1152 1152 0.0330 0.0993 0.0243 0.0719 0.0002 0.0014 0.1077 0.3077 capitalization 1152 0.0759 0.0312 0.0353 0.2402
sovereign yield spread 1152 1.8554 3.1543 -0.120 27.390
sovereign debt 1116 77.945 32.914 25.920 177.50
23 5. Results
Box 5 summarizes the results of the baseline model. The monthly intercepts are excluded for clarity reasons.
Box 5 – Results baseline model
specification (1) (2)
variable domestic government exposure(t)
portfolio composition(t)
capitalization(t) 0.0403
(1.557) -0.551*** (-6.983)
sovereign yield spread(t) -0.000163
(-0.597) -0.00663*** (-7.966) sovereign debt(t) 0.000422*** 0.00167*** (15.67) (17.11) Observations 1116 1116 R-squared 0.382 0.352 R-squared (adjusted) 0.324 0.292 specification (3) (4)
variable Δ domestic government exposure(t)
Δ portfolio composition(t)
capitalization(t) 0.00212 0.00619
(0.853) (0.855)
Δ sovereign yield spread(t) 0.000835***
(6.489) 0.00208*** (5.552)
Δ sovereign debt(t) 1.54e-05** 3.55e-05
(1.995) (1.577) Observations 1116 1116 R-squared R-squared (adjusted) 0.0718 0.151 0.0589 0.139 t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1
The results of the first specification of the baseline model, as summarized in box 5, are mixed. The first specification does not provide significant results for the effect of the level of capitalization on the amount of domestic government bond holdings. In addition, it does not provide a significant coefficient on sovereign yield spread either. It does, however, provide significant results for sovereign debt. The estimated coefficient on sovereign debt is in line with economic intuition and conform expectations: banks tend to hold more government bonds when their sovereign is highly-indebted. This is consistent with the deficit-absorption hypothesis.
The second specification shows significant results for all three variables. The negative coefficient with respect to the level of capitalization is consistent with my hypothesis: the banks’ level of capital is negatively associated with their holdings of domestic government bonds relative to their lending to the private sector. Low capitalized banks tend to invest their money in domestic government bonds rather than lend it to households and firms. The negative coefficient for
24
sovereign yield spread is, though significant, not in line with expectations. The coefficient suggests that yield spread is (albeit limited) negatively associated with the amount of domestic government bond holdings relative to private sector lending. At first, this might seem rather surprising, as it is evidently opposed to the prevailing ‘carry trade hypothesis’, according to which banks increased their bond holdings in response to increasing returns on government debt in order to reap profits. However, the negative relationship between spreads and domestic government bond holdings can be caused by the fact that government bond yields in core countries decreased over the period, as a result of the so-called ‘save haven effect’, while banks in core countries also increased their exposure to their sovereign at the same time. In other words, the carry trade hypothesis might not be able to explain bank behavior in core countries. With the inclusion of interaction between a periphery dummy and sovereign yield spread, specifications 9 and 10 address the possibility of different dynamics between bank behavior and spreads in both groups of countries.
Results for sovereign debt in the second specification are again in line with expectations and consistent with the deficit absorption hypothesis. Between 2007 and 2015, banks that
operated in countries with high government debt, held on average more government debt relative to the amount they lent to the real economy. This result suggests that high government
indebtedness could have implications for the real economy and lead via portfolio decisions of banks to a contraction of bank lending to households and corporations. This question is, however, beyond the scope of this paper.
The third and fourth specifications show no significant results for the level of capitalization. The results for coefficients on the differences of sovereign yield spread and sovereign debt are significant for both specifications, while the former is only significant for the third specification. In contrast with the first and the second specification, the third and fourth specifications show a positive coefficient for swings in returns on government debt. Apparently, banks on average reacted to increases in returns on domestic government debt by increasing their exposure to their sovereign, and vice versa, which is in fact consistent with the ‘carry trade hypothesis’.
The results of the extended model are summarized in box 6. Box 6 – Results extended model
specification (5) (6)
variable domestic government exposure(t)
portfolio composition(t)
capitalization(t-1) -0.0725**
(-2.454) -0.591*** (-6.365)
sovereign yield spread(t) -0.000611**
(-2.248) -0.00683*** (-7.995) government debt(t) 0.000372*** 0.00165*** (13.83) (19.45) periphery(t) 0.0133*** (8.060) 0.00472 (0.822) Observations 1104 1104 R-squared 0.418 0.350 R-squared (adjusted) 0.363 0.289
25
specification (7) (8)
variable domestic government exposure(t)
portfolio composition(t)
capitalization(t-1) -0.186*** -1.170***
(-3.768) (-7.601)
sovereign yield spread(t) -0.000636**
(-2.341) -0.00693*** (-8.186)
government debt(t) 0.000368*** 0.00163***
periphery(t)
periphery capitalization(t-1)
periphery sovereign debt(t)
(13.69) 0.00240 (0.570) 0.164*** (2.844) -1.99e-05 (-0.349) (19.38) -0.0524*** (-3.923) 0.842*** (4.689) 3.70e-05 (0.208) Observations 1103 1104 R-squared R-squared (adjusted) 0.422 0.367 0.364 0.302 specification (9) (10)
variable domestic government exposure(t)
portfolio composition(t)
capitalization(t-1) -0.201*** -1.331***
(-4.060) (-9.003)
sovereign yield spread(t) 0.00602** 0.0649***
government debt(t) (2.463) 0.000344*** (8.854) 0.00137*** (12.24) (16.28) periphery(t) 0.00380 (0.898) -0.0373*** (-2.948) periphery capitalization(t-1) 0.186*** (3.210) 1.083*** (6.170)
periphery sovereign debt(t) -1.38e-05 0.000102
(-0.243) (0.601)
periphery sovereign yield
spread(t) -0.00647*** (-2.739) -0.0698*** (-9.876) Observations 1104 1104 R-squared R-squared (adjusted) 0.427 0.371 0.420 0.364 t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1
Inclusion of the periphery dummy variable changes the coefficients of the model estimates of the first and second specification. The periphery dummy itself has a positive effect on the amount of
26
government bond holdings as a share of total assets, though only statistically significant in specification 5. The positive coefficient of the periphery dummy is consistent with the patterns in sovereign debt holdings described in section 2. Specification 6 suggests that the amount of
domestic government bond holdings of banks relative to their lending to the real economy between 2007 and 2015 was not statistically different between periphery countries and core countries.
All specifications of the extended model show a negative association between the level of banks’ capitalization and the amount of domestic government bond holdings. These results points out that, on average, low capitalized banks in the euro area held more government bonds between 2007 and 2015 than high capitalized banks. This applies for both domestic government bond holdings as a share of total assets as well as for domestic bond holdings as share of lending to the real economy. These results are consistent with my hypothesis. The interaction between the periphery dummy and the level of capitalization yields a positive coefficient for all specifications. This suggests that, although a negative relationship between the level of capitalization and bond holdings exists on average for both groups, the negative effect of capitalization on bank behavior was more pronounced for banks in core countries. This is noteworthy, as behavior of low
capitalized banks was thus far particularly associated with investments in high-yielding government debt. The results point out that low capitalized banks in core countries also tend to hold more low-yielding domestic government debt.
The results of the interaction between the periphery dummy and sovereign debt is, on the contrary, not statistically significant. It can therefore not be concluded that banks in periphery countries reacted differently to high levels of government debt than banks in core countries. The overall effect, which is captured in specifications 5 and 6, is nonetheless positive and statistically significant. This is consistent with the ‘deficit absorption hypothesis’, according to which bank behavior regarding domestic government bonds is mainly a result of the amount of domestic government debt. The results, however, are not conclusive about whether the surge in domestic government bonds is caused by financial repression or moral suasion by the respective
government, as suggested by some authors (see literature part for more details).
The interaction between the periphery dummy and sovereign yield spread is included in specifications 9 and 10. The coefficient on the interaction between the periphery dummy and sovereign yield spread is negative for both specifications, which implies that banks in periphery countries on average reacted negative to increases in government bond yields compared to banks in core countries. The results are somewhat counterintuitive, as banks in periphery countries on a country aggregated level continued to increase their government bond holdings in the wake of rising risk premiums during the European debt crisis. However, is should be noted that the investigated period (2007 to 2015) is characterized by both upswings and downswings in bond yield spreads. From 2012 onwards, bond yield spreads in the euro area decreased again. Hence, banks also increased their domestic government bond holdings in the wake of decreasing bond yields, whereas the decrease in sovereign bond yields after the summer of 2012 was obviously more pronounced in periphery countries. Altogether, the results of the extended model do not suggest carry trade behavior on a country aggregated level from 2007 to 2015.
