1
Bachelor Thesis
BSc. Economics and Business – Finance and Organization
Field: Finance
The determinants of Euro banks’ domestic sovereign debt holdings
A European analysis of country level drivers of domestic sovereign exposure during and after
the European debt crisis
Thomas van den Berg
10551778
University of Amsterdam (UvA)
26
thof June, 2018
2
Abstract
This thesis studies the determinants of banks’ domestic sovereign debt holdings in Euro countries during and after the European debt crisis. It analyses the regulatory capital drivers of domestic sovereign exposure embedded in European law on the one hand and uses multiple random effects panel data regressions to find other macroeconomic determinants of sovereign exposure on the other. Evidence suggests that for both stressed as non-stressed countries, low capitalized banks tend to increase sovereign exposure. In addition, higher government debt has a positive effect on
domestic exposure in stressed countries. No conclusive evidence is found to support the carry trade hypothesis but nevertheless results point towards a search for yield of banks in non-stressed countries. For policy makers this means that the withdrawal of EU regulation allowing banks to deviate from the non-zero risk weight for sovereign bond holdings could be a strong move towards breaking the sovereign bank nexus since it increases the banks’ capitalization and decreases sovereign debt holdings.
Statement of Originality
This document is written by Student Thomas van den Berg who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are 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.
3
Table of content
1. Introduction ... 4
2. Literature review ... 5
2.1 Regulatory capital requirements ... 7
2.2 Other drivers of sovereign exposure ... 8
2.3 Hypotheses ... 12 3. Data ... 13 4. Methodology ... 15 5. Results ... 17 5.1 Policy implications……… 22 6. Conclusion……… 23 7. Reference list……… 24
4
1. Introduction
The European debt crisis (EDC) showed that government bonds contained more risk than anticipated before (Alsakka, ap Gwilym, & Vu, 2014). After this, literature about the so called sovereign bank nexus emerged, describing the risk transmission between both the state and banks as vice versa (Podstawski & Velinov, 2017; Pagano, 2016). Little attention is paid to the determinants of banks’ sovereign debt holdings, while these holdings resemble the most important transmission channel of risk in the sovereign bank nexus. This proves the analysis of sovereign debt holdings is key.
More reasons to focus on the sovereign debt holdings and its determinants are firstly that research already showed the impact sovereign exposure could have. It has the tendency to create inefficient equilibria known as diabolic loops: pessimistic views about government solvency could lead to sovereign debt revaluations at banks with high exposure (Pagano, 2016; Acharya, Drechsler, & Schnabl, 2014). Secondly, Faia (2017) finds consequences for funding cost of banks for highly exposed banks. In addition, it has been established that EU banks increased their sovereign exposure over the years (BIS, 2013). Moreover, contemporary regulation such as Basel II and III makes
sovereign debt more attractive for banks in ways that will be explained in the theoretical framework (Ibid.).
This thesis will therefore expand knowledge on the underexposed determinants of sovereign debt holdings of EU banks during and after the EDC. This leads to the following central question: What are the determinants of the amount of banks’ domestic sovereign debt holdings in the Eurozone during and after the European debt crisis? Using panel data regressions the influence of sovereign yield, sovereign debt, bank capitalization, market size, crisis and stressed countries on sovereign debt holdings are tested. This might help policy makers in trying to get the overall sovereign exposure down, since they ideally want that the sovereign banks nexus ceases to exist to restrict the influence of a financial crisis.
Main findings are that seemingly no new policy is needed to start the breakdown of the sovereign bank nexus. This thesis find that low capitalized banks in stressed and mainly non-stressed countries are a key driver of increasing sovereign debt holdings. Also, government debt plays an important role in stressed countries. The thesis concluded that by revising existing exceptions
enabling banks to assign a zero risk weight on sovereign bonds the capital position of banks improves and the attractiveness of sovereign bonds decreases.
The thesis is structured as followings. The literature review elaborates first on capital regulations and then on other drivers of sovereign exposure which are tested in regressions. It ends with composing hypotheses. Furthermore, the data section explains what the variables used in the models consist of. The methodology explains more about the specific regression models used and in
5
the results section the outcomes are discussed and linked to policy implications. This thesis ends with a summarizing conclusion and suggestions for further research.
2. Literature review
The literature review starts with a brief explanation about the emergence of the sovereign bank nexus literature to clarify the position of this thesis in it. The remaining part of the review is structured by a distinguishment between roughly the two main drivers of sovereign exposure. First the regulatory driver is discussed and in the second part the focus shifts towards non-regulatory factors that possibly increase sovereign debt holdings. This part will also emphasize on which aspects this thesis differs from existing literature. The literature review will conclude with summarizing the main expectations and transform them to statistical hypothesis.
As mentioned in the introduction, the sovereign bank nexus literature emerged during (and slightly before, although not under this umbrella concept) the EDC. According to Acharya, Drechsler and Schnabl (2014), prior to the financial crisis in 2008 there were no signs of sovereign credit risk in developed countries. The common view was even that there was an almost zero probability that such risk would be a concern in the future. The aftermath of the financial crisis and the EDC showed that especially in Europe sovereign credit risk became a severe problem and highlighted the connection between banks and governments (Drago & Gallo, 2017). Podstawski and Velinov (2017) describe how these events incentivized interest in the transmission of risk between governments and banks via the so-called diabolic loop. Within this loop, authors focused on different aspects which are briefly outlined to understand the place of this thesis in the sovereign bank nexus literature.
Podstawski and Velinov (2017) focus for example on the risk transferring from banks to the sovereign and conclude that rising sovereign debt holdings increase default risk in stressed countries while higher sovereign exposure had a stabilizing effect on default risk in core countries. In contrast with this thesis however, they do not consider domestic sovereign debt holdings and focus on the transmission of risk from banks to the sovereign. Drago and Gallo (2017) research the other side of sovereign bank nexus, consistent with this thesis, and state that the risk transmission from the sovereign to banks is almost completely explained by the asset (holdings) channel, funding channel and rating channel. Also, they conclude that regulation effects amplify the risk transmission which is why this thesis elaborates more on this topic in the next part of this review.
In addition, some authors such as Bernoth and Erdogan (2011) and Battistini, Pagano and Simonelli (2014) focus on the role of government default and credit rating agencies in this discussion, while Drago and Gallo (2017) and Alsakka et al. (2014) research the influence of credit rating
6
transmission channels are defined which together maintain the diabolic loop. This thesis pertains on the transmission of risk from the sovereign to the banking sector. In contrast to the above mentioned literature I focus merely on the most important transmission channel: the asset holdings channel. Now, all channels transferring risk from the sovereign to banks are briefly outlined.
The Bank for International Settlements (2011), was the first in identifying four main channels: the collateral/funding channel, rating channel, guarantee channel and asset holdings channel. Firstly, increased sovereign risk can reduce the value of a banks’ collaterals in the collateral channel.
Sovereign securities are often used as collateral to secure wholesale funding and if the sovereign risk rises, the collateral value decreases due to reduced availability and eligibility of the collateral. Assuming the asset is already used as collateral, it could lead to margin calls. Secondly, a sovereign downgrade may influence the domestic bank rating negatively as proved by Drago and Gallo (2017) and Alsakka et al. (2014). In this rating channel, Alsakka et al. (2014) find transmission of risk between governments and banks in merely stressed countries. Moreover, they find correlation with the sovereign exposure of banks indicating that the sovereign debt holdings, main instrument in the asset holdings channel, also play a role in this channel. Thirdly, in the guarantee channel the lower funding costs due to both implicit (too big to fail) and explicit government guarantees is key for risk transmission. Higher sovereign risk will reduce guarantee value, worsen the fiscal position of banks and raise funding issues (BIS, 2011).
Fourthly, the asset channel is in the literature treated as the most important channel to transfer sovereign risk and of main interest for this thesis (ESRB, 2015). Increases in sovereign risk could effect banks in the most direct way: through their sovereign bond holdings on the balance sheet. Losses on sovereign portfolios weaken banks’ balance sheet and rise the cost of funding on the one hand (Drago & Gallo, 2017; BIS, 2011). On the other hand, most securities are held at market value and hence a fall in value of sovereign bonds has direct effects on banks’ income statement and leverage (BIS, 2011; ESRB, 2015). Finally, Afonso, Fuceri and Gomes (2012) find that spillover effects of sovereign rating downgrades were only transmitted via sovereign bond holdings on banks’ balance sheets. Therefore, this thesis targets the sovereign bond holdings of banks and hence the asset holdings channel.
To conclude the literature overview, an elaborated part of the transmission (asset) channel literature focuses on the consequences of sovereign debt holdings on banking activity, lending and profits (Drago & Gallo, 2017; Podstawski & Velinov, 2017; Angeloni & Wolff, 2012). However, barely any research exists about the drivers of banks to increase their sovereign exposure. Most authors such as Battistini et al. (2014) only partly touch upon the drivers of sovereign debt holdings and do not address it as the main topic. Furthermore, articles who do solely focus on the determinants of sovereign debt holdings often have a limited scope. Buch, Koetter and Ohls (2016) for example only
7
take into account German banks. This thesis contributes to the asset holdings channel literature by partway filling the gap of information on the determinants of sovereign exposure. It provides a more complete analysis of the sovereign debt holdings in Euro countries by including countries such as Luxembourg, Malta, Slovakia and Slovenia. The methodology on the other hand, consisting of panel data regressions, is commonly used in the literature (Drago & Gallo, 2017; Buch et al., 2016).
There is a possible explanation why there is little research about the determinants of
sovereign exposure. As briefly mentioned before, the increasing sovereign exposure during and after the EDC roughly has two types of causes. On the one hand, changing regulatory capital requirements play an important role as a determinant of enlarged sovereign exposure. On the other hand,
macroeconomic events and carry trade could explain the amounts of sovereign debt holdings on bank’ balance sheets. To give a complete overview of the determinants of sovereign exposure the literature review will be completed by first elaborating on the regulatory requirements making sovereign bond holdings more attractive. Secondly, research about possible other determinants, which are tested in regressions is outlined and based on this hypotheses are formulated.
2.1 Regulatory capital requirements
The Basel Accords require banks to assign different risks to different classes of assets and obliges them to keep a certain percentage of capital against these risk weighted assets. The Capital Requirements Directive (CRD), which is used to implement the Basel Accords in European law, contains several clauses making sovereign debt holdings more attractive for banks than other assets. Firstly, the CRD makes it possible for banks to assign a zero risk weight to government bonds issued in the domestic currency (Popov & van Horen, 2013). This means banks are not obliged to keep capital against sovereign bond holdings since they are not part of the risk weighted assets. Secondly, the CRD exempts the Euro government debt from the general 25% limit on large exposure which normally applies to all asset holdings. In Europe, most of the debt is issued in Euro and hence banks hold a significant amount of government bonds without raising their risk weighted assets. Moreover, this CRD regulation enables banks to assign a zero risk weight to every government bond they buy in Euro countries regardless of the sovereign risk. In practice, this means that banks hold between 75% of Tier I capital for Italian and German banks in sovereign bonds to over 200% in Belgium (BIS, 2011). In other words, the preferential treatment of sovereign debt compared to loans to companies and households induces banks to search for the more risky government bonds rather than loan to the real economy (Pagano, 2016).
8
loans. The Basel III rules require banks to build up short term liquidity buffers and this is measured with the Liquidity Coverage Ratio (LCR). To meet the LCR, assets are divided into less liquid and highly liquid assets. The less liquid assets could only represent maximum 40% of the total liquidity buffer. Government bonds however are declared as highly liquid and hence could be unlimited included in the liquidity buffer to make sure banks never exceed the 40%.
On the other hand, the Bank of International Settlement (2013) argues that the Basel capital requirements do not simply prescribe a zero risk weight for sovereign bond holdings. In short, banks may adopt two methodologies regarding the treatment of sovereign debt holdings: the Standardized Approach (SA) and the Internal Ratings-Based (IRB) approach. Both approaches, in rule, assign a non-zero risk weight to sovereign debt. In the US the IRB is mandatory but in Europe authorities allowed banks that use the IRB approach to use the Standardized Approach for their sovereign bond holdings. However, in applying the SA approach EU authorities decided to set the risk weight of sovereign exposure to EU member states at zero. In short, although both approaches in rule assign a positive risk weight to sovereign exposure, regulation imposed by EU authorities makes it still possible for banks to assign the sovereign debt holdings zero risk. Pockrandt and Radde (2012) add that in CRD II and CRD III directives the zero risk weighting remained unchallenged.
To summarize, banks have many reasons to value government bonds more than other loans due to European regulation. The European Parliament (2017) showed that banks’ sovereign debt holdings have indeed been increasing until 2015 in Europe and are still at a high level. In addition, Popov and van Horen (2013) show that banks’ domestic sovereign exposure (exposure to the sovereign where the bank is situated) almost doubled between 2011 and 2015.
2.2 Other drivers of sovereign exposure
Now it has been established why sovereign debt is attractive for banks from a regulation point of view, this part elaborates on other drivers of sovereign exposure. Furthermore, it is explained in more detail how this thesis differs from other sovereign bank nexus literature and in the last part the expectations posed below are transformed to statistical hypothesis. Roughly, the current research describes the following forces as drivers of sovereign exposure: the carry trade hypothesis, home bias and macroeconomic factors. Research about these forces is outlined below.
9 A. Carry trade hypothesis
According to Altavilla et al. (2017, p. 2113), the carry trade hypothesis can be described as a search for yield by banks. They argue that poorly capitalized banks are more incentivized to buy high yield government bonds to “gamble for resurrection”. Also, they show that banks with lower regulatory capital grow high-yield sovereign bonds more than other banks, which is consistent with their explanation of the carry trade hypothesis. Moreover, banks attempt to profit from borrowing money against low interest rates and reinvest this in high yield sovereign bonds (Acharya & Steffen, 2014; Battistini et al., 2014). The last reason for undercapitalized banks to search for the highest yield bonds is that they try to meet the capital regulations without the need to issue economic capital next to raising profit (Acharya & Steffen, 2014).
Acharya and Steffen (2014) also find that GIIPS banks increased their sovereign exposure to stressed countries while non-GIIPS banks decreased their exposure to stressed countries. GIIPS banks are banks in Greece, Italy, Ireland, Portugal and Spain: in this thesis classified as stressed countries as will be explained in the data section. In addition, some stressed country banks used the ECBs Long Term Financing Operations (LTROs) in 2011 and 2012 to increase their domestic exposure which implies that banks in stressed and non-stressed countries could response differently to high yields. Buch et al. (2016) confirm these differences by concluding that they found no evidence for a search for yield of German banks.
The distinguishment between stressed and non-stressed (core) countries is also in general key in the debate about sovereign exposure because many authors believe this distinguishment could give insight in the amplifying effects on rising sovereign debt holdings. As mentioned above, Buch et al. (2016) find other drivers of sovereign exposure for German banks (a non-stressed country) compared to similar research on banks in stressed countries. Battistini et al. (2014) for example find strong evidence of banks’ search for yield in stressed countries and in contrast with Buch et al. also in non-stressed countries. This implies that in stressed countries the search for yield is at least a stronger driver of sovereign exposure than in non-stressed countries. Testing the
amplifying effects of stressed countries is not unique in the sovereign bank nexus literature, although it is new in research solely focused on the determinants of sovereign exposure. Pagano (2016) for example argues that sovereign exposure in stressed countries amplifies the transmission of risk between the sovereign and the banking sector. However, this is partly due to higher sovereign exposure in stressed countries. In addition, Altavilla et al. (2014) and Popov and van Horen (2013) find that the consequences of sovereign exposure for bank performance in stressed countries are stronger than in non-stressed countries.
10
both stressed as in non-stressed countries have more sovereign exposure. Moreover, it implies that the sovereign yield spread incentivizes banks to buy more sovereign debt. Finally, considering the amplifying of stressed countries, it is expected that low capitalized banks and the search for yield show stronger effects in stressed countries.
B. Home bias
The concept home bias in the sovereign bank nexus literature refers to banks buying more
government bonds of their home country than other countries. Home bias can also explain some of the dynamics behind increasing sovereign exposure (Pagano, 2016). Home bias can be a result of moral suasion, government guarantees and hedging. Moral suasion is financial repression of a government towards its domestic banks in order to increase demand for sovereign debt in times when it is low (Battistini et al., 2014). In addition, Acharya and Steffen (2013) argue that this is most common for stressed countries since they are exposed to more risk. Gennaioli, Martin and Rossi (2014) add that bondholding patterns and moral suasion change during a sovereign crisis: mainly large banks are forced by governments to buy high-yield bonds because typically in crisis times demand is weak. These high-risk governments want to drive their yield down to signal a healthy low-risk status. Moreover, this implies that for banks in stressed countries the home bias of their
sovereign exposure increases more.
Secondly, government guarantees could grow a banks’ sovereign exposure. Recall that sovereign debt holdings also played a role in the guarantee transmission channel. Governments will implicitly or explicitly support banks of systemic importance for their economy. As a consequence, banks buy the domestic sovereign bonds to increase their importance in the eyes of the government improving the chance they get saved during distress. Both Gianniaoli et al. (2014) and Becker and Ivashina (2014) mention government guarantees as a driver of domestic sovereign exposure although Becker and Ivashina did not find significant effects of government bailouts on domestic exposure.
Thirdly, Battistini et al. (2014) argue that home bias is correlated with the systemic component of sovereign risk, referring to the risk of the collapse of the Euro. In case this happens, both assets and liabilities will be redenominated into a new national currency. So for example Italian banks with domestic (Italian) sovereign bond holding are better hedged against this risk than foreign banks.
Figure 1 and 2 illustrate the development of home bias at the beginning of the EDC, at the end of the crisis and recently. Figure 1 shows that the non-stressed countries in general increased their home bias during the EDC and are at a higher level than before the EDC. Exceptions are Malta
11
and Slovakia but their levels of home bias already were relatively high. Only Finland shows another trend. Figure 2 shows growing home bias and for all stressed countries during the EDC. After the crisis, home bias decreased for all stressed countries except for Cyprus.
Finally, in the sovereign bank nexus literature most papers distinguish between country level and bank level determinants, this paper focuses on country level determinants (Gianniaoli et al., 2014; Buch et al., 2016; ESRB, 2015). On a country level, the only detailed breakdown of sovereign bond holdings possible, is total sovereign bond holdings and domestic sovereign bond holdings. Thus, based on the key role of home bias in explaining the dynamics behind sovereign exposure and the data availability, this thesis only takes the domestic sovereign exposure into account. Therefore, sovereign exposure in this thesis always refers to domestic sovereign exposure.
C. Macroeconomic fundamentals and fiscal imbalances
Buch et al. (2016) conclude that next to the home bias, macroeconomic determinants of sovereign exposure became more important after the fall of Lehman Brothers. An example for which they found significant results for German banks is government debt. The European Systemic Risk Board (2015) define these drivers as macroeconomic fundamentals and fiscal imbalances and summarize it as the deficit absorption hypothesis. They find evidence that fiscal imbalances, especially high government debt, influence the sovereign exposure. In line with this, BIS (2011, p. 14) concludes that “holdings of domestic government bonds as a percentage of bank capital tend to be larger in
countries with high public debt”. Cipollini, Coakley and Lee (2013), argue that sovereign bond holdings are the main cause of European debt integration. They conclude that government debt (to GDP) ratios are the main “macroeconomic fundamentals” in this integration. Therefore, government
0,0% 20,0% 40,0% 60,0% 80,0% 100,0% AU BE GE FI FR LU ML NL SV
Home bias non-stressed countires
2009 2014 2018 0,0% 20,0% 40,0% 60,0% 80,0% 100,0% CY ES GR IE IT PT SL
Home bias stressed countries
2009 2014 2018
12
debt is expected to increase domestic sovereign exposure in this thesis.
Castro and Mencia (2014) stress that government debt is even more important as determinant of sovereign bond holdings than carry trade incentives during financial turmoil. This means that during financial turmoil, such as the EDC, the sovereign exposure could increase more severely. Although this thesis partly researches the same factors as the ESRB and BIS, it entails a more extensive regression. Other authors often focus on a limited time period within the EDC, partly due to a lack of data, but this paper extents the scope till after the EDC and combines different hypothesis discussed in several papers (Alsakka et al., 2014; Bruyckere et al., 2013).
Finally, geography could be a valuable determinant of sovereign exposure. Notice that all of the factors explaining home bias hint to an important role for stressed countries. According to Pagano (2016), a banks’ home bias is key in improving knowledge about the exacerbating of the sovereign bank nexus in stressed countries. In addition, the carry trade hypothesis, as previously discussed, also implies a stronger growth of domestic sovereign exposure for stressed countries since their yield is higher (especially during the EDC). If the regulatory incentives for banks to increase their sovereign exposure are also taken into account, the importance of yield (and thus of the
distinguishment between stressed and non-stressed countries) should be looked into in this thesis.
2.3 Hypothesis
To summarize, based on the carry trade hypothesis it is expected that on the one hand lower capitalized banks have more government exposure which leads to the first hypothesis. On the other hand, the search for yield is reflected by the second hypothesis: higher sovereign yield spread incentivizes banks to increase their sovereign exposure. In addition, it is expected that the influence of the above mentioned variables on government exposure is stronger for stressed countries, as can be seen in the third hypothesis. Although this is not an official statistical hypothesis, it clarifies what is expected in the regression output discussed in the results section. Below, all hypothesis are displayed and the beta’s correspond to the beta’s in the methodology section where the regression models are introduced.
1. 𝐻0: 𝛽1= 0 𝐻1: 𝛽1< 0 2. 𝐻0: 𝛽2= 0 𝐻1: 𝛽2 > 0
3. 𝐼𝑡 𝑖𝑠 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑡ℎ𝑎𝑡 𝛽7> 𝛽1 𝑎𝑛𝑑 𝛽8 > 𝛽2 𝑖𝑛 𝑚𝑜𝑑𝑒𝑙 5 𝑎𝑛𝑑 6
Taking into consideration the macroeconomic fundamentals and home bias paragraphs, a fourth hypothesis can be added stating that higher government debt also leads to more government exposure. Moreover, a fifth hypotheses is formed to measure if sovereign exposure grew more
13
during the EDC. Finally, also for government debt it is expected that the effect on sovereign exposure is amplified by stressed countries which is measured by the last hypothesis.
4. 𝐻0: 𝛽3= 0 𝐻1: 𝛽3 > 0 5. 𝐻0: 𝛽5= 0 𝐻1: 𝛽5 > 0
6. 𝐼𝑡 𝑖𝑠 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑡ℎ𝑎𝑡 𝛽9> 𝛽3 𝑖𝑛 𝑚𝑜𝑑𝑒𝑙 5 𝑎𝑛𝑑 6
3. Data
In this section the data description, data sources and descriptive statistics are outlined. First, the dependent variables are defined and then the independent and control variables are explained. Before this, a few general remarks are made about the data. First of all, the data used in this thesis is aggregated balance sheet data of banks per country, originating from the ECB Statistical Data
Warehouse. In addition, Eurostat is used as data source for the country data such as the GDP and government debt. All data is monthly between January 2009 and December 2017. The balance sheet data is used of Monetary Financial Institutions (MFIs) excluding the ESCB (European System of Central Banks). Moreover, the MFI definition of the ECB, which is in line with the definition in EU law, includes credit institutions and money market funds (MMFs) but the number of MMFs is negligible (ECB, 2018). The European Systemic Risk Board (ESRB, 2015) used the same ECB data of MFIs in their analysis.
Secondly, the country data retrieved from Eurostat (government bond yields and GDP) were only quarterly available. To transform these quarterly figures into monthly figures the commonly used cubic spline interpolation method is applied using Matlab (Columbia, 2018). For this method quarterly data until December 2017 is used but due to the specific way of calculation monthly data of November and December 2017 could not be calculated. Hence, regression are performed on data between January 2009 and October 2017.
Dependent variables
The dependent variable is composed of two measurements: domestic exposure and lending composition. Domestic exposure is measured as total domestic sovereign bond holdings divided by total assets. Lending composition is measured as total domestic sovereign debt holdings divided by total lending to the real economy (non-MFIs and non-governmental organizations). Both variables are seen as domestic sovereign exposure measurements. The second dependent variable is included for two reasons. First of all, the amount of assets in general has been increasing over the researched
14
time period which could influence the domestic exposure measurement. Many researches found crowding out effects of increasing sovereign exposure, meaning that when domestic sovereign bond holdings rise, lending to the real economy declines (Acharya & Steffen, 2014; Altavilla et al., 2017). These portfolio changes are captured in the alternative dependent variable lending composition. Using two ways to calculate sovereign exposure is new in this literature. Note that significant effects on this dependent variable could also imply crowding out, though it is not in the scope of this thesis. Finally, the total sovereign debt holdings include bonds of both general as local and state governments (Buch et al., 2016). The total lending to the real economy consists of loans to non-financial corporations and households and non-profit organizations supporting households. Descriptive statistics of the dependent variables can be found in the table below.
Table 1: Descriptive statistics dependent variables
Variable Observations Mean Std. Dev. Min Max
Domestic exposure 1696 0.047708 0.044824 0.00024 0.22025 Lending composition 1696 0.115879 0.084504 0.00195 0.41961 Independent variables
Descriptive statistics of the independent and control variables can be found in table 2. The first independent variable is capitalization. Capitalization is hard to measure at a country aggregated level since it involves the capital adequacy ratio which requires balance sheet data of individual banks. To still capture the effect of capitalization a proxy is used, calculated by dividing the capital and reserves by total assets. Acharya and Steffen (2014) also used a leverage ratio to measure capitalization. The second independent variable is the sovereign yield spread. As in many other articles the yield spread is calculated relative to the 10-year German Bund, which is seen as the bond closest to risk free (Bernoth & Erdogan, 2011; Maltiz, 2011). This means that the yield spread for Finland is negative in a few months because the yield on Finnish bonds was lower than the yield on German bonds which explains the negative minimum value for this variable in table 2.
Thirdly, government debt is used as an explanatory variable. In line with Buch et al. (2016), government debt is measured as a percentage of GDP. Since this data was only quarterly available, the cubic spline interpolation method is used to create monthly data.
The fourth variable measures market size as total loans provided to the real economy divided by GDP. According to Gennaioli et al. (2014), this proxy can control for the size of the
15
in contrast with other research. Because GDP was only quarterly available, the cubic spline
interpolation is applied again to transform this quarterly calculated ratio into a monthly ratio. Naceur and Omran (2010) also use this proxy to measure the importance of bank financing in the economy. Finally, two dummy variables are added. The first dummy variable is the stressed countries variable. Altavilla et al. (2017), classify a country as stressed when the yield on government bonds exceeds 6% or the sovereign yield spread exceeds 4% in the period 2009-2017. This dummy variable displays 1 if a country is perceived as stressed. The other dummy variable is the crisis dummy, which is 1 between January 2009 and December 2013: the period defined as the EDC (Alsakka, ap Gwilym, & Vu, 2016; Acharya & Steffen, 2014). Between January 2014 and October 2017 the dummy is 0.
Table 2: Descriptive statistics independent variables
Variable Observations Mean Std. Dev. Min Max
Gov. debt 1696 0.83170 0.35092 0.148 1.809 Capitalization 1696 0.09212 0.04535 0.03681 0.25875 Market size 1696 4.8143 1.9913 1.7754 12.5252 Yield spread 1696 0.019123 0.028201 -0.0028 0.2739 Stressed 1696 - - 0 1 Crisis 1696 - - 0 1 4. Methodology
In this section, the method used to conduct this research is explained. The different regression models are outlined and clarified. To answer the central question: “What are the determinants of the amount of banks’ domestic sovereign debt holdings in the Eurozone during and after the European debt crisis?” multiple panel data regressions are conducted. The first step is to find the best regression type for this dataset. Therefore, the Breusch and Pagan Lagrangian test is conducted to choose between the OLS regression and the random effects. Then the Hausman test is conducted to choose between the random effects and fixed effects regression (Cipollini et al., 2015). The results are stated below:
1. OLS – Random effects chibar2(01) = 42410.27
Prob > chibar2 = 0.0000
2. Random effects – Fixed effects chi2(8) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 7.19 Prob > chi2 = 0.5158
16
For the first test, the H0 states the OLS model is better and is can be seen above the p-value is below 1% so H0 can be rejected and the conclusion is that the random effects model should be preferred above the OLS regression. In the second test, H0 states that the differences in coefficients are not systemic. In other words, it is not necessary to use a fixed effects model and the random effects model will be a better fit. The p-value for this test is above 0.05 so the null hypothesis cannot be rejected and hence the random effects model should be used.
The remaining part of the methodology section explains the reasoning behind the different regressions that are performed. Only the main explanatory variables (capitalization, yield spread and government debt) are tested first for both domestic exposure as for lending composition. The extended baseline model, also shown in box 1, adds the amplifying dummy variables of stressed countries and the EDC. Also, the control variable market size is added.
(1) 𝐷𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒𝑖 = 𝛽0+ 𝛽1𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖+ 𝛽2𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑦𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖+ 𝛽3𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡𝑖 + 𝜀𝑖 (2) 𝐿𝑒𝑛𝑑𝑖𝑛𝑔 𝑐𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛𝑖 = 𝛽0+ 𝛽1𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖+ 𝛽2𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑦𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖+ 𝛽3𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡𝑖 + 𝜀𝑖 --- (3) 𝐷𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒𝑖 = 𝛽0+ 𝛽1𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖+ 𝛽2𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑦𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖+ 𝛽3𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡𝑖 + 𝛽4𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖+ 𝛽5𝐶𝑟𝑖𝑠𝑖𝑠𝑖+ 𝛽6𝑀𝑎𝑟𝑘𝑒𝑡 𝑠𝑖𝑧𝑒𝑖+ 𝜀𝑖 (4) 𝐿𝑒𝑛𝑑𝑖𝑛𝑔 𝑐𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛𝑖 = 𝛽0+ 𝛽1𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖+ 𝛽2𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑦𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖+ 𝛽3𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡𝑖 + 𝛽4𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖+ 𝛽5𝐶𝑟𝑖𝑠𝑖𝑠𝑖+ 𝛽6𝑀𝑎𝑟𝑘𝑒𝑡 𝑠𝑖𝑧𝑒𝑖+ 𝜀𝑖
17
As the hypothesis displayed at the end of the literature review reflect, it is also expected that being a stressed country could strengthen the effect on domestic exposure. To measure this,
multiple interaction variables are added to the extended baseline model. The interaction between all explanatory variables, capitalization, sovereign yield spread and government debt with stressed countries is researched. There is no interaction variable between market size and stressed countries since there is no economic intuition found in the literature to assume this. These regressions are displayed below in box 2. Finally, in the results section both robust as non-robust regressions are shown to control for heteroskedasticity.
5. Results
In this section the regression models introduced in the previous section are conducted and the results are discussed. Also, the hypotheses in the literature review are elaborated on. Notice that for clarity reasons the intercept is not included in the result tables, since it is not important for this thesis. In addition, be aware that both domestic exposure as portfolio composition measure domestic sovereign exposure.
(5) 𝐷𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒𝑖 = 𝛽0+ 𝛽1𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖+ 𝛽2𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑦𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖+ 𝛽3𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡𝑖+ 𝛽4𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖+ 𝛽5𝐶𝑟𝑖𝑠𝑖𝑠𝑖+ 𝛽6𝑀𝑎𝑟𝑘𝑒𝑡 𝑠𝑖𝑧𝑒𝑖+ 𝛽7𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖× 𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖 + 𝛽8𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑦𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖 × 𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖+ 𝛽9𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡𝑖 × 𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖+ 𝜀𝑖 (6) 𝐿𝑒𝑛𝑑𝑖𝑛𝑔 𝑐𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛𝑖 = 𝛽0+ 𝛽1𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖+ 𝛽2𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑦𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖+ 𝛽3𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡𝑖+ 𝛽4𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖+ 𝛽5𝐶𝑟𝑖𝑠𝑖𝑠𝑖+ 𝛽6𝑀𝑎𝑟𝑘𝑒𝑡 𝑠𝑖𝑧𝑒𝑖+ 𝛽7𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖× 𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖 + 𝛽8𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑦𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖 × 𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖+ 𝛽9𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡𝑖 × 𝑆𝑡𝑟𝑒𝑠𝑠𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖+ 𝜀𝑖
18
Table 3 – Regression output baseline model
The results of the baseline regression models are summarized in table 3. What strikes out first, is that there are some differences between the robust and non-robust model. Sovereign yield spread for example is not significant in both robust regressions, while it is significant at a 1% level in the non-robust regressions. This could indicate heteroskedasticity issues, which makes the results less reliable. Therefore, in the remaining part of the result section, only the robust regressions are summarized. Although the negative sign for sovereign yield spread is unexpected and goes against the first hypothesis, the coefficient is not significant. Model 5 and 6 however address the possibility of different dynamics for sovereign yield spread since the interaction variables are included. Secondly, the results regarding capitalization are in line with the first hypothesis for both dependent variables. The negative sign and significance shows that sovereign exposure measured both to total assets as to total loans to the real economy is higher for low-capitalized banks. Also, the significant results regarding government debt make it able to reject the H0 of the fourth hypothesis, in line with the deficit absorption hypothesis: increasing government debt is a positive determinant of sovereign exposure.
In table 4 the (only robust) results of baseline model extended with the dummies and control variables are summarized. In contrast with the second hypothesis, stating that growing sovereign yield spread should increase exposure, these coefficients are again not significant. Moreover, the
Variables Domestic exposure (1) Domestic exposure (1) Portfolio composition (2) Portfolio composition (2) Capitalization -0.197511*** (0.016816) -0.197511** (0.095139) -0.46868*** (0.042123) -0.46868* (0.273946)
Sovereign yield spread -0.067229***
(0.0200468) -0.067229 (0.079504) -0.306357*** (0.050376) -0.306357 (0.207506) Government debt 0.079486*** (0.010735) 0.079486*** (0.030661) 0.170773*** (0.009393) 0.170773** (0.076920) Observations 1696 1696 1696 1696 Overall R-squared 0.0003 0.0003 0.0051 0.0051 Within R-squared 0.2196 0.2196 0.1713 0.1713 Between R-squared 0.0063 0.0063 0.0002 0.0002
Robust No Yes No Yes
This table shows the results for regressions (1) and (2) in box 1. Note that the robust standard errors are given in parentheses. *** significant at 1%, ** significant at 5%, * significant at 10%. Also note that it is taken into account that Stata provides two-tailed p-values.
19
coefficients and significance of capitalization and government debt remain more or less the same. Both dummy variables, stressed countries and crisis, seem to have no significant effects on sovereign exposure and portfolio composition. The latter implies that sovereign exposure did not increase more during the financial turmoil in the EDC. The complete model will show to what extent this is true. Finally, it seems that the sovereign exposure decreases in size of the domestic credit market. The intuition behind this could be that for big domestic financial markets, the difference in incentive for banks to choose sovereign bond holdings over loans to the real economy is not that big and maybe even mitigated. Since Naceur & Oman (2010) use this variable as a proxy for the importance of bank financing it could also imply that demand of the real economy is higher and hence banks need to issue more loans to the real economy in their portfolio.
Variables Domestic exposure (3) Portfolio composition (4)
Capitalization -0.185934**
(0.089814)
-0.417214* (0.251155)
Sovereign yield spread 0.048919
(0.050925) 0.034598 (0.116042) Government debt 0.071695*** (0.021873) 0.151879*** (0.051847) Stressed countries 0.008204 (0.019575) 0.017913 (0.035988) Crisis 0.001812 (0.005739) 0.011043 (0.013319) Market size -0.008269** (0.003967) -0.027184*** (0.007788) Observations 1696 1696 Overall R-squared 0.0536 0.1540 Within R-squared 0.3285 0.3667 Between R-squared 0.0354 0.1302
Robust Yes Yes
This table shows the results for regressions (3) and (4) in box 1. Note that the robust standard errors are given in parentheses. *** significant at 1%, ** significant at 5%, * significant at 10%. Also note that it is taken into account that Stata provides two-tailed p-values.
20
Most importantly, table 5 gives an overview of the results of the complete regression model. First of all, although the coefficient for capitalization is still significant in the domestic exposure model, its effect is completely adopted by the interaction variable with stressed countries in the portfolio model. On the one hand, the fifth regression implies that low capitalized banks in both stressed as in non-stressed countries are more exposed to sovereign bonds. On the other hand, the sixth regression shows only significant effects for low-capitalized banks in stressed countries. It might be that crowding out effects in stressed countries are higher than in non-stressed countries and hence the portfolio composition of low-capitalized banks in non-stressed counties does not change that much. So both models support the third hypothesis and partly the reasoning behind the carry trade hypothesis, concluding that low capitalized banks in stressed countries are stronger drivers of sovereign exposure than in non-stressed countries and it cannot be ruled out that low capitalized banks in non-stressed countries also have more domestic sovereign exposure.
Secondly, the sovereign yield spread shows, in contrast with the first four regressions, significant positive effects on domestic exposure and portfolio composition in line with the second hypothesis. However, counterintuitively the sign of the interaction variable between sovereign yield spread and stressed countries switches to a negative sign. This goes against the third hypothesis based on carry trade (and moral suasion) that argued that banks in stressed countries take part in a search for yield. A possible explanation is that the yield spread was very volatile during the crisis. In the second half of the crisis, yields already went down and recall from table 2 that most stressed countries increased domestic sovereign exposure during the whole crisis period, which could lead to this negative sign. In addition, the regression also contains the post crisis period and figure 2 showed that all stressed countries (except for Cyprus) decreased their home bias. Both trends could explain the negative sign of the interaction variable.
Moreover, both regressions in table 5 find significant effects for the search for yield in non-stressed countries which is in contrast with the findings of Buch et al. (2016). On the other hand, it pertains on findings of Battistini et al. (2014). This means that according to these results, the search for yield of banks in non-stressed countries is present. All conclusions regarding the sovereign yield spread should however be interpreted with caution, since the switch of sign could indicate
21
Thirdly, the regressions in table 5 show that the interaction effect between stressed countries and government debt took over the significance of the government debt variable in the first 4 specifications. By proving that higher government debt only for stressed countries induces higher government exposure it both supports the sixth and denies the third hypothesis partly. In other words, these results imply that the deficit absorption hypothesis only applies for stressed countries. This could also point to moral suasion or financial repression but there is more detailed
Variables Domestic exposure (5) Portfolio composition (6)
Capitalization -0.058039**
(0.029569)
-0.025876 (0.097278)
Sovereign yield spread 1.29167***
(0.409384) 3.31135*** (0.557573) Government debt -0.002149 (0.022312) 0.005620 (0.083153) Stressed countries -0.029441 (0.028631) -0.039151 (0.076891)
Stressed countries * Capitalization -0.309731***
(0.110202)
-0.897538*** (0.224475)
Stressed countries * Sovereign yield spread -1.32110***
(0.419419)
-3.481374*** (0.604197)
Stressed countries * Government debt 0.098702***
(0.031647) 0.210294** (0.100328) Crisis -0.005145 (0.004303) -0.006846 (0.010493) Market size -0.006268** (0.003043) -0.022585*** (0.005692) Observations 1696 1696 Overall R-squared 0.1447 0.2676 Within R-squared 0.4335 0.4894 Between R-squared 0.1127 0.2320
Robust Yes Yes
This table shows the results for regressions (5) and (6) in box 2. Note that the robust standard errors are given in parentheses. *** significant at 1%, ** significant at 5%, * significant at 10%. Also note that it is taken into account that Stata provides two-tailed p-values.
22
research needed about demand for government debt to confirm that.
The market size variable behaves the same as in model 3 and 4: negative small significant coefficients meaning less domestic sovereign exposure for bigger or stronger credit markets. In addition, for all regressions the within R-squared is bigger than the between R-squared. This means that most of the variation in the dependent variables is caused by differences within countries. This is not surprising since this analysis entails a turbulent crisis period with volatile macroeconomic
circumstances. In addition, in the last 4 specifications, the overall R-squared is bigger for portfolio composition regressions than for regressions with domestic exposure as dependent variable. The explanatory and control variables in these regressions hence explain more of the variance in portfolio composition than in sovereign exposure. Finally, it is not surprising that the overall R-squared
remains relatively low since the increased sovereign exposure is partly caused by European capital regulations as explained in the literature review. The effects of these regulations are not captured in the regression specifications.
5.1 Policy implications
Gennaioli et al. (2014) stress the importance of breaking the sovereign bank nexus to prevent spillover during crisis. Knowing that the asset holdings channel is the most important channel to transfer the risk from the sovereign to the banking sector, the determinants of sovereign bond holdings could play a key role in breaking the sovereign bank nexus. The results discussed above provide policy makers with the knowledge where to focus. First of all, this thesis identifies
government debt as an important driver of sovereign exposure in stressed countries. Secondly, low capitalized banks are an important cause of increasing sovereign exposure in the banking sector, especially in stressed countries but also to a lesser extent in non-stressed countries. In addition, the relatively low explanatory power of these regressions hint towards the key role of European policy incentivizing banks to hold sovereign debt. Combining this knowledge, policy makers are able to tackle two problems with one measure: assign mandatory risk weight to sovereign bonds, without exception rules. This will make sovereign bond holdings for banks less attractive compared to other loans, decreasing sovereign exposure on the one hand . On the other hand, banks are automatically obliged by existing regulation to hold more tier 1 capital against these risk weighted bond holdings which improves their capital position. Finally, although this thesis in contrast with other literature does not find sufficient evidence for the search for yield or banks, this increased risk weight discourage banks to perceive high-yield sovereign bonds as an investment without downside. Capturing these upsides does not even require creating new policy, only cancelling exceptional EU regulation allowing banks to deviate from the non-zero risk weight for sovereign bond holdings.
23
6. Discussion and conclusion
The EDC highlighted the need to understand the underexposed determinants of sovereign exposure. This thesis complements the asset channel literature by shedding light on these determinants with multiple panel data regressions on country aggregated level, reflecting on a broad scope of Euro countries and variables. Both the regulatory as the macroeconomic drivers are outlined in this thesis. The analysis show that low capitalized banks in non-stressed but mainly in stressed countries are an important determinant of sovereign exposure, partly supporting the carry trade hypothesis. In addition, higher government debt in stressed countries increased sovereign debt holdings. In contrast with other literature, this thesis did find evidence for the search for yield in non-stressed countries but contradicting the carry trade hypothesis not in stressed countries. This result however needs more research. Finally, a large domestic credit market decreases sovereign exposure.
To deteriorate the sovereign bank nexus and thereby preventing spillovers to the banking sector during another financial (sovereign) crisis, policy makers need to prevent the possibility for banks to assign a zero risk weight to sovereign debt holdings. This will be beneficial in two ways. First of all, sovereign bonds are made less attractive for banks which decreases sovereign exposure and increases lending to the real economy. Secondly, the capital position of banks improves since they need to hold more capital against non-zero risk weighted sovereign bonds .
This thesis concludes with suggestions for other research. The contrasting results regarding the sovereign yield spread could possibly be caused by the fact that the regression entails the final crisis period and post crisis period, where yields were already rising. This could have mitigated the effects of the regression. Therefore, in other research separate country individual regressions should be conducted both during and after the crisis, to gain more detailed knowledge about the variations in determinants. Moreover, similar analysis should be performed on individual balance sheet data of banks to extend the knowledge on bank level determinants of sovereign exposure. Finally, research on a more sophisticated breakdown of a banks’ sovereign bond holdings per country could give more insight in the so far contradicting behaviour of non-stressed country banks regarding exposure towards stressed countries.
24
7. Reference list
Acharya, V. V., & Steffen, S. (2015). The “greatest” carry trade ever? Understanding eurozone bank risks. Journal of Financial Economics, 115(2), 215-236.
Acharya, V., Drechsler, I., & Schnabl, P. (2014). A pyrrhic victory? Bank bailouts and sovereign credit risk. The Journal of Finance, 69(6), 2689-2739.
Afonso, A., Furceri, D., & Gomes, P. (2012). Sovereign credit ratings and financial markets linkages: application to European data. Journal of International Money and Finance, 31(3), 606-638.
Alsakka, R., ap Gwilym, O., & Vu, T. N. (2014). The sovereign-bank rating channel and rating agencies' downgrades during the European debt crisis. Journal of International Money and Finance, 49, 235-257.
Alsakka, R., ap Gwilym, O., & Vu, H. (2017). Differences of opinion in sovereign credit signals during the European crisis. The European Journal of Finance, 23(10), 859-884.
Altavilla, C., Pagano, M., & Simonelli, S. (2017). Bank exposures and sovereign stress transmission. Review of Finance, 21(6), 2103-2139.
Angeloni, C., & Wolff, G. B. (2012). Are banks affected by their holdings of government debt? (No. 2012/07). Bruegel working paper.
BIS (2011). The impact of sovereign credit risk on bank funding conditions. Retrieved on June 26 from https://www.bis.org/publ/cgfs43.pdf
BIS (2013). International banking and financial market developments. Retrieved on June 26 from https://www.bis.org/publ/qtrpdf/r_qt1312.pdf
Becker, B., & Ivashina, V. (2014). Financial repression in the European sovereign debt crisis. Review of Finance: working paper.
Bernoth, K., & Erdogan, B. (2012). Sovereign bond yield spreads: A time-varying coefficient approach. Journal of International Money and Finance, 31(3), 639-656.
Bruyckere de, V., Gerhardt, M., Schepens, G., & Vander Vennet, R. (2013). Bank/sovereign risk spillovers in the European debt crisis. Journal of Banking & Finance, 37(12), 4793-4809.
Buch, C. M., Koetter, M., & Ohls, J. (2016). Banks and sovereign risk: A granular view. Journal of Financial Stability, 25, 1-15.
25
Castro, C., & Mencía, J. (2014). Sovereign risk and financial stability. Revista de Estabilidad Financiera, 26, 73-107.
Cipollini, A., Coakley, J., & Lee, H. (2015). The European sovereign debt market: from integration to segmentation. The European Journal of Finance, 21(2), 111-128.
Drago, D., & Gallo, R. (2017). The impact of sovereign rating changes on the activity of European banks. Journal of Banking & Finance, 85, 99-112.
European Systemic Risk Board (2015). Report on the Regulatory Treatment of Sovereign Debt. Retrieved on June 26 from
http://www.esrb.europa.eu/pub/pdf/other/esrbreportregulatorytreatmentsovereignexposures0320 15.en.pdf?c0cad80cf39a74e20d9d5947c7390df1
ECB (2018). List of financial institutions. Retrieved on June 26 from
https://www.ecb.europa.eu/stats/financial_corporations/list_of_financial_institutions/html/index.e n.html
Faia, E. (2017). Sovereign risk, bank funding and investors’ pessimism. Journal of Economic Dynamics and Control, 79, 79-96.
Gennaioli, N., Martin, A., & Rossi, S. (2014). Banks, government bonds, and default: What do the data say? International Monetary Fund: working paper.
Naceur, S. B., & Omran, M. (2011). The effects of bank regulations, competition, and financial reforms on banks' performance. Emerging markets review, 12(1), 1-20.
Pagano, M. (2016). The Sovereign-Bank Nexus and the Case for European Safe Bonds. European Economy, 76(1), 129.
Pockrandt, J., & Radde, S. (2012). Need for reform of EU banking: Decoupling the solvency of banks and sovereigns. DIW Economic Bulletin, 2(11), 11-18.
Podstawski, M., & Velinov, A. (2018). The state dependent impact of bank exposure on sovereign risk. Journal of Banking & Finance, 88, 63-75.
Popov, A., & Van Horen, N. (2013). The impact of sovereign debt exposure on bank lending: Evidence from the European debt crisis. De Nederlandsche Bank: working paper.
