• No results found

Bank riskiness and the slope of the yield curve: Empirical evidence from the Eurozone

N/A
N/A
Protected

Academic year: 2021

Share "Bank riskiness and the slope of the yield curve: Empirical evidence from the Eurozone"

Copied!
31
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

1

Bank riskiness and the slope of the yield curve:

Empirical evidence from the Eurozone

Thijs Scholte Albers

1

Master thesis MSc Finance and MSc International Economics & Business University of Groningen, The Netherlands

Abstract

This study examines the influence of the slope of the yield curve on domestic and foreign banks' riskiness by using a fixed effects model on a dataset of banks operating in the Euro area during the sample time period 2005-2014. This findings of this study are that there is strong evidence for a positive relation between the steepness of the slope of the yield curve and bank riskiness. However, this relation tends to be weaker for foreign banks. In addition, this study also provides (limited) evidence that due to intra-group funding flows the marginal effects of the slope of the yield curve on foreign bank riskiness becomes weaker.

Keywords: Bank riskiness, yield curve, domestic banks, foreign banks, internal capital markets, non-performing loans

JEL Classification: E43, E44, G21

“When the music stops, in terms of liquidity, things will be complicated, but as long as the

music’s playing, you’ve got to get up and dance. We’re still dancing.”

Charles Prince, former CEO Citi-group, Financial Times, July 2007

1.

Introduction

The pre-crisis period lending boom was characterised by rapidly growing credit expansion stimulated by the prosperous business cycle, financial innovation and central banks' monetary expansionary policies (Acharya & Richardson, 2009). While central banks purely focused on maintaining price stability in order to foster economic growth, the financial system became more unstable and fragile as banks increased their balance sheets and loosened lending standards (Angeloni et al., 2015). Eventually, excessive bank risk-taking fanned the flame to the global financial crisis which had first-order effects on worldwide financial and economic stability (Laeven & Levine, 2009). In light of these developments

1Student MSc Finance and MSc International Economics & Business, University of Groningen, student number:

1798561, e-mail: T.D.Scholte.Albers@student.rug.nl, address: Europaplein 117-1, 1079AX Amsterdam. Supervisors: dr. J.O. (Jochen) Mierau and dr M.J. (Michiel) Gerritse.

(2)

2 the central question to policymakers and economists remained: What contributed to the build-up of these financial imbalances?

A recent line of literature states that central banks' monetary policy actions have direct consequences for the financial stability through the existence of the so-called "bank risk-taking channel" of monetary policy. The risk-taking channel operates through two transmission channels. In the first channel, monetary expansion influences bank riskiness through a leverage accumulation on banks' liability side (Adrian & Shin, 2009). Short-term deposits become a cheap source of funding and favored over capital and other stable sources. This makes banks more vulnerable to bank runs and illiquidity (Diamond & Dybvig, 1983). The second channel, the so-called "search-for-yield" operates through banks' assets side (Rajan, 2005). Low interest rates decrease adverse selection problems, which increase both competition and credit expansion. As a result, banks loosen their lending standards and take on more risky projects in order to expand activities and increase their profit margins (Dell'Ariccia & Marquez, 2006).

Differences arise between domestically operating banks and foreign banks. The functioning of the internal capital market between global bank groups insulates foreign banks from host country monetary actions (e.g. Wu et al., 2011; Cetorelli & Goldberg, 2012; De Haas & Lelyveld, 2014). Furthermore, according to the diversification theory foreign banks operate under less bank risk exposure as their operations are diversified through cross-border subsidiaries (Laeven & Levine, 2007). In addition, due to intra-group funding flows international banks operate as a group and are not constrained to local business and monetary conditions, and are therefore better able to stick to the most efficient risk-return investments (Morgan et al., 2004).

Prior research to the bank risk-taking channel of monetary policy focuses on the level of short-term interest rates (Jiménez et al., 2009; Ioannidou et al., 2009; Delis & Kouretas, 2011; Maddaloni &

(3)

3 In this thesis I investigate bank riskiness, in line with the second strand of literature, by focusing on the relation between the slope of the yield curve and ex-post bank credit risk exposure.23 As the slope of the yield curve represents the spread between long-term and short-term interest rates it is a leading indicator of the business cycle (Ang et al., 2006). In addition, it is an important driver of a bank's profitability due to its reliance on maturity transformation (Gambacorta, 2009; Mink, 2011). Furthermore, I focus on differences between domestic and foreign banks and open a new field of research by investigating the direct influence of internal capital flows on foreign bank riskiness. The research topics correspond with the following central research question:

"What is the relation between the slope of the yield curve and bank riskiness of domestic and foreign banks?"

Using a fixed effects model on a large panel dataset including macroeconomic variables, intra-group funding flows and annual balance sheet information and credit risk measures, I find empirically significant evidence of a positive relation between the slope of the yield curve and bank riskiness. In addition I provide (limited) evidence that this relation is less pronounced for foreign banks, due to the dampening effect of the internal capital markets on bank riskiness.

The remainder of this thesis is organized as follows. Section 2 discusses the related relevant theory behind bank risk-taking, empirical evidence for the existence of the risk-taking channel of monetary policy and ends with the developed research hypotheses in order to test the validity of the theory. Section 3 describes the dataset and descriptive characteristics. In section 4 the econometric model is described, followed by section 5 wherein the main results will be presented. Finally, in section 6 the conclusions of this paper are presented.

2.

Theoretical framework

As described in the introduction the main focus of this thesis is on the relation between the slope of the yield curve and bank riskiness. In this section I will discuss the most relevant theories behind bank risk exposure, the slope of the yield curve and internal capital markets. I will start by explaining the underlying theory of the bank risk-taking channel, followed by prior empirical evidence

2 This thesis empirically investigates the relation between bank riskiness and the slope of the yield curve and is

therefore closely related to earlier work regarding bank riskiness and the related risk-taking channel. However, it is important to stress that balance sheet data is limited as it only provides insights in the level of bank (credit) risk exposure in relation to the slope of the yield curve. This data source does provide further insights whether an increase in bank riskiness is due to actual changes in risk-taking behaviour or driven by changes in assets values or other external circumstances.

3

(4)

4 for the existence of the bank risk-taking channel of monetary policy. Finally, I describe the importance and hypothetical link between the slope of the yield curve, intra-group flows and bank riskiness.

2.1 Bank risk-taking channel of monetary policy

Prior to the global financial crisis macroeconomic and monetary literature mainly focused on the classical interest rate channel which describes the impact of monetary policy on the real economy through investment decisions of firms and household consumption.4 Monetary expansion directly affects investment decisions, and therefore firms' demand for credit (Friedman & Schwarz, 2008). However, financial and credit conditions also play an important role in the business cycle. Due to frictions in the credit market monetary policy can influence the ability and incentives of banks to provide credit. Through the so-called "credit channel" the effect of real interest rate changes on borrowers, banks’ balance sheets and the credit supply is amplified by information asymmetry (Miskhin, 1996). The monetary transmission of the credit channel can be broken down into two conduits of policy influence: the balance-sheet channel and the bank-lending channel (Bernanke & Gertler, 1995).5

The credit channel describes the nexus between the central banks' monetary policy and loan supply. It does not address the quality and riskiness of the bank lending activities. Borio and Zhu (2008) opened this black box and described a third monetary transmission mechanism, the bank risk-taking channel. This mechanism describes the impact of monetary policy actions on bank risk-taking and is defined by Borio and Zhu (2008) as:

"The impact of changes in policy rates on either risk perceptions or risk-tolerance and hence on the degree of risk in the portfolios, on the pricing of assets, and on the price and non-price terms of the extension of funding".

In contrast to the credit channel the bank-risk taking channel does actually focus on the quality and riskiness of bank lending activities. Given our current financial and regulatory system a clear understanding of this mechanism is becoming more important for policy makers.

The risk-taking channel can be broken down into two different transmission mechanisms through which it operates: (i) bank risk is intensified through increasing balance sheets as result of accumulation of leverage, and (ii) the link between low levels of interest rates and return.

In the first channel the accumulation of leverage through the banks' liability side plays an important role. Banks increase their balance sheets through leverage during times of favorable financing

4 e.g. Romer & Romer (1989); Bernanke & Blinder (1992); and Christiano et al. (1994).

5 The balance-sheet channel refers to the intensified impact of short-term interest rate changes on borrowers'

(5)

5 conditions (Adrian & Shin, 2009). A loosened monetary policy influences the funding source on the liability side, thereby affecting the structure of a bank's liabilities. As short-term funding becomes a cheaper option banks will favor this source of funding over capital and other stable funding sources. In this regard, banks' lending capacities are increased by loosened monetary policy as it influences the present value of income, resulting in more favorable forward-looking capital measures (Adrian & Shin, 2010). However, the greater reliance on (risky) short-term deposit funding make banks more vulnerable to bank runs and sudden illiquidity (Diamond & Dybvig, 1983).

The second channel the so-called "search-for-yield" operates through increased riskiness in the banks' asset side. Lower interest rates mitigate adverse selection problems which leads to higher competition and credit expansion. Consequently, the asset-side composition is altered towards a riskier mix by relaxing their lending standards (Rajan, 2005). Hence, as result of a fading separating equilibrium between good and bad borrowers, mispricing risk will increase (Dell'Ariccia & Marquez, 2006). This would result in an increased amount of new loans to (riskier) borrowers.

2.3 Empirical evidence

In monetary economics and finance studies the bank risk-taking channel is a relatively new topic. Only a limited amount of research empirically tested the existence of the bank risk-taking channel. Most empirical literature focuses on the level of short-term interest rates which stem directly from theoretical proposition of Dell'Ariccia and Marquez (2006), Rajan (2006), Borio and Zhu (2008) and Dell'Ariccia et al. (2014). Prior research can be divided into two different strands of research. One strand of research focuses on the quality and riskiness of loans provided through information from lending standard surveys and credit information from financial institutions. Hereby, this strand directly investigates banks' changing behaviour towards risk. The other strand of research focuses on ex-post bank riskiness using market based or balance sheet risk indicators. This thesis will be related to the second strand of research by focussing on the ex-post riskiness of banks using balance sheet risk indicators.

(6)

6 theoretical composition (Dell'Ariccia & Marquez, 2006). These results are also in line with the outcomes of Maddaloni and Peydró (2011). They use lending standards of both US banks and banks in the Euro area from 2002-2008 and provide evidence that low short-term interest rates relax lending standards. For long-term interest rates no such evidence is found.

Other studies focus on ex-post exposure bank riskiness and follow a different approach using bank level data. Altunbas et al. (2010) use a benchmark rate and a market based indicator, the expected default frequency (EDF),6 and provide evidence for an increase in bank riskiness as the short-term interest rate falls below the benchmark rate.7 Using a similar approach Gambacorta (2009) provides evidence that bank riskiness is also a function of the central banks' time-span of monetary easing policies. Banks' credit risk exposure increases when the interest rates fall below the benchmark rate for a longer period. Ozsuca and Akbostanci (2012) test the influence of interest rates on the Turkish banking sector following the EDF approach and several accounting based risk indicators. Their results provide significant evidence for different bank risk measures. Delis and Kouretas (2011) assess the existence of the relation between short-term interest rates and bank riskiness in Europe over the period 2001-2008. The ratio of risky assets to total assets and the ratio of non-performing loans (NPLs) to total loans are defined as indicators of bank risk. Their study also provides a significant negative association between bank riskiness and interest rates.

Given that the bank risk-taking channel is seen as a relatively new research topic there are still many more unaddressed questions. The above described empirical evidence solely focuses on bank riskiness in relation to monetary policy actions whereas I am interested in the relation of bank riskiness (credit risk exposure) and the spread between long-term and short-term interest rates, also known as the slope of the yield curve.

2.4 Traditional banking, the slope of the yield curve & hypotheses

Banks fulfil a vital role in the economy by ensuring money circulation in the economy. Acting as financial intermediaries banks mitigate information asymmetry problems due to their expertise and superior screening techniques and channel funds from i.e. private savers to investors looking for large sums of financial resources (Leland & Pyle, 1977). This traditional form of banking relies on the concept of maturity transformation as bank assets (loans offered) tend to be of longer maturity than the liabilities (deposits) of these funds. Due to this maturity mismatch banks rely on the spread between the long-term and short-term interest rates and are thus exposed to repricing and yield curve risk (Alessandri & Nelson, 2015).

6 EDF is a forward-looking market-based indicator of credit risk. It measures the expected probability of default of

banks within a one-year time frame.

7

(7)

7 Given this vital link between a bank's profitability and the yield curve, it is interesting to research how underlying changes impact bank riskiness. Where central banks' monetary policy actions serves as an important underlying assumption, it only presents half of the story given that it mostly reflects the funding side of banks. The yield curve directly reflects the interest margin, and thus a bank's profitability (Gambacorta, 2009). Furthermore, the yield curve is not only influenced by monetary policy actions, but also influenced by macroeconomic conditions (e.g. Afanasieff et al., 2002).

The theoretical proposition of Mink (2011) is strongly related to the main research purpose of this thesis. Mink's (2011) model provides a theoretical connection between micro-level bank riskiness and the macro-level stance of the economy via the slope of the yield curve. The model shows that bank riskiness increases during macroeconomic favourable periods, which is reflected by a steeper slope of the yield curve.8 A steeper yield curves implies greater profitability which gives banks increased incentive to engage more maturity transformation through leverage and softening of lending standards. This is in line with earlier work of i.e. Adrian & Shin (2009) who state that banks increase their balance sheets through leverage during times of favorable financing conditions, and Dell'Ariccia & Marquez (2006) who provide the link between softening lending standards and bank risk-taking.

Empirical evidence of Alessandri & Nelson (2015) shows that banks' profitability through maturity matching activities and the slope of the yield curve are positively related. Furthermore, their data suggest that in order to prevent a fall in return on assets, banks increased their leverage. In line with this evidence, Gambacorta (2009) shows that a steeper yield curve increases bank profits because of the maturity transformation function. Besides that, their results show a positive, but insignificant relation between the slope of the yield curve and bank riskiness proxied by the EDF. Maddaloni and Peydró (2011) find evidence for the existence of a relation between bank risk-taking and the slope of the yield curve by using Euro Area Bank Lending Survey answers. In their approach, they measure banks' (changing) behaviour towards loosening lending standards. However, this does not directly indicate whether the actual level of banks' riskiness increases.

To fill this gap I want to measure the relation between the slope of the yield curve and bank riskiness. Based on the related theory of the risk-taking channel, the direct link between the slope of the yield curve and the reliance of banks on maturity transformation, Mink's (2011) theoretical proposition and empirical evidence of e.g. Maddaloni and Peydró (2011), there is a strong theoretical and empirical foundation to investigate the relation between the slope of the yield curve and bank riskiness. Based on this I come to the following research hypothesis:

HYPOTHESIS I: There is a positive relation between the slope of the yield curve and the level of bank riskiness

(8)

8 Foreign banks played a major role in the transmission of contagion and financial instability (Jeon et al. 2014). Given the special setting in the Eurozone where the ECB sets the monetary policy, while the individual member states are responsible for bank supervision, it is important for policymakers to understand how bank risk differs between domestic and foreign banks. Therefore, it would be relevant to investigate potential differences between domestic and foreign banks with regard to hypothesis I.

The diversification theory explains how international banks operate under lower levels of bank risk in comparison to domestic banks (Cetorelli & Goldberg, 2012). Through foreign affiliates international banks diversify their business (Laeven & Levine, 2007). In addition, the holding company can efficiently manage liquidity and allocate capital among the group using internal capital markets (De Haas & Lelyveld, 2010).9 These internal capital markets are particularly active in financially integrated areas such as the EU (Navaretti et al., 2010). Hence, subsidiaries would be less reliant to fluctuations in their own internally generated funds and capital (e.g. Desai et al., 2004; Wulf, 2009; Hovakimian, 2011).

Foreign banks are less impacted by the bank lending channel of monetary transmission mechanism as intra-group capital flows buffer the effect of host country monetary actions while domestic banks are directly impacted (e.g. Wu et al., 2010; Cetorelli & Goldberg, 2012). This mainly arises due to the structural differences in the stability of the funding side of their lending activities, which causes foreign and domestic banks' lending activities to develop in a significantly different matter (e.g. de Haas & Lelyveld, 2014). Foreign banks rely for a great extent on the financial liquidity management of their (global) banking conglomerate operating through their internal capital markets (Reinhardt & Riddiough, 2014). In contrast, domestic banks rely to a great extent on more instable procyclical funding from external sources such as local and inter-bank funding, international capital markets (if sizeable enough) and their own internally generated funds. Therefore, domestic bank loans expansions is procyclical given its direct link with the business cycle, whereas foreign bank loan expansion is more stable through the business cycles (Dahl et al., 2002).

Morgan et al.'s (2004) model states that intra-group capital is reallocated fulfilling a "support" and "substitution" function in order to ensure overall stable return on capital among countries.

According to the so-called "support effect" intra-group funds flow to less profitable subsidiaries that generate less internal funds. This is in line with Jeon & Wu (2014) who find evidence that intra-group flows increases to these banks during economic down turn as their liquidity positions weaken and internally generated funds decline. Furthermore, the "substitution effect" states that excessive capital is reallocated to subsidiaries in countries with favorable conditions, in order to maintain the most efficient risk return trade-off. Therefore, global parent banks have the power to shift liquidity where it is most efficiently needed (Cetorelli & Goldberg, 2012).

9

(9)

9 Due to the functioning of internal capital markets, I expect foreign banks' riskiness to be less affected by a change in the slope of the yield curve than domestically operating banks. As the slope steepens, banks become more profitable and generate more internal funds open for further investments. Domestic banks are constrained to the local business conditions and try to maximize their activity during prosperous cycles (de Haas & Lelyveld, 2010). This search-for-yield results in more exposure to uncompensated credit risk (Borio & Zhu, 2011). However, international banks are not bounded to local business conditions and can simply reallocate excessive funds to investments in other countries in order to maintain optimal risk-return trade off and avoid an (unnecessary) increase in credit risk exposure (Jeon et al., 2012). This suggests that excessive internally generated funds will be reallocated as the slope steepens. Thereby avoiding mispricing risk and inefficient risk-return tradeoffs, and hence less increase in bank riskiness. In order to empirically test this I will use the following hypothesis:

HYPOTHESIS II: The positive relation between bank riskiness and the slope of the yield curve is of greater magnitude for domestic banks than for international banks

Furthermore, in order to explain this dampening effect for foreign banks economically, I directly investigate the influence of the internal capital markets on foreign bank riskiness using the following hypothesis:

HYPOTHESIS III: The positive relation between bank riskiness and the slope of the yield curve is negatively influenced by intra-group funding flows

3.

Data & descriptives

In this section I provide an overview of the dependent variables used to proxy banks' exposure to credit risk, the independent variables and control variables used in the econometric estimation in order to test the in section 2 presented hypotheses. Furthermore, the descriptive statistics are discussed at the end of this section.

3.1 Data overview

(10)

10 Latvia and Lithuania are restricted from the sample given the fact that these countries joined the Euro area in 2014 and 2015 respectively. In addition, for Estonia no sovereign bond data is available, and therefore it is not included in the estimation. Furthermore, member states Cyprus, Malta, Slovakia and Slovenia are partly restricted from the sample as these countries did not use the Euro over the whole sample period. Cyprus, Malta, Slovakia and Slovenia joined the Euro area in 2008, 2008, 2007 and 2009 respectively. The dataset only includes yearly bank-level data collected from the balance sheet and income statements to proxy bank riskiness from unconsolidated accounts where available and otherwise the consolidated accounts (leaving out unconsolidated company accounts) in order to avoid double accounting. The sample period is restricted to the years 2005-2014 as the bank-level data is very limited over a longer time period. Following i.a. Delis & Kouretas (2011) banks with missing data values are not excluded from the sample, and an outlier rule is applied to the main variables in order to drop extreme values.

3.2.1 Dependent variables: Non-performing loans & z-Score

In literature two different types of bank risk measures are used. Indicators that rely on lending surveys and loan data. These measures give direct insight in the actual risk-taking behaviour towards loan quality and lending standards. Other proxies are constructed with balance sheet and income statement data in order to measure the level of bank riskiness. In this thesis I focus on the latter by using accounting data given the limited time available. In order to measure banks riskiness two different accounting based credit risk proxies are used, non-performing loans and z-Score.10

The first credit risk proxy which will be used for the main estimations is, Non-performing loans (NPLS). NPLS, is the ratio non-performing loans to gross loans and is a widely used credit risk measure (e.g. Jimenez et al. 2007; Delis & Kouretas, 2011). Non-performing loans are defined as loans on which the borrower is not repaying the principal or not making interest payments and therefore close to default or in default. They include doubtful loans and loans 90 days overdue. Given this, NPLS, is a measure of banks' loan quality outstanding and reflects the exposure to earnings and is therefore a direct proxy for credit risk. High NPLS values indicate a great amount of non-performing loans in banks' portfolio and therefore a high level of credit risk exposure.

The second proxy, the z-Score (ZSCORE), has also been frequently used as measure for individual bank fragility and is related to NPLS (e.g. Laeven & Levine, 2009; Altunbas et al., 2010; Maddaloni & Peydró, 2011; Drakos et al., 2014). I will use ZSCORE as the dependent variable for robustness purposes. It is calculated as follows:

𝑧-𝑆𝑐𝑜𝑟𝑒𝑖=

𝑅𝑂𝐴𝑖+ 𝐸/𝑇𝐴𝑖

𝜎(𝑅𝑂𝐴𝑖)

(1)

(11)

11 where 𝑅𝑂𝐴𝑖 represents the return on assets of bank i, 𝐸/𝑇𝐴𝑖 is the ratio of equity to total assets of bank

i and 𝜎(𝑅𝑂𝐴𝑖) represents the standard deviation of the return on assets of bank i. Following e.g. Drakos

et al. (2014), in order to count for changes in return volatility, the standard deviation of the return on assets is calculated over a 3-year rolling time window. The ZSCORE changes due to movements in earnings stability, profits or capital level. It can therefore be seen as a proxy for bank riskiness, whereas a lower score indicates less stability and thus higher bank risk.

3.2.2 Independent variables: Slope of the yield curve, intra-bank funding flows and foreign dummy

Main dependent variable is the macroeconomic SLOPE, representing the slope of the yield curve. SLOPE, is calculated as the difference between the long-term and short-term interest rates. Whereas a steeper slope represents a wider spread between the long-term and short-term interest rates. All interest rates are annual averages which are calculated of their monthly relevant time period.11 The short-term rate is proxied by the 3-month Euro Interbank Offer Rate (EURIBOR) which is widely used in literature (e.g. Maddaloni & Peydró, 2009; Altunbas et al., 2010; Kouretas & Delis, 2011). Besides the EURIBOR, the Euro Overnight Index Average (EONIA) is used to test the robustness of the results. In the Euro area, the EONIA can float within a predetermined corridor set by the ECB. Therefore, EONIA represents a useful variable for the monetary policy stance in the Eurozone and hence to test for robustness (see e.g. Maddaloni & Peydró, 2009). Data for both the EURIBOR and the EONIA rates are obtained from Bloomberg.

The most commonly used debt instrument for the long-term interest rates are individual member states' 10-year government bonds (e.g. Maddaloni & Peydró, 2009; Altunbas et al., 2010; Kouretas & Delis, 2011). The annual averages calculated over monthly time series are used. The 10-year government bond rates for most countries are obtained from Bloomberg. However, as the long-term rates for Cyprus, Luxembourg, Malta and Slovenia were not available on Bloomberg they are extracted from ECB's Statistical Data Warehouse.12

In order to measure intra-group bank funding flows I use aggregate country data reported by foreign banks located in the specific Euro member states, extracted from Bank for International Settlements (BIS), which is widely used for empirical research on cross-border flows (e.g. Peek and Rosengren, 2000; Van Rijckeghem & Weder, 2001; Cerutti et al., 2014). The data represents the net aggregate cross-border intra-group claims of the banks located in the specific member states.13 Ideally, individual bank level intra-group lending data would be preferable, however this is not available.

11 For Cyprus, Malta, Slovenia and Slovakia the relevant 10-year government bonds are used from the year in

which they entered the Euro area.

12 For Luxembourg a harmonised 10-year government interest rate is used from 2010. Before this period,

Luxembourg did not have sovereign debt securities with maturities close to ten years. Therefore, pre-2010 the 10-year long-term bond yields of high creditworthy private institutions were used (ECB's Statistical Data Warehouse).

13

(12)

12 Therefore data on macro-level is used in order to measure aggregated intra-group funding flows. The variable, INTRA, represents the annual change of net aggregate intra-group cross-border claims per specific Eurozone country. The annual changes are calculated over quarterly available data. The data is only available for the following countries: Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg, the Netherlands and Spain.

The variable, FOREIGN, represents the dummy variable that equals 1 if the operating bank is a foreign subsidiary, and 0 otherwise. Foreign subsidiary is a bank where at least 50% of the shares are owned by foreigners. For the construction of this variable Bankscope information is used.

3.2.3 Control variables

Besides the independent variables, I need to control for variables that influence the dependent variables and are correlated to bank riskiness. Important macroeconomic conditions that drive the dependent variable SLOPE and are correlated to bank riskiness are the following domestic variables GDP growth, inflation, unemployment rate and the debt to GDP ratio (e.g. Afanasieff et al., 2002; Demirgüç-Kunt et al., 2006; Claeys & Vander Vennet, 2008; De Haas & Van Lelyveld, 2010). During more favorable economic conditions (GDP growth) and low inflation projects become more profitable. Banks increase their lending activities in search for yield by reducing screening activities, which increases bank credit risk exposure (Afonso et al., 2010). On the other hand, increasing unemployment and a rising debt-to-GDP ratio will cause an increase in credit risk exposure due difficulties in loan repayments by borrowers. Furthermore, to control for INTRA besides GDP growth and inflation the CDS spread will be used. The CDS spread is used as it measures the degree of (future) risk, uncertainty and attractiveness from international investors perspective. It therefore drives both the interaction variables INTRA and SLOPE, and the dependent variable bank riskiness (e.g. Aizenmann et al., 2013;

Demirgüç-Kunt et al., 2013).

Besides macroeconomic variables also individual bank-level variables are widely used in order to test which bank-specific characteristics influence bank riskiness. However, this thesis has a macroeconomic focus (slope yield curve and internal capital markets) on microeconomic bank riskiness. Therefore, individual bank-specific variables are beyond the scope of this thesis and will be omitted in order to avoid over fitting of the model. Hence, only variables GROWTH, INFLATION,

UNEMPLOYMENT, DEBT and CDS, will be included. The variables represent the member states yearly

(13)

13

3.3 Descriptive statistics

Table 1 shows an overview of the descriptive statistics. The panel set is balanced with a total number of 8,148 observations for all variables. A remarkable observation is the negative minimum value of the variable SLOPE. During the crisis years 2007 and 2008 most countries experienced an inverted yield curve (negative slope) as result of the deteriorating economic situation. The dependent variable NPLS follows a normal distribution. As described in the data section, variables INTRA and

CDS are not available for all Euro area member states and therefore show a deviating number of

observations.14 Additionally, formal tests show signs of autocorrelation and heteroscedasticity are present, which may cause biased standard errors.15 Heteroscedasticity could be the result of non-normality of the dependent variable. Non-non-normality does not produce biased coefficients it does, however, lead to efficiency problems, as the errors might be biased. However, non-normality is not an issue in the main dataset. 16 To control the estimation procedure for heteroscedasticity and autocorrelation Hubert/White errors will be used.

Table 1

Descriptive statistics.

Variable #Obs Mean St. dev. Min Max

NPLS 8,148 0.049 0.027 0.0095 0.115 SLOPE 8,148 0.018 0.014 -0.006 0.056 INTRA 3,629 -0.023 0.134 -0.330 0.581 GROWTH 8,148 0.003 0.022 -0.147 0.107 INFLATION 8,148 0.020 0.009 -0.002 0.106 UNEMPLOYMENT 8,148 0.077 0.033 0.028 0.261 DEBT 8,148 0.901 0.211 0.037 1.323 CDS 7,359 2.235 3.410 2.235 25.212

This table shows the summary statistics for the variables used in the empirical estimation. NPLS is the ratio non-performing loans to gross loans, SLOPE is the spread between the annual average of the 10-year interest rate and the EURIBOR, INTRA is the annual aggregate change in net intra-group positions, GROWTH is the annual GDP growth, INFLATION is the annual change in inflation, UNEMPLOYMENT is the annual unemployment rate, DEBT represents the annual debt to GDP ratio and CDS is the annual change in the difference between the 5-year sovereign CDS bid and ask quotes. The period under consideration is 2005-2014.

Table 2 presents additional descriptive statistics. It gives an overview of the number of domestic and foreign banks operating per Euro country. Important to note is that Germany represents

14

The total number of INTRA observations is related to the total sample. Taking only the subsample of relevant foreign banks into considerations, only 280 observations remain.

15 The modified Wald-test for heteroscedasticity shows a p-value lower than 0.05, whereas the Wooldridge test for

autocorrelation shows a p-value lower than 0.05, indicating both heteroscedasticity and autocorrelation are present.

16 Robustness variable ZSCORE shows signs of non-normality. Therefore it will be transformed to its natural

(14)

14 almost half of all banks in the Eurozone with 2,696 operating banks. The amount of foreign banks is more evenly spread over the Euro member states.

Table 3 provides an overview of the correlation coefficients of the variables used. Strong correlation levels between independent variables can lead to multicollinearity. This can cause errors in the statistical interpretation as the variables' coefficients and the significance are affected by the standard errors. However, based on a inspection of the correlation matrix there are no signs of potential multicollinearity, as all variables have correlation coefficients far below 90 percent.

Table 2

Overview banks per country.

Country Domestic banks Foreign banks Country Domestic banks Foreign banks

Austria 380 32 Italy 910 14 Belgium 106 13 Luxembourg 148 63 Cyprus 29 6 Malta 15 5 Finland 47 4 Netherlands 74 14 France 491 47 Portugal 62 11 Germany 2696 51 Slovakia 33 12 Greece 32 1 Slovenia 28 7 Ireland 43 13 Spain 266 16

This table shows the summary statistics of the number of local and foreign banks operating in the 16 Euro area countries under consideration over the sample period 2005-2014.

Table 3

Correlation matrix.

Variables test (I) NPLS SLOPE GROWTH INFLATION UNEMPL DEBT

NPLS 1 SLOPE 0.373*** 1 GROWTH -0.192*** -0.437*** 1 INFLATION -0.044*** -0.148*** 0.2444*** 1 UNEMPL 0.215*** 0.562*** -0.234*** -0.060*** 1 DEBT 0.404*** 0.486*** -0.337*** 0.027*** 0.145*** 1

Variables test (II) NPLS SLOPE GROWTH INFLATION INTRA CDS

NPLS 1 SLOPE 0.360*** 1 GROWTH -0.174*** -0.543*** 1 INFLATION -0.117*** -0.266*** 0.387*** 1 INTRA -0.158*** -0.196*** 0.293*** 0.368*** 1 CDS 0.353*** 0.784*** -0.558*** -0.058*** -0.029*** 1

(15)

15

4.

Estimation approach

In this section I will discuss the methodology used to estimate the bank risk-taking channel in relation to the slope of the yield curve. First, I look at the different methodologies used in prior literature. Thereafter, I discuss the estimation approach and the model set-up used to test the hypotheses.

4.1 Fixed effects regression model

The empirical approach used in related prior research mostly relies on a series of panel regressions in order to analyse bank riskiness and deal with time-invariant bank specific characteristics. Within this panel approach related literature distinguishes between static panel data regression methods and dynamic panel data methods. The most widely used methods are the fixed effects model for static panel data, and the Generalized Methods of Moments (GMM) in order to estimate a dynamic panel data model.17

The GMM methodology is a popular approach for dynamic panel data which in prior research, includes both macro-level as micro-level bank specific variables as it account for risk persistence and endogeneity of the bank-specific controls. The bank specific characteristics can be related to two main identification challenges, namely potential endogeneity of the interest variable and the persistence of bank risk (e.g. Delis & Kouretas, 2011; Drakos et al., 2014). Besides that, the GMM estimation makes datasets with a large number of banks over a very small time span suitable for estimation (Arellano & Bover, 1998; Blundell & Bond, 1998). It is used to deal with endogenous bank-level variables, while macro-economic conditions are considered as exogenous (see e.g. Delis & Kouretas, 2011). However, micro-level variables are beyond the scope of this thesis, as I am purely interested in the relation between macroeconomic conditions and bank riskiness (see section 3).

I will use a (static) panel regression with fixed effects. Fixed effects are added to the panel regression as differences between individual banks may bias the results. This approach controls for time-invariant characteristics of the banks and therefore the predicted variable's net outcome can be assessed.18 Following the literature, time-effects are also added to the estimation, in order to control for any unexpected variation or events over time that may affect the estimations.19

17 See e.g. Altunbas et al., 2010; Maddaloni & Peydró, 2011; Delis & Kouretas, 2011; Dell'Ariccia et al., 2013;

Drakos et al., 2014.

18 A Hausmann test confirms fixed-effects preferred over random effects, which means that there is no correlation

between time-invariant characteristics.

19

(16)

16

4.2 Estimation approach

In order to estimate the hypotheses discussed earlier, the following basis fixed effects regression model (regression I) is used:

𝑟𝑖𝑗𝑡 = 𝛼 + 𝛽1𝑆𝐿𝑂𝑃𝐸𝑗𝑡 + 𝜃𝑖+ 𝜏𝑡+ 𝜀𝑖𝑗𝑡 (1)

where the dependent variable, 𝑟, represents the level of bank riskiness defined as variable, NPLS, or

ZSCORE for bank, 𝑖, in country, 𝑗 at time, 𝑡. The dependent variable is written as a function of variable, 𝑆𝐿𝑂𝑃𝐸, which represents the slope of the yield. Furthermore, 𝜃, and, 𝜏, represent the fixed effects and time effects respectively, included into the regression in order to control for time-invariant characteristics, whereas, 𝜀, is the error term.

𝑟𝑖𝑗𝑡 = 𝛼 + 𝛽1𝑆𝐿𝑂𝑃𝐸𝑗𝑡 + 𝛾𝐺𝑅𝑂𝑊𝑇𝐻𝑗𝑡 + 𝛿𝐼𝑁𝐹𝐿𝐴𝑇𝐼𝑂𝑁𝑗𝑡 + 𝜗𝑈𝑁𝐸𝑀𝑃𝐿𝑂𝑌𝑀𝐸𝑁𝑇𝑗𝑡 + 𝜇𝐷𝐸𝐵𝑇𝑗𝑡 + 𝜃𝑖+

𝜏𝑡+ 𝜀𝑖𝑗𝑡 (2)

As discussed in the previous section the variable, SLOPE, is driven by macroeconomic conditions that are correlated to bank riskiness. Therefore, I should include variables GROWTH,

INFLATION, UNEMPLOYMENT and DEBT into equation (2) in order to avoid omitted variables bias,

which can lead to over- or underestimation of variable SLOPE.

𝑟𝑖𝑗𝑡 = 𝛼 + 𝛽1𝑆𝐿𝑂𝑃𝐸𝑗𝑡 + 𝛽2𝑆𝐿𝑂𝑃𝐸𝑗𝑡 ∗ 𝐹𝑂𝑅𝐸𝐼𝐺𝑁𝑖+ 𝛾𝐺𝑅𝑂𝑊𝑇𝐻𝑗𝑡 + 𝛿𝐼𝑁𝐹𝐿𝐴𝑇𝐼𝑂𝑁𝑗𝑡 + 𝜇𝐷𝐸𝐵𝑇𝑗𝑡 +

𝜗𝑈𝑁𝐸𝑀𝑃𝐿𝑂𝑌𝑀𝐸𝑁𝑇𝑗𝑡 + 𝜃𝑖+ 𝜏𝑡+ 𝜀𝑖𝑗𝑡+ (3)

Finally, in order to measure the specific effect of the slope on bank riskiness of foreign banks the dummy variable, 𝐹𝑂𝑅𝐸𝐼𝐺𝑁, is added to the equation.

4.3 Intra-group funding flows foreign banks

In order to test the third hypothesis, I will perform a subsample test (regression II) and only include the foreign banks present in the sample. The sample is further restricted as variable INTRA is not available for all Eurozone countries.20 The following estimation set-up will be used:

𝑟𝑓𝑜𝑟𝑒𝑖𝑔𝑛𝑖𝑗𝑡 = 𝛼 + 𝛽1𝑆𝐿𝑂𝑃𝐸𝑗𝑡 + 𝛽2𝐼𝑁𝑇𝑅𝐴𝑗𝑡 + 𝛽3𝑆𝐿𝑂𝑃𝐸𝑗𝑡 ∗ 𝐼𝑁𝑇𝑅𝐴𝑗𝑡 + 𝜃𝑖+ 𝜏𝑡+ 𝜀𝑖𝑗𝑡 (4)

20

(17)

17 where the dependent variable, 𝑟𝑓𝑜𝑟𝑒𝑖𝑔𝑛, represents the level of riskiness of foreign banks defined as,

NPLS, and ZSCORE for bank, 𝑖, in country, 𝑗, at time, 𝑡. The dependent variable is a function of interaction variable, SLOPE*INTRA, which stands for the interaction between the slope of the yield curve and the intra-group capital flows, and the error term, 𝜀. Whereas the individual (interaction term) variable, FOREIGN, is absorbed due to the bank fixed effects, variable INTRA needs to be added to the equation in order to measure the interaction effect properly (Brambor etl al., 2006). As earlier mentioned, the above fixed-effects, 𝜃, and time effects, 𝜏, are added to the equation, in order to control for special time events and time-invariant effects.

𝑟𝑓𝑜𝑟𝑒𝑖𝑔𝑛𝑖𝑗𝑡 = 𝛼 + 𝛽1𝑆𝐿𝑂𝑃𝐸𝑗𝑡 + 𝛽2𝐼𝑁𝑇𝑅𝐴𝑗𝑡 + 𝛽3𝑆𝐿𝑂𝑃𝐸𝑗𝑡 ∗ 𝐼𝑁𝑇𝑅𝐴𝑗𝑡 + 𝛿𝐺𝑅𝑂𝑊𝑇𝐻𝑗𝑡 +

𝜇𝐼𝑁𝐹𝐿𝐴𝑇𝐼𝑂𝑁𝑗𝑡 + 𝜗𝐶𝐷𝑆𝑗𝑡 + 𝜃𝑖+ 𝜏𝑡+ 𝜀𝑖𝑗𝑡 (5)

As discussed in the previous paragraph, control variables GROWTH, INFLATION and CDS, which influence both SLOPE and INTRA and are correlated to bank riskiness, are added to the equation in order to avoid omitted variable bias.

5.

Empirical results

In this section the main results of the fixed effects methodology will be discussed. Table 4 reports the results of equations 1-3 using the complete dataset. Table 5 presents the results of equations 4 and 5 for the sub-samples tests of foreign banks and the influence of intra-group capital flows. The tables present the results in three panels: (I) the basic fixed effects regression, (II) the baseline fixed effects regression (control variables included) and (III) the fixed effects regression with Hubert/White errors. At the end of this section different kinds of robustness checks will be discussed.

5.1 General results

(18)

18 results of panel II and III show highly significant (p-value<0.01) results for all control variables except for GDP growth (GROWTH). The coefficients' signs of the control variables are as expected.

5.2 Foreign banks

Table 5 presents the results of the sub-sample results (regression II) for foreign banks operating in the Euro area countries. Variable SLOPE is highly significant (p-value<0.01) and positively related to bank riskiness. There is no significant relation between intra-group funding flows (INTRA) and bank riskiness. However, the results in Panel I-III show highly significant coefficients (p-value<0.01) for the interaction term INTRA*SLOPE. The significant interaction term implicates that the effect of one variable on the dependent variable NPLS changes, as the value of the other variable changes. Given the fact both variables INTRA and SLOPE are continuous, the results presented in table 4 only throw limited light on the hypothesis.

Table 4

Main results regression (I) - Dependent variable NPLS (I) FE (II) FE Baseline (III) FE Hubert/White errors SLOPE 0.534*** (11.91) 0.398*** (5.74) 0.398*** (4.30) FOREIGN*SLOPE -0.228*** (-3.75) -0.254*** (-4.27) -0.254** (-2.40) INFLATION -0.241*** (-2.81) -0.241*** (-2.60) GROWTH 0.041 (1.07) 0.041 (0.94) DEBT 0.039*** (3.92) 0.039*** (3.03) UNEMPLOYMENT 0.243*** (7.22) 0.243*** (5.52) #Obs. 8,131 8,131 8,131 F-Value 234.93 216.09 63.39 R-sq within 0.31 0.36 0.36

(19)

19 For better interpretation of the interaction variable a graphical illustration is needed that illustrates how the marginal effect of the slope of the yield curve changes across the observed range of INTRA (Brambor et al., 2006).

Figure 1 presents the average marginal effects of the interaction variable, and illustrates that the marginal effects of SLOPE decrease for increasing values of INTRA. However, the marginal effects are only significant up to a value of circa 0.09 INTRA. Figure 1 provides significant (weak) evidence in support of the third hypothesis that the positive relation between bank riskiness and the slope of the yield curve is negatively influenced by intra-group funding flows.

Table 5

Main results regression II (foreign banks) - Dependent variable NPLS (I) FE (II) FE Baseline (III) FE Hubert/White errors SLOPE 0.642** (2.25) 1.283*** (2.45) 1.283* (1.72) INTRA 0.441 (0.95) 0.049 (1.01) 0.049 (1.12) INTRA*SLOPE -7.347*** (-4.07) -4.845*** (-2.83) -4.845*** (-2.80) INFLATION -1.990*** (-2.98) -1.988* (-1.78) GROWTH 0.549 (1.56) 0.549* (1.63) CDS -0.000* (-1.73) -0.000 (-1.36) #Obs. 280 280 280 F-Value 132.85 116.67 110.07 R-sq within 0.274 0.323 0.323

(20)

20 Furthermore, in the baseline regression (II) all control variables except GROWTH are significant and the coefficients signs are in line with prior research and the previous presented results. When dealing with autocorrelation and heteroscedasticity (Panel III), INFLATION and GROWTH are significant albeit at a marginal level (p-value<0.10), while CDS becomes insignificant.

Figure 1. Average marginal effects interaction variable SLOPE*INTRA. This figure shows the interaction term's

marginal effects of the conducted regressions. The dependent variable is the ratio of non-performing loans to gross loans (NPLS). Independent variables used for the interaction term are as follows: SLOPE, is calculated as the spread between the long-term 10-year government rate and 3-month EURIBOR interest rate and INTRA, the average annual aggregate change per member state of net outstanding internal capital market positions. The period under consideration is 2005-2014. Note: significance at 5% is denoted by the red line.

5.3 Robustness

In order to test the estimated results for robustness, several checks are performed in order to see how different choices affect the results. An overview of the robustness results are presented in Appendix section B and C.

The first robustness test checks whether the results of regression (I) remain in line with the main estimations by substituting the dependent variable, NPLS, with a different proxy of bank riskiness,

ZSCORE. The main results show a negative and highly significant relation between the slope of the

yield curve and bank riskiness measured by ZSCORE. Given the fact that a lower ZSCORE indicates higher bank riskiness, the results are in line with the main estimations. Important to note is the negative and significant coefficient of SLOPE*FOREIGN. A positive coefficient was expected and would be in line with the main results. This difference might be caused given the fact ZSCORE is a constructed bank (credit) risk proxy, and therefore also influenced by other factors, whereas NPLS is considered as a true credit risk proxy (Ozsuca & Akbostanci, 2012. The coefficients of the robustness test for regression (II)

Intra E ff ec ts o n lin ea r p red ictio n

(21)

21 are in line with the main results. However, only variable SLOPE is significant. The interaction effect turns out to be insignificant. Inspection of the margin effects (see Appendix C, figure 1) shows that the relation is insignificant and therefore not robust.

The second robustness check is performed by using the EONIA rate instead of the EURIBOR as the short-term interest rate for the construction of the slope of the yield curve. For both regression (I) and regression (II) the main results are robust.

Given the fact that the time period under consideration contains the severe years of the global financial crisis, two sub-sample test are performed. The sample period will be split up into a 'crisis-period' (2007, 2008, 2009, 2010) and a time period excluding the crisis years. For both regression (I) and regression (II) the main variables (except for SLOPE*FOREIGN) are insignificant during the crisis period. The marginal effect graph for the crisis period (see Appendix C, figure 1) does not show a relation between the interaction variables and is highly insignificant. For the second sub-sample time period (excl. crisis years) the results remain in line with the main results for both regression (I) and regression (II).

According to Delis & Kouretas (2011) there might be problems regarding reverse causality. In order to limit these effects I test for robustness using lagged variables. Remarkably, for regression (I) the main variables remain highly significant. In regression (II) the main variable, SLOPE*INTRA, remains significant albeit at a lower level of significance (p<0.05) (see Appendix B, table 2). The graphical presentation of the marginal effects is similar to the graph of the main results, however less values are significant (see Appendix C, figure 1).

Finally, a robustness test is conducted for regression (I) by dropping German banks from the regression sample. Given the fact German banks represent almost half of the total sample set, they might have a strong influence on the underlying results. Dropping German banks from the regression does not lead to significant different results. In other words, the results remain in line with the main regression.

6.

Conclusion

(22)

22 The impact of short-term interest rate changes on bank risk-taking has recently been much documented in literature. This thesis investigates the macro-micro relation between the spread of the long-term and short-term interest rate, bank riskiness, and the dampening role of the internal capital markets on foreign banks' exposure to risk. Determining how the slope of the yield curve and internal capital markets influence banks' exposure to risk is important for policymakers in order to determining widespread financial stability among banks.

Using a fixed effects model upon a large panel dataset of banks operating in the Euro area over the period 2005-2014, I exploit information on the riskiness of banks by analyzing how macroeconomic factors impact banks' exposure to credit risk. In addition, I specifically analyze the effect of intra-group funding flows on foreign bank riskiness.

I present new evidence linking the slope of the yield curve and bank riskiness: a steeper slope of the yield curve (widened spread between long-term and short-term interest rates) is positively related to the level of bank riskiness. This slope-to-bank riskiness nexus is in line with Mink's (2011) theoretical proposition that banks' credit risk exposure intensifies due to an increase in maturity transformation driven by leverage and softening lending standards. In addition, the results provide evidence that the effects are stronger for domestically operating banks than for foreign banks.

Furthermore, I open a new strand of literature by directly linking the functioning of internal capital markets and its stabilizing effect on the level of (foreign) bank riskiness. The results provide (limited) evidence of a negative relation between the interaction of foreign banks net intra-group claims and the slope of the yield curve and bank riskiness. In other words, net intra-group funding flows stabilize bank riskiness as the slope steepens. This provides evidence for the underlying hypothesis and theory based on i.a. diversification theory and Morgan et al.'s (2004) model, that international operating banks reallocate funds across borders to maintain efficient risk-return investment and avoid unnecessary credit risk. Hence, through this mechanism, the positive relation between bank riskiness and the slope of the yield curve is less profound for foreign banks (as compared to domestic banks).

(23)

23 focusses purely on the spread between the short-term and long-term interest rate. Therefore, it does not take into account how (extreme low) short-term interest rate levels might solely drive the results. Finally, the theory and data used for measuring the effect of the internal capital markets in relation to the slope of the yield curve and bank riskiness might be endogenous to an underlying theory regarding mismanagement in foreign operating banks, as subsidiary managers may have less entrepreneurial incentive to take risk.

(24)

24

7.

References

Acharya, V. V., & Richardson, M. (2009). Causes of the financial crisis. Critical Review, 21(2-3), 195-210.

Adrian, T., & Shin, H. S. (2009). Money, liquidity, and monetary policy. FRB of New York Staff Report, (360).

Adrian, T., & Shin, H. S. (2010). Financial intermediaries and monetary economics. FRB of New York Staff Report, (398).

Afanasieff, T. S., Lhacer, P. M., & Nakane, M. I. (2002). The determinants of bank interest spread in Brazil. Money Affairs, 15(2), 183-207.

Aizenman, J., Hutchison, M., & Jinjarak, Y. (2013). What is the risk of European sovereign debt defaults? Fiscal space, CDS spreads and market pricing of risk. Journal of International Money and Finance, 34, 37-59.

Alessandri, P., & Nelson, B. D. (2015). Simple banking: profitability and the yield curve. Journal of Money, Credit and Banking, 47(1), 143-175.

Altunbas, Y., Gambacorta, L., & Marques-Ibanez, D. (2010). Bank risk and monetary policy. Journal of Financial Stability, 6(3), 121-129.

Ang, A., Piazzesi, M., & Wei, M. (2006). What does the yield curve tell us about GDP growth? Journal of Econometrics, 131(1), 359-403.

Angeloni, I., Faia, E., & Duca, M. L. (2015). Monetary policy and risk taking. Journal of Economic Dynamics and Control, 52, 285-307.

Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of error-components models. Journal of Econometrics, 68(1), 29-51.

Barrell, R., Davis, E. P., Karim, D., & Liadze, I. (2010). Bank regulation, property prices and early warning systems for banking crises in OECD countries. Journal of Banking & Finance, 34(9), 2255-2264.

Barth, J. R., Caprio Jr, G., & Levine, R. (2013). Bank Regulation and Supervision in 180 Countries from 1999 to 2011. Journal of Financial Economic Policy, 5(2), 111-219.

Bernanke, B. S. (2005). The global saving glut and the US current account deficit, (77).

Bernanke, B. S., & Gertler, M. (1995). Inside the black box: the credit channel of monetary policy transmission. National Bureau of Economic Research Working Paper, (5146).

Bernanke, B. S., & Blinder, A. S. (1988). Credit, money, and aggregate demand. National Bureau of Economic Research Working Paper, (2534).

Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87(1), 115-143.

(25)

25

Brambor, T., Clark, W. R., & Golder, M. (2006). Understanding interaction models: Improving empirical analyses. Political analysis, 14(1), 63-82.

Buch, C. M., Eickmeier, S., & Prieto, E. (2014). In search for yield? Survey-based evidence on bank risk taking. Journal of Economic Dynamics and Control, 43, 12-30.

Claeys, S., & Vander Vennet, R. (2008). Determinants of bank interest margins in Central and Eastern Europe: A comparison with the West. Economic Systems, 32(2), 197-216.

Cetorelli, N., & Goldberg, L. S. (2012). Banking globalization and monetary transmission. The Journal of Finance, 67(5), 1811-1843.

Cordella, T., & Yeyati, E. L. (2002). Financial opening, deposit insurance, and risk in a model of banking competition. European Economic Review, 46(3), 471-485.

Dahl, D., Shrieves, R. E., & Spivey, M. F. (2002). Financing loan growth at banks. Journal of Financial Services Research, 22(3), 189-202.

Diamond, D. W., & Dybvig, P. H. (1983). Bank runs, deposit insurance, and liquidity. The Journal of Political Economy, 401-419.

Diamond, D. W.,& Rajan, R. G. 2006. "Money in a theory of banking."American Economic Review,

96(1), 30-53.

Dell'Ariccia, G., Igan, D., & Laeven, L. (2009). Credit booms and lending standards: Evidence from the subprime mortgage market. Journal of Money, Credit and Banking, 44(2-3), 367-384.

Delis, M. D., & Kouretas, G. P. (2011). Interest rates and bank risk-taking. Journal of Banking & Finance, 35(4), 840-855.

DellʼAriccia, G., Laeven, L., & Marquez, R. (2014). Real interest rates, leverage, and bank risk-taking. Journal of Economic Theory, 149, 65-99.

Dell'Ariccia, M. G., Laeven, M. L., & Suarez, M. G. (2013). Bank leverage and monetary policy's risk-taking channel: Evidence from the United States (No. 13-143). International Monetary Fund.

Dell'Ariccia, G., & Marquez, R. (2006). Lending booms and lending standards. The Journal of Finance, 61(5), 2511-2546.

Demirgüç-Kunt, A., Detragiache, E., & Tressel, T. (2008). Banking on the principles: Compliance with Basel core principles and bank soundness. Journal of Financial Intermediation, 17(4), 511-542. Demirgüç-Kunt, A., & Huizinga, H. (2013). Are banks too big to fail or too big to save? International evidence from equity prices and CDS spreads. Journal of Banking & Finance, 37(3), 875-894. Desai, M. A., Foley, C. F., & Hines, J. R. (2004). A multinational perspective on capital structure choice and internal capital markets. The Journal of Finance, 59(6), 2451-2487.

Drakos, A. A., Kouretas, G. P., & Tsoumas, C. (2014). Ownership, interest rates and bank risk-taking in Central and Eastern European countries. International Review of Financial Analysis.

Friedman, M., & Schwartz, A. J. (2008). A monetary history of the United States, 1867-1960. Princeton University Press.

(26)

26

Gropp, R., Hakenes, H., & Schnabel, I. (2010). Competition, risk-shifting, and public bail-out policies. Review of Financial Studies, (114).

De Haas, R., & Van Lelyveld, I. (2010). Internal capital markets and lending by multinational bank subsidiaries. Journal of Financial Intermediation, 19(1), 1-25.

De Haas, R., & Lelyveld, I. (2014). Multinational banks and the global financial crisis: Weathering the perfect storm? Journal of Money, Credit and Banking, 46(1), 333-364.

Hovakimian, G. (2011). Financial constraints and investment efficiency: Internal capital allocation across the business cycle. Journal of Financial Intermediation, 20(2), 264-283.

Ioannidou, V. P., Ongena, S., & Peydró, J. L. (2009). Monetary Policy, Risk-Taking and Pricing: Evidence from a Quasi-natural Experiment. European Banking Center Working Paper, (2009-04). Jiménez, G., Ongena, S., Peydró, J. L., & Saurina Salas, J. (2009). Hazardous times for monetary policy: what do twenty-three million bank loans say about the effects of monetary policy on credit risk-taking? Banco España Working Paper, (0833).

Jeon, B. N., & Wu, J. (2014). Global banks and internal capital markets: Evidence from bank-level panel data in emerging economies. Journal of Multinational Financial Management, 28, 79-94. Kashyap, A. K., & Stein, J. C. (1994). Monetary policy and bank lending. The University of Chicago Press in Monetary policy, 221-261.

Keeley, M. C. (1990). Deposit insurance, risk, and market power in banking. The American Economic Review, 1183-1200.

Jeon, B. N., Lim, H., & Wu, J. (2014). The impact of foreign banks on monetary policy transmission during the global financial crisis of 2008-2009: Evidence from Korea. Bank of Korea Working Paper, (8).

Laeven, L., & Levine, R. (2009). Bank governance, regulation and risk taking. Journal of Financial Economics, 93(2), 259-275.

Laeven, L., & Levine, R. (2007). Is there a diversification discount in financial conglomerates? Journal of Financial Economics, 85(2), 331-367.

Larrain, B. (2006). Do banks affect the level and composition of industrial volatility? The Journal of Finance, 61(4), 1897-1925.

Brealey, R., Leland, H. E., & Pyle, D. H. (1977). Informational asymmetries, financial structure, and financial intermediation. The Journal of Finance, 32(2), 371-387.

Maddaloni, A., & Peydró, J. L. (2011). Bank risk-taking, securitization, supervision, and low interest rates: Evidence from the Euro-area and the US lending standards. Review of Financial Studies, 24(6), 2121-2165.

Männasoo, K., & Mayes, D. G. (2009). Explaining bank distress in Eastern European transition economies. Journal of Banking & Finance, 33(2), 244-253.

Mink, M. (2011). Procyclical bank risk-taking and the lender of the last resort. DNB.

(27)

27

Morgan, D., Rime, B., & Strahan, P. (2004). Bank integration and state business cycles. Journal of Econonomics, 119(4), 1555-1585.

Navaretti, G. B., Calzolari, G., Pozzolo, A. F., & Levi, M. (2010). Multinational banking in Europe– financial stability and regulatory implications: lessons from the financial crisis. Economic Policy, 25(64), 703-753.

Ozsuca, E. A., & Akbostanci, E. (2012). An empirical analysis of the risk-taking channel of monetary policy in Turkey. Economic Research Center Working Paper, (1208).

Peek, J., & Rosengren, E. S. (2000). Collateral damage: Effects of the Japanese bank crisis on real activity in the United States. American Economic Review, 30-45.

Rajan, R. G. (2005). Has financial development made the world riskier? National Bureau of Economic Research Working Paper, (11728)

Reinhardt, D., & Riddiough, S. J. (2014). The two faces of cross-border banking flows: an investigation into the links between global risk, arms-length funding and internal capital markets. Arms-Length Funding and Internal Capital Markets (May 2014).

Stein, J. C. (1995). An adverse selection model of bank asset and liability management with implications for the transmission of monetary policy. Journal of Economics, 29(3): 466-486. Stiglitz, J., & Greenwald, B. (2003). Towards a new paradigm in monetary economics. Cambridge University Press.

Stiroh, K. J., & Rumble, A. (2006). The dark side of diversification: The case of US financial holding companies. Journal of Banking & Finance, 30(8), 2131-2161.

Van Rijckeghem, C., & Weder, B. (2003). Spillovers through banking centers: a panel data analysis of bank flows. Journal of International Money and Finance, 22(4), 483-509.

Wu, J., Luca, A. C., & Jeon, B. N. (2011). Foreign bank penetration and the lending channel in emerging economies: Evidence from bank-level panel data. Journal of International Money and Finance, 30(6), 1128-1156.

(28)

28

APPENDIX A

Table 1 Variables.

Variable Measure Source

SLOPE 10 year German Bund – 3 month EURIBOR.

Reflects the slope of the yield curve

Bloomberg / ECB

NPLS Ratio non-performing loans to total loans. Reflects

the quality of bank assets.

Bankscope

ZSCORE Measure of credit risk, calculated as

(ROA-Equity/Total Assets) / Std dev. ROA

Bankscope

GROWTH Annual GDP growth IMF

INFLATION Annual change in inflation IMF

INTRA Annual aggregate change of net intra-group bank

capital positions

BIS

FOREIGN Index, international banks equals 1 and domestic

equals 0.

Bankscope

UNEMPLOYMENT Annual unemployment rate per member state Eurostat

DEBT Annual debt to GDP ratio per member state Eurostat

CDS Annual change in member states' monthly 5-year CDS spread average

(29)

29 APPENDIX B

Table1

Main results robustness tests (regression I) all banks ZSCORE (I) SLOPE-II (II) Crisis period (III) Excl. crisis years (IV) Lagged (V) No Germany (VI) SLOPE -13.922*** (-5.52) 0.404*** (4.37) -0.351 (-0.42) 0.196** (2.13) 0.919*** (8.20) 0.594*** (6.49) FOREIGN*SLOPE -6.273** (-2.03) -0.260** (-2.43) -0.245*** (-2.96) -0.299* (-1.65) -0.162* (-1.64) -0.255*** (-2.69) INFLATION -5.967** (-2.37) -0.241*** (-2.61) 0.089 (0.29) -0.383*** (-4.23) -0.416*** (-4.03) -0.157* (-1.75) GROWTH -8.680*** (-7.84) 0.042 (0.97) -0.256*** (-2.80) -0.024 (-0.54) 0.088* (1.76) 0.078 (1.52) DEBT 0.973*** (2.71) 0.038*** (3.01) 0.002 (0.03) -0.001 (-0.40) -0.01 (-0.53) -0.018 (-1.18) UNEMPLOYMENT -6.90*** (-6.75) 0.243*** (5.54) 0.086 (1.04) 0.614*** (7.94) 0.163*** (3.19) 0.240*** (6.01) #Obs. 25,400 8,131 1,413 6,020 5,615 5,543 F-Value 66.85 63.55 56.09 58.26 58.58 64.45 R-sq within 0.05 0.36 0.33 0.32 0.42 0.40

(30)

30 Table 2

Main results robustness tests (regression II) foreign banks ZSCORE (I) SLOPE-II (II) Crisis period (III)

Excl. crisis years (IV) Lagged (V) SLOPE -16.981*** (-2.83) 1.192* (1.63) -0.177 (-0.12) 1.841** (2.23) 0.294 (0.63) INTRA 0.041 (0.08) 0.077 (1.43) 0.045 (1.38) 0.051 (0.59) 0.033 (0.80) INTRA*SLOPE 1,288 (0.06) -6.432*** (-2.68) 1.649 (1.20) -6.407** (-1.87) -5.241** (-1.97) INFLATION -0.021 (-0.00) -2.027** (-1.79) -0.323 (-0.48) -2.207* (1.59) -3.286** (-2.02) GROWTH -1.711 (-0.54) 0.523* (1.53) -0.226 (-0.67) 0.623 (1.15) -0.746* (-1.89) CDS 0.002* (1.91) -0.001 (-1.28) 0.008** (2.59) 0.000* (1.67) -0.000 (-1.22) #Obs. 1,031 280 72 203 256 F-Value 23.08 120.59 2.39 76.92 47.70 R-sq within 0.03 0.32 0.43 0.31 0.31

Referenties

GERELATEERDE DOCUMENTEN

Generally, the results are in line with the benchmark panel VAR that was estimated with quarterly data: a sovereign debt shock negatively affects both the output

The data includes the following variables: output measured by real gross domestic product, inflation using the price deflator for private consumption, the short-term interest

As mentioned above the sovereign spread of GIIPS countries do not react different to changes in the debt to GDP ratio, credit rating or US yield compared to the other

The third dependent variable, loan loss provisions to total assets, shows also a significant and positive value between the short-term, long-term and bank-level lending

All of the Best fit models of the countries included in this research contain (macro)economic variables and have much higher explanatory power (measured in adjusted R-squared)

This table presents regression results in the years before the crisis (pre: 2000-2006) and during and after the crisis (post: 2007- 2014) of the effect of yield curve movements and

They are as follows: z-score is the distance to insolvency (measured as the natural logarithm of [ROA+CAR]/st.dev.[ROA]), NPL ratio is the ratio of non-performing loans to total

In contrast to Niu (2012) for the American banking sector, I find for only the nonperforming loans ratio significant evidence of an inverted U-shaped relationship between charter