The effect of capital shocks on bank lending : an empirical study

(1)

Master Thesis

Faculty of Economics and Business

University of Amsterdam

Date: 7th_{of July, 2016}

Student name: Beaudine Besem

Student number: 11096373

Track: Business Economics: Finance

Current title of Thesis: The Effect of Capital Shocks on Bank Lending: An Empirical Study Academic supervisor: Dr. S.R. Arping

(2)

Statement of originality

This document is written and published by Beaudine Besem who declares to take full responsibility for the contents of this thesis. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study empirically analyzes the impact of a capital shock on the loan growth of European banks. Moreover, the focus lies on the way banks respond to a capital shock. When the capital ratio (which is represented by the equity-to-assets ratio) is decreasing, banks still need to ensure that they hold the required percentage of capital on hand. Using Bureau van Dijk Osiris data from 120 individual European banks during 2005-2014, five hypotheses are tested on the relation between the loan growth and the equity-to-assets ratio, the retained earnings, the net income, the total dividend, the common shares and the return on equity. One important finding is that there exists a positive and significant relationship between the banks that are close to the threshold, which experience changes in their capital and the loan growth. Moreover, several robustness checks are performed. After executing the robustness checks, it can be concluded that the general results are driven by the time frame that is used and the size of the banks.

JEL classifications: G01, G18, G21, G28, G38, H25

(4)

Introduction

This study investigates how European banks respond to capital shocks when they are under- or overcapitalized. Having built a complete and accurate dataset, a couple of panel regressions are run on the on forehand-invented variables. The main variables that are used are the loan growth and the equity-to-assets ratio. Moreover, some extended other regressions are made to see which underlying variables drive the relationship the most.

1.1 Research topic

The financial crisis made clear that highly indebted financial institutions can cause negative externalities that harm the worldwide economy. Many researchers have investigated the situation and have identified insufficient capital buffers of banks as one of the main reasons for the onset of the crisis. Other researchers have diagnosed the financial distress of financial institutions as the main cause for the creation of the financial crisis. Eventually both researchers share the same opinion: a situation of financial distress needs to be avoided. To find a way to avoid financial distress, one needs to know the theoretical meaning of the phenomenon. In theory, financial distress occurs when a financial institution does not fulfill its obligations optimally. The costs that arise in situations of financial distress are equal to the additional loss from economic distress of an indebted bank versus an identical other institution (Admati, Demarzo, Hellwig, & Pfleiderer, 2013). In case of a capital shock, both institutions will experience costs in the form of economic distress. Nevertheless, the leveraged institute will experience a greater loss because of the increased risk to go bankrupt (Berger, Herring, & Szegö, 1995). Furthermore, a central consequence of financial distress is the rise of systemic risk.

As we could have seen during the recent financial crisis: financial distress of one company can easily spill over into other financial institutions. The risk that arises in such a situation is called ‘systemic risk’. As the rise of systemic risk could lead to a credit crunch, it is important to concentrate on the avoidance of this risk. One way to lower the systemic risk is to give banks the requirement to ensure that they have enough capital on hand. Moreover, regulators argue that banks need to fund themselves with more equity instead of more debt. By increasing the equity within the firm, the potential systemic risk will decrease. Banks will have more skin in the game (Admati et al., 2013). Other fundamental issues and complaints that have emerged from the financial crisis were also mainly concentrated on the role of bank capital. Regulators see capital as a form of ‘bank survival’: a mechanism to avoid a new distressed situation. For this reason, regulators argue that a new upcoming crisis can only be avoided if the banking sector is strengthened so that it can deal with large unexpected capital shocks. Moreover, the financial restrictors state that the quality of the bank’s assets needs to be in line with the amount of capital a bank holds on the balance sheet. In order to meet the regulatory expectations, the Basel Committee has created some regulatory requirements. The Basel Committee on Banking Supervision (BCBS) is responsible for creating requirements that meet the

(6)

bank safety objective. The requirements should provide more stability and security in the financial system (Cohen & Scatigna, 2016). The first official Basel Accord was established in 1988. This accord and its amendments required capital to be above a certain threshold. The capital requirements were designed to ensure that banks had enough capital to get through a potential crisis. In other words, the aim of the capital requirements was to require financial institutions to maintain enough capital to absorb upcoming losses without causing systemic disruptions and to level the playing field on an international base (Noss & Toffano, 2016). The total capital ratio, which was considered as the sum of the Tier 1 ratio and the Tier 2 ratio, needed to be equal to or higher than eight per cent of the risk-weighted assets (BIS, 2010). The second Basel Accord was implemented in 2004. This accord was designed to address the shortcomings of the Basel I Accord. Some of the shortcomings were the coarse risk categories and the bad reflection of the credit risk exposure. Therefore, the second Basel Accord changed the risk weights and the way financial institutions could assess the risk within the company (Blundell-Wignall et al., 2014). After the introduction of the newest Basel Accord, one negative externality have been originated: the creation of moral hazard. As the banks felt to such an extent insured by the safety net, moral hazard came into the system. Consequently, some banks invested their money in too risky projects and hold too few liquid assets. To avoid the uprising of new negative externalities, another version of the Basel Accord have been designed. The latest version of the Basel Accord, the Basel III accord, has been gradually phased in since 2013, and will continue untill the end of 2019. The focus of this new accord lies on the recognition of liquidity risk, the recognition of the risk of a bank run, the introduction of the minimum leverage ratio, stress testing and the implementation of designated additional buffers for global systematically important banks (Blundell-Wignall et al., 2014). After addressing the development of all regulatory restrictions one can conclude that financial regulators are constantly searching for ways to improve the banking stability in the economy.

As already mentioned, the capital requirements are designed to ensure that banks have enough capital to get through a downturn in the economy. In order to develop new improved capital requirements it is important to know whether the actual rules that are being taken have the desired effect. In particular, understanding whether a higher capital level has a significant effect on the survival likelihood of a bank and how this effect differs depending on the level of the bank’s actual capital ratio are important details for regulators who try to deliver the desired level of banking stability. This empirical study investigates how European banks respond to unexpected capital shocks that arise in the economy. As the outstanding loans form a significant part of the bank its assets, this study will focus on the response to this particular variable. A research question is designed to give this research an empirical and scientific focus point.

(7)

1.2 Research question

This study investigates whether the level of the capital ratio do affect the way European banks adjust their outstanding loans when they experience capital shocks. Moreover, the focus lies on the dispersion between the banks that are close to the minimum level of the required capital ratio and the banks that are further away from this level. The following question is of interest for this thesis: ‘How does the response of European banks to capital shocks depend on the level of their capital ratio?’ 1.3 Relation to existing literature

Several studies have already investigated what the effect is of a change in the regulatory capital requirements (or a change in the capital ratio) on the lending of a bank. The authors of the article ‘The impact of Capital Requirements on Bank Lending’ (2014) state that the regulatory capital requirements influence the level of the capital ratios held by banks. Besides, the authors conclude that a change in the regulatory capital requirements will lead to a change in the total outstanding loans. In addition to the existing literature, it is interesting to see whether banks that are close to the minimum level of the required capital behave differently from banks that are further away from this level. In advance, the authors Buch and Prieto (2014) researched whether better capitalized banks lend on average more than the less-capitalized banks in Germany during the period from 1965 until 2009. The authors found statistical evidence that confirmed the positive relation between the two variables. As the research of Buch and Prieto (2014) focused on one European country, it is relevant to see whether the same conclusion can be drawn for Europe. Furthermore, Bernanke and Lown (1991) investigated the same for the US credit crunch during the nineties. The authors concluded that states with less-capitalized banks experienced less loan growth. Besides, the authors Peek and Rosengren (1996) did a comparable research for the Japanese banking sector. The relevant results of all existing literature are addressed in the literature review of this study.

1.4 Contribution

This thesis contributes to the existing literature for at least two reasons. First, regulators will see what the indirect effect is of the introduction of the capital requirements. Since the capital ratio of a bank needs to be above the eight per cent, official actions are required when the capital level falls below this point. This research will show whether banks that are close to the critical threshold behave differently from the banks that are further away from the threshold. Another reason why this subject contributes is because a high level of uncertainty exists about how banks respond to unexpected decreases in their capital ratio. This uncertainty will affect the real economy depending on the state of the business cycle (Noss & Toffano, 2016). Moreover, as capital acts as the ‘fundament of the bank its business’, a decrease in capital could have consequences for the actual worthiness and profitability of a bank. When we analyse the behaviour of banks according to their capital ratio, we may find patterns and correlations that explain the behaviour of banks properly. Conclusively, this empirical

(8)

investigation is of great interest to policymakers, as this research gives more insight in the behaviour of European banks.

1.5 Introduction methodology and data

The methodology consists of several panel data regressions that measure the relation between the dependent variable ‘the growth rate of the total outstanding loans’ and the independent variable ‘the equity-to-assets ratio (ETA)’. Consequently, the beta1 shows the relation between the changes in the

capital ratio and the loan growth. Moreover, a binary dummy is included to show the difference in effect between banks that are close to the threshold and banks that are further away from the threshold. In general, the minimum threshold value is equal to the required capital ratio of eight per cent. After including the binary dummy, an interaction term is added to show the relation between the banks that are close to the threshold that experience changes in the capital ratio. As stated in the introduction, the emphasis of this research lies on this particular coefficient. After executing the regression on the loan growth and the equity-to-assets ratio, some other regressions are performed on the retained earnings, the net income, the total dividend, the common shares and the return on equity. These additional regressions are executed to see which underlying variables drive the relation the most. The Bureau van Dijk Osiris Banks Financials database is used to get access to the data that we need for this research. Moreover, the focus of this thesis is on the time period from 2005 to 2014.

1.6 Outline thesis

The structure of this thesis is as follows. In the second chapter, an overview of all the theories and literature on this topic will be presented. In the third chapter, the data and methodology will be discussed and explained. After the data and methodology are discussed, the results and conclusion will be presented. All the tables, figures and illustrations referred to in the thesis can be found in the appendix.

(9)

Literature review

This literature review addresses all important theories and models according to this study. As capital forms a crucial element of the business of banks, some regulatory constraints are adopted to restrict the amount of capital a bank must hold. As a consequence, some banks hold more capital than other banks. This leads to the fact that there exist significant differences between banks given their capital level. If the capital ratio comes closer to the restricted capital level of eight per cent, the banks need to extend their capital level.

In general, a lot of the authors refer to the theory of Modigliani and Miller (1958) when evaluating the effect of a capital shock. The financial benchmark predicts that a change in the composition of the liabilities of a bank won’t change the overall funding costs of a bank. The authors assume that the risk level of the assets stays unchanged and they assume that the market is ‘perfect’. So, without a change in the funding costs of a bank there is no reason to change the bank its capital ratio. And eventually, there shouldn’t be a change in the price and quantity of the bank its credit. However, given the fact that the market is triggered by, for example the long-term friction effect of the tax deductibility of debt payments, the Modigliani-Miller theory is not always the most accurate benchmark to use. Evidence is given by the authors Bahaj et al. (2016) who state that the Modigliani-Miller theory isn’t valid as the asset overhang problem and/or subsidies from the government affect the bank decisions that the entities make.

Given the fact that the Modigliani-Miller theory is not completely in line with the functioning of the real economy, some authors have created alternative capital definitions. At first, Admati et al. (2013) have created a definition for the term ‘capital requirement’. According to Admati et al. (2013), a capital requirement refers to the way that banks are funded and thus the mix of debt and equity within the banks. Berger et al. (1995) define the market capital requirement as ‘the capital ratio that maximizes the value of the bank in the absence of regulatory capital requirements but in the presence of the rest of the regulatory structure that protects the safety and soundness of banks’. The ratio should be equal to the perceived long-run capital ratio of a bank in absence of the regulatory requirements. As banks have the highest level of leverage in the financial industry, regulatory institutions have introduced the requirements to secure and safeguard the economic system (Berger et al., 1995).

When following the banking literature, most authors refer to the capital ratio as the equity-to-assets ratio (Berger, 1995; Bernanke & Lown, 1991; Board et al., 2010; Foos et al., 2010). The authors see the equity-to-assets ratio as a measure for identifying the probability of insolvability of a bank. It calculates whether the bank is able to cover all unexpected losses in the current state (Foos et al., 2010). The lower the capital ratio, the higher the expected costs of financial distress and the higher the

(10)

probability of insolvency. In a traditional setting, banks hold a certain amount of capital in the absence of the regulatory capital requirements. When a bank holds an amount of capital in excess of the required capital it needs to hold, this is called ‘Capitalization’ (Gambacorta et al., 2004; Berger et al., 1995). But, why does the market requires financial institutions to hold capital? In general, banks need to hold capital for a number of reasons. First, the most important reason is that banks finance their business different from other financial institutions. Banks perform a so-called ‘maturity transformation’: they invest their money in long-term risky assets and fund this with short-term liquid liabilities. Consequently, the banks their profit is reflected by the risk premium. This profit can fluctuate by a change in the confidence in the economy. If the market confidence decreases, the economy will experience a downward trend. As a consequence, capital acts like a buffer in times of a downward economy. Another reason why banks need to hold a certain amount of capital is because of the demandable nature of the deposits. Demandable deposits have two general facets. First, the depositors can ask their money back at any time in the future. Furthermore, these depositors have the privilege to be given priority over other claims (Diamond & Rajan, 2000). One may think that the potential for a deposit run serves as a discipline for banks. Nevertheless, the authors Diamond and Rajan (2000) share another opinion on this particular subject. The researchers state that bank capital acts like a cushion in distressed situations, but bank capital may also be costly since the liquidity creation will be reduced. As additional capital makes it hard for well-capitalized banks to commit monitoring, the bank’s ability to create liquidity will be restricted. Conclusively, capital forms an important element of the balance sheet of a bank, especially during a downturn (Kapan et al., 2013). Additionally, Admati et al. (2013) their research investigates the function of capital on the bank’s balance sheet. In their research, the authors conclude that the strength of the balance sheet of a bank is measured by the amount of equity on the balance sheet. Equity represents ‘the ownership in the form of common shares’. It functions as a buffer because the holders of the shares do not have a hard claim as the debt holders have (Admati et al., 2013). For this particular reason, regulators forbid distressed banks to pay out large portions of dividend and to purchase new stocks. As equity serves as a stable fundament for banks during these moments, the regulators and market participants will feel the effect of a shock. Another function of equity is that it has a long maturity and cannot be collected during a period of crisis in general (Berger et al., 1995). As a result, the stronger the balance sheet, the more banks will be able to sustain their lending in times of a financial crisis (Kapan & Minoiu, 2013).

2.1 Regulatory constraints

The main objective of financial regulation is to avoid the existence of systemic risk and the associated social costs. In other words, regulators need to safeguard the functioning of the financial system (Admati et al., 2013; Berger et al., 1995). Hanson et al. (2011) define the goal of capital regulation as: ‘the way to force banks to internalize losses and thereby protecting the deposit insurance fund and mitigating moral hazard’. The authors conclude that financial regulation is used to control the social

(11)

costs that are associated the balance shrinkage of financial institutions that are hit by an unexpected shock. A positive externality of the regulatory restrictions is that the investors’ confidence in the market will improve. When a bank allows to decrease the funding costs, the level of capital will increase without decreasing the level of lending (Noss & Toffano, 2016). To realize the stated goal, the Basel Committee has published a global regulatory framework of the underlying Basel Accords in relation to the banking systems. In the end, all the regulated banks need to have enough capital on hand to meet the financial regulatory requirements.

One central element of the global regulatory framework is formed by the capital requirements. The total capital requirement can be separated into three levels: the Common Equity Tier 1 capital (CET), the Additional Tier 1 capital (AT1) and the Tier 2 capital (T2) (Berger et al., 1995). The Common Equity Tier 1 capital ratio is equal the book value of equity relative to the risk-weighted assets. This value consists of the issued common shares, the stock surpluses, the retained earnings, other comprehensive incomes and disclosed reserves. The Additional Tier 1 capital is denoted as the amount of preferred shares and the high-trigger contingent convertible bonds (coco’s). The Tier 2 capital is defined as the total of all loan loss reserves, hybrid capital, convertible debt securities, low-trigger coco’s and subordinated debt instruments. Currently, the CET needs to be equal to or higher than 4.5 per cent. The combined value of the CET and AT1 together needs to be equal to or higher than 6 per cent. And the total capital ratio, the value of the CET, AT1 and T2 together, needs to be equal to or higher than 8 per cent (BIS, 2010).

One other regulatory element that influences the way banks behave according to their capital is the ‘Safety net’. The safety net refers to all the government-related actions to safeguard the stability in the financial system other than the capital requirements. This includes the legal deposit insurance, the unconditional payment guarantees and the available discount window. As the safety net reduces the incentive for banks to have above average capital ratios, this measure insulates the regular market discipline of banks (Berger et al., 1995). In addition, one should know that some countries do have additional Basel standards that safeguard the balance of their local economy (Dionne & Harchaoui, 2008). As these requirements are part of the Basel III Accord that is not yet completely implemented, these restrictions will not be enclosed in this research.

2.2 Capital and bank lending

This section starts with a short note to clarify the definition of the ‘capital ratio’ within this research. The existing literature on the relationship between the capital of a bank and the lending behaviour can be separated into two underlying groups: the studies that investigate the impact of shocks to the capital resources on bank lending and the studies that investigate the impact of shocks to the capital requirements on bank lending. The first category focuses on the shocks to the observed capital levels of banks. The other studies focus mainly on the impact of shocks to the regulatory capital requirements on the total lending of banks. Within this literature review, the focus lies on the studies

(12)

that investigate the impact of shocks to the observed capital levels of banks.

Several studies have conducted an enquiry to see the relation between the observed loan growth and changes in the capital level. Gambacorta & Mistrulli (2004) conclude that a decline in the capital ratio leads to a downward turn in the loan supply. Adrian et al. (2008) and Bridges et al. (2014) confirm the same statistical relation in their research. Next to that, Peek and Rosengren’s research (1995) shows the magnitude of this relation within the English market during the early nineties recession. The focus of this research is mainly on the link between the regulatory activities and the lending behaviour of the English banks. Instead of using the change in outstanding loans, the authors performed a regression on the amount of new loan extensions and the capital ratios. Within the study, the authors corrected for the charge-offs, the transfers of foreclosed real estate loans and net loan sales to see the net flow of new loans. The central conclusion of the research is that there existed a correlation between the shrinkage of the loan growth and the capital ratios of the English banks during the early nineties recession. Next to that, the authors conclude that the well-capitalized banks invest more money in new loan extensions as they realize more growth. Moreover, they conclude that the less-capitalized banks are more affected by the downturn in the economy than the well-less-capitalized banks. This is due to the ‘too big to fail’ idea that large banks could have. In the end, the authors expect that the correlation between the shrinkage of the bank loans and the capital ratio can be caused by the voluntary actions performed by the undercapitalized banks that try to improve their capital ratio. In another research, the authors Peek and Rosengren (1996) investigate the linkage between the regulatory requirements, the capital ratio of a bank and the declination of the banks’ lending. They use a natural experiment to address the effects of capital shocks on the bank its lending. The focus of this study lies on the Japanese banks located in the United States. One of the findings is that when the risk-based capital ratio falls with a one percentage point, this causes a four percentage point decrease in the loan growth. The authors Furfine et al. (2000) also investigate the linkage between the loan growth and the capital ratios of banks in the United States. The authors build a theoretical model that reflect the significant relation between the capital requirements, the change in outstanding loans and the change in the capital of a bank. In the end, the authors conclude that capital regulation explains the decline in loan growth and the rise in bank capital. Bernanke and Lown (1991) draw the same general conclusion: there exists a significant correlation between capital ratios and the shrinkage of a bank. The authors find that a shortage of equity capital limited banks’ ability to extend new loans in some regions within the United States. At first, the authors were sceptical about the fact that the credit crunch played a major role in worsening the recession. In the end, they admitted that the relation between the bank shrinkage and the capital ratio played a decisive role.

Buch and Prieto (2014) analyze the link between bank capital and bank loans in the long run by using a panel data regression. In their research, they have observed German banking groups over the last 44 years (1965–2009). One of their findings is that bank capital and business loans are

(13)

cointegrated. So, banks with a higher capital level lend more to private firms. Another finding is that the long-term impact of bank capital on the total loans is positive. One per cent increase in the level of bank capital leads to a 0.22 per cent growth in the bank loans. The response of the total loans to an increase in bank capital becomes negative for the companies that have an equity-to-assets ratio of 33 per cent or higher.

2.3 Differences in the capital level

The regulatory capital requirements are implemented to ensure that banks make better economically appropriate decisions that create social value (Admati et al., 2013). But given the fact that no bank is completely identical to another bank, the capital requirements will affect all banks differently in the business cycle stages. Gambacorta et al. (2004) investigated the situation and concluded that smaller banks with a large asset-liability maturity mismatch experience most effects. Moreover, banks with a higher Common Equity Tier 1 capital ratio reduce their lending less than identical banks with a lower level of Common Equity Tier 1 capital ratio (Kapan & Minoiu, 2013). The authors Dionne & Harchaoui (2008) also examined the differences between banks given their capital level. The investigators pointed out that the banks with a high capital ratio were less able to compete. A justification for this outcome was that the equity costs were higher than the costs associated with debt. The main motivation for keeping additional capital was that banks held additional capital as a hedge against having to raise new equity or debt in a short time.

Consequently, a well-known phenomenon is that banks hold additional capital on their balance sheet. The authors Berger, DeYoung, Flannery, Lee, & Öztekin (2008) investigated why banks hold excess capital on their balance sheets. They created three hypotheses to test the central question ‘Why do banks hold excess capital?’ The first hypothesis focused on the earnings retention. The lower the earnings retention, the more capital the bank will hold. The second hypothesis focused on the economic capital. The expectation was that the more volatile companies needed to have more capital since they were more sensitive to default risk. Given the riskiness of these companies, these companies could also be the more valuable charters. The second hypothesis was linked to the market-to-book ratio. A low market-market-to-book ratio was associated with lower capital ratios. The third hypothesis focused on the asset size of banks. The expectation was that the asset size affected the banks’ preferred capital ratio. The more extended companies tended to be more diversified, which eventually would led to economies of scale and less costs that are linked to the raise of new equity. Eventually, all hypothesis were tested and all null hypotheses were accepted (Berger et al., 2008). Overall, a thing that proved to be true is that there exist significant differences between banks given their capital level. At first, the well-capitalized banks will be better in making lending decisions in contrast to the banks that are less capitalized. The banks that are well capitalized are less constrained by the regulatory capital requirements and have therefore wider opportunities to extend their lending portfolio. Moreover, the credit supply of the less-capitalized banks will be more dependent on the

(14)

business cycle. So, when there occurs a downturn in the economy, the well-capitalized banks will experience fewer losses in the loan exposure and the banks their capital ratio will change less in contrary to less-capitalized banks (Gambacorta & Mistrulli, 2004). As a consequence, banks that have enough capital on hand will not pass up valuable lending chances (Admati et al., 2013).

A subsequent argument is that banks with a higher capital ratio supply more loans over time (Buch & Prieto, 2014). As a consequence, the well-capitalized banks increase their lending during a boom. If these banks are in a bust, the banks will only increase their lending if their capital ratio is decreasing and close to the critical threshold. As a consequence, the well-capitalized banks will decline their lending less in contrast to the banks that are less capitalized (Kapan & Minoiu, 2013). Another connected consequence is that banks with more equity will be better in absorbing losses. According to Gambacorta et al. (2004), the capital level of a bank will influence the way a bank reacts to GDP shocks in the economy. A well-capitalized bank will be better in absorbing financial difficulties of its clients and lending relationships. For this reason, the profits of well-capitalized bank are less exposed to changes in the business cycle, as the portfolio choices will be different from the banks that are less capitalized (Gambacorta & Mistrulli, 2004). The more capital the bank has, the more perspective on efficiency in the lending segment the bank will have (Admati et al., 2013). An associated advantage of having enough capital is that the banks are able to accommodate capital losses without reducing the total assets. Less-capitalized banks will actively manage their capital ratios to maintain a constant ratio, as a reduction of their capital ratio could easily lead to a reduction of their total assets and hence their lending (Board et al., 2010).

At last, the banks that are well capitalized are less willing to invest in excessive risky investments that give a benefit to shareholders and the managers of a bank. Since the debt of well-capitalized banks is safer and less sensitive to information, this debt is potentially better for granting liquidity within the bank. Moreover, well-capitalized banks will need less debt since they have enough capital on hand to fund their growth (Admati et al., 2013). In general, all banks will prefer equity finance above debt finance. In the article, the authors Admati et al. (2013) refer to the ‘Pecking order theory’ to explain this phenomenon. At last, the research of Berger et al. (1995) shows that banks that have a high equity-to-assets ratio have a lower chance to fail in the future. However, the strength of this relation changes over time and is not equally significant in all periods.

2.4 Adjustment of the capital ratio

If the capital ratio of a bank is close to the critical threshold of eight per cent, the bank needs to adjust its capital level. Noss and Toffano (2001) investigated the way how banks could adjust their balance sheet to perceive the required capital ratio. When a bank is inadequately capitalized, an official action to increase the capital level is needed. Such an action will increase the confidence and resilience in the market. Moreover, the costs of funding will decrease. Eventually, this will have a positive effect on the outstanding loans. One central conclusion the authors draw is that an increase in capital does not

(15)

necessarily have to lead to a decrease in the outstanding loans. One underlying condition is that the funding costs must stay the same. When the funding costs are increasing, the additional costs will be passed on to the borrowers by raising interest and reducing the quantity of the loans. When the outstanding loans are decreasing, this will lead to a reduction in the return on equity and thus a reduction of the profitability of the bank. In case the market risk is low, the funding costs are relatively insensitive to the amount of capital. In a scientific exploration of Memmel & Raupach (2010), the authors conclude that the private commercial banks and the banks with a high amount of liquid assets adjust their capital ratio faster than other banks. Next to that, the authors show that banks with high asset volatility do have higher capital ratios.

The adjustment of the capital ratio can be accomplished in several ways. One general way to increase the capital ratio is by increasing the capital, by reducing the risk-weighted assets in proportion to the total assets of the bank and/or by decreasing the total assets of the bank (Dionne & Harchaoui, 2008). According to the article of Admati et al. (2013), a bank can change its capital ratio in three ways. First, a bank can ‘delever its balance sheet’ by scaling back the size of the balance sheet. This is called ‘Asset liquidation’. This method can cause pressure on the asset market, which eventually will lead to falling asset prices. Moreover, the outstanding loans will decrease. The deleveraging of assets becomes only more logic when the bank has a higher capital ratio. Another possibility is called ‘Recapitalization’. This means that a bank issues additional equity. At last, a bank may raise equity and assets in equal proportions. This is called ‘Asset expansion’ (Admati et al., 2013; Buch & Prieto, 2014; Cohen & Scatigna, 2016).

(16)

Data and Descriptive statistics

This chapter describes the data by using a descriptive statistics. Next to that, this chapter gives an interpretation of the relationships between the epsilon (Y) and the drivers. In this study, data of European commercial banks between 2005 and 2014 is used. All data is collected from the Bureau van Dijk Osiris database. Before starting with interpreting the relationships and correlations between the variables, the way to structure this chapter will be introduced. The SAS Institute Inc. introduced a way to mine data in logic and functional way. The tool is called ‘SEMMA’ and consists of five underlying steps: Sample, Explore, Modify, Model and Assess.

3.1 Data overview

The SEMMA tool emerges with finding the right dataset. One essential element of this process is that the sample needs to be extended enough to be valid for the research. If the dataset is not unfolded enough, then the results are not representative for drawing general conclusions. Five steps have been taken to optimize the dataset of this study. First of all, a dataset is created that includes all available variables in the period from 2005 until 2014. Secondly, two conditional statements are implemented to optimize the data sample. The first condition concentrates on the currency type. As this research examines only European countries, this study only works with the currency ‘EURO’. When there are European countries with another monetary unit in the sample, these values are translated to euros by using the exchange rate on a yearly base. The second condition focuses on the total loan volume. We assume that only European banks with a total loan exposure larger than zero are able to add value to this research. Thirdly, all non-European countries as Benin, Burkina Faso, Côte D’Ivoire, Niger, Saint Kitts and Nevis, Saint Lucia and Senegal are excluded. At this stage, the dataset includes data of 20 countries. Fourthly, all the financial institutions other than the commercial banks are excluded. Within this research, the focus lies on commercial banks since these banks are organized in a way to handle the financial transactions on a daily base. Moreover, the issuance of capital is of interest for these banks. All the other financial institutions have another business model and therefore focus on another business field. Lastly, the last step is focused on the unconsolidated and consolidated statements. The Osiris database allows for downloading both consolidated and unconsolidated statements. This study focuses on the unconsolidated statements, since the capital behavior of the underlying banking companies will show more diversified effects. So, in order to have no duplicates in the dataset, this study only works with the U1 and U2 statements. All the other statements, as the C1 and C2 statements, are deleted out of the sample.

Hereafter, the final dataset entails 928 observations in which 120 European banks are included. The panel dataset can be described as ‘unbalanced’ since the number of observations per country and bank vary. Next to that, we can conclude that there are missing data items in the sample. For this reason, one other restriction has been put in place. When there are extreme data points in the sample,

(17)

these data items have been dropped. One example would be that the total assets observations feature a percentage growth below or above the hundred percent. In this case, it’s likely that the variable will be driven by exogenous factors. This restriction has been put in place to improve the robustness of the findings. Consequently, 16 observations are dropped.

To revise the dataset even more, macroeconomic data is merged with the existing dataset. The macroeconomic control variables include the real GDP growth rate, the inflation rate (CPI) and the Euribor rate (3 months). The GDP growth rate and the inflation rate are collected from the World Bank dataset. The inflation rate (CPI) is used instead of the inflation rate (GDP), since the inflation rate as a percentage of the customer price index may influence the speed at which lending is growing. Moreover, the Euribor rate is collected via DataStream. The interbank offered rate of the ECB is calculated by measuring the weighted average rate of the daily interbank deposits on a yearly base.

Given the fact that this study focuses on European banks, it’s important to see how dispersed the European banks are. If it is the case that all banks are situated in one country, this will influence the overall results. For this reason, an overview is made to see where the banks are located. The next table states that the banks in the sample are mostly situated in West-Europe and South-Europe. Table 1 is added to the appendix.

[Table 1: Representation per country]

When evaluating the table, three main conclusions can be drawn. Firstly, the table shows that only the countries Bulgaria, Latvia, Finland, Ireland, Slovakia and Estonia are situated in the north and east of Europe. All the other countries are located in South-Europe and West-Europe. Approximately 87.5 per cent of the countries are located in the south and west of Europe. Secondly, a conclusion can be drawn on the number of banks per country. In table 1, all included countries are categorized by selecting the land volume in cubic meters. One would expect that the largest countries would have most banks in the sample. However, this is not the case in this particular dataset. Italy and Montenegro have proportionally more banks than expected given their ranking in the list. Moreover, Finland has less observed banks in the sample. After this analysis, we can conclude that the banks are not divided on a base of land volume. Lastly, it can be concluded that a large proportion of the banks in the sample are situated in Italy.

3.2 Summary statistics

A descriptive statistic is designed to see the median, the average, the standard deviation and the percentiles of each variable. The notes, which are stated under the descriptive table on page 47, give a description of all the stated values.

(18)

[Table 2: Descriptive statistics sample]

Three interesting elements can be discussed after analyzing the statistics. The first element that catches the analyst’s eye is that the dataset represents an unbalanced panel. A large proportion of the included banks have a high total assets value. The mean lies at 82 billion euros, whereas the 50th

percentile lies at 6.6 billion euros. As a consequence, we can conclude that most banks can be categorized as ‘large banks’. Secondly, the values of the equity-to-assets ratio of the banks are dispersed. The mean lies at 8.87 per cent, whereas the 25th_{percentile lies at 4.36 per cent and the 75}th

percentile at 10.35 per cent. A third element concentrates on the fact that some banks do not have enough capital on hand or didn’t have enough capital on hand. The 4.36 per cent can only be realized before the Basel Committee introduced the Basel III Accord, since this accord persists that the capital ratio needs to be above the eight per cent. Consequently, the distribution of the capital ratios is not constant over time. One brief note needs to be made on this last argument. Since the equity-to-assets ratio is not calculated in the way that the required capital ratio is calculated, we must be careful in drawing conclusions or expectations on this particular aspect.

3.3 Cross correlation table

A correlation matrix is created to address all underlying correlation coefficients between the variables. [Table 3: Cross correlation table]

Table 3 shows the degree of linear association between the variable pairs within the sample. The numbers that are written in bold script are of interest for this study. As not one of the correlations is close to minus or plus one, it can be concluded that there doesn’t exist a strong relationship between one of the variable pairs. The largest correlation exists between the size indicator and the lagged value of the retained earnings. This middle-strong correlation exists since the size indicator is affected by the level of the lagged value of the retained earnings and vice versa. When we regress the retained earnings on the size indicator, there exists a large chance that the coefficient on the size indicator will be positive and significant.

3.4 Exploration of the dataset

The next step entails the ‘exploration of the dataset’. In order to understand the data, the data is explored by addressing the theoretical explanation behind each relation. This paragraph gives the statistical and theoretical explanation of the drivers and epsilons. In the end, all the relevant scatterplots and underlying statistical tests are included in the appendix of this thesis.

The main focus of this study is on the relationship between the loan growth and the equity-to-assets ratio (ETA) of a bank. The epsilon in this model is the ‘Loan growth’. Ideally the real value of the total new loan extensions would be used, however this data point is not in the sample. For this reason,

(19)

this study uses the same method as in the article ‘The Credit Crunch’(1991) is used for calculating the lending growth rate. Instead of using the real value of the new loan extensions, we use the nominal growth rate of the outstanding loans. This is measured by the ΔL/L*100% calculation. After making a correlation table, it is made clear that the impact of the loan loss provisions on the loan growth is relatively small and insignificant. For this reason, we eliminate the variable out of the dataset. The key driver within this model is the equity-to-assets ratio. When looking to other researches on this topic, this ratio is mostly used to measure the solvency of a bank (Berger, 1995; Bernanke & Lown, 1991; Board et al., 2010; Foos et al., 2010). Another possible way to undertake this research was to use the capital ratio (calculated according to the Basel restrictions). As the equity-to-assets ratio is divided by the total assets instead of the risk-weighted assets, the values of both ratios differ. Since the number of observations of the total capital ratio is below the 150, the reliability of the data point is however questionable. Consequently, the decision is made to focus on the equity-to-assets ratio. The descriptive values of this variable are in between the 1.49 and the 38.24 per cent. Moreover, the median is 6.74 per cent. The ratio calculates whether the bank is able to recover if it experiences losses. In the current state of the economy, European banks need to have a required capital level on hand. As the required capital ratio is currently eight per cent, this threshold is used for splitting the sample in two groups: close to the threshold versus not close to the threshold. All banks that have an equity-to-assets ratio between the 8 and 8.67 are categorized as ‘close to the threshold’. As we have already seen in the descriptive statistics, most banks have significantly more capital. If there are values below the 2.5 per cent – quartile and the 97.5 per cent – quartile, these values were considered as outliers and are excluded out of the sample.

Besides the fact that the equity-to-assets ratio plays a central role, it is important to differentiate the banks by their capital level to see the differences in behavior. Two main situations can arise between well-capitalized banks and less-capitalized banks. Firstly, banks that are less capitalized can rely on an underpricing strategy since their poor solvency stadium does not allow for the acceptance of high-risk borrowers. These banks will ‘dehigh-risk’ their portfolio to keep the capital ratios high enough. Secondly, another situation that can arise is that banks with a high ETA ratio will grow by setting low loan spreads. The well-capitalized banks benefit from the high capital buffer that allows for a temporary decline in interest. Consequently, the banks will take on more risk to increase their expected return. Next to that, the well-capitalized banks take on more risk because their management skills are more developed. They know exactly how they need to do business and how to realize gains. Eventually, seventeen scatterplots are created to illustrate the statistical relation between the epsilon and the drivers that are used in this study.

(20)

[Graph 1&2: Relation Loan Growth, the lagged value of the ETA ratio & Δ ETA-ratio] Graph 1 shows the relation between the loan growth and the lagged value of the equity-to-assets ratio. The scatterplot shows a positive relationship between the loan growth and the lagged value of the ETA ratio. The higher the lagged value of the ETA ratio, the more money is invested in capital. Consequently, the higher the lagged value of the equity-to-assets ratio the more the bank can be invested in loans in the future. If the current value of the ETA ratio is going upwards, the loans of well-capitalized banks will increase in t+1. Moreover, the second graph focuses on the relation between the loan growth and the percentage change in the ETA ratio. The higher the change in the ETA ratio, the more money is invested in capital. As a consequence, we can conclude that a negative change in the ETA ratio is correlated with a positive loan growth in period t. One explanation for this negative relation is the change in the balance of the equity-to-assets ratio. It could be the case that additional money is invested in new loan extensions, which results in a higher degree of assets. To optimize the dataset, some relevant control variables are included. Specifically: the growth rate of the GDP, the Euribor rate, the inflation rate (CPI), the natural logarithm of the total assets and a risk indicator are included.

[Graph 3, 4 & 5: Relation Loan Growth, GDP growth rate, inflation rate & the Euribor rate] The growth rate of the GDP serves as a macroeconomic control variable. The relation between the loan growth and the growth rate of the GDP is positive: an increase in the GDP level leads to an increase in the outstanding loans. The second macroeconomic variable that is included is the inflation rate (CPI). The relation between the loan growth and the inflation rate (CPI) is positive: the higher the inflation, the lower the purchasing power parity (PPP) and the more people will lend from the bank. The last macroeconomic variable that is included is the Euribor rate. The higher this rate, the better the macroeconomic situation. Eventually, more people will invest their money in new loan extensions. So, the relationship between the loan growth and the Euribor rate is positive.

[Graph 6: Relation Loan Growth and the size indicator]

The natural logarithm of the total assets is used to control for the bank size. The way of calculating the size indicator is confirmed in the research of Peek & Rosengren (1995). The larger the bank, the more the bank is constrained by the restricted capital ratios. Moreover, the more extended the bank is, the more difficult it becomes to realize more loan growth in percentages. Given the outcome of scatterplot 6, it can be concluded that there exists a weak negative relationship between the variable ‘bank size’ and the loan growth.

(21)

The last control variable that is incorporated is the risk indicator, which is represented by the Z-score. The Z-score is regularly used in empirical financial researches to identify the bank’s probability of insolvency. Hannan et al. (1988) have also explored the bank risk by this particular approach. According to this article, the Z-score makes an accurate assessment of the individual bank risk and the overall financial stability in the economy.The Z-score is calculated by the following formula: [equity-to-assets ratio + return on assets] / standard deviation of the return on assets. As already mentioned in the first part of this chapter, the expectation is that the well-capitalized banks will take on more risk to gain a higher return. For this reason, the assumption is that there exists a weak positive relationship between the risk indicator and the loan growth.

[Graph 8 t/m 13: Relation Loan Growth, RE, NI and DIV variables]

To see the effects of a change in the ETA ratio on the loan growth in more detail, some additional variables are included. Instead of using the ETA ratio, the four underlying drivers that affect the equity level are used. To start with the retained earnings (RE), which include the change in net income (NI) and the change in dividend (DIV). The scatterplots on the lagged value of the net income, the lagged value of the retained earnings and the lagged value of the total dividend are around zero. Since the three scatterplots identify a relation that lies around zero, we can assume that the relation between the variable pairs is very weak. The relations between the loan growth and the change in net income and the retained earnings are strong: the larger the absolute growth in retained earnings and net income, the larger the loan growth. The statistical relation between the loan growth and the absolute change in dividends is weakly positive.

[Graph 14 & 15: Relation Loan Growth & common shares variables]

Another way to replace the ETA ratio is to change this ratio by using the growth rate of the common shares (CS). The variable ‘Common shares’ of the Bureau van Dijk Osiris database is used to test whether a change in the common shares influences the loan growth of a bank. Graph 14 and 15 show the scatterplots that are associated with this statistical relationship. The relation between the lagged value of the common shares and the loan growth is around zero. Moreover, the absolute change in the common shares influences the loan growth positively.

[Graph 16 & 17: Relation Loan Growth & ROE variables]

One other element that is incorporated in this research is the return on equity ratio (ROE). The regular way of measuring the profitability of a bank is by using the return on equity ratio. Graph 16 and 17 show the statistical relation between the loan growth and the profitability of a bank. This ratio is calculated by dividing the net income by the stockholders’ equity. We expect that the higher the return on equity, the higher the growth rate of the total net loans.

(22)

3.5 Modifying the dataset

After the exploration of the dataset, the dataset can be modified. The modification phase is the phase where the outliers are deleted, the number of variables may be reduced and the most significant variables are selected. All variables are winsorized at a 2.5 per cent level to exclude the largest outliers. Besides, a histogram is drawn for all variables to see the shape of the normal distribution. When the distribution was left-shaped, only the low values were winsorized. In case the distribution was right-shaped, only the high values were winsorized.

3.6 Model and assessment

The final model can be constructed by linking all the variables in a three-based model. The assessment of the data is done in a later stage. After explaining the methodology of this study, the assessment takes place.

%LOAN GROWTH

ETA SIZE & RISK

EQUITY ASSETS MACRO ECONOMIC FACTORS ENTITY & TIME FIXED EFFECTS INFLATION GDP EURIBOR RETAINED EARNINGS NET INCOME DIVIDEND COMMON SHARES

(23)

Methodology

In the methodology, the hypotheses and the underlying regression equations are presented. Moreover, as mentioned in the data section, this chapter incorporates the last step of the SEMMA tool designed by the SAS Institute. First, this study addresses the hypotheses of this study. Afterwards, all seven regression equations and the endogeneity issues are illustrated. As the previous section has already focused on data, there is no subsection about data in this chapter.

As stated in the introduction of this study, the main objective of this research is to investigate how banks behave when their capital ratio changes. The prediction is that there exists a significant difference between the banks that are close to the minimum capital requirement and the banks that are further away from this limit. The banks that are close to the threshold are forced to change their risk-weighted assets or change their equity. Within this research, the main focus lies on European banks in the period from the end of 2005 until the end of 2014. Given the fact that the most significant differences arise at bank-level stadium, the focus lies on the bank-level differences within Europe. To answer the research question optimally, a couple of hypotheses have been drawn.

4.1 Hypotheses

To give a statistical and accurate answer to the question whether the behavior of banks is dependent of the level of their capital ratio, five hypotheses have been drawn. This chapter presents all hypotheses and the arguments for performing the hypotheses.

Hypothesis 1: There exists a significant relation between the loan growth and change in the ETA ratio The first hypothesis focuses on the relation between the loan growth and the change in the equity-to-assets ratio (ETA). This hypothesis is designed to find out whether there exists a statistical relation between the two variables. As some authors investigated this relation already for other countries (Peek & Rosengren, 1993; Peek & Rosengren, 1996; Furfine, 2000; Bernanke and Lown, 1991), it is of interest to discover whether this statistical relationship is also present in the dataset of this research. Moreover, the outcome will say something about the way banks re-invest their equity money in new assets, dividend pay-outs and/or the accumulation of the reserves. Afterwards, there can be concluded whether a change in the ETA ratio will affect the actual loan performance of the banks.

Hypothesis 2: The banks that are close to the threshold are more likely to respond to changes in their capital ratios than the banks that are far from the threshold.

The second hypothesis forms the main aspect within this research. The hypothesis looks whether the banks that are close to the required minimum capital ratio behave differently from the banks that are

(24)

far away from the eight per cent threshold. When evaluating the literature on this subject, most researchers believe that well-capitalized banks are less constrained by the regulatory capital requirements (Admati et al., 2013). Therefore, these entities have wider opportunities to extend their loan exposure (Buch & Prieto, 2014). Next to that, the expectation is that the well-capitalized banks will change their equity-to-assets ratio less than the banks that are less capitalized. This will be due to the fact that the credit supply of a less-capitalized bank is more dependent on the business cycle (Gambacorta & Mistrulli, 2004). Eventually, the expectation is that banks that are close to the threshold will respond more to changes in their ETA ratio than the banks that are far away from the restricted threshold level of eight per cent.

Hypothesis 3: The significance of the response of banks to changes in their capital ratios is correlated to the size of the bank

The third hypothesis investigates whether the size of a bank influences the statistical relation between the loan growth and the equity-to-assets ratio. As stated in the literature review, there exist differences between banks according to their capital level and size. Gambacorta et al. (2004) concluded that small banks with a large asset-liability mismatch experience most effects. Since the used dataset entails significant more large banks than small banks, the focus lies on the larger banks in this hypothesis.

Hypothesis 4: The underlying drivers of the bank its equity do influence the change in the loan growth significantly

Moreover, the next hypothesis focuses on the equity drivers that could influence the growth rate of the outstanding loans. As the equity-to-assets ratio is used to represent the capital ratio within this research, the drivers that could drive this ratio can be identified. Within this hypothesis, the focus is on the nominator of the formula: the total equity. A change in the equity level can be influenced by a change in the net income, a change in the total dividends, a change in the retained earnings and a change in the common shares. After performing the regressions, it will become clear how most European banks adjust their capital ratio and which drivers influence the loan growth. Most banks can adjust their capital level by performing asset liquidation, recapitalization or an asset expansion (Admati et al., 2013; Buch & Prieto, 2014; Cohen & Scatigna, 2016).

Hypothesis 5: The level of profitability of a bank is correlated with the loan growth of a bank.

The last hypothesis concentrates on the link between the profitability and loan growth of the banks. Noss and Toffano (2001) did some research to the correlation between the return on equity and the loan growth of banks. In the research, the authors conclude that a decline in the outstanding loans

(25)

leads to a decrease in the return on equity ratio. This last hypothesis tests whether there exists a correlation between the return on equity ratio and the loan growth of European banks.

4.2 Research design

The proposed research design for testing these hypotheses are several panel data regressions. The utilization of panel data regressions is appropriate in this case because it takes into account all the individual bank behaviours and is flexible in providing more degrees of freedom. The overall sample size of this research is formed by 120 European banks. These banks are situated in twenty European countries. Within the regressions, some sub-groups are diversified to see the difference between these levels. At first, a diversification is made between the banks that are close to the minimum regulatory capital ratio and the banks that have a capital ratio far above this level. This difference is addressed by splitting the banks that have an equity-to-assets ratio between the 8 and 8.67 per cent (=1) and the banks that have a capital ratio under or above this rate (=0). Another way to specify the groups is to split the large, medium and small banks in the dataset by asset size. This particular aspect is executed in the third regression equation.

4.3 Statistical model

4.3.1. Regressions on the equity-to-assets ratio

At first, seven fixed effect regression equations will be presented for testing the first three hypotheses. Within the following regressions, all the potential determinants of the variable ‘loan growth’ along with entity and time fixed effects are stated. The first equation tests the causal relation between the loan growth and the lagged equity-to-assets ratio. This study executes this regression since the lagged value of the equity-to-assets ratio is included as a coefficient in the regressions for answering the second and third hypothesis. Within this study, the focus lies on the first lagged value (t-1), since this variable has explanatory power according to the loan growth of a bank. Moreover, the loan growth of period t is influenced by the lagged value of the ETA ratio. The first fixed effect regression equation is stated as following:

ΔL/Li,t = αi + β1 * ETAi,t-1 + λt + εi,t (1)

The dependent variable ΔL/Li,t-1 is the growth rate of the total loans outstanding. The ΔL/Li,t-1 denotes

the loan growth by bank i in period t. The ETAi,t-1 represents the equity-to-assets ratio at the end of

period t-1. The ETAi,t-1 is relevant since the current capital level must meet the required ratio. Next to

that, the αi represents the entity fixed effects and the λt represents the time fixed effects.

Before addressing the second regression equation, some leading regression assumptions are presented. Four assumptions are relevant for the included fixed effects regressions. At first, there can be assumed that the error term εi,t has a conditional mean zero: E(εi,t |Xi,1, Xi,2,.., Xi,t, αi). Moreover, the second

(26)

distributed draws from the joint distribution. At last, we assume that large outliers are unlikely and that there is no perfect multicollinearity.

ΔL/Li,t = αi + β1 * ΔETAi,t + λt + εi,t (2)

The second regression equation is used to test whether the change in the ETA ratio correlates with the change in the loan growth of a bank. Consequently, this equation is used to find the answer on the first hypothesis. The only difference with the first regression equation is the coefficient on the β1. The

lagged value of the equity-to-assets ratio is replaced by the change in the ETA ratio. The change in the ETA ratio is calculated by: ETAT – ETAT-1.

ΔL/Li,t = αi + β1 * ETAi,t-1 + β2 * ΔETAi,t + λt + εi,t (3)

Regression three combines the drivers of regression (1) and (2) in one regression. So, the value of the equity-to-assets ratio at t-1 and the change in the ETA ratio are enclosed in the equation. Regression (3) is performed to see how the statistical significance of the coefficients changes when both variables are included. Moreover, by showing all underlying regressions, the statistical validity of the outcomes can be improved.

ΔL/Li,t= αi + β1 * ETAi,t-1 +β2 * ΔETAi,t + β3 * Di,t + β4 * (Di,t * ΔETAi,t) + φ1 * Ln(Ai,t)+ φ2 * Ri,t + λt + εi,t (4)

The fourth regression focuses on the main question whether banks that are close to the threshold behave differently from the banks that are far from the threshold. In contrary to the first equations, this equation is extended with a couple of dummies, macroeconomic variables and control variables to optimize the outcomes. Equation (4) is included for the same reason as equation (3) is included: by showing all regressions step by step, the validity of the outcomes can be analysed.

As mentioned in an earlier stage, the ΔETAi,t represents the absolute change in the

equity-to-assets ratio of bank i from last year. The β1 shows the relation between the change in loan growth and

the equity-to-assets ratio at t-1 of a bank. The β2 shows the relationship between the epsilon and the

change in the ETA ratio. Moreover, a binary dummy Di, t is included to show if banks are close to the

threshold (=1) or far from the threshold (=0). In general, the regulatory capital ratio of eight per cent is used in this study for addressing the threshold. The banks that have an equity-to-assets ratio between the 8 and 8.67 per cent are seen as ‘Banks that are close to the regulatory capital level of 8%’.The interaction term between the change in the ETA ratio and the dummy variable shows whether a change in the capital ratio has a stronger effect on the banks that are close to the threshold. If the β4 is significant this means that banks that are close to threshold react differently than the banks

that are further away from the threshold. Some important control variables that are included are the natural logarithm of the total assets Ln (Ai,t) and the risk indicator Ri,t-1. The natural logarithm of the

total assets is incorporated to control for the size of the European banks. As larger banks will have more difficulty with increasing their loan growth (%) on average, there need to be controlled for this