Procyclicality of Dutch Banks Leverage

Transcription

1 Procyclicality of Dutch Banks Leverage Did banks in the Netherlands display a procyclical policy? W.J.N. van Diepen July 216 MSc thesis Economics Supervisor: Dirk Veestraeten Second reader: Kostas Mavromatis

2 Abstract This study examined the degree of procyclicality of Dutch banks leverage. This research will give a clear overview whether Dutch banks displayed a procyclical policy in the period from 21 to 214. In this paper, procyclical leverage means that leverage growth increases (decreases) when asset growth increases (decreases). The regression results suggest that asset growth has a positive effect on leverage growth, so there is a procyclical leverage. The coefficient of asset growth is positive (.97) and significant. The results do not suggest that GDP growth has effect on leverage growth. This paper contributes to the discussion about the sustainability of banks. In the paper I focus on the Dutch banks, because there is little research about leverage procyclicality of Dutch banks. Statement of Originality This document is written by Student Wietse van Diepen who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3 Content 1. Introduction Dutch banking sector characteristics and banking regulation Leverage: What it is and how it works Empirical analysis Data and descriptive statistics Econometric analysis Regression model Results Conclusion References Appendix... 23

4 1. Introduction The credit crisis began in 27 when US home prices started to drop. Since this price decline, a part of the homeowners were unable to make timely payments on their mortgages. As a result, the risk of mortgage default increased. Especially homeowners with a bad credit history or with a undocumented income had an increased risk of mortgage default. This kind of mortgages were called subprime mortgages. By 27, there were 32 billion dollars worth of these kind of mortgages. These mortgages were packaged by banks and were sold to other banks and investors. These bundled mortgage loans declined rapidly in value in 27 and 28. The financial institutions which had bought these products had to write off tens of billions of dollars. This created a huge distrust between banks, because it was not clear which institutions had bought these packages. As a result, the banks did not feel safe to lend money to each other. The interbank lending market rates rose very fast. On September 28, Lehman brothers filed for bankruptcy. The collapse of Lehman Brothers contributed to the erosion of close to 1 trillion dollar in market capitalization from global equity markets in October 28. Worldwide, central banks had to intervene to prevent a total collapse of the financial system. The crisis in the financial sector had also an impact on the real economy. The economy of the United States and several economies in the Eurozone went into recession. To halt the downturn of the economy, many countries decided to conduct an active stimulation. This stimulation included lower interest rates to stimulate firms and households to borrow more money. Central banks hoped that this would give an incentive to invest and consume more. Also, central banks bought government bonds and corporate bonds to lower bond rates. The Federal Reserve alone bought more than three trillion dollar of assets. Despite this stimulation, they could not avoid a deep recession in many countries. During the crisis many banks were quite fragile, because these banks suffered huge losses. Therefore, they were likely to become insolvent. Many banks needed a bail-out or were nationalized. The question is why banks are vulnerable to financial shocks and how this problem can be solved. One of the main reasons for this vulnerability was the high leverage ratio of many banks. The equity relative to the total assets was too small to absorb negative 1

5 financial shocks. For that reason the Dutch government (partially) nationalized three major banks: ING, ABN AMRO and SNS Bank. These three banks built up huge losses when some products, such as United States subprime mortgages, turned out less worthy than they were before the crisis started. To absorb these losses without a bankruptcy, they were bailed out by the government or were (partly) nationalized. When a bank was nationalized, the government gave these banks enough money to avoid a bankruptcy. A failure of one of these banks would be a disaster for the Dutch economy. This is because there were just five large banks in the Netherlands. The three banks as mentioned above and the Rabobank and Fortis. These banks are too big to fail. So it is very useful to investigate these banks. They have a huge influence on the Dutch financial markets and they are important for the Dutch economy. Therefore it is important to know how the leverage ratios of these Dutch banks changed. The main question of this paper is: Is the leverage ratio of Dutch banks procyclical? The definition of procyclicality in this paper is: when total assets increase, the leverage ratio increases also and vice versa. So in this paper procyclicality is not directly linked to the business cycle. The leverage ratio of a bank is L = A/E, where L is leverage, A is total assets and E is equity. In this paper I will use a non-risk-based leverage ratio. The non-risk-based leverage ratio means that the assets are not risk weighted. That means that these kind of assets are included in the calculation of total assets for 1 percent. An example of an asset which is not included for 1 percent is specific government bonds. Government bonds of countries with a credit rating of AAA, that means that countries has a very low default risk, had a risk weighting of percent. As a result, these assets were subtracted from total assets for purposes of calculating, for example, the leverage ratio. Another example are residential mortgages. These loans have a risk weight, for example, of 8 percent. When a bank has 1 billion euro of these loans on its balance sheet, then the bank has a risk weighted asset on its balance sheet of 1 *.8 = 8 billion euro. So it can use 8 billion to calculate the leverage ratio. If this is the only assets on the balance sheet, then the bank has 8 billion dollar of risk weighted assets. When we assume that the bank has an equity of 1 billion dollars, the risk weighted leverage ratio is 8 / 1 = 8. When a non-risk-based leverage ratio would be used, the leverage ratio would be 1 / 1 = 1. So a risk-based leverage ratio has a decreasing 2

6 effect on the leverage ratio. It was common to use this way of leverage calculation, but the crisis of 28 has shown that this way of leverage calculation underestimate the risks of insufficient capital. When it seems that the default risk of specific assets is higher than expected, then there is too little capital to absorb losses. Then it is possible that a bank will go bankrupt. Therefore the Basel Committee of Bank Supervision (BCBS), the institution which develops banking guidelines, uses in its latest regulatory framework a non-risk-based leverage ratio (BIS, 21). After the crisis the leverage ratios were studied by policymakers and academics. Since 28, several papers have been written about the problems and risks of increasing leverage ratios. Academics and policymakers studied banks in many countries and regions and the general conclusion was that leverage behaved in a procyclical manner. However, there were a few differences. There was a difference between the procyclicality of commercial banks and investment banks. In Europe, only the investment banks showed a procyclical leverage (Baglioni et al., 212). Also Adrian and Shin (21) concluded that there was a strong correlation between growth of the leverage ratio and the growth of the balance sheet. They studied the American investment banks in the period from 1994 to 28. That was the period before the credit crunch of 28. In this period the leverage ratio of banks increased and that was one of the main reason of the financial crisis. However, it is also important how the growth of assets was financed. When total assets increase, they have to be funded by the liability side of the balance sheet. It matters which kind of funding banks use. When banks use retail deposits as funding, the leverage ratio is less procyclical. This kind of funding cannot be changed quickly, so the balance sheet cannot increase quickly. Therefore the leverage ratio cannot increase quickly. When banks use primarily wholesale funding, the leverage ratio is more procyclical. This is because wholesale funding can be changed quickly. Therefore the balance sheet can be increased or decreased rapidly (Damar, Meh and Terajima, 213). Nuño and Thomas (214) have had the same conclusion: banks which were financed with long term debt and retail deposits could not increase their balance sheet very rapidly, so in that situation the leverage ratio is less procyclical. 3

7 Another factor that could play an important role in changing the leverage ratio is the use of fair value accounting. Fair value accounting means that assets are valued on the balance sheet at the price they have on the balance sheet date. By using fair value accounting the balance sheet can increase or decrease when prices change. So this changes the leverage ratio. Laux and Rauter (215) found no evidence that fair value accounting caused a procyclical leverage ratio. In several countries academics and policy makers studied the degree of procyclicality of leverage ratios. They studied banks across whole Europe, but there is not a study about only the Dutch banks. The Netherlands has a huge banking sector relative to GDP. Before the crisis the balance sheets of several banks increased very rapidly. According to aforementioned studies this could have increased the leverage ratio. In addition, ABN AMRO (in 28) and SNS Bank (in 213) were nationalized and ING Bank received a capital injection of 1 billion euro in 28, because the capital buffers were insufficient to absorb the (expected) losses. It therefore makes sense to examine the leverage ratios of Dutch banks from 1993 to 214. This may contribute to understand the vulnerabilities of banks during the crisis of 28. The structure of this thesis is as follows. Section 2 presents an overview of the Dutch banking sector. In section 3, I will explain how leverage works. I will also use some examples of balance sheets. Section presents the data and descriptive statistics, section explains the methodology which I will use and section presents the results of the regression analyses. I conclude in section 5. 4

8 2. Dutch banking sector characteristics and banking regulation In this chapter, I will give an overview of the Dutch banking sector. I will briefly explain the most important characteristics of this sector and how these characteristics have changed. Furthermore, the regulation from the government, the central bank and the Basel Committee of Bank Supervision (BCBS) will be discussed. 2.1 Overview Between 1945 and 196, each Dutch bank had its own focus. There were banks for mortgages, banks for loans to large firms, banks for loans to small and medium firms, etc. However, in the 6s the first merger wave commenced. Small banks were acquired or two or more banks merged. As a result, banks started operating in more than one type of banking. The effect of this development was that banks became more similar. Today, specializations of a bank are still visible. In this paper, I will focus on five banks: ING Bank, Rabobank, ABN AMRO, SNS Bank and Fortis. ING Bank is part of the ING Group. In 1991, ING Group was created by a merger between NMB-Postbank and Nationale-Nederlanden. Between 1991 and 27, the group took over several banks and insurance companies and became a worldwide operating bank. The Rabobank Group is created in 1972 after a merger between Coöperatieve Centrale Raiffeisen-Bank and Coöperatieve Centrale Boerenleenbank. The Rabobank Group took over several banks and insurance companies. The ABN AMRO Bank was created in 1991 after a merger between ABN and AMRO Bank. Also the ABN AMRO Bank has taken over several foreign banks and as a result, it became the largest foreign bank in the United States in the 9s. However, the bank was taken over by a consortium including Fortis in 27. In 28, when Fortis suffered a huge loss and risked a bankruptcy, most of the former ABN AMRO Bank was nationalised by the Dutch government. The fourth bank I will use in this paper is SNS Bank. This bank was created in 1817 and was a part of SNS Reaal for a long time. Fortis is the last of the five banks in this study. Fortis was created by a merger between the Dutch insurance company AMEV and the Dutch VSB Bank. This was followed by numerous acquisitions in the Benelux. In 27, Fortis took over ABN AMRO Bank, but one year later 5

9 Fortis was almost bankrupt and the Dutch part of Fortis was taken over by the Dutch government. In the 8s, the Dutch banking sector consisted largely of six mid-sized banks. At the end of the 8s and the early 9s, the banking industry saw several acquisitions which caused a high market concentration. As a result, the Dutch banking sector became one of the most concentrated banking sectors of Europe (DNB, 214). Foreign banks play a small role in the Netherlands. About 1 percent of the total assets was held by foreign banks at the end of 212. Compared with other European countries this percentage is low. There are several reasons for this. One of them is the high loan-to-value ratio for Dutch mortgages. Foreign banks regarded high loan-to-value ratios as too risky. At the end of 211 the largest three Dutch banks held 75 percent of all Dutch banking assets (Ministerie van Financiën, 213). 2.2 Basel Accords Before 199, banks were tested for sustainable bank policy and undesirable developments of the credit system. Thereby, the central bank focused on the liquidity and the solvability and tried to avoid undesirable market concentration. On 1 January 199, the central bank changed its policy. The policymakers assumed that competition between banks would be guaranteed by the entry of foreign banks. However, these foreign banks barely entered the Dutch market. In contrast, Dutch banks have become very active abroad. As a result, several Dutch banks grew to major internationally operating banks, which were too big to fail during the crisis in 28. From 199 to 28, the total assets of Dutch banks increased from 44 billion to 29 billion euro, and decreased to 24 billion euro after the crisis. The BCBS in Basel is an institution which creates guidelines for the banking sector. These guidelines are called Basel Accords. Central banks have the responsibility to implement these guidelines. The first Basel Accord, known as Basel I, was issued in This agreement included that banks must maintain capital equal to at least 8 percent of their risk weighted assets, whereby at least 4 percent of the highest quality (Tier 1) (Smolders, 21). However, the default risk of some assets was estimated too low. Therefore banks bought risky assets 6

10 with low capital requirements. As a result, banks had too little capital relative to assets. During an economic downturn, banks could become insolvent, then banks can go bankrupt. To solve that problem, the BCBS presented Basel II in 24. This agreement was implemented in the European Union in 28 and was based upon three pillars: minimum capital requirements, supervisory review and market discipline. The minimum capital requirement of 8 percent is maintained, but it is more risk sensitive (Smolders, 21). That means that policymakers focus more on the risks of bank assets. They make a distinction between credit risk, market risk and operational risk. Banks may use internal risk models, but the supervisor assesses whether the models are a true reflection of reality. The financial crisis of 28 showed that Basel II was no longer sufficient. The capital held by banks was in some cases not sufficient. This was partly due to financial innovations and the complexity of some products. Therefore it was difficult to estimate the risks of some products and how much capital banks need to absorb possible losses. Also off balance sheet exposure was a cause of too less capital. As a result, Basel III was introduced in the Netherlands on 1 January 214. This happened step by step to avoid undesirable effects. The Basel III Accord consist of the following measures: increase of the quality of capital, a better risk coverage, the introduction of a leverage ratio, a reduce of the procyclicality and addressing systemic risk. Especially the introduction of the leverage ratio is important in this paper, because I will study the leverage ratio of the Dutch banks. A Basel Accord includes for the first time a leverage ratio. Banks are required to hold at least 3 percent of equity relative to their total assets (BIS, 213). The Dutch supervisor requires that banks have a minimum capital ratio of 4 percent. The definition of leverage used by the BIS is a little bit different from the definition used in this paper. The BCBS uses the Tier 1 Capital as equity measure and in this paper the total equity is used as equity measure. To measure the total assets the BCBS uses both the on balance sheet assets and the off balance sheet assets. In this paper I used only the on balance sheet assets, because the data about the off balance sheet assets was not available. 7

11 3. Leverage: What it is and how it works In this section, I will explain the definition of leverage and the causes and effects of procyclical leverage. I will explain it on the basis of some examples of balance sheet. The leverage ratio of a bank is L = A/E, where L is the leverage, A the total assets and E the equity. Total assets includes all assets of a bank, such as cash, financial assets, intangible assets, loans and property. Off balance sheet assets are not included in this study, because this data was not available. Equity includes shareholder s equity and minority interests. Subordinated debt can be seen as equity, but it will be seen as debt in this paper. Procyclical leverage implies a positive relation between the growth of leverage ratio and the growth rate of assets. Below is a simplified balance sheet of a bank: Assets Liabilities Total Assets 2 Equity 1 Debt 19 The leverage ratio of this bank is 2 / 1 = 2. This means that the total assets are financed by equity of 1. The other part of the assets are financed by debt. So the total assets are 2 times the size of the equity, which means that the leverage ratio is 2. When there are losses on the asset side of the balance sheet, then the equity will reduce. Losses can include, for instance, non performing loans and declining real estate prices. When the losses are bigger than 1, the bank has no equity of even negative equity. During a economic boom the prices of assets generally increase. For example, commercial real estate. When economic growth rises, the demand for commercial real estate rises also. As a result, the price of this real estate increases, so the value of this real estate on the balance sheet increases also. Another example are government bonds. When the demand for bonds increases, the interest on bonds declines. The price of bonds change in the opposite direction. So when the price increases, the value of a bond on the balance sheet rises. When banks use fair value accounting or marked-to-market accounting, then a change in asset prices cause an immediate change in the value of the total assets on the balance sheet. With this accounting method the price of assets on the balance sheet is the market price at the balance sheet date. This can be very volatile, so the balance sheet can expand 8

12 and shrink very quickly. In this example, we assume that the value of these assets increases with 5. Because of this price increase the equity increased by 5. When the asset side of the balance sheet increases, the liability side has to increase also. Debt does not change, so equity has to increase. Price reductions reduce equity and price increases increase equity. The result is the following balance sheet: Assets Liabilities Total Assets 25 Equity 15 Debt 19 The price increase has changed the leverage ratio: 25 / 15 = 13,67. This ratio has declined from 2 to 13,67. This means the bank does not actively manage its balance sheet. Actively managing your balance sheet means that banks want to keep their leverage ratio at a certain level. Banks can do this by keeping the ratio of equity and debt equally. This did not happen in above example. As a result, the leverage ratio decreases. When the value of the total assets decreases, the leverage ratio will increase. In above example the bank leverage was not procyclical. In the following example the bank leverage is procyclical. That means that debt increases more than equity. Equity has risen with 5 percent on the previous balance sheet. To achieve a procyclical leverage ratio, debt has to increase with more than 5 percent, as shown in the following balance sheet: Assets Liabilities Total Assets 315 Equity 15 Debt 3 Debt has increased with 11 units, or 58 percent. The leverage ratio is 21 now. By actively managing the balance sheet the ratio has increased from 2 to 21. The impact of a rising leverage ratio is that the equity/debt ratio declines, so there is less equity relative to total assets. When there are losses on the asset side, then there is less equity (relative to total assets) to absorb the losses. Therefore the chance of insolvency becomes bigger, that means that the amount of debt is larger than the value of the total assets. Then the bank has no equity. 9

13 The benefits of a high leverage ratio is the possibility of an increasing return on equity. When a bank pays 6 percent interest on its debt on average, and the total assets gives an 8 percent return, the difference (2 percentage points) is available to pay out as dividend for the shareholders, the providers of the equity. During an economic boom the asset prices increases. When the total amount of debt remains the same and the equity increases (because of the increasing value of assets), the leverage ratio will decline. To restore the initial leverage ratio, banks have to increases their debt. With this debt they can buy some assets. Therefore the banks will expand their balance sheet. In addition, during an economic upturn there are more investment opportunities, so many banks buy more assets and increase their balance sheet. So the demand for assets increases and this has an inflationary effect. Banks may want to keep their leverage ratio constant, because a lower leverage ratio means less profit for the shareholders. That is explained above. The inflationary effect is shown by Adrian and Shin (21) in the following figure. If financial markets are not perfectly liquid so that greater demand for assets tend to put upward pressure on their price, then an increase in demand of assets will increase the asset prices. Then the leverage will decline, and the bank will expand its balance sheet to increase the leverage. So this creates an asset price boom. When the leverage is procyclical, this effect is even stronger. This mechanism works exactly in reverse in downturns. Figure 4: Adjustment of leverage Source: Adrian and Shin (21) 1

14 4. Empirical analysis In this chapter I will discuss the empirical part of this study. Section 4.1 explains the methodology. This includes two parts. Section presents the descriptive statistics and the data and presents the econometric analysis. This analysis includes two parts: section explains the regression model, and section presents the results. 4.1 Methodology This section describes the methodology used in this paper. This consists of two parts. The first part consists of figures and tables. This is data of five Dutch banks which are used in this study. The second part is a regression analysis for three banks, which are ING Bank, Rabobank and SNS Bank. This part includes three banks instead of five banks, because there is only data of three banks for the period from 21 to 214. For the other two banks, there is no data for this entire period Data and descriptive statistics I obtained the bank-level data from the banks. In their annual reports I could find the data about the equity and the total assets. The total assets of a bank, including foreign subsidiaries, includes all assets of a bank, without risk weighting. The off balance sheet assets are not included in this study. The equity of banks consists of shareholder s equity and minority interests. Subordinated loans are not included, when these loans were mentioned separately. The GDP data is obtained from the database of the Centraal Bureau voor de Statistiek. The data which is used in this study comes from the largest five banks in the Netherlands, which held collectively more than 8 percent of the Dutch banking assets. It was impossible to use data of the five largest banks that covers the period from 199 to 214. That data was not available. This study examined whether bank leverage is procyclical. However, the data does not only consist of bank data. The data of ING Bank and SNS Bank is only bank data, but the data of 11

15 Rabobank, ABN AMRO and Fortis is both bank data and data of the insurance part of the company. Unfortunately, these data could not be separated. So this study is not purely about bank activities. I have data from the ING Bank, Rabobank and SNS Bank on equity and total assets for the period from 21 to 214. From ABN AMRO, I have data for the period from 1993 to 27 and from Fortis I have data for the period from 1997 to 27. This is yearly data. Five banks are examined in this study. But these banks have data of different time periods. So the data of three banks is useful in a regression analysis. In this part I will describe the data of all banks. This data consists of some statistics. Figures 1a-1e shows the leverage ratio and the total assets of each bank. The data for each bank covers different time periods. These figures give a good overview of the leverage and total asset of each bank. All these five banks generally show a strong growth of total assets. ING Bank is the only bank which shows a drop in the amount of assets during and after the financial crisis, as shown in figure 1a. ING Bank had to divest several division worldwide as part of the approval by the European Commission in November 29 for the State support from the Dutch government during the financial crisis. The support was necessary to avoid a bankruptcy. This was the most important reason for the decline in total assets. (ING Bank, 215) The leverage ratio increased to almost 4 before the crisis and declined to almost 2 after the crisis. So the financial crisis had a large impact on both the total assets and the leverage ratio of ING Bank. The other two banks (SNS Bank and Rabobank), which are used in the regression analysis, show a growth in total assets before as well as after the crisis. They had only a small drop in assets after 212. SNS Bank had a huge drop in leverage, from 6 in 212 to 25 in 214 (see figure 1b). This was caused by both an increase in equity and by a drop in total assets. In 213, SNS Bank was nationalized, so that increase in equity was caused by a capital injection of the government. Figure 1c (Rabobank) shows a relatively stable leverage ratio. The financial crisis had little impact on this ratio. 12

16 Total assets Leverage ratio Totaal assets Leverage ratio Total assets Leverage ratio Figure 1a: Leverage ratio (total assets / equity) and total assets (in billion euro) ING Bank, Figure 1b: Leverage ratio (total assets / equity) and total assets (in billion euro) SNS Bank, Total assets Leverage Total assets Leverage Source: own calculations on the basis of balance sheet data of ING Bank Source: own calculations on the basis of balance sheet data of SNS Bank Figure 1c: Leverage ratio (total assets / equity) and total assets (in billion euro) Rabobank, Total assets Leverage Source: own calculations on the basis of balance sheet data of Rabobank Figure 1d shows the graph of ABN AMRO. The data covers the period from 1993 to 27. The total assets increased in this period and the leverage ratio fluctuated with a small rising trend. There is no data over the period after the crisis, so it is not possible to compare data before and after the crisis. Figure 1e shows the graph of Fortis. This data covers the period from 1997 to 27. The years before the crisis, from 24 to 27, the leverage ratio declined. Before that period, the leverage ratio generally increased. The total assets increased the whole period of time. That is the same trend which can be seen by other banks. 13

17 Total assets Leverage ratio Total assets Leverage ratio Total assets Leverage ratio Figure 1d: Leverage ratio (total assets / equity) and total assets (in billion euro) ABN AMRO, Figure 1e: Leverage ratio (total assets / equity) and total assets (in billion euro) Fortis, Total assets Leverage Total assets Leverage Source: own calculations on the basis of balance sheet data of ABN AMRO Source: own calculations on the basis of balance sheet data of Fortis Figure 1f is a graph about the leverage ratio and total assets of the three banks together which are used in the regression analysis. That gives a clear picture of the Dutch banking sector, because these banks held more than 8 percent of total of Dutch banks assets. After the crisis, total assets did not grow anymore, but they declined a little bit. The leverage ratio of these banks declined also. Figure 1f: Leverage ratio (total assets / equity) and total assets (in billion euro) ING Bank, Rabobank and SNS Bank, Total assets Leverage Source: own calculations on the basis of balance sheet data of ING Bank, SNS Bank and Rabobank Figures 2a-2e shows the correlation between the asset growth rate and the leverage growth rate for each bank. These figures show the scatter plots. The line in these scatter plots is the 14

18 Leverage Leverage.2 line of best fit, also called the regression line. This line shows the correlation between asset growth and leverage growth. That does not mean that there is a causal relationship. The regression line of ING Bank, Rabobank, ABN AMRO and SNS Bank is an increasing line, so there is a positive relationship between the asset growth and the leverage growth. When the asset growth increases, the leverage growth will generally increase as well. Fortis is the only bank with a decreasing regression line. This means that when the asset growth increases, the leverage growth decreases. Figure 2a: Yearly change in leverage and total assets of ING Bank (22-214) Total Assets Fitted values?leverage (LN) Source: own calculations on the basis of balance sheet data of ING Bank Figure 2b: Yearly change in leverage and total assets of Rabobank ( ) Total Assets Source: own calculations on Fitted the basis values of balance?leverage sheet (LN) data of Rabobank 15

19 Leverage Leverage Leverage.5 Figure 2c: Yearly change in leverage and total assets of SNS Bank (2-214) Total Assets Source: own calculations on Fitted the basis values of balance?leverage sheet (LN) data of SNS Bank Figure 2d: Yearly change in leverage and total assets of ABN AMRO ( ) Total Assets Source: own calculations on Fitted the basis values of balance?leverage sheet (LN) data of ABN AMRO Figure 2e: Yearly change in leverage and total assets of Fortis ( ) Total Assets Source: own calculations on Fitted the basis values of balance?leverage sheet (LN) data of Fortis 16

20 Figures 3a and 3b show the correlation between these variables for the three banks which are used in the regression analysis and for the full sample. Both graphs show a strong positive correlation between asset growth and leverage growth. In the regression model, which is described further below, I will look if there is a definite relationship between these variables. I found the same correlation, between leverage growth and asset growth, in other studies. Damar, Meh and Terajima (213) found a strong correlation between these variables. They studied Canadian banks and their study uses data that covers the period from 1994 to 29. The procyclicality was higher among banks that are more dependent on wholesale funding. Also Laux and Rauter (215) found this correlation, they studied US commercial and savings banks. Their study used data that covers the period from 199 to 213. Figure 3a: Yearly change in leverage and total assets of ING Bank, Rabobank and SNS Bank (22-214) Source: own calculations on the basis of balance sheet data of ING Bank, Rabobank and SNS Bank 17

21 Figure 3b: Yearly change in leverage and total assets of all banks ( ) Source: own calculations on the basis of balance sheet data of all banks The data used in this paper covers different time periods. There was no data for ABN AMRO available after 27, because of a acquisition in 27. Therefore there was no suitable data. Also, there was no data for Fortis available after 27, because this bank was acquired by the Dutch and Belgian governments during the crisis in 28. As a result, there was no connection anymore between the data before 28 and after Econometric analysis Regression model The second part of the methodology includes the regression analysis. I will use the following banks for this analysis: ING Bank, Rabobank and SNS Bank. There is data available for the period from 21 to 214. Table 1 describes all the variables which are used in this paper. The model which I will use in this study is a time-fixed random effects model. This model measures whether the leverage growth is procyclical. That is the main goal of this study. I will use a panel regression analysis. I estimate the following regression: Δ Leverage i,t = α + β Δ Total Assets i,t + γ Δ GDP t + δ Leverage i,t-1 + ε i,t 18

22 The dependent variable is Δ (Leverage) i,t. This is the leverage growth of bank i at year t and is defined as ln(leverage i,t ) ln(leverage i,t-1 ). The most important independent variable is Δ Total Assets i,t. This is the total asset growth and is defined as ln(total assets i,t ) ln(total assets i,t-1 ). In this study these variables are the main variables, because I want to know the effect of asset growth on leverage growth. That is the way to test procyclicality. The main coefficient of interest is β. The coefficient captures the relationship between changes in leverage growth and changes in total asset growth. It is the degree of leverage procyclicality. If this coefficient is positive, leverage is procyclical. The regression model will be extended to include some other variables. The first variable which I will add is Δ GDP t. This variable controls for macroeconomic conditions, so whether macroeconomic conditions have an effect on leverage growth. Δ GDP t is equal to the real GDP growth per year. This variable is for all banks the same. Another variable which will be added to the regression model is leverage i,t-1. This variable measures the leverage at the start of a period, so at the start of a year in this case. This variable measures whether the level of leverage has an effect on leverage growth. In this model there are yet two other variables: a constant and an error term. These are indicated by α and ε i,t. I have used a Hausman test to decide between an individual specific fixed effects model or an individual specific random effects model. This test showed that an individual specific random effects model is the best model. The result of this test was.53, this is more than.5. When the value is more than.5, the random effects model is the best option. When the value is less than.5, the fixed effects model is the best option Results Table 2 shows the output of the regression analysis. The most important coefficient is β. In all four regression equations the coefficient of Δ Total assets, β, is statistically significant at a 1 percent level. Without control variables, the value of β is.917. When I add the control variables leverage t-1 and Δ GDP, the value of β decreases to.97. So the variables leverage t-1 and Δ GDP have a little effect on the dependent variable Δ Leverage. The variable leverage t-1 and the constant are significant, but the variable Δ GDP is not significant at a 5 19

23 percents level. This means that GDP growth has no effect on the leverage growth in this study. The value of β, which is shown in regression equation [4], is,97. That means there is a positive correlation between the change of total assets and the change of leverage. A stronger increase in assets, causes a stronger increase in leverage. When total assets increase 1 percent faster, then the leverage increases 9.7 percent faster. The p-value is.5, so the result is significant at a 5 percents level. The value of the coefficient of leverage t-1 is -.5. This means that a higher leverage ratio has a declining effect on leverage growth. This effect is not very large, but the results of the regression analysis are significant at a 5 percent level. So banks with a high leverage ratio show a slightly less procyclical pattern. 2

24 5. Conclusion I studied the degree of procyclicality of the leverage of the Dutch banking sector in the period from 21 to 214. I used yearly data about equity and total assets. Three banks are used in the regression analysis, ING Bank, SNS Bank and Rabobank. Two other banks, ABN AMRO and Fortis, are used to give a more comprehensive overview of the Dutch banking sector. A disadvantage of this regression analysis is the low number of observations. The total number of observations are 39. The main question was: Did banks in the Netherlands display a procyclical policy? My findings suggest that Dutch banks displayed a procyclical policy. To come to this conclusion, I used a time-fixed random effects model. The results of this model were very significant. Therefore, I can conclude that the leverage growth increases, when the total asset growth increases. That is the definition of procyclicality in this paper. The variable leverage i,t-1 was negative and significant, that means that the leverage growth is smaller when the leverage ratio is higher. I did not find evidence that Δ GDP has effect on the leverage growth. The general conclusion of these results is that Dutch banks displayed a risky policy when asset growth increases. A risky policy means an increase of the leverage growth. This is possibly one of the reason for the onset of the financial crisis. 21

25 References Adrian, T. and Shin, S.H. (21), Liquidity and Leverage, Federal Reserve Bank of New York Staff Reports, no Baglioni, A., Beccalli, E., Boitani, A. and Monticini, A. (212), Is the leverage of European banks procyclical?, Catholic University of Milan. BIS (21), Basel III: a global regulatory framework for more resilient banks and banking systems, December 21 BIS (213), Revised Basel III leverage ratio framework and disclosure requirements, Consultative Document, June 213 Damar, H.E., Meh, C.A. and Terajima, Y. (213), Leverage, balance-sheet size and wholesale funding, Bank of Canada. DNB (214), Perspective on the structure of the Dutch banking sector, DNB Study. ING Bank (215), Herstructurering ING Bank, Laux, C. and Rauter, T. (215), Procyclicality of US Bank Leverage, Journal of Economic Literature. Ministerie van Financiën (213), Kabinetsvisie Nederlandse bankensector, Nuno, G. and Thomas, C. (214), Bank leverage cycles. Smolders, N. (21), Impact van de nieuwe kapitaalvoorstellen uit Bazel, Rabobank, Themabericht 21/6 22

26 Appendix Table 1: Definition of regression variables Variable Total Assets i,t Leverage i,t GDP t Definition Book value of all assets on the balance sheet of bank i at the end of year t Total Assets i,t / Total Equity i,t Real Dutch groos domestic product at the end of year t Δ Leverage i,t ln(leverage i,,t ) ln(leverage i,t-1 ) Δ Total Assets i,t ln(total Assets i,t ) ln(total Assets i,t-1 ) Δ GDP t (GDP t / GDP t-1 ) 1 Table 2: Leverage regressions This table shows the results of the regression model. The dependent variable is the yearly growth rate of leverage (Δ Leverage). The independent variables are the yearly growth rate of total assets (Δ Total Assets), the lagged leverage (Leverage t-1 ) and the real GDP growth (Δ GDP). Significance is indicated by: *** <.1, ** <.5 and * <.1. [1] [2] [3] [4] Δ Leverage Δ Leverage Δ Leverage Δ Leverage Δ Total Assets.917***.917***.867***.97*** (.293) (.342) (.28) (.326) Leverage t-1 -.6** -.5** (.3) (.3) Δ GDP (1.648) (1.569) Constant -.56* -.536* (.31) (.33) (.8) (.81) 23