The effect of Basel II on listed companies
P.J. Hagoort1University of Amsterdam, Amsterdam, The Netherlands July 6, 2015
Abstract: This paper examines the effect of Basel II on the leverage ratio of listed companies in the Netherlands. Using a fixed effects regression, it can be concluded that Basel II does not have a significant negative influence on the leverage ratio of listed companies. There are different factors important in explaining the variation of the leverage ratio of the three different indexes in the Netherlands.
Keywords: Basel II; Leverage ratio; Companies; AEX; AMX; AScX
Index
1. Introduction 3
2. Literature review 4
2.1 History of financial intermediaries 4
2.2 Regulation and Basel I 6
2.3 Basel II 7
3. Factors that influence the leverage ratio 9
4. Research hypotheses 11
5. Methodology 12
5.1 The variables 13
5.2 The model 15
6. Research Results 16
6.1 Testing the assumptions 16
6.2 The regression results 17
6.3 Conclusion 26
7. Discussion 27
8. Literature 29
1. Introduction
A company that wants to invest in a new project has several options to finance it. The project can be financed internally (internal financing), or it can be financed externally. For external finance, two options arise: option one is to go to a bank and apply for a loan, option two involves issuing new shares to raise capital. When a company decides to raise capital through private debt, a bank loan, its financial statements will be examined carefully. The bank will assess the risk attached to this new loan. Capital requirements oblige the bank to assign a risk-weight to the loan. These capital requirements have been drawn up by the Basel Committee on Banking Supervision and are combined together as the Basel Accord. The first Basel Accord was implemented in 1988. Basel I obligated banks to hold a minimum of eight per cent of its risk-weighted assets in capital. Assets were grouped into five categories according to their risk. Each group was assigned a weight and accordingly a weighted value (Mishkin et al., 2013). In 1999 new proposals were put forward and the existing accord was reformed to the second one: Basel II, which is based on three pillars. The asset groups of Basel I were imprecise; regardless of a company’s risk, a commercial loan was assigned a weight of 100 per cent. No distinction was made among companies and the risk they carry. The first pillar of Basel II does distinguish between risks; market risk, credit risk and operational risk, linking capital requirements more closely to actual risk (Mishkin et al., 2013). In January 2008 Basel II was officially implemented in the European Union. This new accord raised questions about the effect on lending.
Heid (2007) has investigated what the effect of Basel II would be on the macro economy. He concluded that it had a pro-cyclical effect on lending. This means that if the economy is in a downturn the lending supply will fall and if the economy is in a boom the lending supply will increase. He also found that if a company’s outlook weakens, negatively affecting the credit quality, the bank granting credit to this company faces higher capital charges and decreases its loan supply.
Schmidt, Hackethal & Tyrell (1997) have investigated the importance of banks in the financial sector. They found that banks, in the years prior to Basel II, were shifting their focus from collecting and investing towards investing alone. Nonbank financial intermediaries, such as insurance companies, were focusing more on collecting savings and passing these on to banks. Basel I made no distinction
between the risks of companies. This is no longer the case for Basel II; it specifies more categories of assets with different risk weights. These new categories and risk weights might increase the amount of capital that a bank has to hold. Since banks were focussing more on issuing loans, Basel II might have a large effect on the supply of loans.
This study will examine what the effect of Basel II is on the leverage ratio of listed companies. The leverage ratio is defined as the total amount of liabilities divided by the total amount of shareholders’ equity. As Heid (2007) concluded, as an effect of Basel II, the loan supply might decrease, making it more difficult for companies to obtain credit. Companies might turn to other sources of finance such as internal financing and equity financing. Control variables affecting the leverage ratio such as the interest rate, relative firm size and GDP growth rate will be included in the model.
Many studies have investigated the effect of Basel II on loan supply of banks but a research about the effect of Basel II on companies, and consequently the economy, has not yet been done. This study will contribute to the investigation of the economic effects of bank regulation. The economic effects are important to consider because it influences the stability of the economy, the employment and consequently the prosperity of a country.
2. Literature review
2.1 History of financial intermediariesThe financial system of a country has not always been an important topic. A number of years ago, financial institutions and financial contracts did not play a part in the theoretical structure of general equilibrium analysis (Schmidt, Hackethal & Tyrell, 1997). These general equilibrium analyses were based on a perfect world. In a perfect world there would be no need for financial institutions and intermediaries because the market is perfect. A perfect market entails no frictions, such as transaction costs, and symmetric information. Unfortunately, we do not live in a perfect world. Our world has no perfect market; transaction costs exist and information is asymmetric. Asymmetric information occurs when one party does not have enough information about the other party to make accurate decisions, or one party has less information than the other party (Mishkin, Matthews & Giuliodori,
2013). This explains, in part, why financial intermediaries play an important role in our financial sector.
An example of a financial intermediary is a bank. Households and other savers put their money in the bank and the bank transforms this into loans for enterprises or for the government. Banks can reduce information asymmetries in a way that could not have been done by direct financing or by equity financing (Schmidt et al., 1997). Adverse selection and moral hazard are problems created by asymmetric information. Adverse selection occurs when the borrower who poses a great risk to the bank is the one who actively seeks out a loan and who is also the one with the largest probability of getting it (Mishkin et al., 2013). Banks are equipped with tools to screen out these bad borrowers. Bad borrowers have the highest probability of default. When the probability of default increases the credit risk increases as well. Credit risk stems from the uncertainty in a counterparty’s (the borrower) to meet its financial obligations (Mishkin et al., 2013). Due to the screening ability of banks, the adverse selection problem will be reduced. Moral hazard is the problem that arises when one party changes its behaviour negatively with respect to the other party after the transaction occurs (Mishkin et al., 2013). The borrower might invest in projects with a great probability of failure, which is not desirable for the lender. Banks also minimize this problem because they have the tools to monitor the parties they lend to (Mishkin et al., 2013).
The importance of banks within the financial sector is different among countries. Schmidt et al., (1997) have investigated within three European countries (France, Germany and the UK) whether there was a trend towards disintermediation, securitization and whether banks were losing importance to the markets. The financial systems of members of the EU had been exposed to regulatory changes. These regulatory changes were made to integrate their financial systems. It is likely that, because of these changes, the role of banks had been altered. The financial systems of the largest economies of the EU; France, Germany and the UK might exhibit disintermediation and securitization (Schmidt et al., 1997). Their research used data starting at the early 1980s until 1996. Differences among the three countries were found but they could not reject the hypothesis that banks were sill important. Rather they found evidence of a changing role of banks as intermediaries within the financial sector. Banks were focussing more on granting credit to borrowers and monitoring
them than on collecting savings. Nonbank financial intermediaries, such as insurance companies, have replaced this task. They had become important collectors of savings and passed these on to banks. This means that the supply chain of funds has changed (Schmidt et al., 1997).
They concluded that the financial system in France has had a tendency to move away from a ‘bank-based financial system’ where banks play an important role in the allocation of funds towards a ‘market-based financial system’ where financial markets play an important role. There is no such finding for Germany or for the UK. Germany was still relying on banks for funds and the UK still relied on the financial markets to supply funds (Schmidt et al., 1997).
As mentioned in the introduction, the change in the supply chain of funds might have a large effect on the loan supply. The shift in the focus of banks occurred in the years prior to Basel II. In 1999, Basel II made its introduction and took effect on the financial statements of banks. By increasing the number of asset categories and risk weights, a bank might face higher capital requirements. The result might be a smaller supply of loans. This might have a large effect on the leverage ratio of companies since, according to Schmidt et al., banks focused on granting credit to companies.
2.2 Regulation and Basel I
As shown in the research of Schmidt et al. (1997), banks are still important as financial intermediaries and providers of funds. To ensure the soundness of the financial system banks are subject to regulations (Mishkin et al., 2013). One of the reasons why they are regulated is asymmetric information. Depositors do not know the quality of the loans granted by the bank to other clients and might, in a panic, withdraw their funds. The bank (good or bad) has to liquidate its assets, and file for bankruptcy, to meet the increasing demand for cash. This may cause a bank panic; the failure of one bank causes the failure of others (Mishkin et al., 2013). Another effect of asymmetric information is that depositors do not know where a bank manager invests their money. They may be reluctant to put money in the bank, making the banking institutions less viable. In the Netherlands, the government has put forward a solution for this problem: one will be paid off in full on the first €100.000 they have deposited in a bank if it fails. This is called the deposit insurance (Mishkin et al.,
2013). The deposit insurance will decrease the probability of a panic and increase the chance that one will deposit money in the bank (Mishkin et al., 2013).
A negative effect of this deposit insurance is moral hazard. A bank might take on more risk, increasing the probability of default, at the expense of the depositor. The Basel Accord is a direct result of this problem. It tries to limit the excessive risk taking by banks by dealing with risk-based capital requirements. Assets facing high credit risk are assigned larger weights and consequently higher capital requirements than assets facing little credit risk. Banks have to hold more capital which will make them want to pursue less risky activities since they have more to lose (Mishkin et al., 2013). This capital could also function as a cushion to absorb any losses caused by shocks, which will make the failure of the bank less likely. The first Basel Accord (Basel I) has been implemented in 1988. Banks had to hold a minimum of eight per cent of its total risk-weighted assets in capital. Assets were grouped into five categories according to their risk. Each group was assigned a weight and accordingly a weighted value. The weighted value is this weight multiplied by the value of the asset. The bank sums up these weighted values and holds eight per cent in capital (Mishkin et al., 2013).
2.3 Basel II
In 1999 the Basel Committee released proposals to reform the current agreement. As mentioned above, the asset categories of Basel I were to crude and insufficiently differentiated. It also led to regulatory arbitrage; banks include risky investments, with the same risk-based capital requirement, in their financial statements but exclude riskless investments (Mishkin et al., 2013). This leads to risk taking, which is the opposite of its intent. These new proposals led to Basel II, which is based on three pillars: the first pillar links capital requirements more closely to actual risk, the second pillar emphasizes on strengthening the supervisory process and the third one focuses on improving the transparency of a bank’s credit exposure (Mishkin et al., 2013). Pillar one distinguishes between three types of risk: market risk, credit risk and operational risk. In addition, it specifies many more asset categories with different risk weights. These new categories with new risk weights might increase the amount of capital that a bank has to hold. Outstanding relatively risky loans, which at first had a risk weight of one, now have a risk weight of more
than one. In contrast to relatively riskless loans, which also had a risk weight of one assigned to it now have a risk of weight less than one. In addition, sophisticated banks are allowed to use an internal ratings-based approach that permits them to use their own models to assess credit risk. Other banks can use the approach that uses credit ratings from rating agencies, such as Standard & Poor, to assess credit risk.
Heid (2007) has investigated what the effects of this new Basel Accord would be on the macro economy. He assumed that the risk weights were a function of gross domestic product (GDP), minimal capital requirements and changes in the risk sensitivity of these minimal capital requirements. In addition, assuming that the amount of capital is fixed at the beginning of a period. Heid (2007) concluded that banks adjust their capital to loan ratios in a cyclical way. This means that if the economy is in a downturn the loan supply decreases and if it is going up the loan supply increases. If the minimal amount of capital increases, the capital to loan ratio rises and since capital is fixed at the beginning of the period the loan supply decreases. Meaning that if a company’s business outlook weakens, negatively affecting the credit quality, the bank granting credit to this company faces higher capital charges and decreases its loan supply. In addition he did an analyses using balance sheet data from banks operating in OECD countries in the year 2004. He posed three different scenarios to simulate changes in lending and capital. In the first scenario he analysed a drop of ten per cent in total assets and no change in the required capital. The assets of a bank consist of reserves, outstanding loans and securities (Mishkin et al., 2013). This first scenario can be interpreted as a reflection of capital regulation under the first Basel Accord. In the second scenario he analysed a 30 per cent change in capital requirements and in the last scenario he analysed the combined effect. The second scenario can be interpreted as a reflection of capital regulation under the second Basel Accord. The loan supply under scenario one (Basel I) would decrease by three per cent while under scenario two and three the supply would decrease by respectively 10 and 13 per cent. Under Basel II (the second scenario) the loan supply would decrease more than under Basel I (the first scenario). Heid (2007) concluded that under Basel II cyclicality would increase.
3. Factors that influence the leverage ratio
Over the past years there have been many developments on how to assess credit risk. This has been in response to an increase in the number of bankruptcies, the increase in off-balance sheet instruments and the decline in the value of assets (Altman & Saunders, 1998). In the early years, bankers used the ‘4C’s of credit’ to assess the characteristics of the borrower. These four C’s stand for: character (reputation), capital (leverage), capacity (volatility of earnings) and collateral (Altman & Saunders, 1998). One of the ‘4C’s’: collateral, has been a large determinant in assessing credit risk and has caused a lot of debate. Collateral can mitigate the problem of moral hazard. A borrower is less likely to pursue risky investments since he will lose this collateral if he defaults on his loan. Moral hazard was also one of the reasons to introduce the first Basel Accord, as mentioned above.
There are two alternative interpretations in the use of collateral for obtaining credit. Boot, Thakor & Udell (1991) concluded that the borrower who had the highest probability of default (the bad borrower) would pledge collateral and the borrower who had the lowest change of default (the good borrower) would not. They explained this by assuming that the borrower can choose two sets of actions: a high effort and a low effort. The good borrower has a larger probability of investing in a successful project and consequently has a smaller marginal productivity of effort than the bad borrower. This means that if the good borrower chooses an amount x of effort, his probability of success will increase by less than when the bad borrower chooses the same amount x as effort. In equilibrium, the bad borrower will choose a high effort and a collateralized loan. This is because he will only lose his collateral if he defaults and the change of default will be lower if he chooses the higher effort. The good borrower will choose the low effort and no collateral. It is not efficient to choose the higher effort because of his lower marginal productivity (Boot et al., 1991).
This is the opposite view of Stiglitz & Weiss (1981) who state that wealthy borrowers, who have the lowest change of default, would be willing to pledge the most collateral. Wealthy borrowers are less risk averse than non-wealthy borrowers and will, therefore take more risk. In this case, there should be a negative relationship between collateral and loan default (Stiglitz & Weiss, 1981). But they all agree on the fact that banks use collateral to assess whether a borrower is a good one or a bad one and consequently which one poses the highest credit risk to the bank. A company with
small credit risk has a higher probability for obtaining a loan and has to pay a smaller price for it. Therefore, credit risk influences the leverage ratio of a company.
A determinant important to consider when the factors influencing the leverage ratio are examined is monetary policy. Just after the beginning of the financial crisis, the question arose whether the low interest rates caused banks to increase their risk taking. A low interest rate may cause a search for a higher interest rate somewhere else (Jiménez, Ongena, Peydró & Saurina, 2014). Jiménez, et al. (2014) have done research on this topic by examining data from Spain. They found that this is the case; a lower overnight rate causes banks to increase their risk taking by granting more credit to companies with higher risk. The average bank increased its grants to risky companies by seven per cent for a one-percentage point decrease in the overnight interest rate. This is especially true for banks that had a low amount of capital. These low capitalized banks granted larger loans that were more likely to be uncollateralized (Jiménez et al., 2014).
Another factor influencing the leverage ratio of a company is the industry in which the company is located. The financial data, such as cash flow, of a company is compared to the norm of its industry (Treacy & Carey, 2000). Companies active in a declining industry are riskier than companies active in an increasing industry (Treacy & Carey, 2000). A decreasing industry is an industry that faces a shrinking demand for its products. The probability of default in a declining industry is, consequently, larger than in an increasing industry. A risky company poses a higher credit risk to the bank. Therefore the probability of obtaining a loan is reduced, which will reduce to leverage ratio. Treacy & Carey (2000) also mention that even if companies are active in an increasing industry there might be a difference in risk assigned to these companies. Medium and small size companies are relatively more risky than larger companies. Consequently, a large company is assigned a smaller credit risk than a medium or small company. In addition, a large company has easier access to external finance, such as equity finance, than a medium or small company (Treacy & Carey, 2000). Easier access to equity finance reduces the leverage ratio. However, equity finance has its drawbacks. Research has proven that managers will only issue equity to obtain credit as a last resort (Berk & DeMarzo, 2011). This is because stock prices decline on the announcement of en equity issue, diluting the value of the shares of the
current shareholders. Debt, on the other hand, has an advantage since interest payments can be deducted before paying taxes (Berk & DeMarzo, 2011).
A company could also finance its projects internally instead of applying for a loan or issuing equity. Internal financing reduces the leverage ratio since credit does not have to be obtained from external sources. Berk & DeMarzo (2011) conclude that the vast majority of companies in the US, in the years 1995-2008, fund their projects using retained earnings instead of debt or equity.
The leverage ratio of a company is also affected by its credit rating. A credit rating is used to assess the credit worthiness of the borrower or its public debt, such as commercial bonds. These credit ratings are assigned by an external rating company, for example Standard & Poor. Credit ratings are used by small banks to assess the credit risk of a company. However, not every company has a credit rating. This is because a company has to pay the rating agency for obtaining a rating. A high credit rating (for example AAA or AA) is associated with low credit risk and a low credit rating (for example BB) with high credit risk. Consequently, a company with a high rating has a greater probability to obtain credit than a company with a low rating (Treacy & Carey, 2000).
Finally, the leverage ratio of a company is affected by business cycle fluctuations. In an expansion, employment, production and sales are rising. If the economy is expanding, more profitable investment opportunities are available than when the economy is contracting (Mishkin et al., 2013). Therefore, a company is more likely to apply for a loan to invest in a project when the economy is expanding than when it is not. This causes the leverage ratio to increase in a boom (an expansion) and to decrease or stay the same in a bust (a contraction).
4. Research hypotheses
Many researches have investigated the effect of Basel II on banks. So far, no researcher has examined the effect of Basel II on companies. The first Basel Accord assigned fixed risk weights to each of the five asset categories. The new Basel II Accord has submitted more asset categories with matching risk weights. Under Basel II, more differentiation among different asset classes is possible. Therefore the risk-weighted value of the assets of a bank will change. This will affect the required amount of capital a bank has to hold. It might decrease the available loan supply.
Another effect caused by the newly submitted asset categories is companies who, at first, had no trouble obtaining credit might encounter difficulties now. Research indicated that the total loan supply offered by a bank decreases if one company’s outlook weakens, making it more difficult for another company to obtain credit. Obtaining credit might even be more difficult if a bank has just experienced defaults
on other loans, decreasing the value of its assets.
Companies are assessed by their credit risk before obtaining a loan. Earlier research suggests that factors affecting the credit risk of a company are: collateral, size, the industry in which it is active and a possible rating. A factor that affects the loan supply is the overnight interest rate. Banks assess the credit risk before granting a loan to a company. As a result, the amount of credit risk is an important factor that determines the outstanding leverage of a company. In addition, companies can finance their projects internally instead of applying for a loan. They are more likely to apply for a loan when the economy is expanding than when it is contracting. Suggesting that the leverage ratio will increase when the economy is expanding. Internal financing and business cycle fluctuations are also factors to account for when the leverage ratio of a company is examined. With the available information, the following hypothesis
will be examined:
Hypothesis: Basel II has a significant influence on the leverage ratio of listed companies in the Netherlands.
5. Methodology
This section discusses the data and the model that will be used in this research. The database used for this study is Compustat Global Fundamentals. Data was found for every listed company during the years 2000 up to and including 2010 in the Netherlands. Companies that were listed at the AEX, the AMX or the AScX were part of this study2
. The AEX contains the 25 (1-25) companies with the largest amount of traded shares and the highest turnover in the Netherlands. The AMX contains the subsequent 26-50 companies and the AScX contains the subsequent 51-75 companies.
The composition of one of the indexes changed once a year. In the years prior to 2008 this revising was done in March. However, in 2008, Euronext announced that they would be revising the index twice a year from now on. In order to prevent a bias, this study will only look at the revising in March and will leave out the potential revising after March. This means that companies who leave the index in the revising of September 2009 will be left out of the index in 2010 but are still part of the index in 2009. Financial data of companies who were acquired by, or merged with another company following the month March, in for example 2006, were not taken into account for that year. This is because no annual report is available for the year 2006 since the company does not exist anymore. This is also the case for a company who files for bankruptcy in the months following the month March. Using Excel, data was altered to fit the variables of the regression. Some variables of companies were stated in dollars, using the data on exchange rates available in the databases of ‘De Nederlandsche Bank’, they were converted into euros.
The model that is used in this research is the fixed effects regression model. The fixed effects regression model controls for omitted variables in panel data when these variables differ among entities but stay the same during the sample period. The fixed effects regression model sets n different dummy variables that absorb the influences of all these omitted variables (Stock & Watson, 2012). Examples of omitted variables are industry dummies. As Treacy & Carey (2000) mentioned, the industry in which a company is active is an important factor. The fixed effects regression model takes these industry differences into account.
5.1 The variables
Stiglitz & Weiss (1981) and Boot et al. (1991) agreed on the fact that collateral is used to screen new borrowers. However, they drew opposite conclusions stating which borrower would pledge collateral to obtain credit. As a result, collateral is one of the variables included in this model. The value of the collateral within a company is estimated by the amount of fixed assets. Fixed assets include property, plant and equipment that can be sold to repay a loan in case of default. In the research of Boot et al. (1991), fixed assets is also referred to as an estimator of collateral. The second variable that is part of the model is the overnight interest rate. The overnight interest rate related to the Netherlands is the Euro OverNight Index Average. Jiménez
et al. (2014) used the annual change in the overnight interest rate to predict the amount of loans granted by banks. This is the reason why, in this study, the change in the annual overnight interest rate is used as a variable instead of the absolute value. The variable concerning the change in the overnight interest rate is calculated by using the data available at the database of the European Central Bank. The third variable that affects the leverage ratio and changes over time is the credit rating. Since companies have to pay for these ratings not all of them would do so. Small companies might not think of it as profitable or do not have the money available to invest in these ratings. This seems to be the case; only eight companies have invested in them over the sample period3
. These eight companies are all listed at the AEX. In this research, the credit rating assigned by the rating agency Standard & Poor on the long-term debt of a company is used. Each grade had been assigned a numerical value to be able to incorporate it as a variable4
.
A factor that might affect the shareholders’ equity and consequently the leverage ratio is the size of the company. In this research the size of a company is compared to the size of the industry. A company that has millions of euros worth of sales can be seen as a big company. On the other hand, if you look at the industry it operates in (it has been proven that banks do this), it might still be a small company with great risk. This is the reason why the relative size of a company is included as a variable instead of the absolute size. The size of a company is estimated by means of the annual amount of sales. This is compared to the total supply of goods and services within the industry to form the variable5
. The data about the industries was found on the site of ‘het Centraal Bureau voor Statistiek’.
The second last variable included in this research is retained earnings. Retained earnings is used to estimate the possibility of a company to finance projects internally. Retained earnings is the amount of earnings retained in the company to invest in projects or to repay debt. It is calculated by adding the net income from this year to the retained earnings from last year and subtracting the paid out dividends form this year.
The last variable included in the fixed effects regression model is the gross domestic product growth rate. The value of final goods and services produced by a
3 Table 2 in the Appendix displays the companies with a credit rating
4 Table 3 in the Appendix assigns credit ratings to a numerical value
country plus product taxes and minus subsidies not included in the value of the goods equals the gross domestic product. The annual gross domestic product growth rate (GDP growth rate) is the percentage growth- or contraction of the GDP during the year. It is used to describe the business cycle fluctuations in the Netherlands (Mishkin et al., 2013). If the economy is in an upward movement, GDP will rise and if it is in a downward movement GDP will fall. In addition, Heid (2007) concluded that Basel II is cyclical; the loan supply will increase if the economy is in a boom and it will decrease when the economy is in a bust. Meaning, the variable GDP will be correlated with the variable for basel. It is also a determinant in explaining the variation in the leverage ratio of listed companies. To control for omitted variables, the GDP growth rate is included as a variable. Using the database available at the World Bank, this variable had been calculated.
5.2 The model
The model that will be tested is:
𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 𝑟𝑎𝑡𝑖𝑜
= 𝛽!𝐹𝑖𝑥𝑒𝑑 𝑎𝑠𝑠𝑒𝑡𝑠!"+ 𝛽!𝑅𝑒𝑡𝑎𝑖𝑛𝑒𝑑 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠!"+ 𝛽!𝑆𝑖𝑧𝑒
+ 𝛽!∆𝑂𝑣𝑒𝑟𝑛𝑖𝑔ℎ𝑡 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑟𝑎𝑡𝑒 + 𝛽!𝐶𝑟𝑒𝑑𝑖𝑡 𝑅𝑎𝑡𝑖𝑛𝑔
+ 𝛽!𝐺𝐷𝑃 𝑔𝑟𝑜𝑤𝑡ℎ 𝑟𝑎𝑡𝑒 + 𝛾𝐵𝑎𝑠𝑒𝑙 + 𝛼!+ 𝑢!"
With i = 1, . . , 83 for all the companies, t = 1, . . , 11 for the years that have been accounted for and α1, . . , α83 are the company-specific fixed effects. Basel is a stepwise dummy; stepwise until 2008 and equal to one from 2008 up to and including 2010. The reason for this is that since 1999 the first Basel Accord had been reforming to Basel II. Banks needed to make changes in their capital before the actual date of implementation on January the first of 2008.
The hypothesis becomes:
Hypothesis: 𝐻!: 𝛾 = 0 𝐻!: 𝛾 ≠ 0
The coefficient γ will be tested on a significance level of five per cent. When it is significant, the null hypothesis will be rejected and significant influence of Basel II on the leverage ratio can be proved. The coefficient γ will be tested first using the data available for all the companies listed at the indexes together. Second, the coefficient will be tested using the data available for companies listed at an index separately. The hypotheses will be tested five times; once for the entire data set, once for companies
listed at the AEX, once for companies with an outstanding rating, once for companies listed at the AMX and once for companies listed at the AScX. The variable credit rating is only available for a few companies listed at the AEX, as mentioned above.
The overall significance of the model will be tested using a F-test at the significance level of five per cent. It tests the null hypothesis that all the coefficients, except the constant term, are zero. This means that if the null hypothesis cannot be rejected, none of the variables explain the variation in the leverage ratio.
6. Research results
6.1 Testing the assumptionsBefore using the fixed effects regression model to test the hypotheses the four assumptions underlying the regression model have to be met (Stock & Watson, 2012). These four assumptions are:
1. uit has a conditional mean zero.
2. All variables included in the model are independent and identically distributed draws from their joint distribution.
3. Large outliers are unlikely.
4. There is no perfect multicollinearity.
The first assumption is tested and the errors have, approximately, a conditional mean of zero6
. A problem that might arise in panel data is autocorrelation. Autocorrelation, in this study, would mean that the fixed assets in year one are correlated with the fixed assets in year two for the same company. This would violate the second assumption. The variables have been tested for autocorrelation within panels and heteroskedasticity across panels. Heteroskedasticity has been tested using a feasible generalized least squares regression and autocorrelation has been tested using a woolridge test. No autocorrelation is observed, however there is evidence of heteroskedasticity. Meaning that the variance of the conditional distribution of the error term given the variables is not constant and depends on the variables (Stock & Watson, 2012). The option ‘robust’ is added to control for the heteroskedastic errors. The data has also been tested for cross-sectional correlation. This study contains data over a small time series (10 years) therefore, within the dataset, too few common
observations are observed and cross-sectional correlation does not seem to be of an issue. It can be concluded that the second assumption is met since there is no sign of autocorrelation or cross-sectional correlation. Plotting each variable against the other variables tests the third assumption. All the data seem to be in range, except for a few small outliers7
. To test the last assumption, the variance inflation factor is measured. The variance inflation factor measures how much the variance of an estimated regression estimator is inflated, which is caused by collinearity. If the variance inflation factor is larger than 10 it indicates multicollinearity. There is no sign of multcollinearity in this study, all factors are smaller than 108
. However, the variables retained earnings and fixed assets have a high variance inflation factor, which means that they could be highly correlated which in turn affects the estimated regression coefficients. To compare the regression results, two regressions; one including retained earnings and one excluding retained earnings are estimated.
Including the variable credit rating causes a reduction in the number of observation. This is due to the fact that the fixed effects regression model leaves out observations that do not have a value for every included variable. This smaller sample is also tested for the assumptions underlying the model and every assumption is met except the fourth one. Adding the variable credit rating causes multicolinearity among the variables. In order to meet the assumption again, the variable retained earnings is left out of the regression.
6.2 The regression results
Table 1 shows the estimates for the fixed effects regression model using the entire dataset. As stated in the research hypotheses, this research expects to find a significant influence of Basel II on the leverage ratio of companies listed at one of the indexes during the years 2000 up to and including 2010. The first regression, shown in table 1, estimates the value of the coefficient γ, corresponding with the dummy variable basel. The value is negative and it is not significant at the five per cent level; its p-value is larger than 0.05. The null hypothesis, stating that Basel II does not have a significant influence on the leverage ratio of listed companies in the Netherlands, cannot be rejected.
7 Figure 2, 3 and 4 in the Appendix
Table 1
Regression Output for the entire Dataset
* p < 0.05 ; ** p < 0.01
In the second regression, also shown in table 1, the variable retained earnings is omitted. The F-statistic increased a little bit, making the model a better fit. This could be due to the fact that the variable retained earnings and fixed assets are highly correlated. The R2
in the second regression has decreased compared to the first one. The regression R2
measures the fraction of explained variance in the leverage ratio. Given the formula of the R2
, the lower value in the second regression can be explained. When a variable is added to the model, the R2
always increases (Stock & Watson, 2012). The second regression omits the variable retained earnings, logically, therefore the R2 decreases. Despite the increase in the F-statistic, the joint null hypothesis stating that all the coefficients equal zero cannot be rejected. The overall model does not explain the variation in the leverage ratio of listed companies. In addition, in the second regression, the coefficient γ, corresponding with the dummy
Variable (1) (2) Fixed assets 0.000 0.000 (.098) (0.62) Retained earnings -0.000 (0.62) Size 0.107 0.152 (0.70) (1.16) Overnight interest rate 0.031 0.041 (0.14) (0.19) GDP growth rate -2.493 -2.602 (1.30) (1.36) Basel -0.557 -0.646 (1.28) (1.49) Constant 2.426 2.427 (8.04)** (7.95)** R2 0.02 0.01 Number of observations 498 81 F- statistic 0.77 0.93 (0.5954) (0.4686)
variable basel is still insignificant. Again, the null hypothesis cannot be rejected at the five per cent significance level. Basel II does not have a significant influence on the leverage ratio of listed companies in the Netherlands.
Table 2 shows the estimates for the fixed effects regression model using only the companies listed at the AEX during the years 2000-2010. In the first regression, retained earnings is included in the model and in the second regression it is omitted. The first regression shows the coefficient γ, corresponding with the dummy variable basel, is insignificant; its p-value is larger than 0.05. Within this regression the hypothesis cannot be rejected. Basel II does not have a significant influence on the leverage ratio of the companies listed at the AEX. This is also the case for the second regression where retained earnings is omitted. To evaluate the overall model, a F-test is performed on the coefficients of the included variables. The p-value of the F-statistic is, for both of the regressions, larger than 0.05, which does not result in the rejection of the null hypothesis stating that all the coefficients equal zero. None of the variables included in the model explain the variation in the leverage ratio of listed companies at the AEX.
Table 2
Regression Output for the Companies listed at the AEX
Variable (1) (2) Fixed assets 0.000 0.000 (0.93) (0.63) Retained earnings -0.000 (0.82) Size 0.151 0.085 (0.87) (0.66)
Overnight interest rate 41.935 -21.993
(1.13) (0.53) GDP growth rate -32.189 1.430 (1.37) (0.06) Basel -1.213 -1.315 (1.47) (1.58) Constant 2.525 2.701 (3.72)** (4.27)** R2 0.02 0.04 Number of observations 192 192 F- statistic 1.15 1.44 (0.3160) (0.2401) * p < 0.05 ; ** p < 0.01
Table 3 shows the estimates for the fixed effects regression model using only the companies listed at the AEX during the years 2000-2010 with a credit rating. As mentioned before, there are only eight companies in this sample. In addition, the variable retained earnings is omitted because of multicollinearity. This regression does show a significant effect, at the five per cent level. Therefore the hypothesis can be rejected and significant influence of Basel II on the leverage ratio can be proved. It is also shown, in table 3, that the coefficient γ, corresponding to the dummy variable basel is smaller than zero, however it is not significantly smaller than zero. This implies that the introduction of Basel II has lowered the leverage ratio of the included companies. The decrease in the leverage ratio could be caused by a decrease in
outstanding leverage or an increase in outstanding equity. As mentioned above, companies will only issue equity as a last resort, since it has its drawbacks. Companies will issue equity as a way of funding when the market believes that the stock is worth more than it actually is. If the stock prices are too high, issuing equity is a relatively cheap way of funding. However, this was not the case in the time span of this study. During the crisis, which erupted in Europe in 2008, the market was careful and stock prices were decreasing in value. The remaining conclusion is that Basel II has caused the outstanding leverage to decrease. This seems to be a strange result. The companies included in this regression are quite large and active all over the world. Hence, they can obtain capital from other countries besides the Netherlands who are not subject to the Basel Accords. If banks in the Netherlands are not eager to grant credit to them, they can turn to other countries such as the United States to obtain credit. In addition, they have easier access to the public debt market than smaller companies and consequently can obtain credit through that channel. This would mean that Basel II should not have a significant effect on the leverage ratio of companies listed at the AEX with a credit rating.
The fixed effects regression model controls for omitted variables when the omitted variables vary across the companies but do not change over time. It does not control for omitted variables that vary across the companies and change over time. An example of an omitted variable is the credit rating. In the regression output it is shown that the error term is negatively correlated with the variables. By adding the variable credit rating, the correlation has decreased and the negative sign is gone. The negative correlation of the error term with the variables could be due to the correlation between credit rating and retained earnings. This causes a bias in the coefficient of retained earnings. The correlation between the two variables equals -0.726, causing the coefficient β2 to decrease together with its t-value. This can be seen in the first regression in table 2; the coefficient β2 of the variable retained earnings is very small and almost equal to zero.
Table 3
Regression Output for the Companies listed at the AEX with a credit rating
Variable (1) Fixed assets 0.000 (1.35) Retained earnings Size -0.128 (0.90)
Overnight interest rate -56.927 (1.28) GDP growth rate 26.126 (1.11) Basel -2.611 (2.68)* Constant 4.601 (1.19) R2 0.34 Number of observations 81 F- statistic 3.13 (0.0807) * p < 0.05 ; ** p < 0.01
When the variable credit rating is added to the model, retained earnings is left out because of multicollinearity, the model becomes a better fit. Meaning that the p-value of the F-statistic of the regression shown in table 3 is smaller than the regressions in table 2. The reduction in the p-value does not cause the rejection of the joint null hypothesis stating that all the coefficients equal zero. This model does not explain the variation in the leverage ratio of listed companies at the AEX with a credit rating.
Table 4 shows the estimates for the fixed effects regression model using only the companies listed at the AMX during the years 2000-2010. Also for the companies listed at the AMX, Basel II does not have a significant influence on the leverage ratio. The coefficient γ, corresponding with the dummy variable basel, is insignificant; its
p-value is larger than 0.05. Consequently, the hypothesis cannot be rejected. However, the coefficient 𝛽!, corresponding with the variable size, is significant at the five per cent level and even at the one per cent level. This means that the size of a firm relative to the entire industry explains part of the variation in the leverage ratio. When the total sales of a company listed at the AMX increases with one euro relative to its industry, its leverage ratio increases with 1.628 euros in the first regression and with 1.680 euros in the second. It can be concluded, from the regression, that the size of a company relative to its industry is an important factor in explaining the variation of the leverage ratio for companies listed at the AMX. A relative small company faces large competitiveness and therefore has no influence on prices. It cannot increase prices to obtain higher revenues, which can be used to pay interest on its outstanding debt. A relative large company is less sensitive to competitiveness and does have some influence on prices. It can increase prices to obtain higher revenues, which can be used to pay interest on its outstanding debt. In addition, the higher revenues decrease the probability of default. Hence, a relative large company can more easily obtain credit since its credit risk is lower than a relative small company. This also explains why the coefficient 𝛽! is positive; if the relative size of a company increases, it can more easily obtain credit and its leverage ratio will increase. It can also be concluded that if a company’s size increases it will increase its debt instead of issuing equity. If it would increase its outstanding equity when its size increases, the coefficient 𝛽! would be negative. A reason for this might be that issuing equity is more costly than increasing debt, as mentioned above.
To evaluate the overall model the F-statistic has been calculated. As shown in table 2, both the p-values are smaller than 0.05. The null hypothesis stating that none of the variables explain any of the variation in the leverage ratio of companies listed at the AMX can be rejected. The smaller p-value of the F-statistic of this regression compared to the AEX regression could be due to the large significant effect of the variable size. All the other coefficients are still insignificant.
Table 4
Regression Output for the Companies listed at the AMX
Variable (1) (2) Fixed assets -0.000 -0.000 (0.19) (0.75) Retained earnings -0.000 (0.94) Size 1.628 1.680 (3.27)** (3.39)**
Overnight interest rate -85.039 -82.111
(0.87) (0.86) GDP growth rate 48.028 45.800 (0.89) (0.88) Basel -0.219 -0.373 (0.50) (0.88) Constant 0.932 1.113 (0.79) (1.06) R2 0.03 0.03 Number of observations 207 207 F- statistic 3.70 4.16 (0.0048)** (0.0036)** * p < 0.05 ; ** p < 0.01
Table 5 shows the estimates for the fixed effects regression model using only the companies listed at the AScX during the years 2000-2010. In this regression, the coefficient γ, corresponding with the dummy variable basel, is insignificant; its p-value is larger than 0.05. Again, the hypothesis stating that Basel II does not have a significant influence on the leverage ratio of listed companies cannot be rejected. However, in the first regression, the coefficient 𝛽! of the variable fixed assets is significant at the five per cent level and even at the one per cent level. This means that the amount of collateral that a company has is an important factor in explaining the variation of the leverage ratio for companies listed at the AScX. If the amount of fixed assets of a company increases with one euro, the leverage ratio increases with 0.004 euros. This is a small effect.
In the second regression, without retained earnings, the coefficient 𝛽! of the variable fixed assets is also significant at the five per cent level but not at the one per cent level anymore. If the amount of fixed assets of a company increases with one euro, the leverage ratio increases with 0.003 euros. It can be concluded that indeed the variables retained earnings and fixed assets are correlated. When retained earnings is left out of the regression, the t-value of the coefficient 𝛽! decreases making it insignificant at the one per cent level.
Companies listed at the AScX are the ones with the smallest turnover and least amount of shares traded. Among the companies listed at the AMX, was a difference in size and consequently leverage ratios. However, there is no such finding among the companies listed at the AScX. It can be concluded that these companies are all too small to influence prices in order to increase revenues. Therefore, the variable size does not have an influence on the leverage ratio. Since they cannot influence prices, they pose a higher risk to banks. To decrease this risk, a bank will look at the amount of collateral. If a company defaults on its loan, the bank will pledge this collateral. When the value of the collateral is high the risk of the loan will be smaller than when the value of the collateral is small. Meaning, a company with a large amount of fixed assets will have easier access to credit than a company with a small amount of fixed assets.
The overall model explains the variation in the leverage ratio properly. Both the p-values of the F-statistics are smaller than 0.05, therefore the joint null hypothesis that all the coefficients equal zero can be rejected.
Table 5
Regression Output for the Companies listed at the AScX
Variable (1) (2) Fixed assets 0.004 0.003 (3.28)** (2.26)* Retained earnings -0.006 (1.79) Size 7.062 3.079 (0.90) (0.42)
Overnight interest rate -4.373 -9.786
(0.88) (1.94) GDP growth rate 2.189 3.752 (0.74) (1.31) Basel -0.452 -0.507 (0.30) (0.30) Constant 0.765 1.134 (0.94) (1.20) R2 0.13 0.07 Number of observations 101 101 F- statistic 3.20 2.96 (0.0138)* (0.0258)* * p < 0.05 ; ** p < 0.01 6.3 Conclusion
Almost every regression shows no significant effect for the dummy variable basel on the leverage ratio. It can be concluded that Basel II has not had a significant influence on the leverage ratio of listed companies in the Netherlands. However there are differences among the three indexes.
No variable has a significant influence on the leverage ratio of companies listed at the AEX and consequently cannot explain the variation in the leverage ratio. The size of a company, relative to its industry, does have a significant influence on the leverage ratio of companies listed at the AMX. A relative large company can influence prices to increase its revenue. This ‘power’ reduces its probability of default
and consequently increases its probability of obtaining credit. When the relative size of a company increases, it is more likely to increase its leverage than to increase its outstanding equity. A reason might be that issuing equity is more costly than increasing debt.
Finally, the amount of fixed assets (collateral) has a significant influence on the leverage ratio of companies listed at the AScX. A small company cannot influence prices and consequently poses a higher risk to a bank. This risk will be decreased if a company has a large amount of fixed assets. Meaning, if the value of fixed assets of a company increases, its leverage ratio will increase, since it has easier access to capital.
Another conclusion is that the credit rating of a company is an important factor in explaining the variation of the leverage ratio. Including this variable causes the coefficient γ, corresponding with the dummy variable basel to be significant. Within this smaller sample, Basel II has had an influence on the leverage ratio of the included companies. This can be seen as a strange result, since the included companies can obtain credit from other sources than the banks in the Netherlands, who are subject to the Basel Accords.
7. Discussion
The fixed effects regression used in this research omits important variables. One of these variables is the borrow-lender relationship. Access to the necessary data was not possible because of privacy reasons. If a company has a close relationship with a bank they might have easier access to credit. They can obtain more credit and have to pay a lower price for it (Jiménez & Saurina 2004). On the other side, the bank has a certain degree of market power over the company if he is the only one who provides credit to it (Jiménez & Saurina 2004).
Large companies can obtain credit through private debt, which is considered in this research, and public debt, which is not considered in this research. The private debt market is much larger than the public debt market, however it is a way of financing projects. In this research only some companies, such as Shell, are able to issue public debt. Consequently, including the public debt sector as a variable might not have a large effect on the estimates in this research.
Another issue regarding this research is about the industries. Each company was assigned to only one industry, while it might be active in more than one. An example is the company Unilever that is active in both the food and household products industry. In this research it has been assigned to the food industry but might as well be assigned to the household products industry. This has an affect on the variable size. Implying that the variable size might not represent the actual size of a company.
A further research on this topic could include the missing variables: the borrower-lender relationship and the public debt market. This could make the fixed effects regression model a better fit for the data. The influence of Basel II on the leverage ratio might become significant, rejecting the hypotheses after all.
Literature
Altman, E. I., & Saunders, A. (1997). Credit risk measurement: Developments over the last 20 years. Journal of Banking & Finance, 21(11), 1721-1742.
Berk, J., & DeMarzo, P. (2011). Corporate Finance. Edinburgh Gate, Pearson Education.
Boot, A. W., Thakor, A. V., & Udell, G. F. (1991). Secured lending and default risk: equilibrium analysis, policy implications and empirical results. The
Economic Journal, 458-472.
Jiménez, G., & Saurina, J. (2004). Collateral, type of lender and relationship banking as determinants of credit risk. Journal of Banking & Finance, 28(9), 2191-2212.
Mishkin, F.S., Matthews, K., & Giuliodori, M. (2013). The economics of money, banking & financial markets. Edinburgh Gate, Pearson Education.
Schmidt, R. H., Hackethal, A., & Tyrell, M. (1997). Disintermediation and the role of banks in Europe: An international comparison. Journal of Financial
Intermediation, 8(1), 36-67.
Stiglitz, J. E., & Weiss, A. (1981). Credit rationing in markets with imperfect information. The American economic review, 393-410.
Stock, J.H., & Watson, M.M. (2012). Introduction to econometrics. Edingburgh Gate, Pearson Education.
Treacy, W. F., & Carey, M. (2000). Credit risk rating systems at large US banks. Journal of Banking & Finance, 24(1), 167-201.
Appendix
Table 1
Companies that took part in this study assigned to each index
AEX AMX AScX
Ahold CSM Brunel
Akzo Nobel Vopak Exact Holding
ASML Wessanen Grontmij
DSM OCÉ Unit 4
Heineken Imtech InnoConcepts
KPN Nutreco Pharming
Philips Ordina Sligro
Unilever Stork Ten Cate (=KTC)
Wolters Kluwer Randstad Mediq (=OPG)
TNT Super de Boer (=Laurus) Draka
Hagemeyer AM Imtech
Numico P&O Nedlloyd BE Semiconductor
Getronics Vedior Boskalis
Gucci Volker Wessels Jetix Europe
Vendex KBB SBM Offshore (=IHC
Caland)
Ordina
Baan Libertel USG People
Shell Corp Express (=Buhrman) Univar
Corus CMG Ballast Nedam
Corp Express (=Buhrman) CapGem SA Snowworld (=Fornix)
KPNQ West Corus Super de Boer (=Laurus)
SBM Offshore (=IHC Caland)
Logica Smit Internationaal
Vedior Air France – KLM Arcadis
TomTom KLM Wegener
Arcelor Mittal ASM Int. Kendrion
Randstad PinkRoccade Beter Bed
BAM Groep Vendex KBB Macintosh
Fugro Draka Wavin
Air France – KLM Fugro Lavide (=Qurius)
Boskalis Heijmans Accel Group
UPC USG People Gamma
Boskalis TKH Group
BAM Groep Eriks
Aalberts Spyker Cars
Crucell Sopheon Pharming Univar AMG Arcadis Wavin Ten Cate (=KTC) Mediq (=OPG) Smit Internationaal Unit 4 BE Semiconductor
Table 2
Companies that took part in the regression including the variable credit rating
Companies with a credit rating Ahold Akzo Nobel Philips Unilever Wolters Kluwer TNT Shell Arcelor Mittal Table 3
Credit ratings and their numerical values
Credit rating Numerical value
AAA 100 AA+ 94 AA 88 AA- 82 A+ 76 A 70 A- 64 BBB+ 58 BBB 52 BBB- 46 BB+ 40 BB 34 BB- 28 B+ 22 B 16 B- 10 CC 4 Table 4
Company assigned to its industry
Industry Company
Extraction of crude petroleum and natural gas Shell
SBM Offshore Food Unilever Numico CSM Nutreco Wessanen
Laurus (=Super de Boer)
Beverage industry Heineken
Textiles, clothing and leather Gucci
Chemical Akzo Nobel DSM
Metal Arcelor Mittal
AMG Aalberts
Metal products Corus
Electronics ASML
TKH Group Imtech
Electrical appliances Philips
Macintosh Kendrion OCE Transportation TNT Boskalis Air France-KLM Univar KLM P&O Nedlloyd Mediq
Furniture Corporate Express
Other Fugro Smit Internationaal InnoConepts Accel Group Fornix=Snowworld Eriks
Employment agencies Vedior
Brunel Randstad USG People IT services Getronics Baan TomTom Ordina PinkRoccade CMG Logica Unit 4 Cap Gemini Sopheon Qurius (=Lavide) Exact Holding Publishers Wegener Wolters Kluwer
Film, radio and television Jetix
Telecommunications KPN
UPC Libertel Draka
KPNQ West
Retail Ahold
Gamma Beter Bed Vendex KBB
Wholesale and trade Hagemeyer
Sligro General construction and product development BAM
AM Volker Wessels Heijmans Arcadis Ballast Nedam Grontmij Pharmaceuticals Pharming Crucell
Rubber and plastic products Wavin
Machinery ASM Int.
BE Semicond. Reparation and installation of machinery Stork
Warehousing and support activities for transportation Vopak
Cars Spyker Cars
Figure 1. The conditional mean of the errors is approximately zero 0 .1 .2 .3 .4 D en sit y -10 0 10 20 Residuals Kernel density estimate Normal density
kernel = epanechnikov, bandwidth = 0.2869
-1 0 -5 0 5 10 15 e ( yl eve ra g era tio | X ) -20000 0 20000 40000 e( x1fixedassets | X ) coef = .00004706, se = .00001316, t = 3.58 -1 0 -5 0 5 10 15 e ( yl eve ra g era tio | X ) -30000 -20000 -10000 0 10000 20000 e( x5retainedearnings | X ) coef = -.00005856, se = .00001862, t = -3.14 -1 0 -5 0 5 10 15 e ( yl eve ra g era tio | X ) -10 -5 0 5 10 e( x6grootte | X ) coef = -.14627644, se = .07399053, t = -1.98 Figure 2. The variable fixed assets plotted against the leverage ratio
Figure 3. The variable retained earnings plotted against the leverage ratio
Figure 4. The variable size plotted against the leverage ratio
Table 5
Test for multicollinearity
Variable VIF 1/VIF
Fixed assets 8.49 0.118
Retained earnings 6.97 0.144
Size 4.50 0.222
Basel 1.03 0.975
Overnight interest rate 1.00 0.998
