1
An investigation into the capital structuring trends of
financial institutions across cultures.
December 2015
University Of Amsterdam
Benjamin Wright
Student Number: 10621962
Thesis Supervisor: Dr J.K (Jens) Martin
14/12/2015
2
Table of Contents Page
Abstract 3
Introduction 3
Literature Review 5
Motivation for Study 11
Data 13
Variable Descriptions 16
Variable Summary 19
Regression Specifications (part 1) 21
Results (i) 22
Results (ii) 24
Results (iii) 25
Results (iv) 28
Readjustment Hypothesis - Introduction and Description 29
Regression and Variable Description (part 2) 31
Data 32
Results (part 2) 33
Conclusion & Table of Findings 36
Appendix (figures) 38
Appendix (tables) 43
3
Abstract:This is an investigation into the capital structures of financial firms from across 4 regions of the world – Asia, South America, Europe and North America – of two main parts. Overall, there are clear differences in the makeup of financial institutions’ capital structuring trends around the world. Through empirical analysis of the behaviour of these capital structures, I discover the following. Firstly, the EU is the only region in my sample in which there is any evidence of firm value, measured by stock price, reacting to changes in capital structure. Secondly, when periods before and after the financial crisis (2008) were observed, there were no further relationships to observe, meaning that public perception had surprisingly not changed as a result. However, there is some evidence for changes in investor attitude when comparing firms in the lower quartile of equity holdings against others. Lastly, I find some evidence for a ‘readjustment hypothesis’ of deposit levels in the overall capital structure of financial firms. I make use of fixed effects modelling for the first part of my analysis, as well as a modification of the Fama-MacBeth two step regression model for the second section.
Introduction:
Capital structuring theory has come a long way since Modigliani and Miller (1958) published the first work on the subject. Since its inception, many departures from this theory have been attempted. Most notably of these are the pecking order and the trade-off theory, although corporate finance is still undecided on the determinants of capital structure. Financial services, however, have long been regarded as a special case when it comes to capital structure theory. It’s widely accepted that the financial industry exhibits leverage ratios far beyond what is the norm in most other industries. Most firms would not dream of having leverage ratios as high as most financial service firms do and enjoy the luxury of solvency. Due to this, it’s commonplace to exclude financial institutions from empirical studies into capital structuring. The question of finding the optimal capital structure for financial services has to involve the delicate practice of balancing the need for financial services to stay competitive whilst being able to withstand economic shocks of the capacity seen in recent history. Staying competitive through having enough leverage whilst having sufficient capital in times of economic crisis is the delicate balance to be struck. Regulation has helped guide this process over
4
the course of the last decade through setting of minimum capital ratios.
This ability to withstand economic shocks has been exposed to devastating effect in the past decade. The naivety of the financial sector as a whole witnessed in 2007 demanded a regulatory update across the board, through the Basel III agreement. This requires a minimum capital buffer for all banks in good times as well as bad times. For a long time there has been an issue of moral hazard within the banking industry. Due to the unique service that banks provide, it is within the
government’s best interests to keep banks solvent at all costs. This has led to overleveraged balance sheets as a result of this reckless risk appetite in the past. Higher leverage benefits firm performance in the absence of financial distress. When times of financial distress do arise, there is the backstop of government bailouts. How the market reacts to different approaches to capital structuring choices is a trend which has not been given much attention in past empirical studies.
Most empirical literature to date has focused on non-financial firms. When financial firms have been included, these have been from North America, and possibly the EU. My investigation will span both these continents, as well as South American and Asian countries. This is relevant because vastly different capital structure make-ups of banks from these areas of the world will be observed when compared to the more developed regions of the world. This will focus on primarily how the market reacts to different capital structure choices of banks over the course of the last decade. As well as this, it will be observed whether market reactions have changed at all after the financial crisis in 2008. It’s a logical conclusion to draw that there is the possibility that the market will have a
different attitude to highly levered banks after 2008. Banks provide a third factor of capital structure to observe beyond leverage ratios. Leverage can be decomposed into straight debt and customer deposits. Customer deposits are a function of the main income source of a bank. Therefore, I will also be able to observe how these ratios change between regions of the world, and over time. There shall be two parts to this study;
Firstly, through analysis of panel data and the use of fixed effects (within) estimation techniques, I shall look at four sub-studies. Firstly, does changing capital structure affect market value, and how does this manifest itself? Secondly, does this relationship change when the studies are split into pre and post 2008 time periods, possibly exposing greater investor awareness to the dangers of
overleveraging. Thirdly, which role do customer deposits have to play in capital structures of financial institutions? This is a third type of financing that is unique to financial institutions – it’s a
5
form of debt which is income generating, however it can also be withdrawn at short notice. This could be a reason for the absence of financial institutions in capital structuring studies in the recent past. The added complication of a third type of financing unique to only one industry makes it very difficult to fit into trends of other industries. I will also look at how investor’s react differently to those firms with the lowers equity ratios compared to others. The question here is, do investors react more favourably to an increase in equity/leverage ratios when the firm has extraordinarily low holdings of equity as opposed to other firms in the industry?
Secondly, I shall be using and modifying the capital structuring readjustment hypothesis as observed in Welch (2004). This is a study, using the Fama-MacBeth two step regression technique, which investigates the readjustment tendencies of capital structure ratios. In the study of Welch (2004), the relationship of debt issuance or repurchasing in response to equity changes resulting from stock price fluctuations is investigated. This changing of debt levels in response was labelled the
‘readjustment hypothesis. Welch (2004) observed all non-financial firms. I shall not be using this methodology to be measuring the readjustment hypothesis of capital structure to market value of equity changes however, instead I shall be investigating the dynamics of deposit and debt
movements in the capital structure. Due to my study being one focused only on financial firms, one part of the capital structure which gets little attention is the role of deposits. These are not a normal type of debt, and are not found in any other type of firm so are unique to the financial sector. As well as this, deposits fluctuate in accordance with factors very much out of the individual firms’ control. Where factors such as market sentiment or company earnings control (market value) equity fluctuations, interest rates and inflation rates control deposit fluctuations. Hence, where Welch (2004) investigated the readjustment hypothesis of the capital structure to stock price changes, I shall be investigating the readjustment hypothesis of the capital structure to deposit fluctuations.
Literature Review on Capital Structuring Theories: i)
Modigliani & Miller (1958) published the first noticeable work on the subject of capital structuring theory. In the absence of taxes, bankruptcy costs, agency costs and asymmetric information, it was concluded that the value of the unlevered firm (pure equity) is equal to that of the levered firm (debt & equity financed). This is known as the capital structure irrelevance theory. However, this was the earliest research into the field and there have been many departures since. One implication of
6
this theory can instantly be refuted from a first glance of the data in this study. The notion that capital structures should vary randomly across industries and firms because of its irrelevance can be discarded looking at appendix (1a). We can see a very clear differentiation of capital structure patterns across different regions of the world. If the capital structure irrelevance theory is to hold, then we should see no such impact of capital structure changes on firm value. Any evidence to refute this hypothesis will require a statistically significant effect of changes in leverage ratios on the value of the firm, or the net income. It’s seen that this is the case as changes on the ELR affect net income in three of the four regions examined. As well as this, the fact that there are very clear trends across the regions show that capital structuring is not random.
The two main departures of this theory are the trade-off theory and the pecking order theory of capital structuring. The pecking order theory proposes that a firm will always prefer internal financing as a first resort, thereafter resorting to safe debt, risky debt and at the bottom of the pecking order is equity. A firm will often forego valuable investment opportunities in the interest of existing shareholders if equity has to be raised (Myers & Majluf 1984). This theory was based on asymmetric information between internal firm management and outside investors, and the assumption that a firms’ management has inside information that investors don’t have, and
therefore will always prefer internal financing. As pointed out by Myers (1984), this is a hard theory to measure because of the fact there are two types of equity, internal and external. Internal equity is at the top of the pecking order, whereas external is at the bottom. The best way to empirically investigate this would be to look at IPOs or equity offerings and measure these impacts on firm value through an event study, as this is exclusively external equity. If the pecking order theory explains the firms’ preferences, then an increase in outside equity relative to debt would tell us that a firm is in bad health. This is because the firms would have been unable to raise finance through internal means or through debt and would have had to resort to its least favoured option.
The trade-off theory proposes that the optimal capital structure comes down to the weighing up of the benefits of debt tax shields, and the costs of bankruptcy (Myers 1984). Because of the tax shields associated with debt, it’s always the preferred choice of financing. Firms should be financed
exclusively with debt, if not for the consequences of financial distress. The levels of debt and equity firms hold should be at the optimum level which would balance the benefits of tax shields and dangers of financial distress, and therefore work to this level to maximise firm value. If this theory is to be believed however, this leads to the implicit assumption that firms in similar situations should all be at very similar capital structure ratios. The assumption that all financial institutions across
7
regions are at the same optimal levels can immediately be dismissed looking at figure I, however, this will be expanded on later. The notion that there is an optimal capital structure which mediates between the costs and advantages of debt can be examined in this study. I will be looking at firms close to the regulatory minimums of equity/leverage levels. This behaviour should be different to firms comfortably above the regulatory minimum.
According to the trade-off theory, this will manifest itself in one of two different ways. The first way would be where the firms close to regulatory minimums will exhibit no relationship with firm value and other firms will exhibit a negative relationship. This will tell us that the optimum equity ratio exists at the bare minimum equity levels, and all other firms will benefit from decreasing equity ratios. Alternatively, the firms with equity/leverage levels just above the regulatory minimum will exhibit a statistically significant positive relationship with firm value. The rest of the firms will exhibit no statistically significant relationship. This will tell us that the optimum leverage level is at a level which is comfortably above the regulatory minimum. In this case, firms with the bare minimum equity will benefit from an increase in equity/leverage levels.
The neutral mutation hypothesis is a relatively unpopular theory, not given much attention over time, and was proposed by Miller (1977), when investigating what role taxes have to play in capital structures. This theory is based on the assumption that there is no real value in capital structures, and management simply falls into habits over time. Therefore, the capital structure isn’t based on any value maximising theory in particular, but a result of human nature and inertia. To use this as a null hypothesis would be easy, as it could be attributed to virtually any unexplained trend. It would also be almost impossible to refute, and would need definitive results pointing to another theory to do so. However, looking at the consistency of the equity ratios across the different cultures in appendix (1a), it’s hard to explain these differences. We can attribute this to ‘cultural neutral mutation’. To refute this, we would need to see results which suggest a much stronger positive relationship between firm value and equity ratios in South America than other parts of the world. We don’t see this relationship at a statistically significant level, so therefore can make an argument for turning to cultural habits unchanged over time.
The pecking order and static trade-off theories were empirically tested against one another using Industrial firms (Sunder & Myers, 1998). This study used data from 1971 – 1989, and favoured the pecking order theory as opposed to the trade-off theory. When both were included in the same regression, the pecking order theory remained robust and held its original explanatory coefficient
(R-8
squared 0.86), whereas the trade-off theory fell to a third of its value. It was pointed out, however, with higher growth firms in industries of less tangible assets (financial sector), this would be unlikely to perform to the same standard.
A slightly more modern take on capital structure theory is the market timing theory (Baker & Wurgler, 2002). As opposed to the previous theories, this draws on market valuations of equity being a key driver in determining capital structuring decisions. It’s proven that firms show a tendency to issue equity when market valuations are high, and repurchase equity when market valuations are low. However, only firms with IPO dates from 1968 – 1998 are taken, and only those with book value assets over $10 million, as well as the exclusion of financial firms. Although
insightful, this is a restrictive sample of firms and doesn’t address financial institutions. A survey conducted by Graham & Harvey (1999) adds credibility to this theory. A total of 392 CFO’s were surveyed on this topic, and a significantly large proportion of which admitted that stock price undervaluation or overvaluation was a key instrument in capital structure decision making. Drawing further on stock values to predict capital structuring decisions, Welch (2004) investigated the readjustment of (non-financial) firms’ capital structures in response to market value equity changes. It was concluded that firms do little to rebalance debt/equity ratios in relation to stock price fluctuations (Welch, 2004). This study used only North American firms, from 1962 – 2000, and from non-financial industries. Masulis (1980) also investigate the effect of stock returns on capital structure changes, and finds that there is a significant effect on firm stock price when capital structure changes are announced, especially when these changes are predicted to cause either a ‘corporate tax shield’ or ‘wealth distribution’ affect.
ii)
Banks have a unique capital structure compared to other industries. The main reason for this is due to the fact that debt is not only a source of financing for the firm, but a main source of income for the firm. Through customer deposits banks generate interest income; therefore it’s natural to assume that banks will have leverage ratios far higher than those of other industries. As opposed to other industries, this also makes the strong argument that higher leverage increases firm
profitability. Not only is this achieved through tax reductions or other reasons coming from traditional capital structure theories, but it is also a direct result of gains in interest income for the firm.
9
Due to the unique capital structures of banks, this industry is usually excluded from empirical studies. It’s widely accepted that banks have far lower equity ratios than those firms from other industries. In 2004, median bank leverage was 93% compared with 23% for non-financial firms (FED, 2011) in North America. This will have increased since then because of Basel III capital requirements, however the gap remains vast. Berger (1995) identifies this and points to unique institutional factors such as the Central Bank deposit insurance safety netas a reason for financial institutions having the lowest capital rates of all industries. There is a ‘too big to fail’ moral hazard issue which the banking institution exhibits which is absent in all other industries. Unlike other industries, the failure of a financial institution which holds deposits from the public has a devastating effect on the wider economy, so this will not be allowed by the government. Flannery and Rangan (2008) find evidence to support this claim, drawing on the change in bank capital structures around times of regulation changes. During the late 1980’s, where the U.S government amended the regulations for failing banks, equity levels subsequently increased to the highest level in 50 years (mid 1990’s). This shows some evidence supporting the notion that regulations do have an effect on bank capital structures. Although, evidence from the past 20 years shows that banks do not appear to hold the minimum amount of equity required by regulators (FED, 2011). There are more stringent regulations enforced in developed nations, especially after 2008, therefore I expect to see a far greater increase in equity ratios after 2008 in developed nations. Looking at appendix (1a), it’s clear to see that developed nations, especially the EU have the lowest equity levels. Regulatory effects look not to be the driver in developing nations, and have more of an impact in developed nations which operate far closer to ‘critical’ levels. North America deviates from EU levels, and equity ratios increase dramatically throughout the course of the 1990’s. By 2001, North American banks were holding 75% more capital than regulatory minimums (Flannery & Rangan, 2008).
Capital structure choices in publicly traded banks for the EU15 were found not to be driven by regulatory minimums (Gropp & Heider 2008). Evidence shows that banks optimise capital structure choices in the same way that non-financial firms do, apart from when equity levels come close to the regulatory minimum. Furthermore, evidence also shows that banks capital structures are
determined by some unobserved time invariant bank fixed effect. Put in more simple terms, an unobserved factor specific to the individual bank which drives capital structuring choices over the long term. This latter trend which points to some unobserved fixed effect points to the neutral mutation hypothesis.
10
Looking at capital structuring from a corporate governance standpoint, it’s important to mention the agency costs which result from capital structure decisions in the financial industry. With ownership and management separation, different leverage levels lead to different types of agency costs. Banks with high leverage levels and low equity levels can decrease agency costs by aligning management interests with those of shareholders. A firm operating at high leverage ratios close to bankruptcy encourages management to undertake safer investment choices. Failed investments will affect management through loss of income among the other consequences of firm bankruptcy. Less severe consequences of financial distress will also affect firm management such as a downgrading of credit risk. However, there are also agency costs with debt holders. At higher levels of leverage, increased risks of bankruptcy costs leads to risk shifting towards debt holders. Increases in leverage, whilst reducing agency costs for equity holders, is submerged into the greater effect of increased overall agency costs (Berger & Patti, 2002). Whether these trends will significantly affect firm value through stock price return is questionable. There will be other factors such as increased interest income which override the concerns over agency costs, impacting negatively on firm value.
iii)
Capital structuring in developing countries is not a topic which has had much research in the past. Booth et al (2001) investigates the trend of capital structuring in 10 developing countries and compare to developed countries. A key finding is that firms in developing countries’ capital structures are determined as a result of the same variables as in developed countries. Specifically, the pecking order hypothesis is seen as a consistent explanatory factor. This is deduced from the observation that on a consistent basis, the more profitable the firm, the less the debt ratio, therefore leading to the assumption that whenever possible the firm will pay down debt. One finding not consistent with developed countries is the scope to which macroeconomic factors have an influence over capital structures. Factors such as GDP growth, inflation and interest rates play a much larger role, most likely due to the fact that these economies are more unstable and therefore a smaller change will see a larger reaction. Chen (2003) examined how differences across
institutional settings would affect capital structures by studying the differences between Western firms and Chinese listed companies. While Booth et al studied developing economies, the countries studied had very much western institutional values underpinning their economies, whereas China contrasted to Western economies in this sense. A key finding was that firms in China bore far less
11
long term debt on their books and had a strong preference to short term debt financing. However, neither the pecking order nor the static trade-off model had any significant explanatory power in explaining the key determinants of capital structures due to the fact that these models were developed on western institutional ideologies. These differences could arise from different legal or corporate governance structures, and also the degree to which the government controls and regulates the economy. This is an example of how cultural differences can heavily influence capital structuring choices.
Motivation and Relevance:
The first part of my investigation into the relationship between capital structure choices and firm value consists of the evaluation of the effects these different choices have on firm value through its stock price. Two main subsections for the analysis are:
1. Geographically across 4 regions: Asia, South America, North America and Europe
2. Before and after 2008: Retaining the geographical division, I will also appraise pre and post 2008 relationships.
One motive for this examination is that the observation of a relationship could result in highly beneficial findings for investors and firm management. Both a positive or negative relationship would reveal much about how the market reacts to capital structuring decisions. The implication of these findings would be firm management acquiring the tools to alter something which they in large part control through bond or equity issuance/repurchasing. From the literature above it can be seen that this subject continues to be ambiguous with no one set of determinants for capital structure clearly agreed upon by empiricists. Furthermore, previous studies indicate the existence of two clear findings in many cases:
1. Firms consistently show differing levels of debt and equity.
2. The movement of debt and equity levels are not always towards a specific target level, and can often appear erratic.
The reasons for the prevalence of the above trends have no real explanation and to an extent show market inefficiency. There is the possibility of firm capital structure choice to have a signalling effect to investors. In a period following a financial crisis which was sparked by excessive leverage, it would seem illogical not to investigate whether or not the investing public have picked up on the capital
12
structure ratio as an indicator of firm health. The best way in which to measure this is through publicly traded stock prices. This is the closest one can get to an indicator of public opinion in corporate finance.
The theories above (in literature) are torn as to whether higher or lower debt levels are preferable. The pecking order theory suggests that debt is the most favourable due to information asymmetry. The trade-off theory likewise suggests that to an extent, increased debt levels lead to tax savings and generally a high level of debt is seen as the source for higher profits.The advantageous effect of debt may be seen especially in banks compared to other firms, although its cause may not be initially apparent through its omission of the two main theories (above). The reason is its unique role in mediating deposits and loans, through which a bank generates its main source of income. This unique make-up of banks’ capital structures have long seen these institutions discriminated against when it comes to studies within this field of finance.
Whereas ‘normal’ firms have debt made up of issued corporate bonds, banks take on a high level of debt, with the majority of debt often from customer deposits. This in turn provides banks with more capital with which to make loans, in which higher interest rates will be charged (compared with the deposits held). Therefore, higher levels of debt (or even just higher levels of customer deposits) would result in a greater level of income generated from the interest (differential). Whereas in most firms capital structure choice is simply a vehicle to finance other income generating activities, in banks it is the income generating activity. Consequently it’s plausible to hypothesise a negative correlation between equity/leverage ratios and stock prices.
A brief example would demonstrate how both a financial and a non-financial firm can use debt. We assume both hold $100m in liabilities, with the debt bond type being a 5 year 3% corporate bond. The ‘normal’ firm would be minus $3m at the end of the 5 years, due to its 100% bond debt liability make-up. The financial firm however would typically hold 70% liabilities in deposits (subsequently loaned out), and 30% in corporate bonds. The net interest income for a bank could typically be 2% per annum. Calculated over 5 years, the financial firm would actually make $5.79m from its liabilities.
However, never before has there been such a readily apparent risk of over-leveraging for banks. This leads to a dilemma for the investing public. Where is the tipping point, or where do investors pick up on excessive, troubled leverage, as opposed to higher income generating leverage? Or is it the case, as moral hazard has dictated in the past, that the average investor will not be able to probe the risks
13
of leverage and only recognise its benefits until the occurrence of another financial crisis, or until the firm begins to have trouble paying off its debts?
If we combine the income generating benefits with information asymmetry and tax shielding
benefits of debt, one would presume that banks aim to hold as little equity as possible. This brings us full circle to ask the question of why financial firms hold more than the required amount ofcapital as stated by regulatory bodies such as Basel II, III. Furthermore, why do all banks hold varying amounts of equity? Presuming market efficiency, taking into account the above arguments, firms holding more than the median amount of equity would be pushed out of the market.
Cultural Segregation:
Cultural trends inadvertently affect our choices in everyday life. This can extend to corporate finance, and will affect both how firm management set their capital structures demonstrating differing risk appetites. This will also extend to investor reactions. As will be outlined below, in order to gauge this impact I have attempted to keep this study as culturally representative as possible. We can also observe to an extent whether an extra buffer needs to be present in less developed areas of the world to attract investment. Even though the financial crisis began in the USA (or earlier with Northern Rock), it’s firms in the developing world which exhibit the biggest risk to investors. Political and macroeconomic risks in these areas of the world make it plausible to assume firms in developing countries would need to hold even more equity in their capital structure. This
heightened risk is seen every day in methodologies used in investment banks, where it is accepted that firm values are altered to include a risk premium in emerging markets (Damodaran, 2009). This might affect the equity/leverage ratio and firm value relationship and it might make a capital structure ‘signal’ more necessary in these parts of the world.
However, investors are also generally less aware and less sensitive to changes in firm health in less developed countries. The availability of easy buying and selling of stocks is reduced through less access to the market, as well as less news coverage and real-time information. For these reasons it is assumed that any changes in a firm’s financial health will have a much more severe reaction in a developed country where there is an immediate reaction through news outlets and the internet, and the markets are heavily traded every day. Hence, higher liquidity to start with exacerbates a liquidity crisis when negative investor sentiment is prevalent.
14
Data:The data was obtained for the financial statements via the Bureau van Dijk bankscope database via WRDS. The data for the security prices was obtained from a range of sources due mainly to the fact that for some of the less developed countries, this data was harder to find and therefore not available on mainstream websites (below is an overview).
By financial institutions, what is meant is retail deposit banks. Hence, financial institutions focused solely on corporate investment (i.e investment banks), were excluded. This is because the make-up of these could be vastly different to a bank where main profit sources come from taking and issuing deposits and loans respectively.
Three main prerequisites existed to qualify for the datasets:
1) The bank is a large, nationwide bank: Because this study is across developed as well as
underdeveloped countries, to try make the samples consistent for comparisons across regions, only the largest banks of each country were taken. This is because of two reasons: Firstly, even if the smaller banks in the more developed countries are publicly traded or have available data, it's highly unlikely that the smaller banks in underdeveloped countries will have stock price data. Secondly, there have been studies which suggest that there is a trend finding that larger firms have a lower Equity to Leverage ratio, but when comparing between the largest firms, the trend was absent - therefore comparing a the largest banks only in one continent against a mix of large and smaller banks in another isn't consistent.
2) The bank is publicly traded: Because I am investigating whether changes in ELRs have an effect on value, stock price data is a prerequisite. Moreover, only firms which are traded on a frequent enough basis are taken - there were some only with OTC stock, prices very rarely changing due to absent liquidity.
3) The bank is not a branch of another parent in another country/continent: This was because I did not want to include banks in a sample which only have a small number of overseas offices in the country, and are controlled from another country. For example, an ING branch located in South America. Even if data were available in this situation, I can't be confident that this is representative of South America, not Europe.
15
There are four groups being investigated, which represent four different 'cultures': 1) North America: USA & Canada
2) Europe: EU15 (minus Luxembourg)
3) Eastern Asia: India, South Korea, Japan, Thailand, Indonesia, Japan, Singapore 4) South America: Mexico, Brazil, Colombia, Peru, Argentina, Chile
China was excluded mainly due to institutional factors which made certain data very hard to access in this country. Secondly, Mexico is a North American country, but was included in my South America dataset due to the fact that indicators representing levels of development and cultural factors were suited more to South America. Thirdly, there are notable countries left out from all datasets (apart from North America):
Eastern Europe: This region was excluded because even though it's in the same continent, I adjudged this to be a separate region in terms of development levels (income, gdp etc.). Therefore it wouldn't fit in with the 'sample' or 'culture' I have in mind for Europe. As well as this, there is the problem that many of the countries don't have reported information.
South America: Many countries are excluded such as Venezuela, Bolivia etc. This is once again due to the lack of reliable, or any, reporting on banks. As well as this, the second criteria of having the financial institution floating on an exchange (public) was either non-existent in these countries, or extremely hard to obtain reliable stock price data.
Africa: This region was excluded as a whole because of lack of reporting information and lack of publically traded banks.
One issue which could lead to biased accuracy across the cultures was the sample size. Europe consisted of many more financial institutions than the South American sample where there were far fewer financial institutions which fit the two criteria.
16
Due to the fact that my sample was further reduced to only the largest banks (by TA) in the respective countries as well as public banks, for the sample selected, mostly there were full sets of data on all variables needed.
To normalise nominal variables such as Net Income, Total Assets and Stock Price, I transferred them into log form. This took care of the discrepancies which would arise across currencies, for example - a 10 US dollar change in stock price and a 1000 Mexican Peso change in stock price are obviously vastly different. Attention was paid to make sure that variables of the same bank (unit) were all in the same currency however.
When filtering down my sample into firms which were publicly traded stocks, firms with only over the counter, extremely illiquid stocks were also excluded - this wouldn’t have been representative and fair to compare against a highly traded security which would have a much higher price
sensitivity and a more true measure of the public response.
My data was taken over only annual periods which was a result of including underdeveloped countries, otherwise it would be helpful to have a higher frequency (quarterly), which would increase the accuracy of the data.
The time period observed was predominantly 2000-2013 for the majority of firms. This was a slightly unbalanced panel dataset as a consequence of taking firms from underdeveloped areas of the world where either the bank did not become public until recently, or there were no financial reports until recent years.
Variable Descriptions:
Note – Winsorized Variables: All variables shown below were winzorised at the 5% / 95% levels, to
remove the influence of outliers. In the data sample there were large outliers which, when removed, altered the outcome of the regression estimates.
Log Stock Price: This was used as a measure of firm value. It should be noted that the natural
logarithm of stock price was used for the entire first section, even when described only as ‘Stock Price’. Because the observed relationship was the impact of firm value on capital structure choices, stock price is the best way to gauge a public’s reaction to a firm choice. Stock price was obtained on
17
an average annual basis. More frequent reporting is desirable but due to the annual reporting of the other variables, the average was taken.
𝑃𝑟𝑖𝑐𝑒 = ln ( (𝑃𝑟𝑖𝑐𝑒𝑡+ 𝑃𝑟𝑖𝑐𝑒𝑡+1365… … + 𝑃𝑟𝑖𝑐𝑒𝑡+364))
Equity Leverage Ratio: This was used to measure firm leverage. Other studies such as Welch (2004)
commonly use Equity/Total Assets as a ratio for leverage when investigating financial institutions. Equity/Leverage was used to measure not only equity levels, but these in relation to debt and deposits.
𝐸𝐿𝑅 = ( 𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝐸𝑞𝑢𝑖𝑡𝑦
𝐵𝑉 𝐸𝑞𝑢𝑖𝑡𝑦 + 𝐵𝑉 𝐷𝑒𝑏𝑡 + 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠) ∗ 100
Deposit Ratio: This ratio is used to measure the change in deposits relative to firm leverage. Total deposits are made up of both bank and consumers.
𝐷𝑒𝑝𝑜𝑠𝑖𝑡 𝑅𝑎𝑡𝑖𝑜 = ( 𝐵𝑎𝑛𝑘 𝑑𝑒𝑝𝑜𝑠𝑖𝑡𝑠 + 𝐶𝑜𝑛𝑠𝑢𝑚𝑒𝑟 𝑑𝑒𝑝𝑜𝑠𝑖𝑡𝑠
𝑇𝑜𝑡𝑎𝑙 𝐷𝑒𝑝𝑜𝑠𝑡𝑠 + 𝐵𝑉 𝐸𝑞𝑢𝑖𝑡𝑦 + 𝐵𝑉 𝐷𝑒𝑏𝑡) ∗ 100
Control Variables:
Log Net Income: Natural log of Net Income was taken to normalise across firms and ensure normal
distribution. Net Income is used as a control variable for the main regression because it’s a large explanatory variable for stock price changes. The addition of this increased R-squared and reduced standard errors significantly. This is also used as a dependant variable in one regression below. 𝐿𝑜𝑔 𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒 = ln (𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒)
Log Index: The natural log of the stock index was used as a control variable when measuring stock
price as the dependant variable. Log form was taken to normalise across currencies and ensure normal distribution. Stock index is a very significant explanatory variable of stock price, reducing standard error and increasing R-squared. The below example is North America:
18
Lag (ln) Stock Price: Because of the autocorrelation of stock price over time, a lagged variable is used to control for the stock price.
𝐿𝑎𝑔 𝑃𝑟𝑖𝑐𝑒 = ln (𝑃𝑟𝑖𝑐𝑒)𝑡−1
Lag Equity Leverage Ratio: Firms are often working towards a target capital ratio, hence making the ELR auto correlated. Therefore, lag ELR is used as a control variable. This accounts for the previous periods Equity ratio, so the movements of Equity ratio in the present period can be more accurately captured.
𝐿𝑎𝑔 𝐸𝐿𝑅 = 𝐵𝑉 𝐸𝑞𝑢𝑖𝑡𝑦𝑡−1
𝐵𝑉 𝐸𝑞𝑢𝑖𝑡𝑦𝑡−1+ 𝐵𝑉 𝐷𝑒𝑏𝑡𝑡−1+ 𝑇𝑜𝑡𝑎𝑙 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑡−1
Fixed Effects: For my regressions, I used fixed firm and time effects estimators. This method is used to capture firm and time specific effects which could not be controlled for by independent variables, and would have an impact on the standard error if left unaccounted for.
Lagged Variables:
Two lagged variables were included to control for the non-stationary tendencies of stock price and ELR.
Investors will often use historical price trends to predict future prices because stock price
momentum is such an important factor. When a firm’s stock price is increasing, there is an incentive to go with the crowd a ‘momentum investor’. If we look at the median stock prices of the firms across the four samples in appendix (3a), trends can be seen especially strongly in Asia and South America, not so much more in Europe and North America. This is primarily due to the faster growth of emerging markets, although this would be controlled mostly by the index. Because of this,
supported by the Dickey-Fuller statistics in appendix (3b), I have chosen to include Stock price (t-1) in my final regression.
19
For the ELR, there are many theories which suggest that firms work towards a capital structuring target, meaning that this year’s debt or equity issuance was not done irrespective of last years. Combined with the Dickey-Fuller tests for this variable as well, I have decided to include the lag. Both lags, when included also increase the adjusted R-squared.
Clustered standard errors are used in all estimates. This is because of the prevalence of panel data and the danger of errors being underestimated otherwise.
Fixed Effects:
Fixed firm and time effects were used. Fixed firm effects were used due to the fact that there can be unobserved factors which affect a firms stock price which change from firm to firm and country to country, which make up the final sample. For example, banking regulations can differ throughout countries in a sample or even stock trading regulations, but these will stay constant over time and affect a firm’s value. To build upon this, when I ran the Hausman test to compare the random vs fixed effects models, the results were large enough to warrant using only the fixed effects model. Although the figure was significantly larger in the EU (328) compared with NA (23), for consistency, I decided to use the firm fixed effects models. I ran both random effects and firm fixed effects, and the comparisons can be seen in tables (I – IV).
I chose to include time invariant dummy variables in my model. This was to guard against the effect of fluctuations in unobserved effects changing over time but the same for all firms. For example, if we are to think about how the access to trading accounts has changed since the year 1991 to 2013, along with the access to information through internet, this will have a significant impact on the relationship. The individual investor now has as much access to financial statements as institutional investors, as well as the same access to a trading account. This means a much higher sensitivity to stock value to good/bad shocks today compared with 1991.
Variable Summary Discussion:
Looking at the line graph in appendix (1a), it's clear to see that median Equity/Leverage ratios are consistently the lowest in Europe. They are fairly similar in Asia and North America, and consistently
20
increasing in North America. South America seems to show by far the highest equity/leverage ratios for the financial firms in the dataset on a consistent basis. Looking at the means (1b), Asian firms have the highest mean ELR by a significant margin, but also a standard deviation of 87.3, much larger than the rest of the groups. The reason for this is because the Asian dataset is a lot more widespread and diverse stretching from India to South Korea, it could be argued these countries are at different levels of development and therefore exhibit vastly different tendencies, hence the reason for the much greater spread in this dataset – we will take the median to get a truer reading of the average. Equity/Leverage (ELR) ratios consistently fluctuate around the 12.5-15% level for South America (SA) for the period observed. The trends are strikingly similar for North America (NA) and Asia, both exhibiting steady increases in ELR in an almost linear fashion finishing around the 10% level as of 2013. Europe (EU) has the undisputed lowest levels of ELR, and what is interesting to note is no significant change after the financial crisis either.
Where there is the gradual increase of ELRs in North America and Asia, especially after 2008, in Europe there is no trend. This increase in equity ratios is predicted due to tighter regulations after the financial crisis, but this was absent in Europe. Also notable is the much higher equity ratios in South America compared to the other regions. One reason for this could be that the institutions exist in a developing region of the world. A consistently higher level of equity is needed to give investors’ confidence. An example of this in practice is the developing country risk premiums put on discounted cash flow analysis of firms in developing regions of the world by investment banks. To attract investment and deposits, these banks need higher capital buffers as an added level of security for investors.
Looking at appendix 1b, it’s clear to see there are also large fluctuations in the deposit ratios of the banks. Of the four regions, South America exhibits the lowest proportions of deposits in its leverage make-up. This is followed by the EU, and then North America and Asia. Whilst one would presume that North American banks would behave similarly to European banks, this is not the case. In both equity/leverage ratios and the components of leverage, North America exhibits more characteristics of Asia than of Europe. It’s possible that this is due to institutional factors. Japan and South Korea had their financial systems put in place by North America during post war economic recoveries, whereas EU systems have a much richer history. Therefore, these trends seem to be affected more by institutional factors than similar levels of growth or development.
21
Because equity/leverage ratio and stock price both directly affect each other, this makes measuring this relationship in only one direction especially difficult. This is a problem of simultaneous causality, which if goes untreated can affect the reliability of results. To attempt to address this I have
implemented a fixed effects model (above), and I will also be including lag variables for the
dependant variable to try and reduce the effect of price momentum. It’s also important to note that in ELR (equity/leverage) ratios, it is the book value of equity which is used. This is a key factor because book value and market values of equity do not move together, and exist at differing levels. This decreases the chance of simultaneous causality existing in many regressions where the market value equity (stock price) and the book value of equity exist as dependant and independent
variables.
Regression Specifications (Part 1):
Regression specification to measure changes in ELR on Value (table I –V, VII-VIII, XII-XVII):
Log(SP)i,t= α + ELRi,t+ ELRi,t−1+ SPi,t−1+ ln(NI)i,t+ ln(IND)c,t+ TIMEt+ FIRMi+∈
Regression specification to measure changes in ELR on Net Income (table VI):
Log(NI)i,t= α + ELRi,t+ ELRi,t−1 + TIMEt+ FIRMi+ ϵ
Regression specification of changes in deposit on value (table IX):
Log(SP)i,t= α + DepRatioi,t+ SPi,t−1+ ln(NI)i,t+ ln(IND)c,t+ TIMEt+ FIRMi+∈
Regression specification of changes in deposit on ELR (table XI):
ELR𝑡 = α + DepRatio𝑡+ TIME𝑡+ FIRM𝑖+ ϵ
Regression specification of changes in ln (deposit) on Net Income (table X):
Ln(Net Income)𝑡 = α + ln (deposits)𝑡+ TIME𝑡+ FIRM𝑖+ ϵ
For all of the above;
𝛼 = 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡
22
𝑇𝐼𝑀𝐸 = 𝑇𝑖𝑚𝑒 𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝐹𝑖𝑥𝑒𝑑 𝐸𝑓𝑓𝑒𝑐𝑡𝑠
∈ = 𝐸𝑟𝑟𝑜𝑟 𝑇𝑒𝑟𝑚 (𝑐𝑙𝑢𝑠𝑡𝑒𝑟𝑒𝑑 𝑠𝑡. 𝑒𝑟𝑟𝑜𝑟𝑠 𝑓𝑜𝑟 𝐹𝐸 𝑝𝑎𝑛𝑒𝑙 𝑑𝑎𝑡𝑎 𝑢𝑠𝑒𝑑) 𝑎𝑡 𝑡𝑖𝑚𝑒, 𝑡, 𝑓𝑜𝑟 𝑓𝑖𝑟𝑚, 𝑖
Results:
Results part 1 – Comparison of 4 Cultures (table V, app):
Overall, the highest Equity/leverage ratios exist in the developing countries, and the lowest ratios exist in EU. At first glance, this suggests that low equity ratios exist in the most competitive of environments. The EU is the most densely populated area in this study in terms of top tier banks and has by far the lowest ratio, whereas it’s the opposite in South America. This makes sense wherein it will tie in with the pecking order theory that debt issuance is always the first option of any firm. This shows that the most competitive firms cannot afford to have as much equity as those in less dense environments. That brings us to the question of equity ratios and their effect on firm value.
Table V in appendix displays the respective final regressions of equity leverage ratios on firm value. It’s clear that in all, taking the results with all independent variables included as well as fixed firm and time effects manages to increase r-squared and t-ratios as high as possible. Looking at the results, North America, South America, and Asia exhibit no significant relationship whatsoever. In addition to coefficients being small, we cannot take them with any level of significance, therefore must discard these figures.
Europe is the only region which exhibits a significant coefficient, of -0.0296 (p<0.05). According to these results, in the EU, financial institutions, stock price will increase 2.96% for every 1% decrease in the equity/leverage ratio:
(ELR = Equity/Leverage x 100 & stock price = ln(price)) Log-Linear Model: 1 unit change in βi = 100 x βi % change in Yi
1 unit increase in ELR = [(Equity/Leverage) x 100] = [100 x -0.0296]% decrease in stock price This result goes directly against a signalling hypothesis, and works in favour of a pecking order theory (assuming investors react to this). This also raises the question of why there is no relationship in other parts of the world if equity decreases do increase value in Europe in line with the pecking
23
order theory. As well as this theory of capital structuring, due to the fact it is banks being analysed, there is even more reason for lower equity ratios of equity to be beneficial - a banks’ main source of revenue is through deposit interest income, which is part of the non-equity portion of liability (results part 3). Hence, an increase in deposits could be argued as almost a proxy for increase in net income. The issue with this line of thinking is exploited in table VI, in which a panel of regression results is shown. This was a fixed effects regression of (log) net income (instead of stock value), on the equity/leverage ratio. What we can see is that in the EU which founded the previous only significant relationship, the coefficient for here between ELR and (log) Net Income is 0.0200, with a 0.41 t-statistic, making it statistically insignificant. All other regions also display statistically
insignificant results. If this is the case for the EU, then investors on the stock market react favourably to lower equity ratios whilst these have no actual effect on income statement measures of
profitability, net income In this case.
It could be that the perceived advantages for firm efficiency of a low equity leverage ratio is counterweighted by the aversion to investing in risky, over leveraged companies in other parts of the world for investors. When considering an investment, the safety of the overall economic climate could be higher up on the list of importance to the investor than the corporate structure of the firm. There is also a potential bias to take into account when considering these results. The European dataset contains a very high density of large, trustworthy banks. For example, the largest in each of 15 countries were taken for the EU which are therefore all industry leaders, whereas for the USA only the top five measured by total assets will be real industry leaders. Hence, the relationship exists in the EU only because the dataset was at such a density that it contained only the largest banks which are trustworthy no matter what the leverage ratio. The lower proportions of equity in the capital structure leads investors to see only greater profits rather than higher risk investments in Europe, whereas in other parts of the world the opposite is true. A large part of the reason for this is also down to the security in the European economic climate which comes with a stable economy and the guarantor of a reliable central bank.
24
Results part 2 – Before and after the financial crisis: (See tables VII, VIII)
As stated in my initial set of aims and hypothesis, because of the prevalence of the largest financial banking system collapse this century, it’s plausible to hypothesise that there might be a positive relationship between ELR and firm value post - 2008, due mostly to one main factor – investor awareness. As frequently studied in the field of behavioural finance, there is generally a lot of ‘noise’ in the market. This is investors picking up on signals that are not really relevant to company
performance but more based on emotion than inputs. The emotions could be more averse to risky investments with the financial crisis fresh in the mind of the average investor.
There are no findings which can be reported as statistically significant. However, it is interesting to observe the starkly different coefficients for the EU, Pre and Post 2008. There is a much stronger negative relationship after the banking collapse of 2008, -0.0721, as opposed to a positive coefficient in the period preceding the collapse, 0.0211. Even though we cannot state these as being statistically significant, this runs opposite to logic. After a financial collapse caused primarily by overleveraging financial institutions, the relationship between lower leverage and a higher stock price grows to very strong levels.
Results Part 3:
Now we have seen that the EU is the only region where a significant negative relationship of this kind exists, we need to try and understand why. Interest income is the greatest income generator for banks, so there is a question of the relationship between the proportion of deposits in liabilities and some measure of value generation.
An advantage of looking at financial institutions is that because they’re unique in their make-up, we can delve further into their capital structures and break down the debt into two parts. As can be seen in table IX, there is no statistically significant relationship between customer deposits and firm value, apart from Europe where we observe a positive relationship of 0.338. This reinforces the argument made in part 1 of my discussion – the negative ELR, firm value relationship exists because
of the proportion of banks’ income generated from deposit interest income. A statistically significant
positive relationship would seek to reinforce this, as investors recognize that when deposits increase, profits also increase. Ironically, if we observe figures II – V, Europe has the lowest
25
proportion of deposits compared to the other regions. Asia exhibits among the highest deposit/liability ratios of all regions observed (figure IV). The ratio used here is total customer deposits/total leverage, capturing the amount of total leverage is made up of customer deposits. Looking at the graphical representation of this ratio across the 4 cultures in figure VII, as seen with the other ratios, there is a large and consistent difference across the regions. The capital structures of banks in the EU are very evenly weighted between customer deposits and other forms of
liabilities, floating around the 50% mark. In Asia however, deposit ratios are much higher, floating around the 80% mark from 2005 onwards. For the other two regions (NA, SA), these are growing in a very linear fashion from EU levels (2000) to breach Asian levels by 2013.
There are very deliberate differences in deposit levels across the four regions, and seemingly non-random. There are two possible explanations for this huge difference. Either there are differences in the relationship between deposit ratios and market firm values across the cultures, however this is not the case as seen by three out of the four regressions. The alternative is that corporate finance simply hasn’t agreed upon the most profitable level of customer deposit/total liability level to maximise profits or we would see all regions would be floating around a similar level. The most prevalent finding here seems to be that deposit ratios are massively dictated by the firms’ country of origin. This is an example of how cultures have impacted capital structures in a sizeable way. It is surprising to see just how consistently and significantly EU banks are different from other areas of the world. This is possibly a reflection of the more regulated environment in which European firms operate.
The pecking order is a theory of capital structure which hypothesises that firms fund their operations using safe, junk debt and then equity only as a last resort. It’s an assumption based on the premise that debt is favourable to equity because it’s cheaper primarily due to tax savings. Within financial firms however, we have three types of financing – equity, deposits, and those liabilities not included in deposits – primarily issued debt. As we have already seen, there is a big gulf in ratios of these across the cultures. What we will now look at is how the overall leverage ratio (Equity/Liabilities) and the deposit ratio (Total Customer Dep/Total Liability) interact with each other. Hence, I am looking for a pecking order within a pecking order. If firms choose debt over equity as a first source of funding, then which kind of debt?
The issue with the hypothesis that firms prefer either deposit or non-deposit debt is that firms cannot control customer deposits like they can with the issuance or repurchase of debt/equity. In
26
other words, firms cannot choose to have more customers deposit money, they can only react when this happens. The main drivers of deposit levels are out of the control of commercial banks, and in the control of central banks and governments – factors such as interest or inflation rates. Just as firms cannot control their stock prices, however it would always be preferable to have a higher stock price, deposits are similarly dictated by outside macroeconomic influences. However, as is
commonly accepted, the more deposits, the more beneficial this is to the profits of the bank. This is shown in the results table (x) in appendix, wherein there are statistically significant positive
relationships in all regions of an increase in ln(total deposits) on ln(Net Income). The second strongest relationship in Asia makes sense. Asian firms hold mean approx. 80% of liabilities in customer deposits (figure IV & VII), so therefore the relationship shown in table (X) is
understandable – in Asian banks, a 1% increase in total deposits results in a 0.54% in firms’ net income. In the US, and particularly the EU however which has the lowest ratios hovering around the 50% mark, firms would surely benefit from an increased deposit/liability ratio. An increase of 1% in deposits results in a 0.46% increase in net income (EU) and 0.46% increase in net income (NA), both significant at p<0.05.
27
Further study into the EU:
Because of the size and diversity across the EU, it’s appropriate to break down the EU into separate regions to examine whether there are any contrasting findings in different areas of the EU. Appendix table XII shows the EU split into north and south regions. The reason for this split is due to the fact that in economic terms, it is common knowledge that the north and south of Europe have stark differences in cultural norms. Northern Europe is seen as the driver of EU’s growth, as well as being the driver of regulation in the financial sector, whereas Southern Europe is well known to be the weaker of the two regions.
These two sets of regressions (table XII) show the results when the EU is split into North and South. What’s clear here is that although there are no significant results to observe, there is also no significant difference in the relationship being observed either. The gulf in performance and the cultural aspects of northern European economies and southern European economies does not translate into having any effect on the relationship being examined.
28
Results part 4:Firms closer to the capital regulatory levels have been seen to behave differently to other banks in previous studies. Results tables XIV-XVII show the results of these investigations this relationship across regions. To proxy for firms close to the regulatory minimum in each region, those firms with equity/leverage ratios in the bottom quartile (25%), are contrasted with all others, hence firms in the region in the top 75% of equity/leverage ratios. The results are insightful.
The EU shows no statistically significant negative relationship founded earlier in either the quartile or top 75% alone.
North America shows quite extreme results. The top 75% of firms exhibit no significant positive or negative relationship. The lower quartile of firms show that a 1% increase in ELR would result in a 22.3% increase in stock price (p<0.01).
We know these results aren’t driven by outliers because the data is winsorized at the 5%, 95% levels. The sample size is not large enough to draw definitive conclusions here, however it does provide some interesting trends to observe. What this tells us is that it’s possible that investors do recognise when firms are dangerously low to the regulatory minimum. An increase in equity ratios tell investors that the firm is more stable, less prone to financial distress. Therefore, the lack of a positive relationship up to this point is now being observed, but only when firms’ equity/leverage ratios fall into the bottom quartile.
South America shows a 1% increase in ELR results in a 6.67% (p<0.05) increase in stock price for the bottom quartile of firms. The top 75% shows no significant relationship. In this region, along with North America, it’s only firms in the bottom 25% of equity/leverage ratios which see market value benefits from an increase in equity in their capital structures.
29
Readjustment Hypothesis:
To test the readjustment hypothesis of capital structures, I shall draw on the methodology of Welch (2004). This paper tests the readjustment of capital structures using Fama-MacBeth two step regressions to obtain a beta coefficient for a firm’s readjustment of debt issuance in reaction to changes in the market value of equity. As in Welch (2004), there will be two factors measured; implied debt ratio and actual debt ratio. The original methodology used in Welch (2004) is below; 𝐴𝐷𝑅𝑡+𝑘= 𝛼0+ 𝛼1∙ 𝐴𝐷𝑅𝑡+ 𝛼2∙ 𝐼𝐷𝑅𝑡,𝑡+𝑘+ 𝜖 Where; 𝐴𝐷𝑅𝑡 = (𝐷𝑒𝑏𝑡𝑡) (𝐸𝑞𝑢𝑖𝑡𝑦𝑡+𝐷𝑒𝑏𝑡𝑡) 𝐴𝐷𝑅𝑡+𝑘= (𝐷𝑒𝑏𝑡𝑡+1) (𝐸𝑞𝑢𝑖𝑡𝑦𝑡+1+𝐷𝑒𝑏𝑡𝑡+1) 𝐼𝐷𝑅𝑡,𝑡+𝑘=(𝐷𝑒𝑏𝑡(𝐷𝑒𝑏𝑡𝑡) 𝑡 +𝐸𝑞𝑡∙𝑥𝑡,𝑡+𝑘,) 𝑥𝑡,𝑡+𝑘= 𝑠𝑡𝑜𝑐𝑘 𝑝𝑟𝑖𝑐𝑒 𝑐ℎ𝑎𝑛𝑔𝑒 𝑓𝑟𝑜𝑚 𝑡𝑖𝑚𝑒 𝑡, 𝑡 + 𝑘
Where 𝛼1= 1, 𝛼2= 0, it represents a perfect readjustment hypothesis. Where 𝛼1= 0, 𝛼2= 1, it represents a perfect non-readjustment hypothesis.
In Welch (2004), capital structure ratios were constructed using data from non-financial firms only. As is the case in non-financial firms, there is either debt or equity to consider. As I am investigating firms in the banking sector, there are also customer deposits. This three dimensional aspect to capital structuring introduces complications. There are two approaches to deal with the addition of deposits; either leaving customer deposits as part of the debt, or discarding customer deposits from the ratios all together.
In my investigation, I shall not be replicating the methodology above, merely drawing on the fairly unique application of the Fama MacBeth two stage method for my own investigation. In the first part of my investigation into capital structuring, I did not draw too much on the segregation of deposits from debt in types of liabilities. For the most part, these were treated simply as ‘debt’ or ‘liabilities’.
In this part of the investigation I shall be investigating the components of debt – mainly deposit type debt and non-deposit. The non-deposit part shall be named simply as ‘debt’. The reason behind this
30
investigation is due to the fact that deposits are a part of the debt structure of financial institutions, and unique only to financial institutions. The level of deposits are largely out of the bank’s control as customers can deposit/withdraw at their discretion, so carry a higher level of risk as a type of financing for future activities. Factors which influence the levels of deposits do not come from a bank specific policy, but from wider macroeconomic factors such as interest rate changes or inflation expectations. Therefore, the question of whether a bank readjusts the debt/deposit ratio due to a change in deposit level should be investigated.
The change in the level of deposits shall be treated like the change in stock price was in Welch (2004)’s paper. Hence, where Welch found that firms largely allow their debt/liability ratios to drift with stock price fluctuations, I shall be looking to see whether firms allow their debt/liability ratios to drift with deposit level fluctuations. The two factors of stock price and deposit levels are similar in the way that they are both fluctuating parts of the capital structure which are largely out of the firms control, and therefore lead to a reactionary actions in response to changes in them. I will run the regressions in two forms; firstly measuring the readjustment of equity and debt (liabilities minus deposits), as well as only debt (liabilities minus deposits and equity). My alterations to the methodology are below:
31
Methodology (II):
i) First Regression Specification:
𝐴𝐿𝑅𝑡+𝑘= 𝛼0+ 𝛼1∙ 𝐴𝐿𝑅𝑡+ 𝛼2∙ 𝐼𝐿𝑅𝑡,𝑡+𝑘+ 𝜖 Where, 𝐴𝐿𝑅𝑡= (𝐷𝐵𝑡+ 𝐸𝑄𝑡) (𝐷𝑃𝑡+ 𝐸𝑄𝑡 + 𝐷𝐵𝑡) 𝐼𝐿𝑅𝑡 = (𝐷𝐵𝑡+ 𝐸𝑄𝑡) (𝐷𝑃𝑡∙ (1 + 𝑦𝑡,𝑡+𝑘) + 𝐷𝐵𝑡+ 𝐸𝑄𝑡) 𝐷𝐵𝑡= 𝑇𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 − 𝐸𝑞𝑢𝑖𝑡𝑦 − 𝑑𝑒𝑝𝑜𝑠𝑖𝑡𝑠 𝐷𝑃𝑡 = 𝑇𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 − 𝐸𝑞𝑢𝑖𝑡𝑦 − 𝑁𝑜𝑛 𝑑𝑒𝑝𝑜𝑠𝑖𝑡 𝑑𝑒𝑏𝑡 𝐸𝑄𝑡 = 𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝐸𝑞𝑢𝑖𝑡𝑦 𝑦𝑡,𝑡+𝑘= (𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑡+𝑘− 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑡)/𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑡 (𝐷𝐵𝑡+1+ 𝐸𝑄𝑡+1) (𝐷𝑃𝑡+1+ 𝐸𝑄𝑡+1+ 𝐷𝐵𝑡+1)= 𝛼1 (𝐷𝐵𝑡+ 𝐸𝑄𝑡) (𝐷𝑃𝑡+ 𝐸𝑄𝑡+ 𝐷𝐵𝑡)+ 𝛼2 (𝐷𝐵𝑡+ 𝐸𝑄𝑡) (𝐷𝑃𝑡∙ (1 + 𝑦𝑡,𝑡+𝑘) + 𝐷𝐵𝑡+ 𝐸𝑄𝑡) +∈
ii) Second Regression Specification:
𝐴𝐷𝑅𝑡+𝑘= 𝛼0+ 𝛼1∙ 𝐴𝐷𝑅𝑡+ 𝛼2∙ 𝐼𝐷𝑅𝑡+ 𝜖 Where, 𝐴𝐷𝑅𝑡 = (𝐷𝐵𝑡) (𝐷𝑃𝑡+ 𝐷𝐵𝑡) 𝐼𝐷𝑅𝑡 = (𝐷𝐵𝑡) (𝐷𝑃𝑡∙ (1 + 𝑦𝑡,𝑡+𝑘) + 𝐷𝐵𝑡) 𝐷𝐵𝑡= 𝑇𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 − 𝐸𝑞𝑢𝑖𝑡𝑦 − 𝑑𝑒𝑝𝑜𝑠𝑖𝑡𝑠 𝐷𝑃𝑡 = 𝑇𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 − 𝐸𝑞𝑢𝑖𝑡𝑦 − 𝑁𝑜𝑛 𝑑𝑒𝑝𝑜𝑠𝑖𝑡 𝑑𝑒𝑏𝑡 𝑦𝑡,𝑡+𝑘= (𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑡+𝑘− 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑡)/𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑡
32
Where 𝛼1= 1, 𝛼2= 0, it represents a perfect debt readjustment hypothesis. Where 𝛼1= 0, 𝛼2= 1, it represents a perfect debt non-readjustment hypothesis.
Data:
The dataset used for the EU, South America, North America and Asia is equal to the dataset used in the first set of investigations in this paper (see page 14).
Less emphasis is placed on segregation of regions here as this is an investigation into the general behaviour of financial institutions on this subject.
In the investigation of Welch (2004), the time horizons (k) of 1 year, 3 years, 5 years and 10 years are used. I shall be using mainly the 1,2 and 3 year time horizons. The reason for this is that the time horizon of my dataset (1999-2013) and the sample size would decrease to such an extent for any time horizon beyond 5 years that it would make the results unreliable. This will also provide
opportunity to observe at a greater frequency the adjustment, if at all, of debt ratios in response to deposit level changes in the capital structure.
