Influence of Determinants on Capital Structure: An Empirical Investigation of the Energy, Industrial Machinery and Chemical & Applied Product Industry, utilizing Panel Data from 2005-2008.

(1)

Influence of Determinants on Capital Structure:

An Empirical Investigation of the Energy, Industrial Machinery and

Chemical & Applied Product Industry, utilizing Panel Data from 2005-2008.

Author: Fabian O. Umole

Student Number: 1551582

E-Mail: f.o.umole@gmail.com

Double Degree Program:

MSc International Financial Management

MSc Economics and Business

Groningen, February 2010

Rijksuniversiteit Groningen Faculty of Economics and Business

Landleven 5, 9747 AD Groningen, The Netherlands &

Uppsala Universitet Department of Economics

P.O.Box 513

SE-751 20 Uppsala, Sweden

(2)

Page | ii

Acknowledgment

This study is done on behalf of the Treasury department of Nederlandse Gasunie in Groningen, Netherlands and conducted to conclude the MSc study International Financial Management at the Rijksuniversiteit Groningen in the Netherlands and the MSc program Economics and Business at Uppsala Universitet in Sweden. The thesis1 would not have been successful without the support of several people, whom I would like to show my appreciation to.

First and foremost, I would like to thank God for his mercy, love and guidance he has shown me throughout this period of life. My supervisor Professor L.J.R (Bert) Scholtens from Rijksuniversiteit Groningen, Thank you. His support, suggestions, remarks and timing came a long way with helping me finish the MSc thesis. My appreciation also goes to my supervisor at Nederlandse Gasunie Jan van Esch for his supports, suggestions, time, knowledge and second look at my thesis. Lammertjan Dam, your assistants towards the success of this thesis is greatly appreciated. Thanks to my colleagues at Nederlandse Gasunie and my friends, especially those that stood by me during difficult moments. I just want to let you know you are “awesome”. Last but not least, my family, their unconditional support, love, advice and confidence has been the bedrock that saw me through to this stage of life. Thank you all.

Groningen, February 26th 2010

Fabian O. Umole.

1

The style guide line for this thesis is according to the Journal of Finance

(3)

Page | iii Abstract

This study examines the influence determinants have on capital structure. Moreover, the difference in capital structure between Energy Industry, Industrial Machinery Industry and Chemical and Applied Product Industry is investigated utilizing data from 2005 to 2008. The determinants of capital structure used in this study are based on previous empirical research. With leverage ratio as the dependent variable, the results from the pooled OLS regression show that profitability, size, tangibility, growth, regulatory quality have a significant influence to the dependent variable leverage ratio. While non-debt tax shield and capital intensity show an insignificant influence to leverage ratio. All pooled regression results are consistent with the results of time fixed effect estimation. Difference in capital structure is also found between Energy Industry, Industrial Machinery Industry and Chemical and Applied Product Industry. The results are generally robust using the measurement of leverage ratio, but the significance of some variables changes. Most of this study results are in line with the capital structure prediction of pecking order and trade off theory.

(4)

Page | iv

Tables

Table I: Past Empirical Studies on Capital Structure……….. 10

Table II: Definition of Variables and Predicted Signs of Explanatory Variables………...… 26

Table III: Geographical Sample Spread……….. 29

Table IV: Descriptive Statistics Sample Variables………. 31

Table V: Correlation Matrix……… 33

Table VI: Panel Least Square Pooled and Time Fixed Regression………. 37

Table VII: Panel Least Square Time Fixed Regression and Wald-Test……….. 43

Table VIII: Panel Least Square Pooled and Time Fixed Regression……….. 46

(6)

Page | 1 Capital structure debate has gained widespread attention over the last 50 years. Starting from the year 1958 with Nobel Prize winners Miller & Modigliani, who showed under strong assumptions that the manner a company is financed does not affect the firm value. However, many researchers in the past decades after the publication of Miller & Modigliani have provided evidence that capital structure choice does affect the value of a firm. Theories evolved through time and among such theories are the trade off theory and pecking order theory. The trade off theory assumes that capital structure preference is based on a tradeoff between the present value of debt and the present value of the cost of financial distress. The pecking order theory states that firms prefer some finance options above others due to asymmetric information between insiders and outsiders of the firm. Furthermore, the pecking order theory also suggests that there is no optimal debt level for a firm. The trade off theory focuses on the degree of tangibility of assets, size and non-debt tax shield (Titman & Wessels (1988); Kale, Noe & Ramirez (1991) & Sayilgan, Karabacak & Kucukkocaoglu (2002)). While the pecking order theory focuses on the flows of cash variables like profitability, capital expenditure, and dividend payout just to mention a few (Shyam-Sunder & Myers (1999)).

(7)

Page | 2 “How do variables in terms of profitability, size, tangibility, growth, capital intensity, regulatory quality and non-debt-tax shield influence capital structure for capital intensive sector firms? Are the variables positively or negatively related?”

“Are there differences in capital structure among industries that operate in the capital intensive sector?”

In order to answer the research questions, a panel data-set is utilized, which allows for measurement across time and between firms. 288 firms in 34 countries over 6 different continents are used in this study. The 288 firms are separated into 3 different industrial types namely: Energy Industry, Industrial Machinery Industry and Chemical and Applied Product Industry. All data from the 288 firms are book values and is collected from World Scope, Thomson-One Financials and World Bank data-base, while a time fixed regression technique and a Wald-test is used to show the results that answers the research question above. The variables selected are based on previous research were researchers concluded that variables such as profitability, size, tangibility, non-debt tax shield, capital intensity and growth are good determinants of capital structure. Institutional variable is also included in this study as control variable to analyze the influence institutional environment has on firms’ capital structure.

This paper contributes to existing literature due to the use of a world sample, which covers 6 different continents and focuses on industries that have never been researched. The approach is unique in the field of capital structure research. The panel data regression also adds an additional advantage to this paper because it allows the analysis of either fixed time or fixed firm effect.

(8)

Page | 3

Section I. Literature Review

A. Introduction

The main assumption of Modiglian and Miller of a perfect capital market is no transaction costs and no taxation. According to Modiglian and Miller under their

assumptions, firm value is independent from its capital structure. Modigliani and Miller

(1958) proved that the choice between debt and equity financing has no material effects on the firms’ value. However, when some of the assumptions of Modigliani and Miller are relaxed, such as a world with corporate taxes, firm value will be maximized by using as much

debt as possible (Modigliani & Miller (1963)). The theoretical aspects (assumptions of

factors in capital structure decision) of Modigliani and Miller’s (1958 &1963) models are

widely accepted (Myers (2001); Deesomsak, Paudyal, & Pescetto (2004)). But following the

study of Modigliani and Miller, considerable amount of research has emerged to determine the factors that influence a firm’s capital structure. These studies have revised and extended Modigliani and Miller’s works not only to consider the impact of taxes and bankruptcy costs, but also the impact of agency problems, asymmetric information and other frictions and

deviations from perfect markets. Existing research includes the pecking order theory and the

trade off theory, which will serve as a foundation in this study. The pecking order theory and the trade off theory will be used to answer the research questions of this thesis.

B. Pecking Order Theory

The model of the pecking order theory uses different dependent variables that describe the flow of funds and investment opportunities, including free cash flow, capital

expenditures, dividend payout and profitability (Huang & Song (2006)). In addition, Heinkel

and Zechner (1990) show that component of optimal capital structure is preferred stock because preferred stock increases a firm’s debt capacity. An increase of a firms’ debt capacity is a positive outcome because more debt leads to optimal investment decisions. More debt

also has an advantage due to the option of debt holders to liquidate investments (Grinblatt &

Titman (2002)).

(9)

Page | 4

and outsiders of a firm. When firms plan to make financial investments, they place their financing choices in an order of ranking. The order of ranking is with internally generated

funds on top of the ladder followed by debt issues and finally with equity issue (Grinblatt &

Titman (2002)).

Jensen and Meckling (1976) and Harris and Raviv (1991) suggest two arguments why firms prefer to use debt over equity. Their first argument comes when managers do hold less than 100% of equity. In such a case, the manager does bear all the cost of wealth creation for its shareholders but does not receive the entire wealth for his/her effort. This leads to incentives for management to make investments that are sub-optimal for the value of the

company. As a result, the more leverage a firm uses, lesser are the incentives the managers

will take on sub-optimal investments. More debt simply means a higher fixed claim on free cash flow that makes the free cash flow smaller. Smaller free cash flow gives less incentive

for managers to undertake on sub-optimal investments. Jensen (1986) also considers higher

fixed claim on free cash flow as one of the benefits of debt financing. The second argument why debt is preferred over equity comes from the conflict between debt and equity holder because of the difference in debt and equity contracts. On the one hand, debt holders receive a fixed pay not based on performance measures. While on the other hand, equity holders receive a pay based on the profitability of the company. If an investment yields high return (over face value of debt), equity holders capture most of this premium. However, if the investment fails, the debt holders pay the price. This means equity holders capture the premium of risky investments, while debt holders are responsible to bear the risk. The difference between debt and equity contracts leads to sub-optimal investments by equity holders in the sense that they invest in risky projects even when such projects does not create extra value for the firm. Equally, investment in projects that do not create extra value for the

firm occurs when a firm is likely to go bankrupt. There is little or no incentive for equity

holders to invest in value increasing projects because the largest part of the earned profit will

go to the debt holders (Meyers (1977)). Besides the conflict between debt and equity holder

that show why firms will prefer debt over equity, other incentives also play a role why firms will prefer debt over equity.

The signalling effect is another incentive for firms to choose debt over equity.

(10)

Page | 5

corresponds to high quality firms. Furthermore, since low quality firms cannot imitate these high debt level firms, high debt level acts as a powerful signalling tool. Other authors like

Castanias (1983) provide similar evidence that show lines of business with relatively higher failure rates have lower debt levels. He went further to conclude that the ex ante bankruptcy cost is high enough to impose and that there is enough benefits and prospects to hold an optimal mix of debt and equity.

Concluding, the pecking order theory can also be explained from the perspective of asymmetric information between managers and the market and the existence of transaction

costs (Myers (1984)). Agency cost of equity due to asymmetric information can have

consequence on the optimal capital structure of a firm, which could lead to internal financing been favoured. Whether this is true or not is investigated with variables describing the available internal funds of firms (profitability). According to the pecking order theory, these variables would be negatively related to leverage because firms’ prefer internal financing to external financing.

C. Trade Off Theory

The trade off theory suggests that the optimal capital structure is decided by a trade off between the tax benefits of debt and the cost of financial distress. The cost of financial distress include transaction cost, administrative cost of bankruptcy, agency monitoring and

contracting cost, and the cost of moral hazard (Grinblatt & Titman (2002)).

In addition to the trade off between tax benefit of debt and the cost of financial distress, a firms’ optimal capital structure is also based on the balance between the present

value of tax shield and the present value of the cost of financial distress (Meyers & Majluf

(1984)). Due to the presence of this trade off, there is an optimal mix of debt and equity where the benefit of debt comes from tax deductibility of interest payments. Therefore, an increase in debt to a firms’ capital structure helps to bring down the tax liability and raises the

after tax cash flow available to providers of capital (Drobetz & Fix (2003)). In other words,

the optimal target financial debt ratio maximizes the value of the firm. Despite the fact that the use of debt reduces the agency costs of free cash flow thereby raising the after tax cash

flow, it also incurs its own agency costs (Jensen & Meckling (1976)). Once debt is issued,

(11)

Page | 6

That means manager will tend to invest in risky projects, which offer high returns but also come with the likelihood of bankruptcy. If such risky project is successful, debt holders cannot get above the sum of debt issued and the interest payments, but the existing shareholders will receive majority of the gains. On the other hand, if the risky project is not successful, shareholders exposure are limited, while debt holders bear majority of the burden and might lose everything. The consequence is often referred to as the moral hazard and

adverse selection problem2 (Jensen & Mekling (1976)). Moreover, Myers (1977) proposes

that a highly leveraged firm may forgo some positive net present value investment due to

debt overhang problem3.

Bankruptcy cost is also associated to the debt overhang problem, as debt overhang problem is most likely to occur when a firm uses excessive debt and is unable to meet the interest and principal payment. Bankruptcy cost comes in two forms: as direct and indirect bankruptcy costs. Both types of bankruptcy cost influence the capital structure decision by setting equilibrium between the costs associated with debt and the benefit of tax resulting from debt. Direct bankruptcy costs come in the form of legal and administrative costs or from selling the assets of the firm below the market value. Indirect bankruptcy costs can be significant to firms in the sense that, when a firm runs into financial distress, its financial condition can affect the firm’s investment policies, which automatically reduces the firms’ value. The reason behind the reduction of the firms’ value is the decrease in research and development, maintenance, advertisements, and promotional expense due to poor financial

positioning of the firm (Grinblatt & Titman (2002)). So the basic distinction with the trade off

theory and the pecking order theory is that the pecking order theory suggests that there is no optimal capital structure and that possible differences in capital structure occur when there are not enough internal funds available to finance the investment.

Several authors have conducted research to determine the influence of cost and determinants on optimal capital structure. Amongst the authors are Rajan and Zingales (1995), Pandey (2001), Prasad, Green and Murinde (2003), and Buferna, Bangassa and Hodgkinson (2005), that show a positive relationship between size and leverage. However, Titman and Wessels (1988) who examine the determinants of optimal capital structure for US firms identifies different independent variables namely: volatility of earnings, size of the

2

Adverse selection and moral hazard problem refers to asymmetry of information among parties.

3

(12)

Page | 7 company, uniqueness of firms’ products, profitability, industry, growth and the value of assets. The dependent variables he identifies were classified into three categories: short and long-term debt, and convertible debt. Titman and Wessels (1998) conclude that debt levels are negatively related with the uniqueness (firm size) of firms and that transaction costs do have an important influence on debt levels. Miller (1977) also had the same conclusion. The difference in the results of uniqueness of firms found by the authors mentioned above is attributed to the chosen method of estimation (Van der Wijst & Thurik (1993)). Furthermore, Graham and Harvey (2001) who survey 392 CFOs found out that the potential costs of

financial distress are not important for firms, but firms are often concerned about their credit

rating. They conclude that credit rating can be seen as a concern for distress because firms with high rated debt view credit rating as important determinant for leverage policy.

Furthermore, Taub (1975) uses a slightly different methodology with the focus on the choice

of new financing rather than existing debt and equity levels. He found little evidence to

support conventional theories and proposes a focus on theories that are more general rather

than the use of existing ones. Just like Titman and Wessels (1988), Bradley, Jarell and

Han-Kim (1984) used mostly the same determinants to investigate 851 US firms in 25 different industries. They provided strong support for industry differences between target debt ratios and industry debt differences by utilizing industrial dummies. They concluded that there is an inverse relationship between the volatility of profits and a firms’ leverage. This explains inter

as well as intra industry variations in leverage ratios in their research. Bowen, Daley and

Hubner (1982), Scott (1972), Hamberg (2004), Scott and Martin (1975) also show evidence for differences in debt ratios among industries. The authors found that different industries with different operational risks have different financial structures. Additionally, the authors emphasize that it would be unwise to neglect industry differences when examining capital

structure. However, Balikrisham and Fox (1993) conclude that debt ratios do not significantly

differ across industries.

Finally, the trade off theory should prove a positive relationship between the costs associated variables in this study, as well as differences in capital structure among industries operating in the capital intensive sector.

D. Capital Intensive Sector

(13)

Page | 8 capital involved is much higher than the proportion of labour. The reason why the proportion of capital is much higher than the proportion of labour is because the industrial structure and industrial type of capital intensive sector requires high value investment in capital assets. Generally, a capital intensive industry generates high profits due to the large amount of capital invested in these industries; it produces a relative high rate of return, which is used for more capital investment4.

Additionally, capital intensive industry also incurs high fixed cost, which in turn requires a high level of risk because the capital intensive industry will have to generate enough income to match its high fixed cost. If sales volume declines, profits earned by the industry also experience a sharp decline due to the inability to reduce or remove the fixed cost. So in general, if market demand declines, capital intensive industries suffer more loss compared to labour intensive industries. For example industrial machinery, chemical, and oil/energy refinery industries are basically capital intensive industries, which require large capital investment for starting up and running the business5. Since capital intensive industry requires a large volume of financial resources for starting up, the number of new entrants to a capital intensive industry is relatively low compared to most labour intensive industry. Due to the relatively low entrance of new firm into the capital intensive sector, one of the benefits of capital intensive sector is that it promises a high level of productivity. High level of productivity is possible because the capital investments made are used to equip the industry with essential tools and high tech machinery (firm-specific assets) so that the use of these equipments raises the productivity of labour and thus a greater output.6

Investments in firm-specific assets do not only raise the productivity of labour and output but also enhance the firms’ uniqueness and competitive advantage (Mocnik (2001)). Firm-specific assets improve the firms’ uniqueness and competitive advantage because investments in firm-specific assets that are tailored to a firms’ strategy and technology can reduce cost, improve quality and enable differentiation of the firms’ products and services from its competitors (Mang (1998)). The firm-specific asset generates greater firm value when used by that firm than when used for other purpose because many firm-specific asset are intangible (for example advertising and R&D) and difficult to measure and evaluate (Balakrishnan and fox (1993)). For the capital intensive sector, specialized assets create both

4

Nederlandse Gasunie internal document

5

http://www.economywatch.com/world-industries/capital-intensive.html

6

(14)

Page | 9 a problem and an opportunity (Williamson (1975), (1985)). Capital intensive firms may have problems financing firm specific assets that are encumbered by debts because of the nature of the asset7 (Williamson (1985)). Nevertheless, the firms have an opportunity to create governance structures8 in a way that mitigates the problem of assets redeployment (Mocnik (2001)).

The nature of the firms’ assets mentioned by Williamson (1985) has adverse effect on the firms’ ability to borrow because often firm-specific assets cannot be redeployed to another location or used as collateral for borrowing (Mocnik (2001)). Furthermore, firm-specific assets have an impact on borrowing (debt ratio) because of bankruptcy costs that reflect the loss in firm value due to the likelihood of financial distress. In the event of bankruptcy or liquidation of a firm, its specific assets will lose a lot of value. This is why specific assets of capital intensive sector firms cannot be used as collateral. If lenders nevertheless decide to finance specific assets, the costs of finance will be higher because lenders have only limited ability to control a manager’s investments and activities. So, the transaction involving firm-specific asset is affected by informational asymmetry between insiders and outsiders of the firm. Information asymmetry and the inability for easy redeployment of capital intensive sectors assets, offer a poor security for lenders (Mocnik (2001)). Williamson (1988); Balakrishnan and Fox (1993); and Mang (1998) conclude that investment in assets by capital intensive sector firms’ are more likely to be financed with equity rather than debt. Vilasuso and Minkler (2001) also support this conclusion as the authors state that investment projects that require highly specific assets should be financed with equity, while debt financing should be used for assets that are more generic and can be easily redeployed. Therefore, capital intensity variable should be negatively related to leverage ratio in this study.

Finally, table I below is a summary of previous empirical studies that relates to capital structure. The summary of previous empirical studies serves as a motivation for the choice of variables used in this study. The authors of the research, the sample researched, the period of the research and the variables used in the research is presented below.

7

Nature of the asset means the ability to redeploy the assets

8

(15)

Page | 10

Table I

Past Empirical Studies on Capital Structure

In this table blow, a summary is given of various studies on the determinants of capital structure. Of each study, the name of author is presented, the sample frame researched, the period of research, the methodology used in testing variables, the dependent variable of the model and the independent variable of the research.

Authors Sample Research period Methodology Dependent variable Independent variable Richardson & Lanis (2007) Publicly listed Australian firms 1997-1999 2001-2003 Regression analysis Effective tax rate Firm size, financial leverage, capital intensity, inventory intensity, R&D intensity, return on assets, industry sector (dummy) Beattie (2006) 831 U.K. listed

companies 2000 Survey regarding target debt level and responsibility of setting target debt level - -

Huang & Song (2006) 1200 Chinese listed company 1994-2003 Regression analysis Leverage Size, profitability, tangibility, tax, non-debt tax shield, volatility, growth opportunities, managerial ownership, ownership structure Buferna, Baugassa, Hodgkinson (2005) Libyan business environment 1995-1999 Regression analysis

Debt level Profitability, growth, tangibility, size Chen (2004) Chinese listed

(16)

Page | 11 Hamberg (2004) Inter-industry comparison of capital structure: forest companies & Swedish clothing companies Annual report 2000 and 2002 - - - Prasad, Green, and Murinde (2003) Thai and Malay companies Average period of 5.5 years Regression analysis

Long and short term debt ratio

Tangibility, growth, profitability, size, non-debt tax shield Huang and Song (2002) Chinese listed companies 1994-2003 Regression analysis Leverage Size, fixed assets, profitability, non-debt tax shield, growth opportunities Sayilgan, Karabacak, Kucukkocaogl u (2002) 123 Turkish manufacturing firms 1993-2002 Regression analysis Total debt/total equity Size, growth opportunities, profitability, tangibility, non-debt tax shield Pandey (2001) Malaysian firms 1984-1999 Regression analysis

(17)

Page | 12 Graham & Harvey (2001) 392 CFOs of U.S. and Canadian companies 1999 Survey - Financial flexibility, credit rating, level of interest rates, target debt ratio, changes in stock price, bankruptcy/dist ress call, insufficient internal funds, foreign regulations Vilasuso & Minkler (2001) Transportation equipments, printing and publishing industries 1987-1997 Non linear least square regression analysis Optimal capital structure Assets specificity, agency cost Mocnik (2001) Slovene manufacturing firm 1991-1996 Questionnaire, regression analysis Short term debt. Long term debt (book value) Specific assets, profitability, volatility, firm size Wiwattanakant ang (1999) Non-financial firms 1996 Regression analysis

Leverage non-debt tax shield, tangibility, profitability, business risk, size Shyam-Sunder, Meyers (1998)

157 U.S. firms 1971-1989 Regression analysis Optimal debt ratio Net debt/assets, gross debt/assets, change in debt ratio, assets, R&D, tax, volatility earnings LLSV et al. (1997)

49 countries - - Legal rules and quality of law enforcement

(18)

Page | 13

Balakrishnan, Fox (1993)

295 U.S. firms 1978-1987 Regression analysis Leverage, log leverage Earning volatility, growth opportunity, R&D, depreciation, industry Harris, Raviv (1991) - - Comparing different papers based on agency cost, asymmetric information, product/input markets, corporate control Capital structure Capital structure agency cost, asymmetric information, product/input markets, corporate control Titman and Wessels (1988) 469 US firms 1974-1982 Regression analysis, correlation matrix Short term debt, long term debt, convertible debt Asset structure, non-debt tax shield, growth, uniqueness, industry classification, size, earnings volatility, profitability Jensen (1986) - - Agency cost

theory - - Bradley, Jarrell, Kim (1984) 25 industries 851 U.S. firms 1962-1981 Analysis of variance with industrial dummy Industry firm leverage Advertising, R&D, non-debt tax shield, firm volatility Castanias (1983) 36 lines of business, 18732 sampled firms 1940-1977 Regression analysis, Pearson coefficient test

Failure rates Long term debt/net worth, net worth/total assets, total liabilities/net worth, return on total assets, cash flow/long term debt, total assets

Meyers (1983) - - Model where an equilibrium is established to optimise firm value depending on over undervaluation and NPV projects

(19)

Page | 14 Bowen, Daley & Hubner (1982) 1800 firms from different industry 1951-1969 Multivariate & univariate analysis of variance Two different measurement of Leverage -

Meyers (1977) - - - Firm value Growth

opportunities, amount of debt Miller (1977) - - Model with

full deductible corporate and personal tax

Leverage Personal and corporate tax rate Jensen, Meckling (1976) - - Indifference curves Agency cost, market value Fraction firms financed with outside claims, market value for manager’s expenditure on non-pecuniary claims Lim (1976) West Malaysian manufacturing firm 1972 Correlation & Regression analysis Capital intensity Employment, value added per employee, capital utilization Taub (1975) 89 U.S. firms 1960-1969 Regression

analysis Issuance bonds, equity dummy Size, tax, solvency, volatility future earnings, difference return on capital and pure interest rate

Scott & Martin (1975) 12 different industries 1967-1972 One-way analysis of variance Leverage ratio - Scott (1972) 12 different industries 77 U.S. firms 1959-1968 One way analysis of variance Debt/Equity Industry Schwartz (1967) None U.S. firms in 4 different industries 1923-1963 F-test of the variation of in common equity

Industry class Equity, long term debt, short term debt, preferred stock Archer, Faerber (1966)

238 U.S. firms 1960-1962 Regression analysis

Compensation to investors, flotation rate, cost of capital

Firm size, size issue, leverage, firm age, variation of past earnings Miller & Modigliani (1958)

Oil and electric utility firms

1947-1953 Regression analysis

Debt/Equity Cost of capital, common stock yield,

(20)

Page | 15

Section II. Methodology and Hypotheses

A. Variable Construction and Hypotheses

The selection of independent variables is primarily guided by the results from previous empirical studies (mentioned above) in the context of capital structure of firms. The focus of this study is to examine the influence the independent variables has on leverage policy of firms operating in capital intensive sector rather than identifying new variables. In a comparative cross country study, Rajan and Zingles (1995) find the following four variables: size, tangibility, profitability and growth important to leverage policy. Many other studies like Prasad, Green and Murinde (2003);Beattie et al. (2006);Huang and Song (2006);Huang and Song (2002); Richardson and Lanis (2007);Booth et al. (2001) also show non-debt tax, shield, capital intensity and institutional variables as important determinants of leverage policy. This study uses book values for measurement of the variables mentioned above as independent variables and discusses below the theoretical and empirical considerations underlying each one of them.

A.1 Firm Size

(21)

Page | 16 when the value of that firm decreases. Consequently, one might expect larger firms to have more leverage due to relatively low bankruptcy cost.

On the other hand, equity might just be too expensive for small firms. In accordance to that, I expect smaller firms to use more leverage as well. Smith (1977) has shown that the size and the relative cost of an equity issue are inversely related. Consequently, smaller firms might have more severe information asymmetry problems between managers and potential lenders, due to lack of adequate and reliable financial information (Titman & Wessels (1988); Hall, Hutchinson & Michealas (2004)). Hence, lenders can reduce the risk of financing to smaller firms by restricting the maturity of the loan. Thus, it is believed that smaller firms will employ less long term debt and possible more short term debt than larger firms (Marsh (1982); Titman & Wessels (1988); Stohs & Mauer (1996); Hall, Hutchinson, & Michealas (2004); Pandey (2001)). Most empirical studies have included size as an explanatory variable in their regression model, while the natural logarithm of sales or total assets is usually used as a proxy for firm size (Rajan & Zingales (1995), Booth et al. (2001)). Due to the high correlation between the two measures, it is advised not to use both. The variable size will be defined in the regression model as the natural logarithm of total assets LN (totalasset). In order to test the variable size in this study, the following hypothesis is taking into account:

H1a: Size will be negatively related to leverage ratio.

A.2 Tangibility of assets

(22)

Page | 17 a positive relationship between tangibility and financial leverage. Pandey (2001) and the trade off hypothesis confirms this positive relationship of tangibility to leverage as the former concludes that firms with higher tangible assets are expected to have higher level of leverage and the latter states that tangible assets act as collateral and provide security to lenders in the event of financial distress.

Nevertheless, creditors might want better terms on a guarantee on loans which creates another incentive to use equity instead of debt (Jensen and Meckling (1976)). However, Jensen (1986) states that this problem can be resolved when equity levels are high (debt low) due to close monitoring by bondholders. This means companies with little debt will want to secure more debt to resolve the conflict of interest between shareholders and managers. As a result, tangibility of assets is represented in the regression model as tangi,t with the statement of the hypothesis test as:

H2a: Tangibility will be positively related to leverage ratio.

A.3 Profitability

This study defines profitability in line with Rajan and Zingales (1995) as earnings before interest, taxes and depreciation (Ebitda) as a function of total assets (Ebitda i, t / Total asset i, t). Ebitda is calculated by taking the pre-tax income and adding back interest expense on debt and depreciation, depletion and amortization then subtracting interest capitalized.

(23)

Page | 18 taxes (DeAngelo & Masulis (1980)). Studies such as Rajan and Zingales (1995), and Wald (1999) confirm the negative relationship between profitability and leverage. Thus, it seems paramount to include profitability as a variable in this study since previous empirical studies have used profitability and all studies found interesting results regarding profitability and leverage.

The reason for choosing Ebitda as an indicator for profitability over net income is simply because of accounting difference in each country. Due to accounting differences, some firms can expense more debt and assets in the form of depreciation and amortization than others. This will lead to different net income numbers but with the use of Ebitda this problem is overcome because I assume Ebitda measures operation income before depreciation and amortization using identical accounting data. In addition, different industries have different depreciation ratios depending on the amount of fixed asset. The use of Ebitda also assists with overcoming this problem. The influence of profitability on leverage is tested with the following hypothesis in this study:

H3a: Profitability will be negatively related to leverage ratio.

A.4 Non-Debt Tax Shield

Following studies like Huang and Song (2002), and Prasad, Green and Murinde (2003), non-debt tax shield is define in this study as accumulated depreciation over total asset (Accumulated depreciation i, t / Total asset i, t). Accumulated depreciation represents the expense related to the fixed assets still carried on the books of the companies. So, accumulated depreciation includes accumulated depreciation, accumulated depletion, accumulated amortization, depreciation on leasehold improvements, amortization of property, plant and equipment under capitalized lease obligations and finally, excess depreciation for non United States corporations. Non-debt tax shield is an important determinant of capital structure for firms (Bowen, Daley & Hubner (1982)). The more non-debt tax shields, the lower the relative value of debt-tax shields. Also DeAngelo and Masulis (1980) argue that firms can use non-interest items such as depreciation, investment related tax shields as substitutes for tax benefits of debt financing. Therefore, an inverse relationship between debt and non-debt tax shield might be expected.

(24)

Page | 19 related tax shield on leverage depends on the trade off between income effect9 and substitution effect, which relates to an increase in optimal investment. If the income effect is larger than the substitution effect, firms with a larger amount of non-debt tax shield would take up a greater amount of debt. Therefore, non-debt tax shield is shown in the regression model as ndt i, t and test the following hypothesis:

H4a: Non-debt tax shield will be negatively related to leverage ratio.

A.5 Asset Growth

Asset growth increases the market value of firms because of higher expected future cash flow. Therefore, firms which grow rapidly are often quite dynamic and capital intensive. So, firms with higher growth rate place high demand on internal funds and therefore retain more earning (Michaelas, Chittenden and Poutziouris (1999); Pandey (2001); Cassar & Holmes (2003)). According to the trade off theory, the retained earnings of firms with relatively high asset growth increase will issue more debt to sustain the target debt ratio. Similarly, according to the pecking order theory, once retained earnings have been exhausted, rapidly growing firms would prefer debt rather than equity. Therefore, a positive relation between asset growth and leverage is expected. This study defines asset growth as the change in total asset (Total asset t / Total asset t-1)-1), similar to Titman and Wessels (1988) and tests the following hypothesis:

H5a: Asset growth will be positively related to leverage ratio.

A.6 Institutional Variables

Economic institutional difference play an important role in capital structure decisions (Booth et al. (2001); Beattie, Goodacre and Thomson (2006); Huang & Song (2006)). The authors conclude that besides firm characteristics, the institutional economic environment plays an important role. Since the data sample in this study includes firms from different countries with different institutional differences, it becomes important to include institutional variable in the model.

Regulatory quality is one of the institutional variables added to the regression model. It is chosen because each country has different regulation within which they can access debt from bank or external source. Also, regulatory quality shows the ability for each respective government in the sample to implement sound policies and regulations that permit and promote private sector development such as banks or financial institution. So, if a country

9

(25)

Page | 20 exhibits high score on regulatory quality, there is enough reason to believe that firms situated in that region will have better access to loans due to sound erected policies and laws. Regulatory quality data in this study is collected from the World Bank data-base and it is measured on a scale from -2.5 to +2.5, with higher values corresponding to higher regulatory quality and represents in the model as regqualityi,t. This study tests the following hypothesis:

H6a: Institutional variables will be positively related to leverage ratio

A.7 Capital Intensity

Lim (1976) argue that the ratio of capital to labour is a good proxy to measure capital intensity. In this study capital intensity is measured as a ratio of capital to labour. Capital represents the funds used to acquire fixed assets. Capital also includes additions of property, plants and equipment as well as investment in machinery and equipments. Capital does not include funds associated with acquisition. Labour represents the number of both full and part time employees of the firms’ in the sample. Labour does not include seasonal employees and emergency employees. Capital intensity is included in this study because research such as Prasad, Burton and Merikas (1997); Richardson and Lanis (2007); Vilasuso and Minkler (2001) and Mocnik (2001) have shown that capital intensity is a significant determinant of capital structure for firm. They argued that the fundamental characteristic of capital intensive sector is the investment in firm-specific assets. The investment in firm-specific assets according to Mocnik (2001); Williamson (1988); Balakrishnan and Fox (1993); Mang (1998) and Vilasuso and Minkler (2001) are more likely to be financed with equity rather than debt. They stated that investment projects that require highly specific assets should be financed with equity, while debt financing should be used for assets that are more generic and can be easily redeployed. Since this study includes industries from the capital intensive sector, it is expected that capital intensity will be negatively related to leverage ratio. Thus, capital intensity in shown in the model as Cinti, t with the hypothesis statement as:

H7a: Capital intensity will be negatively related to leverage ratio.

(26)

Page | 21 Machinery Industry and Chemical and Applied Product Industry), the following hypothesis is tested:

H8a: There is a significant difference in leverage ratio among industries operating in the

capital intensive sector.

A.8 Dependent Variables

Myers (1984) has argued that book values can be used as proxies for the asset value of the firm. So, all company specific variables in this study are based on book values. This study defines leverage ratio as total debt as a function of total asset, (Total debt i, t / Total asset i, t,), in the modellev i, t. Total debt represents all interest bearing and capitalized lease obligations. It is the summation of long and short term debt. Using this calculation for leverage ratio is convincing that it is a good proxy for changes in leverage ratio as Haung and Song (2006) use the same method as a dependent variable and conclude that it is a good alternative for measuring leverage changes within firms.

To increase the quality of this study’s result and check the robustness of the model, an additional dependent variable is introduced into the models. The dependent variable is defined as total debt over total shareholders’ equity (Total debti, t / Total shareholders equityi, t). It shows in the regression model as lev2i, t. The variable is included in this study because Sayilgan, Karabacak and Kucukkocaoglu (2002) used the same variable as a dependent variable in their study and conclude it is a good proxy for measuring leverage ratio.

Furthermore, to analyze the interaction between capital intensity and leverage ratio because the industries used in this study are from the capital intensive sector, capital intensity is used as a dependent variable. Although the analysis of the interaction between capital intensity and leverage ratio is beyond the scope of this paper, the results are however presented but future research will need to elaborate more on the interactions.

B. Estimation Method

(27)

Page | 22 while cross-sectional analysis measures the changes between individuals or firms in this case. In order to blend the two characteristics of both time-series and cross-sectional analysis, a panel regression analysis becomes suitable since it measures both across time and between individuals or firms (Hsiao (1985)). This panel regression method is chosen because according to Hsiao (1985), panel data allow us to study in depth complex economic and related issues which could not be treated with equal severity using time-series or cross-sectional data alone. While Brooks (2008) also mentioned some advantages of using panel data:

 The ability to address a broader range of issues while tackling more complex problem as opposed to using pure time-series or cross-sectional data.

 The advantage of increasing the degrees of freedom and thus the power of the test by combining both time-series and cross-sectional data.

 The capability of removing the impact of certain forms of omitted variable bias in the regression results.

Subsequently, a regression analysis that is restricted to just time-series or cross-sectional data is more likely to reflect inter-individual rather than intra-individual differences. Furthermore, time-series or cross-sectional regression analysis disregards inter-individual differences because it does not take into account cross-sectional variances, while normal time-series data does not only take into account cross-sectional variances but also ignores their dynamics over time.

The combination with the richness of information provided over time and across space reduces the gap between information requirements of a sample and information provided by the data-set (Hsiao (1985)).Thus, time-series or cross-sectional data are not very suitable in this study because the hypotheses are based on demographic and institutional difference that varies over time and space.

To see the way panel data works, I consider a common panel data regression model which looks as follows:

(28)

Page | 23 similar to a dummy variable model in one dimension. While in a random effect model, the error term is assumed to vary stochastically over i and t requiring special treatment of the

error variance matrix (Hsiao et al. (1999)). So, if the covariance between the dependent variable is zero, the error term is evenly distributed over i and t with zero mean and constant

variance. Then the regression analysis will have an unbiased and consistent outcome for all its coefficients (homoskedasticity10). However, when there is heteroskadasticity11, the standard error of the different variable is underestimated. This means that the result could underestimate the confidence internal of the tested variable and automatically lead to committing a type II error. Under normal time-series or cross-sectional regression analysis, the problem of heteroskadasticity can be resolved using the White’s test (Field (2009)). However, this is not the case under panel regression analysis because the residuals can suffer from heteroskedasticity over time and across section. Therefore, other ways of calculating standard errors must be used when residuals are permanent or temporary12 (Brooks (2008)). In this study, the homogeneity of variance assumption is tested with a statistical program and presented in the Appendix B. In the residual statistics table, the mean of the error term is zero and scatter plot table below the residual statistics table confirms the assumption of homoskedasticity.

In order to test the above hypothesis, a pooled regression will be performed using the following equation:

Levi,t = αi + β1(profiti,t) + β2(sizei,t) + β3(tangi,t) + β4(growthi,t) + β5(ndti,t) + β6(regqualityi,t) + β7(Cinti t) + εi,t (2) Where α, β and ε are the intercept term, parameters that explains the independent variable, and error term respectively. The sub-script i and t represents the firm and time period accordingly. Additionally, capital intensity is excluded from model 2 to determine if there will be a change to the result. Thus the equation is as follows:

Levi,t = αi + β1(profiti,t) + β2(sizei,t) + β3(tangi,t) + β4(growthi,t) + β5(ndti,t) + β6(regqualityi,t) + εi,t (3) According to Brooks (2008), using pooled regression as described above has certain limitations. Pool regression assumes that the average values of the variables and the relationships between them are constant over time and across all of the cross-sectional units

10

The variance term in the model is constant.

11

There are variations in the variance term in the model.

12

(29)

Page | 24 in the sample. It is therefore possible to estimate separate time-series regression for each firm, but this is likely to be sub-optimal as this approach will not take into account any common structure present in the time-series of interest (Brooks (2008)). Alternatively, a cross-sectional analysis can also be done for each time period, but this might not be intelligent if there is some common variation in the series over time. To guide against the above mentioned problems and test the models in this study, a time fixed effect regression model is used. The time fixed effect regression model allows the intercept in the regression model to differ sectionally and over time, while all the slope estimates are fixed both cross-sectionally and over time.

The way the time fixed effect regression analysis works is to decompose the error term in equation (1) into an individual specific effect, µi, and the remainder error term as ϋi,t. Therefore, the individual specific effect encapsulates all the variables that affect the dependent variable cross-sectionally but does not vary over time. It is difficult to identify which effect is present in the dataset due to the small time frame of 4 years. Therefore, the assumption is made that the effect is temporary, which is in line with previous research in this field. Among others, Woodridge (2007) assumes temporary effects because correlation of standard error dies out over time. He went further to conclude that the effect is temporary because residuals in succeeding years can be severely high; however, their correlation is not expected to remain persistent. Woodridge (2007) suggests that for example 2005 to 2006 residual correlation is much lower than 2006 to 2007 residual correlation.

Furthermore, to compare different industries that operate in the capital intensive sector if their capital structures are similar or different, the sample is divided into 3 industrial categories which are: Energy, Industrial Machinery, as well as Chemical and Applied Product industries. Thus, estimating the model using a dummy variable as state below:

(30)

Page | 25 To compare if there is an actual difference in leverage ratio among industries operating in the capital intensive sector, a test statistics (t-test) will be employed. Two approaches in conducting a test statistics are the likelihood test (x-square) and the Wald-test (F-statistics). In this study, the Wald-test gains higher preference over the Likelihood test because the Wald-test is usually considered preferable for finite samples (just like this study) and the Wald-test is also sensitive to sample size unlike the likelihood test (Brooks (2008)). Therefore, a significant p-value under the Wald-test gives enough reason to reject hypothesis 8a. Finally, in the table II below are the definition of variables in the model and the predicted

(31)

Page | 26

Table II

Definition of Variables and Predicted Signs of Explanatory Variables

Table II show in the first column the full names of all the variables in the model and in parentheses the way the variable names appear in the model. The second column shows the definition of each variable in the model, while the third column states the predicted sign of the variable before the regression analysis is performed. “+” means that leverage increases with the factor. “-” means that leverage decreases with the factor. “?” means that no clear prediction or empirical study results from the literature.

Variable Definition Predicted sign

Leverage (Lev) Total debt over total assets Dependent variable

Leverage (Lev2) Total debt over total

shareholders’ equity Dependent variable

Profitability (profit) Ebitda over total assets -

Size (lnsize) Log total assets -

Tangibility (tang) Property, plant & equipments over total assets +

Growth (growth) Change in total assets +

Non-debt tax shield (ndt) Accumulated depreciation over

total assets -

Regulatory quality (regquality) Measure on a scale from (-2.5)

to (+2.5) +

Capital Intensity (Cint) Capital over labour -

Chemical and Applied Product firms (C.firms)

Dummy, (1) for energy firms and (0) otherwise

Difference in capital structure from the other industries Industrial Machinery firms

(M.firms)

Dummy, (1) for machinery firms and (0) otherwise

Difference in capital structure from the other industries

(32)

Page | 27

Section III. Data

A. Sample Design and Availability

To investigate whether company specific variables have an impact on capital structure choice for firms operating in capital intensive sector, this study uses balanced panel data from world-wide sample. The capital intensive industry covered in the world sample are; Energy Industry (comprise of gas distribution firms and oil and gas extraction firms), Heavy Industrial Machinery and Equipments Industry, as well as Chemical and Applied Product Industry. The world data sample is collected from World Scope, Thomson-One Financials, and the World Bank data-base. All company specific data is collected from World Scope except earnings before interest, taxes, depreciation, and amortization (Ebitda), which is obtained from Thomson-One Financial.

(33)

Page | 28 outliers which are legitimate13. The first type should always be removed or corrected in the data set, but with the second type, simple removal is not always the best option as removal of legitimate data points will lead to loss of valuable data.

The data set has two outliers which were legitimate outliers. These outliers could have arisen due to the fact that in our sample, it contains different firms from different countries. These outliers in the data set are corrected following the “mean plus 2 standard deviations method” (Field (2009)). This method is simply multiplying the standard deviation by 2 and adding it to the mean of the sample. However, since the outlier in the data set is only two and I consider it to be small regarding the size of the data-set, there is no reason to analyze a data set without the outliers and enough reason to believe the substantial effect on the coefficient in the regression will be negligible.

Table III below shows the geographical spread of the firms in the sample. Also, it includes the percentage term each region covers in the entire sample. Europe emerges as the region with the highest firms in the sample with 47.1%, while Africa and Oceania region has the least with 2.9% each.

13

(34)

Page | 29

Table III

Geographical Sample Spread

Table III show in the first column the different continents firms in the sample belong to. The number of countries in each particular continent is presented in the second column. The third to fifth column show the industrial categories of firms in the sample. Column six and seven depicts the total number of firms and the percentage of firms in the sample respectively.

Continent No of countries Firms Energy Industry Firms Machinery Industry Firms Chemical Industry No. Of firms % of firms in sample Europe 16 23 25 19 67 23.3% Asia 10 29 51 44 124 43.1% N. America 2 50 12 16 78 27.1% S. America 4 2 4 5 11 3.8% Oceania 1 1 1 - 2 0.7% Africa 1 - 1 5 6 2.1% Total 34 105 94 89 288 100%

In order to understand and describe the main features of the data-set in quantitative terms, a descriptive statistics and normality test is analyzed. In table IV below, the descriptive statistics of the variables are presented. The table includes the mean, median, standard deviation, maximum and minimum values. This way, it is easily seen how the variables are represented in the data set. The variables in this study show no extraordinary statistics. For example, the standard deviation in the second column does not deviate largely from the mean therefore, it can be concluded that outliers do not significantly affect the data set. Also, the negative value in the fourth column (Minimum) is acceptable since it is possible for firms to have negative growth, profit, and quality of regulation in their respective countries. However, Leverage ratio 2 in the fourth row has a negative sign in the fourth column because of the negative sign of the denominator (total shareholders’ equity) when constructing this variable. The negative sign on shareholders equity occurs due to the accounting method of dealing with accumulated losses from previous years. These losses are viewed as liability carried forward until cancelation (Thomson-One financial).

(35)

Page | 30 have been possible to get significant results which would deviate from normality (Field (2009)). Similarly, Brooks (2008) also mentions that for sample size that is relatively large, violation of the normality assumption is virtually inconsequential because of the central limit theorem14. Thus, there is enough reason to believe that even a deviation from normality in this study is not enough to bias the results of the regression analysis. Also, a (P-P plot) normality graphical representation of each variable is shown in appendix B, which shows the normal distribution of most variables in this study.

14

(36)

Page | 31

Table IV

Descriptive Statistics Sample Variables

Table IV (panel A) show the descriptive statistics of the whole firms in the sample. The mean, standard deviation, median, minimum and maximum is present in column two to size respectively. Mean shows the average value in the data-set, the minimum and maximum shows the lowest and highest value in the data-set. 1152 observation are generated coming from 4 years annual data for 288 firms. Table IV only refers to firms that meet the criteria and the sample size (n) =1152

Panel B show the descriptive statistics of the firms in the sample that operates in the energy industry. The mean, standard deviation, median, minimum and maximum is presented in column two to six respectively. Mean shows the average value in the data-set, the minimum and maximum shows the lowest and highest value in the data-set. 420 observations are generated coming from 4 years annual data for 105 firms. Panel B only refers to firms that meet the criteria and the sample size (n) =420.

Panel C show the descriptive statistics of the firms in the sample that operates in the Industrial Machinery industry. The mean, standard deviation, median, minimum and maximum is presented in column two to six respectively. Mean shows the average value in the data-set, the minimum and maximum shows the lowest and highest value in the data-set. 376 observations are generated coming from 4 years annual data for 94 firms. Panel C only refers to firms that meet the criteria and the sample size (n) =376

Panel D show the descriptive statistics of the firms in the sample that operates in the Chemical and Applied product industry. The mean, standard deviation, median, minimum and maximum is presented in column two to six respectively. Mean shows the average value in the data-set, the minimum and maximum shows the lowest and highest value in the data-set. 356 observations are generated coming from 4 years annual data for 89 firms. Panel D only refers to firms that meet the criteria and the sample size (n) =356

Panel A

Variable Mean St. deviation Median Minimum Maximum Reg. quality 1.002 0.745 1.230 - 0.770 2.000 Lev. Ratio 0.258 0.164 0.257 0.000 1.520 Profit 0.129 0.100 0.117 - 0.360 0.880 Tang 0.563 0.205 0.568 0.090 0.960 Ndt 0.368 0.328 0.275 0.010 2.100 Growth 0.158 0.290 0.091 - 0.790 1.950 Lnsize 3.230 0.781 3.210 1.350 5.370 Lev. Ratio 2 0.710 0.554 0.605 -0.550 2.950 Cint 0.186 0.455 0.220 -0.001 2.97 Panel B

(37)

Page | 32

Panel C

Variable Mean St. deviation Median Minimum Maximum Reg. quality 0.907 0.759 1.120 -0.480 2.000 Lev. Ratio 0.223 0.152 0.208 0.000 0.780 Profit 0.106 0.988 0.104 -0.356 0.480 Tang 0.429 0.165 0.408 0.087 0.838 Ndt 0.271 0.241 0.187 0.015 1.940 Growth 0.134 0.278 0.086 -0.791 1.832 Lnsize 2.898 0.683 2.716 1.489 4.802 Lev. Ratio 2 0.630 0.548 0.509 -0.441 2.923 Cint 0.069 0.219 0.008 -0.01 2.005 Panel D

Variable Mean St. deviation Median Minimum Maximum Reg. quality 0.748 0.785 1.100 -0.480 1.930 Lev. Ratio 0.233 0.163 0.228 0.001 1.523 Profit 0.135 0.115 0.123 -0.289 0.882 Tang 0.522 0.163 0.539 0.090 0.840 Ndt 0.362 0.256 0.283 0.330 1.237 Growth 0.159 0.299 0.086 -0.729 1.757 Lnsize 3.129 0.769 2.966 1.830 4.869 Lev. Ratio 2 0.577 0.484 0.483 -0.545 2.565 Cint 0.107 0.341 0.107 0.000 2.666

In this study, where the goal is to understand how the different independent variables influence the dependent variable, multi-collinearity is a problem. Multi-collinearity affects the outcome; it can lead to larger standard errors for the variables involved. This effect has the tendency of leading to a type II error, a rejection of H1, while the coefficients are

significant. The correlation matrix for the variables of this study is included in table V below.

(38)

Page | 33

Table V Correlation Matrix

Table V (panel A) show the correlation matrix of variables for all firms in the sample. The sample consists of 1152 observations from a 4 year time period. All variables in the sample show moderate correlation effect. With a negative sign in front of the number meaning the variables are negatively correlated and a positive sign in front of the number means the variables are positively correlated. * Means the values is significant at 0.05 significance level and ** means the values are significant at 0.01 significance level.

Panel B show the correlation matrix of variables for Energy firms in the sample. The sample consists of 420 observations from a 4 year time period. All variables in the sample show moderate correlation effect. With a negative sign in front of the number meaning the variables are negatively correlated and a positive sign in front of the number means the variables are positively correlated. * Means the values is significant at 0.05 significance level and ** means the values are significant at 0.01 significance level.

Panel C show the correlation matrix of variables for Industrial Machinery firms in the sample. The sample consists of 376 observations from a 4 year time period. All variables in the sample show moderate correlation effect. With a negative sign in front of the number meaning the variables are negatively correlated and a positive sign in front of the number means the variables are positively correlated. * Means the values is significant at 0.05 significance level and ** means the values are significant at 0.01 significance level.

Panel D show the correlation matrix of variables for Chemical and Applied product firms in the sample. The sample consists of 356 observations from a 4 year time period. All variables in the sample show moderate correlation effect. With a negative sign in front of the number meaning the variables are negatively correlated and a positive sign in front of the number means the variables are positively correlated. * Means the values is significant at 0.05 significance level and ** means the values are significant at 0.01 significance level.

Panel A

Variable Cint Reg.

(39)

Page | 34

Panel B

Variable CINT Reg.

quality Lev. Profit Tang Ndt Growth Lnsize Lev.2 CINT 1.00 Reg. quality 0.146 ** 1.00 Lev. -0.088 0.001 1.00 Profit 0.239** 0.40 -0.313** 1.00 Tang 0.266** 0.183** 0.344** 0.094 1.00 Ndt 0.022 -0.130** 0.119* 0.083 0.333** 1.00 Growth -0.082 -0.015 -0.004 0.67 -0.132** -0.195** 1.00 Lnsize 0.556** 0.220** -0.057 0.253** 0.343** 0.095 -0.143** 1.00 Lev.2 -0.054 0.060 0.820** -0.401** 0.232** -0.031 -0.026 0.11 1.00 Panel C

Variable CINT Reg.

(40)

Page | 35

Panel D

Variable CINT Reg.

Influence of Determinants on Capital Structure: An Empirical Investigation of the Energy, Industrial Machinery and Chemical & Applied Product Industry, utilizing Panel Data from 2005-2008.