Impact of dynamic environment on the capital structure of Dutch corporations

Impact of dynamic environment on the capital structure of Dutch

corporations

Sophie A. van Kleef University of Groningen Faculty of Economics and Business

S.A.van.Kleef@student.rug.nl September 30, 2009

ABSTRACT

This paper uses a panel model regression to empirically test whether a dynamic environment has an impact on the capital structure choice of Dutch corporations. The periods researched are 1985 to 1989 (covering Black Monday) and 1997 to 2002 (covering the Dot-com Bubble), respectively consisting of 39 and 83 companies. The firm-specific and macro-economic variables show results that are not in-line with previous empirical work. Also, the interpretations of results depend crucially on what measure of leverage is used in the model, book value versus market value. However, the Wald test does indicate that some coefficients during stable periods significantly differed from those during dynamic periods. When investigating whether the level of leverage has a different impact on performance during a stable or dynamic environment, the Black Monday sample showed some support for my hypothesis that during stable periods a higher leverage ratio leads to increased profit efficiency, while during dynamic periods a higher leverage ratio has a less positive or negative impact on the profit efficiency. However, the Dot-com Bubble sample results were less significant and suggested the exact opposite. Based on my research it cannot be supported that a dynamic environment, when compared to a stable environment, has a different impact on the Dutch capital structure choice; neither can it be concluded what impact the dynamic environment has on the capital structure’s effect on performance.

Keywords: Capital structure, financial crisis, dynamic environment, asymmetric information, agency theory, profit efficiency

JEL-codes: E44, G32, G33

Section 1: Introduction

determinants of capital structure. While several studies suggest that factors like agency costs, growth, profitability and firm size determine a firm’s financial leverage (e.g. Jensen and Meckling, 1976; Myers, 1977), some empirical results often contradict this (e.g. Chen et al. 1999). Many firms have experienced severe financial distress and even bankruptcy because of wrong decisions on debt policy (Vasiliou and Daskalakis, 2006). So the financing decision between equity and debt is of great importance.

Presently a huge number of countries is experiencing financial crisis. Crisis can be considered an example of a dynamic environment, where the dynamic is defined as a period with unpredictable, unsystematic and uncertain changes in which many companies get in financial distress, go bankrupt or grow disproportionately fast (Das, Das and Lim, 2009). Because of uncertainty, people may be reluctant to take risks. In this case, companies have difficulty raising equity and they have to consider debt financing. However, additional debt increases the risk of default. A trade-off exists between equity and debt financing that might be affected by the state of the environment. An interesting question is therefore whether the uncertainty among consumers, investors and firms caused by a dynamic environment has an impact on the corporate capital structure. Few empirical studies have looked into this matter. Former research that did look into the effect of the environment on capital structure mainly focused on the Asian crises of 1997 (Das, Das and Lim (2009), while many more dynamic periods in other regions have occurred that did not receive much attention in this field of research.

Knowing how a company reacts to a dynamic environment1 regarding its capital structure choice and how this subsequently affects profit efficiency2, can be of particular interest to investors. For example, if it is found in this study that companies are likely to issue more debt during dynamic times and if their profit efficiency decreases due to the increased use of leverage, the risk of default of a company increases and investors might demand more return on their investment. My research might therefore provide additional insights to investors when and how to invest.

1

In this research environment is thought of as all physical and social factors that are taken directly into consideration in the decision-making behaviour of individuals in the organization (Duncan, 1972). 2

Booth, Aivazian, Demirguc-Kunt and Maksimovic (2001) tested several determinants of capital structure on ten developing countries and found persistent dissimilarities across countries, indicating that specific country factors, like institutional differences, impact on debt-equity ratios too. Knowing that institutional differences between countries lead to dissimilarities of corporate capital structures, it is interesting to research a country like the Netherlands that differs a lot from the much researched US and Asia. A better understanding of the capital structure determinants in a relatively small yet open industrialized economy like the Netherlands is fundamental not only for enriching empirical studies in this field, but also for the purpose of cross country evaluation (Chen et al., 1998). Subsection 2.4 will explain how the particular characteristics of the Dutch system might affect the capital structure of firms. Next to the fact that researching capital structure for the Netherlands will provide understanding of the Dutch case, it will show whether theories mostly tested on other countries are also supported by data from the Netherlands.

Finally, I myself am Dutch. The Netherlands is therefore of particular interest to me and I have relatively easy access to Dutch data.

The first objective of this study is to research to what extent capital structure of Dutch corporations is different during a dynamic environment in comparison to stable periods3. The second objective is to investigate whether the level of leverage, measured as a ratio of debt to equity, during stable and dynamic environments has a differing effect on the profit efficiency. This second objective is researched because researchers like Berger and Di Patti (2002) proposed that a more profit efficient firm chooses a lower equity ratio than other firms, since higher profit efficiency reduces the expected costs of bankruptcy and financial distress. Pratomo and Ismail (2007) find evidence for Islamic firms consistent with the hypothesis that higher leverage or a lower equity capital ratio is associated with higher profit efficiency. However, the above-mentioned research has not taken into account effects of the environment. In a stable environment, it might be better for a firm to use debt rather than equity financing when taking into account the

3

A dynamic period is characterized by unpredictable, unsystematic and uncertain changes in which many companies get in financial distress, go bankrupt, or grow disproportionately fast (Das, Das and Lim, 2009). A stable environment means that a firm is able to develop a fixed set of routines for dealing with

availability of lower cost debt financing. However, when the environment changes to dynamic, equity financing should be preferred to reduce the transaction costs that arise from increased risk (Li and Simerly, 2002).

The two dynamic periods my research will focus on are Black Monday and the Dot-com Bubble. Macro-economic data, like money supply provided by the DNB, is only accessible from 1983 onwards. Due to this historic data limitation, I have to focus on more recent periods for my research. Section 4 will illustrate why Black Monday and the Dot-com Bubble can be considered dynamic environments.

By applying known empirical work and theories about capital structure, I will investigate the two above mentioned objectives. Since no former research has been done on the impact that a dynamic versus a stable environment can have on the capital structure of Dutch firms, this study will contribute to the existing literature.

In the next sections former research and theories on capital structure will first be discussed. Section 3 and 4 will provide the methodology and data that is used. Then I will give an overview of the results and finally a conclusion is presented.

Section 2: Literature review

To better understand the purpose of this paper and the empirical results my models will show, I will first give an overview of the most influential theories to date and the empirical research that has been done regarding capital structure. Then I will present my hypotheses for this research. At the end, a brief discussion is given of the institutional settings in the Netherlands that are relevant to the capital structure choice and a summary is given of previous research regarding capital structure for Dutch companies.

and Miller are not quite realistic under dynamic conditions when business risks, unpredictability, and uncertainty are high.

The traditional view on capital structure is that it can influence the cost of capital and thereby affect the value of the company (Ariff, Shamsher and Taufiq, 2008). The cost of capital is the expected rate of return required to attract funds to a particular investment; it thus reflects the attitudes of investors toward risk (Grabowski and Pratt, 2008). If the investors become more risk averse, companies will have difficulty raising equity and may need to cancel or defer some investments. A moderate use of debt will reduce the overall cost of capital initially and so increase company value. However, when leverage becomes too high, the cost of capital will start to increase. This is because additional debt increases the risk of default and so increases the interest rate that has to be paid by the company to borrow money. Company value will consequently decrease (Ariff, Shamsher and Taufiq, 2008). How to measure a moderate level of debt is not precisely identified. Common measures are the moving average of historical levels or an industry ratio (Ariff and Lau, 1996).

However, while the traditional approach as described above is appealingly simple, its predictions cannot always be substantiated in the data. The behavioural finance framework has emerged, responding to the problems faced by the traditional model (Barberis and Thaler, 2003). The behavioural finance theory contradicts Modigliani and Miller’s efficient market theory on certain aspects. It tries to explain the motivation behind managers’ and investors’ behaviour and why they do not always act rationally. Psychological and social aspects may influence the managerial decision making, thereby affecting the way capital is structured. (Baker et al, 2004; Graham and Harvey, 2001; Bancel and Mittoo, 2004; and Vasiliou and Daskalakis, 2006). Although behavioural finance has led to some interesting findings regarding capital structure, my model will not focus on the psychological and social aspects that can influence managerial decision making.

2.1: Capital structure in relation to environmental dynamism

studied the determinants of capital structure, with most work finding a significant relationship between diversification strategy and debt financing (e.g. Lowe, Naughton and Taylor, 1994). However, most research on the determinants of capital structure did not take into account what the effect of environmental uncertainty can be. Increasing environmental uncertainty can occur because of lack of information, an inability to predict outcomes or an inability to predict how environmental conditions will affect outcomes (Duncan, 1972). According to some researches (e.g. Milliken, 1987), increasing uncertainty is associated with greater environmental dynamism. Environmental dynamism is a period characterized with unpredictable, unsystematic and uncertain changes in which many companies get into financial distress, go bankrupt, or grow disproportionately fast (Das, Das and Lim, 2009). During a dynamic environment, companies find it harder to assess both the current and future state of the environment accurately and managers cannot easily determine the potential impact of decision-making on present and future business activities. Whether the environment is dynamic or not thus impacts on financing decisions by managers, which is why the capital structure may differ during stable and dynamic environments.

A summary of the researches that did take into account environmental dynamism is shown in Table 1.

2.2: Capital structure in relation to profit efficiency

Table 1: Several studies that did take into account environmental dynamism. Author(s)

(year of publication)

Model Method Result

Li and Simerly (2002)

Linking the interaction between environmental dynamism and capital structure with firm

innovation. Using US firms for the period 1988-1994.

A multiple regression methodology is used with an interactive term between leverage and environmental dynamism. To produce a standardized index of ED, they regressed the industry value of shipments over 5 years against time, and used the standard error of the regression coefficient related to a time dummy variable divided by the average value of industry shipments.

The level of environmental dynamism interacts with capital structure.

Deesomsak, Paudyal and Pescetto

(2004)

Investigating the determinants of capital structure of firms operating in the Asia Pacific region; and researching whether there were significant changes in the coefficients of the explanatory variables due to the financial crisis of 1997.

Firm’s leverage ratios are modeled as a function of firm-specific factors in a cross-sectional framework. Wald statistics are estimated to test for significant changes in the coefficients due to environmental dynamism.

The dynamic environment during the financial crisis of 1997 is found to have had a significant but diverse impact on a firm’s capital structure decision across the region.

Das, Das and Lim (2009)

Researching the influence of related and unrelated product diversification on a firm’s level of debt financing arguing that this link is moderated by the

environment. Using Asian firms for the period 1995-2000.

Using SAS PROC MIXED they fit a mixed-effects model to the dataset. To measure environment they collected time-series data on the weekly levels of the Singapore Stock Exchange Straits Times Index. The total period of the study was split into a pre-crisis period to reflect a stable environment and a crisis and recovery period to reflect a dynamic environment. Separate models were estimated for the two sub-periods.

The asymmetric information theory implies that managers know more about the prospects, risks and values of the company than the outside investors. This can influence choices between external and internal financing and equity or debt. For example, Myers (1984) suggests using a low risk debt to finance a new investment instead of issuing new equity. When a new investment has to be financed, investors can be sceptical with respect to its return due to asymmetric information. In that case the investors might under value a possible equity issue. In addition, Myers and Majluf (1984) argue that issuing new equity can be perceived as a negative signal by the market leading to a drop in the value of the company when announcing a new stock issue. If risk-free debt is issued, on the other hand, company value should not fall. Issuing risk-free debt is equivalent to having plenty financial slack, and plenty slack insures that the firm will take on all positive NPV projects (Dann and Mikkelson, 1982). So a decision to issue risk-free debt and invest, conveys the information that the company has a positive NPV project. This may lead toa positive value change unless the project’s existence was known before the announcement of the issue. A risk-free debt issue is not very likely, but if the probability of default on the issue was small, the negative information effect should also be small.

The above-mentioned studies by Jensen and Meckling (1976) and Myers and Majluf (1984) both suggest that a higher leverage ratio leads to superior financial performance, like higher stock market returns. Berger and di Patti (2002) support the arguments of Jensen and Meckling and Myers and Majluf, and propose the efficiency-risk hypothesis under which the more efficient companies choose lower equity ratios than other companies, since higher efficiency reduces the expected costs of bankruptcy and financial distress. The hypothesis suggests both that profit efficiency is positively correlated with expected returns and that the higher expected returns from high efficiency can be used as substitutes for equity capital in protecting the company against future distress. The results of Berger and di Patti (2002) show that this efficiency-risk hypothesis dominates the hypothesis that a higher equity ratio reduces the likelihood of bankruptcy and financial distress.

Table 2: Studies that researched the effects of leverage on profit efficiency. Author(s)

(year of publication)

Model Method Result

Berger and di Patti (2002)

Testing whether leverage affects agency costs and thereby influences firm performance. Data

consists of the US banking industry for 1990-1995.

Profit efficiency4 is used as the dependent variable in a simultaneous equations model accounting for reverse causality from performance to capital structure.

Findings suggest that higher leverage or a lower equity capital ratio is associated with higher profit efficiency. The effect is economically, as well as statistically, significant.

Pratomo and Ismail (2007)

Attempting to prove the agency cost theory for Islamic Banks in

Malaysia. Period tested: 1997-2004.

The method is comparable to that of Berger and di Patti (2002).

Findings are consistent with the agency

hypothesis. The higher leverage is associated with higher profit efficiency.

2.3: Hypotheses

Former research (e.g. Li and Simerly, 2002) suggests that a stable versus a dynamic environment can have a different impact on the capital structure. Therefore I hypothesize:

Hypothesis 1: During a dynamic environment, the capital structure choice (measured by a leverage ratio) of Dutch corporations is significantly different from that during a stable period.

4

In a dynamic environment, the company experiences unpredictable, unsystematic and uncertain changes. Hence, unlike in the stable case, a manager must adapt to new situations and overcome possible unpredictable problems. Li and Simerly (2002) suggest that when companies operate in a stable environment, it is better for the companies to use debt rather than equity financing when taking into account the availability of lower cost debt financing. However, when the environment changes to dynamic, equity financing should be preferred, to reduce the transaction costs that arise from increased risk. This would mean that the agency and asymmetric information theory hold during stable environments, but the theories are not applicable during a dynamic environment. This leads to my next hypothesis:

Hypothesis 2: During a stable environment, increased leverage has a positive impact on profit efficiency, while during a dynamic environment equity financing, (decreased leverage), has a positive impact on profit efficiency.

2.4: Research based on capital structure for Dutch companies.

A better understanding of the capital structure determinants in a relatively small yet open industrialized economy like the Netherlands is fundamental not only for enriching empirical studies in this field, but also for the purpose of cross country evaluation (Chen et al., 1999). Former analysis on the capital structure of Dutch companies by Chen et al. (1999) showed that Dutch companies have a relatively strong preference for internal funds over external funds. In addition, when companies need external finance, they prefer to use debt finance rather than equity finance. These two findings are very comparable to the situation in many other industrialized countries. However, the Dutch system has some special characteristics which makes it interesting to research capital structure in the Netherlands further. Relative to other industrialized countries, for example, the Dutch shareholding of companies is more widely spread. Therefore, relevance of the agency theory for Dutch companies has to be assessed by considering the role of Dutch equity holders.

2.4.1. Tax considerations

Modigliani and Miller (1958) showed that when interest payments on debt are tax deductible, corporate taxation implies that the irrelevance theory of capital structure no longer holds. The corporate tax rate in the Netherlands, compared to other countries, is relatively low. Furthermore it is characteristic for the Dutch system that the interest is tax deductible for corporate taxes, dividend is taxed as income and capital gains are tax exempt. However, investors can decide to take on a dividend re-investment option, under which dividend becomes tax free. Still, in general, the Dutch tax system prefers debt over equity financing due to which companies prefer debt over share issues (Chen et al., 1999).

2.4.2. Agency costs

The theory on agency costs initiated by Jensen and Meckling (1976) considers debt as an important method to mitigate the conflicts between equity holders and managers. To assess whether the agency theory is relevant for explaining the capital structure choice of Dutch companies, the role of Dutch equity holders has to be considered. The Dutch shareholding of companies is more widely spread than in other industrialized countries. In other words, there are many small shareholders, who do not have a lot of incentives to monitor the corporations. Additionally, a system of cooptation in which new members of the supervisory board are elected by the current members of the board, is typical for the Dutch corporate governance (Chen et al., 1999). Generally this system reduces the corporate governance role of stockholders. Moreover, it is one of the reasons why pension funds do not have an important role in corporate governance, although these funds hold a substantial percentage of total shares (about 8%) (Chen et al., 1999).

Because of this widespread shareholding and the system of cooptation, shareholders do not seem to play an important role in corporate governance, which might indicate that agency problems between shareholders on the one hand and debt holders and managers on the other barely exist.

2.4.3. Asymmetric information

issuance of debt for example is perceived as a signal of management’s confidence in the company. A company will initially try to finance a new investment with internal funds, then with low risk debt, and finally with equity. This is the so-called pecking order theory (Myers, 1984). Findings by De Haan et al. (1994), surveying Dutch companies, suggest that asymmetric information is relevant for explaining capital structure of companies in the Netherlands.

A summary of the studies that researched capital structure for Dutch companies is given in Table 3. These studies give evidence of relevant determinants for Dutch capital structure. De Haan (1992) was the first to research the effect of developments external to companies, like inflation and interest, on the capital structure of Dutch companies. However, he did not include firm-specific factors in his research. These factors can, according to theory, be determinants of capital structure and may have different weights when an environment changes from stable to dynamic. In addition, he did not research whether other periods also supported his findings.

Table 3: Studies that researched capital structure for Dutch companies. Author(s)

(year of publication)

Model Method Result

Chen, Lensink and Sterken

(1999)

Investigating whether and to what extent the main capital structure theories can explain capital structure choice of Dutch firms. (1984-1995)

A panel data model is estimated that explains both the absolute level of leverage with respect to various factors and the year-to-year changes in leverage with respect to the changes of various factors.

Evidence suggests the

relevance of the pecking order hypothesis in explaining the financing choice of Dutch firms. Factors based on agency costs and corporate control are argued to be relatively unimportant for the Dutch case.

De Haan, Koedijk and Vrijer

(1994)

Getting more insight into the microeconomic motives behind the macroeconomic liquidity behaviour of the Dutch business sector in the 1980s.

Interview survey: sending questionnaires to non-financial Dutch companies.

About 75% of the responding firms stated they have a certain financial hierarchy. 54% of the respondents favoured internal financing, 18% debt and 3% preferred share issues. Suggesting that asymmetric information is relevant for explaining capital structure of firms in the Netherlands. De Haan

(1992)

Researching whether macro-economic factors affect the capital structure of Dutch companies during 1970-1990.

Using a sequential regression analysis the relation between the capital structure and some macro-economic variables is investigated.

The capital structure of Dutch firms during the seventies and eighties can be well explained by development of the costs of equity, interest, inflation and changes in the tax

Section 3: Methodology

In this section, I show in greater detail how the hypotheses are addressed. The empirical model to test the relationship between capital structure, its determinants and the environment will be given (sub-section 3.1). Next I describe how to test whether leverage has a different impact on performance during a stable or dynamic environment (sub-section 3.2).

3.1: Capital structure and its determinants in a stable versus dynamic environment. As discussed in section 2, the capital structure for a company depends on several factors explained by theory and on overall economic conditions. The panel data model should therefore reflect both firm specific and macro-economic variables, like the one developed by Ariff, Taufiq and Shamsher (2008).

∑

∑

+

= i it i t it

it a a X Y

D 1 β ε (1)

Where Dit is the debt level, firms are represented by subscript i, and subscript t represents

the month. Furthermore, Xi represents the firm specific variables and Yt the

macro-economic variables.

The dependent variable is given by the level of debt financing. Two measures for financial leverage will be used (Chen et al., 1999):

1. Total debt divided by equity book value (LEVB). 2. Total debt divided by equity market value (LEVM).

Although it might have been better to use the market value of debt, I have to use the book value, since the market value is not available to me. However, according to Bowman (1980) the cross-sectional correlation between the book and market value of debt is very large, so the possible miss-specification due to using book value is probably quite small. Equity market value is the product of year-end stock price and the number of shares outstanding. Both book value and market value leverage are measured since several capital structure theories have not explicitly stated which leverage measure should be used. Also, most empirical studies have used both book and market value.

Table 4: Explanatory firm specific variables and their expected relationship with the leverage measure

Variable Name variable Description Expected

sign

Based on

X1 Company size Natural logarithm of total assets + Titman and Wessels (1988)

X2 Profitability Return on assets (ROA) - Titman and Wessels

(!988) X3 Tangibility Ratio of tangible assets to total

assets

+ Johnson (1997)

X4 Non-debt tax

shields

Ratio of depreciation over total assets

- Titman and Wessels (1988)

X5 Earning volatility Absolute value of the first difference of percentage change of operating income

- Chen, Lensink and Sterken (1999)

Company size (SIZE) is expected to have a positive relationship with leverage. Larger companies can more easily diversify their investment projects and so limit their risk to cyclical fluctuation in a particular production line. So the risk of financial distress is expected to be lower for larger companies. In addition, a positive relationship also supports the asymmetric information theory (Myers and Majluf, 1984). The bigger a company is, the less information asymmetry is present, that is outsiders have more information about the company. Larger corporations will therefore get easier access to debt finance.

Profitability (PROF) should have a negative sign if the pecking order theory holds. The better the past profitability and hence the more retained earnings a company has, the better a corporation is in the position to finance future projects with internal funds instead of external debt financing. Since previous profitability will impact on the current leverage, ROA for the month preceding t (i.e. t-1) will be used in the regression.

ensure that their loans are backed up by some collateral assets (Chen and Jiang, 2001). So corporations with less tangible assets will find it more difficult to raise funds by debt financing. Tangible assets are measured as the sum of fixed assets and inventories.

Non-debt tax shields (NDTS) are expected to be negatively related to leverage. DeAngelo and Masulis (1980) argued that tax deductions for depreciation and investment tax credits are substitutes for the tax benefits of debt financing. So when companies have large non-debt tax shields, they include less debt in their capital structures. As an indicator for the non-debt tax shield (NDT), the ratio of depreciation over total assets is used.

Earnings volatility (EVOL) can have a negative impact on leverage. Financial markets usually consider a company’s volatile earnings as a result of poor management and so discount the company’s stock price and demand a premium if such a company wants debt finance. These corporations find it harder to obtain external financing. (Chen, Lensink and Sterken, 1999)

Table 5. Explanatory macroeconomic factors and their expected relationship with the leverage measure

Variable Name variable Description Expected sign Based on

Y1 Economic growth Real Gross Domestic Product growth

+ Ariff, Shamsher and Taufiq (2008)

Y2 Money supply Annual change of M3 - Ariff, Shamsher and

Taufiq (2008) Y3 Interest rate Base lending rate of

commercial banks

- Henderson, Jegadeesh and Weisbach (2004) Y4 Inflation rate Consumer price index + Ariff, Shamsher and

Taufiq (2008)

Economic growth (GDP) is expected to have a positive impact on leverage. During periods of growth, companies resort to debt financing to finance their expansion projects.

Interest rate (INT) increases are considered to reduce the amount of long-term debt firms resort to.

Inflation rate (INFL) has a positive relationship with leverage. During inflation, companies employ more debt as the real cost of debt falls.

Figures one to four in the appendix show graphs of how the above macro-economic factors developed over the last 25 years.

Several studies have supported the theory that the type of industry a company is in, is an important determinant of capital structure (e.g. Bowen et al., 1982; Collins and Sekely, 1988), while others have not (e.g. Naidu, 1983). Since one of the research periods in my study concerns the Dot-com bubble, the IT industry might naturally be more affected than any other industry. Therefore, the type of industry involved might matter in this case and so the research uses the Nomenclature statistique des Activités économiques dans la Communauté Européenneas (NACE) code, an industrial classification to categorize companies into different industries. Cools (1993) and De Haan et al. (1994) also suggest that the capital structure choice differs between groups. Just like Kuo and Wang (2005) I will examine whether there is an industry effect on firm-specific financial ratios by estimating the following equation:

∑

∑

+

= i it i it i in it

it a a X Y I

D ₀ β γ ε (2)

where In is a dummy variable to identify the industry classification. Table 6 shows the

industries and their NACE codes that I distinguish. For example, to research whether IT companies have a significant different level of financial leverage, a company that has a NACE code beginning with J will be labeled as IT industry, I=1; in all other cases, a firm is labeled as non-IT industry, I=0.

Wald test will be applied in this research to examine whether the explanatory variables have significant changes in the coefficients for a stable versus dynamic environment.

Table 6: Industry classifications according to NACE

Industry NACE code

Agriculture, forestry, fishing, mining and quarrying. A and B

Manufacturing C

Electricity, gas, steam, air conditioning and water supply; sewerage, waste management and remediation activities

D and E

Construction F

Wholesale and retail trade; repair of motor vehicles and motorcycles

G

Transportation and storage H

Accommodation and food service activities I

Information and communication J

Financial, insurance and real estate activities K and L

Professional, scientific and technical activities M

(Public) administrative and support service activities; social security

N and O Education, human health, social work, arts, entertainment and

recreation

P, Q and R

Other service activities S, T and U

Note: All NACE codes can be found on the website of the European Union: http://ec.europa.eu/

3.2: Performance affected by capital structure in a stable versus dynamic environment.

it i t i it a a D a S R = 0 + 1 ,−1+ 2 +ε (3)

where Rit measures the profit efficiency, companies are represented by subscript i, and

subscript t represents the month; Dit-1 is the debt level of the month preceding month t; Si

is a control variable for company size.

Since Berger and di Patti (2002) suggest that profit efficiency and expected returns are strongly positively correlated, I will use return measures as a proxy for Rit.

Three return measures (Kuo and Wang, 2005) are used to make sure that results are robust:

1. Return on total assets (ROA) = after-tax income before interest / total assets. 2. Operating profit margin (OPM) = operating income / net sales.

3. After-tax profit ratio (APR) = after-tax income / net sales.

Furthermore, the level of debt (Dit-1) and firm size (Si) are measured in the same way

as in equation (1).

Again a Wald test is applied to research whether there are significant differences in the variables when the environment is stable or dynamic.

Section 4: Data

The definition of environmental dynamism applied in this research is a period characterized with unpredictable, unsystematic and uncertain changes in which many companies get in financial distress, go bankrupt or grow disproportionately fast (Das, Das and Lim, 2009). A stable environment, in contrast to a dynamic environment, means that a company is able to develop a fixed set of routines for dealing with environmental elements (Aldrich, 1979).

more recent periods for my research. Figure 2 to Figure 5 show the development of real GDP growth, annual change of money supply, interest rates and the inflation since 1980.

For my research I include all companies that were listed on the Dutch stock exchange during a particular period. Making use of listed companies makes it interesting to look at the development of their market values (Figure 6). The first remarkable development is a huge decrease in October 1987. Then again in 2002 the market value fluctuates heavily. These developments can be explained by the stock market crash of 1987 and the so called Dot-com bubble that started around 1997. When looking at Figure 8 to Figure 11 in the appendix, it can be seen that because of the stock market crash in 1987, consumer confidence became negative and market value of listed companies, which reflects investors’ confidence, also decreased. Both of these factors suggest an uncertain economic period. The same applies for the period of the Dot-com bubble during which consumer trust and market value of listed companies also started to fall at the end of 2000. Since I define a dynamic environment as a period characterized by unpredictable, unsystematic and uncertain changes, both Black Monday and the Dot-com bubble can be called dynamic periods.

In the following two sub-sections, a description will be given of Black Monday and the Dot-com Bubble, and what periods are chosen to reflect a stable and dynamic environment.

4.1: Stock market crash of 1987

After a period of worldwide strong economic optimism in the mid 1980s, a crash occurred on October 19, 1987, known as Black Monday. It was a climax of the market decline that had begun five days earlier. Up to 1987, this crash was the greatest single-day loss that Wall Street had ever suffered. However, it was not limited to the US. The AEX lost 12% in only one day, which was a start of a large deterioration. In November 1987 the AEX had lost more than 46%. The rapid increase of the short-term interest rates contributed significantly to this loss (InfoNu.nl). Although many feared a repeat of the 1930s Great Depression, the market recovered quite quickly. The Dow, for example, took only two years to regain all of the value it had lost in the crash (InfoNu.nl).

into a pre-crisis period to reflect a stable environment (January 1985- September 1987), and a dynamic environment that includes the stock market crash and recovery period (October 1987 – June 1989). This recovery period is chosen to reflect the time it took to restore market values of listed companies to their pre-crisis levels. Recovery is included in the dynamic period because, during recovery, governments or central banks might adjust money supply and interest rates. These adjustments have uncertain effects on the environment and so it is only until investor confidence is restored that one can speak again of a stable environment.

4.2: Dot-com bubble

During the period of 1997 to 2001, stock market values in the Western world increased rapidly due to growth in the new Internet sector and related fields. Many internet-based companies, so called dot-coms, were founded during this period. Due to a combination of rapidly increasing share prices, great stock speculation by individuals and widely available venture capital, the environment became fairly optimistic and businesses ignored standard business models. The focus was on market share at the expense of attention to net income (Kraay and Ventura, 2005).

In 2001 the bubble burst, which caused a small worldwide recession that lasted longer than expected for most Western countries. As can been seen in Figure 10 and Figure 11, both consumer and investor confidence started to drop at the beginning of 2001 and continued to fall, which is why January 2001 will be used as the start of the dynamic environment. The terrorist attacks that took place in the US on 11 September 2001 increased uncertainty among investors and consumers even more, which is why this period should be included in the dynamic environment.

A sensitivity analysis will show whether the specification of these particular stable and dynamic periods has an effect on the outcomes.

4.3: Data description

The sample is constructed using Thomson Datastream and covers the periods 1985 to 1989 and 1997 to 2002. All the selected companies are incorporated in the Netherlands. Companies that have incomplete records on required accounting issues for a certain year are excluded. In addition, financial companies like banks and insurance companies will be excluded from research because of their different accounting methods. They are highly regulated by central banks and have to comply with various legal requirements on financing.

For Black Monday the sample consists of 39 companies; for the Dot-com Bubble the sample consists of 83 companies. Figure 7 shows to which industry the companies in the samples are classified. Most companies are in the manufacturing or wholesale and retail sector. 25 companies that are in the Black Monday sample are also in the Dot-com Bubble sample. The other companies either did not survive or data was no longer available.

Furthermore, the aggregate macro-economic data such as (i) economic growth, (ii) money supply, (iii) interest rates, and (iv) inflation rates are gathered from databases of the Central Bureau of Statistics (CBS), De Nederlandsche Bank (DNB), the International Financial Statistics (IFS) of the International Monetary Fund (IMF) and from statistics of the Organisation for Economic Co-operation and Development (OECD).

The macro-economic variables show that the inflation rate was much higher for the 1997-2002 period, than for the 1985-1989 period, while the interest rate was lower for the Dot-com Bubble period.

Tables 16 and 17 in the appendix give the summary statistics for the Stable versus Dynamic period.

Table 7: Summary statistics

Black Monday firms Dot-com Bubble firms

Mean Median Std Dev Mean Median Std Dev

SIZE 13.27 13.21 1.411 13.03 13.17 2.093 PROF 7.10% 6.93% 0.042 6.18% 6.16% 0.179 TANG 32.74% 34.19% 0.235 43.46% 38.83% 0.308 NDTS 5.94% 4.60% 0.084 5.30% 5.40% 0.151 EVOL 19.81% 5.78% 0.585 197.75% 7.17% 37.831 GDP 0.25% 0.25% 0.001 0.26% 0.32% 0.001 MS 8.16% 7.40% 0.035 10.36% 9.68% 0.038 INT 8.91 8.75 1.183 4.97 5.13 1.142 INFL 0.76 0.80 1.020 2.83 2.40 0.933 OPM 4.94% 5.29% 0.071 13.38% 6.48% 0.550 APR 4.11% 3.67% 0.050 7.35$ 4.07$ 0.798

SIZE is the natural logarithm of total assets. PROF (profitability) is the ratio of after-tax income before interest to total assets. TANG (tangibility) is the ratio of tangible assets to total assets. NDTS (non-debt tax shield) is a ratio of depreciation to total assets. EVOL (earnings volatility) is the absolute value of the first difference of percentage change of operating income. GDP (gross domestic product) is the real gross domestic product growth. MS (money supply) is the yearly change in M3. INT (interest rate) is the base lending rate of commercial banks. INFL (inflation rate) is the consumer price index. OPM (operating profit margin) is the ratio of operating income to net sales. APR (after-tax profit ratio) is the ratio of after-tax income to net sales.

Table 8: Mean values of Market and Book Leverage

Black Monday firms Dot-com Bubble firms

Stable Dynamic Stable Dynamic

LEVM 269.40% 442.58% 41.61% 82.20%

LEVB 78.63% 58.87% 92.45% 132.47%

LEVM (market leverage) is the debt to equity market value ratio. LEVB (book leverage) is the debt to equity book value ratio.

Source: DATASTREAM

Figure 1: Mean values of Market and Book Leverage over time

Black Monday sample Dot-com Bubble sample

Source: DATASTREAM

Table 8 shows that the leverage ratios are quite different for stable and dynamic periods. As can be seen in the figure, for the Black Monday period the leverage ratio increased steeply in 1987 and started to decrease again in 1988. The decrease was probably due to the rapid increase in short-term interest rates. From 1997 to 2000 the leverage ratio increased steadily. However, in 2001 a sudden increase appears and again in 2002. The graphs show that the leverage ratio, and so the capital structure of the Dutch firms, definitely responded to the market crash in 1987 and the burst of the Dot-com Bubble in 2001.

While non-debt tax shield and tangibility have relatively high correlation with market leverage in the Black Monday sample, the correlations with book leverage are lower. This suggests that there will not be a serious spurious effect. However, looking further at the correlation matrix of Black Monday, one notices that change of money supply, real Gross Domestic Product and interest rate are all highly correlated with each other, namely corr(GDP,MS)=0.952, corr(INT,GDP)=0.719, and corr(MS,INT)=0.822. This suggests that in my multivariate model, none of the coefficients of inflation, money supply and interest rate individually has a significant impact on the choice of a company’s leverage. However, this does not mean that as these variables all vary together, there is no significant statistical relation between a corporation’s leverage and each specific variable. If, for example, a theory suggests that company size is a significant determinant of capital structure choice without specifying the environment, including the correlation with other variables, then the findings of my model estimation is not necessarily inconsistent with the theory (Chen and Jiang, 2001).

The data is checked for normality using Shapiro-Wilks. The results suggest that the hypothesis of normality can be rejected for all variables. The non-normal distribution is a potential caveat, because of which my estimates might not be completely robust. Although the condition that the explanatory variables should be normally distributed is not necessary for obtaining consistent and unbiased OLS estimates, there should be sufficient variability in these to ensure that the estimators are defined. Otherwise an identification may follow as a result of multicolinearity stemming from the constant term and the respective variable with insufficient variance.

Section 5: Results

5.1: Capital structure in relation to environmental dynamism

The method of Ordinary Least Squares (OLS) does not take into account unobservable firm heterogeneity, which is why the estimates could have an omitted variable bias. To test whether I should use a fixed effects model, I have performed the Hausman specification test. The test statistics that I calculated for the Black Monday firms for the models with market leverage and book leverage as dependent variables are 68.33 and 83.96 respectively; and for the Dot-com Bubble firms 18.04 and 116.07 respectively. The random effects models can be rejected at all critical levels.

The estimation results of the panel fixed effects model for market leverage and book leverage are reported in Table 9 and Table 10. Since the companies in the panel differ in size, the error terms probably will not have the same variance and heteroscedasticity may be present. Therefore, the t-statistics are the t-values adjusted for heteroscedasticity consistent standard errors. The first column presents the results over the whole sample period, while columns two and three present the results pertaining to the stable period and dynamic period, respectively. The findings over the whole period are discussed first and an analysis of stable and dynamic period differences follows.

5.1.1. Results over the entire Black Monday period (1985-1989) and entire Dot-com Bubble period (1997-2002)

When not taking into account the different impact the stable versus dynamic environment can have on the capital structure choice, the companies in the Black Monday sample only give three significant estimates. When market leverage is the dependent variable, company size and tangibility have a negative coefficient. However, tangibility has a positive significant coefficient when book leverage is the dependent variable.

Results for the Dot-com Bubble period show more estimates that are significant. Profitability and money supply have positive coefficients for both leverage measures. Tangibility, real GDP and interest rate have negative estimates in both cases. Furthermore, when book leverage is the dependent variable, company size has a significant negative coefficient, and non-debt tax shield has a significant positive coefficient. Inflation has contrasting signs for the market and book leverage case.

capital structure and whether they are significant is dependent on how capital structure is measured. Market leverage gives different results than when using book leverage as the dependent variable.

Table 9 and Table 10, however, may be presenting inefficient estimates. An arbitrary shock affecting one company may affect other companies too because of relations between the companies. Hence, it may be the case that errors are correlated with one another. Doing the panel regression of the model in first differences solves the problem of auto- and cross-correlation of the disturbances, and therefore improves the robustness of the OLS estimation results (Chen, Lensink and Sterken, 1999).

The Hausman specification test is performed again, to test the fixed versus random effects. The test statistics that I calculated for the model with the first difference of market leverage (∆LEVM) and the first difference of book leverage (∆LEVB) as dependent variables are for the Black Monday sample 1.97 and 0.47 respectively, and for the Dot-com Bubble sample 0.81 and 9.54 respectively. The random effect models cannot be rejected at any critical level. In Table 11 and Table 12, results for the first differences panel model with random effects are showed.

The Black Monday companies have a negative significant estimate for tangibility and earnings volatilty, and a positive significant estimate for inflation when market leverage is the dependent variable. However, when book leverage is the dependent variable only non-debt tax shield has a significant estimate with a postive sign.

Dot-com Bubble results for both leverage measures show a significant negative estimate for company size, real GDP, interest rate and inflation. Also, in both cases a positive significant coefficient is found for profitability and non-debt tax shield.

Table 9: Equation (1). Panel data regression: fixed effects. For Black Monday firms, stable covers the period 1985 to September 1987; Dynamic is the period October 1987 to 1989. For Dot-com Bubble firms, stable covers the period 1997 to 2000; Dynamic is the period 2001 to 2002. Market leverage is the dependent variable.

Black Monday Dot-com Bubble

Full sample Stable Dynamic Full sample Stable Dynamic SIZE Wald test -20.316 (2.89)* 1.843 (0.38) -12.141 (1.83)*** (4.42)** 0.026 (1.06) -0.269 (11.79)* 0.175 (1.53) (0.93) PROF Wald test 39.346 (0.92) 40.305 (1.52) 5.553 (0.17) (1.07) 0.249 (3.07)* -1.459 (8.98)* 0.631 (3.60)* (59.99)* TANG Wald test -108.516 (8.20)* -34.515 (2.96)* -73.643 (19.89)* (111.65)* -0.066 (1.82)*** -0.042 (1.96)** 1.078 (5.38)* (1.41) NDTS Wald test 31.049 (1.00) 537.083 (9.49) -439.938 (6.81)* (229.00)* 0.060 (0.64) -1.208 (8.62)* -0.750 (1.56) (0.08) EVOL Wald test -0.013 (0.01) 0.155 (0.20) -0.850 (0.97) (1.32) 0.000 (0.16) 0.001 (5.19)* 0.072 (2.06)** (1.45) GDP Wald test 3285.468 (0.80) 2542.36 (0.79) 10244.1 (2.52)** (3.59)*** -102.397 (10.65)* -400.644 (14.11)* Dropped MS Wald test -13.441 (0.12) -60.532 (0.59) -124.463 (1.59) (0.67) 0.633 (2.49)** 0.471 (2.09)** 3.322 (5.11)* (4.16)** INT Wald test -0.942 (1.43) -0.201 (0.83) -1.331 (1.04) (0.78) -0.084 (10.88)* -0.127 (20.90)* -0.191 (5.48)* (0.81) INFL Wald test -0.520 (0.54) 1.238 (1.00) -2.867 (1.08) (2.38) 0.024 (1.77)*** 0.102 (5.97)* -0.244 (4.25)* (40.79)* R² 0.3617 0.650 0.058 0.047 0.024 0.000 F-test value 1619.94* 2499.03* 2006.16* 76.39* 70.42* 19.02* Note: it t t t t t i t i t i t i t i t i INFL INT MS GDP EVOL NDTS TANG PROF SIZE LEVM ε β β β β α α α α α α + + + + + + + + + + = ₋ 4 3 2 1 , 5 , 4 , 3 1 , 2 , 1 0 ,

The t-statistics are the t-values adjusted for heteroscedasticity consistent standard errors. The absolute value of t-statistics is reported in parentheses below the coefficient estimates. Next to F-test value is the F-statistic for the hypothesis that all

coefficients are zero. Wald test gives the statistic for testing if estimates for the stable and dynamic period are equal. See Table 7 and section 4 for the definition of the variables.

Table 10: Equation (1). Panel data regression: fixed effects. For Black Monday firms, stable covers the period 1985 to September 1987; Dynamic is the period October 1987 to 1989. For Dot-com Bubble firms, stable covers the period 1997 to 2000; Dynamic is the period 2001 to 2002. Book leverage is the dependent variable.

Black Monday Dot-com Bubble

Full sample Stable Dynamic Full sample Stable Dynamic SIZE Wald test 1.887 (1.44) -0.977 (0.28) 6.192 (1.05) (1.47) -0.424 (7.42)* -0.818 (11.83)* -3.054 (11.93)* (1.43) PROF Wald test 1.549 (0.42) -34.692 (1.07) 10.791 (0.98) (17.21)* 1.179 (6.25)* -0.086 (0.18) 11.673 (29.65)* (12.68)* TANG Wald test 1.631 (2.91)* -1.592 (0.15) 2.122 (1.16) (4.10)*** -0.580 (6.93)* -0.355 (5.50)* -3.014 (6.70)* (0.57) NDTS Wald test 3.191 (0.69) -15.806 (0.36) -9.912 (0.72) (0.18) 1.583 (7.18)* 0.372 (0.87) -6.585 (6.11)* (0.94) EVOL Wald test -0.205 (0.95) -0.086 (0.25) -0.358 (0.67) (0.26) 0.000 (0.92) 0.001 (3.36)* 0.132 (1.69)*** (0.73) GDP Wald test -689.069 (1.00) -589.394 (1.08) -962.529 (0.71) (0.08) -229.482 (10.26)* -1133.332 (13.17)* Dropped MS Wald test 0.077 (0.02) -8.920 (0.41) -2.177 (0.16) (0.23) 1.613 (2.73)* 3.723 (5.45)* -3.314 (2.27)** (5.59)** INT Wald test 0.058 (0.82) 0.010 (1.44) 0.007 (1.22) (0.16) -0.125 (6.97)* -0.286 (15.52)* -0.010 (0.13) (478.12)* INFL Wald test -0.115 (0.76) -0.009 (0.98) -0.004 (0.16) (0.04) -0.102 (3.22)* 0.271 (5.25)* 0.019 (0.15) (36.48)* R² 0.003 0.057 0.026 0.000 0.000 0.006 F-test value 100.74* 6.78* 1.45 29.23* 42.36* 124.96* Note: it t t t t t i t i t i t i t i t i INFL INT MS GDP EVOL NDTS TANG PROF SIZE LEVB ε β β β β α α α α α α + + + + + + + + + + = ₋ 4 3 2 1 , 5 , 4 , 3 1 , 2 , 1 0 ,

The t-statistics are the t-values adjusted for heteroscedasticity consistent standard errors. The absolute value of t-statistics is reported in parentheses below the coefficient estimates. Next to F-test value is the F-statistic for the hypothesis that all coefficients are zero. Wald test gives the statistic for testing if estimates for the stable and dynamic period are equal. See Table 7 and section 4 for the definition of the variables.

Profitability shows positive estimates when significant. Though, because of the pecking order theory, the coefficient was expected to be negative. A possible explanation can be the argument put forward by Ross (1977) that a firm’s capital structure choice signals information of good performance to investors, in which case investors take larger debt levels. Also known as the signalling theory.

The relationship between leverage and tangibility is found to be negative in most cases. This is surprising since one would expect firms with more tangible assets and therefore collateral value to borrow more easily than firms with lower tangibility.

Results for non-debt tax shield show a positive sign. This does not support the argument that firms with large non-debt tax shields include less debt in their capital structures.

The estimates of earnings volatility show differing signs and most estimates are not significant. It cannot be supported that companies with higher volatility of operating income have more difficulty obtaining external financing. Companies may ignore the earnings volatility if the risk and costs of entering into liquidation are low (Deesomsak et al., 2004).

The macroeconomic variables in the Black Monday sample barely show significance. This was already expected due to the high correlations found in Table 16. For the Dot-com Bubble companies almost all estimates are significant, however, only interest rate is found to have the expected relationship with leverage. When rates increase, companies reduce the amount of long-term debt they resort to.

Although it is expected that during periods of growth companies resort to debt financing to finance their expansion projects, the significant estimates for real GDP are negative.

The annual change of money supply has some positive significant estimates. Suggesting that when the money supply increased, companies increased their level of leverage. This is in contrast to expectations based on Ariff, Shamsher and Taufiq (2008).

Table 11: Equation (1) in first differences. Panel data regression: random effects. For Black Monday firms, stable covers the period 1985 to September 1987; Dynamic is the period October 1987 to 1989. For Dot-com Bubble firms, stable covers the period 1997 to 2000; Dynamic is the period 2001 to 2002. The first difference of market leverage is the dependent variable.

Black Monday Dot-com Bubble

Full sample Stable Dynamic Full sample Stable Dynamic SIZE Wald test 4.237 (0.72) 1.099 (0.95) -7.031 (2.51)** (8.39)* -0.191 (3.85)* -0.392 (1.55) 0.204 (0.92) (7.41)* PROF Wald test -29.421 (1.43) 4.106 (1.08) -1.911 (0.14) (0.20) 0.521 (4.57)* -0.721 (1.83)*** 0.230 (0.48) (3.90)** TANG Wald test -21.536 (3.34)* -0.476 (0.16) -35.191 (20.49)* (408.68)* 0.052 (1.09) -0.035 (0.38) 0.857 (1.88)*** (3.84)** NDTS Wald test 148.420 (1.58) 71.806 (6.59)* 273.367 (14.88)* (120.45)* 0.810 (5.00)* -0.188 (0.52) 0.449 (0.69) (0.88) EVOL Wald test -0.601 (2.06)** 0.122 (0.88) 0.137 (0.38) (0.00) 0.000 (0.15) 0.000 (0.37) 0.036 (0.86) (0.74) GDP Wald test 1379.593 (1.33) -847.185 (2.46)** 1808.099 (0.40) (0.35) -31.965 (2.04)** -144.425 (2.00)** -55.851 (1.30) (4.19)** MS Wald test -28.389 (1.27) 12.862 (1.14) -23.085 (0.24) (0.14) 0.729 (3.01)* 1.317 (1.20) 1.570 (1.75)*** (0.08) INT Wald test -0.142 (0.77) 0.189 (0.68) -0.113 (0.25) (0.45) -0.050 (4.04)* -0.073 (2.31)** -0.036 (0.81) (0.70) INFL Wald test 0.912 (2.23)** 0.457 (1.39) 0.832 (0.51) (0.05) -0.050 (3.22)* 0.008 (0.85) -0.155 (2.02)** (4.52)** R² 0.097 0.022 0.193 0.011 0.039 0.014 Wald χ²(9) 24.78* 351.70* 1220.48* 7.02* 14.63 13.81 Note: it t t t t t i t i t i t i t i t i INFL INT MS GDP EVOL NDTS TANG PROF SIZE LEVM ε β β β β α α α α α α + + + + + + + + + + = ∆ ₋ 4 3 2 1 , 5 , 4 , 3 1 , 2 , 1 0 ,

The z-statistics are the z-values adjusted for heteroscedasticity consistent standard errors. The absolute value of z-statistics is reported in parentheses below the coefficient estimates. Next to Wald χ²(9) is the Chi-statistic for the hypothesis that all coefficients are zero. Wald test gives the statistic for testing if estimates for the stable and dynamic period are equal. See Table 7 and section 4 for the definition of the variables.

Table 12: Equation (1) in first differences. Panel data regression: random effects. For Black Monday firms, stable covers the period 1985 to September 1987; Dynamic is the period October 1987 to 1989. For Dot-com Bubble firms, stable covers the period 1997 to 2000; Dynamic is the period 2001 to 2002. The first difference of book leverage is the dependent variable.

Black Monday Dot-com Bubble

Full sample Stable Dynamic Full sample Stable Dynamic SIZE Wald test 1.875 (1.12) 0.048 (0.02) 5.319 (1.19) (1.39) -1.640 (21.08)* -1.989 (2.08)** -2.872 (2.18)** (0.40) PROF Wald test -7.008 (0.59) -35.076 (1.15) 6.799 (0.92) (31.78)* 4.900 (27.36)* 0.707 (0.59) 6.642 (2.11)** (3.62)*** TANG Wald test 1.203 (1.20) 2.940 (0.43) 1.101 (1.05) (3.07)*** -0.681 (9.17)* -0.620 (1.38) -2.604 (1.34) (1.03) NDTS Wald test 9.078 (1.76)*** 8.075 (0.32) -1.669 (0.16) (0.82) 5.463 (21.49)* 2.304 (1.69)*** -3.036 (0.57) (1.03) EVOL Wald test -0.108 (0.95) 0.087 (0.25) -0.213 (0.76) (1.15) 0.001 (3.18)* 0.001 (1.80)*** 0.019 (0.17) (0.02) GDP Wald test -899.236 (1.45) -642.080 (1.32) -1208.646 (0.58) (0.07) -284.538 (11.57)* -717.613 (2.73)* -183.979 (1.29) (14.10)* MS Wald test 5.195 (0.45) -4.993 (0.32) 5.075 (0.14) (0.08) -0.492 (1.30) 2.903 (1.13) -1.618 (0.96 (7.38)* INT Wald test 0.034 (0.36) 0.011 (1.01) 0.000 (0.02) (0.95) -0.099 (5.12)* -0.229 (3.08)* 0.003 (0.68) (2151.90)* INFL Wald test -0.122 (0.69) -0.003 (0.42) -0.005 (0.18) (0.00) -0.136 (5.54)* 0.029 (2.55)** -0.005 (0.53) (7.61)* R² 0.041 0.1051 0.081 0.156 0.197 0.215 Wald χ²(9) 9.16 4.26 4.38 117.48* 10.60 8.54 Note: it t t t t t i t i t i t i t i t i INFL INT MS GDP EVOL NDTS TANG PROF SIZE LEVB ε β β β β α α α α α α + + + + + + + + + + = ∆ ₋ 4 3 2 1 , 5 , 4 , 3 1 , 2 , 1 0 ,

The z-statistics are the z-values adjusted for heteroscedasticity consistent standard errors. The absolute value of z-statistics is reported in parentheses below the coefficient estimates. Next to Wald χ²(9) is the Chi-statistic for the hypothesis that all coefficients are zero. Wald test gives the statistic for testing if estimates for the stable and dynamic period are equal. See Table 7 and section 4 for the definition of the variables.

All these findings combined, no decisive answer can be given as to what the determinants of Dutch capital structure are and how the different variables impact on the capital structure. Although there are some significant coefficients, they are not consistent for the two measures of capital structure, market and book leverage, and they are not consistent for the two samples.

5.1.2. Results by stable and dynamic periods

Performing the Wald test to the fixed effects panel model and the random effects panel model of first differences shows that some coefficients have significant differences between estimates during the stable and dynamic period. This implies that changes to the overall economic environment may significantly alter the determinants of company’s decissions. Results are in Table 9 to Table 12. Taking an example, the tables show that in most regressions the coefficient of profitability in the stable period is significantly different from the one in the dynamic period. The profitability coefficient has a negative sign for the stable period, which is predicted by literature (see Table 4), but the sign becomes positive for the dynamic period. This suggests that when return on assets increases during a dynamic environment, companies increase their leverage ratios. However, although most differences between the coefficients are significant, the estimates themselves are not.

For Black Monday, when leverage is measured by its book value, none of the coefficient estimates are significant during the stable or dynamic period. On the other hand, when market leverage is used some significant estimates are found. The coefficient for non-debt tax shield, for example, is significant for both the stable and dynamic period, but with different impact.

Companies in the Dot-com Bubble sample show that earnings volatility has a positive significant impact on leverage in the stable period, but an even higher positive impact in the dynamic period. More significant estimates can be found in the tables above, but none of the differences between stable and dynamic estimates is consistent for both measures of capital structure, market and book leverage.

consistency between the Black Monday and Dot-com Bubble samples. Hypothesis 1 cannot be suported.

5.1.3. Industry effects

To research whether there is an industry effect on the leverage ratios, dummies were added to the regression, resulting in equation (2). All industry dummy estimates were significantly positive and very similar. Therefore, it cannot be stated that for example the IT-industry has been affected more by the dynamic environment during the Dot-com Bubble, than did any other industry. A possible explanation for the reduction of industry influence comes from an overall reduction of industry distinctions. Over the past several decades large, multinational companies are more and more engaging in activities in more than one industry. Hence, like Bowen et al. (1982) suggest, "there is potential conglomeration effect which would tend to favour the null hypothesis of no industry influence".

5.1.4. Sensitivity analysis

To test whether my research was sensitive to the particular specification of the stable and dynamic period, I have performed the regressions with first differences again with different periods for the stable and dynamic environment. Results can be found in Tables 20 and 21. I have not found a reason to reconsider the particular stable and dynamic periods used.

Up to this moment I have considered recovery to be part of the dynamic period. In Tables 22 and 23 results can be found for the first differences regression when the sample is instead divided into three periods consisting of a stable, dynamic and recovery period. However, the results show that only a couple of the estimates are significant, which is why a three period segregation is not considered relevant.

5.2: Performance affected by capital structure in a stable versus dynamic environment.

variables. For both samples all Hausman statistics that I calculated showed that a fixed effects model has to be used. The random effects models can be rejected at all critical levels.

All estimation results of the fixed effects panel models are reported in Table 13 to Table 15. For each sample, the first column presents the results over the whole sample period, while columns two and three present the results pertaining to the stable period and dynamic period, respectively. The findings over the whole period are discussed first and an analysis of stable and dynamic period differences follows.

5.2.1. Results over the entire Black Monday period (1985-1989) and entire Dot-com Bubble period (1997-2002)

For Black Monday firms over the entire 1985-1989 period, the results for operating profit margin (OPM) as the dependent variable are all insignificant. However, for the other 2 profit efficiency measures some significant results are found. First of all, firm size (SIZE) has a negative impact on both return on assets (ROA) and the after-tax profit ratio (APR), suggesting that larger firms are less profit efficient. Furthermore, market leverage (LEVM) has a negative significant effect on return on assets while it has a positive significant impact on the after-tax profit ratio. Finally, book leverage (LEVB) only has a positive significant coefficient in relation to return on assets.

For Dot-com Bubble companies over the entire 1997-2002 period, none of the coefficient estimates are significant.

Results for the full samples do not give decisive answers as to how performance is affected by the capital structure.

5.2.2. Results by stable and dynamic periods