A Broader Perspective on the Pecking Order Theory
Maurits. G. T. Tiemstra
s1652192
Master Thesis
University of Groningen
Faculty Economics and Business
MSc Business Administration
Specialization Finance
August 2012
Supervisor: B.A. Boonstra
Abstract:
In 2008, a financial crisis started, which drastically slowed down the European economy. Since then Banks and European governments are at a stalemate concerning the role and preparedness of banks to fund firms through bank credit in uncertain economic circumstances, where the
relationship between investments and economic growth at the one hand and financing decisions of firms at the other are not always clear. As a contribution to this issue this paper intends to shed light on the investment and financing decisions of firms through the use of two models: firstly a basic pecking order model and secondly an extended pecking order model which includes Nominal
GDP and EBITDA. Based upon individual industries and an aggregate data set the results of this
analysis show that in all but the Health Care Industry, the outcomes of the basic pecking order model are not consistent with the pecking order theory. However, when including variables
Nominal GDP and EBITDA into the aggregate data set, this extended pecking order model shows an
increase of explanatory power.
JEL classification: G30, G32
2
Contents
1. Introduction ... 3
2. Literature Review ... 5
3. The Pecking Order Models ... 9
4. Data description and descriptive statistics... 13
3
1. Introduction
The financial crisis that started in 2008 has provided the opportunity to add a whole new dimension to understanding investment and financing decisions. European banks are keeping a tight grip on their credit due to the difficult economic conditions, making it harder for firms to finance their investments. Banks believe that the economy must first pick up before they can safely provide credit. Governments across Europe, on the other hand, are trying to stimulate banks to make credit available to firms in order to contribute to economic recovery.
In this paper a further understanding of the relationship between investment decisions and
financing is provided. This understanding can give both banks and European governments a better insight into the drivers needed to boost the European economy. When looking at the financing methods that firms use, it is useful to look at the much debated pecking order theory. This theory, which is based on the capital structure theory presented by Modigliani and Miller (1958), was first introduced by Myers and Maljuf in 1984. It states that internal funding is preferred to external financing and if external financing is used debt is favored over equity.
The pecking order theory helps to explain the decision process regarding the way firms finance their investments. To get a better understanding of the functioning of the pecking order theory, in this paper a basic pecking order model will be developed to test this theory.
In order to be able to look at investment and financing decisions in a broader perspective, this paper will subsequently create an extended pecking order model by adding two additional control variables to the basic model. These two additional control variables are Nominal GDP to reflect the macro-economic environment in which the firm operates and EBITDA (Earnings Before Interest, Taxes, Depreciation and Amortization) as the parameter that best reflects the operational cash flow generation of the firm before investments and movements in working capital.
This approach helps to answer the research question of this paper:
Can the basic and extended pecking order models help to understand investment and financing decisions of firms?
4 This paper offers first an overview of the pecking order theory, including findings from other research papers, after which the hypotheses are introduced. Then a pecking order theory model is presented, including the details on how this model can be adjusted to include the variables Nominal
GDP and EBITDA. The dataset and methodology follow subsequently, after which the results of the
5
2. Literature Review
As explained by Frank and Goyal (2007), the theory of business finance in a modern context starts with the Modigliani and Miller (1958) capital structure irrelevance proposition. Modigliani and Miller found that the value of a firm is solely based on the riskiness of its future cash-flows,
assuming that the financial markets are perfect, meaning there are no tax or transaction costs. This means that the value of a firm is indifferent to the capital structure it has.
However, as explained in a later paper by Modigliani and Miller (1963), the capital structure does have an important influence on the value of a firm and on how investments are financed, due to: 1. the absence of perfect capital markets; 2. the presence of transaction costs; and 3. the tax
deductibility of interest payments. Frank and Goyal (2007) explain that, as a result of the existence of these market imperfections, the Modigliani-Miller capital structure irrelevance proposition does not provide a realistic description of how firms finance investments in the real world. It only clarifies why financing may matter. As a result of the limited explanation power of Modigliani and Miller (1958) researchers have searched for theories that do shed more light on how firms finance investments.
Three theories, that analyze financing decisions of firms and are based on Modigliani and Miller (1958), seem to stick out:
1. The agency theory, presented by Jensen and Meckling (1976), states that debt can be used as an instrument to ease conflicts between a firms' managers and their principal equity
holders by aligning each other’s interest.
2. The static trade-off theory, presented by Myers (1984) states that an optimal level of leverage is influenced by three factors, namely taxes, cost of financial distress and agency costs. It is further argued by this theory that when using debt, the costs are offset by the benefits of debt financing, such as a tax shield.
6 Of these three theories the pecking order theory is the only one that solely focuses on the financing method of investments and is therefore best suited to be used in addressing the research question in this paper.
As explained before, the pecking order theory tries to bridge the gap between the Modigliani and Miller capital structure irrelevance proposition and the capital structure choices that firms actually make. The purpose of this approach is to understand the decision-making process of firms, which can help predict companies’ future financing judgments.
The pecking order theory came to light over two decades ago. Since then scores of researchers have attempted to analyze whether and when the pecking order theory correctly explained observed financing behavior. Ariff, Shamsher and Taufiq(2008), Bharath, Puasquariello and Wu (2009), Chirinko and Singha(2000), Fama and French (2005), De Haan (1992), Jung, Kim and Stulz (2005), Frank and Goyal (2003), Leary and Roberts (2010), Lemmon and Zender (2004), and Shyam and Myers (1999) have all analyzed the explanation power of the pecking order theory. These papers found results which conflict with each other.
Shyam-Sunder and Myers (1999) compare traditional capital structure models such as the static trade-off model with a pecking order theory model. Its goal is to see which model has a better ability to explain financing decisions. They concluded that the pecking order theory provides a good description of broad financing decisions. This is in stark contrast to the findings of Frank and Goyal (2003) which found the opposite when they analyzed whether the pecking order theory can be used to explain financing decisions of a broad cross-section of publicly traded American firms from 1971 to 1998.
Another interesting conflict concerning the pecking order theory can be found between the findings of Lemmon and Zender (2004) on the one hand and Fama and French (2005) on the other. Lemmon and Zender (2004) analyzed the impact of explicitly incorporating a measure of debt-capacity in recent tests of competing theories of capital structure. They found that if the pecking order theory takes into account financial distress costs, then this “modified“ model would be a good descriptor of financing behavior. Fama and French (2005) found the opposite when they also took the issuance of equity into consideration.
7 is said in Jung et al. (2005) which analyzes the workings of the pecking order model, the agency model and the timing model.
Thus, a great deal of conflict surrounds the pecking order theory. Leary and Roberts (2010) tries to analyze these discrepancies. They argue that the divergence of findings concerning the pecking order theory is due to two main reasons. In the first place many of the papers mentioned above use testing strategies that are hampered by problems of statistical nature. This is due to the fact that their regressions are based on Shyam – Sunder and Myers (1999) and as Chirinko and Singha (2000) analyzed, that model has no power to discriminate among alternative explanations. This makes it difficult to prove the pecking order theory. The second problem is that researchers focus on the modified pecking order which Myers (1984) describes as “grossly oversimplified and under-qualified”.
After having identified the problems with past pecking order theory research Leary and Roberts (2010) try to eradicate the conflicting evidence by presenting a new approach to pecking order theory research. Leary and Roberts (2010) goal is to “shed light on this debate by quantifying the empirical relevance of the pecking order and its variants using a novel empirical model and testing strategy that addresses the relevant power concerns”. With this approach they find that the pecking order theory can explain financing decisions. This is however dependent on how the pecking order theory is interpreted. The stricter the interpretation the less explanatory power the pecking order theory has.
The findings and method of Leary and Roberts (2010) brings the analyzes of the pecking order theory to a new level due to the thorough evaluation of past research and its well structured model. Therefore, Leary and Roberts (2010) is used as a basis for the research in this paper. The workings of this model, and hence how the Leary and Roberts approach works, is explained in the model section of this paper.
In the context of its intention to extend the basic pecking order model with the addition of variables of a macroeconomic nature this paper has done research on where this has been done before in literature. Two examples of relevant studies have been summarized below:
8 interest rates and changes in the tax administration. He explains that through the use of the pecking order theory the macro-economical variables, which he analyzed, were useful in explaining the changes that had occurred in the 1970s and 1980s concerning capital structure of companies. Also Ariff et al.(2008) uses macroeconomic variables when researching how capital structure adjusts dynamically during financial crises. The paper looks at the factors that drive capital
structure adjustment of financially distressed and of healthy firms, using, among other theories, the pecking order theory. It is interesting to note that Ariff et al. include macro-economical variables such as gross domestic product, money supply, interest rates, exchange rates and inflation rates. The variables mentioned are used as control variables because they believe that it will be useful to take these matters into account during times of financial crises. Ariff et al. (2008) continue to explain that researchers rarely actually analyzed the influence of macro-economical variables on capital structure and suggest that further research should be done on this issue, especially in times of financial crisis.
Both these papers show the importance of placing the pecking order theory into a broader perspective and the need for further research. Therefore, this paper plans to add the control variables Nominal GDP and EBITDA to the basic pecking order model to do further research on its implementation. This might give the banks and European governments additional insights regarding how firms finance their investments.
Hypotheses Stating
Based on the pecking order theory this paper hypothesizes that:
Hypothesis 1: There is a negative relationship between Investments and Cash; Hypothesis 2: There is a positive relationship between Investments and Debt.
After considering the literature review and the first two hypotheses it is relevant to look at the pecking order theory in the broader perspective by including the control variables Nominal GDP and EBITDA. Based on the inclusion of these two variables this paper hypothesizes that
9
3. The Pecking Order Models
Let us first summarize the basic pecking order model which is based on the Leary and Roberts (2010) paper.
The first equation of this model explains the scenario when a firm is not in need of external funding but is planning to invest, where Investment represents all expansive investments, as well as all re-investments resulting from and are assumed to be equal to depreciation and amortization.
(1) Equation (1) shows that in first instance financing an investment will be done using cash up to point . It assumes that the investment takes place at the closing date of the period presented by it, where i indexes the firm and t represents the year. Furthermore, represents a mean zero random variable. represents the capacity for a firm to keep cash on its balance-sheet without investing it. This refers to the ability for a firm to maintain a cash management policy. The decision between the amount of internal and of external financing ( ), which is seen as the first rung of the pecking order model, is represented in this equation:
{
(2) where .
Equation (2) dictates that investments are to be financed externally if meaning that internal funds are insufficient to fund investment needs. Otherwise, if the firm solely relies on internal funds to finance investment and Equation (1) is sufficient.
When external resources ( ) are needed, first debt is used for financing. The incorporation of debt is represented by the following equation,
( ) ( ). (3) Hence, Equation (3) represents the situation when a firm has exhausted all its internal resources
10 point represented by , where represents a mean zero random variable. refers to the ability for a firm to maintain a debt management policy.
This choice between debt and equity is represented by the following equation: {
(4) where ( ) ( ).
If the remainder of external funds needed to finance the investment is done with issued equity.
This paper follows a strict interpretation of the pecking order theory, meaning that equity is never issued. Due to this reason the basic pecking order model in this paper is based on Equation (3). In the methodology section of this paper it is explained how this will be tested.
Extended pecking order model
In order to give the basic pecking order model a broader perspective it is adjusted by adding
Nominal GDP and EBITDA. These two variables will function as control variables. To understand
how this works an overview is first given of how cash flows through a firm. On the basis of that explanation new variables are added and finally combined with the basic model.
According to the pecking order theory a firm first finances its investment using internal resources, which can be defined as the cash balance (excluding debt) of the firm. It is important to understand the components, which make up the Cash Balance of a firm. This is reflected by the following equation:
. (5) where i indexes the firm and t represents the year.
The variables in Equation (5) are defined as follows:
is the amount of cash that a firm has at the beginning of the year;
is cash paid to shareholders.
this variable is the change in operational working capital.
11
o . (6) represents the cost accrued for holding debt.
is the firm’s operating income before depreciation and presents the opportunity to more easily compare various companies. Due to the fact that the capital structure of a company, nor matters such as taxes,
depreciation and amortization are not taken in to consideration. Having presented the equation for CashBalance, which is also found in the Equation (3), adjustments can be made to incorporate other variables as can be seen below.
The EBITDA can be broken down using this formula,
. (7)
Sales is the total sales of a company and EBITDAmargin is the operating profitability of a company,
which is assumed to be constant in this paper.
Viewing the change in the variable EBITDA in the period between it-1 and it is relevant for this model and is therefore represented by:
. (8) This paper assumes that any change in Sales, and therefore in EBITDA, is directly and exclusively correlated to a change in Nominal GDP. In the equation above, Nominal GDP is represented by where P denotes the price level and Y denotes the level of current real GDP. Taking into account the changes made above CashFlow is represented as:
. (9) To sum up the equation for internal funds, or in other words the CashBalance is:
. (10) For the purpose of this paper Dividends, Working Capital and Interest Costs are assumed to be
constant. This will give us the opportunity to better look at the influence that Nominal GDP and
EBITDA have on the financing needs of firms, which are according to this paper of essential
12 Finally, to sum up the extended pecking order model we combine Equation (3), which is the basic pecking order model, with Equation (9) where dividends, working capital and interest costs are constant. This gives us the following equation.
13
4. Data description and descriptive statistics
The data that is used to test the hypotheses is gathered from Thomson Financial DataStream. The hypotheses are tested using non-financial EUROSTOXX Companies. The reason that financial firms are not included in the data set is because this paper focuses on how firms finance their
investments which are financed by financial firms. The EUROSTOXX index was used because the focus of this paper is on Europe, as explained in the introduction.
In total the data set consists of 250 EUROSTOXX Companies with a time frame of 1985-2010. This time frame was chosen because it gives a broad window in which firms make investment- and financing-choices. The 250 companies are subdivided in 9 different industries which are Auto, Chemicals, Energy, Food, Health Care, Media, Retail, Technology and Construction. The data from each of the 9 industries is discussed separately. However, the main focus of this section is on the data set which includes all 9 industries together, in other words the aggregate data set. In addition to the company data, Nominal GDP data was obtained from the ECB. The figures found in the tables below represent Nominal GDP growth for the entire Euro area. All variables are represented by year-on-year percentage change. This makes comparison between firms easier, because it makes differences between firms irrelevant of the size of the firm.
The data series have been split into two different sets: one which includes outliers and another that excludes outliers. This split is necessary due to the large fluctuations that can be observed in data of certain companies, mainly as a result of the impact of acquisitions and disposals. Certain firms have an extremely large increase in EBITDA and other metrics due to these acquisitions and other companies have extremely large decreases in these same metrics due to certain company sections being sold off. Hence, two data sets will be used: the full set and one where the top and bottom 2.5% of the data is removed. The use of these two sets of data gives the possibility to get a more in-depth analysis of the regression done and more importantly gives the opportunity to test the robustness of the model used in this research.
14
Variables Investments Debt Cash EBITDA Nominal GDP
Dickey-Fuller Unit Root test: t-statistic (Prob.) -59,872 (0,0001) -5,831 (0,0000) -2,8478 (0,0329) -6,007 (0,0001) -3,619 (0,0135) Mean (%) 66,00% 818,40% 1798,50% 816,00% 2,10% Median (%) 5,90% 3,40% 6,90% 7,60% 2,40% Maximum (%) 158741,00% 1545238,0% 378340700,0% 2456125,00% 44,67% Minimum (%) -100,00% -100,00% -100,00% -100,00% -4,10% Std. Dev (%) 2619300,00% 28924,20% 66789,10% 41024,80% 1,60% Skewness 59984,00 47,24 53,15 57,88 -1,79 Kurtosis 3632,39 2370,26 2976,29 3440,55 7,17 Observations 3710 3876 3464 3738 5175 Table 1
Descriptive statistics including all 250 companies and outliers. Variables are represented by year on year percentage change.
When analyzing Table 1 it is interesting to see that it shows several extremes. The difference between the Maximum and Minimum values is very large, which also leads to a disparity between the Mean and the Median. The Skewness and Kurtosis hint at a significantly non-normal data set. This implies that, if this data are to be used in a regression, caution should be taken when
considering the validity of the outcomes. However, on the positive side using a Dickey-Fuller test, no unit root was found in any of the variables given that they are all significant at a level of at least 5%. A unit root test is used to investigate whether a variable is non-stationary.
The descriptive statistics of the data set with all companies, but excluding outliers, is less extreme as can be seen in Table 2.
Variables Investments Debt Cash EBITDA Nominal GDP
Dickey-Fuller unit Root test: t-statistic (Prob.) -49.429 (0.0001) -54.678 (0.0001) -54.838 (0.0001) -48.681 (0.0001) -3.619 (0.0135) Mean 10,20% 19,60% 27,80% 10,30% 2,10% Median 5,90% 3,40% 6,90% 7,60% 2,40% Maximum 110,60% 418,00% 569,90% 82,50% 4,47% Minimum -25,70% -61,10% -78,30% -26,60% -4,10% Std. Dev 19,50% 60,30% 83,20% 16,10% 1,60% Skewness 1,788 2,985 2,548 1,338 -1.791 Kurtosis 7,74 14,603 11,619 6,131 7.173 Observations 3532 3505 3287 3287 4089
The descriptive statistics of the data set including all 250 companies but excluding outliers meaning the top and bottom 2,5% of the data set. Variables are represented as year on year precentage change.
Table 2
15 Maximum and Minimum are therefore considerably smaller. The Mean and Median are also nearer to each other, indicating a more normal distribution of the data. This is also backed by the Skewness and Kurtosis which shows signs of a normal distribution. Overall this data set seems more suited for use in a regression than the set including outliers.
Comparing Table1 and Table 2 confirms that there is good reason to focus on the data excluding outliers, as this is more likely to give robust results. The use of the data including outliers is therefore used to check the robustness of the model and to make sure that no insights in the data are lost.
Having discussed the aggregate data sets it is important to now mention the descriptive statistics of the individual industries. It is a diverse group covering industries that in combination present a proxy for how the total of non-financial firms might behave regarding investment and financing dynamics. They form a mix of service/production oriented industries with different capital intensity for which EUROSTOXX has separate indices available.
When looking closely at these industries we expect to be better able to understand under what conditions the pecking order theory will hold. When looking at the descriptive statistics which can be found in Appendix 1 we see a similar picture as explained above. The data set including outliers is one of extremes showing that it is not normally distributed. The data excluding outliers does however show that the variables in each industry are relatively normally distributed. For all variables in both data sets no unit root was found.
16
5. Methodology
Having presented the data it is important to understand the methodology used in this paper. The main structure is a Pooled Least Square regression. Considering this paper uses panel data, the panel data equation method is used. This method differs from cross section and time series
methods. This difference can be found in the fact that cross sectional methods only regard multiple observations at one point in time, whereas time series method analyses how a single variable changes over time. The power of panel data is that it combines these two and looks at multiple variables over a stretch of time.
According to Hsiao (2006) the fact that panel data combine both cross-sectional and time series data is not the only advantage of this method. He explains that panel data model parameters are also subject to more accurate inference, come with additional sample variability and have more degrees of freedom. Also when testing more complicated behavioral hypotheses, panel data has a significant edge when compared to cross-sectional and time series data. Furthermore, Hsiao (2006) explains that panel data considerably decreases the impact of omitted variables. This is due to the fact that panel data contain information on intertemporal dynamics which helps control for unobserved and missing variables.
Regression
As explained before the two hypotheses relating to the basic pecking order theory are tested first. Thereafter, Nominal GDP and EBITDA are added in order to investigate the influence of changes in these variables on Investments. Moreover, a time-lag is added to control for the dynamic element. This is expected to lead to a better understanding of the interactions between the variables over multiple years.
17 In the two equations above the variables are defined accordingly:
Inv – The firm’s investments C - The constant factor
Cash – The firm’s cash balance Debt – The firm’s total debt EBITDA – The firm’s total EBITDA GDP – The nominal GDP of Europe ε – The error term
The Δt characterizes the fact that variables are represented by percentage change over time. The pecking order theory explains Debt or Cash as the dependent variable on the left hand side of the equation. In the basic and extended models used in this paper, Investment is the dependent variable, as this is the variable that is to be explained in relationship to financing decisions and credit requirements that are connected to economic growth. Further reasons for this choice are:
1. There is a causal managerial relationship between investment and related financing. A firm will first decide whether to invest in a project before it will subsequently look at the
possibility and best method to finance it.
2. When considering this issue from an accounting perspective a movement in assets (caused by the Investment), which takes place on the left side of the balance sheet, triggers a corresponding and equal change in the net debt (debt minus cash) on the right side of the balance sheet, or in the cash balance on the left side of the balance sheet if net debt is positive.
3. The Leary and Roberts (2010) model as presented in the theory section of this paper also describe Investment as their dependent variable.
18 When using these types of regressions there are certain matters that need to be taken into account in order to assure that the estimates of coefficients which are used to test the hypotheses are valid. The matters that need to be tested are heteroscedasticity, autocorrelations in the residuals and multicollinearity .
According to Brooks(2008) the use of a Pooled Ordinary Least Square regression might not be appropriate when heteroscedasticity and autocorrelations in the residuals are found in a
regression, as this might lead to hypotheses being tested using invalid results, which may lead to conclusions that are misleading. The presence of heteroscedasticity and autocorrelations in the residuals can be resolved by replacing a Pooled Ordinary Least square regression with a Generalized Least Squares regression.
Also when discussing the data it is imperative to point out that multicollinearity might be present. When multicollinearity is present it means that there is a large correlation between variables. This can then lead to invalid results. This paper has chosen to test for multicollinearity by using Variance Inflated Factors, which are presented together with the regressions in the results and discussion section.
Outliers:
As explained before this paper uses two data sets: one with outliers and one without. All
19
6. Results
This section presents the results of the regressions done to test the hypotheses as explained in the methodology section. These outcomes are presented in the following manner: Firstly the regression using Equation (12) is shown and analyzed. Secondly, the regression using Equation (13) is
analyzed. These two sections contain two data sets each: one data set including outliers and one excluding outliers. The data set excluding outliers can be found below and is discussed and analyzed in detail. At the end of this section an analyzes can be found concerning the data set including outliers. The results of those regressions can be found in Appendix 3.
Basic Pecking Order Model
In Table 3 the regression results of Equation (12), as explained in the methodology section, can be found. The regression includes investments as dependent variable. The independent variables are: investments including a lag, cash balance, cash balance including a lag, debt and debt including a lag. This can be used to better understand the workings of the pecking order theory and to test whether or not hypothesis 1 and 2 can be rejected.
Aggregate Auto Chemicals Energy Food Health Care Media Retail Technology Construction Investments(-1) 0,099 0,520 0,834 0,968 0,808 0,938 0,867 0,796 0,614 0,794 [4,155] [7,595] [7,595] [7,595] [7,595] [7,595] [7,595] [7,595] [7,595] [7,595] Investments(-2) 0,022 0,181 0,058 0,000 0,025 -0,111 -0,031 -0,028 0,112 0,037 [2,941] [-2,404] [2,051] Cash 0,018 0,011 -0,009 0,002 -0,003 -0,009 0,010 -0,002 0,003 -0,003 [3,516] [1,729] [-1,675] [-2,848] [3,147] Cash(-1) 0,008 -0,003 0,012 -0,001 0,001 0,007 -0,002 0,003 0,008 -0,009 [1,638] [2,013] [2,336] [-1,616] Cash(-2) 0,004 -0,003 0,001 -0,005 0,005 -0,010 0,003 0,005 -0,003 -0,006 [-3,447] Debt 0,124 0,050 -0,006 0,007 0,014 0,013 -0,003 0,024 0,033 0,017 [17,695] [3,166] [3,052] [2,601] [2,430] [2,683] [2,400] Debt(-1) -0,008 -0,002 0,003 -0,002 0,001 -0,011 0,000 0,003 0,017 -0,004 [-2,019] [1,325] Debt(-2) -0,001 -0,006 0,012 -0,012 -0,005 0,005 0,004 -0,001 0,009 -0,001 [-1,999] C 0,056 -0,003 -0,011 -0,017 -0,014 -0,008 -0,028 -0,001 -0,023 -0,011 [11,268] [-2,721] [-5,828] [-5,481] [-2,022] [-6,186] [-2,244] [-3,027] R-squared 0,173 0,516 0,762 0,970 0,894 0,906 0,788 0,708 0,630 0,927 Adj. R-squared 0,169 0,498 0,752 0,968 0,891 0,903 0,781 0,700 0,616 0,924 Sample (adjusted): 4087 423 425 275 598 811 756 621 632 508 Included observations: 1852 221 192 138 273 225 263 288 225 226
The PLS regressions using Equation (12). Investments is the dependent variable. Vertically the explanatory variables and horizontally the various industries being tested are shown. The table shows the coefficient and t-statistics being represented between [ ] with only the significant ones of at least 10% showing.
Table 3
Results Panel Least Square Regressions
20 The results of these tests can be found in Appendix 2. The conclusion is that heteroscedasticity is present, and hence this problem was corrected for by using White’s modified standard error estimates. No autocorrelation was found. Multicollinearity is present in certain industries.
What strikes most about the variables in Table 3 is that a number of them are significant in various regressions. These are Investment(-1), Cash, Cash(-1) and Debt. These are the variables which are important for understanding and validating the pecking order theory. These variables will be discussed one by one with the Aggregate regression leading the discussion and the individual industry giving an opportunity for further analysis.
In the first place we see that the variable Investment(-1) has a positive coefficient for all industries and is significant at a 1% level. To look at the Aggregate regression we see that it has a coefficient of 0,099 and a high t-statistic of 4,155 meaning that it is significant at a 1% level. This implies that when a company invests, it continues to do so in the following year. This is true for all the regressions. This does not directly have an influence on the pecking order theory but it is interesting to see that this phenomenon occurs, as it gives a better understanding of investment behavior.
When looking at the Cash in Table 3 we see that it has a coefficient of 0,018 with a t-statistic of 3,516 meaning it is significant at the 1% level. It is remarkable that the sign of this variable is positive which conflicts with hypothesis 1 stating that this relationship should be negative. This positive and significant relation is also found in the Auto and Media industry. This means that in these industries and in the Aggregate dataset, the cash balance increases when investments are done by a firm. This is the opposite of what the pecking order theory predicts, as it predicts that a firm will first use up all of its cash funds before it turns to external funding. Therefore, on this basis we reject hypothesis 1. However, when looking at the industries not yet discussed, we see that Chemicals and Health care industry have coefficients, which are negative with a significance of 10% and 1% respectively, implying a negative relationship between cash and investments. This shows that in these two industries the hypothesis 1 is not rejected. Also the Food and Retail industry have a negative coefficient. These are however not significant, but worth mentioning to get a better understanding of what is happening in these industries.
21
Cash(-1) is another variable which draws the attention. When first looking at the Aggregate
regression we see that it has a coefficient of 0,008 with a t-statistic of 1,638 meaning it is significant at a level of 10%. An increase of the cash balance leads to investments the next year. As you would expect with the pecking order theory the variable Cash in the next year would then have a negative coefficient as the investments would be paid by cash. However, this is not the case with the
Aggregate regression as explained above. This phenomenon can however be seen in the industries Chemicals and Healthcare, giving extra strength to the idea that the pecking order theory can explain a finance decision.
Another important variable when analyzing the pecking order theory is Debt. Once again the Aggregate regression will be analyzed first. When evaluating these results we see that Debt has a coefficient of 0,124 with a high t-statistic of 17,695 making it easily significant at 1%. This pattern of a positive coefficient with a significance level at 1% is true for 6 other industry regressions. Only the Chemicals, Energy and Media industry do not have a significant coefficient. However, for the rest of the regressions it means that when a firm in that industry is investing then debt is being used to finance this investment. This is in line with the pecking order theory and also with
Hypothesis 2. On the bases of the results just explained in the 6 industries and Aggregate regression Hypothesis 2 is not rejected.
Having discussed the variables of the various regressions and before summing up the findings it is important to also interpret the R2 of each regression, which represents the explanatory power of a
model. To begin with, we see that the Aggregate regression has a relatively low R2 of 0,173. This
however does not mean that the analyzed results are less significant. It means that other factors also play a role when looking at investments. It is interesting to see that the regressions of the individual industries do have noteworthy higher R2 with the lowest being 0,516 in the Auto
industry and highest being 0,970 in the Energy industry. This higher R2 means that the variables
included in the models are good indicators for investment behavior. However, caution is needed when considering such high R2 given that these can be misleading. In the first place it could be that
a unit root is present in the data which leads to an elevated R2. As can be seen in the data
description this is however not the case. Another important matter to consider is the influence of multicollinearity. As can be seen in Appendix 2 this is present in the industries Energy, Food, Health Care, Media and Construction. This could help explain the high R2 in these industries, considering
22 appears reasonable to assume that multicollinearity is part of the problem. Another explanation can be that the Health Care industry have a strong indicative power over investments in this regression. The reason for this being that almost all variables are significant at the 1% level.
The findings so far lead to mixed conclusions. Hypothesis 1 is rejected in some industries and in others it is not, meaning that we can only conclude that this hypothesis is industry related and further research would be needed to examine this more closely. When analyzing hypothesis 2 we see that in a total of 7 regressions this hypothesis is not rejected meaning that it can be concluded that in most industries debt is used to finance investments. When taking these findings into account and placing them in the light of the pecking order theory we see a mixed picture. Only in one
industry (Health Care) the pecking order theory holds. Here a negative relationship between cash and investments and a positive relationship between debt and investment were found.
This discovery is in line with what has been said before by the Leary and Roberts (2010) paper, namely that many papers are finding conflicting findings concerning the pecking order theory. If we limit ourselves to the findings above we would need to conclude that the pecking order theory can only explain financing decisions in a limited number of industries. This conclusion might change when looking at the next section where the pecking order theory is adjusted to also take into account control variables Nominal GDP and EBITDA.
Extended Pecking Order Model: including Nominal GDP and EBITDA
23
Aggregate Auto Chemicals Energy Food Health Care Media Retail Technology Construction
INVESTMENTS(-1) 0,069 0,469 0,799 0,870 0,808 0,924 0,880 0,741 0,627 0,801 [2,807] [6,509] [10,709] [10,312] [15,189] [16,270] [13,390] [13,121] [8,921] [16,120] INVESTMENTS(-2) -0,001 0,166 0,029 0,058 0,029 -0,126 -0,073 -0,008 0,126 0,039 [2,666] [-2,642] [1,953] CASH 0,010 0,010 -0,012 -0,001 -0,004 -0,009 0,012 -0,003 0,003 -0,008 [2,066] [-2,077] [-2,869] [3,297] [-1,714] CASH(-1) 0,008 -0,002 0,014 -0,001 0,004 0,008 -0,005 0,005 0,011 -0,006 [1,743] [2,287] [2,642] [-1,444] CASH(-2) 0,002 -0,004 0,004 -0,004 0,006 -0,010 0,003 0,006 -0,004 -0,004 [1,818] [-3,290] DEBT 0,099 0,044 -0,007 0,004 0,013 0,013 -0,001 0,015 0,031 0,000 [14,687] [2,852] [3,053] [2,517] [1,655] [2,402] DEBT(-1) -0,012 -0,001 0,008 0,006 0,001 -0,010 -0,006 0,002 0,015 -0,022 [-1,645] [-1,970] DEBT(-2) -0,004 -0,010 0,015 -0,002 -0,004 0,004 0,002 0,009 0,007 -0,025 EBITDA 0,392 0,167 0,093 0,017 -0,031 0,049 0,016 0,205 -0,046 -0,068 [13,852] [2,919] [2,361] [-1,628] [2,339] [4,714] EBITDA(-1) -0,013 0,002 -0,007 0,014 -0,001 -0,008 0,149 -0,089 0,008 0,153 [3,772] [-1,881] EBITDA(-2) -0,018 0,013 -0,035 -0,012 -0,002 -0,008 0,017 -0,003 -0,021 -0,003 NOMINAL GDP 0,402 0,206 0,046 -0,088 0,202 0,110 0,184 0,033 0,363 0,019 [2,046] [2,010] NOMINAL GDP(-1) 0,145 0,493 0,249 -0,126 0,270 0,033 -0,121 0,332 -0,141 0,006 [2,568] NOMINAL GDP(-2) -0,053 -0,736 0,390 0,128 0,104 0,258 0,123 0,075 0,144 -0,085 [-2,204] [1,990] [1,763] C 0,026 -0,007 -0,022 -0,015 -0,022 -0,014 -0,038 -0,017 -0,022 -0,008 [3,508] [-3,297] [-3,072] [-5,342] [-2,546] [-5,356] [-2,162] R-squared 0,270 0,574 0,762 0,968 0,899 0,908 0,793 0,757 0,620 0,927 Adj. R-squared 0,264 0,544 0,742 0,964 0,894 0,902 0,780 0,744 0,593 0,922 Sample (adjusted): 4087 423 425 275 598 811 756 621 632 508
Included observations after adj.: 1757 217 180 126 257 221 242 269 213 217
Table 4
Results Panel Least Square Regressions
24 As before the outcomes of Table 4 have been checked for heteroscedasticity, autocorrelation and multicollinearity. These results can be found in the Appendix 2. The conclusion is that
heteroscedasticity is present. This problem was corrected for by using White’s modified standard error estimates. No autocorrelation was found. Multicollinearity is present in certain industries. When analyzing Table 4 it is easy to identify the variables, which stand out in the regressions. These are just as before Investment, Cash, Cash(-1), Debt and now also Nominal GDP and EBITDA. At first glance the variables that were significant in the basic pecking order model as represented in Table 3 are also significant here, meaning Investment across all regressions have a positive coefficient and are significant at a 1% level coefficient, Cash has both positive and negative significant coefficients depending upon the industries analyzed, the same for Cash(-1) and Debt. There are however some changes among the regressions namely that the regression for the Construction industry gained a negative significant cash variable, although its Cash(-1) and Debt lost its significance. The Media regression gained a significant Cash(-2) with a negative coefficient. Also the Auto regression lost its significant Cash variable. This shows that these regressions are not very robust when looking at these specific variables. However, when looking at the Aggregate, Chemicals and Health Care we see that these variables still have the same sign and significance, showing robustness given that two variables have been added.
Having shortly discussed the new regressions compared to the basic pecking order theory it is important to take a close look at the added value of the two new variables namely Nominal GDP and
EBITDA . These are the variables that place this model into a broader macroeconomic perspective.
25 Just like the EBITDA variable, Nominal GDP also has an interesting story to tell. When looking at Table 4 we see that only in two regressions a significant coefficient can be found, namely in the Aggregate and Food industry. Both of these coefficients are positive and are significant at a 5% level. What is interesting to take note of is that apparently in all but the companies in the Food industry the economic climate does not have an influence on their investment decisions. This is strange as you would expect that in times of economic prosperity more investments would be done for a number of reasons. In times of economic growth more demand is to be expected and as
explained before this implies more production and hence more investments in production capacity. In addition, debt might be more readily available to fund these investments if necessary. In negative economic times it is expected that no investments are done, for the opposite reasons. On the basis of these considerations it is expected that a correlation would be present.
The absence of correlation between Nominal GDP and Investment in other industries is explained by the fact that the Aggregate regression does show a positive significant coefficient. This is
understandable given that combining all industry data together makes it possible to see the correlation between Nominal GDP and Investment. The correlation will therefore be more obvious due to the larger amount of firms found in the data. In industries themselves there might be some companies that go against the trend leading to no significant correlation. When however putting firms together these companies are lost in the crowd of companies that do have a correlation. The discussion concerning R2 is also present with Table 4 because they are significantly high once
more. The explanation for this phenomenon is the same as with the basic pecking order model. However, what is interesting to see is that all regressions stayed around the same high R2 except for
the Aggregate regression which increased by .10. This means that the extended pecking order model has a better explanatory power for investments than the basic model. This is good news for hypothesis 3, considering the Aggregate regression, this hypothesis is not rejected due to an significant increase in explanatory power. In the Auto and Retail industry hypothesis 3 is also not rejected however here the increase of R2 was less significantly high when compared to the
26
Aggregate Auto Chemicals Energy Food Health Care Media Retail Technology Construction
Investments(-1) Basic model +*** +*** +*** +*** +*** +*** +*** +*** +*** +***
Extended model +*** +*** +*** +*** +*** +*** +*** +*** +*** +***
Investments(-2) Basic model +*** -*** +**
Extended model +*** -*** +*
CASH Basic model +*** +*** -* -*** +***
Extended model +** +** -*** +*** -*
CASH(-1) Basic model +* +** +** -*
Extended model +* +** +***
CASH(-2) Basic model -***
Extended model +* -***
DEBT Basic model +*** +*** +*** +*** +** +*** +**
Extended model +*** +*** +*** +*** +* +**
DEBT(-1) Basic model -**
Extended model -* -**
DEBT(-2) Basic model -*
Extended model EBITDA Basic model
Extended model +*** +*** +*** +* +** +***
EBITDA(-1) Basic model
Extended model +*** +*
EBITDA(-2) Basic model Extended model
NOMINAL GDP Basic model
Extended model +** +**
NOMINAL GDP(-1) Basic model
Extended model +***
NOMINAL GDP(-2) Basic model
Extended model +** +* +*
C Basic model +*** -*** -*** -*** -*** -*** -** -***
Extended model +*** -*** -*** -*** -** -*** -**
R-squared Basic model 0,173 0,516 0,762 0,970 0,894 0,906 0,788 0,708 0,630 0,927
R-squared extended model 0,270 0,574 0,762 0,968 0,899 0,908 0,793 0,757 0,620 0,927
Table 5
Overview results / Comparison Table
27 The Aggregate regression in Table 5 shows that the Extended Pecking Order model is better at explaining financing decisions of firms than the Basic Pecking Order Model. Only the Cash variable is less significant in the extended model, which is compensated by a highly significant EBITDA and a significant Nominal GDP. That combined with the increased R2 makes the Extended Pecking Order
model a stronger model.
Data set including outliers
When comparing the results with those of the data set including outliers we see a similar picture (Tables are found in Appendix 3). The biggest difference is that Investment(-1) does not seem to be significant in any regression which is interesting when compared to the data set excluding outliers just discussed. Furthermore, we see that the Cash in all industries has a positive relation to
Investment which would mean that hypothesis 1 is rejected in all regressions. Debt also has a
positive coefficient in all industries and is significant in all but three industries. This means hypothesis 2 is not rejected when looking at the extended pecking order model. The same can be concluded from Table 3. The addition of the variables Nominal GDP and EBITDA has given the model more indicative power. However, the model was already very strong in the basic model with an R2 of .97. Not a great deal of validity can be associated with these results given that there are
many cases of extremely high multicollinearity and also the data set is not normally distributed. This has led to extremely high t-statistics and R2 in certain industries. Therefore, the function of the
28
7. Conclusion
This paper analyzes whether placing the pecking order theory into a broader macroeconomic perspective leads to a better understanding of financing decisions for investments made by
European firms. Firstly, an analysis was done of past research. This resulted in the conclusion that a great deal of conflicting findings surrounds the pecking order theory. The conflicting findings are explained by Leary and Roberts (2010) and the basic pecking order model used in this paper was therefore based on their model. Another finding concerning past literature is that some papers referred to the fact that hardly any research, which include macroeconomic variables, has been performed on the pecking order theory. Hence, in our extended pecking order model Nominal GDP and EBITDA are included to give the pecking order theory a broader perspective.
Using Pooled Least Square Regressions, the workings of the basic and extended pecking order models were investigated, using both an aggregate data set as well as individual industry data sets. These data sets were split in one data set including and one data set excluding outliers. The
outcomes of the basic pecking order model with the data set excluding outliers showed that only in the Health Care industry both hypothesis 1 and 2 are not rejected, meaning that a significant negative correlation was found between Investments and Cash and a significant positive correlation was found between Investments and Debt respectively. This indicates that only in this industry the pecking order theory holds. In the other industries and in the Aggregate data set, hypothesis 1 was rejected since a significant positive correlation was found between Investments and Cash, instead of the predicted negative correlation. However, in many industries and with the Aggregate data set hypothesis 2 was not rejected meaning that the pecking order theory is only partially consistent. When analyzing the results of the extended pecking order model, we see that the addition of
Nominal GDP and EBITDA in the Aggregate regression increases R2 by .10 and both added variables
are significant at 5% and 1% respectively. The addition of these two variables therefore does help to better understand the decision making process regarding financing of investments by European firms. This significant increase of R2 is not observed in the individual industries. This implies that
hypothesis 3 (which states that adding Nominal GDP and EBITDA to the basic pecking order model improves the explanatory power of the pecking order theory) is only true for the Aggregate data set and hence for the universe of non-financial European firms at large.
Taking the results of both models into account, this paper concludes that incorporating Nominal
29 financing decisions of non-financial European firms. This can possibly be of practical relevance for European governments in understanding the relationship between growing and funding of the European economy.
Limitations and relevant future research
When doing research there are always factors that lead to limitations in this research and suggestions that arise for further investigation. This thesis is no different, as it has a relatively narrow scope. This is due to the use of only one theory, combined with the fact that the variables
dividends, working capital, interest cost and EBITDAmargin are kept constant. Even though this has
given this research the opportunity for an in-depth analyzes, future research should include changes in these variables to fully understand the choices that firms make with respect to investment and financing decisions. Another limitation is that the pecking order theory is predominantly based on micro-economical considerations making it less suitable for macro-economic studies.
30 8.
References
Ariff, M., Shamsher, M., Taufiq, H., 2008. How capital structure adjusts dynamically during financial crisis. CARF Working paper, F130.
Brooks, C., 2008. Introductory Econometrics for Finance, Cambridge University Press.
Bharath, S., Pasquariello, P., WU, G., 2003. Does asymmetric information drive capital structure decisions?. Review of Financial studies 22, 3211-3243.
Chirinko, R., Singha, A., 2000. Testing static tradeoff against pecking order models of capital stucture: a critical moment. Journal of Financial Economics 58, 417-425.
Fama, E., French, K., 2005. Financing decisions: Who issues stock?. Journal of financial Economics 76, 549-582.
Frank, E., Goyal, V., 2003. Testing the pecking order theory of capital structure. Journal of Financial
Economics 67. 217-248.
Frank, E., Goyal, V., 2007. Tradeoff and Pecking Order Theories of debt. University of British
Columbia working paper.
Haan, de L., 1992. Solvabiliteit en pikordegedrag (Solvency and pecking order behavior),
Economisch Statistische Berichten, 77, 344-348.
Hsiao, C., 2006. Panel Data Analyzes – Advantages and Challenges, Test 16, 1-22.
Graham, J., Harvey, C., 2001. The theory and practice of corporate finance: Evidence from the field,
Journal of Financial Economics, 60, 187-243.
Jensen, M., Meckling, W., 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305-360.
Jung, K., Kim, Y., Stulz, R., 1996. Timing investment opportunities, managerial discretion and the security issue decision, Journal of Financial Economics 42, 159-185.
Korajczyk, R,A., Levy, A., 2001. Capital Structure Choice: Macroeconomic Conditions and Financial Constraints, Kellogg working paper, No. 279.
Leary, M., Roberts, M., 2010. The pecking order, debt capacity, and information asymmetry. Journal
of financial Economics 95. 332-355.
Lemmon, M., Zender, L., 2004. Debt capacity and tests of capital structure theories. Unpublished
31 Modigliani, F., and Miller, M., 1958. The cost of capital corporation finance, and the theory of
investment, American Economic review 48, 261-297.
Modigliani, F., and Miller, M., 1963. Corporate income taxes and the cost of capital: A correction,
American Economic review 53, 433-443.
Myers, S., 1984. The capital structure puzzle. Journal of finance 3, 575-592.
Myers, S., Maljuf, N.,1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13, 187-221.
O’Brien, R., 2007. A caution regarding rules of thumb for variance inflation factors, Quality &
Quantity 41, 673-690.
32
Appendix 1
Descriptive statistics for individual industries with industry and data set being indicated in the top left hand corner. Including meaning including outliers and excluding meaning excluding outliers, as explained in the data description section of this paper. The unit root test mentioned is a Dickey-Fuller Test. All variables are represented by year-on-year percentage change. This makes
comparison between firms easier because it makes differences between firms irrelevant of size.
Construction
(including) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -21,807 (0,000) -20,231 (0,000) -21,348 (0,000) -17,945 (0,000) -21.645 (0.000) Mean 18,84% 43,71% 34,82% 12,54% 2,06% Median 8,31% 7,09% 4,67% 7,79% 2,43% Maximum 569,43% 2455,53% 1187,03% 399,07% 4,47% Minimum -92,68% -99,76% -85,43% -48,58% -4,08% Std. Dev. 54,28% 219,16% 130,88% 30,09% 1,72% Skewness 6,374 8,474 6,316 6,814 -1,791 Kurtosis 55,584 85,590 49,283 76,680 7,173 Observations 428 390 426 428 525 Construction
(excluding) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
33
Autos (including) Investments Cash Debt EBITDA Nominal GDP
Unit Root Test:
t- statistic(Prob.) -19,004 (0,000) NA -18,393 (0,000) -19,327 (0,000) -20.455 (0.000) Mean 447,24% 13682,85% 102,67% 59,84% 2,06% Median 5,08% 6,61% 1,25% 6,60% 2,43% Maximum 158741,00% 3783407,00% 21527,85% 7411,60% 4,47% Minimum -100,00% -100,00% -100,00% -100,00% -4,08% Std. Dev. 8263,50% 208761,10% 1209,76% 574,79% 1,72% Skewness 19,127 17,581 16,329 11,200 -1,791 Kurtosis 366,886 316,695 281,968 128,751 7,173 Observations 369 338 356 374 425
Autos (excluding) Investments Cash Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -12.330 (0,000) -16,387 (0,000) -16.216(0.000) -15.537(0.000) -20.455 (0.000) Mean 7,82% 27,96% 16,95% 7,43% 2,06% Median 5,08% 6,61% 3,15% 6,60% 2,43% Maximum 109,28% 679,09% 352,03% 48,47% 4,47% Minimum -19,14% -78,54% -43,85% -26,13% -4,08% Std. Dev. 15,03% 92,55% 51,28% 11,93% 1,72% Skewness 2,048 3,643 3,485 0,419 -1,791 Kurtosis 10,871 21,812 18,233 4,071 7,173 Observations 351 320 354 356 425 Chemicals
(including) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -18,494 (0,000) -18,007 (0,000) -19,224 (0,000) -18,661 (0,000) -20.455 (0.000) Mean 27,40% 69,76% 22,97% 15,40% 2,06% Median 4,44% 4,22% 0,78% 4,15% 2,43% Maximum 5682,06% 9353,37% 1156,01% 2840,39% 4,47% Minimum -96,07% -98,97% -66,15% -85,61% -4,08% Std. Dev. 322,17% 567,90% 102,55% 157,86% 1,72% Skewness 16,233 14,387 6,450 16,776 -1,791 Kurtosis 278,079 227,852 57,716 297,382 7,173 Observations 348 321 334 348 425 Chemicals
(excluding) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
34
Energy (including) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -15,686 (0,000) -21,589 (0,000) -15,162 (0,000) -17,578 (0,000) -15.609 (0.000) Mean 22,34% 39,26% 32,64% 13,17% 2,06% Median 8,57% 9,19% 6,99% 9,67% 2,43% Maximum 1747,64% 1360,27% 1287,39% 229,37% 4,47% Minimum -68,08% -94,32% -91,31% -58,13% -4,08% Std. Dev. 118,04% 136,24% 122,95% 29,95% 1,72% Skewness 13,170 5,640 6,516 2,465 -1,791 Kurtosis 189,800 45,458 55,344 16,432 7,173 Observations 244 250 268 244 300
Energy (excluding) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -11.059 (0.000) -14.346 (0.000) -13.011 (0.000) -14.170 (0.000) -15.609 (0.000) Mean 12,19% 23,52% 18,90% 11,48% 2,06% Median 8,57% 9,19% 6,99% 9,67% 2,43% Maximum 75,59% 250,82% 245,60% 74,78% 4,47% Minimum -19,08% -60,95% -49,11% -34,83% -4,08% Std. Dev. 17,50% 58,90% 46,59% 20,29% 1,72% Skewness 1,085 1,450 2,334 0,639 -1,791 Kurtosis 4,546 5,507 9,952 3,904 7,173 Observations 230 236 254 230 300
Food (including) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -26,346 (0,000) -21,932 (0,000) -21,442 (0,000) -22,140 (0,000) -24.234 (0.000) Mean 52,81% 379,81% 500,58% 24,06% 2,06% Median 4,69% 5,85% 3,55% 7,22% 2,43% Maximum 11531,55% 949437500,00% 1545238,00% 253397300,00% 4,47% Minimum -100,00% -99,63% -100,00% -96,41% -4,08% Std. Dev. 651,80% 4923,50% 80025,40% 179,73% 1,72% Skewness 15,818 16,985 17,234 12,673 -1,791 Kurtosis 257,603 313,027 312,469 169,162 7,173 Observations 474 447 473 482 600
Food (excluding) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
35 Health Care
(including) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -15,675 (0,000) -19,548 (0,000) -25,308 (0,000) -21,078 (0,000) -23.050 (0.000) Mean 16,58% 147,74% 735,54% 21,35% 2,06% Median 8,03% 12,51% 4,81% 9,68% 2,43% Maximum 306,21% 14195,28% 108324,10% 2322,92% 4,47% Minimum -100,00% -100,00% -100,00% -88,32% -4,08% Std. Dev. 39,92% 975,64% 7871,97% 115,54% 1,72% Skewness 3,509 12,121 12,768 18,399 -1,791 Kurtosis 20,988 161,256 170,179 365,510 7,173 Observations 422 393 417 433 825 Health Care
(excluding) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -11.734 (0.000) -14.911 (0.000) -16.639 (0.000) -15.762 (0.000) -23.050 (0.000) Mean 13,04% 46,10% 23,81% 14,56% 2,06% Median 8,03% 12,51% 4,48% 9,68% 2,43% Maximum 108,48% 852,67% 750,10% 105,09% 4,47% Minimum -20,72% -76,64% -61,40% -35,93% -4,08% Std. Dev. 21,20% 121,84% 73,14% 19,20% 1,72% Skewness 1,771 3,077 4,311 1,752 -1,791 Kurtosis 6,883 15,069 32,672 7,887 7,173 Observations 400 371 396 413 825
Media (including) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -20,868 (0,000) -20,564 (0,000) -21,464 (0,000) -21,565 (0,000) -25.053 (0.000) Mean 9,93% 1967,02% 440,85% 5025,43% 2,06% Median 2,84% 3,58% 2,90% 6,43% 2,43% Maximum 1185,98% 767881,40% 144673,10% 2456125,00% 4,47% Minimum -100,00% -100,00% -100,00% -100,00% -4,08% Std. Dev. 63,94% 36042,00% 6605,50% 110842,80% 1,72% Skewness 13,480 21,120 21,268 22,090 -1,791 Kurtosis 239,401 449,194 463,988 488,994 7,173 Observations 487 456 492 491 775
Media (excluding) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
36 Health Care
(including) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -15,675 (0,000) -19,548 (0,000) -25,308 (0,000) -21,078 (0,000) -23.050 (0.000) Mean 16,58% 147,74% 735,54% 21,35% 2,06% Median 8,03% 12,51% 4,81% 9,68% 2,43% Maximum 306,21% 14195,28% 108324,10% 2322,92% 4,47% Minimum -100,00% -100,00% -100,00% -88,32% -4,08% Std. Dev. 39,92% 975,64% 7871,97% 115,54% 1,72% Skewness 3,509 12,121 12,768 18,399 -1,791 Kurtosis 20,988 161,256 170,179 365,510 7,173 Observations 422 393 417 433 825 Health Care
(excluding) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -11.734 (0.000) -14.911 (0.000) -16.639 (0.000) -15.762 (0.000) -23.050 (0.000) Mean 13,04% 46,10% 23,81% 14,56% 2,06% Median 8,03% 12,51% 4,48% 9,68% 2,43% Maximum 108,48% 852,67% 750,10% 105,09% 4,47% Minimum -20,72% -76,64% -61,40% -35,93% -4,08% Std. Dev. 21,20% 121,84% 73,14% 19,20% 1,72% Skewness 1,771 3,077 4,311 1,752 -1,791 Kurtosis 6,883 15,069 32,672 7,887 7,173 Observations 400 371 396 413 825
Media (including) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
t-statistic(Prob.) -20,868 (0,000) -20,564 (0,000) -21,464 (0,000) -21,565 (0,000) -25.053 (0.000) Mean 9,93% 1967,02% 440,85% 5025,43% 2,06% Median 2,84% 3,58% 2,90% 6,43% 2,43% Maximum 1185,98% 767881,40% 144673,10% 2456125,00% 4,47% Minimum -100,00% -100,00% -100,00% -100,00% -4,08% Std. Dev. 63,94% 36042,00% 6605,50% 110842,80% 1,72% Skewness 13,480 21,120 21,268 22,090 -1,791 Kurtosis 239,401 449,194 463,988 488,994 7,173 Observations 487 456 492 491 775
Media (excluding) Investments Cash Total Debt EBITDA Nominal GDP
Unit Root Test:
37
Appendix 2
Basic Pecking Order Model
The presence of heteroscedasticity in a data set can best be found using a White’s test, when looking at the table below you can find the results of such a test for each regression found in Table 3.
Aggregate Auto Chemicals Energy Food Health Care Media Retail Technology Construction
White's Test 5,341*** 5,652*** 1,473** 2,486*** 14,491*** 12,004*** 13,356*** 4,925*** 1,483*** 36,793***
Table 6
The results of a White's test for each regression presented in Table 3. The F-statistic is shown where *,**,*** indicates a significance of respectively 10%, 5%, 1%.
When looking at these results and at all the f-statistics we see that all but two are significant at a 1% level and the other two are significant at a 5% level. Knowing that heteroscedasticity is present when a significance level of 5% is measured suggests the existence of heteroscedasticity in each of the regressions. This problem was corrected for by using White’s modified standard error
estimates, these are heteroscedasticity consistent and will result in valid results.
It is important to also look at the presence of autocorrelation in the residuals. The test that can best be used to look for its incidence is a Durbin Watson test, in Table 7 the results for such a test can be found for each of the regressions found in Table 3.
Aggregate Auto Chemicals Energy Food Health Care Media Retail Technology Construction
Durban Watson 1,831 2,215 1,918 2,045 1,307 1,690 1,946 2,187 2,319 1,448
Table 7
The results of the Durbin Watson test for each regression presented in Table 3.
When analyzing the results of a Durbin Watson test it is important to determine for what particular value this test suggests no auto-correlation. According to Brooks (2006) the values need to be between 1.65 and 2.35 for no autocorrelation to be present. This is true for all but the Food and Construction industries for which the values say that the test is inconclusive. This means that no autocorrelation is found in the regressions, except for Food and Construction, for which the test is inconclusive.
