What is the effect of the financial crisis on the determinants of the capital structure choice of SMEs?

What is the effect of the financial crisis on the determinants of the capital structure choice of SMEs? Master Thesis presented to Tilburg School of Economics and Management Department of Finance by Apostolos-Arthouros Karampateas ANR 370636 Supervisor: Dr. Peter de Goeij 16th August 2013

ABSTRACT In this study the capital structure determinants of small and medium sized enterprises (SMEs) is analyzed on a sample of four European countries, which experienced severely and were mostly affected by the recent Financial Crisis: Greece, Italy, Portugal and Spain. The research is conducted for the period 2003-2011 in order to examine whether there are any differences in the determinants of the capital structure of entrepreneurial companies pre- and post-crisis. We apply panel data methods to the sample of firms for the period 2003 to 2011. We assess the extent to which the capital structure choice of a firm depends on its asset size, profitability, tangibility and growth of total assets. We also introduce some macroeconomic country-specific variables in our model for robustness test, such as corruption and unemployment rate. Our findings indicate that the SMEs in all countries exhibit similarities in their capital structure choices. Profitability and tangible assets have a negative relationship with level of total debt, whereas firm size and growth opportunities are positively related to their debt to assets ratio. Consequently, we find evidence that the composition of capital structure of SMEs is in line with the pecking order theory. In addition to that, we find that the Financial Crisis affected critically the importance of the determinants of capital structure decision for all firms with ROA exhibiting an increased importance regardless of the country of firm incorporation. 2

ACKNOWLEDGEMENTS I would like to express my appreciation and gratitude to my supervisor Dr. P.C. (Peter) de Goeij for his patience and enthusiasm and for providing me continuous support of my Master Thesis. His guidance helped me substantially in my research in order to complete this study. 3

TABLE OF CONTENTS 1. Introduction 6 2. Current State of Literature 7 2.1 Capital Structure Theories 7 2.2 Capital Structure and the Determinants of Leverage 9 2.3 Hypotheses Development 12 3. Data Description and Research Methodology 14 3.1 Data Description 14 3.2 Econometric Model 18 4. Empirical Results 18 4.1 Full Sample Estimates 20 4.2 Industry Effects 22 4.3 Robustness 27 5. Conclusion 31 6. References 32 7. Appendix 35 4

LIST OF TABLES Table 1: Capital structure theory and expected sign on leverage of the determinants Table 2: Descriptive Statistics for all firms with country panels Table 3: Leverage: Panel data regressions including crisis-dummy Table 4: Leverage: Panel data regressions with Industry fixed effects Table 5: Industry Effects: Dummy variable coefficients of average debt ratios for six industries. Table 6: Robustness: Panel data regressions including corruption and unemployment 5

1. INTRODUCTION The composition of capital structure is a crucial decision for every firm and has attracted the interest of many authors. There is a considerable amount of literature that investigates the factors that determine the level of total debt a firm chooses compared to other sources of financing. This study explores the impact of the recent financial crisis on the main determinants of capital structure in small and medium sized enterprises with a focus on the South European countries that experienced severely this crisis. There has been an impressive growth of SMEs the last years in Europe and these firms account now for a large and constantly increasing number of job positions. According to the European Commission report, 85% of net new jobs in the EU between 2002 and 2010 were created by SMEs. Along with the fact that SMEs represent the majority of firms in European countries, this reflects their importance for the current state of the economy. Despite that, access to capital is more difficult for small unlisted companies. The capital sources for SMEs are limited and information asymmetries are faced when they look for financing. The goal of this paper is to contribute to the literature regarding the capital structure choice of SMEs. The economic and financial environment has altered substantially the last years and we expect that this has an impact on the factors that determine the composition of capital structure of SMEs. To our knowledge, this is the first study that looks at this particular time period and investigates the effect of the crisis. The main questions that are addressed are the following: 1) What is the composition of SMEs capital structure during the recent financial crisis? 2) Is the composition of SMEs capital structure different than before the financial crisis? Moreover we focus on cross country differences. So, we also address the following questions: Are the capital structure determinants of SMEs in the aforementioned countries driven by similar factors? Are potential differences driven by country-specific or industry-specific factors? To perform our study, we employ a large dataset from Greek, Italian, Portuguese and Spanish SMEs and test whether the predictions of traditional capital structure theories apply for all countries during the examined period. Cross-country differences are expected due to different law system and regulations, entrepreneurship facilitations, bureaucracy and dissimilar costs and benefits that firms face in each country. 6

Although many studies exist on the determinants of leverage of SMEs, to our knowledge, no study has examined the impact of the financial crisis on the capital structure. We aim to provide evidence regarding the capital structure theory that becomes more important under difficult circumstances and in periods of recession. In this paper we study the factors that determine the composition of capital structure of SMEs in Greece, Italy, Portugal and Spain and investigate the role of both industry and country characteristics. We choose to do our research on these countries because we expect at least here to be some effect of the crisis and they have been very popular in the press the last years. Firms incorporated in Ireland are not in our sample due to limited observations and data availability. Our aim is to find whether the financial crisis has led to significant differences in the factors that affect the level of leverage of a firm. For this reason, we divide our sample into two sub-periods, the first period from 2003-2007 and the second from 2008-2011. We want also to compare the determinants across the four countries and between industries and investigate whether the same variables determine the composition of capital structure in every country. We find that there is a significant negative effect of the financial crisis on the amount of total debt SMEs choose on average. More specifically, a considerable decrease is observed in the short term debt ratio which indicates that SMEs alter their capital structure decisions due to the crisis. This result holds for all countries although our findings report differences concerning the importance of the determinants and the change of their effect on the ultimate capital structure decision. ROA appears to have significantly higher importance after the financial crisis for all countries. The remainder of this paper is organized as follows: In Section 2, we review the literature and formulate hypotheses. Section 3 presents the data as well as the applied panel data methodology. The empirical results are discussed in Section 4 and robustness tests are performed. Finally, Section 5 concludes. 2. CURRENT STATE OF LITERATURE 2.1 Capital structure theories When we refer to the capital structure of a firm, we mean the composition of the capital owned by the firm and used for its operations and activities. More specifically, a firm can distinguish between two general ways of financing: equity financing and debt financing. Each way of financing has its pros and cons and thus the firm has to consider carefully the optimal 7

capital structure. It is a critical decision for the company as it might determine the future opportunities as well as the growth and profit potential. Much research has been conducted about the capital structure of entrepreneurial firms. As a result, there are some theories that have prevailed in the literature. Modigliani and Miller (1958) were the first that examined the determinants of capital structure. They derived the first theorem that set the basis for further research. According to Modigliani-Miller Theorem, the value of the firm is not affected by the proportion of debt and equity it holds, under some restrictions (e.g. absence of taxes, agency costs etc.) that lead to an efficient market. Their Theorem is often called as the capital structure irrelevance principle. Modigliani and Miller (1958) concluded that a firm cannot increase its value by using leverage as part of its capital structure. This outcome has provided the base with which to study whether capital structure is relevant, that is, a company's value is affected by the capital structure it employs and try to find the factors the determine the leverage of a company. As we mentioned, the MM theorem applies to a perfect market, but the economy has many imperfections which lead to an association between capital structure and the value of a firm. Quite a lot of theories have appeared to explain capital structure in a more realistic world. The first theory for leverage is the Trade-off Theory: capital structure is chosen in such a way that the tax and incentive advantages of debt exactly offset bankruptcy costs at the margin. Kraus and Litzenberger (1972) were the first that introduced the theory. This theory argues that a firm has a target leverage ratio, the level of which is determined by the trade-off between the costs and benefits of debt financing. According to this theory, there is an optimal level of debt which occurs when the marginal benefit equals the marginal cost of an additional unit of debt. This means that once the firm reaches the target, it issues, retires, and repurchases debt or equity to keep its leverage at the optimal level. Franck and Goyal (2009) find evidence consistent with the trade-off theory. An alternative theory of capital structure is the Pecking Order Theory. It was first suggested by Donaldson (1961) and modified my Myers and Majluf (1984). The main prediction of this theory is that there is a hierarchy of financing sources. It suggests that companies prefer to use internal equity to pay dividends and finance new investment. Hence, firms prefer to use retained earnings as their first financing source, followed by debt and, lastly, by equity because of higher costs and effort required due to asymmetric information. The pecking order theory predicts that firms that have historically been more profitable will end up with lower leverage because they had not needed to issue securities to finance investments. This means that firms with more available 8

internal funds should use less external funding. But if external financing is needed, debt is preferred compared to equity. Both the trade-off and the pecking order theories suggest that the capital structure of a firm should be tailored to the characteristics of the firm s assets. For example, profitable firms with stable cash flows should have high leverage, since they are better at utilizing debt tax shields and have lower probabilities of financial distress. Furthermore, costs of financial distress are likely to be higher for firms with more investment opportunities and more intangible assets. Another theory that can explain the relevance of the capital structure of a firm is the Agency Cost Theory presented by Jensen and Meckling (1976). According to this corporate governance theory, there are three types of agency costs between the owner (shareholders) and the manager of the firm: asset substitution, debt overhang and free cash flows. The agency cost theory presents debt as a governance device useful in diminishing this conflict because it reduces the amount of free cash flows available to managers and imposes financial discipline. The threat of default when the firm is highly leveraged prevents managers of taking wasteful decisions by forcing them to maximize firm s value. Asset substitution effect and debt overhang problem relate capital structure with managers incentives to undertake negative and positive NPV projects respectively. To summarize, capital structure is considered a very important part of the field of corporate finance. In spite of the fact that a lot of research has been done so far, each study contributes to our understanding of the determinants of leverage covering different periods and examining different companies and countries. Because SMEs are usually privately held, the most relevant theories for this study are the trade-off theory and the pecking-order theory. Therefore our empirical tests will focus on those two theories. The predictions regarding the sign of the expected effect of the explanatory variables on leverage are summarized in the next table for both theories. Table 1 : Capital structure theory and expected sign on leverage of the determinants. Pecking order theory Trade-off theory Firm size + + ROA + Tangibility + + Growth + 9

2.2 Capital structure and the determinants of leverage There have been many studies recently about the determinants of the composition of capital structure of small and medium sized enterprises that test the above-mentioned theories. See for example Hall et al. 2004, Heyman et al. 2008, Degryse et al. 2012, Mateev et al. 2013. Those studies have resulted in some variables that are considered significant as determinants of the capital structure. We present in short some relevant studies and their findings. To begin with, the most noticeable determinant of capital structure which is often used in empirical research is profitability. Profits seem to be linked with the level of leverage chosen by a company but there are contradicting views about the sign of their connection. For instance, Jensen (1986) claims with his free cash flow theory, that more debt should be used if the firm experiences an increase in its profits. Hence, he expects a positive relationship between profitability and leverage. On the other hand, the opposite effect is predicted by the peckingorder theory. Regarding SMEs, most relevant is the pecking-order relationship because for most of them there is no considerable agency conflict between managers and shareholders and thus overinvestment is unlikely to occur. Empirically, SMEs studies indeed support the pecking-order prediction of a negative relationship between profits and leverage. Degryse et al. (2012) find evidence that document this negative relationship. Also Mateev et al. (2013) test SMEs from seven countries and find a significant negative correlation between profitability and leverage. Similar are the findings of Myers (1984), Sogorb-Mira (2005), Heyman et al. (2008), Van der Wijst and Thurik (1993), Michaelas et al. (1999). Second, a factor that is considered as a determinant of the level of debt a firm chooses is its size. As Bhaduri (2002) mentions, large firms are less likely to face financial problems because they tend to be more diversified. Generally, most of the SME literature (Mateev et al., 2013, Artic et al., 2011, Beck et al., 2008, Frank and Goyal, 2009) finds evidence that small and medium firms face higher barriers to external financing than large firms do. Consequently, a positive association between firm's size and leverage is expected. Sogorb-Mira findings for Spanish SMEs and Voulgaris et al. (2007) for Greek firms are also in line with this statement. Another important factor that determines capital structure is the firm s asset structure, meaning the proportion of tangible assets that is employed by a company. Tangibility is defined as the ratio of tangible assets over the total assets of the firm and is assumed to be positively correlated with debt because tangible assets are easy to collateralize and so they reduce the agency costs of debt. Rajan and Zingales (1995) and Heyman et al. (2008) find evidence of this positive relationship. In general, firms seek to follow the maturity-matching principle, according to which 10

firms aim to finance current assets with short-term financing and fixed assets with long-term financing. Therefore, under the trade-off theory, a positive relationship between collateral and leverage is predicted. Mateev et al. (2013) find that long-term debt is positively correlated with asset structure, whereas this relationship becomes negative when it comes to short-term debt. Sogorb-Mira (2005) and Heyman et al. (2008) also argue that firms behave according to the maturity-matching principle. Furthermore, the pecking-order theory states that a positive effect of growth on the level of leverage should be expected. Growth is defined as the difference in total assets in two subsequent years divided by the total assets of the first year. Sogorb-Mira (2005) and Mateev et al. (2013) find evidence that support the pecking-order theory. But again there is an opposing finding in the literature. Fama and French (2002) and Heyman et al. (2008) find a negative relationship between asset growth and leverage. They explain their findings with the argument that growing firms opt to retain the level of their leverage low so that they can have a low risk debt capacity in the future. There are also some studies that investigated whether some macroeconomic factors might affect firms capital structure. Hanousek and Shamshur (2011) have investigated whether there is any relationship between leverage ratios and the economic environment by using data from European emerging markets. They concluded that the leverage ratios remain relatively stable and this stability is unrelated with changes in the economy due to credit constraints. They claim that stability relies on an unobserved time-invariant firm-specific factor. Similar are the findings of Lemmon et al. (2008) which indicate that corporate capital structures do not vary a lot over long periods of time. Recent research has incorporated country-level characteristics as determinants of capital structure. Hall et al. (2004) perform a cross-country study and show that there are variations in both SME capital structure and the determinants of capital structure between the countries surveyed. Gungoraydinoglu and Öztekin (2011) analyze the capital structure determinants using data from all non-financial firms for 37 countries. They find strong support for both the trade-off and the pecking-order theory between the country-level determinants and leverage and establish a new association between country-specific variables and the traditional theories. McClure et al. (1999) find that firms' capital structures are still significantly different by nationality for the G7 countries. A possible interpretation for these differences in the level of leverage lies in macroeconomic factors such as economic growth, interest rates and inflation. Daskalakis and Psillaki (2008) focus on SMEs incorporated in France and Greece and find that there exist both some similarities and some differences in the sign and the magnitude of the coefficients of the 11

capital structure determinants between the two countries. Profitability has a negative relationship with leverage whereas size is positively related for both countries. On the other hand, growth has a significant positive relationship with the level of debt only for French companies. Michaelas et al. (1999) conduct an investigation about the determinants of SMEs capital structure in UK. Their results indicate that pecking-order theory and trade-off theory apply to their dataset. Moreover, they provide evidence that the capital structure of small firms is industry and time dependent. Both total level and maturity of debt are influenced by industry-specific effects. Brav (2009) examines public and private firms in UK from 1993 to 2003. He finds that private firms have significantly higher leverage ratios than their public counterparts and that there are differences in the maturity structure of the debt. These differences are attributed to the higher cost of equity financing that private firms face. Kalemli et al. (2011) study bank and firm leverage patterns before and after the financial crisis. Their results indicate that leverage ratios increased aggressively before the crisis for the whole sample and only banks incorporated in emerging markets were able to maintain around the same level during the crisis. In spite of focusing on banks, this study is related to our research because they find that leverage was lowered for the average firm due to higher cost of external financing and decreasing collateral. These findings demonstrate that disproportionate risk taking before the crisis was not easily detectable since the risk involved more the quality rather than the amount of investment. Kisgen (2006) investigates the effect of a country s financial creditworthiness on leverage and shows that credit ratings and potential for an upgrade as well as a downgrade directly affect capital structure decisions made by managers. 2.3 Hypotheses Development In this study, we consider four factors as the main proxies for the analysis of the level of leverage of SMEs: firm size, profitability, tangibility and growth of total assets. Based on these proxies we formulate our hypotheses. In addition to leverage, we look at the effect of the determinants on the long-term debt ratio and the short-term debt ratio. We also add two macroeconomic factors as explanatory variables for robustness test: the level of corruption of a country and the unemployment rate. Because of the fact that we are investigating the capital structure of SMEs, most of the hypotheses we develop are in line with the pecking order theory which is considered the most appropriate theory for this case. Under the pecking order theory, firm size is assumed to have a positive relationship with leverage. The rationale behind this idea is that large firms can afford 12

high levels of leverage because they have easier access to borrowing and thus do not face the danger of insolvency at the level smaller firms do. Additionally, larger firms are more diversified which results in a mitigation of the information asymmetry problems and a decrease in the cost of debt relative to other financial sources. Trade-off theory also argues that a positive connection between a firm s size and its level of debt should be expected. Therefore: H1: Larger firms have higher leverage. There have been found significant differences in the effect that size has on long-term debt and short-term debt. Many studies indicate a clear positive relationship between size and long-term debt but there are variations concerning the relationship between size and short-term debt. Therefore, we support the first hypothesis with the following sub-hypothesis: H1a: Larger firms have more long-term debt. According to the pecking order theory, firms that are more profitable use less debt as a source of financing. This happens because debt financing is used only when internal financing sources such as operational profits are depleted. Since the firm is profitable, generated internal financing is preferred so, as profitability rises, the proportion of debt financing is reduced. H2: Profitability has a negative effect on leverage. Next, we form a hypothesis for tangibility. Both the trade-off and the pecking order theory suggest that tangible assets have a positive effect on the amount of debt employed by the firm. Tangible assets are used as collateral and mitigate the agency problems with debtholders. Besides that, they reduce also the information asymmetry problem which is mostly relevant for the SMEs as their operations are not as transparent as those of large companies. Consequently, our hypothesis regarding collateral is consistent with the one Frank and Goyal (2009) developed: H3: Tangible assets have a positive effect on leverage. A positive relationship between growth and level of debt chosen is predicted by the pecking order theory. A firm with bright growth opportunities is more likely to be able to meet its obligations and thus can raise more debt. In this research a common variable will be used as a proxy for growth opportunities which is the growth of total assets of the company. So, our hypothesis is: 13

H4: Growth has a positive effect on leverage. Apart from the hypotheses connected with specific determinants of leverage, we are interested to investigate the link between the financial crisis and the composition of capital structure of the firms. This is actually the most important part of the paper. Apparently, the economy of the countries examined in this research suffered a deep recession after 2007. We expect to see a difference in the capital structure choice of the typical firm towards being less levered. This decrease in debt should mostly be due to a decrease in short-term debt because firms are expected to avoid risky short-term financing during periods of high uncertainty. So, the purpose of this paper is to give an answer to the following research question: Composition of capital structure of SMEs during the crisis is different than before. Our intention here is to perform a cross-country study and to identify the potential differences in the capital structure choice across countries. The macroeconomic environment is different in each country and thus we expect it to affect in a different way the optimal capital structure of the firms incorporated in a specific country. As proxies for the macroeconomic environment we use the unemployment rate and the corruption perception index. Subsequently, we expect that country effects play an important role in affecting the determinants of capital structure and we aim to investigate the next research question: The determinants of leverage vary across countries before and during the crisis. The above research question implies that the impact of the financial crisis is different among countries. We will also investigate empirically the importance of industry effects to address the question to what extend capital structure difference between firms is caused by industry characteristics rather than country characteristics. 3. DATA DESCRIPTION AND RESEARCH METHODOLOGY 3.1 Data description The data collected in order to conduct the empirical research are provided by the Orbis Bureau van Dijk database. The database contains detailed financial information for European companies. We selected all non-financial companies originated in Greece, Italy, Portugal and Spain for the period 2003-2011. For all companies the following variables were downloaded: NACE, Total Assets, Number of Employees, Operating Revenue, Long-Term debt, Current Liabilities, ROA, Tangible Fixed Assets and Other Fixed Assets. NACE is used to distinguish 14

the industry a firm belongs to so that industry differences can be considered. Companies with missing data in one of the aforementioned variables are eliminated only for the years with missing values so that our sample is not biased. The same accounts for companies with negative values in Total Assets, Long-Term debt, Current Liabilities and Tangible Fixed Assets. We adjusted our dataset in order to be appropriate for a study on small and medium enterprises. Observations that did not meet the SME criteria as specified by the definition of the European Commission were dropped. According to this definition, as SMEs are considered the companies that satisfy the next three criteria: 50 < Employees <250 10 mln < Operating revenue (Turnover) < 50 mln 10 mln < Balance sheet total (Total assets) < 43 mln Once these criteria are implemented in the dataset, we ended up with unbalanced panel data. We created the variables Size, Leverage, Long-term debt ratio, Short-term debt ratio, Tangibility and Growth, which are necessary for the research. Information about the definition of the variables is provided in the Appendix. In addition to that, data for the yearly Unemployment Rate per country is obtained from Datastream. The Corruption Perception Index is obtained from Transparency International as a measure of corruption in each country. The Index takes values from 0 to 100, with 0 being highly corrupt and 100 indicating no corruption at all. Table 2 shows descriptive summary statistics per country for all variables. The time period is divided into two sub-periods using a dummy variable: 2003-2007 and 2008-2011. These subperiods are introduced for crisis effect investigation. By construction, missing variables are created for the first year for growth and as a consequence year 2003 is not included in table 2. As we see in the table, there is a considerable increase in the number of observations concerning Portuguese firms in the second sub-period in comparison to the first sub-period although both periods include 4 years of observations. NACE number is used to classify the firms of our sample into industries. Following Orbis clarification of NACE numbers, six industry dummies are created: one for each of the six different industries. The six industries are energy supply, manufacturing, construction, wholesale and retail trade, transport and services. 15

Table 2 : Descriptive Statistics The table presents the variable mean, standard deviation, maximum, minimum values and number of observations for all firms categorized by country. The table consists of 4 panels, one for each country. All data are divided into the two sub-periods examined. Panel A: Greece 2004-2007 2008-2011 Variable Obs. Mean Std. Dev. Min Max Obs. Mean Std. Dev. Min Max Total debt 12248 0.6346 0.2346 0.02 4.06 13776 0.6296 0.2539 0.01 3.12 Long-term debt 12248 0.0763 0.1267 0.00 1.42 13776 0.1137 0.1538 0.00 1.49 Short-term debt 12248 0.5583 0.2433 0.01 4.06 13776 0.5159 0.2548 0.00 3.03 Size 12248 8.8218 0.6920 7.60 10.67 13776 8.8802 0.7169 7.60 10.67 ROA 12248 3.7971 7.5832-86.29 80.85 13776 1.3806 8.5564-94.09 89.42 Tangibility 12248 0.2540 0.2100 0.00 0.98 13776 0.2676 0.2209 0.00 0.98 Growth 12248 0.1141 0.2378-0.72 5.94 13776 0.0197 0.1964-0.80 2.55 Panel B: Italy Total debt 88689 0.6448 0.2066 0.00 2.90 97137 0.6046 0.2204 0.00 3.04 Long-term debt 88689 0.0717 0.1038 0.00 1.01 97137 0.0754 0.1035 0.00 1.21 Short-term debt 88689 0.5731 0.2057 0.00 2.65 97137 0.5292 0.2157 0.00 3.04 Size 88689 8.9617 0.6983 7.60 10.67 97137 9.0074 0.7294 7.60 10.67 ROA 88689 1.9628 6.0750-98.26 96.62 97137 1.2977 7.1240-99.14 95.33 Tangibility 88689 0.1987 0.1805 0.00 0.99 97137 0.2320 0.2058 0.00 0.99 Growth 88669 0.0816 0.2283-0.94 9.84 97137 0.0574 0.2333-0.86 7.23 Panel C: Portugal Total debt 3092 0.6381 0.1909 0.03 1.48 10647 0.6236 0.2121 0.01 2.63 Long-term debt 3092 0.1418 0.1567 0.00 1.11 10647 0.1606 0.1594 0.00 1.97 Short-term debt 3092 0.4962 0.1985 0.01 1.43 10647 0.4631 0.2063 0.01 1.87 Size 3092 8.8005 0.7064 7.60 10.65 10647 8.7910 0.7122 7.60 10.67 ROA 3092 3.0264 5.8275-46.30 67.05 10647 1.8241 7.2988-93.95 66.89 Tangibility 3092 0.2580 0.1881 0.00 0.97 10647 0.2555 0.1893 0.00 0.99 Growth 3092 0.0940 0.2019-0.56 1.88 10647 0.0468 0.2081-0.73 2.74 Panel D: Spain Total debt 47510 0.5916 0.2418 0.01 8.20 48898 0.5395 0.2620 0.00 6.07 Long-term debt 47510 0.0972 0.1512 0.00 7.94 48898 0.1138 0.1636 0.00 5.11 Short-term debt 47510 0.4944 0.2181 0.01 2.81 48898 0.4257 0.2289 0.00 2.78 Size 47510 8.9323 0.7094 7.60 10.67 48898 8.9631 0.7180 7.60 10.67 ROA 47510 4.9928 7.3532-95.49 98.22 48898 1.9660 8.6325-99.60 95.01 Tangibility 47510 0.2182 0.1924 0.00 0.99 48898 0.2356 0.2031 0.00 1.00 Growth 47510 0.1191 0.2714-0.82 12.03 48898 0.0053 0.2230-0.84 10.49 16

Table 2 provides the number of observations for each country in the two sub-periods. Obviously, the number of companies is different due to repeated observations of the same firms over the years. In all countries we have slightly more unique SMEs in the second sub-period: 3582 in the first period and 4479 in the second period for Greece, 30651 and 32286 for Italy, 2570 and 3163 for Portugal and 14141 and 14212 for Spain. Some conclusions about our data can be made from the above table. Firstly, it is obvious that in all countries of our sample, firms face a sharp decline in their profitability. This occurs especially after 2007 as can be seen in the line graphs provided in the appendix. As expected, the effect is larger for the average Spanish and Greek firms where the crisis is deeper and hit earlier. Similar are the findings for growth. We notice a decrease in the average growth of total assets in all countries and again the largest effect is observed in Spain and Greece. More specifically, we find that average growth is negative after 2009 both for Spain and Greece which clearly shows the harsh repercussions of the crisis on the prospects and growing of SMEs. As far as the level of total leverage is concerned, a small but steady decline in the mean leverage is observed over time. This is attributed more to a change in short-term debt than in long-term debt which remains relatively stable. Our interpretation is that firms cannot adjust their long-term debt levels so quickly to the new economic environment because they are long-term liabilities. Therefore the total usage of debt as a financing source is decreased by decreasing the short-term debt levels. In addition to the cross-country comparison, we want also to compare the behavior of the top 10% and worse 10% firms in each country. ROA is used as an indication of the performance of a firm. We calculated the total debt, long term debt and short term debt ratios for the best and worst 10% and the results are presented in the graphs in the appendix. As can be observed, the firms with better performance and higher profitability opt for less debt compared to the average firm throughout the entire time period examined. The opposite is true for the worst firms which have a larger proportion of debt included in their financing sources. This indicates that less levered firms have higher chances to survive the crisis and be more profitable. In other words, a high level of debt constricts the firm s viability. This finding holds also if long term and short term debt ratios are considered. Especially for short term debt, we observe a difference of approximately 20% between the best and worst firms. More specifically, there is a further considerable decline in the short term debt level of the best performing firms after 2008 in all countries. This fact suggests that these firms, in order to keep being profitable and survive in a competitive and financially uncertain environment choose to employ even less debt and rely more on internal financing sources. 17

3.2 Econometric model We employ a panel data analysis in each of the four countries samples to test the capital structure theories and examine the effect of the crisis as we have an unbalanced data set with observations over several years. We run all regressions for three different dependent variables: total debt ratio, long-term debt ratio and short-term debt ratio. The independent variables included are the size of the firm, profitability, tangibility, growth as well as a crisis-dummy which takes the value of 0 for all years for the period 2003-2007 and the value of 1 for the period 2008-2011. The crisis-dummy is multiplied with all the dependent variables leading to four additional dependent variables in the regressions. Because we have panel data we can use either random or fixed effects regression. We use a Hausman test in order to choose between them. In general, the Hausman test tests the null hypothesis that the coefficients estimated by the efficient random effects estimator are the same as the ones estimated by the consistent fixed effects estimator. In our case, the value of the probability is very low which indicates a significant p-value and leads to the conclusion that fixed effects should be used. The general regression model can be written as y it = β 0 + β i x it + α i + ε it i= 1,,N and t=1,..,9 years where y it is the dependent variable, x it the aforementioned independent variables, ε it the error term, α i measures the firm effect, β 0 is the constant term and β i the slope coefficients. The regression estimates the impact of the four determinants on the capital structure choice of the firms as well as the effect of the crisis and whether the importance of the determinants has changed significantly after 2008. We apply the above regression to each country separately in order to examine country-specific differences. Additionally, we test differences across industries for every country. Again the same determinants of leverage are used but industry fixed effects are added. The regression model is the same as before but now the factor α i measures the industry effect. 4. EMPIRICAL RESULTS Section 4.1 discusses the main results of our regressions using all our data but divided in four datasets by country. For every firm characteristic the corresponding hypotheses are reviewed. In section 4.2 results of the industry effects are discussed. Section 4.3 provides the results of our robustness check, which is done by including additional variables and testing the difference in the significance of the coefficients.

Table 3: Leverage: Panel data regressions including crisis-dummy Total debt Size 0.0857* (10.67) GREECE ITALY PORTUGAL SPAIN Long Short Total Long Short Total Long Short Total Long term debt term debt debt term debt term debt debt term debt term debt debt term debt 0.0772* 0.0085 0.0525* 0.0407* 0.0118* 0.0286* 0.0496* -0.0211 0.0687* 0.0416* (15.10) (1.03) (21.23) (27.13) (4.75) (2.54) (5.02) (-1.88) (18.28) (13.48) Short term debt 0.0270* (7.88) ROA -0.0026* (-6.18) -0.0001 (-0.52) -0.0025* (-6.38) -0.0039* (-27.79) -0.0006* (-9.74) -0.0033* (-23.38) -0.0043* (-8.69) -0.0013* (-3.17) -0.0029* (-5.70) -0.0029* (-13.45) -0.0010* (-5.60) -0.0019* (-10.46) Tangibility -0.0802* (-4.03) 0.1710* (10.20) -0.2510* (-12.07) -0.2280* (-29.37) 0.1560* (27.57) -0.3830* (-52.90) -0.0604 (-1.79) 0.1700* (6.65) -0.2310* (-7.06) 0.0147 (1.56) 0.1680* (17.49) -0.1540* (-16.37) Growth 0.0418* (6.56) -0.0019 (-0.42) 0.0437* (6.08) 0.0422* (12.66) -0.0090* (-6.61) 0.0512* (14.54) 0.0502* (5.40) -0.0233 (-1.88) 0.0734* (5.49) 0.0610* (12.67) 0.0088* (3.30) 0.0522* (13.49) Constant -0.0867 (-1.25) -0.6430* (-14.46) 0.5560* (7.72) 0.2320* (10.56) -0.3200* (-23.88) 0.5510* (25.13) 0.4160* (4.10) -0.3300* (-3.82) 0.7460* (7.45) -0.0139 (-0.42) -0.3070* (-11.18) 0.2930* (9.53) Crisis-dummy -0.0760* (-2.97) 0.0610* (2.79) -0.1370* (-4.72) -0.0385* (-4.16) 0.0225* (3.69) -0.0610* (-6.64) -0.0110 (-0.45) 0.0671* (2.20) -0.0781* (-2.38) -0.1250* (-8.25) -0.0181 (-1.41) -0.1070* (-7.51) Size crisis 0.0057 (1.95) -0.0043 (-1.71) 0.0100* (3.02) 0.0011 (1.09) -0.0020* (-2.91) 0.0031* (3.03) -0.0004 (-0.14) -0.0054 (-1.54) 0.0049 (1.33) 0.0079* (4.72) 0.0040* (2.79) 0.0039* (2.45) ROA crisis -0.0035* (-7.45) -0.0012* (-5.52) -0.0023* (-5.10) -0.0028* (-17.67) -0.0004* (-6.57) -0.0023* (-14.80) -0.0031* (-5.97) -0.0003 (-0.82) -0.0028* (-5.12) -0.0038* (-14.46) -0.0021* (-9.61) -0.0017* (-8.11) Tangibility crisis 0.0020 (0.19) 0.0014 (0.15) 0.0006 (0.06) -0.0841* (-17.81) -0.0524* (-14.97) -0.0316* (-7.35) -0.0155 (-1.36) -0.0257 (-1.80) 0.0102 (0.68) -0.0105 (-1.69) -0.0202* (-3.45) 0.0097 (1.66) Growth crisis 0.0162 (1.71) -0.0282* (-4.19) 0.0444* (4.26) -0.0069 (-1.75) -0.0064* (-3.17) -0.0006 (-0.13) 0.0028 (0.22) 0.0259 (1.72) -0.0231 (-1.34) 0.0048 (0.64) -0.0044 (-1.08) 0.0092 (1.39) Observations 26024 26024 26024 185806 185806 185806 13739 13739 13739 96408 96408 96408 Adj. R 2 0.133 0.115 0.118 0.222 0.051 0.231 0.198 0.038 0.111 0.212 0.075 0.193 Note: This table provides the estimation results for each country dataset with firm fixed effects regressions. The t-statistics are presented in the parentheses. A * indicates the estimate is significant at the 5% level. 19

4.1 Full sample estimates The results of the firm fixed effects panel data regressions for total debt, long term debt and short term debt are presented in Table 3. The results are presented for each country to facilitate comparison. In all three models most of the individual parameters are statistically significant. As the above table reports, the first hypothesis H1, that larger firms have higher leverage, is confirmed for all countries. Obviously, size plays an important role in determining the level of total debt. Looking closely and more specifically to hypothesis H1a which refers to the association that size has with long term debt, it is clear that this hypothesis is confirmed for all countries, even at the 1 significance level. This demonstrates that large firms use more long term debt and less short term debt to finance their operations and that the increase in long term debt outweighs the decrease in short term debt. Our result is consistent with what Sogorb-Mira (2005) find for Spanish firms. Our hypothesis about profitability is confirmed as well. More profitable firms employ considerably less debt, which is in line with the pecking order theory. By comparing the magnitude of the effect we see that there are no big differences between countries: one standard deviation increase in profitability lowers the total debt ratio by 2% for Greece, 2.4% for Italy, 2.5% for Portugal and 2.1% for Spain. For the calculations we used the standard deviation for the years 2004-2007, which is presented in the previous summary statistics tables. Our findings are consistent with previous studies, as with Sogorb-Mira (2005), Heyman et al. (2008) and Mateev et al. (2013). The negative relationship applies for long term and short term debt as well, with the effect on short term debt being larger. Hypothesis H3 concerning tangibility is rejected. Collateral appears to be an important determinant of leverage for all countries but the relationship with the level of debt a firm selects is significantly negative which comes in contrast to our hypothesis. Only for Spain we observe a positive but insignificant relationship. A possible explanation for this finding is that tangible assets are often illiquid especially in periods of recession. As a consequence, if a high proportion of total assets consist of tangible assets, firms are likely to face financial distress problems. As a result, tangible assets do not provide safety and this fact justifies the negative relationship between the amount of collateral a firm owns and the level of debt selected. A different interpretation relies on the structure of our data. The coefficient of tangibility on total debt is a weighted average of long term and short term debt. The table suggests that this result is driven by the negative coefficient of short term debt, so it might be the case that the percentage of short term debt is substantially higher. We calculated the averages of both ratios for all countries and 20

we found that short term debt is approximately three times higher than long term debt, which supports our explanation about the rejection of our hypothesis. Our results indicate a strong and significant positive relationship between growth of total assets and total debt. Accordingly, hypothesis H4 is confirmed. These findings provide again strong support for the pecking order theory. Firms grow and as a consequence opt for higher levels of leverage because they can afford a riskier capital structure composition due to their expansion and development. From the table it can be observed that the effect of growth is more pronounced on short term than on long term debt. Now we concentrate on the effect of the crisis and whether it changed the importance of the determinants. First of all, looking at the coefficient of crisis-dummy we observe that there is a significant decline in total debt after 2008 for Greece, Italy and Spain of 7.6%, 3.85% and 12.5% on average respectively. This means that the first research question is confirmed. For all countries, this result is driven mostly by a decline in short term debt. For Portugal the signs of the estimates remain the same but are not significantly different in the second sub-period. A possible explanation is that the crisis affected Portugal at a later stage. Moreover, we do not have many observations for Portuguese firms compared to the other countries. Overall, leverage has changed significantly because of the crisis and there is less debt as a benchmark. Firms adapt to the new, tougher economic conditions and environment and change their capital structure composition towards less risky financial resources. Looking at each country separately, we answer our second research question. First of all, we observe that the only determinant that has become more important for all countries because of the crisis is profitability. Clearly, profitable firms employ even less debt than they did before which is in line with the pecking order theory. It is interesting to compare the magnitude of the effect after the crisis: one standard deviation increase in profitability lowers the total debt ratio by additional 3 percentage points for Greece, 2% for Italy, 2.3% for Portugal and 3.3% for Spain. Standard deviation for the period 2008-2011 obtained from Table 2 is used for the calculations. This finding accounts for both long term and short term debt with the exception of Portugal where long term debt ratio is not affected significantly by the crisis. Again a reasonable explanation is that Portugal was hit by the crisis at a later stage. For Italy, this result is driven mainly by the effect ROA has on short term debt. 21

Investigating in depth the table, we see that growth has become a more important determinant for Greek firms due to the crisis and growing firms employ further more debt. The same accounts for the size of Greek firms. In Italy, collateral has a bigger significant effect in this period. This holds also when we look at long term and short term debt. Spain is the only country where size has become significantly more important. We interpret this result with respect to the fact that Spain is bigger compared to the other countries of our sample and this might affect also the size of the companies incorporated there. The coefficient of tangible assets is significant only for long term debt and thus does not affect significantly the total debt ratio. For Portuguese firms, no other significant change in the determinants of leverage is observed. 4.2 Industry effects We apply again the same fixed effects panel data method as in Section 4.1 but this time we aim to control for industry effects that play a role in affecting the importance of the determinants of capital structure choice. Industry effects are implemented by including in the regression the dummy variable industry which takes values from 1 to 6, indicating the different industries. Our firms are allocated to the following six industries: energy supply, manufacturing, construction, wholesale and retail trade, transport and services. The results of the regressions are presented in Table 4 below. The method was applied also for long term and short term debt as dependent variables. As our findings indicate, profitability and growth are still statistically significant determinants of total debt level for all countries and confirm hypotheses H2 and H4. Regarding hypothesis H1, size has a positive impact on total debt in all countries and this effect appears also to be significant everywhere apart from Portugal. This might be due to the fact that crisis did not affect Portuguese firms at that early stage. Hypothesis H1a for long-term debt is also confirmed in all countries. Concerning collateral and hypothesis H3, our results are similar to those derived from firm fixed effects regressions before. To summarize, these results suggest that the pecking order theory is most relevant even when we control for industry effects. Thus, our previous findings are robust after controlling for industry fixed effects. Looking at the coefficient of the crisis-dummy, our findings are comparable with section 4.1. Crisis has affected in a significantly negative way the total amount of debt employed in the average firm, regardless of the country of incorporation. Profitability exhibits again increased importance for all countries in favor of the pecking order theory. As far as the other explanatory variables are concerned, the coefficients are similar with those in Table 3. Size is more important only for Spanish firms, collateral for Spanish and Italian firms and growth only for Greek firms. 22

Table 4: Leverage: Panel data regressions with Industry Fixed Effects Total debt Size 0.0553* (10.72) GREECE ITALY PORTUGAL SPAIN Long term Short term Total debt Long term Short term Total debt Long term Short term Total debt Long term debt debt debt debt debt debt debt 0.0543* -0.0094* 0.0258* 0.0248* -0.0055* 0.0082 0.0345* -0.0325* 0.0472* 0.0344 * (18.72) (-1.96) (16.62) (33.34) (-3.74) (1.46) (8.21) (-6.28) (17.06) (18.36) Short term debt 0.0064* (2.68) ROA -0.003* (-7.53) -0.0004* (-2.16) -0.0029* (-7.79) -0.0046* (-35.51) -0.0008* (-14.89) -0.0041* (-31.20) -0.0056* (-12.76) -0.0022* (-5.62) -0.0043* (-9.03) -0.0036* (-16.99) -0.0015* (-7.78) -0.0025* (-14.54) Tangibility -0.1550* (-9.54) 0.1940* (17.31) -0.3720* (-24.10) -0.2020* (-33.41) 0.1930* (49.19) -0.3820* (-69.91) -0.0241 (-1.04) 0.2250* (12.69) -0.2310* (-10.82) 0.0024 (0.29) 0.1970* (25.85) -0.1920* (-24.35) Growth 0.0482* (7.48) 0.0045 (1.00) 0.0475* (6.63) 0.0522* (14.83) -0.0036* (-2.79) 0.0591* (15.97) 0.0593* (6.54) -0.0119 (-0.99) 0.0798* (6.23) 0.0664* (13.51) 0.0141* (5.30) 0.0556* (14.19) Constant 0.1200 (1.72) -0.4020* (-9.17) 0.6160* (9.22) 0.4640* (28.33) -0.2000* (-24.85) 0.7180* (46.86) 0.6060* (9.44) -0.2220* (-4.82) 0.8790* (13.98) 0.1410* (4.40) -0.2660* (-13.13) 0.4680* (16.90) Crisis-dummy -0.0938 (-1.90) 0.0264 (0.87) -0.1130* (-2.16) 0.0078 (0.70) 0.0364* (5.02) -0.0247* (-2.26) -0.0166 (-0.53) 0.0763 (1.88) -0.0952* (-2.09) -0.1120* (-5.97) -0.0304 (-1.81) -0.0828* (-4.30) Size crisis 0.0052 (1.79) -0.0033 (-1.33) 0.0087* (2.68) 0.0006 (0.56) -0.0022* (-3.34) 0.0025* (2.52) -0.0010 (-0.35) -0.0049 (-1.45) 0.0040 (1.11) 0.0075* (4.40) 0.0039* (2.71) 0.0035* (2.19) ROA crisis -0.0033* (-7.23) -0.0010* (-4.66) -0.0022* (-5.06) -0.0024* (-15.79) -0.0002* (-3.51) -0.0020* (-13.32) -0.0024* (-5.05) 0.0001 (0.21) -0.0021* (-4.02) -0.0034* (-13.08) -0.0018* (-7.91) -0.0014* (-6.73) Tangibility crisis -0.0050 (-0.43) 0.0061 (0.66) -0.0074 (-0.63) -0.0918* (-19.72) -0.0538* (-15.71) -0.0383* (-9.05) -0.0167 (-1.37) -0.0275 (-1.81) 0.0136 (0.85) -0.0173* (-2.62) -0.0092 (-1.49) -0.0041 (-0.68) Growth crisis 0.0281* (2.99) -0.0188* (-2.93) 0.0504* (4.95) -0.0018 (-0.43) -0.0040* (-2.09) 0.0028 (0.64) 0.0079 (0.62) 0.0270 (1.88) -0.0206 (-1.24) 0.0120 (1.52) -0.0029 (-0.70) 0.0163* (2.36) Observations 26024 26024 26024 185806 185806 185806 13739 13739 13739 96408 96408 96408 Adj. R 2 0.168 0.193 0.307 0.167 0.160 0.259 0.193 0.145 0.162 0.115 0.183 0.177 Note: The table provides the estimation results for each country dataset when industry effects are included. The t-statistics are presented in parentheses. A * indicates significance at the 5% level. 23