Influence of the Financial crisis on Capital Structure
Abstract - This paper has tried to answer the question whether corporate capital structures have become more conservative after the financial crisis of 2007 with data from a subsample of listed corporations in Germany and the Netherlands. It was hypothesized that leverage ratios declined after the crisis and regression results show that the accumulated effect of the post-crisis period on leverage is indeed negative when control variables do not change in value (ceteris paribus). The stronger negative effect of corporate risk seems to contribute a lot and can be explained by investors of capital demanding more quality after the start of the crisis, which increased costs of possible financial distress, thereby decreasing leverage ratios. Corporations that followed an aggressive financial strategy before the crisis, experienced an extra decline in their leverage ratios. This too can be explained by investors of capital demanding more quality after 2007. The result are robust to several changes in methodology. (JEL: G01, G32)
Key words: leverage; financial crisis; capital structure
Student number: s2004631
Name: Carolien Deiman
2
1.
Introduction
More than a decade has gone by since the financial crisis, which started during the second half of 2007, converged into the Great Recession. Being considered the worst recession since the Great Depression of the 1930s (Kahle and Stulz, 2013), it entailed large banks incurring massive losses on their securities and mortgage portfolios. They had to reduce their assets or raise more capital in order to maintain their solvability, which made them more vigilant in issuing corporate loans. Not just bank lending, but credit supply in general was affected negatively. Because asset values fell, corporations witnessed their net worth and that of their collateral diminish. On top of that, they experienced a shock in consumer demand and an increase in uncertainty about future demand (Kahle and Stulz, 2013). These events can affect capital structure decisions of corporations, which form part of the domain of the board, and empirical studies have shown that leverage ratios of corporations, at least in the short run, were indeed affected by the crisis (see, e.g., Kahle and Stulz, 2013; Iqbal and Kume, 2014; Harrison and Widjaja, 2014).
Currently, the economy is booming again and corporate boards have returned to their businesses as usual. But have they really? Malmendier, Tate and Yan (2011) argue that macroeconomic shocks like the financial crisis, not only have an immediate effect on capital structure decisions, but also on future capital structure policies. They provide evidence that boards that have experienced major events like crises, are affected in their beliefs and risk attitudes: they display an increased aversion to access external sources of capital, making them more conservative in their capital structure preferences. Before them, Hackbarth (2008) found that alongside other determinants, managerial traits like growth and risk perception can indeed help explain differences in capital structure decisions. A financial crisis like the one that started in 2007 can make boards more pessimistic and uncertain about their future earnings, which can affect their decisions with regard to corporate financial policy in a way that makes them less willing to take on debt. Therefore, the following research question is composed: “Have corporate capital structures become more conservative after the financial crisis of 2007?” This paper shall try to answer this question by looking at the development of capital structures from non-financial, unregulated, listed corporations in the Netherlands and Germany from 2002 until 2017 by means of panel regression analysis.
network-3 oriented, just as most countries in continental Europe, indicating that corporate financial capital is, for a substantial amount, mobilized by institutions such as banks, families, insurance companies and national states. Within the group of countries that have network-oriented corporate systems, the Netherlands and Germany can be categorized as Germanic (Moerland, 1995). In the group of Germanic countries, and especially in Germany itself, the banking sector plays a prominent role in corporate finance and control (Moerland, 1995). Other neighbouring countries such as Belgium, France, and Luxembourg can be classified as Latinic, in which family, financial holding and state-ownership is more important. Because the economies of the Netherlands and Germany are so intertwined and because corporate financial structures are similar in these countries in that banks play a prominent role, it makes sense to investigate the impact of the financial crisis in these countries together, especially since it were the banks that got into trouble during the financial crisis.
This paper contributes to literature by investigating the assumption made by Malmendier et al. (2011) that corporate capital structures have become more conservative after the financial crisis that started in 2007, thereby providing additional insight and empirical evidence on how a financial crisis can affect capital structure decisions of corporations in the short and in the longer run.
The remainder of this paper is organized as follows. Section 2 provides a literature review with regard to the factors determining capital structure and on how the financial crisis could have had an effect on capital structure. Section 3 then discusses the methodology followed by section 4 that discusses the data. Section 5 then presents the results and section 6 concludes.
2.
Literature review
Even though Modigliani and Miller already stated in 1958 that corporate value is not affected by the way in which corporations are financed, capital structure still matters because markets are not always perfect and frictionless. Different theories exist about how choices between equity and debt financing are made and they differ in the way they weigh and interpret factors like information asymmetry between insiders and outsiders, taxes and agency costs (Myers, 2001). Pecking order theory, trade-off theory, free cash flow theory and market timing theory all have different views on the matter. These more general theories provide information about possible explanatory variables and their relation with financial leverage, where the size and direction of the relation can be dependent on the theory applied.
General theories on capital structure
4 1991; Myers, 2001; Harrison and Widjaja, 2014). The reasoning behind this is that outside investors infer information from the decision to issue equity: the company is overvalued. This leads to falling stock prices when equity issuances are announced. For debt, the information asymmetry does not matter as much because debt holders have prior claim to equity holders, making the impact of debt issuance on share prices less severe. Internal financing is preferred over both because information asymmetry does not play a role here. According to pecking order theory, there is no notion of an optimal leverage ratio (Frank and Goyal, 2009). Another theory is trade-off theory. This theory emphasizes the tax factor and predicts that corporations will pursue more debt until the tax advantages of additional debt are balanced against the additional costs of possible financial distress, like bankruptcy costs and agency costs caused by dubious corporate creditworthiness (Frank and Goyal, 2009; Harrison and Widjaja 2014). Trade-off theory rationalizes moderate debt ratios (Myers, 2001). Pecking order theory and trade-off theory assume that managers act in the interests of existing shareholders, as opposed to the third theory: free cash flow theory (Myers, 2001), which some consider part of trade-off theory (Frank and Goyal, 2009). This theory emphasizes agency costs and predicts that despite the costs of possible financial distress, debt can increase corporate value when corporations are inclined to invest their operating cash flows in bad projects (Frank and Goyal, 2009; Myers, 2001). The final theory discussed here is market timing theory. This theory states that managers are indifferent between internal and external financing, and within the latter between debt and equity financing. They simply use the source that comes at the least cost or is otherwise most favourable at the time the corporation seeks financing (Frank and Goyal, 2009; Harrison and Widjaja, 2014).
Factors explaining capital structure
Now that the theories are discussed, this section turns to factors that explain financial leverage. What determines financial leverage has remained elusive despite all the research that has focussed on the subject. However, many agree that a corporations tangibility, growth opportunities, size, profitability, risk and the industry in which it is active, have some explanatory power (see, e.g., Frank and Goyal, 2009; Harris and Raviv, 1991;) and many more that at least four or five out of these six factors play a role (see, e.g., Harrison and Widjaja, 2014; Malmendier et al., 2011; Rajan and Zingales, 1995). So what do the general theories of capital structure say about how these factors can explain financial leverage?
5 The amount of growth opportunities is a second factor in explaining financial leverage. When corporations have extensive growth opportunities, the costs of possible financial distress are expected to be higher as well and free cash flow problems are reduced. Therefore, a negative relation is to be expected, consistent with trade-off theory and free cash flow theory. Another explanation for a negative relation would be that firms tend to issue stock when they are overvalued, consistent with market timing theory. However, according to pecking order theory, growth opportunities (implying more investments) lead to corporations taking on more debt. Therefore, this theory predicts a positive relation. Despite ambiguous theoretical predictions, empirical evidence seems to be in support of the negative relationship (see, e.g., Frank and Goyal, 2009; Malmendier et al., 2011; Rajan and Zingales, 1995).
A third factor with explanatory power is size. Size may serve as a proxy for the inverse probability of default and can therefore be positively related with leverage when costs of possible financial distress are low, consistent with trade-off theory. However, pecking order theory predicts that because information asymmetry between insiders and outsiders is expected to be lower for large corporations, they should have lower debt levels. Frank and Goyal (2009), Malmendier et al. (2011) and Rajan and Zingales (1995) find empirical evidence that supports the positive relation in general. However, in their German subsample, Rajan and Zingales (1995) found a negative relationship between size and leverage, giving empirical support for pecking order theory in this specific country.
Profitability is a fourth factor that can help explain financial leverage. According to trade-off theory, profitability decreases costs of possible financial distress and increases the value of tax shields. Therefore, a positive relation can be expected. Free cash flow theory predicts the same for a different reason: profitable corporations are more inclined to have problems related to an extensive amount of free cash flow. However, according to pecking order theory, profitability should influence leverage in a negative way because corporations prefer to finance with internal funds rather than debt. The empirical evidence is supportive of the negative relation in several studies (see, e.g., Frank and Goyal, 2009; Harrison and Widjaja, 2014; Malmendier et al., 2011; Rajan and Zingales, 1995).
Corporate risk is the fifth factor to be discussed. When corporations have more volatile cash flows, the possibility that tax shields can be fully used is reduced and costs of possible financial distress increase. Trade-off theory thus predicts a lower leverage ratio when corporate risk is higher. Pecking order theory predicts the opposite since high volatility of stock returns also leads to those corporations suffering more from adverse selection. Empirical evidence with regard to the effect seems to favour the negative relationship in some studies (Harris and Raviv, 1991). However, Frank and Goyal (2009) find that the effect is not consistent across alternative treatments of the data.
6 corporations in different industries and these similar characteristics lead to similar leverage ratios. Therefore, corporations in industries with a high (low) average leverage ratio tend to have a high (low) leverage ratio as well. This is consistent with trade-off theory. A different interpretation is that corporations adjust their leverage ratios actively towards the industries average. Empirical evidence for corporations in the same industries having similar leverage ratios is provided by Frank and Goyal (2009) and MacKay and Phillis (2005).
Macroeconomic shocks during the financial crisis
Firm and industry-specific factors alone cannot explain capital structures. The macroeconomic environment plays an important role as well (Frank and Goyal, 2009; Harrison and Widjaja, 2014). Therefore, a lot of research has focussed on the relationship between macroeconomic conditions and the way in which corporations are financed (see, e.g., Erel, Julio, Kim, and Weisbach, 2012; Harrison and Widjaja, 2014; Kahle and Stulz, 2013; Iqbal and Kume, 2014; Mendoza, 2010). Several theories exist about the way in which macroeconomic conditions affect capital structure. Some focus on the supply side of capital, some focus on the demand side of capital and some on collateral. These theories can affect both the quantity of capital raised as well as the type of security used to raise this capital (Erel et al., 2012). Kahle and Stulz (2013) have tried to identify the separate effects of these theories during the financial crisis that started in 2007. They find that corporate borrowing (and capital expenditures) decreased during the crisis, especially after the fall of the bank Lehman Brothers, and put the theories to the test to see whether they can explain the empirical evidence they find.
7 However, Kahle and Stulz (2013) find no empirical evidence that supports the credit supply shock theory as a dominant causal factor for the changing corporate financial policies during the financial crisis of 2007.
They do however, find empirical evidence for the demand shock theory (Kahle and Stulz, 2013). This theory states that corporate desired investment decreased as well as corporate demand for capital to finance investment because of, among others, a decrease in consumer credit, a decrease in housing wealth, and the panic that followed after the collapse of Lehman. The way in which this theory can affect corporate leverage ratios is ambiguous. Kahle and Stulz (2013) argue that the uncertainty, substantiating the postponement of investments, leads corporations to have higher cash holdings and lower optimal levels of debt. Frank and Goyal (2009) suggest that taxable income is less during recessions, which also rationalizes lower optimal levels of debt from the perspective of the corporation. Other mechanisms originating from the demand side of capital, do not necessarily predict lower optimal levels of debt. Erel et al. (2012) state that changes on the demand side are based on changes in information asymmetry between insiders and outsiders. Pecking order theory predicts that during times of economic expansion, leverage declines because internal funds increase, holding everything else equal (Frank and Goyal, 2009). If adverse selection costs associated with asymmetric information increase when the macroeconomic environment deteriorates in terms of business conditions, then according to pecking order theory, the preference for issuing debt over equity becomes stronger. This mechanism too can lead to corporations having to offer more security to investors. Of course, internal funds, if sufficient, are still preferred over external funds. However, the empirical evidence provided by Erel et al. (2012) is more in support of supply of capital having an impact on corporate financing during crises than demand for capital.
8 In general, Iqbal and Kume (2014) found the following empirical evidence with regard to how the financial crisis had an impact on leverage: leverage ratios increased from the pre-crisis period, 2006-2007, to the period they labelled pre-crisis period, 2008-2009, but then decreased again to their pre-crisis values in the period they named post-crisis, 2010-2011. Thus, despite some theories that are in support of the claim from Malmendier et al. (2011) that corporations de-leverage and shift towards self-sufficiency, the studies mentioned so far do not give clear-cut empirical evidence for this.
Factors of capital structure and the financial crisis
The previous section provided a theoretical basis on how leverage in general can change during crises. But how are underlying factors influenced? Harrison and Widjaja (2014) investigated the effect of the financial crisis on the influence of four of the six factors considered earlier that can determine capital structure: tangibility, growth opportunities, size and profitability. They did so by comparing regression results from their total sample with two subsamples and between the two subsamples. Their dependent variable was book leverage. The total sample consisted of data from 2004 to 2011, one subsample contained data from 2004 to 2007, which they labelled the pre-crisis period, and one sample had data from 2008 to 2011, their post-crisis period. Unfortunately, they did not consider industry effects and the effect of corporate risk.
They find that in the period after the financial crisis, tangibility has substantially more explanatory power than in the period before the financial crisis. The size of the effect as well as the relative explanatory power has increased. Apparently, in the period after the financial crisis, corporations with more tangible assets were inclined to have higher leverage. This is consistent with the theory that the financial crisis caused suppliers of capital to flight to quality. Harrison and Widjaja (2014) also find that in their sample, the negative influence of growth opportunities, as proxied by the market-to-book ratio, weighs substantially heavier after the crisis than before the crisis in terms of size of the negative effect as well as in terms of relative explanatory power. This means that the market valuation of a corporation has a greater influence and this market valuation was low in the period after the crisis compared to the period before the crisis. Therefore, Harrison and Widjaja consider it possible that corporations are more leveraged in the post-crisis period (2014).
The influence of size was considered too, but it turned out to be insignificant in either period (Harrison and Widjaja, 2014).
9 Managerial traits and capital structure
Now that attention is given to how leverage can change during crises and how underlying factors explaining capital structure might be influenced, this section turns to an explanatory factor of capital structure not considered yet in the previous sections. This factor cannot only explain how leverage changes when a crisis hits, but also how a crisis can have a long-lasting influence.
10 Hypothesis
Concluding, the literature reviewed in this section has provided some interesting insights in answering the research question as mentioned in the introduction. First, attention was given to general theories and factors explaining capital structure. Then, literature was explored regarding how macroeconomic conditions could have influenced capital structure during and shortly after the crisis. The capital supply shock theory was linked to the trade-off and the free cash flow theory, which provides a theoretical basis for rational debt levels to decline during the financial crisis. Then, the influence of the financial crisis on the demand side of capital was explored. A smaller demand for credit, a decrease in the value of tax shields and an increase in information asymmetry between insiders and outsiders can all lead to declining leverage ratios, but an increase in information asymmetry can also lead to the increase of leverage ratios during the crisis. Finally, the balance sheet multiplier effect increases costs of possible financial distress, again indicating that rational debt ratios should decrease. However, empirical evidence to support these theories are often inconclusive. Then, attention was given to how the factors determining capital structure could have been influenced by the financial crisis. At last, literature is reviewed with regard to how financial crisis can have a long-term effect on managerial risk and growth perception, which in turn can affect leverage. The predictions of Hackbarth’s (2008) model may work the other way round as well: when top executives underestimate their future returns and/or overestimate the riskiness of their future returns, they choose lower debt levels than unbiased top executives. It then seems plausible that an important event like the financial crisis, can lead to corporations taking on less debt, which would be consistent with the claim from Malmendier et al. (2011) that corporations experienced de-leveraging after the crisis. To put these theories to the test, the following hypothesis shall be assessed in the sequential sections:
Hypothesis Corporate leverage ratios have declined after the financial crisis compared to the period before the financial crisis.
3.
Methodology
To test whether corporate leverage ratios have decreased after the financial crisis, the following regression is run for the period from 2002 until 2017
𝐿𝐸𝑉𝑖𝑡 = 𝛼 + 𝛽1𝑇𝑖𝑡+ 𝛽2𝑀𝑇𝐵𝑖𝑡+ 𝛽3𝑆𝑖𝑡+ 𝛽4𝑃𝑖𝑡+ 𝛽5𝑅𝑖𝑡+ 𝛽6𝐴𝐺𝐸𝑖𝑡+ 𝑢𝑖 + 𝛽7𝐷1𝑖𝑡+ 𝛽8𝐷2𝑖𝑡+ 𝛽9𝑇𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽10𝑀𝑇𝐵𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽11𝑆𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽12𝑃𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽13𝑅𝑖𝑡∗ 𝐷1𝑖𝑡+
𝛽14𝐴𝐺𝐸𝑖𝑡∗ 𝐷1𝑖𝑡 + 𝛽15𝐷1𝑖𝑡∗ 𝐷3𝑖+ 𝑣𝑖𝑡 (1)
11 value of assets. This choice is made because corporate boards focus on these values more than say market value of equity (Frank and Goyal, 2009; MacKay and Phillips, 2005), which makes it more interesting for this paper because the literature section provided some theoretical evidence that indicated that top executives may want to decrease leverage after the crisis. The coefficients 𝛽1 t/m 𝛽6 estimate the impact of six control variables as mentioned in the literature section. 𝛽1 represents the influence of tangibility, 𝑇𝑖𝑡. Tangibility is calculated as the ratio of the book value of net property, plant and equipment (PPE) to the book value of assets. The expected sign of 𝛽1 is positive, in line with previous empirical evidence. 𝑀𝑇𝐵 is the market-to-book ratio, the market value of equity divided by the book value of equity. In accordance with empirical evidence from previous studies, the expected sign of 𝛽2 is negative. 𝑆𝑖𝑡 is the size of corporation 𝑖 at time 𝑡 and is included in the regression as well. Size is measured as the natural logarithm of net sales. A different option would have been to take total assets as a measure, this however could have distorted results because some corporations (e.g. service) use human capital more than assets. The expected sign of 𝛽3 is ambiguous. However, considering the empirical evidence provided by Rajan and Zingales (1995) with regard to corporations in Germany it could be negative. Next, the control variable 𝑃𝑖𝑡, the profitability of corporation 𝑖 at time 𝑡, is included. Profitability is calculated as the earnings before interest and taxes divided by the book value of the assets. This measure is chosen to prevent leverage from affecting profitability, which leads to an endogeneity problem. Considering empirical evidence from previous studies, the expected sign of 𝛽4 is negative. The impact of the fifth control variable 𝑅 represents the impact of corporate risk. Corporate risk is proxied by asset volatility and is calculated as the standard deviation of weekly returns times the ratio of book value of equity to book value of assets. This proxy is rather crude considering that liabilities hold risk too, but it serves the purpose in this paper. In accordance with the literature section, the expected sign of 𝛽5 is negative. The sixth control variable represents the influence of the average age of the CEO and CFO of the corporation. The choice for CFO as the next most influential executive is made because it seems plausible that this executive has a lot of authority in financial leverage decisions. The expected sign of 𝛽6 is negative. In accordance with findings from previous research by Lemmon et al. (2008), the outcome is controlled for firm fixed effects, 𝑢𝑖, because the majority of the total variation in capital structure is due to time-invariant factors. Industry effects are fixed and are therefore not separately included in the regression of model (1).
12 about the sign of 𝛽8. Next to just including this dummy variable, interaction variables between the six control variables and 𝐷1 are included as well, to see whether the financial crisis had an impact on their explanatory power. In accordance with Harrison and Widjaja (2014), 𝛽9 is expected to be positive, 𝛽10 is expected to be negative, 𝛽11 is expected to be negative and 𝛽12 is expected to be positive. Harrison and Widjaja (2014) do not provide empirical evidence on the signs of 𝛽13 and 𝛽14. Another interaction variable is added to the regression model because Iqbal and Kume (2014) provided evidence that the financial strategy before the crisis mattered in whether the crisis had an impact. To check whether this holds for our sample as well, an interaction variable between 𝐷1 and 𝐷3 is included. The dummy 𝐷3 takes the value 0 when corporations have an average leverage ratio before the crisis that is below the industry mean, considering their financial strategies to be conservative. It takes the value 1 when the corporations have an average leverage ratio before the crisis that is above the industry mean, indicating that their financial strategies are more aggressive. In accordance with Iqbal and Kume (2014), the expected sign of 𝛽15 is negative.
To test for robustness, another regression is run where leverage is calculated differently to see what the impact of that decision was. Leverage, here named market leverage, is now measured as a ratio of book value of liabilities to the sum of book value of liabilities plus the market value of the equity of the corporation. Results may change because the market value of equity decreased and consequently the market leverage increased during the crisis years. It is therefore expected that the coefficient of 𝛽8 increases. The regressions take the following form:
𝑀𝐿𝐸𝑉𝑖𝑡 = 𝛼 + 𝛽1𝑇𝑖𝑡 + 𝛽2𝑀𝑇𝐵𝑖𝑡+ 𝛽3𝑆𝑖𝑡+ 𝛽4𝑃𝑖𝑡+ 𝛽5𝑅𝑖𝑡+ 𝛽6𝐴𝐺𝐸𝑖𝑡 + 𝑢𝑖 + 𝛽7𝐷1𝑖𝑡+ 𝛽8𝐷2𝑖𝑡 + 𝛽9𝑇𝑖𝑡 ∗ 𝐷1𝑖𝑡+ 𝛽10𝑀𝑇𝐵𝑖𝑡 ∗ 𝐷1𝑖𝑡 + 𝛽11𝑆𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽12𝑃𝑖𝑡∗ 𝐷1𝑖𝑡+
𝛽13𝑅𝑖𝑡∗ 𝐷1𝑖𝑡 + 𝛽14𝐴𝐺𝐸𝑖𝑡∗ 𝐷1𝑖𝑡 + 𝛽15𝐷1𝑖𝑡∗ 𝐷3𝑖+ 𝑣𝑖𝑡 (2)
Then, a second robustness test is performed with regard to controlling for the turbulence in the crisis years. In the first regression model, 𝐷2 took the value 1 in 2008 and 2009, and 0 in the other years. The amount of years included in this crisis dummy can impact regression results a lot if markets have not stabilized enough by the end of 2010. Therefore, it makes sense to check whether regression results start to deviate a lot from the results from model (1) when 2010 is included in the crisis dummy and excluded from the post-crisis dummy. The third model thus replaces dummy 𝐷1 by 𝐷4, that takes value 0 from 2002 to 2010 and 1 from 2011 until 2017 and replaces 𝐷2 by 𝐷5, that takes the value 1 in 2008, 2009 and 2010 and 0 in any other year:
𝐿𝐸𝑉𝑖𝑡 = 𝛼 + 𝛽1𝑇𝑖𝑡+ 𝛽2𝑀𝑇𝐵𝑖𝑡+ 𝛽3𝑆𝑖𝑡+ 𝛽4𝑃𝑖𝑡+ 𝛽5𝑅𝑖𝑡+ 𝛽6𝐴𝐺𝐸𝑖𝑡+ 𝑢𝑖 + 𝛽7𝐷4𝑖𝑡+ 𝛽8𝐷5𝑖𝑡+ 𝛽9𝑇𝑖𝑡∗ 𝐷4𝑖𝑡+ 𝛽10𝑀𝑇𝐵𝑖𝑡∗ 𝐷4𝑖𝑡+ 𝛽11𝑆𝑖𝑡∗ 𝐷4𝑖𝑡+ 𝛽12𝑃𝑖𝑡∗ 𝐷4𝑖𝑡+ 𝛽13𝑅𝑖𝑡∗ 𝐷4𝑖𝑡+
13 As a third robustness test, fixed effects are replaced by industry dummies and interaction variables between the industry dummies and the post-crisis dummy are added. This is done because of the possibility that change in leverage is industry specific. To avoid perfect multicollinearity, the intercept and post-crisis dummy itself are excluded from the regression:
𝐿𝐸𝑉𝑖𝑡 = 𝛽1𝑇𝑖𝑡+ 𝛽2𝑀𝑇𝐵𝑖𝑡+ 𝛽3𝑆𝑖𝑡+ 𝛽4𝑃𝑖𝑡+ 𝛽5𝑅𝑖𝑡+ 𝛽6𝐴𝐺𝐸𝑖𝑡 + 𝛽7𝐷6𝑖𝑡+ 𝛽8𝐷7𝑖𝑡 + 𝛽9𝐷8𝑖𝑡+ 𝛽10𝐷9𝑖𝑡+ 𝛽11𝐷10𝑖𝑡+ 𝛽12𝐷11𝑖𝑡+ 𝛽13𝐷12𝑖𝑡+ 𝛽14𝐷13𝑖𝑡+ 𝛽15𝐷2𝑖𝑡+ 𝛽16𝑇𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽17𝑀𝑇𝐵𝑖𝑡∗ 𝐷1𝑖𝑡 + 𝛽18𝑆𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽19𝑃𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽20𝑅𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽21𝐴𝐺𝐸𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽22𝐷6𝑖𝑡∗ 𝐷1𝑖𝑡 + 𝛽23𝐷7𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽24𝐷8𝑖𝑡∗ 𝐷1𝑖𝑡 + 𝛽25𝐷9𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽26𝐷10𝑖𝑡 ∗ 𝐷1𝑖𝑡+ 𝛽27𝐷11𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽28𝐷12𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽29𝐷13𝑖𝑡∗ 𝐷1𝑖𝑡+ 𝛽30𝐷1𝑖𝑡 ∗ 𝐷3𝑖+ 𝑣𝑖𝑡 (4)
A last robustness test is performed with respect to the post-crisis period. The literature section provided theoretical evidence that the impact of the financial crisis could be long term. However, this may not be the case and therefore it makes sense to check whether regression results from model (1) change a lot when data from the last two years of the post-crisis period, 2016 and 2017, are excluded from the sample.
For an overview of the variables and dummies used in the regressions of this section and how they are determined, see appendix I.
4.
Data
For this study, a sample of corporations listed on the Frankfurt Stock Exchange and the Euronext Amsterdam is used. This sample consists of stocks included in the AEX (Amsterdam Exchange index), the AMX (Amsterdam Midcap Index), the DAX (Deutscher Aktienindex) and the MDAX (Midcap DAX). The AEX and DAX are large cap indices with 25 respectively 30 constituents. The AMX and MDAX are midcap indices with 25 respectively 60 constituents. The initial sample thus consists of 140 corporations. Annual data with regard to leverage ratios and control variables are obtained from the Thomson Reuters Datastream database for the period from 2002 until 2017, thereby using constituent lists LAMSTEOE (AEX), LAMSMKAP (AMX), LDAXINDX (DAX) and LMDAXIDX (MDAX). Data with respect to CEO and CFO age is retrieved from the BoardEx database.
14 are distributed among indices and industries. The sample seems rather dispersed among the indices and industries.
TABLE 1 - Distribution of sample among different indices and industrial classes
Industry AEX AMX DAX MDAX Total
Oil and gas 1 2 - - 3
Basic materials 3 - 1 5 9 Industrials 3 5 5 8 21 Consumer goods 2 2 7 2 13 Healthcare 2 - 4 3 9 Consumer services 2 2 1 5 10 Telecommunications 1 - 1 2 4 Technology 2 2 2 4 10 Total 16 13 21 29 79
This table shows how the 79 corporations of the sample are distributed among indices (AEX, AMX, DAX and MDAX) and industries. The Industry Classification Benchmark (ICB) is used to categorize the corporations.
Table 2 gives insight into the relation between leverage ratios and industry classes in different time periods. Some industries have higher average leverage ratio’s than others and
TABLE 2 – Leverage and industry Industry Book leverage
in % Rank ↑ = 1 Market leverage in % Rank ↑ = 1 ∆leverage / book leverage in % Period 2002-2017
Oil and Gas 59.30 3 44.90 4 24.28
Basic materials 56.30 6 47.10 2 16.34 Industrials 67.00 1 52.40 1 21.79 Consumer goods 56.60 5 39.80 6 29.68 Healthcare 44.50 7 30.40 7 31.69 Consumer services 57.40 4 41.00 5 28.57 Telecommunications 61.40 2 45.20 3 26.38 Technology 42.40 8 24.00 8 43.40 Period 2002-2007
Oil and Gas 62.40 2 37.70 6 39.58
Basic materials 61.60 3 49.00 2 20.45 Industrials 67.40 1 53.80 1 20.18 Consumer goods 58.90 4 40.30 5 31.58 Healthcare 45.80 7 33.10 7 27.73 Consumer services 56.70 5 40.60 4 28.40 Telecommunications 56.60 6 42.60 3 24.73 Technology 42.30 8 26.60 8 37.12 Period 2010-2017
Oil and Gas 57.50 4 50.30 1 12.52
Basic materials 52.30 6 45.10 3 13.77 Industrials 66.50 1 49.20 2 26.02 Consumer goods 53.40 5 36.60 5.5 31.46 Healthcare 44.10 7 27.50 7 37.64 Consumer services 58.20 3 36.60 5.5 37.11 Telecommunications 62.20 2 43.70 4 29.74 Technology 41.00 8 19.50 8 52.44
15 differences exist among the periods. In the total sample period and in the pre-crisis period, corporations in the industrial sector have the highest book and market leverage ratios. In the post-crisis period, the industrial sector has the highest book leverage ratio, but the oil and gas sector has the highest market leverage ratio. In all periods, book and market leverage ratios of the technology and healthcare sector are lowest. When looking at the relative difference between book and market leverage, this is largest for the oil and gas sector in the pre-crisis period. Post-crisis, the opposite is true: the difference is smallest for the oil and gas sector. After the crisis, the technology sector shows the largest difference between book and market leverage.
TABLE 3 - Descriptive statistics of variables Book leverage Market leverage Tangi-bility Market-to-book Size Profita-bility Risk Age Period 2002-2017 Mean 0.566 0.416 0.224 2.549 22.282 0.081 0.020 52.147 Median 0.569 0.392 0.194 2.059 22.250 0.083 0.017 52.500 Maximum 1.279 0.979 0.727 49.962 26.615 0.686 0.150 69.000 Minimum 0.056 0.024 0.003 -23.842 14.904 -2.387 -0.021 32.000 St. deviation 0.177 0.217 0.162 2.545 1.859 0.117 0.014 5.476 Skewness -0.076 0.367 0.757 5.488 -0.319 -8.352 2.738 -0.195 Kurtosis 3.033 2.323 2.988 112.063 3.131 168.140 17.600 3.363 Jarque-Bera 1.263 51.693 119.007 623294.0 22.060 1429175. 12612.79 14.729 Probability 0.532 0.000 0.000 0.000 0.000 0.000 0.000 0.001 Observations 1245 1245 1245 1245 1245 1245 1245 1245 Period 2002-2007 Mean 0.575 0.425 0.250 2.536 22.024 0.083 0.021 51.305 St. deviation 0.170 0.218 0.165 1.834 1.969 0.155 0.017 6.504 Observations 458 458 458 458 458 458 458 458 Period 2008-2009 Mean 0.588 0.494 0.216 1.909 22.254 0.062 0.028 51.274 St. deviation 0.171 0.215 0.155 1.437 1.826 0.124 0.013 5.419 Observations 157 157 157 157 157 157 157 157 Period 2010-2017 Mean 0.554 0.390 0.207 2.717 22.476 0.084 0.018 52.977 St. deviation 0.183 0.212 0.159 3.118 1.761 0.076 0.010 4.461 Observations 630 630 630 630 630 630 630 630
16 Then, table 3 shows descriptive statistics of the independent and explanatory variables used in the regressions. The average leverage ratio is 56.6% with a standard deviation of 17.7%. Market leverage is on average lower: 41.6% with a standard deviation of 21.7%. The table also shows the average values and standard deviations of the variables in the period before and after the financial crisis. The average leverage ratio before the crisis was 57.5%, which is higher than the average ratio of 55.4% after the crisis. This difference is statistically significant at a 10% level. Market leverage shows a similar decline, from 42.5 to 39.0%. This decline is statistically significant at a 1% level. Such a decrease in leverage is consistent with the hypothesis that was formed in section 2.
Table 4 provides correlation statistics of the independent and explanatory variables. Book leverage and tangibility correlate positively, which is consistent with the theory from section 2. The negative correlations between book leverage and market-to-book ratio, book leverage and profitability, and book leverage and risk are also consistent with theory. A negative correlation coefficient was also expected for age. However, this correlation coefficient turns out positive. Size and book leverage also correlate positively in this sample. The correlation coefficients of market leverage with the explanatory variables have the same sign as correlation coefficients of book leverage with the explanatory variables, albeit in a varying extent. The correlation coefficient of market leverage with the market-to-book ratio is considerably larger than the correlation coefficient of book leverage with the market-to-book ratio, which makes sense because the values of both market leverage and the market-to-book ratio depend on the market value of equity. The negative correlation coefficient of market leverage with profitability is inflated relative to the correlation coefficient of book leverage with profitability and the negative correlation coefficient of market leverage with risk has become smaller.
TABLE 4 - Correlation statistics Book leverage Market leverage Tangi-bility Market-to-book Size Profita-bility Risk Age Book leverage 1.000 Market leverage 0.779 1.000 Tangibility 0.178 0.291 1.000 Market-to-book -0.031 -0.413 -0.183 1.000 Size 0.453 0.478 0.285 -0.181 1.000 Profitability -0.162 -0.319 -0.048 0.167 0.088 1.000 Risk -0.594 -0.363 -0.181 -0.040 -0.527 -0.134 1.000 Age 0.137 0.213 0.148 -0.096 0.339 0.037 -0.242 1.000
This table reports the correlation statistics of variables used in this study from 79 corporations listed on the AEX, AMX, DAX and MDAX in the years 2002 to 2017. All numbers are rounded to three decimal places.
17
5.
Results
In this section, regression results from the models described in the methodology section are discussed and interpreted using the literature described in the literature section.
Main model
Table 5 shows regression results from the main model (1) in which the post-crisis period is represented by the years 2010 until 2017. From the control variables, only three have statistical significance in explaining book leverage. The market-to-book ratio is one of them and has a positive coefficient, indicating that corporations with higher market-to-book ratios tend to have higher leverage ratios. This is not consistent with previous empirical evidence. However, it is consistent with pecking order theory. In this sample, when the market-to-book ratio increases with 1 percentage point, leverage increases with 0.01 percentage points. Profitability is the second and has a negative effect, indicating that corporations tend to have lower leverage ratios when they are more profitable. This is in accordance with previous empirical evidence. When profitability increases with 1 percentage point, leverage decreases by 0.22 percentage points.
TABLE 5 - Regression results model (1)
Variable Coefficient P-value
Constant 0.621 0.008 *** Tangibility -0.058 0.231 Market-to-book 0.009 0.008 *** Size 0.001 0.933 Profitability -0.217 0.000 *** Risk -2.938 0.000 *** Age 0.000 0.928 Post-crisis dummy 0.245 0.020 ** Crisis dummy 0.033 0.001 *** Tangibility post-crisis 0.010 0.729 Market-to-book post-crisis -0.007 0.065 * Size post-crisis -0.001 0.734 Profitability post-crisis -0.131 0.110 Risk post-crisis -3.617 0.000 *** Age post-crisis -0.002 0.015 ** Aggressive post-crisis -0.056 0.000 *** Adjusted R-squared 0.800 F-statistic 54.471 Prob(F-statistic) 0.000 *** Durbin-Watson stat 0.781
Total panel (unbalanced) observations 1245
18 The third control variable that is statistically significant is risk, with an expected negative coefficient. This indicates that when corporations are more risky, they tend to have lower leverage ratios. An increase in risk of 1 percentage point leads to an average decrease in leverage of 2.94 percentage points. The effects of market-to-book, profitability and risk are significant at a 1% level. Coefficients of tangibility, age and size are not statistically significant.
The coefficient of the crisis dummy has a positive value of 0.03 and is significant at a 1% level. Apparently, corporate leverage ratios were 0.03 percentage points higher during the crisis. This coefficient is not as large as the positive coefficient from the post-crisis dummy, which has a value of 0.25 and is significant at the 5% level. This seems to contradict the hypothesis from the literature section. However, when attention is given to the interaction of the control variables with the post-crisis dummy, a different image is created. From the interaction variables, only interaction of the post-crisis dummy with market-to-book, risk, age and an aggressive financial strategy have statistically significant coefficients. They indicate that there is a statistically significant difference between the period before and after the crisis in the coefficients of the control variables.
The coefficient of the interaction variable of market-to-book and the post-crisis dummy is negative, indicating that the positive effect of an increasing market-to-book ratio is weaker in the post-crisis period as compared to the pre-crisis period. The negative sign is consistent with previous empirical evidence provided by Harrison and Widjaja (2014). The relative explanatory power of the variable market-to-book is also less post-crisis. Apparently, growth opportunities are less important in explaining leverage after the crisis. In the post-crisis period, the coefficient is only 0.002 and this coefficient may not be statistically significant at all.
The interaction effect of risk with the post-crisis dummy is also negative. Post-crisis, an increase in risk of 1 percentage point leads to an average decrease in leverage of 6.56 percentage points, as compared to 2.94 percentage points pre-crisis. The relative explanatory power of risk is also larger in the period after the financial crisis as compared to the period before the financial crisis. The credit supply shock theory states that investors of capital flight to quality (i.e. security) during the financial crisis. That the relative explanatory power of risk has increased post-crisis could be an indication that this still has an effect on leverage after the crisis.
The interaction variable of age and the post-crisis dummy has a negative coefficient too, indicating that the influence of age was significantly more negative in the post-crisis period compared to the pre-crisis period. This can be seen as an indication that there is indeed a change in risk perception. However, the total coefficient of age is still only -0.002 post-crisis and this total coefficient may well be statistically insignificant.
19 be consistent with the assumption that investors of capital demand more quality after the crisis.
When the coefficients of the post-crisis dummy and the significant interaction variables are all taken into consideration, the accumulated effect of the post-crisis period on book leverage is negative when control variables do not change in value (ceteris paribus). This is consistent with the hypothesis that was formed in section 2. In whole, the model is statistically significant at a 1% level and with an adjusted R-squared of 0.80, it has a good fit.
Market leverage
Will these results change a lot when market leverage is chosen as dependent variable instead of book leverage? Table 6 shows the regression results from regression model (2). Coefficients of five out of six control variables show statistical significance. The coefficients of tangibility, market-to-book, size, profitability and risk are all statistically significant at a 1% significance level, even though they do not always have the expected sign. Tangibility, insignificant in model (1), has a negative coefficient where a positive coefficient was expected. In this sample, when the tangibility ratio increases with 1 percentage point, market leverage decreases with
TABLE 6 - Regression results model (2)
Variable Coefficient P-value
Constant 1.191 0.000 *** Tangibility -0.203 0.002 *** Market-to-book -0.042 0.000 *** Size -0.027 0.033 ** Profitability -0.331 0.000 *** Risk -2.037 0.000 *** Age 0.001 0.271 Post-crisis dummy -0.268 0.009 *** Crisis dummy 0.052 0.000 *** Tangibility post-crisis 0.180 0.000 *** Market-to-book post-crisis 0.033 0.000 *** Size post-crisis 0.012 0.001 *** Profitability post-crisis -0.310 0.002 *** Risk post-crisis 0.707 0.293 Age post-crisis -0.002 0.032 ** Aggressive post-crisis -0.031 0.006 *** Adjusted R-squared 0.828 F-statistic 65.332 Prob(F-statistic) 0.000 *** Durbin-Watson stat 0.956
Total panel (unbalanced) observations 1245
20 0.20 percentage points on average. This is more in line with pecking order theory than with trade-off theory. Market-to-book, size, profitability and risk all have negative coefficients too, as expected. Size was insignificant in model (1) and the sign of the market-to-book coefficient has changed compared to model (1). According to model (2), when the market-to-book ratio increases with 1 percentage point, market leverage decreases with 0.04 percentage points on average and when net sales increase with 1%, market leverage decreases with 0.03 percentage points times 1% on average. Profitability and risk were significant in model (1) too and have the same sign as in that model. The effect of age remains statistically insignificant.
The crisis and post-crisis dummy are both statistically significant at a 1% level in this model. The crisis dummy has a similar effect on market leverage as on book leverage, albeit in a different extent. During the crisis period, market leverage was on average 0.05 percentage points higher than in the pre-crisis period, as compared to 0.03 percentage points in model (1). This makes sense, considering that the market value of equity decreased during the crisis years, thereby increasing market leverage. The post-crisis dummy however, has a different effect on market leverage. The sign has changed compared to model (1) and the coefficient now takes a value of -0.27, indicating that during the post-crisis period, market leverage was on average 0.27 percentage points lower than in the pre-crisis period. This seems to support the hypothesis from section 2. Six out of seven interaction variables have statistically significant coefficients too. The coefficients of the interaction of the post-crisis dummy with tangibility, market-to-book, size, profitability and the aggressive financial strategy dummy are statistically significant at a 1% level. The interaction effects of the post-crisis dummy with size and tangibility were insignificant in model (1). The influence of age remains statistically significant at a 5% level and the interaction effect of the post-crisis dummy with risk has lost all significance in this model, despite being of great importance in model (1).
The interaction effect of tangibility with the post-crisis dummy is positive, indicating that the negative influence of tangibility is weaker post-crisis. The relative explanatory power of the tangibility ratio has decreased as well. An increase in tangibility of 1 percentage point post-crisis, only leads to an average decrease in market leverage of 0.02 percentage points, as compared to 0.20 percentage points pre-crisis. This is consistent with an enduring effect of the supply shock theory in that tangible assets can help attract external financing after the crisis because they can serve as collateral in an environment where investors search for quality.
21 The negative effect of size also seems weaker in the post-crisis period. Because the interaction variable of the post-crisis dummy with size has a positive coefficient, the effect of a 1 percentage point increase in net sales leads to a 0.02 percentage point times 1% decrease in market leverage post-crisis, as compared to 0.03 percentage points times 1% pre-crisis. The relative explanatory power of the variable size decreased as well in the period after the crisis.
The coefficient of the interaction variable of profitability with the post-crisis dummy is negative, indicating that the negative effect of increasing profitability is significantly stronger post-crisis. An increase in profitability of 1 percentage point leads to an average decrease in market leverage of 0.64 percentage points post-crisis, as compared to 0.33 percentage points pre-crisis. The increase is not only in terms of size of the effect, but also in terms of relative explanatory power. This is consistent with pecking order theory: more profitable corporations generate their own funding so there is no need to attract external financing. After the financial crisis, the preference to be self-sufficient in financing may be even larger because of the increase in information asymmetry between insiders and outsiders.
The financial strategy before the crisis helps to explain market leverage post-crisis. Corporations that followed an aggressive financial strategy pre-crisis experienced an extra decrease in leverage of 0.03 percentage points post-crisis. The effect was similar in model (1). Overall, the accumulated effect of the post-crisis period on market leverage is negative when control variables do not change in value (ceteris paribus). This means that this result is robust to whether market leverage or book leverage is chosen as dependent variable. The post-crisis dummy by itself contributes a lot to this effect in this model, where it did not in model (1). Model (2) is statistically significant at a 1% level and has an adjusted R-squared that with a value of 0.83, is similar to the adjusted R-squared of model (1).
Crisis years defined differently
In model (3), the crisis and post-crisis period are defined differently than in model (1). The crisis period is extended with one year, 2010, and the post-crisis period starts later, in 2011. Table 7 presents the regression results of model (3). The results are similar to the results from the main model. Coefficients do not show considerable changes in values and signs have not changed. It is interesting to see what happens to the crisis dummy. The coefficient of the crisis dummy was statistically significant at a 1% level in model (1) where it is only statistically significant at a 10% level in this model. This could be an indication that 2010 is correctly considered post-crisis in model (1).
22 As a whole, the model is statistically significant at a 1% significance level and the adjusted R-squared is similar to that of model (1). All in all, the results of model (1) seem to be robust to the decision whether or not to include 2010 in the crisis years or in the post-crisis period.
TABLE 7 - Regression results model (3)
Net a Variable Coefficient P-value
Constant 0.610 0.011 *** Tangibility -0.059 0.222 Market-to-book 0.010 0.005 *** Size 0.001 0.916 Profitability -0.221 0.000 *** Risk -2.646 0.000 *** Age 0.000 0.948 Post-crisis dummy 0.300 0.005 *** Crisis dummy 0.013 0.098 * Tangibility post-crisis 0.010 0.732 Market-to-book post-crisis -0.007 0.043 ** Size post-crisis -0.003 0.471 Profitability post-crisis -0.160 0.051 * Risk post-crisis -3.773 0.000 *** Age post-crisis -0.003 0.009 *** Aggressive post-crisis -0.055 0.000 *** Adjusted R-squared 0.796 F-statistic 53.091 Prob(F-statistic) 0.000 *** Durbin-Watson stat 0.841
Total panel (unbalanced) observations 1245
This table reports the regression results from model (3). Book leverage is the dependent variable. The sample consists out of 79 corporations over the time period from 2002 to 2017. The regression outcome is controlled for firm fixed effects. The post-crisis dummy takes value 1 in the years 2011-2017 and 0 otherwise. The crisis dummy takes value 1 in 2008, 2009 and 2010 and 0 otherwise. The model is estimated using White standard errors. ***, ** and * represent statistical significance of the coefficients at the 1%, 5% and 10% level respectively. All numbers are rounded to three decimal places.
Industry effects
Then in model (4), firm fixed effects are replaced by industry dummies and interaction variables between the industry dummies and the post-crisis period are added. Table 8 shows the regression results. Now that fixed effects are no longer controlled for, the coefficients of the control variables tangibility and size become statistically significant at a 10% and 1% level respectively, but do not have the expected signs. The market-to-book ratio loses all statistical significance in explaining leverage. Profitability and risk are significant at a 1% level and their coefficients are similar to but larger than in model (1). The coefficients of four industry dummies are statistically significant. All their signs are positive indicating that corporations in these sectors have higher average leverage ratios than corporations in the other sectors.
23 and is significant at a 1% level. The coefficients of the interaction variables of the post-crisis period and profitability, risk and age are statistically significant. The former was insignificant in model (1) and the total effect post-crisis of both profitability and risk is larger. Post-crisis, an increase in profitability of 1 percentage point leads to an average decrease in leverage of 0.59 percentage points in this model, as compared to 0.22 percentage points in model (1), and an increase in risk of 1 percentage point leads to an average decrease in leverage of 9.21 percentage points, as compared to 6.56 percentage points in model (1). The interaction variable of the post-crisis dummy and market-to-book has become insignificant in this model and the interaction variables of the post-crisis period and tangibility and size remain so.
TABLE 8 - Regression results model (4)
Variable Coefficient P-value
Tangibility -0.063 0.095 * Market-to-book 0.007 0.115 Size 0.021 0.000 *** Profitability -0.301 0.001 *** Risk -4.674 0.000 *** Age 0.001 0.154
Oil & Gas dummy 0.173 0.071 *
Basic materials dummy 0.174 0.070 *
Industrials dummy 0.226 0.015 **
Consumer goods dummy 0.140 0.135
Healthcare dummy 0.084 0.374
Consumer services dummy 0.146 0.122
Telecommunications dummy 0.187 0.052 * Technology dummy 0.104 0.293 Crisis dummy 0.037 0.004 *** Tangibility post-crisis 0.023 0.625 Market-to-book post-crisis -0.003 0.540 Size post-crisis -0.010 0.077 Profitability post-crisis -0.285 0.023 ** Risk post-crisis -4.534 0.001 *** Age post-crisis -0.004 0.010 ***
Oil & Gas dummy post-crisis -0.024 0.563
Basic materials dummy post-crisis -0.074 0.013 **
Industrials dummy post-crisis -0.026 0.308 Consumer goods dummy post-crisis -0.046 0.120 Healthcare dummy post-crisis -0.031 0.278 Consumer services dummy post-crisis -0.015 0.665 Telecommunications dummy post-crisis -0.009 0.810
Technology dummy post-crisis 0.034 0.005 *** Aggressive post-crisis 0.085 0.000 ***
Adjusted R-squared 0.599
Durbin-Watson stat 0.496
Total panel (unbalanced) observations 1245
24 In two industries, leverage has changed significantly in the period after the crisis as compared to the period before the crisis. The interaction effect of the post-crisis dummy with the ‘basic materials’ industry is negative and is significant at a 5% level. Apparently, the average leverage ratio in the basic materials industry is 0.07 percentage points lower in the period after the crisis as compared to the period before the crisis. In the technology sector, the opposite has happened: the average leverage ratio is 0.03 percentage points higher in the period after the crisis as compared to the period before the crisis. This coefficient is statistically significant at a 1% level.
Overall, model (4) does not provide us with evidence that contradicts the main result from model (1) and it therefore seems robust to the decision whether to include firm fixed effects or industry effects. Model (4) has a lot less explanatory power than model (1), with an adjusted R-squared of only 0.60 as compared to an adjusted R-squared of 0.80 of model (1). This is an indication that the industry effect is only one of more firm fixed effects that have explanatory power in determining leverage. By excluding the other firm fixed effects, the coefficients of other variables can become inflated due to an omitted variable bias. This would explain the larger coefficients for the effects of risk and profitability.
TABLE 9 - Regression results model (1) with data until 2015
Variable Coefficient P-value
Constant 0.616 0.014 ** Tangibility -0.064 0.244 Market-to-book 0.008 0.030 ** Size 0.001 0.913 Profitability -0.235 0.000 *** Risk -2.868 0.000 *** Age 0.000 0.892 Post-crisis dummy 0.236 0.041 ** Crisis dummy 0.031 0.002 *** Tangibility post-crisis 0.017 0.570 Market-to-book post-crisis -0.003 0.425 Size post-crisis -0.003 0.484 Profitability post-crisis -0.191 0.020 ** Risk post-crisis -3.887 0.000 *** Age post-crisis -0.002 0.120 Aggressive post-crisis -0.054 0.000 *** Adjusted R-squared 0.805 F-statistic 49.365 Prob(F-statistic) 0.000 *** Durbin-Watson stat 0.841
Total panel (unbalanced) observations 1088
25 Adjusted sample
As a last robustness check, model (1) is used again but this time the sample is adjusted, excluding data from the years 2016 and 2017. The results are shown in table 9. The coefficients of the interaction variables of the post-crisis dummy with market-to-book and age have lost statistical significance. The interaction variable of profitability with the post-crisis dummy has become statistically significant, indicating that increasing profitability has a stronger negative effect on leverage in the post-crisis years up until 2015, as compared to the pre-crisis years. Apart from that, not a lot has changed compared to the regression results from table 5. Coefficients do not show considerable changes in values. The adjusted R-squared of 0.81 is similar to the R-squared from table 5 and the model as a whole remains significant at a 1% level. The main result from table 5, that the accumulated effect of the post-crisis period on book leverage is negative (ceteris paribus), is robust to the decision whether or not to include 2016 and 2017 in the post-crisis period.
6.
Conclusion
This paper has tried to answer the question whether corporate capital structures have become more conservative after the financial crisis of 2007 with data from a subsample of listed corporations in Germany and the Netherlands. General theories on capital structure and theories on how the financial crisis could have had an effect on capital structure, are used to substantiate the hypothesis that leverage ratios have declined after the financial crisis of 2007. With these theories, a model is build that can provide an answer to the research question. Regression results from the model show that the accumulated effect of the post-crisis period on book leverage is indeed negative when control variables do not change in value (ceteris paribus). This is because certain variables that explain leverage have a stronger negative effect on leverage in the post-crisis period compared to the pre-post-crisis period. The stronger negative effect of corporate risk seems to contribute a lot and can be explained by investors of capital demanding more quality after the start of the financial crisis, a phenomenon of the credit supply shock theory. As a consequence, costs of capital and costs of possible financial distress increased and in such circumstances, leverage ratios are expected to decline, consistent with trade-off theory. Corporations that followed an aggressive financial strategy before the crisis, experienced an extra decline in their leverage ratios. This too can be explained by investors of capital demanding more quality after 2007. However, when interpreting these results one should bare in mind that results may be consistent with one theory even when they are actually generated by another (Myers, 2001).
26 shortened. Despite some differences in regression results, these tests found no empirical evidence that the hypothesis should be rejected.
It was mentioned in the introduction that the financial crisis could have a long-lasting effect on leverage because of a change in risk perception of the ones in charge. No clear-cut evidence was found to support this theory. However, the proxy that is used to measure risk perception, age, is very crude and may not be tenable at all. Therefore, it may still be very interesting to look at the possible long-lasting effect of the financial crisis on risk perception and thereby on leverage in future research.
Another limitation of the paper is the amount of corporations included in certain industrial classes of the dataset. Especially the oil and gas sector and the telecommunications sector include only few corporations. This could have biased results, especially in model (4), because the characteristics of the included corporations have a large influence in these sectors but may not be representative for the sector.
References
Erel, I., Julio, B., Kim, W., Weisbach, M. 2012. Macroeconomic conditions and capital raising. The Review of Financial Studies 25, 341-376.
Frank, M., Goyal, K. 2009. Capital structure decisions: which factors are reliably important? Financial Management 38, 1-37.
Hackbarth, D. 2008. Managerial traits and capital structure decisions. The Journal of Financial and Quantitative Analysis 43, 843-881.
Harris, M., and Raviv, A. 1991. The theory of capital structure. The Journal of Finance 46, 297-355.
Harrison, B., Widjaja, T. 2014. The determinants of capital structure: comparison between before and after financial crisis. Economic Issues 19, 55-82.
Iqbal, A., Kume, O. 2014. Impact of financial crisis on firms’ capital structure in UK, France, and Germany. Multinational Finance Journal 18, 249-280.
Kahle, K., Stulz, R. 2013. Access to capital, investment, and the financial crisis. Journal of Financial Economics 110, 280-299.
Lemmon, M., Roberts, M., Zender, J. 2008. Back to the beginning: persistence and the cross- vooralsection of corporate capital structure. The Journal of Finance 63, 1575-1608.
27 Malmendier, U., Tate, G., Yan, J. 2011. Overconfidence and early-life experiences: the effect of managerial traits on corporate financial policies. The Journal of Finance 66, 1687-1733. Malmendier, U., Nagel, S. 2011. Depression babies: do macroeconomic experiences affect risk taking? The Quarterly Journal of Economics 126, 373-416.
Mendoza, E. 2010. Sudden stops, financial crisis, and leverage. American Economic Review 100, 1941-1966.
Modigliani, F., Miller, M. 1958. The cost of capital, corporate finance, and the theory of investment. American Economic Review 48, 261-297.
Moerland, P. 1995. Alternative disciplinary mechanisms in different corporate systems. Journal of Economic Behavior and Organization 26, 17-34.
Myers, S. 2001. Capital structure. Journal of Economic Perspectives 15, 81-102.
Rajan, R., Zingales, L. 1995. What do we know about capital structure? Some evidence from international data. The journal of Finance 5, 1421-1460.
28
Appendix I – variables
Symbol Variable Database/
calculated Measure ment Datastream code Description
AGE Age Calculated Annual Average of CEO age and CFO
age per year end, in absence of either of them only the one that is known is used to determine age.
LEV Book leverage Calculated Annual The book value of liabilities
divided by the book value of assets, per year end.
Book value of liabilities
Datastream Annual WC03351 The book value of liabilities
represents all short and long term obligations expected to be satisfied by the company, per year end.
Book value of assets
Datastream Annual WC02999 The book value of assets
represents the sum of total current assets, long term receivables, investment in unconsolidated subsidiaries, other investments, net property plant and equipment and other assets, per year end.
Book value of equity
Calculated Annual The book value of assets
minus the book value of liabilities, per year end.
CEO age BoardEx Annual Age per year end of the CEO
or when absent, the chairman of the executive board.
CFO age BoardEx Annual Age per year end of the CFO
or when absent, the one in charge of finance in the executive board.
EBIT Datastream Annual WC18191 Earnings before interest and
tax (EBIT) represent the earnings of a company before interest expense and income taxes. It is calculated by taking the pre-tax income and adding back interest expense on debt and subtracting interest capitalized.
Industry Datastream Fixed WC07040 ICB code represents an
industry code within the Industrial Classification Benchmark (ICB).
MLEV Market
leverage
Calculated Annual Market leverage is calculated
