Master Thesis
The Global Financial Crisis and the Determinants of
Capital Structure in the United Kingdom
by Daan Aarnink
MSc. Finance
Faculty of Economics and Business
University of Groningen
Author: Daan Aarnink Student nr.: s1888668 Phone: +31 6 55141228
Email: daanaarnink@gmail.com Date: January 16, 2015
1
The Global Financial Crisis and the Determinants of
Capital Structure in the United Kingdom
Daan Aarnink Abstract
This paper studies the determinants of capital structure and the role of the global financial crisis by performing panel data analysis on UK listed firms over the period from 2004 to 2013. Based on the trade-off theory and pecking order theory I test the relation between leverage and asset tangibility, profit-ability, firm size, and growth opportunities. I find a positive relation for tan-gibility and firm size, and a negative relation for profitability. Furthermore, the global financial crisis strengthens these relations. By using Altman’s z-score, I find that financial distressed firms have relatively high debt ratios, compared to healthy firms.
Keywords: Capital Structure, Leverage, Global Financial Crisis, UK. JEL classification: G01, G30, G32
1. Introduction
2
The greater part of the empirical literature about capital structure focuses on the United States. Additionally, several comparison articles exist, with differences and similari-ties of capital structure in countries worldwide. Fan, Titman, and Twite (2012) study these differences over 39 developed and developing countries, with varying results. Rather nar-rowed down is the study by Rajan and Zingales (1995), which compares the G7 countries (United States, Japan, Germany, France, Italy, United Kingdom, and Canada). They find significantly different and surprising results for firms in the United Kingdom and Germany, which on average have lower leverage ratios. Because of these unexpected results and the high availability of financial data (in contrast to Germany), the United Kingdom is used as input for this paper.
The remainder of this paper is organized as follows. In section 2, I provide the prior literature and seminal theories on capital structure. In section 3, I explain the hypotheses and methodology which I use to answer the hypotheses, while in section 4, I describe the data, the sample selection process, and descriptive statistics. In section 5, I present the main results, which I discuss in the final section, where I also report the conclusions and limitations. Finally, I report additional statistics and tests in the appendices.
2. Literature Review
In this section, I try to answer my research question by discussing seminal theories in the existing literature, together with the findings and evidence from related articles. Fur-thermore, I discuss the determinants of capital structure, and provide indications of ex-pected results.
2.1. Capital Structure Theories
as-3
sumptions, the debt to equity ratio should not have influence on a firm’s market value. This theory is used as starting point for the most important papers of capital structure. By con-sidering taxes, Modigliani and Miller (1963) introduce the trade-off theory, where a target leverage amount is determined based on the tax benefits of debt and the bankruptcy costs of debt. A highly profitable firm which is unlikely to be in financial distress, thus facing low bankruptcy costs, is able to be highly levered. In contrast, firms with high costs of financial distress (e.g., as a consequence of a crisis), have to adjust their leverage, to compensate for these costs.
The agency theory is closely related to the trade-off theory, and thoroughly described by Jensen (1986). This theory describes how benefits and costs of debt are being influenced by agency costs. Well known examples are asset substitution (Jensen and Meckling, 1976), where high risk, low return projects are preferred over low risk, high return projects, and debt overhang (Myers, 1977), rejecting profitable projects because of extensive debt levels. The influences of these issues on capital structure are in particular studied by Harris and Raviv (1991).
The pecking order theory is generally considered as the most important theory chal-lenging the trade-off theory. Myers (1984) states that a firm follows the pecking order theo-ry when it prefers internal financing to external financing, and in case of the external fi-nancing, it favors debt over equity. The most common motivation for the pecking order theo-ry is adverse selection. Myers and Majluf (1984) describe in their model how the adverse se-lection problem increases the costs of external financing, due to managers having more in-formation about the firm’s assets and its growth prospects than investors. When managers are selling their assets, investors attempt to reveal the reason for selling these assets. In most of the times, managers of an overvalued company are willing to sell their equity, while managers of an undervalued company will not.
4
funds, but a firm’s current debt to equity ratio is an evolving result of former attempts to time the market.
Although the findings of Baker and Wurgler (2002) seem credible, several articles question their market timing hypothesis. For example, Hovakimian (2006) finds contradict-ing results in his study about equity market timcontradict-ing and capital structure. He states that the influence of past equity issues are rather small and do not have long term effects. Therefore, he finds no convincing evidence of the market timing hypothesis correctly describing the capital structure policy. Furthermore, Kayhan and Titman (2007), also question the find-ings of Baker and Wurgler (2002), by employing an alternative market timing measure. They argue that although the measure used by Baker and Wurgler (i.e., the average mar-ket-to-book ratio), has a significant effect and in the right direction, this merely is caused by other implicit aspects of market-to-book ratio (e.g., growth opportunities).
Hence, considering the strong evidence against the market timing hypothesis, I ex-clude this theory from my analysis. Thereby, I emphasize on the most important debate in the capital structure literature, which is the discussion of the trade-off theory versus peck-ing order theory.
2.2. Empirical Evidence and Determinants of Capital Structure
In the empirical literature there is no consistency about the capital structure deter-minants. In this section, I analyze the most relevant literature, summarize all the corre-sponding factors and compare their similarities and dissimilarities. Finally, I decide on a set of core factors which have the most explanatory power on determining a firm’s capital struc-ture.
5
on capital structure. These variables are asset tangibility, profitability, firm size, and growth opportunities, which are equal to the variables studied by Rajan and Zingales (1995). Additionally, to capture the effect of the financial crisis and to explore a firm’s finan-cial situation, the chance of a firm being in finanfinan-cial distress is added.
Rajan and Zingales (1995) test the relation between leverage and four factors, asset tangibility, profitability, firm size and growth opportunities. Frank and Goyal (2009) expand their study by including several other minor factors (e.g., expected inflation, and stock mar-ket conditions). Out of all variables possibly influencing capital structure decisions, they form a set of core factors. The core factors are estimated with book and market leverage. One key finding is the declining importance of profits. In the period before the 1980s, profits played a powerful role in determining leverage. In the later period, profits became less im-portant in leverage decisions. The reason for this decline is that, during the 1980s and 1990s, equity markets became more willing to fund unprofitable firms with good growth prospects. Hence, the determinants change over time, and most likely have changed more after this period. To extend the model in Frank and Goyal (2009), I analyze the relation be-tween the factors and leverage for the period succeeding their sample period, which is from 2004 to 2013. This period will be divided into two sub periods, one before and one during the financial crisis (i.e., from 2004 to 2007, and from 2008 to 2013). In this way I attempt to re-veal the effects of the financial crisis on the corporate financial policy decisions.
2.2.1. Asset Tangibility
6 2.2.2. Profitability
The theory is somewhat ambiguous about the relation between profitability and lev-erage. According to the trade-off theory there is a positive relation between leverage and profitability. Firms with large profits face lower financial distress. Since financial distress is seen as a cost of debt (Scott, 1976), firms can profit more from the tax benefits of debt by raising their debt levels. High debt levels can also mitigate the free cash flow problem which arise as a consequence of large profits, as described by Jensen (1986). Additionally, lenders are more willing to do business with firms with high cash flows. On the other hand, the pecking order theory predicts a negative relation between profitability and leverage, since high profit firms tend to have large amounts of equity and prefer exploiting these internal funds over external finance.
2.2.3. Firm Size
The theories are also contradicting about the relation between a firm’s size and its amount of leverage. The trade-off theory predicts that large firms generally are more diver-sified and can offset unprofitable company divisions with profitable ones. Furthermore, larger firms face lower agency costs of debt, since they are more mature, more experienced and have built up a better reputation. Therefore the trade-off theory suggests that there is a positive relation between leverage and firm size. On the contrary, according to Rajan and Zingales (1995), firm size can be a proxy for the information outside investors have, which implies a less severe adverse selection problem. A lower chance of adverse selection prob-lems results in a higher preference for equity relative to debt, consistent with the pecking order theory. Moreover, according to Van Binsbergen, Graham, and Yang (2010), small firms are not able to continuously adjust their leverage ratio because of the relatively high fixed costs. To minimize these costs, small firms use external finance less frequent but in larger amounts, resulting in higher debt amounts on aggregate. This concludes in a nega-tive relation between firm size and leverage.
2.2.4. Growth
7
agency theories support this negative relation. Myers (1977) describes the debt overhang problem, which indicates that highly levered firms have to reject positive net present value projects because equity holders get only a small return on their investment. Firms with high growth opportunities should therefore not focus on external financing. Likewise, equity holders have a preference to invest in highly volatile yet negative net present value projects to capture a portion of the project’s benefits as well, instead of being obliged to pass on the complete return to the debt holders. This asset substitution problem, introduced by Jensen and Meckling (1976), becomes even more severe in firms with high growth opportunities, resulting in higher costs of debt. Furthermore, according to the theory of free cash flow, there is a positive relation between leverage and assets in place, since assets in place have high collateral value and therefore create free cash flow problems. As debt can mitigate the-se free cash flow problems, there is reason to believe that growth opportunities and leverage are negatively related. Van Binsbergen, Graham, and Yang (2010) state that firms are being restricted in its ability to optimally invest and exercise growth options, by the inflexibility and extra costs of debt contracts with external financers. In sum, there are several theories underpinning a negative relation between leverage and growth. On the other hand, follow-ing the peckfollow-ing order theory, firms with high growth options and a large number of invest-ments, should build up debt over time. When profitability is left constant, this results in a higher leverage ratio. Therefore the pecking order theory suggests there should be a positive relation between growth opportunities and leverage.
2.2.5. Financial Distress
8 2.3. Leverage
In the empirical literature, different leverage measures have led to varying and con-tradicting results. In their benchmark study, Rajan and Zingales (1995) decide on choosing only one leverage measure, while Frank and Goyal (2009) analyze four different measures for leverage. Welch (2004) states that not the firm’s debt policy, but the variation in a firm’s market value has the biggest influence on the variation of its market leverage. According to Graham and Harvey (2002), managers focus on leverage based on book values rather than leverage based on market values, when making capital structure decisions. On the contrary, Serfling (2014) states that, when determining a firm’s target debt level based on theoretical predictions, market leverage is the better measure. Frank and Goyal (2009) capture in their approach the most important aspects of leverage, which are maturity of debt and the differ-ences in accounting perspective, by using four different measures of leverage. Overall, as in most empirical literature, I will decide on one key measure of leverage, and check this measure for robustness with several other leverage measures.
2.4. Global Financial Crisis
9 2.5. Empirical Evidence from the UK
The UK financial market is broadly similar compared to the US market. The key dif-ferences are the legal tax system and the corporate market size (Rajan and Zingales, 1995). Not surprisingly, the amount of empirical literature about the capital structure decisions in the UK falls short compared to the US. However, similarly to the US, the evidence from UK firms is inconclusive as well. While several papers find significantly important factors which determine the capital structure composition, neither the trade-off theory nor the pecking order theory independently explains the relation between leverage and the factors ade-quately, according to Beattie, Goodacre, and Thomson (2006). Adedeji (2002) compares the trade-off theory and pecking order theory based on UK data, and finds a significantly posi-tive relation between growth and leverage, contrary to the trade-off theory. Ozkan (2001), finds significantly negative relations between leverage and growth opportunities and lever-age and profitability, and a slightly positive relation for firm size. These findings correspond with the findings from US studies, e.g., Lemmon, Roberts, and Zender (2008), and Serfling (2014).
3. Methodology and Hypotheses
In this section, I explain the general methodology, discuss the actual beginning of the global financial crisis and formulate the hypotheses. In addition, I discuss the robustness checks.
3.1. General Method
In the standard model the measure of leverage is based on total debt and book val-ues, which I regress on tangibility, profitability, firm size, and growth opportunities. Eq. (1) shows the standard model, which I estimate using the panel least squares method. Leverage based on book values, , is calculated as
(1) where is the intercept, x are the parameters to be estimated and finally, which
en-10
dogeneity, the explanatory variables are lagged by one period. The effects of variables which are not included in the regression, but do have influence on both dependent and independ-ent variables, are reduced by using differindepend-ent time periods. Hereby, the effect of omitted var-iable bias is minimized. Furthermore, to correct for the effect of heteroskedasticity, I use White cross-section standard errors in all regressions. To check for the possible presence of multicollinearity I construct a correlation matrix of all variables, which you can find in Ap-pendix B. Table B.I shows there are no variables with high correlations, which means that the multicollinearity problem can be neglected1. To test whether fixed effects or random
fects are appropriate, I perform a Hausman test, and to test whether cross-section fixed ef-fects and/or period fixed efef-fects are appropriate, I use the Redundant Fixed Efef-fects test. By including cross-section fixed effects I control for time-invariant omitted variables that affect cross-sectionally (e.g., headquarter location). The inclusion of year fixed effects con-trols for macroeconomic trends that vary over time but not cross-sectionally (e.g., govern-ment regulations, and inflation). Adjusting Eq. (1) for the above govern-mentioned steps results in the following equation. Leverage, , is calculated as
(2) where represents the year fixed effects and represents the firm fixed effects. 3.2. Global Financial Crisis
There are different thoughts about the actual cause of the current financial crisis. Moreover, the origin from where it all started is ambiguous as well. Although the details about the global financial crisis’ origin are beyond the scope of this paper, it is necessary to decide when the global financial crisis hit the UK financial market.
To determine the cut-off point of the pre-crisis period and the period of financial cri-sis itself, I use the gross domestic product (GDP) in the United Kingdom. Fig. 1 reports the quarterly change of UK’s GDP for the period 2004 to 2014. The relative change is based on constant prices of GDP, that is adjusted for inflation, (i.e., true growth in GDP). The data
1 The correlation between variables PROF1 and PROF2, is quite high, 0.889. Normally, this
11
are collected from the Organization for Economic Co-operation and Development (OECD).2
Note that the last period of growth (0.3%) is realized in the first quarter of 2008. The second quarter of 2008 shows a minor deterioration of 0.2%, followed by a serious downfall of −1.7% and −2.2% in the third and fourth quarter of 2008, respectively. Hence, I conclude that the period from 2004 to the end of 2007 can be seen as a period of economic upturn and will be denoted as the pre-crisis period. The actual beginning of the global financial crisis in the UK lies in the first months of 2008, therefore, the period from 2008 to 2013 can be seen as a pe-riod of economic downturn and is denoted as the financial crisis pepe-riod.
Figure 1. UK GDP % change, quarterly. This figure shows the quarterly change of the gross domestic product (GDP) in the United Kingdom. The data are collected from the OECD database. There is a clear decline starting in quarter two of 2008 (−0.2%). The next period of growth is realized in Q3-2009. The period with the highest growth is Q4-2005 (1.4%), whereas the period of highest downfall is Q4-2008 (-2.2%).
3.3. Leverage Measure
All leverage measures used in the empirical literature, can be categorized into four measures. From these four, I choose total debt over book value of assets as my main lever-age measure. Furthermore, I distinct between short term and long term debt, and between
2 The OECD data are from stats.oecd.org
12
book value and market value of assets. The distinction between book value of assets versus market value of assets checks for differences in a historical perspective and a forward look-ing perspective. The second distinction is made in the debt value. Short term obligations have marginal effect on a firm’s long term perspective and its continuity; furthermore, it can be relatively easily managed. Including the short term liabilities in the leverage measure can thus bias the estimates. Therefore, leverage is measured based on total debt, including long term debt and short term debt, and based on long term debt only. Hence, the four measures of leverage are, (1) total debt divided by book value of assets (TDA), (2) total debt
divided by the market value of assets (TDM), (3) long term debt divided by the book value of
assets (LDA), and (4) long term debt over the market value of assets (LDM).
3.4. Explanatory Variables and Hypotheses
The following variables are the set of core factors on which leverage is regressed. Based on existing literature and seminal theories, I predict the relation of these factors with leverage. Furthermore, for each factor I define a hypothesis about the influence of the finan-cial crisis on this relation. Detailed variable definitions and calculations are given in Ap-pendix A.
3.4.1. Asset Tangibility
In most literature the fixed assets to total assets ratio is positively related to lever-age. When debt increases, the portion of fixed assets increase as well. When companies are in financial distress, managers want to compensate for the extra amount of risk. Therefore, the relation of the fixed assets and leverage is expected to be stronger during a financial cri-sis and in a period of financial distress. Hence, I state the following hypotheses.
Hypothesis 1a: There is a positive relation between the firm’s tangibility of assets and its amount of leverage.
Hypothesis 1b: The financial crisis has a positive effect on the relation between the firm’s as-set tangibility and its amount of leverage.
3.4.2. Profitability
13
of financial distress and therefore can bear higher levels of leverage. The free cash flow problems which arise as a consequence of large profits can be mitigated by having larger debt amounts. Therefore, profitability is positively related with leverage. On the contrary, according to the pecking order theory, described in Myers and Majluf (1984), the high avail-ability of internal funds as a consequence of high profits will result in firms preferring these internal funds over debt. The overall alertness generated by a financial crisis mitigates the free cash flow problems. The inspection of firms, in particular firms in financial distress, is more thorough and severe. Hence, the preference for internal funds is more likely than us-ing extra debt. Therefore, I hypothesize that profitability and leverage are negatively relat-ed and that this relation strengthens during a financial crisis.
Hypothesis 2a: There is a negative relation between the firm’s profitability and its amount of leverage.
Hypothesis 2b: The financial crisis has a positive effect on the relation between the firm’s profitability and its amount of leverage.
3.4.3. Firm Size
Larger firms are more diversified and thus can bear higher levels of leverage. They have built up a reputation, have a larger network and are well known by investors. In case of financial distress, large firms profit even more from their reputation and network, with which they can mitigate the financial distress problems. Hence, I hypothesize that there is a positive relation between firm size and leverage and that this relation strengthens in case of financial distress.
Hypothesis 3a: There is a positive relation between firm size and leverage.
Hypothesis 3b: The financial crisis has a positive effect on the relation between firm size and leverage.
3.4.4. Growth
14
relation. However, the pecking order theory assumes profitability is kept constant, which is not likely. According to Rajan and Zingales (1995), the negative relation between growth options (i.e., market-to-book ratio) and market leverage is significant across all G7 countries in their sample. Considering the findings in existing literature and the weak pecking order theory, I state the following hypotheses.
Hypothesis 4a: There is a negative relation between the firm’s growth opportunities and its amount of leverage.
Hypothesis 4b: The financial crisis has a positive effect on the relation between the firm’s growth opportunities and its amount of leverage.
3.4.5. Financial Distress
To measure the chance of firms being in financial distress I use the traditional ver-sion of Altman’s z-score. This z-score, is a weighted average of five factors regarding the firm’s financial situation, and is calculated as
(3) where, is working capital divided by total assets, is retained earnings divided by total assets, is the EBIT over total assets, is the market value of equity divided by total liabilities, and is the net sales divided by total assets. When a firm scores below 1.81, it is considered to be in financial distress, whereas a score above 2.99 shows that a firm is healthy. Consequently, when firms score between the critical values (i.e., ), they belong in a gray area, indicating no financial distress, but not healthy either. Furthermore, according to Altman (1986), the critical value that discriminates best between bankrupt firms and non-bankrupt firms is equal to 2.675. This value is taken as cut-off point for the z-score dummy. The z-score dummy (D_ZSCORE) takes a value 1 for firms in financial distress (i.e., ), and is equal to 0 for healthy firms (i.e., ).
3.5. Robustness Checks
15
of assets, whereas market leverage is measured by long term debt over market value of as-sets.
4. Data and Descriptive Statistics
This section reports the data selection and construction of the data samples. Fur-thermore, the descriptive statistics are given and compared with those of the empirical lit-erature. Finally, the characteristics are put into context, which gives an indication of the expected results.
4.1. Sample Selection
To test the hypotheses I collect data from Bureau Van Dijk’s database Orbis. The da-ta are from publicly listed, industrial companies in the United Kingdom. The dada-ta are col-lected for the period from 2004 to 2013 on an annual basis, resulting in a total of 480 firms. Ideally, I would conduct the data set based on a more frequent basis, e.g., quarterly data. However, this is simply not possible, due to the lack of quarterly data in the Orbis database. I limit the data set on one country, and thereby rule out the country specific effects, which have great influence on a firm’s capital structure, according to Fan, Titman, and Twite (2012). Firms in financial and insurance activities are excluded from the data set, based on Eurostat classifications (i.e., NACE Rev. 2 divisions 64, 65, and 66), since the heavy regula-tion (e.g., capital requirements) restricts them in their capital structure decisions. Also ex-cluded from the data set are firms with a total assets value below $ 1 million, because these firms have a high probability of having unreliable financial data. Furthermore, I remove firms with missing data for calculating book leverage. To adjust for outliers, I winsorize the data at the 1st and 99th percentile. I choose to winsorize the data because this technique
re-sults in more significant estimation rere-sults and a better fit, compared to trimming (also at 1st and 99th percentile). These selections are summarized in Table 1, which gives an
16
Table 1
Sample Construction
This table summarizes the sample construction of the complete data set as of December 2nd 2014.
represents the dependent variable, which is total debt divided by book value of assets.
Sample No. companies
All active companies in United Kingdom 7,909,714
Sample including publicly listed companies only 2,094 Sample excluding financial and insurance companies based on
NACE rev. 2 codes 64, 65, and 66 1,511
Sample including firms with total assets of at least $1,000,000 604 Sample A Sample with nonmissing values for dependent ( ) variable
(i.e., book leverage)
480 No. firm-year obs. Sample B Total firm-year observations in pre-financial crisis period (i.e.,
2004-2007) 1,920
Sample C Total firm-year observations in financial crisis period (i.e.,
2008-2013) 2,880
Sample D Sample including financial distressed firms (i.e., z-score below
2.675) in pre-financial crisis period. 602
Sample E Sample including financial distressed firms (i.e., z-score below
2.675) in financial crisis period 1,388
4.2. Summary Statistics
The summary statistics of the leverage measures and explanatory variables are pre-sented in Table 2. The distinctions between market versus book values and between histori-cal perspective versus future perspective results in four measures of leverage. TDA (1) is the
total debt divided by total assets, TDM (2) is the total debt divided by the market value of
assets, LDA (3) is long term debt divided by the total assets value, and LDM (4) is the long
bank-17
rupt. The summary statistics correspond with the statistics in the empirical literature (Frank and Goyal, 2009; Korteweg, 2010; Serfling, 2014).
Table 2
Summary Statistics
This table provides summary statistics of all dependent and independent variables used in this paper for 2004-2013. TDA represents the total debt over the book value of total assets, TDM is the total debt divided by the market value of assets, LDA represents the long term debt divided by the book value of total assets, and LDM is long term debt divided by the market value of assets. GRWT1 is the market to book ratio as a proxy for growth, GRWT2 is the annual change of the logarithm of total as-sets, also as a proxy for growth. PROF1 is the return on assets based on net income to measure prof-itability, PROF2 is the ratio of earnings before interest and taxes over the book value of assets.
SIZE1 measures firm size by the log of total assets, and TANG1 measures tangibility by the portion
of fixed assets over total assets. ZSCORE is the traditional Altman’s z-score, and measures the chance of a firm going bankrupt. All variables are winsorized at the 1st and 99th percentile.
Variable Mean Med. Min. Max. Std. Dev. Obs.
Dependent Variables TDA 0.451 0.439 0.034 1.273 0.220 4,800 TDM 0.275 0.251 0.007 0.964 0.179 3,898 LDA 0.148 0.096 0.000 0.755 0.168 4,800 LDM 0.090 0.050 0.000 0.521 0.113 3,898 Explanatory Variables GRWT1 1.938 1.630 0.477 6.864 1.024 3,898 GRWT2 0.003 0.003 -0.036 0.056 0.014 4,320 PROF1 2.978 4.649 -57.573 28.620 12.369 4,766 PROF2 0.051 0.066 -0.721 0.342 0.144 4,800 SIZE1 19.672 19.533 15.203 24.997 2.230 4,800 TANG1 0.569 0.578 0.046 0.981 0.235 4,800 ZSCORE 3.004 2.584 -8.305 17.422 3.206 3,825
As shown in Fig. 1, the full sample period is divided into two subsample periods to observe the effects of the financial crisis, sample B from 2004 to 2007 and sample C from 2008 to 2013. Table 3 provides the mean and median values of the leverage measures for both sample B and sample C.
vari-18
ation is the decline in firm valuation by investors during the financial crisis. Since firms cannot adjust their debt levels as fast as investors can adjust their valuation, the portion of debt to market value increases.
Table 3
Descriptive Statistics of Leverage Measures for Different Sample Periods.
This table provides the mean and median values of the leverage measures for sample B and C, where sample B presents the period of economic upturn (i.e., 2004-2007), and sample C represents the peri-od of economic downturn (i.e., 2008-2013). The beginning of 2008 is used as cut-off point since the OECD data show a great downfall in the change of GDP in the first half of 2008. The difference in number of observations is due to higher data availability of book values. TDA represents the total debt over the book value of total assets, TDM is the total debt divided by the market value of assets,
LDA represents the long term debt divided by the book value of total assets, and LDM is long term
debt divided by the market value of assets. All variables are winsorized at the 1st and 99th percentile.
Leverage Mean Median No. Obs
‘04-‘07 ‘08-‘13 ‘04-‘07 ‘08-‘13 ‘04-‘07 ‘08-‘13
TDA 0.462 0.443 0.444 0.436 1,920 2,880
TDM 0.251 0.287 0.231 0.264 1,320 2,578
LDA 0.149 0.147 0.097 0.095 1,920 2,880
LDM 0.079 0.096 0.046 0.053 1,320 2,578
19 4.3. Altman Z-Score
In the empirical literature (e.g., van Binsbergen, Graham, and Yang, 2010; Lemmon, Roberts, and Zender, 2008), several modified versions of Altman’s z-score are used to meas-ure the chance of bankruptcy. A possible adjustment to the traditional measmeas-ure is to remove the leverage factor , from the equation, as this is also the independent variable in the es-timations. However, this also has consequences for the traditional critical values, which are not applicable any longer. The possibility to overcome this problem, by taking above median z-scores for healthy firms (see, van Binsbergen, Graham, and Yang, 2010), does not ade-quately illustrate the changes of healthy firms before the crisis compared to healthy firms after the crisis. Therefore, I calculate Altman’s z-score following the traditional formula, re-sulting in a total of 3,825 firm-year observations, with a mean of 3.00 and median of 2.58. Since firms are in financial distress with a z-score below 1.81, and considered healthy with a z-score above 2.99, the average firm in this sample seems to be doing well. Nonetheless, out of the 3,825 observations, 1,151 firm-years have a value below 1.81. Comparing the mean and median values of this z-score over different periods give confirmatory outcomes. These results are reported in Table 4.
Table 4
Altman Z-Score Statistics
This table shows the number of total firm-year observations with a known z-score in column (1), where column (2) reports the statistics only of firm-years with a z-score below 1.81, indicating finan-cial distress. Columns (3) and (4) show the differences between the last period of economic upturn and first period of economic downturn, only for firms with a z-score below 1.81. Note that the mean and median values are much lower, but there are also more firm-years with a z-score below the criti-cal value of 1.81. ‘04-‘13 ‘04-‘13 ‘05-‘07 ‘08-‘10 (1) (2) (3) (4) - Z < 1.81 Z < 1.81 Z < 1.81 Mean 3.00 0.32 0.41 0.33 Median 2.58 1.06 1.16 1.02 No. Obs. 3,825 1,151 265 468
20
This result indicates that on aggregate the z-score is much lower, but also that more firms are in financial distress.
Finally, Fig. 3 reports the correlation diagrams of book leverage with the four ex-planatory variables. These charts show the mean values of all variables and indicate that leverage is positively related to growth opportunities and tangibility, while negatively relat-ed to profitability and firm size.
Figure 3. Relation between book leverage and explanatory variables. This figure shows the relations between the average values of book leverage and the explanatory variables. TDA is total debt divided by book value of total assets, PROF1 is return on assets, GRWT1 is the market to book ratio as proxy for growth, TANG1 is the ratio of fixed assets divided by total assets, and SIZE1 is the logarithm of total assets. The correlations are based on the full sample. All variables are winsorized at the 1st and 99th percentile.
4.4. Industry effects vs. firm effects
deci-21
sions and can stronger differences. In light of these findings, I compare the average book leverage TDA, for all industries. The industry divisions are based on NACE Rev. 2 classifi-cations, which you see in Table 5. Excluded from the data set are firms in Section K, due to differences in capital requirements and different balance sheets, and firms in Section O, due to heavy governmental legislation. Table 5 shows that, except for Sections A and B, which consist of nearly 7% of the full sample, the mean values per industry do not deviate sub-stantially from the full sample mean. MacKay and Phillips (2005) find that the variation in leverage within industries is considerably higher than the leverage variation between in-dustries. While other studies (e.g., Lemmon and Zender, 2010; Fan, Titman, and Twite, 2012), show that controlling for industry effects or firm fixed effects does not result in signif-icant differences. Therefore, in this paper I emphasize on firm fixed effects and ignore in-dustry fixed effects.
Table 5 Industry Divisions
This table shows the industry divisions according to NACE Rev. 2 classifications. Firms in Section K and O are excluded from the data set due to fundamental differences (e.g., governmental legislation, capital requirements and balance sheet composition).
Section Divisions Title Mean Obs.
A 01 – 03 Agriculture, forestry and fishing 0.224 7
B 05 – 09 Mining and quarrying 0.233 26
C 10 – 33 Manufacturing 0.418 165
D 35 Electricity, gas, steam and air conditioning supply 0.471 5
E 36 – 39 Water supply and waste management 0.567 6
F 41 – 43 Construction 0.530 20
G 45 – 47 Wholesale and retail trade 0.519 37
H 49 – 53 Transportation and storage 0.548 17
I 55 – 56 Accommodation and food service activities 0.548 7
J 58 – 63 Information and communication 0.455 56
K 64 – 66 Financial and insurance activities 0
L 68 Real estate activities 0.436 28
M 69 – 75 Professional, scientific and technical activities 0.467 45 N 77 – 82 Administrative and support service activities 0.576 31
O 84 Public administration and defence 0
P 85 Education 0.432 1
Q 86 – 88 Human health and social work activities 0.406 5
R 90 – 93 Arts, entertainment and recreation 0.461 5
S 94 – 96 Other service activities 0.582 7
T 97 – 98 Activities of households as employers 0
U 99 Activities of extraterritorial organizations and bodies 0
X 00 Other 0.471 12
22
5. Results
In this section, I present the main regression results, compare them with the find-ings in the empirical literature, and, with these results, I attempt to answer the hypotheses. Furthermore, to check whether the estimates are robust, adjustments are made in the vari-ables and samples.
5.1. Empirical Results
Table 6 shows the parameter estimates of the main regression from Eq. (2) by using panel least squares. Column (2) shows the estimated coefficients of book leverage on the ex-planatory variables for the pre-crisis period, whereas column (3) shows the relation between book leverage and the explanatory variables for the financial crisis period. Columns (5) and (6) present the relation between market leverage and the explanatory variables for the pre-crisis period and the financial pre-crisis period, respectively. The parameter estimates for the full sample period, from 2004 to 2013, are represented in columns (1) and (4).
The profitability coefficient is slightly negative and significant, indicating that highly profitable firms have lower debt levels, consistent with empirical literature (Lemmon, Rob-erts, and Zender, 2008; Frank and Goyal, 2009; Serfling, 2014). Furthermore, for both book leverage as market leverage, the profitability estimation coefficient decreases after the start of the financial crisis, which indicates that in a period of financial distress profitable firms have relatively lower debt amounts than in the pre-crisis period. On the other hand, less profitable firms have relatively high debt levels and depend more on lenders in a period of financial crisis compared to a period of economic upturn. These results are correctly predict-ed by the pecking-order theory.
23
than the coefficients in the pre-crisis period, which indicates that the financial crisis has a positive effect on the relation between firm size and leverage.
Table 6
Parameter Estimates
This table shows the Panel Least Squares regression estimates of leverage and the explanatory vari-ables. Book leverage (TDA) is regressed on tangibility (TANG1), profitability (PROF1), firm size (SIZE1) and growth opportunities (GRWT1), for the full sample period 2004-2013 in column (1), and for the sub periods 2004-2007 and 2008-2013, in columns (2) and (3), respectively. Whereas market leverage (TDM) is regressed on the explanatory variables in columns (4), (5), and (6). Hausman tests and Redundant Fixed Effects tests are employed to check if Random Effects or Fixed Effects are ap-propriate. To minimize the effects of endogeneity, the explanatory variables are lagged by one period. Standard errors are corrected for heteroskedasticity. All variables are winsorized at the 1st and 99th
percentile. ***, **, and * denote significance at the 1%, 5% and 10% level, respectively.
Book Leverage (TDA) Market Leverage (TDM)
‘04-‘13 ‘04-‘07 ‘08-‘13 ‘04-‘13 ‘04-‘07 ‘08-‘13 (1) (2) (3) (4) (5) (6) TANG1 -0.040*** 0.019*** 0.026*** 0.004*** 0.072*** 0.035*** PROF1 -0.002*** -0.001*** -0.002*** -0.002*** -0.001*** -0.002*** SIZE1 -0.006*** -0.025*** 0.009*** 0.017*** 0.000*** 0.034*** GRWT1 -0.003*** 0.014*** -0.003*** -0.024*** 0.007*** -0.024*** Intercept 0.599*** 0.904*** 0.261*** -0.010*** 0.208*** -0.355***
Firm Fixed Effects Yes Yes Yes Yes Yes Yes
Year Fixed Effects Yes No3 Yes Yes Yes Yes
Adjusted R² 0.818 0.851 0.848 0.784 0.816 0.815
No. Obs. 3,435 904 2,531 3,425 901 2,524
F-Statistic 35.442*** 15.558*** 32.672*** 28.772*** 12.213*** 26.031*** Although the growth coefficient shows some insignificancy, its relation with leverage is severely affected by the financial crisis. The estimate of 0.014 in column (2) indicates an unexpected positive relation between growth options and leverage, however, this relation changes into a negative -0.003, for the financial crisis period. This effect is even stronger for market leverage, regarding the significantly negative estimate of -0.024, which is consistent to findings in the empirical literature. Hence, while in the pre-crisis period, firms with high growth opportunities are highly levered, during a financial crisis, firms with low growth
3 The Redundant Fixed Effects Test for restricting the year fixed effects to zero shows a
24
portunities have relatively large debt amounts. Consequently, the financial crisis has a neg-ative effect on the relation between the firm’s growth options and leverage.
With only one significant result, for both the basic estimation model and the robust-ness tests, it is difficult to interpret the relation between tangibility of assets and leverage. Panel least squares regression estimates, which are not corrected for firm fixed effects and year fixed effects, show a significantly negative relation between tangibility and leverage, which weakens during the financial crisis. Although, these estimates are highly significant (i.e., p-value < 0.001), they are not of great contribution, since the regressions are performed based on pooled OLS and thus not corrected for the appropriate effects.
However, since fixed effects have a major drawback, which is the exclusion of time-invariant variables, pooled OLS has to be performed when dummy variables are included in the regression model. Table 7 shows the pooled OLS regression estimates of leverage and the explanatory variables including the dummy for Altman’s z-score (D_ZSCORE).
The z-score dummy is fairly positive and significant on the 1% level for all estima-tions, which indicates that firms with a high chance of bankruptcy (i.e., ), have a higher leverage ratio than non-bankruptcy firms. This result is robust for regressing with market leverage instead of book leverage (i.e., both coefficients for the full sample are equal to 0.132). Surprisingly, the estimation coefficients of the explanatory variables show quite some differences compared to the coefficients in the basic model. Where in the basic model the relation between tangibility and leverage is positive, yet insignificant, in Table 7 it shows a significant negative relation. Hence, the inclusion of fixed effects, and thereby con-trolling for time-invariant variables, causes the change of relation-sign from positive to neg-ative between tangibility and leverage. In sum, according to the regression coefficients in Table 7, leverage is positively related to firm size, while negatively related to tangibility, profitability and growth opportunities. Therefore, with or without fixed effects, I find con-tradicting results compared to the empirical literature on asset tangibility.
The financial crisis has a slight influence on the relations between the four regres-sors and leverage, since the differences of the coefficients between columns (2) and (3) and between columns (5) and (6) in Table 7 are lower than the differences in Table 6. Finally, because of the exclusion of firm fixed effects and year fixed effects in the pooled OLS regres-sion, the adjusted R2 is significantly lower in the latter model.
25
Table 7 Financial Distress
This table shows the Panel Least Squares regression estimates of leverage and the explanatory vari-ables including a dummy variable for Altman z-score (D_ZSCORE), which has a value of 1 for firms with a z-score below 2.675 and a value of 0 for firms with a z-score equal to or higher than 2.675. Book leverage (TDA) is regressed on tangibility (TANG1), profitability (PROF1), firm size (SIZE1) and growth opportunities (GRWT1), for the full sample in column (1) and for the periods 2004-2007 and 2008-2013, in columns (2) and (3), respectively. Whereas market leverage (TDM) is regressed on the explanatory variables in columns (4), (5) and (6). Year fixed effects and firm fixed effects are not included since they rule out the influence of dummy variables. To minimize the effects of endogenei-ty, the explanatory variables are lagged by one period. Standard errors are corrected for heteroske-dasticity. All variables are winsorized at the 1st and 99th percentile. ***, **, and * denote significance
at the 1%, 5% and 10% level, respectively.
Book Leverage (TDA) Market Leverage (TDM)
‘04-‘13 ‘04-‘07 ‘08-‘13 ‘04-‘13 ‘04-‘07 ‘08-‘13 (1) (2) (3) (4) (5) (6) TANG1 -0.254*** -0.263*** -0.248*** -0.192*** -0.184*** -0.196*** PROF1 -0.001*** -0.001*** -0.002*** -0.001*** 0.000*** -0.001*** SIZE1 0.021*** 0.027*** 0.019*** 0.009*** 0.010*** 0.009*** GRWT1 -0.013*** -0.009*** -0.018*** -0.065*** -0.056*** -0.069*** D_ZSCORE 0.132*** 0.112*** 0.138*** 0.132*** 0.108*** 0.141*** Intercept 0.141*** 0.040*** 0.172*** 0.271*** 0.231*** 0.281***
Firm Fixed Effects No No No No No No
Year Fixed Effects No No No No No No
Adjusted R² 0.159 0.162 0.165 0.364 0.379 0.360
No. Obs. 3,358 888 2,470 3,357 888 2,469
F-Statistic 127.601*** 035.336*** 098.792*** 385.029*** 109.348*** 278.740***
5.2. Robustness Checks
coef-26
ficients as leverage based on total debt as regressand. In sum, estimation results of the basic model in are robust for different leverage measures.
Table 8
Leverage Robustness Tests
This table shows the Panel Least Squares regression estimates of leverage and the explanatory vari-ables, to check for the robustness of leverage measures, where leverage is measured by long term debt instead of total debt. Book leverage (LDA) is regressed on tangibility (TANG1), profitability (PROF1), firm size (SIZE1) and growth opportunities (GRWT1), for the full sample period 2004-2013 in column (1), and for the sub periods 2004-2007 and 2008-2013, in columns (2) and (3), respectively. Whereas market leverage (LDM) is regressed on the explanatory variables in columns (4), (5), and (6). Hausman tests and Redundant Fixed Effects tests are employed to check if Random effects or Fixed effects are appropriate. To minimize the effects of endogeneity, the explanatory variables are lagged by one period. Standard errors are corrected for heteroskedasticity. All variables are winso-rized at the 1st and 99th percentile. ***, **, and * denote significance at the 1%, 5% and 10% level,
respectively.
Book Leverage (LDA) Market Leverage (LDM)
‘04-‘13 ‘04-‘07 ‘08-‘13 ‘04-‘13 ‘04-‘07 ‘08-‘13 (1) (2) (3) (4) (5) (6) TANG1 0.064*** 0.023*** 0.070*** 0.035*** 0.027*** 0.044*** PROF1 -0.001*** -0.000*** -0.001*** -0.001*** -0.000*** -0.001*** SIZE1 0.010*** 0.005*** 0.023*** 0.014*** 0.005*** 0.027*** GRWT1 0.000*** 0.005*** 0.000*** -0.006*** 0.004*** -0.007*** Intercept -0.089*** 0.025*** -0.345*** -0.190*** -0.049*** -0.436***
Firm Fixed Effects Yes Yes Yes Yes Yes Yes
Year Fixed Effects Yes No4 Yes Yes Yes Yes
Adjusted R² 0.808 0.866 0.824 0.759 0.833 0.770
No. Obs. 3,453 904 2,531 3,425 901 2,524
F-Statistic 33.276*** 17.444*** 27.659*** 25.082*** 13.686*** 20.074***
4 The Redundant Fixed Effects Test for restricting the year fixed effects to zero shows a
27
6. Conclusion
6.1. Main Conclusion
This paper studies the effect of the global financial crisis and the capital structure determinants, by performing panel data analysis on 480 UK listed firms over the period from 2004 to 2013. I hypothesize that the financial crisis has a positive effect and thereby strengthens the relations between leverage and the capital structure determinants. Fur-thermore, I hypothesize that the direct and indirect costs of financial distress, as a direct effect of the financial crisis, will increase firms’ debt levels, thereby making firms more de-pendent on their internal funds.
Consistent with this hypothesis, I find a significant increase in the debt levels for all firms (i.e., firms in financial distress and healthy firms). Furthermore, in both the basic model as financial distress model, leverage is positively related to tangibility and firm size, while strongly negatively related to profitability, which is consistent to my hypotheses, and the findings of Lemmon, Roberts, and Zender (2008), Frank and Goyal (2009) and Serfling (2014). The relation between leverage and growth opportunities is ambiguous, though it tends towards the expected negative relation. For both book leverage as market leverage, the profitability coefficient decreases after the start of the financial crisis, which indicates that due to the crisis, profitable firms rely even more on their equity instead of external fi-nancing. This is consistent to the pecking order theory by Myers and Majluf (1984). On the other hand, less profitable firms, have relatively high debt levels and depend more on lend-ers in a period of economic downturn compared to a period of economic upturn. Therefore, I conclude that indeed the financial crisis caused a stronger negative relation between a firm’s profitability and its amount of debt. This effect is also visible for the relation between firm size. Furthermore, I find contradicting results for both the trade-off theory and pecking order theory, indicating that neither theory adequately predicts the capital structure policy. 6.2. Limitations and Further Research
correla-28
tions and more significant estimation results. When restricting the data with nonmissing market values, market price and shares outstanding, the total number of observations drops to 3,000 firm-year observations. In that case there are no missing values for the leverage variables. Again, this would increase the quality of the regression model. Regretfully, Orbis cannot provide sufficient market data to construct a data set with nonmissing values. Alt-hough these improvements will increase the quality of the analysis, I do not expect dramatic changes in estimation coefficients. Furthermore, by using the traditional Altman’s z-score, including the leverage factor, there is a possibility that the regression estimates including z-score, are incorrectly biased upwards. Further research could focus on Altman’s (1986) mod-els, and improve the critical values, in order for them to be applicable to current financial data.
Acknowledgements
I would like to thank Dr. Silviu Ursu for his helpful feedback and supervision throughout the entire process of writing my master’s thesis.
29
7. References
Adedeji, A., 2002. A cross-sectional test of pecking order hypothesis against static trade-off theory on UK data. Available at SSRN 302827.
Altman, E., 1986. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. Journal of Finance 23, 589-609.
Baker, M., Wurgler, J., 2002. Market timing and capital structure. Journal of Finance 57, 1-32.
Beattie, V., Goodacre, A., Thomson, S., 2006. Corporate financing decisions: UK survey evi-dence. Journal of Business Finance and Accounting 33, 1402-1434.
Campello, M., Graham, J., Harvey, C., 2010. The real effects of financial constraints: evi-dence from a financial crisis. Journal of Financial Economics 97, 470-487.
Fan, J., Titman, S., Twite, G., 2012. An international comparison of capital structure and debt maturity choices. Journal of Financial and Quantitative Analysis 47, 23-56. Frank, M., Goyal, V., 2009. Capital structure decisions: Which factors are reliably
im-portant? Financial Management 38, 1–37.
Graham, J., Harvey, C., 2002. How do CFOs make capital budgeting and capital structure decisions? Journal of Applied Corporate Finance 15, 8-23.
Harris, M., Raviv, A., 1991. The theory of capital structure. Journal of Finance 46, 297-355. Hovakimian, A., 2006. Are observed capital structures determined by equity market timing?
Journal of Financial and Quantitative Analysis 41, 221-243.
Hovakimian, A., Opler, T., Titman, S., 2001. The debt-equity choice. Journal of Financial and Quantitative Analysis 36, 1-24.
Jensen, M., 1986. Agency costs of free cash flow, corporate financing, and takeovers. Ameri-can Economic Review 76, 323-329.
Jensen, M., Meckling, W., 1976. Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305-360.
Kayhan, A., Titman, S., 2007. Firms’ histories and their capital structures. Journal of Fi-nancial Economics 83, 1-32.
Korteweg, A., 2010. The net benefits to leverage. Journal of Finance 65, 2137-2170.
30
Lemmon, M., Zender, J., 2010. Debt capacity and tests of capital structure theories. Journal of Financial and Quantitative Analysis 45, 1161-1187.
MacKay, P., Phillips, G., 2005. How does industry affect firm financial structure? Review of Financial Studies 18, 1433-1466.
Miller, M., 1977. Debt and taxes. Journal of Finance 32, 261-275.
Modigliani, F., Miller, M., 1958. The cost of capital, corporation finance, and the theory of investment. American Economic Review 48, 261-297.
Modigliani, F., Miller, M., 1963. Corporate income taxes and the cost of capital: a correc-tion. American Economic Review, 433-443.
Myers, S., 1977. Determinants of corporate borrowing. Journal of Financial Economics 5, 147-175.
Myers, S., 1984. The capital structure puzzle. Journal of Finance 39, 574-592.
Myers, S., Majluf, N., 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13, 187-221. Ozkan, A., 2001. Determinants of capital structure and adjustment to long run target:
evi-dence from UK company panel data. Journal of Business Finance & Accounting 28, 175-198.
Rajan, R., Zingales, L., 1995. What do we know about capital structure? Some evidence from international data. Journal of Finance 50, 1421-1460.
Scott, J., 1976. A theory of optimal capital structure. Bell Journal of Economics 7, 33-54. Serfling, M., 2014. Firing costs and capital structure decisions. Available at SSRN 2396599. Titman, S., Wessels, R., 1988. The determinants of capital structure choice. Journal of
Finance 43, 1-19.
van Binsbergen, J., Graham, J., Yang, J., 2010. The cost of debt. Journal of Finance 65, 2089-2136.
i Appendix A
Table A.I Variables Defined
This table provides the definitions and calculations of the independent and dependent variables. The right column shows the abbreviations for the variables used by the Orbis database. Size is measured by the logarithm of total assets, all variables are based on ratios, for normali-zation purposes.
Variable Definition Calculation with Orbis abbreviations
TDM Total debt / market value of assets (LTDB + CULI) / ((PRICE*SHARES) + TOAS – CSTK)
TDA Total debt / total assets (LTDB + CULI) / TOAS
LDM Long term debt / market value of assets LTDB / ((PRICE * SHARES) + TOAS – CSTK)
LDA Long term debt / total assets LTDB / TOAS
BVA Total assets TOAS
MVA Market value of equity + total assets – common stock MVE + TOAS – CSTK
MVE Market price * shares outstanding PRICE * SHARES
GRWT1 Market to book ratio = market value of equity / book value of
equity ((TOAS – IFAS – (CULI + NCLI)) – CSTK + (PRICE * SHARES)) / (TOAS – IFAS – (CULI + NCLI))
GRWT2 Change in log assets (Log(TOAS)/Log(TOASt-1)) – 1
PROF1 Return on assets ratio ROA
PROF2 Earnings before interest and taxes / book value of assets EBIT / TOAS
SIZE1 Logarithm of total assets log(TOAS)
TANG1 Fixed assets / book value of assets FIAS / TOAS
ZSCORE Z = 1.2 * (working capital / total assets) + 1.4 * (retained earn-ings / total assets) + 3.3 * (earnearn-ings before interest and taxes / total assets) + 0.6 * (market value of equity / total liabilities) + 0.999 * (sales / total assets)
ii Appendix B
Table B.I Correlation Matrix
This table shows a matrix of the correlations between all dependent and independent variables used in this paper, to check for the presence of multicollinearity. Correlations higher than 0.8 are denoted in bold numbers. Definitions of variables are given in Appendix A. The other high correlations do not affect the results, since they are both used as a proxy for the same measure and therefore never are in the same es-timation. Therefore, excluding variables because of the presence of multicollinearity is not necessary.
TDA TDM LDA LDM GRWT1 GRWT2 PROF1 PROF2 SIZE1 TANG1 ZSCORE
