How capital structure strategy relates to performance February 2013 Cas Timmermans S1614126 Student Msc BA Finance University of Groningen Supervisor: Dr. B.A. Boonstra
Faculty of Economics and Business University of Groningen
2nd supervisor:
Dr. L. Dam
Table of content Abstract ... 1 1. Introduction ... 2 2. Literature ... 4 3. Methodology ... 10 4. Data ... 13 5. Results ... 17 6. Robustness analysis ... 22 7. Conclusion ... 23 References ... 25 Appendixes ... 27 Abstract
This empirical investigation studies explanatory variables for the leverage ratio and whether performance is related to the two capital structure theories used as capital structure strategies, namely the trade-off strategy and pecking order strategy, in the agricultural and food industry. Results show significant relations between leverage and the explanatory variables growth opportunities, size, profitability, and tangibility. Firms with a medium-high performance, measured by the Sharpe ratio, have more resemblance to the pecking order strategy and firms with medium-low or low performances have more resemblance with the trade-off strategy. This indicates that firms in the agricultural and food industry with higher performance have more coefficients in line with the pecking order strategy than the trade-off strategy.
1. Introduction
Year on year around two thousand new McDonald’s restaurants open their doors worldwide, due to the fact that the world population grows and their consumption changes. This is an example of a response to the growing and changing demand in the agricultural and food industry. To serve these changes the supply-chain has been altered significantly over the last century (Dunn, 2010). There has been an ever-growing demand for vertical coordination as well as integration to be able to match the increasing diversity of consumer preferences. This trend has been noticed over the past century and is predicted to carry on over the next decades (Myers et al., 2010). The agricultural and food industry is a rapidly changing industry regarding production methods as well as the level of industry composition. Selecting this industry in the United States (U.S.), could serve as an example for firms in developing markets to increase their performance, because as Booth et al (2002) describe some capital structures can be . This can result in higher production and less shortages, preventing famine (Lybbert & Sumner, 2012). So this is an interesting industry to investigate the relation between performance and capital structure.
Modigliani & Miller (1958) state that in a perfect market the value of firms is not influenced by their capital structure. In the absence of taxes, bankruptcy costs, and information asymmetries, management is indifferent with regard to the capital structure of the firm. However since the perfect markets assumption is strictly theoretical, managers are not indifferent and following Modigliani & Miller (1958) on the irrelevance of the capital structure, Titman & Wessels (1988) investigated the capital structure determinants.
The tax-deductibility of interest costs creates a tax-shield, which favors the use of debt over other means of finance. Due to the imperfectness of markets, managers face a puzzle to find the capital structure that is most beneficial to their firm (Barclays & Smith, 2005). To assist managers in their search for the optimal leverage ratio, the relations between leverage and explanatory variables are considered. An important research in this field is performed by Titman & Wessels (1988).
The aim of this empirical investigation is to examine whether high performing U.S. firms in the agricultural and food industry show the most resemblance to either the trade-off strategy or the pecking order strategy.
By investigating whether a capital structure strategy is related to the performance of a firm managers may solve their capital structure puzzle. One strategy outperforming the other indicates that in the agricultural and food industry firms can optimize their performance by choosing the right strategy.
The relations between leverage and several explanatory variables are investigated, because the directions of the relations can indicate which capital structure strategy is applied by managers within the firms. By analyzing what strategy management uses most in high performing firms and what strategy in low performing firms, an indication can be made as to what strategy outperforms the other. This is an innovative addition to existing literature. To be able to examine whether the data of high performing firms shows more or less resemblance to one of the capital structure strategies, the sample is divided into groups based on their performance measured by their Sharpe ratio.
The research is performed among U.S. listed firms from 1990 to 2010 in the agricultural and food industry. The analysis is restricted to U.S. firms because many leading firms in this industry are based there and the sample is of a significant scale. Besides this, the restriction to U.S. listed firms creates equality with regard to the legal systems and the tax systems that apply to the firms included in the research. The period 1990 until 2010 contains not only growth but also decline in global markets, which gives a realistic overall view of the market movements that may also occur in the future.
2. Literature
Because this is an empirical investigation the literature section is used indicative instead of comparative. First the two capital structure strategies, which can be used by management as capital structure strategies, are explained and second the explanatory variables are discussed. For each variable the direction of the relation with leverage is provided first according to the trade-off strategy followed by the pecking order strategy. The relations are linked to the capital structure strategy based on the results in previous research.
The two main strategies explaining the leverage ratio are the trade-off strategy and the pecking order strategy. According to the trade-off strategy management calculates an optimal leverage ratio in order to optimize the firm’s capital structure. This optimal leverage ratio is found by calculating which mix of debt and equity financing results in having the lowest financing costs. Taking on debt can create a tax shield because the interest can be tax-deductible. But financial distress might increase the cost of debt beyond the point of incremental costs of issuing equity. The optimum is found by calculating the combination of debt and equity with the lowest financing costs. When an issue or retirement of debt or equity takes place, the change of the leverage ratio should return the ratio towards the optimum. The static trade-off strategy has a stable optimum leverage. The optimal leverage ratio or target ratio can not be measured with outsider information. A widely used solution to estimate the target ratio is to calculate the historical mean of the leverage ratio. Managerial adjustments to the leverage ratio should be mean reverting, because the optimum is the mean. (Shyan-Sunder & Myers, 1999).
The pecking order strategy states that a firm’s capital structure is a result of the preference of financing methods in order to make investments. According to the developer of the theory, Donalson (1961), managers prefer retained earnings to fund their investments rather than increasing debt or equity. This means that firms become less levered because the retained earnings increases the value of assets, but does not increase debt. If this method of financing is not sufficient, the amount of investments to be made is higher than the retained earnings or there is no profit from which earnings can be retained, a firm prefers to finance investments with debt rather than equity (Hovakimian, Opler & Titman, 2001). The key difference between the two strategies is that the pecking order strategy takes retained earnings into account where the trade-off strategy focuses on the ideal mix of debt and equity. This difference leads to different directions in the relations between leverage and the explanatory variables of leverage as is shown below.
tax shields, uniqueness, industry classification, and volatility. After reviewing the variables growth opportunities, size, profitability, and tangibility are included my research, the reasons for exclusion of the other variables are explained below. Each variable is briefly explained and the relation between leverage and the explanatory variable is discussed, first using the trade-off strategy and second using the pecking order strategy. The expected relations, as described below, between leverage and the explanatory variables and the differences between strategies are presented in Table 1. The indications, with regard to the positive or negative relation between leverage and the explanatory variables, are depicted by plus and minus signs. The measures used for the variables are described in Section 3.
Growth Opportunities
The possibilities management has to increase the firm value, the firms potential. Titman & Wessels (1988) expect that leverage has a negative relation with growth opportunities. Their results indicate this is the case when leverage is measured using market values, but it is positive when using book values to calculate leverage.
Trade-off strategy
First, Jensen & Meckling (1976) and Myers (1977) show that agency costs are higher when firms have relatively large growth opportunities. When debt is issued managers are stimulated to try to transfer wealth away from shareholders and bondholders. This can be referred to as the underinvestment problem. Managers reject profitable investment opportunities when they need to increase debt to create value for shareholders. Thus due to stronger incentives to avoid underinvestment the trade-off strategy predicts that firms with more growth opportunities have a lower leverage level. Besides this factor there are two more, which state that the relation between leverage and growth is negative. Second, Titman & Wessels (1988) state that growth opportunities add value to the firm. Because firms with high growth opportunities will be higher valued by shareholders. This creates an increase in total assets without increasing debt and thus
Variables Trade-off strategy Pecking order strategy
Growth opportunities - +
Size +
-Profitability +
-Tangibility + + /
-Table 1
Expected relation between leverage and the explanatory variables This table presents the expected relations between leverage and each of
the individual explanatory variables under the assumption of the different strategies according to previous research.
decreasing leverage. Third, the free cash flow theory states that with more growth opportunities there is a smaller need to control the free cash flows through the controlling effect of debt. Because there are enough profitable investment opportunities, the level of free cash flows will be lower and therefore also the level of debt needed (Jensen, 1986).
Pecking order strategy
According to Rajan & Zingales (1995), using the pecking order strategy growth opportunities has a positive relation with leverage. When there are many investment opportunities financing needs are more likely to exceed the retained earnings. To make all positive investments management takes on other means of financing such as debt and according to the pecking order strategy this leads to an increase in leverage.
Size
Regarding the expected relation between leverage and size, Titman & Wessels (1988) acknowledge that it can be a positive as well as negative relation. In general large firms are more diversified and thus have lower bankruptcy costs that leads to an increase in leverage. On the other hand the transaction costs are relatively higher for smaller firms. This leads to higher leverage, because smaller firms have to take on more debt at once. Their results indicate a negative relation.
Trade-off strategy
Size affects the leverage ratio in a positive way. Larger firms are generally regarded as being more diversified, which creates a smaller chance of default and thus lowers the bankruptcy costs. Because, the risk for debt holders decreases resulting in lower cost of debt, management adopts more debt and increases in the leverage ratio. This result is found by, Warner (1977), Ang, Chua & McConnel (1982), Hovakimian, Opler & Titman (2001) and Frank & Goyal (2003).
Pecking order strategy
higher costs, managers of smaller firms have to borrow even more, creating higher leverage for smaller firms. (Chan, Chen & Hsieh, 1985)
Profitability
It is expected that profitability has an important influence on the capital structure, because it determines the amount of earnings available to be retained. The results show a negative relation between leverage and profitability (Titman & Wessels, 1988)
Trade-off strategy
An increase in profitability increases a firms leverage for three reasons. First, for more profitable firms the chance of going bankrupt declines, resulting in lower bankruptcy costs. Due to this declined risk, debt-financing costs decline as well and more debt is used, increasing leverage. Second, Frank & Goyal (2003) find that more profitable firms have a larger taxable income. According to the trade-off strategy, Managers increase the leverage ratio to create a larger tax-shield. Third is the agency problem, because in more profitable firms there are larger free cash flows. Through the controlling effect of debt, managers are less able to expropriate wealth from shareholders, reducing agency costs. So an increase in profitability should lead to an increase in leverage (Jensen & Meckling, 1979 and Jensen, 1986).
Pecking order strategy
Managers prefer financing with retained earnings instead of external means of capital. When a firm is more profitable there are more earnings that can be retained, leading to a lower leverage ratio. This means that more profitable firms have less leverage. Titman & Wessels (1988), Rajan & Zingales (1995) and Fama & French (2002) find this relation in their research.
Tangibility
This variable discussed by Titman & Wessels (1988) is the illiquidity of assets and is an important variable with regard to leverage because investors require lower interest and lend with more ease to a firm that can back its debt with assets (Frank & Goyal, 2003). The coefficients found by Titman & Wessels (1988) indicate that there is a negative relation between leverage and tangibility.
Trade-off strategy
strategy an increase in tangibility leads to an increase in leverage (Hovakimian, Opler & Titman, 2001).
Pecking Order strategy
The explanation given for the trade-off strategy also applies for the pecking order strategy (Shyam-Sunder & Myers, 1999). In addition to this explanation, Harris and Raviv (1991) state that tangibility has a negative relation with leverage. According to them, firms with fewer tangible assets are less transparent to investors, resulting in higher costs of debt. The pecking order strategy states that managers increase debt to invest in opportunities and firms with asymmetric information problems have higher costs of debt, as well as higher cost of equity, and have to increase debt even further.
Non-debt tax shields
This variable is excluded from this paper for two reasons. First, compared to the research of Titman & Wessels (1988), the sample in this research consists of firms from a single country instead of international. There are no differences in tax rates applied to firms. Second, there is no apparent difference between the two strategies.
Uniqueness
This variable indicates how firm or product-specific the knowledge and assets are. This variable is directly linked to bankruptcy costs, because firm specific assets are less valuable to other firms and sometimes can not be sold at all. This increases the bankruptcy costs in case of a default and this has a negative effect on the leverage ratios. A widely used measure for uniqueness is research and development (R&D) expenses. After collecting the data it becomes clear that there was too little available R&D data to include this variable, therefore it is excluded.
Industry classification
Volatility
The volatility of firms operating profit is important to the leverage ratio, because firms with less volatile operating profits are viewed as less risky, than firms with more volatile operating profits with the same mean, by debt holders the cost of debt is lower for firms with a lower volatility (Titman & Wessels, 1988). In the sample not all firms have the same amount of data points, which creates a problem calculating volatility, because using a running volatility over ten years results in the loss of at least half the sample. And the use of dummy for high and low volatile firms creates a problem with the amount of data points per firm, since the firms with the lowest number of data points available have only five points, this creates a large difference in volatility with firms that have twenty points. Therefore volatility is not taken into account due to a lack of data.
Hypotheses
Regarding the relation between leverage and the explanatory variables growth opportunities, size, profitability, and tangibility, it is expected that the regression yields significant coefficients for these variables. Because the results found by Titman & Wessels (1988) indicate significant relations between leverage and the explanatory variables.
The second hypothesis is a large significant regression coefficient for the relation between profitability and leverage, because the variables are measured with the same denominator and a significant change in the regression coefficient when the profitability measure is replaced.
A third expected result is that the regression coefficients of the complete sample show more resemblance with the pecking order strategy than with the trade-off strategy, because the results in the research of Titman & Wessels (1988) show more resemblance to the pecking order strategy.
3. Methodology
Building on Titman & Wessels (1988), I investigate how firm performance relates to the capital structure strategies. This analysis includes the ranking of firms based on their performance, as well as researching in which groups the capital structure strategies occur more. Which strategy managers apply is deducted from the data as is described in Section 2.
To examine whether there is a relation between capital structure and firm performance, most logical was to analyze per firm which capital structure strategy fits its data best and to analyze which strategy outperforms the other. However there are at most 21 data points per firm, resulting in a regression with low statistical significance.1 In order to get sample sizes that
have higher statistical significance, firms are first divided into groups based on their performance. After which a regression is performed and with the coefficients of that regression it is possible to answer whether high performing firms show more resemblance to one of the capital structure strategies. The resemblance is measured by comparing directions of the regression coefficients with the directions of the expected relations according to the each strategy based on previous research (Table 1). The number of equal directions indicates the resemblance.
Sharpe ratio
First the firms are ranked based on their performance, which is measured with the Sharpe ratio. The Sharpe ratio of the firms is chosen, because it is a risk-adjusted return measure that takes into account the trend and is calculated as is shown in (1), (2), (3), (4) and (5) (Hull, 2012). First, returns are calculated,
ln
, (1)
where is the return at time t and is the total return index(level) at time t.
The average return is calculated using (1),
∑
⁄ , (2)
where is the average return, is return at time t and T is the number of years for which data is available.
The estimator for standard deviation of return is calculated as follows,
√ , (3)
1Regressions with leverage as dependent variable and growth opportunities, size, profitability and tangibility as
independent after performing variables on individual firms, the R2-values were often above 0.8. This is very high
where is the estimator for standard deviation, is the standard deviation of the total return index and T is the number of years for which data is available.
Using (1), (2) and (3) the estimator for expected return is calculated as follows, ̂
" ∙
", (4)
where ̂ is the estimator for expected return corrected for the trend, is the average return as described in (2) and is the estimator for standard deviation of the TRI as described in (3).
Hence, this leads to the risk-adjusted return measure, or Sharpe ratio $%, which is calculated as, $% &'( )
' , (5)
where ̂% is the estimator for expected return corrected for the trend for firm i as described in (3), * is the risk-free rate approximated by the average return on 1-year U.S. treasury bonds and % is the estimator for standard deviation for firm i as described in (2).
The Sharpe ratio is calculated in line with Hull (2012), who makes use of estimators to correct for the trend in the underlying returns. The measurement for the risk-free rate is the average rate of return on 1-year United States Treasury bill over the period 1990 to 2010. Furthermore the total return index is used as input to calculate the Sharpe ratio.
All firms are ranked based on their Sharpe ratio and divided into four groups. With the use of four groups a range is created, which exists of high performing, medium-high performing, medium-low performing, and low performing firms. Hence the different groups can show resemblance to the two strategies and are able to support or contradict the expectations.
After the division of firms into groups the regression model is set up to analyze for the relation between capital structure and the different explanatory variables that are also used in the research of Titman & Wessels (1988). They use six different measures for leverage, which are long-term, short-term and convertible debt divided by the book value and the market value of equity. I use a common measure of leverage (Shyam-Sunder & Myers, 1999), which is book value of total debt over book value total assets, as is shown,
+,- ./, (6)
Growth opportunities
This variable is not shown on the balance sheet of firms, but is taken in to account by investors when shares are being traded. Therefore a good indicator to measure growth opportunities is to take the book-to-market equity,
234 8 ∗ :;6<567 ?>7=>7, (7) where 234 is the growth opportunity, @$, is the common shareholder equity, A is the share-price, BC$4 are the number of shares, BVE is the book value of equity and MVE is thus the market value of equity.
Size
This is measured the same way as it is in Titman & Wessels (1988). The logarithm of net sales is used rather than the logarithm of total assets, because firms can try to keep their fixed cost low through outsourcing and lease contracts, which lowers assets but not net sales. Using net sales is therefore less sensitive to managerial and accounting instruments. This is determined as,
$DEF +GHB$ , (8)
where $DEF is the size and B$ is net sales.
Profitability
As in Titman & Wessels (1988) the measure for profitability used is the ratio of operating income over total assets. Because, both the leverage ratio and this profitability measure are divided by total assets a second measure for profitability is taken into account, namely net sales over total assets. These are shown in (9) and (10),
ACI1 /; , (9)
where ACI1 is the profitability, C is operating income and 1 is book value of total assets and,
ACIB$ :6; , (10)
where ACIB$ is profitability, C is operating income and B$ is net sales.
Tangibility
The same measurement of tangibility as in Titman & Wessels (1988) is used, i.e. (11),
1B2 J//, (11)
where 1B2 is tangibility, I1 is book value of fixed assets and 1 is book value of total assets.
denominator of both variables. Assuming this may pose a problem, the data is tested for high correlations and two measures for profitability are included.
After the division of firms into groups based on their Sharpe ratio, the regressions are performed with the variables growth opportunities, size, profitability OI/TA or profitability OI/NS, and tangibility as is shown in (12) and (13),
+,- K L 234 L" $M, LN ACI1 LO 1B2 P, (12) where +,- is Leverage, K is a constant, L, L", LN and LO are the coefficients of the explanatory variables, 234 is growth opportunities, $M, is size, ACI1 is profitability using (9), 1B2 is tangibility and P is the error term. The second regression is performed using,
+,- K L 234 L" $M, LQ ACIB$ LO 1B2 P, (13) where +,- is Leverage, K is a constant, L, L", LO and LQ are the coefficients of the explanatory variables, 234 is growth opportunities, $M, is size, ACIB$ is profitability using (10), 1B2 is tangibility and P is the error term.
The regression performed is an ordinary least squares (OLS) regression, this generates an error term in the regressions, which follows P~ B 0, U" . The standardized residuals indicate whether the OLS regression is accurate. The closer they are to a normal distribution, the more accurate OLS is as a regression technique.
Based on the significant coefficients that result from these regressions, it is determined which strategy fits best to each group. And based on the dispersion of the strategies over the groups it is possible to accept or reject the hypotheses.
4. Data
Table 2 shows the mean of the Sharpe ratio of all firms, which is 0.123. This indicates that overall firms have a positive risk-adjusted return. For the variables following the Sharpe ratio the number of observations show that on average there are almost fifteen data points per firm. The firms in the sample are on average approximately 60% levered, have positive growth potential but a negative profitability (make a loss) of about 70% of the total assets and a negative profitability 40% of the net sales, and the total assets consist for almost 35% of fixed assets. The standard deviations for the variables leverage, growth opportunities, and both profitability measures are relatively large indicating there could be outliers. All Jarque-Bera values are large, considered a value below 5.0 indicates that the data points for the variable show resemblance to a normal distribution. The probabilities in the last column confirm there is no significant resemblance to a normal distribution because all values are very close to zero.
In some cases variables take on relatively extreme values, this is taken into account by creating a sample excluding outliers using four restrictions.
The first restriction regards the leverage ratio. Extremely levered firms are excluded from the regression. Firms with a leverage ratio above two are excluded, because all firms are listed have large total assets and debt exceeding total assets are not common but debt being 200% of total assets is not realistic in this industry (Russo et al., 2000). The second restriction is on growth opportunities. This variable has to be between minus five and five, because in this industry it is not likely that market values differ from book values more than five times their value (positive or negative). The third restriction is on profitability, because these are all listed firms with a relatively large amount of total assets, it is not likely that they will have a operating loss of more than half the total assets. Therefore the restriction is made to only admit the values of profitability above minus a half. The fourth restriction is a restriction with regard to tangibility. All firms in the sample are listed firms from the agriculture and food industry in
Variable Observations Mean Std. Dev Jarque-Bera Prob
Sharpe ratio (-) 101 0.123 0.046 1.822e3 0.000
Leverage (%) 1511 0.596 5.811 1.652e7 0.000
Growth opportunities (%) 1445 0.732 14.630 4.800e7 0.000
Size (-) 1512 5.476 1.283 120.168 0.000
Profitability (%) 1514 -0.688 15.823 9.326e7 0.000
Profitability (%) 1512 -0.414 4.031 7.102e6 0.000
Tangibility (%) 1505 0.339 0.185 33.241 0.000
The descriptives of the data sample including outliers. The variables included are the Sharpe ratio (1), (2), (3), (4) and (5), Leverage (6), Growth opportunities (7), Size (8), Profitability OI/TA (9), Profitability OI/NS (10) and Tangibility (11). For these variables the
number of observations, mean, standard deviation (Std. Dev) and Jarque-Bera value are given.
Table 2
which firms use specific machinery and perform a lot of research and development. Therefore it is very unlikely that firms do not have fixed assets and the restriction is made that firms need to have tangibility above zero to be included in the sample. The descriptives of the sample excluding outliers are shown in Table 3 (Boehije, Gloy & Henderson, 2012)
Comparing the descriptives of the sample including outliers with the sample excluding outliers, as presented in Table 3, it is clear that it still consists of 101 firms but that the number of observations for the variables is reduced by approximately 200 per variable. The average leverage drops to approximately 25% and both measures for profitability turn positive around 10%, and 3,5%. These figures are more realistic for an overview of listed firms in the agricultural and food industry. Also the standard deviations for leverage, growth opportunities and both profitability measures are significantly reduced, due to the exclusion of large outliers. The variables are not normally distributed, because the Jarque-Bera values are relatively high and the probabilities close to zero. Therefore the main research is performed using the sample excluding outliers. To investigate the robustness of the sample the research is repeated with the sample including outliers.
As described in the methodology section the sample is divided into four different groups based on their Sharpe ratio. Table 4 and Table 5 present the descriptives of the separate groups of the sample including outliers and excluding outliers.
As presented it is clear that in each following group the mean of the Sharpe ratio is lower, because this division is based on those values. Again there are significant differences between the sample including outliers and the sample excluding outliers. The first group (1-26) has a leverage ratio of 1.5 what indicates that the total debt is 150% of the total assets. In the agricultural and food industry this is not realistic. When outliers are excluded this value drops to 20% as is realistic for this industry.
Variable Observations Mean Std. Dev Jarque-Bera Prob
Sharpe ratio (-) 101 0.123 0.046 1.822e3 0.000
Leverage (%) 1281 0.255 0.178 44.081 0.000
Growth opportunities (%) 1281 0.892 0.892 1.835e3 0.000
Size (-) 1278 5.729 1.079 36.696 0.000
Profitability (%) 1281 0.093 0.112 1.575e3 0.000
Profitability (%) 1278 0.034 0.592 2.447e7 0.000
Tangibility (%) 1281 0.352 0.172 38.546 0.000
The descriptives of the data sample excluding outliers. The variables included are the Sharpe ratio (1), (2), (3), (4) and (5), Leverage (6), Growth opportunities (7), Size (8), Profitability OI/TA (9), Profitability OI/NS (10) and Tangibility (11). For these variables the
number of observations, mean, standard deviation (Std. Dev) and Jarque-Bera value are given.
Table 3
This is also the case for other variables and that is why the research is based on the sample excluding outliers and the sample including outliers is used to control the robustness of the sample.
Table 5 presents the descriptives of the sample excluding outliers. The number of observations per group varies from 140 observations to 501 observations, large enough to create significant results in the regression. The mean leverage in each group lies between 20% and 30%. For the growth opportunities variable the mean values differ between 0.717 and 1.035, which indicates that the firms values are expected to decline by approximately 30% or to grow by approximately 3.5%. Because when the market value of equity is only 70% of the book value the market expects a negative growth of about 30%, which is the case for group 3. The size variable shows that the third group has the largest firms and that the average size of the firms in other groups are relatively close together. Regarding the profitability variable it is remarkable that the group with the highest Sharpe ratio has the lowest profitability values. The means of the tangibility variable lie around the 30% to 40%, this indicates that about 30%-40% of the assets consists of fixed assets, where the highest tangibility is found in the group with the lowest Sharpe ratio.
The Jarque-Bera value indicates whether a variable is normally distributed, because the lower the value, the more a variable resembles a normal distribution. Excluding the Sharpe ratio, there are five occasions where the Jarque-Bera value is below 5.0 and thus might have a normal distribution. But since these are randomly dispersed among different variables and groups it has no significant influence on the regressions.
5. Results
Building on the research of Titman & Wessels (1988), the first analysis investigates the correlations between the variables. Variables sharing a denominator can create high correlation and high correlations can influence the quality of the regressions (Titman & Wessels, 1988). The correlations found are presented in Table 6.
measure for tangibility has the same denominator as leverage, but does not have a large correlation with leverage. Therefore the most common measure for tangibility, fixed assets over total assets is used.
Using (12), the OLS regression is performed with the sample excluding outliers and the results are presented in Table 7. These results show that for the complete sample (All) that all variables are significant at a one percent level.
The growth opportunities variable has a small negative coefficient, which indicates that if a firm has more growth opportunities the leverage ratio will decrease. For each percentile increase in growth opportunities leverage decreases with 0.021%. This is explained by the fact that if the market value of equity (share price) increases, managers prefer equity financing over debt financing because the market value of equity is overpriced compared to its book value.
The relation between size and leverage is small and positive. This means that larger firms relatively take on more debt, which increases leverage, and is possible because larger
Variables Lev Gr Size Prof TA Prof NS Tang TA Tang Log(FA)
Leverage 1.00 Growth -0.02 1.00 Size -0.15 0.05 1.00 Profitability (OI/TA) -0.69 0.04 0.17 1.00 Profitability (OI/NS) -0.18 0.01 0.32 0.27 1.00 Tangibility (FA/TA) 0.03 0.03 -0.09 0.01 0.00 1.00 Tangibility (Log(FA)) -0.15 0.08 0.93 0.18 0.26 0.17 1.00 Table 6
Estimated correlations between variables
The correlations between leverage and the explanatory variables are shown.
Variable All 1-26 27-51 52-76 77-101 Growth opportunities -0.021 0.013 0.014 -0.086 -0.044 (-3.9)*** (1.1) (1.6) (-7.5)*** (-4.4)*** Size 0.058 0.018 0.009 0.072 0.046 (12.6)*** (1.4) (6.8)*** (7.8)*** (5.0)*** Profitability (OI/TA) -0.522 -0.096 -0.809 -0.810 -0.542 (-11.7)*** (-1.0) (10.3)*** (-8.1)*** (-6.3)*** Tangibility 0.171 0.019 0.289 0.099 0.187 (6.2)*** (0.3) (5.9)*** (1.8)* (4.0)*** R-squared 0.163 0.041 0.350 0.280 0.219 Adj. R-squared 0.161 0.012 0.333 0.274 0.208 Table 7
OLS regressions, sample excluding outliers
The regression as it is presented in (12) LEV = β1 GRWTH + β2 SIZE + β3 PROFTA + β4
TANG + C + ε over the sample excluding outliers. The estimated coefficients are presented
with the T-statistic within the parentheses. The first column shows the values for the complete sample and the the following columns give the values for the different groups
based on their Sharpe ratios.
firms often are more diversified and thus carry less risk. Due to the lower risk, large firms get lower debt financing costs and increase their leverage ratio.
The coefficient estimated for profitability is large and negative, which indicates that for each 1% increase in profitability firms decrease their leverage by 0.522%. The relation is explained by the fact that more profitable firms do not have the need for financial means in the form of debt, because they retain sufficient funds from their earnings. The fact that the coefficient is relatively large resembles the research of Titiman & Wessels (1988) and can accordingly be explained by the common denominator, therefore the regression is also performed with the profitability measure operating income over net sales.
The final variable, tangibility, has a positive coefficient meaning firms with a higher ‘fixed assets over total assets’ ratio have a higher leverage ratio. Indicating that the fixed assets are used as collateral that lower the risk and thus the cost of debt financing, resulting in a higher leverage ratio. The directions (positive or negative) of the coefficients are equal for the variables in all groups, only the significance differs in some groups. Therefore the same explanations as presented above apply to all significant relations.
Looking at the first group (1-26), there is no significant coefficient, indicating that these explanatory variables do not significantly explain movements in the leverage ratio. So for the high performing firms there is no significant relation between leverage and growth opportunities, size, profitability, and tangibility.
The next group (27-51) has one percent significant coefficients for size, profitability, and tangibility. For this group of medium-high performing firms, the coefficients for size and tangibility are both relatively small and positive, thus larger firms with many tangible assets have a larger leverage ratio. The negative coefficient for profitability indicates that firms with higher profitability have lower leverage ratios, but the coefficient is very large giving rise to the question whether the shared denominator might bias this result.
For the third group (52-76) with medium-low performing firms, the results are similar to those of the second group with the exception that the coefficient for growth opportunities is negative and significant at a one percent level, and the coefficient for tangibility is ten percent significant instead of one percent significant. So larger firms with more tangible assets have larger leverage ratios and more profitable firms with larger growth opportunities have lower leverage ratios.
The R-squared value of the regression over all the firms indicates that about 16% of the change in leverage is explained using these variables. Looking at the separate groups the R-squared value of the first group containing high performing firms is relatively low compared to the other groups, this fits with the fact that none of the variables are significant in this group. The R-squared values of the other groups are all over 20%, indicating that more than a fifth of the change in leverage is explained using these variables.
Because, each group contains a relative large coefficient for the relation between profitability and leverage in the results, as found by Titman & Wessels (1988), the regression is also performed using operating income over net sales as a measure for profitability. The results of this regression as described in (13) are presented in Table 8.
When comparing the results from the regression using (13) to the regression using (12) there are three major changes. First, although the significant coefficients for profitability are still negative, the coefficients are significantly smaller. So it is concluded that the large coefficients in the regression using operating income over total assets exist due to the use of the same denominator in the dependent variable as well as the explanatory variable. Therefore the second regression, using operating income over net sales, is used to test the hypotheses. Second, the coefficient for growth opportunities in the second group (27-51) becomes significant and although the direction does not change compared to Table 7, the direction of the relation differs from the other significant results for this variable. This positive coefficient indicates that when growth opportunities increase leverage increases as well. This is explained by firms having a lot of growth opportunities and to act on these opportunities the firms need financing means to
Variable All 1-26 27-51 52-76 77-101 Growth opportunities -0.011 0.015 0.028 -0.072 -0.036 (-2.0)** (1.3) (2.9)*** (-6.1)*** (-3.5)*** Size 0.045 0.009 0.063 0.069 0.044 (9.5)*** (0.7) (5.9)*** (7.2)*** (4.5)*** Profitability (OI/NS) -0.039 0.008 -0.184 -0.544 -0.040 (-4.6)*** (0.4) (-4.8)*** (-4.5)*** (-4.4)*** Tangibility 0.191 0.024 0.292 -0.011 0.281 (6.7)*** (0.3) (5.3)*** (-0.2) (5.8)*** R-squared 0.088 0.034 0.188 0.217 0.162 Adj. R-squared 0.085 0.006 0.178 0.211 0.150
The regression as it is presented in (13) LEV = β1 GRWTH + β2 SIZE + β5 PROFNS + β4
TANG + C + ε over the sample excluding outliers. The estimated coefficients are presented
with the T-statistic within the parentheses. The first column shows the values for the complete sample and the the following columns give the values for the different groups
based on their Sharpe ratios.
Where */**/*** indicate significance levels of respectively 10%/5%/1% Table 8
make the necessary investments, which are acquired by increasing debt. Third, tangibility in the third group (52-76) ceases to be significant and is accordingly excluded from the interpretation of the results.
The R-squared values decrease compared to the first regression results. This is explained by the replacement of the profitability variable with the largest coefficients for the more realistic one with smaller coefficients. However for the three groups with significant results the R-squared values remain above 15%.
The output of the standardized residuals test of the regressions is presented in Appendix 6 and shows whether the residuals have a normal distribution. A normal distribution indicates that the model is a good fit to the data. The results in Appendix 6 show that there is little difference in distribution between the two types of measure for profitability but that the exclusion of outliers improves the fit of OLS as regression technique. The group consisting of medium-high performing firms has the most resemblance to a normal distribution.
Table 9 is a simplified version of Table 8, which presents the direction (positive or negative) of the significant coefficients resulting from the regression performed according to (13).
The significant findings presented in Table 9 are compared to the expected relations according to the trade-off strategy and the pecking order strategy as depicted in Table 1. The number of corresponding coefficients is presented in Table 10, along with the strategy that best matches the results.
Variable All 1-26 27-51 52-76 77-101
Growth opportunities - + -
-Size + + + +
Profitability (OI/NS) - - -
-Tangibility + + - +
Where - and + are negative, and positive relations leverage and the explanatory variable Estimated relations
The estimated relations between leverage and the individual explanatory variables according to the regressions performed, as presented in Table 8.
Table 9
Strategy All 1-26 27-51 52-76 77-101
Trade-off strategy 3 0 2 2 3
Pecking order strategy 2 0 3 2 2
Best match TO NA PO TO/PO TO
Where TO and PO are the trade-off strategy, and the pecking order strategy. Table 10
Resemblance between coefficients from the regression and the expected coefficients according to the two capital structure strategies.
For the group with all firms three significant coefficients comply with the trade-off strategy and two comply with the pecking order strategy. This indicates that the managers of the firms in the complete sample act most in line with the trade-off strategy. The high performing group (1-26) has no significant coefficients so for this group it cannot be determined which strategy is dominant. The medium-high performing group (27-51) shows most resemblance with the pecking order strategy, because three coefficients match with that strategy against two matching the trade-off strategy. The final two groups are most in line with the trade-off strategy, as is presented in Table 10.
These results show that the estimated relations in the group with all firms show the most resemblance with the relations between leverage and the explanatory variables according to the trade-off strategy, because the trade-off strategy has more matching coefficients with the results than the pecking order strategy. So deducting from the data the trade-off strategy is the dominating capital structure strategy used by firms in the agricultural and food industry. The first group is excluded due to a lack in significant results, which leads to the three remaining groups. Group 27-51 with medium-high Sharpe ratios is linked to the pecking order strategy, where the last two groups are linked to the trade-off strategy.
These results indicate that when the capital structure is deducted from the data, firms with a higher Sharpe ratio have more in common with the relations according to the pecking order strategy. This can be explained by the fact that in the agricultural and food industry firms using the pecking order strategy as their capital structure strategy create higher performance measured using the Sharpe ratio. The data of the medium-low, and low performing groups indicate more resemblance with the trade-off strategy. This can indicate that firms that have a lower performance, measured using the Sharpe ratio, in general apply the trade-off strategy as their capital structure strategy.
6. Robustness analysis
The robustness of the above-mentioned findings is tested in two ways. First, by substituting the measurement for profitability from operating income over total assets to operating income over net sales and second, by excluding outliers from the sample.
change in direction and only a few changes occur with regard to significance, indicating the measure for profitability is robust.
Second, the regressions were performed using the sample excluded from outliers. The results with regard to the strategies matched to the groups change substantial when outliers are included in the sample, As is shown in Tables 9 and 10, and Appendixes 3, 4, and 5. This indicates that the sample is not robust.
7. Conclusion
This empirical investigation examines whether high performing firms show the most resemblance to either the trade-off strategy or the pecking order strategy. The results show that the group with the medium-high performing firms shows the most resemblance with the pecking order strategy and the medium-low and low performing firms show the most resemblance to the trade-off strategy. This indicates that when capital structure is deducted from the data in the agricultural and food industry, firms that have a higher performance measured by the Sharpe ratio overall use the pecking order strategy as capital structure strategy.
The first hypothesis regarding significant regression coefficients is confirmed by the results. For the complete sample (All) all variables had a significant coefficient. However for the group with high performing firms no variable had a significant coefficient. This is could be explained by the fact that this group has the fewest data points, what could have influenced the results.
The second expected result was about the large and significant coefficient for profitability. This is also confirmed, because due to the large profitability coefficient I chose to conduct the main research with operating income over net sales as profitability measure. The large correlation and regression coefficient for operating income over total assets may indicate that variables with the same denominator might bias regression results. This problem did not occur using the tangibility measure.
performing firms show the most resemblance to the pecking order strategy as their capital structure strategy. This is as expected in the changing environment of the agricultural and food industry, because retained earnings can be important since there are a lot of investment opportunities. The directions of the relations between leverage and the explanatory variables for medium-low and low performing firms, matches most with the trade-off strategy.
When the measure of profitability, operating income over total assets, is replaced by operating income over net sales the coefficients are still significant and in the same direction but smaller. Indicating this variable is robust. To test the robustness of the sample regressions are performed excluding outliers and including outliers, between these regressions a substantial part of the results change. Indicating the sample is not robust.
Limitations & Further Research
An important limitation is the causality between capital structure and firm performance. Causality is not taken into account because this is an empirical investigation to test whether there is any relation between these variables. A second limitation is the size of the dataset. It consists of 101 listed American firms but could be expanded to global listed firms or firms in emerging markets. Then other variables can be added which is another limitation.
References
Ang, J.S., Chua, J.H., McConnell, J.J., 1982. The administrative costs of corporate bankruptcy: a note. The journal of finance 35, 219-226.
Barclay, M.J., Smith, C.W., 2005. The capital structure puzzle: The evidence revisited. Journal of applied corporate finance 1, 8-17.
Boehije, M.D., Gloy, B.A., Henderson, J.R., 2012. U.S. farm prosperity: The new normal or reversion to the mean. American journal of agricultural economics 95, 310-317.
Booth, L., Aivazian, V., Demirguc-Kunt, A., Maksimovic, V., 2001. Capital strucutures in developing countries. The journal of finance, 87-130.
Chan, K.C., Chen, N., Hsieh, D.A., 1985. An explanatory investigation of the firm size effect. Journal of financial economics, 451-471.
Dunn, S., 2011. The modern food industry and public health: a Galbraithian perspective. Journal of post Keynesian economics, 491-516.
Fama, E., French, K., 2002. Testing tradeoff and pecking order predictions about dividends and debt. Review of financial studies 15, 1-33.
Frank, M.Z., Goyal, V.K., 2003. Testing the pecking order theory of capital structure. Journal of financial economics 67, 217-248.
Harris, M., Raviv, A., 1991. The theory of capital structure. Journal of finance 46, 297-356.
Hovakimian, A., Opler, T., Titman, S., 2001. The debt-equity choice. Journal of financial & quantitative analysis 36, 1-24.
Hull, J.C., 2010. Options, futures, and other derivatives. Prentice Hall.
Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of financial economics 4, 305-360.
Jensen, M.C., Meckling, W.H., 1979. Rights and production functions: An application to labor-managed firms and codetermination. Journal of business 52, 469-506.
Lybbert, T.J., Sumner, D.A., 2012. Agricultural technologies for climate change in developing countries: Policy options for innovation and technology diffusion. Food policy 1, 114-123.
Modigliani, F., Miller, M.H., 1958. The cost of capital, corporation finance and the theory of investment. The American economic review, 261-298.
Myers, S.C., 1977. Determinants of corporate borrowing. Journal of financial economics 5, 147-175.
Myers R.J., Sexton, R.J., Tomek, W.G., 2010. A century of research on agricultural markets. American journal of agricultural economics 92(2), 376-402.
Rajan, R.G., Zingales, L., 1995. What do we know about capital structure? Some evidence from international data. Journal of finance 50, 1421-1460.
Russo, C., Weatherspoon, D., Peterson, C., Sabbatini, M., 2000. Effects of managers’ power on capital structure: a study of Italian agricultural cooperatives. International food and agribusiness management review 3, 27-39.
Titman, S., Wessels, R., 1988. The determinants of capital structure choice. Journal of finance 43, 1-21.
Appendixes
Variable All 1-26 27-51 52-76 77-101
Growth opportunities -2.260e-4 -0.044 0.006 -0.051 0.011
(-0.0) (-1.1) (1.4) (-6.5)*** (1.0) Size -0.080 0.272 0.040 0.080 0.062 (-1.2) (3.9)*** (3.6)*** (8.8)*** (0.2) Profitability (OI/TA) -0.597 -0.575 -0.393 -0.694 -2.784 (-35.3)*** (-75.2)*** (-11.4)*** (-7.1)*** (-10.8)*** Tangibility 1.072 0.650 0.195 0.072 2.695 (2.3)** (1.410) (3.2)*** (1.4) (2.0)** R-squared 0.480 0.972 0.271 0.260 0.281 Adj. R-squared 0.479 0.971 0.263 0.254 0.273 Appendix 1
OLS regressions, sample including outliers
The results of the regression as it is presented in (12) LEV = β1 GRWTH + β2 SIZE + β3
PROFTA + β4 TANG + C + ε over the sample including outliers. The estimated coefficients are
presented with the T-statistic within the parentheses. The first column gives the values for the complete sample and the the following columns give the values for the different groups based
on their Sharpe ratios.
Where */**/*** indicate significance levels of respectively 10%/5%/1%
Variable All 1-26 27-51 52-76 77-101 Growth opportunities -0.006 -0.840 0.008 -0.044 -0.001 (-0.7) (-3.8)*** (1.6) (-5.5)*** (-0.1) Size -0.342 -0.201 0.044 0.076 -0.718 (-3.5)*** (-0.5) (3.5)*** (8.1)*** (-2.3)** Profitability (OI/NS) -0.170 -0.107 -0.282 -0.447 -0.422 (-5.6)*** (-1.7)* (-8.3)*** (-3.8)*** (-3.2)*** Tangibility 0.630 -0.103 0.140 -0.008 2.502 (1.0) (-0.0) (2.2)** (-0.1) (1.6) R-squared 0.044 0.121 0.168 0.208 0.068 Adj. R-squared 0.042 0.102 0.159 0.202 0.057
The results of the regression as it is presented in (13) LEV = β1 GRWTH + β2 SIZE + β5
PROFNS + β4 TANG + C + ε over the sample including outliers. The estimated coefficients are
presented with the T-statistic within the parentheses. The first column gives the values for the complete sample and the the following columns give the values for the different groups based on their Sharpe ratios.
Where */**/*** indicate significance levels of respectively 10%/5%/1% Appendix 2
OLS regressions, sample including outliers
Variable/Strategy All 1-26 27-51 52-76 77-101 Growth opportunities - - -Size + + + + Profitability (OI/TA) - - - -Tangibility + + + + Trade-off strategy 3 0 2 3 3
Pecking order strategy 2 0 2 2 2
Best match TO NA TO/PO TO TO
Appendix 3
Where - and + are negative, and positive relations leverage and the explanatory variable and TO and PO are the trade-off strategy, and the pecking order strategy.
Estimated relations, sample excluding outliers
Variable/Strategy All 1-26 27-51 52-76 77-101 Growth opportunities - -Size - + + -Profitability (OI/NS) - - - - -Tangibility + Trade-off strategy 0 1 2 2 0
Pecking order strategy 2 1 2 1 2
Best match PO TO/PO TO/PO TO PO
Appendix 4
Where - and + are negative, and positive relations leverage and the explanatory variable and TO and PO are the trade-off strategy, and the pecking order strategy.
The estimated relations between leverage and the individual explanatory variables according to the regressions performed. Below the resemblance between coefficients from the regression and the expected coefficients according to the two capital structure
Estimated relations, sample including outliers
Variable/Strategy All 1-26 27-51 52-76 77-101 Growth opportunities -Size + + + Profitability (OI/TA) - - - - -Tangibility + + + Trade-off strategy 1 1 2 2 1
Pecking order strategy 2 1 2 1 2
Best match PO TO/PO TO/PO TO PO
Appendix 5
Where - and + are negative, and positive relations leverage and the explanatory variable and TO and PO are the trade-off strategy, and the The estimated relations between leverage and the individual explanatory variables according to the regressions performed. Below the resemblance between coefficients from the regression and the expected coefficients
