The Impact of Capital Structure on Agency Costs: Evidence from UK Public Companies

Transcription

1 The Impact of Capital Structure on Agency Costs: Evidence from UK Public Companies He Zhang and Steven Li i International Graduate School of Business University of South Australia Adelaide, Australia Abstract This paper aims to provide empirical evidence on the agency costs hypothesis which suggests that increase of leverage may reduce agency costs. Both multivariate tests and univariate tests are employed in this study. The multivariate tests reveal that general relationship between leverage and agency costs is significantly negative. Univariate tests are further used to assess whether agency costs are significantly different when a firm has a relatively higher debt to asset ratio from when it is less leveraged. Similar supporting evidence is found for the agency costs hypothesis. Moreover, results from the univariate tests also indicate that this general negative relationship no longer holds when an extremely high level of leverage is present. Keywords: Agency costs, Leverage, Agency costs hypothesis, and Opposite effect 1. Introduction In their seminal work, Jensen and Meckling (1976) point out that agency costs occur due to incomplete alignment of the agent s and the owner s interests. The separation of ownership and control may generate agency costs. Two types of agency costs are identified in the paper by Jensen and Meckling (1976): agency costs derived from conflicts between outside equity holders and owner-managers, and conflicts between equity holders and debt holders. From then on, a great amount of research has been devoted to demonstrate the interaction between agency costs and financial decisions, governance decisions, dividend policy, and capital structure decisions. Much empirical evidence collected by researchers, for example, Ang et al. (2000), and Fleming et al. (2005), shows that agency costs generated from the conflicts between outside equity holders and owner-manager could be reduced by increasing the owner-managers proportion in equity, i.e., agency costs vary inversely with the manager s ownership. However, the conflicts between equity holders and debt holders would be more complicated. Theoretically, Jensen and Meckling (1976) argue that there should be an optimal capital structure, under which the lowest agency costs of a firm can be deduced from an independent variable --- the ratio of outside equity to the whole outside financing. The locus of agency costs, which is equal to agency costs of outside equity and the ones of debt, would be a convex curve. This implies that agency costs should not be monotonic any more. Some researchers such as Grossman and Hart (1982); Williams (1987), argue that high 1

2 leverage reduces agency costs and increases firm value by encouraging managers to act more in the interests of equity holders. This argument is known as the agency costs hypothesis. Higher leverage may reduce agency costs through the monitoring activities by debt holders (Ang et al., 2000), the threat of liquidation which may cause managers to lose reputation, salaries, etc. (William, 1987), pressure to generate cash flow for the payment of interest expenses (Jensen 1986), and curtailment of overinvestment (Harvey et al., 2004). However, as the proportion of debt in the capital structure increases beyond a certain point, the opposite effect of leverage on agency costs may occur (Altman, 1984 and Titman, 1984). When leverage becomes relatively high, further increases may generate significant agency costs. Three reasons are identified in the literature which can cause this opposite effect: first reason is the increase of bankruptcy costs (Titman 1984). Second reason is that managers may reduce their effort to control risk which result in higher expected costs of financial distress, bankruptcy, or liquidation (Berger and Bonaccorsi di Patti, 2005). Finally, inefficient use of excessive cash used by managers for empire building would also increase agency costs (Jensen, 1986). The main purpose of this paper is to provide some evidence on the effect of the leverage on agency costs by using the UK data. We shall focus on two questions: I. Does the agency costs hypothesis hold and whether higher leverage can reduce agency costs? II. When leverage has already been extremely high, whether further increases will result in the opposite effect of the agency costs hypothesis? The remainder of this paper is organized as follows: In Section 2 we undertake a general literature review. Section 3 discusses the data and variables employed in the tests and the methodology of this study in details. Section 4 presents the empirical results. Finally, conclusions and future research directions will be given in Section 5. 2.Literature Review Jensen and Meckling (1976) identify agency costs derived from conflicts between equity holders and owner-managers as residual loss which means agent consumes various pecuniary and non-pecuniary benefits from the firm to maximize his own utility. Related to this issue, Harris & Raviv (1990), Childs et al. (2005) and Lee et al. (2004) argue that managers always want to continue firm s current operations even if liquidation of the firm is preferred by investors. Also, Stulz (1990), Alvarez et al. (2006) and Kent et al. (2004) suggest the manager always want to invest all available funds even if paying out cash is better for outside shareholders, and conflict between the manager and equity holders cannot be resolved through contracts based on cash flows and investment expenditures. Agency theory becomes more complicated when debt holders interest is considered. As a 2

3 financing strategy, debt is widely discussed in capital structure literatures. Modigliani and Miller (1963) demonstrate that in order to raise the value of a firm, the amount of debt financing should be as big as possible for tax subsidy ii. However, their theory ignores the agency costs of debt. Theoretically, Jensen and Meckling (1976) point out that the optimal utilization of debt is when the debt is utilized to the point where marginal wealth benefits of the tax subsidy are just equal to the marginal wealth effects of agency costs. Generally, the agency costs associated with debt consist of the opportunity wealth loss, which is caused by impact of debt on investment decisions of the firm, bankruptcy costs and monitoring and bonding expenditures (Jensen and Meckling, 1976). The most sever conflict of interests between them would be that they have different claims on cash flows (Jensen 1986). Equity holders have a residual claim on cash flows. What they care about are profits and earnings from the projects they invest on. They may accept any projects which will increase firm s value no matter how risky they are. However, debt holders do not only share profits and earnings with equity holders, but also have a fixed claim on cash flows, which is the interest of debt. Therefore, what they focus on is the security of their claims. This conflict between equity holders and debt holders may affect a firm s decisions on three dimensions: investment, financing strategy and dividend distribution (DeMarzo and Fishman 2007). Debt holders may restrict manager s investment on very risky projects even though they may bring high returns ( Kalcheva and Lins,2007). Furthermore, as soon as the amount of debt increases, debt holders will be more and more powerful, and their interferences in firm s investment decisions will increase accordingly (Margarits and Psillaki 2007). A number of researchers focus on the issue of improvement of firm efficiency by reducing agency costs. Some of them focus on the methods to control managers behaviors. For instance, Fama (1980) conducts a discussion of how the pressure from managerial labor markets helps to discipline managers. He points out that the key condition to acquire absolute control of managerial behavior through wage adjustments is that the weight of the wage revision process is sufficient enough to resolve any potential managerial incentives problems. Another example is Chance s (1997) argument on a derivate substitution of executive compensation. He suggests giving the manager stocks without right to vote, which could be beneficial in preventing an executive from wielding too much control. Other researchers are interested in the optimal capital structure under which value of firms could be maximized while agency costs could be minimized. Based on these observations, the agency costs hypothesis stating that the leverage affects agency costs is put forward. Jensen and Meckling (1976) argue that monitoring activities by debt holders will tend to increase the optimal level of monitoring and therefore will increase the marginal benefits. What s more, banks which are one of the major sources of external funds especially for small firms also play a crucial role in monitoring the activities of managers. As banks generally ask managers to report company statements honestly and to run the business efficiently with profit, bank monitoring complements equity holders monitoring which indirectly reduces owner-manager agency costs (Ang et al., 2000). Moreover, greater financial leverage may affect managers and reduce agency costs through the threat of liquidation, resulting in 3

4 personal losses for managers: their decreasing salaries, violated reputation, perquisites, etc. (Grossman and Hart 1982; Williams, 1987). This will also increase managers pressure to generate more cash flow to pay interest expenses (Jensen, 1986). As assets in place generate cash flow that lead to either overinvestment or the outright diversion of corporate funds (Hart and Morre, 1998), debt is likely to curtail overinvestment (Harvey et al. 2004) because interests need to be paid. Additionally, Fleming et al. (2005) argue that compared to the issue of new equity, issuing debt increases the manager s equity holding as a proportion of equity financing, which then further enhancing alignment of interests. However, as suggested by Jensen and Meckling (1976), the effect of leverage on total agency costs could not be monotonic iii. When the proportion of debt in total capital increases beyond a certain point, the loss would increase due to negative net present value projects, and the firm will not be able to meet current payments on a debt obligation, thus bankruptcy will occur (Terje et al. 2006). Although Haugen and Senbet (1978) argue that bankruptcy costs are an insignificant determinant of a firm s capital structure, Altman (1984) finds that indirect costs associated with bankruptcy are not insignificant when these costs are accounted for the first time. Titman (1984) gives a possible theoretical link between liquidation and capital structure. It links the potentially substantial costs associated with liquidation with the event of bankruptcy. Furthermore, Berger and Bonaccorsi di Patti (2005) suggest that in highly leveraged firms, managers may shift risk or reduce effort to control risk which would also result in expected costs of financial stress, bankruptcy, or liquidation. Additionally, inefficient use of excessive cash which is derived from higher than normal leverage level for empire building would also increase agency costs (Jensen, 1986). Therefore, at low level of leverage, increases of leverage will produce positive incentives for managers and reduce total agency costs by reducing the counterpart of external equity. However, after reaching a certain point, where bankruptcy and distress become more likely and the agency costs of outside debt overwhelm the agency costs of outside equity, any further increases in leverage will then result in higher total agency costs. The subject of the measurements of the agency costs magnitude and firm performance has been widely discussed in the literature. These measurements are usually taken by using ratios fashioned from financial statements or stock market data. Ang et al. (2000) made one of the first attempts to measure the magnitude of agency costs by two ratios from financial statements. First ratio is a proxy for the so-called direct agency costs. In order to facilitate comparisons, it is standardized as operating expenses to sales ratio. Second ratio is a proxy for the loss in revenues attributable to inefficient asset utilization. This type of agency costs is derived from management s shirking or from poor investment decisions. This ratio is calculated by annual sales to total assets. Berger and Bonaccorsi di Patti (2006) take a different approach and employ profit efficiency as the performance measure. They use profit efficiency, rather than cost efficiency to evaluate the performance of managers, since profit efficiency explains how well managers raise revenues while reduce costs and it processes tighter relationship with the concept of value maximization. Additionally, Saunders et al. (1990); Cole and Mehran (1998) use stock market returns and their volatility to measure 4

5 agency costs and firm performance. As suggested by Harris and Raviv, (1990), test of the agency costs hypothesis typically are based on regressions of agency costs measures on the indicator of leverage plus some control variables. However, they argue that regressions of agency costs on a measure of leverage may confound the effect of capital structure on agency costs with the effect of agency costs on capital structure. They conduct a two-equation structural model and estimate it using two-stage least squares (2SLS) because they argue that if agency costs affect the choice of capital structure, then failure to take this reverse causality into account may result in simultaneous-equation bias (Berger and Banaccorsi di Patti, 2006). Univariate tests are also widely used in tests of agency costs determinants. Ang et al. (2000) and Fleming et al. (2005) use the t-test methodology and the Mann-Whitney U-test methodology to test the significance of differences of agency costs between firms managed by owners and firms managed by outsiders. 3. Data and Methodology Data used in this study are drawn from Datastream. 323 UK companies are selected from FTSE ALL SHARE index. We choose UK public companies in this study because of three reasons: First, The UK is a country with mature money and capital markets where debt financing is relatively easy to conduct by companies. Second, maximization of shareholders wealth is the dominant goal of management in the Anglo-American world which is consistent with the theory this study is based on. Third, data of public companies could reflect the effect of leverage on agency costs more accurately and sensitively especially in the efficient markets like the UK. There are five variables used in this study. Table 1 provides a summary of these variables along with definitions. Following Ang et al. (2000) s study, we focus on measuring the direct agency costs which is the ratio of operating expenses to sales. This ratio indicates how effectively the firm s management controls operating expenses and it tends to capture the impact of agency costs such as excessive perquisite consumption. Operating expenses variable here excludes corporate wages, salaries and other labor-related items, interest expense, rent, leasing and hiring expenses, purchases, depreciation and bad assets written off. A series of checks and filters on the data have been conducted to reduce the sample from a maximum of approximately 400 iv firms to a final sample of 323 firms for Year 2004 to Year The top and bottom 5% v are also removed to avoid the possible outlier effect. The measurements of leverage and agency costs are critical. Debt to asset ratio is employed which is total debts divided by total assets. However, we do not differentiate between long-term or short-term debt. Three other variables are considered to control other confounding effects: performance (proxied by return on asset), firm size (proxied by log of sales), and industry classification (13 industry dummies). Note that we include 13 industry dummy variables in this study because the ratio of operating expenses to sales varies across 5

6 industries due to the varying importance of inventory and fixed assets. Variable Operating expenses to sales (denoted as OETS) Debt to asset ratio (denoted as DTAR) Log of sales (denoted as LOS) Return on assets (denoted as ROA) Industry dummy variables Table 1. Variable Definitions Definition This ratio is calculated as operating expenses divided by annual sales. Operating expenses excludes corporate wages, salaries and other labur-related items, interest expense, rent, leasing and hiring expenses, purchases, depreciation and bad debts written off(%) This ratio is calculated as total debt divided by total assets. This is the ratio to measure the leverage. Natural log of annual sales. This variable is used as a control of firm size. Return on assets (%) 13 industry dummy variables as per Table 2 with a value of 1 for the firms in the industry and 0 otherwise. Table 2 displays the ratio of OETS by one-digit industry dummies. It shows that the ratio varies from a low of for real estate to a high of for mining. The comparison of the ratio of OETS among industries is shown in Figure 1 below. Table 2. Industry Classification Industry OETS (%) Construction Health Elect&elec.eq Electricity Chemistry Eng. Machin Food product General Retalers Information Technology Hotel Media Mining Oil gas Real estate The descriptive statistics for the sample of 323 UK public companies for 2004 and 2005 are presented in Table 3. The mean value of the OETS of the sample firms is 0.25 for both

7 and 2005.Highest and lowest values of this ratio are 0.91 (0.93) and 0.02 (0.02) for 2004 (2005), respectively. Normality test also is undertaken for 2004 and The results (Chi^2(2) = [0.0000] *** for 2004 and Chi^2(2) = [0.0000]*** for 2005) suggest that the OETS is not normally distributed. Figure 1. Industry Comparisons O E T S (%) Construction Health Elect&elec.eq Electricity Chemistry Eng. Machin Food product General Retalers Information Technology Hotel Media Mining Oil gas Real estate I n d u s t r y Table 3. Descriptive statistic of variables used to analyze agency costs Mean Maximum Minimum Standard Excess Skewness deviation Kurtosis Panel A: 2004 (N=323) OETS DTAR LOS ROA Panel B: 2005 (N=323) OETS DTAR LOS ROA The agency costs hypothesis states that an increase in leverage or a decrease in the equity to asset ratio should lead to a reduction of agency costs of outside equity. This is tested by following two approaches: Based on the data of 323 UK public companies in year 2004 and 2005, we first run the 7

8 following regression: OETS = f ( DTAR, Z) + (1) vi e i where the vector Z represents three control variables : natural logarithm of annual sales, return on assets, and the 13 industry dummy variables. Additionally, e i is a mean-zero disturbance term. According to the agency costs hypothesis, an increase in leverage, i.e., an increase in DTAR is assumed to raise efficiency, namely, ( OETS ) / ( DTAR) < 0 (2) As a lower leverage tends to reduce the pressure on managers to maximize firm value, thus it tends to aggravate the agency problems between managers and owners and reducing profit efficiency. Therefore, it is natural to run the regression above to assess the relation between OETS and DTAR, and the significance of the parameters. In order to further understand the relation between agency costs and leverage, univariate tests will be conducted as well. To this end, we need to select a subsample from the 323 companies. For a company to be included in the subsample, it must meet the following criteria: There is a significant change of DTAR for the company where the mean value of DTAR during 10 years is used as a benchmark. The 10 years sample period can be divided roughly into 2 subperiods. The DTARs are stable in each subperiod. The one in which the company is relatively less/more leveraged is called lower leverage stage (denote as LLS)/higher leverage stage (denote as HLS). Each stage must last at least 2 years. After extensive investigation of all available data, a sample of 168 companies (from year 1995 to year 2005) is selected for univariate tests. Table 4. Ratios of the SHANKS GROUP from 1995 to 2005 Years DTAR (%) OETS (%)

9 To further illustrate the division of stages, we use SHANKS GROUP in Table 4 as an example. The average of DTAR during the 10 years is 32.65%. The period from 1996 to 1998 is defined as LLS when the SHANKS GROUP is relatively less leveraged with the DTAR ratio 14.43%, and the average OETS ratio 14.62%. The period form 2001 to 2005 can be defined HLS when the SHANKS GROUP is relatively more leveraged with the average DTAR 46.97% and the average OETS 8.46%. We calculate the averages of the OETS ratios of the 168 companies when they are less and more leveraged respectively. The agency costs hypothesis is then tested by comparing the means of debt to asset ratios of the two stages. The average of OETS when firms are less leveraged is presumed to be higher than it is when firms are more leveraged. The t-test is employed to test whether the differences of the averages between these two stages are significant. However, as discussed above, when leverage becomes extremely high, the sign of the relationship may change, as the agency costs of outside debt overwhelm the reduction in agency costs of outside equity, further increase in leverage may result in higher total agency costs. We look for the some empirical evidence related to this statement as well. 4. Results and Analysis 4.1. Multivariate Tests The main purpose of these tests is to assess the general relation between agency costs and leverage, and whether this relation is statistically significant. As suggested by the agency costs hypothesis, the relation is expected to be negative. For this purpose, we use agency costs (proxied by OETS) as the dependent variable and the independent variables including: leverage (proxied by DTAR), and three control variables: firm size (proxied by LOS), firm performance (proxied by ROA), and industry identification (set by 13 industry dummies, IND k =1 for industry k, 0 otherwise, k=1, 2,..13). The following regression is proposed for the purpose: OETS = α + γ DTAR + γ LOS + γ ROA + β IND + ε k = 1 k k i (3) where α is the intercept term, γ, γ, γ, β, β,..., β are coefficients, ε i is the error term industry dummies are included in the regression as well. In addition, we also consider a few simple linear regressions which also include industry dummy variables. 13 OETS = α + γ1dtar + βkindk + i k = 1 ε (4) OETS = α + γ LOS + β IND k k ε i (5) k = 1 9

10 OETS = α + γ ROA + β IND k k ε i (6) k = 1 Before running the regressions, we first consider the correlations among variables in the regressions. Two correlation coefficient matrixes are presented in Table 5. Obviously, there is a negative relation between OETS and DTAR both in Year 2004 and 2005, which is consistent with the agency costs hypothesis. The correlation coefficients are and respectively. Additionally, OETS is negatively related with LOS and ROA with correlation coefficients of and in Year 2004, and in Year What s more, as the values of correlation coefficients are all below 0.5 in the absolute term, there should be not much evidence of sever multicollinearity amongst the variables vii. Therefore, a linear regression is employed for the multivariate tests. Table 5. Correlation Coefficients Matrix Panel A: 2004 (N=323) OETS DTAR LOS ROA OETS 1 DTAR LOS ROA Panel A: 2005 (N=323) OETS DTAR LOS ROA OETS 1 DTAR LOS ROA Now let us turn to the regressions. Table 6 reports the results for multivariate regression (3) and the simple regression (4)-(6). Column 1 identifies the explanatory variables and column 2 to 5 display parameter estimates for the regressions (3)-(6). In columns 2 through 4 the leverage and each control variables are analyzed independently. In Column 5, this study tests whether the independent results stand up when all four variables are included in a single regression. Because of the importance of industry structure as discussed in Section 3, 13 industry dummies are included in all regressions. In Column 2, the relation between the leverage and agency costs is tested, the parameter estimates -0.06* (t = -1.78*, F = 2.96*) in Year 2004 and -0.09* (t = -1.65*, F = 2.73*) in Year 2005 suggest that there is an inverse relationship between agency costs and DTAR as expected. This result supports the agency costs hypothesis, i.e., higher leverage can reduce 10

11 agency costs. The coefficient is only 90% significantly different from zero. Similarly, Column 3 suggests that there is a significant negative relation between firm size and agency costs. Column 4 indicates a negative relation between agency costs and firm performance but the relation is not significant. Column 5 gives the results of Model (3). The leverage coefficient exhibits the predicted negative sign, and the coefficient is still 90% significantly differently from zero. The firm size and firm performance coefficients exhibit the inverse relation with agency costs as well. Table 6. Variables Used in the Linear Regression Model (4) Model (5) Model (6) Model (3) Panel A: 2004 (N=323) Intercept 0.26*** 0.62*** 0.25*** 0.63*** (t-value) (16.7) *** (8.03) *** (20.8) *** (8.13) *** DTAR -0.06* -0.06* (t-value) (-1.78) * (-1.72) * LOS -0.03*** -0.03*** (t-value) (-4.84) *** (-4.86)*** ROA (t-value) (-0.21) industry dummies Yes Yes Yes Yes RSS F-statistic 2.96* 23.39*** *** Panel B: 2005 (N=323) Intercept 0.27*** 0.62*** 0.26*** 0.65*** (t-value) (16.9) *** (8.00) *** (19.7) *** (8.22) *** DTAR -0.09* -0.10* (t-value) (-1.65) * (-1.79) * LOS -0.03*** -0.03*** (t-value) (-4.82) *** (-4.82)*** ROA (t-value) (-0.63) industry dummies Yes Yes Yes Yes RSS F-statistic 2.73* 23.18*** *** *, ** and *** indicate significance at 10%, 5% and 1% level, respectively Univariate tests 11

12 As indicated by the multivariate tests, the relation between agency costs and leverage are significantly negative at 10% level. The results support the agency costs hypothesis. However, the relation revealed in the multivariate tests seems to be too general. In this section, the univariate tests are conducted to further test the effect of leverage on agency costs more specifically. The univariate tests will focus on the significance of the difference in agency costs between more leveraged stage and less leveraged stage of the selected firms. There are two questions need to be addressed: first, whether higher leverage can significantly reduce agency costs? Second, if the leverage is sufficiently high, whether the sign of relationship changes? For this purpose, the 168 companies are divided into 7 groups according to their highest debt to asset ratio in the decade: greater than 1, between 1 and 75%, between 75% and 50%, between 50% and 40%, between 40% and 30%, between 30% and 20%, and between 20% and 10%. As discussed above each firm has been divided into two stages: one stage with higher debt to asset ratios, and the other with lower debt to asset ratios. Therefore, this study investigates the differences of the OETS between these two stages viii : LLS and HLS. The results are displayed in Table 7. Table 7 reveals that the opposite effect of agency costs hypothesis: agency costs of firms when they are in HLS are insignificantly more than the ones when they are in LLS provided that their leverage levels have already been sufficiently great. The average of the OETS of the HLS is higher than the LLS but the difference is not significant when the highest leverages are above 1. When the highest leverages are between 1 and 75%, the difference will decrease to As discussed widely in the literature, the reasons for the results are threefold. Firstly, direct and indirect bankruptcy costs which involved in agency costs may occur if the leverage is too high (Titman, 1984). Secondly, in highly leveraged firms, managers may shift risk or reduce effort to control risk which would also result in expected costs of financial stress, or liquidation (Berger and Bonaccorsi di Patti, 2005). Finally, inefficient use of excessive cash which is derived from higher than normal leverage level for empire building would also increase agency costs (Jensen, 1986). However, the situation reverses when firm s highest leverages fall below 50%. In this case, firm s agency costs are smaller when they are in HLS than the ones when they are in LLS, i.e., leverage can reduce agency costs. The differences are -3.53, -5.64, -2.79, and when the highest debt to asset ratios are between 50% and 40%, between 40% and 30%, between 30% and 20%, and between 20% and 10%,respectively. The differences are also considerably significant. Especially when the top leverages fall below 40%, all the differences are significant at 1% level. These results indicate that the agency cost hypothesis holds. The leverage could reduce agency costs because monitoring activities by debt holders (Ang et al., 2000), the threat of liquidation which may cause managers to lose reputation, salaries, etc. (William, 1987), pressure to generate cash flow for the payment of interest expenses (Jensen 12

13 1986), and curtailment of overinvestment (Harvey et al. 2004). Table 7. Comparison of Agency Costs in Different Leverage Stages The highest DTAR in the decade is The highest DTAR Differences of firm s DTAR Firms OETS in HLS Firms OETS in LLS Differences of the OETS More than 1 Between 1 and 75% Mean Mean Mean Mean Mean (t = 0.97) (t = 0.17) Between 75% and 50% (t = -0.92) Between 50% * and 40% (t = -1.71)* Between 40% *** and 30% (t = -4.80)*** Between 30% *** and 20% (t = -2.82)*** Between 20% *** and 10% (t = -7.04)*** This table reports the mean statistics and t-statistics fort 7 groups of the 168 sample companies. The last column gives the difference between the mean of the expenses to sales ratios of firms when they are at the higher leverage stage and the lower leverage stage. Statistical significance of the differences in the mean ratios is based on the t-statistic from a parametric test (based on the assumption of unequal variance) of whether the difference in the mean ratios of the two stages is significantly from zero. *, ** and *** indicate significance at 10%, 5% and 1% level, respectively. To test whether the effect of leverage on agency costs is stronger if the differences of leverages are larger, we need to divide the sample companies into subgroups based on the difference of the debt to asset ratios between the HLS and LLS. Similarly as before, we divide the sample into 7 groups in which the differences of the debt to asset ratios between the HLS and LLS are more than 1, between 1 and 50%, between 50% and 40%, between 40% and 30%, between 30% and 20%, between 20% and 10%, and less than 10%. The results are displayed 13

14 in Table 8. If the difference of DTAR is more than 1, the firms agency costs are bigger when they are in HLS, because the leverage of HLS is extremely high. The result is similar with the ones in Table 7, but in Table 8, the difference becomes significant at 10% level. Unfortunately, there are no significant differences of agency costs between the two stages from Groups 2 and 7. Therefore, there is no evidence showing that the difference of agency costs may become more significant as the differences of leverage getting larger as well. Table 8. Differences of DTAR between the HLS and LLS Comparison of Agency Costs in the Different Leverage Stages Differences of DTAR OETS in HLS OETS in LLS Mean Mean Mean Mean Differences More than 1 (Group 1) * (t=1.66)* Between 1 and % (Group2) (t=-0.13) Between 50% and 40% (Group 3) (t=0.21) Between 40% and 30% (Group 4) (t=-0.48) Between 30% and 20% (Group 5) (t=-0.72) Between 20% and 10% (Group 6) (t=-0.48) Less than 10% (Group 7) * indicates significant at 10% level (t=-0.04) Basically, there are two reasons for these results: First, it may be due to the limitation of the sample size: the sample size in this study may not be big enough to reflect the reality, although effort has been made to get as much data as possible. Second, the variability of the two stages in each group is too large which results in that the t value becomes relatively small. As we know, in this case, firms are grouped according to the differences of the debt to asset 14

15 ratios between HLS and LLS. However, in each group, some companies themselves may keep a relatively high leverage level during the decade, and some may be lowly leveraged. For example, the differences between HLS and LLS in company CAPITAL & REGIONAL and company ALEXON are 0.29 and 0.22 respectively, indicating that they both belong to Group 5. However as shown in Table 9, leverages and agency costs of these two firms are quite different. Therefore, the variability of two stages in each group may be large, resulting in small t values. Table 9. The Comparison of the Leverages and Agency Costs of Two Firms CAPITAL & REGIONAL ALEXON Year Debt to asset ratio Operating expenses to sales ratio Debt to asset ratio Operating expenses to sales ratio In short, the results of univariate tests suggest that agency costs are negatively related to leverage, and the increase in leverage can reduce agency costs because of monitoring activities by debt holders (Ang et al., 2000), and manager s effort to avoid personal losses to managers of salaries, reputation, perquisites (Grossman and Hart 1982; Williams, 1987), pressure to generate cash flow for the payment of interest expenses (Jensen 1986), and curtailment of overinvestment (Harvey et al. 2004). But when the leverage is extremely high, further increase may switch to the opposite relation due to the increase of bankruptcy costs (Titman, 1984), less effort to control risk (Berger and Bonaccorsi di Patti, 2005), and inefficient use of excess cash (Jensen 1988). However, this relation is not statistically significant. We test whether effect of leverage on agency costs becomes stronger when the differences of leverages of firms at different leveraged stages getting larger. However, we can not find any significant evidence which may be due to the limitation of the sample size, and/or large variability of the data. 5. Conclusion In this paper, we provide empirical evidence for the agency costs hypothesis by conducting both the multivariate and univariate tests based on data on UK listed companies. In the 15

16 multivariate tests, the negative relation between leverage and agency costs is confirmed. The results suggest that the inverse relation is significant at 10% level. In addition, it is found that the firm size is negative related to agency costs with a significant level of 1% and firm performance is inversely related to agency costs but insignificantly. The results in univariate tests are similar to the results in the multivariate tests that the increase of leverage does reduce agency costs. However, when the leverage is sufficiently high, further increase in leverage will result in the opposite effect of the agency costs hypothesis. But this opposite effect is not significant. No significant evidence is found when testing whether the effect of leverage on agency costs becomes stronger when the differences of leverages of firms at different leveraged stages getting larger, In sum, our empirical results appear to support the agency costs hypothesis. Endnotes: i Corresponding author: steven.li@unisa.edu.au, Fax number ii Modigliani and Miller (1958) initially demonstrate that firm s value is independent of the capital structure in the absence of bankruptcy costs and tax subsidies on the payment of interest. They later (1963) argue that in order to raise the value of the firm, the amount of debt financing should be as big as possible for the tax subsidy. Finally, when considering tax rate on income from shares (denote as t-ps ) and bonds (denote as t-pb ), Modigliani and Miller (1977) assert that if t-ps< t-pb, firms will gain fewer benefits from leverage than the tax subsidy of debt. iii See Jensen and Meckling (1976), in their paper, they describe and portray the locus of agency costs on the left part of the convex (on the left of the bottom), agency costs would be reduced as the percentage of debt financing increases. That is to say, the leverage could reduce agency costs. iv Although there are 800 firms in FTSE ALL SHARE index, only nearly 400 firms have the full data which can be obtained in Datastream. v We select the data based on the ratio of expenses to sales: i. e. companies with average top and bottom 5% ratio of expenses to sales are discarded. vi At this stage, we have not arrived at that the relation is linear. A liner relationship will be established gradually after the following discussion. vii viii This study compares the differences of the operating expenses to sales ratios between the two stages rather than seven groups. The reason why 168 companies are grouped based on the highest debt to asset ratios they used to have is because the second goal of this study, i.e. whether the sign of relationship switch if the leverage has been sufficient great. References: Agrawal, A.; C.Knoeber, (1996), Firm performance and mechanisms to control agency problems between managers and shareholders, Journal of Financial and Quantitative Analysis, 31, pp Allen, F.; R. Michaely, (2003), Payout Policy, in GM Costantinides et al (eds) Handbook of The Economics of Finance, 1A, Corporate Finance (Elsevier North Holland), chapter 7, 16

17 pp Alvarez, H.; J. Virtanen, (2006), A class of solvable stochastic dividend optimization problems: on the general impact of flexibility on valuation, Economic Theory, 28, pp Ang, J.S.; R.A. Cole; Lin, J.W., (2000), Agency costs and ownership structure, Journal of Finance, 5, pp Berger S., E. Banaccorsi di Patti, (2006), Capital Structure and firm performance: A new approach to testing agency theory and an application to the bank industry, Journal of Banking & Finance, 30, pp Bhattacharya, U.; B., Ravikumar, (2001), Capital markets and the evolution of family business, Journal of Business, 74, pp Chance, D. M., (1997), A Derivate Alternative as Executive Compensation, Financial Analyst Journal, 53, pp Cole, R.A. ; H., Mehran, (1998), The effect of changes in ownership structure on performance: Evidence from the thrift industry, Journal of Financial Economics, 50, pp Donald, G.; T. Burk; H. Alan, (2003), Innovation in Small Businesses: Culture and Ownership Structure Do Matter, Journal of Developmental Entrepreneurship, 8, pp1-31. DeMarzo, P., M., Fishman, (2007), Agency and Optimal Investment Dynamics, The Review of Financial Studies, 20(1), pp Demsetz, H.; K. Warther, (1998), Dividends, asymmetric information, and agency conflicts: Evidence from a comparison of the dividend policies of Japanese and USA firms, Journal of Finance 53, pp Drew, A.; C. Kelley; Kendrick, T., (2005), CLASS: Five elements of corporate governance to manage strategic risk, Business Horizons, 49, 2, pp Fama, E.; M.C Jensen, (1983), Separation of ownership and control, Journal of Law and Economics, 26, pp Fleming, G.; R. Heaney; R. McCosker, (2005), Agency costs and ownership structure in Australia, Pacific-Basin Finance Journal, 13, pp Grossman, S.J.;O.Hart, (1982), Corporate financial structure and managerial incentives. The economics of information and uncertainty, University of Chicago Press, pp Harris, M.; A. Raviv, (1990), Capital structure and the informational role of debt, Journal of Finance, 45, pp Harris, M.; A. Raviv, (1990), The Theory of Capital Structure, Journal of Finance, 46, pp Jensen, M. C.; W. H. Meckling, (1976), Theory of the firm: managerial behavior, agency costs and ownership structure, Journal of Financial Economics, 3, pp Jensen, M. C., (1986), Agency cost of free cash flow,corporate finance and takeovers. American Economics Review,76, pp Kalcheva, I.; K., Lins, (2007), International Evidence on Cash Holdings and Expected Managerial Agency Problems, The Review of Financial Studies, 20(4), pp Kent.B., M. Smith, (2006), In search of a residual dividend policy, Review of Financial Economics, 15, 1, pp1-18. Lee, P. M., H. M. O'Neill, (2003), Ownership Structures and R&D Investments of U.S. and Japanese Firms: Agency and Stewardship Perspectives, Academy of Management Journal, 46, 17

18 2, pp Lerner, J., (1995), Venture capitalists and the oversight and management, Prentice-Hall, Englewood Cliffs, NJ, pp Margaritis, M.; M., Psillaki, (2007), Capital Structure and Firm Efficiency, Journal of Business Finance & Accounting, 84(9), pp Mauer, D.; S. Sarkar, (2005), Real option, agency conflicts, and optimal capital structure, Journal of Banking and Finance, 29, pp Mauer, D.; H. Ott, (2000), Agency costs, underinvestment, and optimal capital structure: The effect of growth options to expand, In: Brennan, M. J., Trugeorgis, L. (Eds.), Project Flexibility, Agency, and Competition. Oxford University Press, New York, pp Milgrom, P., J. Roberts, (1992), Economics, Organisation and Management, Prentice-Hall, Englewood Cliffs, NJ, pp Miller, M., (1977), Debt and taxes, Journal of Finance 32, pp Modigliani, F. and M. Miller, (1958), The costs of capital, corporation finance and the theory of investment, American Economic Review, 48, pp Modigliani, F. and M. Miller, (1963), Corporate income taxes and the cost of capital: A correction, American Economic Review, 53, pp Moyen, N., (2002), How big is the debt overhang problem?, Working paper, University of Colorado. Stulz, R., (1988), Managerial control of voting rights: Financing policies and the market for corporate control, Journal of Financial Economics, 20, pp Saunders, A.; Strock,; E. and Travlos, N, (1990), Ownership structure, deregulation, and bank risk taking, Journal of Finance 45, pp Seetharaman, A.; Z., Swanson; B., Srinidhi, (2001), Analytical and empirical evidence of the impact of tax rates on the trade-off between debt and managerial ownership, Journal of Accounting, Auditing & Finance 16, pp Terje, L; E., Aasmund; McKee, E., (2006), Bankruptcy theory development and classification via genetic programming, European Journal of Operational Research 169, 2, pp Titman, S.; S. Tsyplakov, (2002), A dynamic model of optimal capital structure, Working Paper, University of Texas. Titman, S.; (1984), The effect of capital structure on a firm s ;liquidation decision. Journal of Financial Economics 13, pp Vernimmen, P.; P. Quiry; Dallocchio, M.; Le Fur, Y.; Salvi, A., (2005), Corporate Finance: Theory and Practice, John Wiley & Sons, pp Williams, J., (1987), Perquisites, risk and capital structure. Journal of Finance, 42, pp