Capital Structure and Payout Policy
The effect of capital structure and payout policy on firm value
Evidence from S&P 500
MASTER THESIS
Faculty of Economics and Business Department of Finance1
Student: Ivar Bijl s3841030 i.bijl@student.rug.nl
Total words count (excl. appendix): 14,094 Words count (excl. tables, references): 10,386
Supervisor:
Prof. dr. Wolfgang Bessler
June 2020
1 Acknowledgement: I am grateful for the guidance, suggestions and comments from my supervisor: Prof. Dr.
ii
ABSTRACT
iii
TABLE OF CONTENTS
Abstract ... ii
Table of contents ... iii
List of tables ... v List of figures ... vi 1. Introduction ... 1 2. Literature review ... 4 2.1 Capital structure ... 4 2.2 Payout policy ... 5
2.3 Seasoned Equity Offerings ... 7
2.4 Cash positions ... 8 2.5 Industry effects ... 8 2.6 Size effects ... 9 2.7 Financial crisis ... 9 3. Methodology ... 10 3.1 Model specifications ... 10 3.2 Multicollinearity ... 11 3.3 Heteroskedasticity ... 11 3.4 Autocorrelation ... 12 3.5 Non-parametric tests ... 12 4. Data ... 13 4.1 Dataset description ... 13 4.2 Different panels ... 15 4.3 Subsamples ... 15 4.4 Limitations ... 16 5. Results ... 17
5.1 Raw data analysis and adjustments ... 17
5.2 Detail analysis ... 25
5.3 Analyses on Dividends, Share Repurchases and Seasoned Equity Offerings ... 27
6. Conclusion ... 30
References ... 32
Appendices ... 35
iv
v
LIST OF TABLES
Table 1: Descriptive statistics ... 16
Table 2: Results raw data analysis, using model I and Panel A ... 18
Table 3: Average actual debt ratio (%) per industry (2001-2018) ... 20
Table 4: Average actual debt ratio (%) per size category (2001-2018) ... 22
Table 5: Analysis on the effect of the financial crisis ... 24
Table 6: Results detail analysis, using model II and Panel B... 26
Table 7: Analysis on Dividends, Share Repurchases, and Seasoned Equity Offerings ... 28
Table B.1: Description of obtained variables ... 39
Table B.2: Correlation matrix ... 40
Table B.3: Average actual debt ratio (%) per sector per firm type ... 41
Table B.4: Frequency table by type of firms, separated by industry... 41
Table B.4: Dividend paying, repurchasing and issuing firms ... 42
Table C.1: Analysis on industry effects (Pooled OLS) ... 43
Table C.2: Analysis on industry effects (OLS regression with Newey-West standard errors) ... 45
Table C.3: Analysis on size effects (Pooled OLS) ... 46
Table C.4: Analysis on size effects (Newey-West) ... 47
Table C.5: Analysis on the effect of the financial crisis ... 48
Table C.6: Analysis on Dividends, Share Repurchases, and Seasoned Equity Offerings ... 49
Table D.1: Results Mann-Whitney Test (industries)... 50
Table D.2: Results Mann-Whitney Test (Crisis) ... 50
Table D.3: Results Mann-Whitney Test (Size) ... 51
Table D.4: Results Mann-Whitney test (all industries) ... 52
Table D.5: Mann-Whitney test (all size categories) ... 53
vi
LIST OF FIGURES
Graph 1: Capital structure on S&P 500 (2001-2018) ... 13
Graph 2: Dividends, Share Repurchases, Seasoned Equity Offerings and Market value ... 14
Graph 3: Average actual debt ratio (%) per industry (2001-2018) ... 19
Graph 4: Average actual debt ratio (%) separated by size (2001-2018) ... 21
Graph 5: Average ADR (%) of firms listed on the S&P 500 (2001-2018) ... 23
Graph B.1: S&P 500 index (2001-2018) ... 37
Graph B.2: Return on S&P 500 (2001-2018) ... 37
1
1. INTRODUCTION
In corporate finance, capital structure is the firm’s ratio of equity or debt financing to total financing or the ratio between debt and equity. For many years the Modigliani and Miller theorem, the trade-off theory (Kraus and Litzenberger, 1973) and the pecking order theory (Myers and Majluf, 1984) have been the basis of numerous studies which tried to find out whether an optimal capital structure exists, which maximizes firm value. So far, researchers cannot clearly define what the optimal capital structure of a particular firm is. The three main theories are as follows. Modigliani and Miller (1958) state that, under the assumption of perfect and complete capital markets the choice of the capital structure has no effect on firm value. While Kraus and Litzenberger (1973) state that the optimal capital structure depends on the size of tax savings and the cost on financial distress. In contrast, Myers and Majluf (1984) state in the pecking order theory that the choice of the capital structure is dependent on the associated costs. However, many studies have challenged the theories of Modigliani and Miller (1958) and (1961), the trade-off theory by Kraus and Litzenberger (1973) and the pecking order theory by Myers and Majluf (1984).
In general, investors would expect that managers always take decisions which increase firm value and that they are transparent in the decisions they make. Hence, all financing decisions made by managers should be value enhancing. However, in practice we do not always observe this expected behavior. The agency theory by Jensen and Meckling (1976) suggests that managers do not always act in the interest of shareholders, and hence decisions do not always lead to an increase in shareholders’ value. Due to information asymmetries and agency-problems, managers usually have more information than investors do and sometimes they make financial decisions, which are not value enhancing.
2
This discussion leads us to the central question of this paper: “Does the decision of managers with respect to capital structure, payout policy and seasoned equity offerings have a significant effect on firm value?” To answer this question, this study analyzes an unbalanced panel of 6,849 firm-year observations of 404 unique firms included in the S&P 500 index in the period 2001-2018. The motivation of the study is due to four reasons. Firstly, since the optimal capital structure is not yet determined, this paper can contribute to the level of empirical evidence for the optimal capital structure. Secondly, capital structure is and will be a hot topic in corporate finance in the coming years as there is still much to investigate. Thirdly, the S&P 500 index is the world’s largest index based on the market capitalization of their constituents. This makes it the best index to find evidence for the relationship between capital structure, payout policy and seasoned equity offerings on firm value for firms listed on the US stock markets. The index is a major player in the global economy and therefore a good representative in order to make inferences in the field of corporate finance in the US. Finally, the S&P 500 index consists of different industries which all have their own structure and financing strategies. Splitting the data into these industries offers us the opportunity to find relationships between capital structure, payout policy and raising equity on firm value.
The aim of this study is to find empirical evidence in the fields of corporate finance, particularly evidence, which supports the relationship between capital structure, payout policy and raising equity on firm value. The focus will be on the firms listed on the S&P 500 index in the period 2001-2018. As such, we begin with an analysis on the full sample, which means 404 firms during the period 2001-2018.
Two different models (I and II) are used to find empirical statistical evidence. Each model is run several times, for two different panel data regression methods. The first method is the pooled Ordinary Least Squared (OLS) regression analysis. This analysis assumes that the intercept and slope coefficient are constant over time and across firms. The second method is the OLS regression with Newey-West standard errors. The Newey-West estimators are consistent estimates when there is autocorrelation in addition to possible heteroskedasticity.
3
4
2. LITERATURE REVIEW
This section discusses some aspects of the corporate finance literature. The focus is on capital structure, payout policy, seasoned equity offerings, cash holdings, industry effects, size effects and the financial crisis effect. Next we discuss the empirical studies and the relationship between our financing variables and firm value.
2.1 Capital structure
Capital structure is defined as the firm’s choice between debt and equity financing. It is a topic that has been discussed for many decades. It starts with the seminal papers of Modigliani and Miller (1958) and (1961). They discuss that, under the assumptions of perfect markets, financing decisions are not shareholder value enhancing. Several other researchers challenge the analysis of Modigliani and Miller. Hamada (1969) and Rubinstein (1973) found the same relationship in terms of systematic risk, still under the same assumption of perfect markets. Although the findings of Modigliani and Miller only hold in a world without capital market frictions, they have become the basis for several corporate finance theories.
In the years after the publications of Modigliani and Miller, researchers started to relax the assumptions of perfect markets. They introduced capital market frictions such as taxes, bankruptcy costs and asymmetric information. Kraus and Litzenberger (1973) introduced taxes and bankruptcy costs into the model and developed the trade-off theory. This theory suggests that the total value of a levered firm equals the value of the firm without leverage, plus the present value of the tax savings, less the present value of financial distress costs. They showed that leverage has both benefits as well as costs, which is the trade-off in the choice of the optimal capital structure.
In contrast, Myers and Majluf (1984) introduced the pecking order theory. This theory states that the managers’ preferences for a specific financing instrument is related on the associated information asymmetry and costs. Managers have a specific financing order and they prefer to finance their activities first with retained earnings, then debt and in the end, they will issue new equity. In practice, it happens that firms do not follow the pecking order theory, firms often issue equity even when borrowing is possible (Leary and Roberts, 2010).
debt-5
ratio dynamics. Empirical evidence of Bessler et al. (2008) suggests, in line with Welch (2004), that a large part of the variation in leverage is caused by stock price movements. However, they also found differences in the dynamics of capital structure behavior and issuing activities. Those are consistent with a modified version, including market timing as short-term factor, of the dynamic trade-off theory. This finding is supported by studies conducted by Leary and Robert (2005), Hovakimian (2006), Alti (2006), Kayhan and Titman (2007) and Flannery and Rangan (2006). They all found that market timing is an important factor in determining the financing activities in the short run. The dynamic pecking order theory suggest that equity financing is preferred when information asymmetry between managers and investors is low (Bessler et al., 2011).
Information asymmetry usually means that managers’ information is superior relative to investors’ information. This asymmetric information may trigger managers to alter their previous capital structure decisions. The adverse selection problem, which means that managers have information that investors do not have, has implications for the capital structure. As earlier discussed, the pecking order hypothesis states that managers are willing to finance positive NPV-projects first with retained earnings, followed by debt and equity as a final source of capital. Therefore, the decisions by managers influence capital structure, and hence firm value. With above theoretical arguments in mind, we posit the following testable hypothesis:
Hypothesis 1: Capital structure has a positive relationship with firm value.
2.2 Payout policy
After successfully carrying out positive NPV-projects, firms have to choose how to allocate generated profits. At first, they can retain the profits and use it to invest in new positive NPV-projects, or they can distribute the cash flows to their investors. In this paragraph we will focus on the latter, the distribution of cash to shareholders.
6
and share repurchases are a signal of managers’ optimistic view on future cash flows. They state that the choice of payout policy is irrelevant, therefore dividends and share repurchases are interchangeable.
In contrast, John and Williams (1985) state that taxes paid on dividend income by investors is the cost of dividends, therefore dividends and share repurchases are not interchangeable. This finding finds support by Allen et al. (2000), who state that dividends attract institutions. Institutions can monitor the firm due to the fact they have better information-gathering abilities. Therefore, they can observe whether a firm is overvalued or undervalued. Since institutions prefer dividends to share repurchases, only firms that are undervalued signal that they are undervalued. Hence, dividends and share repurchases are not interchangeable.
Another reason for firms to change their payout policy is to distribute free cash flows to their investors. According to Grullon and Michaely (2002), in the past decade US firms have spent more money on share repurchases relative to dividends. In contrast to the period in the late 1990s, where the amount spent on dividends and share repurchases were almost equal. Under the assumption of the agency problem, managers may make large, unprofitable investments. The motivation to make these negative-NPV investments, is due to personal interest of managers. For example, managers are overconfident, and they try to build an empire. The free cash flow hypothesis, put forth by Jensen (1986), suggests that excess cash flows induce empire building and leads to unprofitable investments. According to this hypothesis, debt increases firm value because it commits the firm to making future interest payments. Moreover, an increase dividends or share repurchases lowers the excess cash flows of a firm, this may prevent managers to make negative-NPV investments (Oswald and Young, 2008). Nohel and Tarhan (1998) found support for the free cash flow hypothesis while focusing on US repurchasing firms.
7
of perfect and complete capital markets, Modigliani and Miller (1958) and Modigliani and Miller (1958) document that share repurchases do not affect share prices. With above theoretical arguments in mind, we expect positive relationships between payout policy and firm value. Therefore, we present the following testable hypotheses:
Hypothesis 2: Dividends have a positive relationship with firm value;
Hypothesis 3: Share repurchases have a positive relationship with firm value.
2.3 Seasoned Equity Offerings
The pecking order theory of Myers and Majluf (1984) state that firms have three sources they can use to finance their future investments. According to their theory, which is based on information asymmetries and agency costs, firms prefer to finance their operations with retained earnings, followed by debt, lastly, they would issue new equity. In this paragraph, the focus will be on the latter, raising new equity or in other words seasoned equity offerings (SEOs).
Previous research has found that firms do not strictly follow the pecking order theory (Leary and Roberts, 2010). This may explain why firms issue equity while borrowing debt is possible. Normally debt is cheaper and therefore managers will decide to finance their activities with debt rather than equity because it maximizes firm value due to lower costs. However, the trade-off theory explains how firms should choose their optimal capital structure to maximize firm value. Therefore, managers sometimes have to issue new equity to maintain the current optimal capital structure. Empirical evidence by Leary and Robert (2010) shows that it is possible in practice that firms sometimes rather issue equity than that they borrow debt. Baker and Wurgler (2002) found evidence that firms are more likely to issue equity when the market values are high, relative to book and past market values, and to repurchase equity when their market values are low. This is in-line with signaling theories and the findings on overvaluation and related to the market timing abilities of management. The market usually greets the news of a SEO with a price decline, this is consistent with the adverse selection problem. Scholes (1972) argues with his price-pressure hypothesis that equity issuance causes the stock price to decline. Previous research, such as Myers (1984), Choe et al. (1993) and Loughran and Ritter (1995), argue a negative stock price reaction on seasoned equity offerings. Loughran and Ritter (1995) for example, found a price drop of 3% after an SEO announcement. Yet, Dittmar and Thakor (2007) found evidence that when the level of agreement between managers and investors is high, managers only issue new equity when the NPV of a project is positive, hence value enhancing. With above theoretical arguments in mind, we postulate the following testable hypothesis:
8 2.4 Cash positions
Tax costs in the United States have been an important determinant of cash holdings of US companies. These firms and firms in other countries, have to pay tax on foreign income. However, US firms can defer the tax payment until firms repatriate their earnings. This is an incentive for multinationals to keep their earnings abroad, which results in large cash positions. One idea is that these firms could acquire other foreign firms and paying with profits before tax. Simone et al. (2019) state that they found a different pattern in cash holdings for domestic and international (multinational) operating companies. Therefore, there exists a link between the level of cash holdings and foreign and domestic revenues. These large cash holdings are in a way related to market values of firms, since firms can easily use these to pay off debt. Paying of debt with excess cash holdings has an effect on the capital structures and hence on firm value.
2.5 Industry effects
Industry effects are important factors for managers in determining the optimal capital structure. Managers sometimes use industry median debt levels as a target level for their own firms, although this is not their optimal capital structure. Besides, industry effects reflect correlated factors which otherwise would be omitted. For example, rules and regulations differ among industries and influence managers’ decisions on capital structure. Regulated firms tend to have lower expected costs of financial distress; hence a higher debt level is expected. Moreover, research and development (R&D) costs are, according to Binsbergen et al. (2010), also an important factor for capital structure decisions. They found that firms with high R&D costs tend to have low debt ratios. Since they have low current free cash flows, they need little debt to create a tax-shield. This finding is in line with Bah and Dumontier (2001), who state that R&D-intensive firms have lower debt levels than non-R&D intensive firms. Moreover, Binsbergen et al. (2010) also found that low growth, mature firms are likely to have high debt ratios due to stable free cash flows and high intangible assets.
9
real estate, energy and industrials have high debt levels according to Bradley et al. (1984). With these theoretical arguments in mind, we propose the following testable hypothesis:
Hypothesis 5: Capital structure is different among industries
2.6 Size effects
The effect of size on capital structure can be explained in twofold. On the one hand, Titman and Wessels (1988) argue a positive relationship between size and leverage. They argue that large firms are more diversified and have a lower probability of default, hence leverage increases with size. In addition, Warner (1977) argues that bankruptcy costs consist of a fixed part and a variable part, therefore the costs tend to be higher for small firms. Hence, they find a positive relationship between size and leverage. Accordingly, the trade-off theory by Kraus and Litzenberger (1973), predicts a negative relation between size and the probability of default. This implies a positive relationship between size and leverage.
On the other hand, the pecking order theory, suggests a negative relationship between size and leverage. Myers and Majluf (1984) argue that large firms prefer equity financing over debt financing, which suggests large firms to hold lower debt ratios. In addition, under the assumption of information asymmetries, large firms are closely monitored by investors. Therefore, large firms are more capable of issuing equity that is sensitive to information. With these arguments in mind, we expect to find differences between size levels and propose the following testable hypothesis:
Hypothesis 6: Capital structure is dependent on firm size.
2.7 Financial crisis
In times of crisis, firms may appeal to the government for financial assistance. Often, this results in the form of subsidized loans or loan guarantees. The latter were an important part in the government response in the global financial crisis in 2008. These loan guarantees enabled firms to obtain loans against lower interest rates compared to the case with no government help. As a result, capital structure of firms change in times of crisis. Berg and Kirshenmann (2010) argue that the financial crisis had an influence on the credibility of firms. Since banks were also affected by the financial crisis, they are only willing to lend money to firms which score high on credibility. Hence, a firm’s possibility to borrow money in the financial crisis is smaller which influence the capital structure decisions by managers. According to Fosberg (2012), the financial crisis had a major effect on firms’ capital structure. He found that firms significantly increased the amount of debt in the crisis. With these arguments in mind, we propose the following testable hypothesis:
10
3. METHODOLOGY
To investigate the relationship between capital structures and payout policy with firm value, we need different models, which help us to test the hypotheses stated in section 2. The models help to find an answer on the main question: “do capital structures and payout policy have an effect on market value of a firm?”.
3.1 Model specifications
To provide an answer on the central question and for the hypotheses as stated in section 2, we use two different models. The first model (I) is built to investigate economic observation, to compare different industries, and as a robustness check to support the detail analysis. Moreover, model I helps to substantiate the decisions made to build model II, which is used for a more detailed analysis. The model provides clear evidence which supports the questions in this paper. It helps to make decisions on statistical issues as, for instance, trimming, splitting the data and generating other variables.
Moreover, Model I provides us with the opportunity to analyze the relationship between the market-based capital structure (ADR) and the market value of the firm (MKOF). Furthermore, running model I with dummy variables, allows us to investigate the different industries in the S&P 500, different size quartiles, and the pre- and post-crisis periods. Two different panel data regression methods are applied to model I. The first method is the pooled Ordinary Least Squared (OLS) regression analysis. This analysis assumes that the intercept and slope coefficient are constant over time and across firms. The second method is the OLS regression with Newey-West standard errors. The Newey-West estimators are consistent estimates when there is autocorrelation in addition to possible heteroskedasticity. The results of these models are presented in section 5 and in the appendix.
MODEL I:
𝑀𝐾𝑂𝐹𝑖,𝑡 = 𝛼 + 𝛽1 𝐴𝐷𝑅𝑖,𝑡+ 𝛽2 𝐷𝑉𝐶𝑖,𝑡+ 𝛽3 𝑆𝑅𝑃𝑖,𝑡+ 𝛽4 𝑆𝐸𝑂𝑠𝑖,𝑡 (I)
Note: α is the intercept term, 𝑖 is firm’s identity and 𝑡 is time
One can find the descriptions and or calculations of the variables of model I in appendix A.
11
to investigate the effect of capital structure on firm value without considering market fluctuations in share prices and interest rates. In addition, size of the firm (SIZE) which is determined by logarithm of book value of equity is considered as controlled variable. Again, we apply the two different panel regression methods to model II, the pooled OLS regression and the Newey-West OLS regression.
MODEL II:
𝑃𝐵 𝑅𝑎𝑡𝑖𝑜𝑖,𝑡 = 𝛼 + 𝛽1𝐶𝑆𝑖,𝑡+ 𝛽2𝐷𝑉𝐶𝐵𝐸𝑄𝑖,𝑡+ 𝛽3𝑆𝑅𝑃𝐵𝐸𝑄𝑖,𝑡+ 𝛽4𝑆𝐸𝑂𝑠𝐵𝐸𝑄𝑖,𝑡+ 𝛽5𝑆𝐼𝑍𝐸 (II)
Note: α is the intercept term, 𝑖 is firm’s identity and 𝑡 is time
To measure the effect of dividends, share repurchases and equity issuance on firm value, a dummy variable approach is performed. The effect of dividends on firm value is measured with the dummy variable which takes value one for non-dividend firms (non_DVC) and zero otherwise. The effect of share repurchases is measured with dummy variable which takes value one for non-repurchasing firms (non_SRP) and zero otherwise. The effect of equity issuance measured with dummy variables which takes value one for non-issuing firms (non_SEOs) and zero otherwise. The explanation of all dummy variables used are in appendix A.
3.2 Multicollinearity
To get the data, and to find out which statistical model should be used, multiple tests need to be conducted. At first, to test for the presence of multicollinearity, three different methods are used. Combining the results provides sufficient evidence to observe the level of multicollinearity. The first method tests if the independent variables are jointly significantly different from zero. Added to that, observing a high R2 and insignificant variables can indicate the presence of multicollinearity. The second method involves analyzing the matrix of correlations of the individual variables, an observed correlation of 0.8 or higher indicates possible multicollinearity. A more formal method for measuring the extent of multicollinearity is the variance inflation factors method (VIF method). This method quantifies the severity of multicollinearity with an index, the larger the index, the more serious is the collinearity between the explanatory variables. As a rule of thumb, an index below five indicates that multicollinearity is assumed to be negligible. An index greater or equal to five suggests to undertake remedial actions. The results of these tests are presented in appendix D.
3.3 Heteroskedasticity
12
this does not provide statistical evidence. Fortunately, there are number of formal statistical tests for heteroskedasticity such as White’s test. His approach tests the null hypothesis, which states homoskedasticity: if chi2-stastistic is high and p-value is low one has enough evidence to reject the
null hypothesis and infer that the model suffers from unrestricted heteroskedasticity. White’s test is conducted for each model and the results are presented in the tables in section 5.
3.4 Autocorrelation
Panel data is also frequently subject to autocorrelation, a simple way to test for autocorrelation is plotting the residuals against residualst-1. However, graphical methods do not provide statistical
evidence, and hence formal statistical tests should be applied. Wooldridge developed an autocorrelation test for panel data with a null hypothesis that states: no first-order autocorrelation. If the F-statistic is high and the p-value is low, there is sufficient evidence to reject the null hypothesis and infer that the model suffers from autocorrelation. These results are presented in section 5.
Pooled OLS models which suffer from heteroskedasticity and autocorrelation are regressed with the Newey-West standard errors. This regression is an OLS regression with Newey-West standard errors. These estimators are consistent estimates when there is autocorrelation in addition to possible heteroskedasticity. The number of lags are calculated according to an econometrics textbook by Green. The formula is: number of lags = T1/4, where T is the number of years. The panel data used in this study has 18 years. Therefore, the number of lags used in this research is 2.
3.5 Non-parametric tests
13
4. DATA
In this section the dataset is presented in terms of data sources, descriptive statistics, subsamples and limitations. The sample includes yearly financial figures for companies between 2001-2018 included in the S&P 500 index.
4.1 Dataset description
The original dataset, gathered through the Compustat Annual Database, consisted of 8,451 firm-year observations for firms that are included in the S&P500 index. This index is representative, because it is based on the market capitalization of 500 large firms in the United States categorized in 12 industries. Since the data from the year 2019 was incomplete, it has been excluded from the sample. Moreover, the sector Financials (GIC code 40) and the sector Utilities (GIC code 55) have been omitted, because they have a specific nature of business. They are subject to specific rules and regulations, which have an impact on capital structure and payout policies. Therefore, 64 firms categorized as “Financials” and 28 firms categorized as “Utilities” are excluded. Leaving us with 404 unique firms in the period 2001-2018 in an unbalanced panel of 6,849 firm-year observations, which cover 10 different industries. An overview of the obtained variables is presented in table B.1 in the appendix.
Graph 1: Capital structure on S&P 500 (2001-2018)
Note: the graph shows the capital structure on the S&P500 in the period 2001-2018. Capital structure is measured as the market value of Equity divided by the total value of the firm (Market value of Equity + Book value of debt) also called equity-to-value ratio. The blue line is the percentage measured through accumulating all values and then dividing them to arrive at a capital structure for each year. The red line is the percentage of the median values in the sample for each year. The highest accumulated equity-to-value ratio is 72% and the lowest value is 50%. The highest median equity-to-value ratio is 70% and the lowest is 55%. The period between the dashed lines refers to the transition period of the financial crisis.
14
The median value of the capital structure, as presented in graph 1, varies between 61% and 72% for the period 2001-2018. However, this is clearly interrupted by the financial crisis in 2008 with a drop to an equity-to-value ratio of 50%. Furthermore, the capital structure is based on market values of equity, and the interruption is in line with the drop of almost 20% of the S&P 500 index presented in graphs B.1 and B.2, see appendix B. Graph B.3 in the appendix, shows the capital structure for all industries. We observe that all industries are converging to a common range. This trend is again clearly interrupted by the financial crisis in 2008.
Graph 2: Dividends, Share Repurchases, Seasoned Equity Offerings and Market value
Note: the graph shows the median values of the independent and dependent variables. All values are in $millions. Dividends, share repurchases, and seasoned equity offerings are gathered through the Compustat Annual Database. Market value is calculated as the sum of the book value of debt plus the market value of equity, where market value of equity is the product of share price (PRICE) and Common Shares Outstanding (CSHO), both gathered from the Compustat Annual Database. The period between the dashed lines refers to the transition period of the financial crisis. The vertical axis on the left presents the amount of dividends, share repurchases and seasoned equity offerings in million dollars, where the vertical axis on the right presents the amount of market values in million dollars. The horizontal axis shows the period in years.
The median values of dividends, share repurchases and market value, as presented in graph 2, follow a clear upward sloping trend with an interruption around the years of the financial crisis. Dividends increased from $20.7 million in 2001 to $308 million in 2018. However, the financial crisis interrupted this growth by a decrease of $7.71 million in 2009. Where dividends follow a steadier growth, share repurchases are much more volatile. In 2001, share repurchases were only $5 million versus an amount of $257 million in 2007. The impact of the financial crisis is clearly
15
visible since share repurchases decreased to $6.15 million in 2009. After the crisis, share repurchases increased significantly to an amount of $430 million in 2018. Market value increased from $5,039 million in 2001 to $19,687 million in 2018, again with an interruption in the years around the financial crisis. In contrast to the earlier mentioned variables, seasoned equity offerings remained roughly the same. They vary between $27.2 million in 2001 and $68 million in 2013 ending at an amount of $34.9 million in 2018. A more in-depth representation of the variables specified for each sector are presented in the appendix.
Trimming the data at the 5th and 95th percentiles leads to the descriptive statistics as presented in table 1. The average firm in the sample, as presented in table 1, is worth $26,577 million in market value (MKOF). However, the median firms if worth about $16,818 million. The Actual Debt Ratio (ADR), one of the explanatory variables and accounting for the capital structure, has a mean about 33% and a median of about 31%. A quick look at the other explanatory variables presents a mean of $280 million and a median of $89 million for dividends (DVC), a mean of $436 million and a median of $122 million for share repurchases (SRP), and a mean of $99 million and a median of $39 million for seasoned equity offerings (SEOs). The relationship between the variables can be found in correlation table B.2 in the appendix.
Panel B in table 1 reveals an average book value-based debt-to-equity ratio (CS) of 2.25 and a median of 1.38. Dividends (share repurchases) are around 6% (12%) the of book value of equity and seasoned equity offerings are around 4% of the book value of equity. The dependent variable, price-to-book ratio, is on average 4.73 (with a median of 3.18).
4.2 Different panels
The data is split into two different panels, each of them are linked to the two different models as discussed in section 3.1. The first panel (A) consists of 5,783 firm-year observations for the dependent variable, the variables have actual numbers in $millions, the descriptive statistics are presented in table 1. The second panel (B) consists of 5,843 firm-year observations for the dependent variable, the variables are presented as ratios and presented in table 1.
4.3 Subsamples
16
years 2001-2006, and post-crisis period is defined as the years 2010-2018. The years 2007-2009 are financial crisis period and account for the transition period. The crisis started in mid-2007 and ended in early 2009, therefore the full years 2007 and 2009 are specified as crisis.
4.4 Limitations
Cash positions, as discussed in section 2, are not used in this paper. Due to data limitation, no access to Eikon, we were unable to gather data on foreign and domestic revenues which are important determinants in the level of cash holdings in US companies. It is important to note, that cash positions have increased for US multinational firms, but not for domestic operating entities. Therefore, the much more data, which is not available, is required to implement a useful analysis.
Table 1: Descriptive statistics Variable
Name N Mean SD Min Median Max P5 P95
Panel A MKOF 5,783 26,577 27,440 1,000 16,818 164,172 3,040 87,235 ADR 5,783 0.33 0.17 0.01 0.31 0.97 0.08 0.63 DVC 6,396 280 473 0 89 2,768 0 1,320 SRP 6,140 436 700 0 122 3,826 0 2,031 SEOs 6,169 99 144 0 39 782 0 431 Panel B PB-ratio 5,843 4.73 7.14 0.24 3.18 246 1.20 11.90 CS 5,723 2.25 5.58 0.03 1.38 177 0.37 5.73 DVCBEQ 5,939 0.06 0.17 0 0.03 8.35 0 0.19 SRPBEQ 5,667 0.12 0.31 0 0.04 8.69 0 0.45 SEOsBEQ 5,660 0.04 0.09 0 0.02 1.65 0 0.14 SIZE 6,115 8.01 1.14 4.83 8.01 10.50 6.07 9.89
Note: the table presents the winsorized descriptive statistics of the full sample in $millions. MKOF is Market value of Firm, ADR is Actual Debt Ratio and calculated by dividing the book value of debt by market value of firm, DVC is dividends payments on common stock in $million, SRP is Share Repurchases in $millions, SEOs is Seasoned Equity Offerings in $millions. N is the number of observations, SD is the standard deviation, P5 is the 5th percentile in the
sample, P95 is the 95th percentile in the sample. PB-ratio stands for the price-to-book ratio for each company each year
17
5. RESULTS
The results of our empirical analysis are presented in three sections. We first provide an overview of the raw data analysis and discuss the appropriate adjustments. Secondly, a detail analysis is performed to better compare the firms in our sample. Lastly, a dummy variable approach is conducted to analyze the effects of dividends, share repurchases and seasoned equity offerings.
5.1 Raw data analysis and adjustments
The objective of performing a raw data analysis, is to focus on economic observations with respect to the relationship between capital structure, payout policy and firm value. Moreover, the raw data analysis allows us to make the appropriate decisions in the context of statistical models. All findings in this section 5.1 are calculated according to equation (I) and panel A.
To get insights into the consequences of omitting zero-values and data trimming, models 1-4, as presented in table 2, are applied. We observe that omitting zero-values, see model 3 and 4, dramatically decrease the number of observations in the sample to 2,236 (6,140 without omitting). Therefore, we decide to keep the zero-values in the data and use a dummy variable approach instead. This approach allows us to analyze the effects of dividends, share repurchases and equity issuance. This detail analysis is performed with model II and panel B and is presented in section 5.2. Furthermore, to eliminate potential outliers, the data are cut-off at the 5th and 95th percentile and leaves us with 4,980 observations. This adjustment leads to the estimation results of the pooled OLS regression presented in model 5, table 2. As discussed in section 3, panel data is often subject to the presence of heteroskedasticity and autocorrelation. The appropriate tests, White’s test and Wooldridge test, are conducted and their results are presented in table 2. The Pooled OLS models (models 1, 3 and 5) reveal, at a 1% significance level, that we observe heteroskedasticity and autocorrelation. One approach to deal with heteroskedasticity and autocorrelation, is by estimating results with Newey-West regressions (see model 6).
18
relationship supports hypothesis 1, explained in section 2. The estimated coefficients of dividends (DVC) and share repurchases (SRP) are 33.34 and 10.66 $millions respectively. Finding that a one $million increase in DVC (SRP), leads to an increase of 33.34 (10.66) $millions in market value, ceteris paribus. These significantly positive coefficients support hypothesis 2 and 3, as discussed in section 2. The findings are consistent with previous literature and are a plausible explanation for the free cash flow hypothesis by Jensen (1986). The increase in a firm’s payout policy, lowers the
Table 2: Results raw data analysis, using model I and Panel A
(1) (2) (3) (4) (5) (6)
Raw data Trimmed data
(without zero-values) Trimmed data (with zero-values) Pooled OLS model Newey-West model Pooled OLS model Newey-West model Pooled OLS model Newey-West model
VARIABLE MKOF MKOF MKOF MKOF MKOF MKOF
ADR 12,811.868*** 12,811.868** 3,692.642 3,692.642 4,381.275*** 4,381.275** (2,626.113) (5,823.984) (2,499.246) (3,325.479) (1,250.635) (1,904.872) DVC 35.055*** 35.055*** 33.947*** 33.947*** 33.349*** 33.349*** (0.466) (4.840) (0.679) (1.713) (0.569) (1.250) SRP 8.868*** 8.868*** 9.437*** 9.437*** 10.656*** 10.656*** (0.274) (1.450) (0.475) (0.896) (0.367) (0.721) SEOs 14.003*** 14.003*** 24.286*** 24.286*** 23.959*** 23.959*** (0.965) (5.042) (2.478) (4.163) (1.760) (3.248) Constant 8,424.438*** 8,424.438*** 6,518.096*** 6,518.096*** 7,646.631*** 7,646.631*** (1,073.758) (2,360.317) (993.802) (1,242.504) (509.459) (703.100) Observations 6,140 6,140 2,236 2,236 4,980 4,980 R-squared 0.713 0.744 0.612 White's Test 3288 497.4 347.40 Prob > chi2 0.000 0.000 0.000 Wooldridge Test 87.44 80.67 236.96 Prob > chi2 0.000 0.000 0.000
Note: Standard errors in parentheses, significance levels:*** p<0.01, ** p<0.05, * p<0.1. MKOF is the market value of a firm in $millions, defined as the sum of book value of debt plus market value of equity. ADR is the actual debt ratio, calculated as book value of debt divided by market value of the firm, DVC is the amount of dividend payments in $million, SRP is the amount of share repurchases in $millions, SEOs is the amount of seasoned equity offerings in $millions. White’s test is conducted to check for heteroskedasticity with H0 is homoskedasticity, Wooldridge test is
conducted to test for first-order autocorrelation with H0 is no first-order autocorrelation. It reports estimation results of
19
cash position. Therefore, managers are less triggered to invest in negative NPV-projects which results in a positive effect on firm value.
Analyzing the estimated coefficient of seasoned equity offerings, denoted as SEOs, provides us with a positive relationship with market value. We observe a significant coefficient of 23.96 $millions, indicating that a one $million increase in SEOs, increases a firm’s market value with 23.96 $million, ceteris paribus. Although SEOs have a negative influence on share prices, the observed finding is consistent with the idea that the value destruction of the negative price reaction, is only a fraction of the new equity raised, hence firm value increases after SEOs. Moreover, it is consistent with Dittmar and Thakor (2007) who argue that when the level of agreement between managers and investors is high, firms only issue new equity when it is value-enhancing. Therefore, this finding rejects hypothesis 4 that state; seasoned equity offerings have a negative relationship with firm value.
Industry effects
In order to get a thorough understanding of the capital structure decisions of the firms in our sample, graph 3 depicts the capital structure (debt-to-value ratio), separated by industry, over time. The
Graph 3: Average actual debt ratio (%) per industry (2001-2018)
Note: the graph presents the average capital structure of all industries, capital structures are measured with the debt-to-value ratio. Each industry is calculated in the same way. We observe that all industries are converging to a common range, this trend is somewhat interrupted by the financial crisis. The dotted line is the transition period of the crisis.
20
actual debt ratio, denoted as ADR, is on average 34.28% of its market value (with a median of 32.58%). Testing for differences between the capital structures of the different industries (Kruskal- Wallis test), results in a chi2 statistic of 672.81 (p-value 0.000) and suggests that the differences
are statistically significant. A comparison of capital structures across industries reveals that the mean of the energy sector (39.60%), communication sector (41.15%) and real estate sector (40.04%) are much higher than information technology sector (23.89%) and the health care sector (27.95%). These observations are in line with the notion that R&D-intensive firms, such as biotechnology and technology firms, usually have low debt ratios. Whereas mature, low-growth industries, such as real estate and energy, usually have higher debt ratios (Binsbergen et al., 2010). To investigate the observation that the capital structures of all industries converge to a common range, table 3 presents ADR per industry per year. Comparing these averages, supported by the results of the Mann-Whitney test, allows us to analyze the statistical and economic significance. Observing the differences between R&D-intensive industries (HC and IT2) and mature, low-growth industries (EN, IN and RE3), we find significant differences according to the results of the Mann-Whitney test, as presented in table D.1. The average ADR of firms in the energy (information
Table 3: Average actual debt ratio (%) per industry (2001-2018)
Year Industry EN MA IN CD CS HC IT CO RE 2001 49.3 39.2 45.1 33.1 30.7 23.3 22.0 39.7 49.9 2002 53.2 41.8 44.8 34.1 32.0 27.4 27.3 42.0 52.2 2003 48.4 38.1 41.0 28.3 30.8 24.4 19.5 45.2 46.0 2004 40.8 35.7 38.2 26.6 28.8 25.1 18.0 37.9 40.9 2005 36.3 36.7 35.7 26.3 30.3 24.9 18.1 40.3 39.3 2006 35.7 33.6 36.3 27.4 28.8 27.4 19.7 41.2 35.5 2007 32.5 34.4 37.7 31.3 32.1 27.8 22.0 38.7 42.7 2008 46.4 48.2 45.5 42.7 35.5 35.3 32.9 45.3 54.4 2009 37.9 38.4 40.3 34.5 33.9 31.1 24.4 44.3 44.9 2010 34.9 35.5 36.5 30.5 31.3 29.8 23.5 41.2 39.9 2011 38.0 40.5 38.7 32.9 30.8 32.1 27.2 46.1 40.0 2012 38.8 37.4 36.6 30.2 32.5 31.4 27.5 43.4 37.0 2013 34.2 35.5 30.5 26.6 28.6 27.2 22.9 38.8 37.5 2014 36.8 36.4 29.0 26.0 27.8 25.6 24.2 39.3 32.5 2015 41.2 39.7 32.2 29.3 25.9 26.7 25.6 40.2 33.0 2016 34.8 35.9 31.3 31.2 28.2 28.9 25.9 38.6 31.1 2017 33.5 32.9 27.9 29.9 29.0 27.1 24.1 37.5 30.7 2018 39.9 38.3 31.8 35.2 35.3 27.6 25.1 40.6 32.9
Note: the table presents the average actual debt ratio of its market value (%). The ratios are per year per industry where EN is energy, MA is materials, IN is industrials, CD is consumer discretionary, CS is consumer staples, HC is health care, IT is information technology, CO is communication services and RE is real estate.
21
technology) sector is 49.3% (22.0%) in 2001 and 39.9% (25.1%) in 2018. We statistically show, in table D.1, that the dispersion between EN and IT in 2001 is higher than the dispersion in 2018. Moreover, we find the same results by comparing industrials with health care, and information technology with real estate. The average ADR for industrials (health care) is 45.1% (23.3%) in 2001 and 31.8% (27.6%) in 2018. Whereas, the average ADR for information technology (real estate) is 22.0% (49.9%) in 2001 and 25.1% (32.9%) in 2018. Again, table D.1 provides evidence that the differences are statistically significant at the 1% level, which support hypothesis 5. We observe that the clear trend in graph 3 finds statistical support by the results of the Mann-Whitney test. Therefore, these findings might suggest that firms found an optimal capital structure. However, further research should be conducted to see if this observation is true.
Size effects
To investigate if capital structure is dependent on firm size, we divide the sample into four quartiles based on the log of their market value. Each quartile represents a group of firms; smallest 25% (small), 25-50% (small-medium), 50-75% (medium-large) and largest 25% (large). In order to get an understanding how capital structure is distributed separated by size quartiles, graph 4 shows the average ADR over time. The ADR averages 32.3% of its market value (with a median of 31.9%) for the complete sample. Testing for significant differences between the capital structures of the different size categories (Kruskal-Wallis test), results in a chi2 statistic of 91.57 (p-value 0.000)
Graph 4: Average actual debt ratio (%) separated by size (2001-2018)
Note: the graph presents the average capital structure of all size categories, capital structures are measured with the debt-to-value ratio. Each size category is calculated in the same way. The categories are calculated as quartiles of the log value of MKOF. We observe that all categories are converging to a common range, this trend is somewhat interrupted by the financial crisis.
22
and suggests that the observed differences are statistically significant. A comparison of capital structures across different size categories reveals that medium-large firms have a higher average ADR (36.3%) compared to small firms (28.6%). Furthermore, the difference between small-medium firms (32.1%) and large firms (32.2%) is negligible. These observations are consistent with the notion that firm size and leverage are positively related (Titman and Wessels ,1988; Warner, 1977) and support hypothesis 6; capital structure is dependent on firm size.
To analyze the different levels of ADR between the size categories (see table 4), table D.3 presents the statistical results of the Mann-Whitney test. The differences observed in 2001 and 2018 are insignificant. Therefore, we are unable to measure size effects and to analyze the observed trend with this approach. However, we do find significant differences between small and small-medium firms, small and small-medium-large firms, and small and large firms in 2008 and 2018 at the 5% significance level. The average ADR of small (small-medium) firms is 39.2% (46.4%) in 2008 and 24.4% (27.9%) in 2018. Two important aspects of this observation should be highlighted. At first, the dispersion between the small and small-medium firms decreased to 2.5 percent point in
Table 4: Average actual debt ratio (%) per size category (2001-2018) Size category Year SMALL SMALL- MEDIUM MEDIUM- LARGE LARGE 2001 35.1 38.2 43.6 27.9 2002 35.7 40.6 43.9 32.2 2003 32.3 32.8 39.3 28.8 2004 29.1 30.5 35.9 27.7 2005 28.6 31.2 30.9 30.7 2006 28.1 30.0 33.3 30.2 2007 29.9 35.5 34.7 30.0 2008 39.2 46.4 47.1 35.9 2009 32.1 35.4 42.3 33.4 2010 29.5 31.8 35.7 33.1 2011 31.3 34.3 37.8 36.6 2012 28.3 31.3 37.6 36.0 2013 24.5 27.7 29.8 33.5 2014 25.2 25.3 30.2 32.0 2015 21.7 28.3 33.5 33.2 2016 19.6 26.9 31.4 34.5 2017 20.2 23.9 30.8 30.7 2018 24.4 27.9 36.0 32.8
23
2018 (7.2 percent points in 2008). Which supports the trend as observed in graph 4. Secondly, small firms have a lower debt ratio (39.2% and 24.4%) than small-medium firms (46.4% and 27.9%). This is consistent with the notion that leverage and size have a positive relationship (Titman and Wessels, 1988; Warner, 1977). Furthermore, the same results are presented for the difference between small and medium-large firms. The average ADR for small (medium-large) firms is 39.2% (47.1%) in 2008 and 24.4% (36.0%) in 2018. These differences are significant at the 5% level, and support hypothesis 6, as explained in section 2.
Financial crisis
An analysis on the average actual debt ratio is performed to grasp insights in the effects of the financial crisis on capital structure. With models 7 and 8, it is possible to investigate the average ADR for the financial crisis period (2007-2009), for the pre-crisis period (2001-2006) and the post-crisis period (2010-2018). The years 2007, 2008 and 2009 are the transition period of the financial crisis and takes value one in dummy variable (CRISIS) and zero otherwise. The pre-crisis period (2001-2006) takes value one in dummy variable (B_CRISIS) and zero otherwise. While the post-crisis period (2010-2018) takes value one in dummy variable (S_CRISIS) and zero otherwise.
Graph 5 presents the average actual debt ratio of firms each year. The ADR averages 32.7% of a firm’s market value (with a median of 32.3%). Testing for differences between the financial crisis
Graph 5: Average ADR (%) of firms listed on the S&P 500 (2001-2018)
Note: the graph presents the average capital structure of all firms in our sample, capital structures are measured with the debt-to-value ratio. The dotted lines represents the financial crisis period (2007-2009). The average ADR is 32.7% and the median is 32.3%.
24
period and the pre- and post-crisis period, leads to significant results as presented in table D.2. The average ADR in the pre-crisis period is lower (33.0%) compared to the financial crisis period (36.4%). This observation is in line with the fact that firms were able to obtain government assistance, in the form of loan guarantees, in times of crisis. A comparison of capital structures between the financial crisis period and the post-crisis period reveals that the average ADR is lower (31.3%) in the pre-crisis period than in the crisis period (36.4%). These findings support hypothesis 7 which states that the financial crisis increase the firm’s capital structure.
Table 5 contains the results of crisis-model (i) and crisis-model (ii). The objective of running these models is to examine the differences between the crisis period and the non-crisis period (model 7), and the pre-crisis period and the post-crisis period (model 8). Consistent with the results in table 2, significant positive coefficients for ADR, DVC, SRP and SEOs are found in both models.
Table 5: Analysis on the effect of the financial crisis
(7) (8)
Crisis model i
Crisis model ii
VARIABLES MKOF MKOF
ADR 5,381.454*** 5,900.596*** (1,898.609) (2,170.865) DVC 33.137*** 32.646*** (1.238) (1.397) SRP 10.660*** 10.912*** (0.715) (0.817) SEOs 24.287*** 26.521*** (3.234) (3.716) CRISIS -4,302.420*** (584.830) B_CRISIS -2,474.068*** (687.447) Constant 8,063.777*** 8,584.929*** (701.290) (827.646) Observations 4,980 4,148
Note: Standard errors in parentheses, significance levels :*** p<0.01, ** p<0.05, * p<0.1. MKOF is the market value of a firm in $millions, defined as the sum of book value of debt plus market value of equity. ADR is the actual debt ratio, calculated as book value of debt divided by market value of the firm, DVC is the amount of dividend payments in $million, SRP is the amount of share repurchases in $millions, SEOs is the amount of seasoned equity offerings in $millions. White’s test is conducted to check for heteroskedasticity with H0 is homoskedasticity, Wooldridge test is
conducted to test for first-order autocorrelation with H0 is no first-order autocorrelation. It reports estimation results of
25
The estimated intercept of model 7, is statistically significant and suggests that a firm’s market value is on average 8,063 $millions if all explanatory variables are equal to zero. At the same time, the statistically significant coefficient of CRISIS is -4,302, indicating that the average market value for firms in the crisis is 4,302 $millions lower compared to firms in the pre- and post-crisis periods. This finding is in line with the graphical representation of market values presented in graph 2. Investigating the differences between the pre- and post-crisis periods, results in an average market value of 8,584 $millions for firms that do not have debt, dividends, share repurchases and seasoned equity offerings. In addition, the statistically significant negative estimation for B_CRISIS, denotes that firms in the pre-crisis period is 2,474 $millions lower relative to firms in the post-crisis period. This result is, again, consistent with the observations in graph 2.
5.2 Detail analysis
Up to this point, the focus of our analysis has been to examine the relationship between capital structure, payout policy, equity issuance and firm value. Moreover, the focus was on economic observations and to analyze how debt ratios are spread out in selected industries and size categories. In a more detailed analysis, it is possible to investigate the effects of dividends, share repurchases, and equity issuance on a firm’s price-to-book ratio. Using equation (II) and panel B, allows us to examine book value-based debt-to-equity ratio (CS). Moreover, it provides the opportunity to better compare firms in the sample due to the fact that the model uses ratios instead of actual values.
Table 6 contains the estimation results of the Pooled OLS and Newey-West regressions with dividends, share repurchases, and seasoned equity offerings scaled to book value of equity as explanatory variables. In addition, SIZE denotes natural logarithm of book value of equity for firm
i and year t and is a controlled variable in the models. Overall, our results are robust using different
26
12.390 (5.917) suggesting that a one percentage point increase in DVCBEQ (SRPBEQ) increases the price-to-book ratio with 12.39 percentage point (5.92 percentage point), ceteris paribus. Hence, increases in dividends and share repurchases are value-enhancing. Overall, the results from estimating the coefficients DVCBEQ and SRPBEQ support our hypotheses (2 and 3 discussed in section 2) that both dividends and share repurchases have a positive relationship with firm value.
Table 6: Results detail analysis, using model II and Panel B
(9) (10) (11) (12) (13) (14)
Raw data Trimmed data
(without zero-values) Trimmed data (with zero-values) Pooled OLS model Newey-West model Pooled OLS model Newey-West model Pooled OLS model Newey-West model VARIABLES PB-ratio PB-ratio PB-ratio PB-ratio PB-ratio PB-ratio
CS 0.297*** 0.297** 0.228*** 0.228*** 0.409*** 0.409*** (0.009) (0.120) (0.030) (0.078) (0.011) (0.115) DVCBEQ 16.163*** 16.163*** 23.404*** 23.404*** 12.390*** 12.390*** (0.288) (4.001) (0.517) (2.259) (0.532) (2.264) SRPBEQ 3.314*** 3.314** 4.719*** 4.719*** 5.917*** 5.917*** (0.091) (1.467) (0.306) (0.767) (0.202) (0.951) SEOsBEQ -5.962*** -5.962 3.126** 3.126 10.082*** 10.082** (0.677) (9.307) (1.387) (2.846) (0.871) (4.059) SIZE -1.090*** -1.090*** -0.237*** -0.237*** -0.727*** -0.727*** (0.084) (0.206) (0.059) (0.068) (0.058) (0.101) Constant 11.637*** 11.637*** 3.088*** 3.088*** 7.665*** 7.665*** (0.698) (1.933) (0.524) (0.644) (0.482) (0.942) Observations 5,932 5,932 2,160 2,160 4,789 4,789 R-squared 0.806 0.780 0.561 White's Test 4299.3 920.3 835.0 Prob > chi2 0.000 0.000 0.000 Wooldridge Test 29.29 7.96 12.32 Prob > F 0.000 0.005 0.000
Note: Standard errors in parentheses, significance levels:*** p<0.01, ** p<0.05, * p<0.1. PB-ratio stands for the price-to-book ratio for each company each year and is calculated as market value of equity / book value of equity. CS accounts for a company’s capital structure each year and is defined as book value of debt / book value of equity. DVCBEQ is the ratio of dividend payments to book value of equity, SRPBEQ is the ratio of share repurchases to book value of equity, SEOsBEQ is the ratio of seasoned equity offerings to book value of equity. SIZE is used as a control variable and is the natural logarithm of book value of equity. White’s test is conducted to check for heteroskedasticity with H0 is homoscedasticity, Wooldridge test is conducted to test for first-order autocorrelation with H0 is no
27
Accordingly, the significantly positive regression coefficient for seasoned equity offerings (SEOsBEQ) is consistent with the results as discussed in section 5.1. Thus, a one percent point increase in SEOs is associated with an increase in a firm’s price-to-book ratio of 10.08 percentage point, ceteris paribus. Again, this explains the idea that firm’s market value goes up after SEOs, despite the negative stock price reaction. This finding, however, provides sufficient evidence to reject our hypothesis that states; seasoned equity offerings have a negative relationship with firm value.
5.3 Analyses on Dividends, Share Repurchases and Seasoned Equity Offerings
To extent the detail analysis as presented in section 5.2, we perform a dummy variable approach. This approach provides us with the opportunity to analyze whether there are significant differences between dividend paying firms and non-dividend paying firms, between repurchasing firm and non-repurchasing firm, and between equity issuing firms and non-equity issuing firms.
Dividends
28 Share Repurchases
The same analysis is performed for model 16, SRP model, where we investigate whether there are different effects between repurchasing firms and non-repurchasing firms. The dummy variable, Non_SRP, takes value one when firms do not repurchase shares, and zero otherwise. The estimated coefficient non_SRP is highly insignificant, which suggests that the differences between repurchasing firms and non-repurchasing firms is insignificant. The lack of significant differences may be due to the high fluctuations in share repurchases as depicted in graph 2. Another explanation is that shares are usually repurchased as part of a program, a firm may take a year or more to buy
Table 7: Analysis on Dividends, Share Repurchases, and Seasoned Equity Offerings
(15) (16) (17) DVC model SRP model SEOs model
VARIABLES PB-ratio PB-ratio PB-ratio
CS 0.400*** 0.409*** 0.410*** (0.117) (0.115) (0.115) DVCBEQ 15.436*** 12.391*** 12.401*** (2.470) (2.266) (2.260) SRPBEQ 5.703*** 5.922*** 5.917*** (0.942) (1.002) (0.950) SEOsBEQ 9.582** 10.079** 9.988** (4.056) (4.061) (4.081) SIZE -0.507*** -0.726*** -0.724*** (0.093) (0.098) (0.103) non_DVC 1.865*** (0.221) non_SRP 0.013 (0.241) non_SEOs -0.136 (0.245) Constant 5.263*** 7.656*** 7.652*** (0.872) (0.907) (0.949) Observations 4,789 4,789 4,789
29
back the shares. Although the difference is insignificant, we do observe that 22.8% of the firms in our sample are non-repurchasing firms. Table B.4, shows that share repurchasing gained increasing attention in the period 2001-2018. In 2001, 66% of the firms in the sample were active in repurchasing, while 91% of the firms in 2018 repurchased shares. In 2009, the percentage of firms decreased to 65% (was 81% in 2008). This observation contradicts the notion that firms repurchase shares when stock prices are low (market timing), because the S&P 500 hit its lowest level in 2009 (in the period 2001-2018).
Seasoned Equity Offerings
30
6. CONCLUSION
This study has aimed to find new empirical evidence contributing to the corporate finance literature. This paper focusses on the relationships between capital structure, payout policy, equity issuance and firm value. Using a sample of 404 firms included in the S&P 500 index for the period between 2001 and 2018, we document results that provide statistical as well as economic evidence. We present evidence for the relationships between capital structure, payout policy and firm value which are consistent with the existing literature. However, we also present evidence for the relationship between equity issuance and firm value that contradicts the discussed literature.
We use a raw data analysis to provide insights into the relationship between market-based capital structures and firm value. This analysis reveals that firm value goes up in the case when firms increase their debt-to-value ratio. This positive relationship is in line with previous research (Titman and Wessels, 1988; Warner, 1977) and is consistent with our hypothesis 1. The results from our detail analysis provides additional evidence for this hypothesis. These findings highlight that capital structure decisions affect firm value.
Investigating the relationship between dividends, share repurchases and firm value, provides us with significantly positive relationships. The results support our hypotheses 2 and 3, which state that dividends (share repurchases) have a positive relationship with firm value. Although we did not directly test the free cash flow hypothesis, our finding may be plausible evidence for the hypothesis by Jensen (1986). Firms increasing payout policy, lower their cash position and prevent managers to engage in negative NPV-projects. Taking on positive NPV-projects instead, will increase firm value.
In contrast to the existing literature, we find statistical evidence that seasoned equity offerings increase firm value. The significantly positive relationship is found with different regression techniques (pooled OLS and Newey-West) and provides sufficient evidence to reject hypothesis 4. A plausible explanation for the positive relationship is that SEOs increase firm value, but with less than the total amount of equity raised. Since firms tend to raise new equity when share prices are high, negative stock price reactions will follow (Baker and Wurgler, 2002; Choe et al., 1993; Loughran and Ritter, 1995). Therefore, firm value will increase after SEOs, but with less than equity raised because of the negative stock price reactions. Surprisingly, literature so far does not discuss this argumentation. At the same time, it supports the finding by Dittmar and Thakor (2007) who argue that managers only issue new equity when the level of agreement between managers and investors is high.
(low-31
growth, mature industries). These findings are consistent with the results of Binsbergen et al. (2010).
The higher debt-to-value ratios of medium-large and large firms compared to small and small-medium firms, indicate that the largest 50% of firms in our sample hold more debt compared to the smallest 50% of firms. This finding is consistent with the notion that large firms are more diversified and have lower probability of default compared to small firms (Titman and Wessels, 1988). This result supports the hypothesis that states; capital structure is dependent on size (hypothesis 6). Taken together, the size effects and the financial crisis effects, the results explain why firms in the crisis obtained higher debt ratios compared to the pre- and post-crisis period. Accompanied by the drop of stock markets, the ability for firms to obtain debt changed. Financial institutions we less willing to lend their money to firms low on credibility, while the government intervened by offering loan guarantees. The higher debt-to-value ratio in the financial crisis period (2007-2009) relative to the pre- and post-crisis periods (2001-2006 and 2010-2018) indicates that firms obtained different debt levels in the crisis. This supports our hypothesis 7; the crisis increased a firm’s capital structure.
The observed significant difference in price-to-book ratios between non-dividend paying firms and dividend paying firms suggests that non-dividend paying firms on average have a higher price-to-book ratio. An explanation for this finding may be that high growth firms do not pay dividends because they want to use all resources to grow the company. This finds support by the observation that the majority of low-growth firms (energy, materials and consumer staples) pay dividends. The inconclusive analysis on the effects of repurchases and seasoned equity offerings, can be an introduction to further research. Although we did not find significant differences, others may find differences using other, more specified, datasets.
This study has several limitations. Due to data limitations, we were not able to gather data on foreign and domestic revenues, which is an important determinant for the level of cash holdings for US public companies. This limitation is an immediate recommendation for further research. Obviously, other variables could be added to the model to create a more detailed analysis. Relaxing the assumptions made in this study, could also lead to better results.
32
REFERENCES
Alti, A., 2006. How persistent is the impact of market timing on capital structure? Journal of Finance 61(4), 1681-1710.
Bah, R. and Dumontier, P., 2001. R&D intensity and corporate financial policy: some international evidence. Journal of Business Finance & Accounting 28(5/6), 671-692.
Baker, M. and Wurgler, J., 2002. Market timing and capital structure. Journal of Finance 57(1), 1-32.
Berg, G., and Kirschenmann, K., 2010. The impact of the US financial crisis on credit availability for small firms in Central Asia. Retrieved from: https://www.rug.nl/research/globalisation-studies-groningen/research/conferencesandseminars/conferences/eumicrofinconf2011/papers /2a.kirschenmann-berg.pdf.
Berk, J. B. and DeMarzo, P. M., 2017. Corporate Finance. Boston, MA: Prentice Hall.
Bernheim, B. D., and Wantz, A., 1995. A tax-based test of dividend signaling hypothesis. The American Economic Review 85(3). 532-551.
Bessler, W. Drobetz, W., and Grüninger, M. C., 2011. Information asymmetry and financing decisions. International Review of Finance, 11(1), 123-154.
Bessler, W., Drobetz, W. Haller, R and Meier, I., 2013. The international zero-leverage phenomenon. Journal of Corporate Finance 23(), 196-221.
Bessler, W., Drobetz, W., Pensa, P., 2008. Do managers adjust the capital structure to market value changes? Evidence from Europe. Zeitschrift für Betriebswirtschaft 6, 113-145.
Bhattacharya, S., 1979. Imperfect information, dividend policy, and “the bird in the hand” fallacy. Bell Journal of Economics 10(1), 259-270.
Binsbergen van, J. H., Graham, J.R., and Yang, J., 2010. The cost of debt. Journal of Finance 65(6), 2089-2136.
Bradley, M., Jarrell, G. A. and Kim, E. H., 1984. On the existence of an optimal capital structure: Theory and evidence. Journal of Finance 39(3), 857-878.
Brown, J.R., Fazzari, S. M. and Petersen, B.C., 2009. Financing innovation and growth: cash flow, external equity, and the 1990s R&D boom. Journal of Finance 64(1), 151-185.
Choe, H., Masulis, R. W., Nanda, V., 1993. Common stock offerings across the business cycle: Theory and evidence. Journal of Empirical Finance 1(), 3-31.
Dittmar, A. and Thakor, A., 2007. Why do firms issue equity? Journal of Finance 62(1), 1-54. Easterbrook, F., 1984. Two agency-cost explanations of dividends. America Economic Review
74(4), 650-659.
