1 Ciaran Jaras 31st January 2017
10824952 Supervisor: Derya Guler Economics and Business
Finance and Organisation
The relationship between capital structure and the dividend pay-out policy of US manufacturing firms, in times of financial crisis
2 Statement of Originality
This document is written by Ciaran Jaras who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Abstract
The purpose of this research was to investigate the relationship between capital structure and dividend pay-out policy of US manufacturing firms, during the financial crisis of 2007-2009. It built on the existing research of Jensen et al (1992), Fama and French (2001) and DeAngelo (2006), through investigating whether the findings of these studies were present in this paper’s data set. The research concluded that only two variables of the model were significant, ROA and dprt1. Futhermore, the relationship between capital structure and dividend pay-outs was found to be insignificant. It suggests that the variables which were found to be significant in existing research papers, were not found to be so in this paper, for one of many reasons. One possible reason for this is the fact that this study investigated the relationship over the period of the financial crisis. The crisis may have impacted the
dividend pay-out decision of the firms’ management, and as a result, the relationship between capital structure and dividend pay-outs does not hold.
4 Contents 1. Introduction……… 5 2. Theoretical Framework……….. 7 2.1. Financial Crisis 2007-2009……….. 7 2.2. Dividend Policy……….. 9
2.2.1. Bird in the hand……….. 10
2.2.2. Clientele effect………. 10
2.2.3. Tax effect………. 11
2.2.4. Signalling effect……… 11
2.2.5. Agency costs……….. 12
2.2.6. Dividends and a firm’s growth prospect………. 13
2.3. Relationship of dividend policy and financial crises……… 14
2.4. Relationship of capital structure and dividend policy ……….. 14
2.5. Relationship of capital structure and financial crises………. 15
3. Hypotheses……… 15
4. Methodology……… 16
4.1. Model……….. 17
4.2. Regression using Stata………. 19
4.3. Data……….. 19
5. Results……… 22
6. Conclusion……….. 24
5 1. Introduction
This paper seeks to investigate the relationship between the capital structure of a firm and its dividend pay-out policy, during times of financial crisis. Specifically, this paper will focus on the dividend policies of US firms operating in the manufacturing sector, and how these policies were influenced by the firms’ debt-to-equity ratios, during the Great Depression. The economic impact of the Great Depression was wide reaching. With the majority of the world’s largest economies being affected, it was a truly global crisis. For many reasons, including the extent to which its banking sector was implicated in the crisis, the impact on the US economy was catastrophic. Productivity fell by 6%, while the unemployment rate doubled to 10.1%. One sector of the US economy which suffered significant hardship, was the manufacturing sector. Many of the goods produced in the industry, for instance mobile phones and tablets, soon became luxuries that many couldn’t afford. A significant fall in sales led to fewer funds being available to the firms. Furthermore, the automotive industry, which contributes a significant portion of the US manufacturing output, saw a decrease of 60% of car assembly in the years 2008-2009.
During this period of a reduction in sales, it is probable that such a company would see its liquidity diminish. Generally, a firm would keep a liquidity buffer large enough to satisfy its existing obligations in the event of unforeseen stress. One might, therefore, argue that such a firm could reduce its expenditure through pay-outs to shareholders, in order to retain capital for these commitments. However, as discussed in papers by Guttman et al (2010) and Litzenberger and Ramaswamy (1982), a reduction or suspension of dividends can act as a negative signal to the market regarding the firm’s financial prospects, and can therefore have a negative impact on the firm’s share price. This results in the so called “dividend stickiness”, where firm management seek to maintain dividend pay-outs at all costs (Guttman et al, 2010).
In this paper, we further consider the influence of a firm’s capital structure on its decision to pay dividends. During a crisis, a firm with a high debt-to-equity ratio would have a greater uncertainty surrounding its future interest repayments. It may, therefore, decide to reduce
6 its dividend pay-outs in order to retain capital to satisfy these repayments (DeAngelo et al, 2006).
This research seeks to contribute to the works of existing literature, by approaching the topic from a new perspective. Whilst extensive research has been carried out into the relationship between capital structure and dividend policy, as well as research into the influence of the financial crisis on dividend policy, this research follows a unique research question. As of yet, there has been no research carried out on the relationship between capital structure and dividend policy of US manufacturing firms during the crisis. Therefore, this paper will seek to answer the following research question: Did the capital structure of US manufacturing firms impact the firms’ dividend pay-out policy during the financial crisis?
The research relies heavily on the paper of DeAngelo et al (2006). In this paper, DeAngelo et al (2006) highlight many of the company financials which contribute to a firm’s dividend policy decision, some of which can be further traced to Fama and French (2001). As a result, a selection of these financial variables are included in the model of this paper, alongside some other relevant variables which can be attributed to Jensen et al (1992) and Guttman et al (2010).
In order to maximise the extent to which the effect of the crisis on the policy decision is captured, this paper will analyse data from the period 2006-2011. The start date ensures that pre crisis dividend levels are represented in the regression, whilst the extended end date ensures that the effect is fully captured. Specifically, a paper by Campello et al (2009) found that the majority of CEOs interviewed, claimed that they were faced with high borrowing costs at the end of the crisis. This implies that firm management may want to retain capital for internal use, therefore the effect of the dividend decision may not be seen until the years following the crisis. As a result, we consider the years 2006-2011.
The content of this paper has been structured in the following manner. Section 2 will explore the wealth of existing literature on the topics of capital structure, dividend policy and the financial crisis. It will explain the degree to which each paper has been drawn on in the undertaking of this research, and how this research adds value to the findings of this existing literature. This section will culminate in the hypotheses of this paper. Section 3 will
7 provide an explanation into the data and methodology used in this paper. Section 4 will present the results of this research, including tables and figures. Section 5 will introduce a discussion on this paper, with the inclusion of some concluding remarks.
2. Theoretical Framework
2.1. Financial crisis 2007-2009
The Great Recession came about as a result of the collapse of the US housing market. A decade of low interest rates and lending to subprime borrowers, culminated in the sharp decline of housing prices during 2006-2007. Indeed, 30th December 2008 saw Standard and Poor’s Case-Shiller house price index record its largest ever decline. Essentially, the
aggressive mortgage lending to those that were incapable of making a full repayment, led to a steep rise in foreclosures. The increased supply of houses which were for sale (as lenders tried to recoup their money) saw houses prices begin to fall. As a result, many homeowners which were able to make full repayments (prime borrowers) chose not to, as they saw the value of their house fall. This led to an increase in the number of forsaken properties, adding to the problem of excess supply (Reinhart and Rogoff, 2008).
But how does the collapse of the US housing market lead to a global financial crisis? Acharya and Richardson (2009) seek to answer this question in their paper “Causes of the financial crisis”.
In the wake of the Dot-com bubble, the US Federal Reserve set interest rates at close to 1%. Not only did this make mortgages more accessible to subprime borrowers, but it also
encouraged investors to seek alternative investment strategies, which would a higher return (Maddaloni and Peydró, 2011). This led to the indirect investment into mortgages, as banks switched from an “originate to hold” to “originate to distribute” strategy. The “originate to hold” strategy is the traditional process behind lending. For instance, a borrower would approach a bank and request a loan (mortgage) in order to purchase a house. The bank would assess the worthiness of the potential borrower, and would agree to lend to them only if they believed that they were able to repay the loan. In the event that the borrower
8 defaulted on the repayments, the bank would typically repossess the house, and use its sale to recoup its money. Despite the repossession, the bank would suffer financially as a result of the default, and therefore has an apparent incentive to only lend to credible borrowers.
As previously mentioned, as a result of the low interest rates attached to treasury bonds, investors began to seek alternative investment opportunities, namely mortgages and other asset backed securities. The process of investment is as follows. Banks pool together long term, illiquid assets, typically mortgages and credit card loans. Such illiquid assets are not attractive for the bank to hold, as they cannot be used to make further loans. These assets are therefore sold to an investment bank for a lump sum, enabling the commercial bank to lend further. The investment bank, which is now the sole claimant for all interest payments, transfers these rights to a special purpose entity. The special purpose entity is a corporation, therefore its shares can be sold to new investors. The investment bank splits this special purpose entity into different tranches – senior, mezzanine and equity (Brunnermeier, 2009). The tranches are ordered by the priority with which the investors will have a claim on the assets. Investors of the senior tranche will have a priority claim on the repayments, in the event that some borrowers default on their repayments. As such, the senior tranche entails the least amount of risk, therefore has a lower yield. On the contrary, investors of the equity tranche will only have a claim on the residual repayments, after the senior and mezzanine investors have been fully compensated. As such, this tranche has the highest yield. At this stage, investors of the asset backed securities are free to choose which tranche they would prefer to invest in. Typically, more strictly regulated pension funds and mutual funds opt for the senior tranche investments, whilst riskier hedge funds opt for mezzanine or equity tranche investments (Goodhart, 2008).
As many borrowers began to default on their repayments, there were fewer funds available to repay investors of the asset backed securities. As such, many senior tranches lost their AAA rating, and the demand from risk averse investors fell. The degree to which the
commercial and investment banks were reliant on the proceeds of the loan repayments can be seen by the immediate collapse of banks such as Lehman Brothers and Bear Stearns (Ivashina and Scharfstien, 2010).
9 Acharya and Richardson (2009) conclude that the US housing bubble lead to a global
financial crisis, due to two primary reasons. The removal of the asset backed securities from the banks’ financial statements, meant that they were able to manipulate the Basel capital requirements, and as a result had limited protection against the collapse of the housing market. Secondly, as banks could reduce their capital requirement if the assets on their balance sheets were AAA rated, they were able to do so by packaging these assets into tranches, some of which were AAA rated.
It can be seen that the policy of “originate to distribute” lending absolves the lender of any responsibility to ensure the credibility of the borrower, as the burden of any default is borne by the future investors (Purnanandam, 2011).
2.2. Dividend policy
The choice of a company’s dividend policy remains a focal point of corporate finance. It’s interaction with a company’s other financial decisions, as well its use to convey valuable financial information to the market, highlights the pertinent nature of dividend initiation and control. As such, a firm’s decision to pay dividends is a topic that has been researched in great detail.
Modigliani and Miller (1961) paved the way for the debate over dividend pay-out policy. Their paper concluded that under perfect market conditions, a company’s decision to pay dividends or not, or to reduce or suspend dividends, was irrelevant. This is due to the fact that dividend pay-outs and share prices are essentially two sides of the same coin. A high dividend pay-out would provide a cash return for the owner of the share, however the share price would duly fall to accommodate this pay-out, making the individual share owner no better off (Modigliani and Miller, 1961).
In reality, the markets in which firms operate in are very different to the perfect capital market concept, as proposed by Modigliani and Miller (1961). As a result, numerous
theories, some contradictory, have been published on dividend pay-out policy. This section will summarise the key developments in dividend theory since that of Modigliani and Miller (1961).
10 2.2.1. Bird in the hand theory
This follows from the adage that “a bird in the hand is worth two in the bush”. In other words, an investor would prefer a share that paid dividends (guaranteed return) over one that didn’t, and as a result would command a higher price. Indeed, Gordon (1963)
concluded that dividend payments have a greater positive impact on share prices than retained earnings, suggesting that investors are risk averse.
2.2.2. Clientele effect
This theory originated from the relaxing of the perfect capital market assumption of Modigliani and Miller (1961). Essentially, it suggests that the combined preference of a group of investors concerning a number of different shares, can lead to a change in the price of the shares. It is argued by Modigliani and Miller (1961) that these investors will make their investment choice, such that the costs of market imperfections are minimised. Indeed, this theory was built upon by Elton and Grubber (1970) who concluded a situation of self-sorting amongst investors. Investors which belonged to a high tax bracket had a preference for low dividend yields, whilst high dividend yields were preferred by those within a low tax bracket. Furthermore, a similar result can be found in the paper of Richardson Pettit (1977), which concluded a significant relationship between both tax preference and time
preference, and the dividend variance across an investors’ portfolio.
The apparent existence of the clientele effect, suggests that investors choose which firms to invest in based on their individual preference. For instance, a firm that pays high dividends would appeal more to an investor that has a preference for dividend pay-outs over capital gains. Furthermore, Allen et al (2000) conclude that investors which have a relative tax advantage –for instance an institutional investor- has a preference for dividend paying shares.
Richardson Pettit (1977) develops the aspect of clientele theory regarding transaction costs and homemade dividends. The research finds that some investors have a preference for a constant, income-like dividend payments. In other words, if they were reliant on the
11 proceeds of their shares as an income, they would find it costly to repetitively sell a number of (non-dividend paying) shares, in order to replicate an income. Alternatively, they would have a preference for shares which paid out regular dividends. On the contrary, the opposite can be said for investors which do not necessarily rely on the proceeds of their shares
2.2.3. Tax effect
The tax effect stems from the fact that dividends and capital gains are often taxed at different levels. For instance, if dividends were to be taxed at a higher rate than capital gains, it is arguable that investors would prefer to invest in non-dividend paying firms. Indeed, evidence taken from the US would seem to confirm this. Prior to the tax reform act of 1986, dividends were taxed at 50%, while capital gains at 20%. Following the reform, both were taxed equally at 28%. This saw a convergence in the valuation of dividend and non-dividend paying shares. However, it should be noted that even after this reform, there existed a (smaller) preference for non-paying dividend shares.
This is in line with the findings of Litzenberger and Ramaswamy (1982), which concluded that such a preference existed, not only due to the rate at which dividends are taxed, but also the fact that the timing of dividend payments are at the discretion of the firm, whereas capital gains can be deferred by the investor, meaning that such capital gains have an inherent tax advantage. As such, investors in dividend paying firms will demand a pre-tax premium to compensate for the associated tax disadvantage.
2.2.4. Signalling effect
This effect originates from the perceived asymmetric information between a firm’s
management and the investor. The signalling theory suggests that dividend payments act as a signalling device to the market of a firm’s financial prospects. In essence, the market assumes that a firm’s management is better informed about the firm’s finances than the individual investor. Therefore any decision of a firm’s management regarding the dividend
12 pay-out policy of a firm can be linked to the firm’s ability to pay these dividends, and as a result have an effect on a firm’s share price.
Koch and Shenoy (1999) further this idea in their research. Their paper suggests that share prices of a firm will rise following the announcement of a dividend initiation or a dividend increase, and will fall following a dividend decrease or a dividend suspension.
It should be noted that for a dividend to act as an efficient signal, it should be sufficiently costly for a firm to send out false signals to the market. For instance, a firm that is
performing poorly should not find it profitable to mimic a strongly performing firm by increasing its dividends. Rather, the cost of foregone investment opportunities through paying out excess capital, should outweigh the benefits of any share price increase.
It follows that an increase in dividends will signal a sound financial future, while a decrease or suspension of dividends will signal the opposite. This leads to the concept of dividend stickiness proposed by Guttman et al (2010). Their research suggests that a firm’s management is highly reluctant to decrease dividend pay-outs, even when it is costly to maintain them, in order to avoid the negative impact on the firm’s share price.
Alternatively, Lintner (1956) finds that a firm’s management is equally reluctant to increase its dividend pay-outs until it is sure that the firm’s future earnings can maintain them. This led to the so-called process of “dividend smoothing” with which a firm’s management can control its dividend pay-out policy, whilst sending valuable signals to the market.
Whilst the impact of a change in dividend policy on share prices is two sided, research has found that the negative impact of a dividend decrease tends to outweigh the positive reaction to an identical increase (Michaely et al, 1995).
2.2.5. Agency Costs
Agency costs are borne as a result of a misalignment of the incentives of a firm’s management, and the interest of a firm’s shareholders (Jensen et al, 1986). Such a misalignment can lead to the common agency problems of overinvestment, excessive risk taking and empire building. Excessive risk taking, for instance, may see a manager invest in a
13 risky project (over a safer alternative), which they hope will yield a higher firm equity, thus entitling them to some reward, at the expense of the expected firm value. Furthermore, managers may be inclined to “empire build”. This would see them stockpiling a firm’s excess reserves (which could alternatively be used for investment purposes) in order to increase the value of the firm, thus entitling them to a larger compensation.
Rozeff (1982) concludes that dividends act as a method of alleviating some agency costs. By paying out excess reserves to shareholders, managers are no longer able to engage in risky investments, nor to engage in empire building. Additionally, by paying out excess reserves, a firm’s management would be forced to turn to external sources of funding, should it wish to undertake an investment. As such, the decisions of the firm’s management would now be scrutinised by the lender, meaning that any overly risk or value destroying investment, would now not be undertaken (Easterbrook, 1984).
2.2.6. Dividends and a firm’s growth prospect
A study by Fama and French (2001) conclude the existence of a propensity for smaller firms with a greater prospect of growth to pay lower (if at all) dividends, relative to larger, more mature firms. It follows that such firms with greater growth prospects, will also have a great number of investment opportunities. As a result, these firms have a tendency to retain any excess cash, in order to undertake such investments.
This can be also be seen throughout different sectors of the economy. Sectors which inherently consist of high growth firms (i.e. information technology sector) pay the lowest average dividends. Rather, the firms in this sector prioritise funding of research and
development activities, which are more relevant to their industry, compared to the highest dividend paying sectors, such as utilities, materials and industrials.
Denis and Osobov (2008) build on the paper of Fama and French (2001). In addition to concluding a negative relationship between growth opportunities and dividend pay-outs, they also find that firms which have a large retained earnings are more likely to pay dividends.
14 2.3. Relationship of dividend policy and financial crises
The relevance of this topic follows from the idea that dividends can act as a signalling device. In times of crisis, a firm’s management may look to reduce expenditure, in order to ensure that the firm is able to satisfy its existing obligations. One method of reducing expenditure would be retain excess earnings within the company that would otherwise be paid out as a dividend. As previously discussed in this paper, a reduction or suspension in dividends can imply financial uncertainty of the firm, thus having a negative effect on its share price. This issue is central to a number of papers which research the impact on dividend policies of a financial crisis.
Rozeff (1982) concludes a negative relationship between the risk of a firm and the firm’s dividend pay-out. In essence, a firm that is facing a high risk (i.e. in times of financial distress) is more likely to reduce its dividend pay-out.
In addition, DeAngelo and DeAngelo (1990) found that US companies in a period of financial distress (1980-1985) tended to reduce their dividend pay-out. They did, however, find that the number of firms which suspended their dividend pay-out was lower than expected, suggesting that management were more unwilling to cease paying out dividends, rather than reducing them. Furthermore, they find that firms which had a long history of paying dividends were less likely to reduce pay-outs during this period. This can attributed to the signalling effect, which suggests that the market will react unfavourable to a sudden reduction of dividend pay-outs.
2.4. Relationship of capital structure and dividend policy
This topic follows from the principle that the management of a levered firm will feel inclined to retain earnings, in order provide a buffer against uncertainty of interest repayments. DeAngelo et al (2006) concluded that firms which were relatively highly levered were less likely to pay dividends, rather choosing to retain funds within the firm.
15 2.5. Relationship of capital structure and financial crises
Research into this topic seeks to establish a pattern of firms’ leverage during financial crises. During the early 2000s, low interests encouraged investment banks to significantly increase their leverage (Kalemli-Ozcan and Sorenson, 2011).
Ivashina and Scharfstein (2008) find that during the peak period of the crisis, loans to large borrowers fell by 37%. They concluded that banks with access to fewer deposits were more likely to decrease lending, whilst banks which had more exposure to revolving credit lines were also more likely to decrease lending. The cost of borrowing became too expensive for investment banks, and many were unable to service existing debt (Ivashina and Scharfstein, 2008).
Bancel et al (2004) find that firms which rely heavily on external financing, see a reduction in borrowing opportunities during a crisis, therefore reducing the ability of the firm’s
management to pay dividends.
3. Hypotheses
A review of the available existing literature on the topics of dividend policy, capital structure and financial crises, provide an indication of the outcome of this research paper. This paper is based on the following hypotheses:
H0: The dividend pay-out policy of US manufacturing firms wasn’t influenced by the firms’ capital structure, during the financial crisis
H1: The dividend pay-out policy of US manufacturing firms was influenced by the firms’ capital structure, during the financial crisis
16 These hypotheses can also be represented as below;
H0: 𝛽1 = 0 H1: 𝛽1 ≠ 0
Where 𝛽1 is the coefficient of the main explanatory variable, ( 𝐷
𝐸), the debt-to-equity ratio.
These hypotheses enable this paper to draw a conclusion on the behaviour of a firm’s management during times of crises. The research paper by Bancel and Mittoo (2006) would indeed suggest that a firm with a greater leverage would be likely to reduce any existing dividend pay-outs. However, this paper seeks to investigate the same relationship, using a different data set. Many manufacturing firms in the US have a long history of paying dividends. As concluded by DeAngelo and DeAngelo (1990), management of a firm which has such a reputation are reluctant to alter their pay-out policy, even in times of financial distress.
For a firm’s management, the most significant trade-off in this case is between retaining sufficient capital to satisfy existing obligations, and minimising the negative effects entailed by reducing dividends. Due to the magnitude of this particular crisis, it is probable that the immediate concern of a firm’s management would be to ensure the survival of the firm, rather than maintaining a constant dividend pay-out. As such, it is predicted that this paper will conclude with the rejection of the null hypothesis, thus implying a negative relationship between the debt-to-equity ratio of a firm and its dividend pay-out.
4. Methodology
This section will firstly introduce the model which underpins this research paper, before providing an insight into the data collection process.
17 4.1. Model
The model in this paper takes the form of a panel data regression. Using a panel data regression makes it possible to consider a number of observations, over a predetermined time span, of each individual within the sample (Hsiao, 2003). This information is especially relevant in this study, as we seek to find the relationship between a number of company financial variables and the respective company’s dividend pay-out policy, over the period of the financial crisis. Therefore, a panel data regression proves to be most suitable form, more so that, for example, a time-series regression. Panel data typically consist of a relatively small value of t (the number of points in time from which data is taken), and a large value for N (the number of observations). In the case of this research, t represents the number of years considered in the data, six, and N is the total number of observations for each
company over each year, 828.
The dependent variable of the model is dividend pay-out ratio (DPR), a main explanatory variable (debt-to-equity ratio – (D/E)) and a six additional explanatory variables, all of which have been taken from the models of the most relevant existing literature, which have been discussed in the literature review. The model is as follows:
𝐷𝑃𝑅 = 𝛽0+ 𝛽1( 𝐷 𝐸) + 𝛽2(𝑅𝑂𝐴) + 𝛽3( 𝑅𝐸 𝑇𝐴) + 𝛽4( 𝐶𝑎𝑠ℎ 𝑇𝐴 ) + 𝛽5(𝐷𝑃𝑅(𝑡 − 1)) + 𝛽6(𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠) + 𝛽7(𝑅&𝐷) + 𝜀 Where
DPR = Annual dividend per share/earnings per share D/E = Debt to Equity ratio
ROA = Return (net income) as a portion of total RE/TA = Retained earnings to total assets ratio Cash/TA = Cash to total assets ratio
DPR(t-1) = DPR of the previous year
Employees = The number of employees (size)
18 The dependent variable in this model, which used to quantify “dividend pay-out policy” is DPR – the dividend pay-out ratio of a firm. This is the most commonly used measure in existing literature, including that of Jensen et al. (1992).
The main explanatory variable of this model is the debt to equity ratio of the firm. A study by DeAngelo et al. (2006) has provided much of the foundation for this paper’s regression model. This study suggested that firms with higher leverage were less likely to pay dividends due to the uncertainty of their interest payments. Again, this uncertainty is likely to increase in times of crisis, therefore it has been included in the model under (D/E).
The model contains six additional explanatory variables, all of which have been found to be significant in existing literature. Firms with a high a degree of self-financing (RE/TA) and a high profitability (ROA), are more inclined to pay (larger) dividends (DeAngelo et al, 2006). The degree of self-financing becomes especially relevant in times of financial crisis, as those that rely on external financing see a reduction in lending, and therefore are less able to pay dividends (Bancel and Mittoo, 2011). Finally, the study by DeAngelo et al. (2006) found a link between the “cash richness” of a firm, and higher dividend pay-outs. As a result, I have included the study’s variable (Cash/TA) in the model.
Following on from the work of Fama and French (2001), it is expected that the dividend pay-out policy is dependent on the growth opportunities available to the firm. The maturity of a firm can be measured by its size. A firm with a large number of employees (proxy for size) is likely to be a mature firm, therefore having fewer growth opportunities. Size is included in the model under (Employees).
A study by Guttman et al. (2010), concluded a stickiness of dividend levels – due to the negative impact on share prices of reducing/suspending pay-outs. The variable (DPR(t-1)) has, therefore, been included in the model. This represents a firm’s dividend pay-out ratio (if any) of the previous year.
Finally, a study by Jensen et al. (1992) introduced a new variable. A firm’s investment and more specifically, investment into research & development (R&D).
19 4.2. Regression using Stata
The model was tested using Stata, a statistical software. A panel data analysis was carried out on the data of all 138 firms. The following Stata output made it possible to test the individual significance of the seven control variables. This was done by carrying out t-tests on the coefficients of each variables, and verifying if any are significantly different from zero. The t-tests were carried out with significance levels of 1%, 5% and 10%. These values are commonly used in existing literature, and therefore provide a suitable range of
significance levels to test the coefficients of the model. Below is the formula for the t-test (Stock and Watson, 2007). The variable X represents the coefficient of each variable, whilst 𝜇 is equal to zero.
𝑡 =𝑋 − 𝜇 ( 𝑆
√𝑁)
As the alternative hypothesis of this paper predicts that the coefficient of the debt-to-equity ratio (𝛽1) will not be equal to zero, a two sided t-test will be carried out in order to capture movement in either direction. Furthermore, despite not being included in the hypotheses, the coefficients of all explanatory variables will be tested in the same manner.
4.3. Data
The data in this research paper was collected from the Compustat North American
database, via the Wharton Research Data Services. The entire database was searched, with the condition that the value of the NAICS (North American Industry Classification System) was between 31 and 33, therefore only considering firms from the manufacturing industry. All relevant financial data (as described by the model) was collected through the database. Data was collected over the period 2006-2011. This ensured that the full influence of the financial crisis on the dividend pay-out decision was captured. It is apparent that any
changes in dividend policy as a result of the crisis were not realised until the years following the crisis, hence the extended period of the data set.
20 This paper makes use of annual data. Annual data was preferred, primarily due to its
availability. Whilst more frequent observations, such as quarterly data, may prove more informative for such a timeframe used in this paper, it poses the problem of missing
observations. For some variables (in particular R&D expenditure) only a small fraction of the firms used in this research had data available on a quarterly basis. Using such data would have made it necessary to remove a large proportion of firms from the sample, or to remove the variables from the model for which data was missing. In order to retain a sufficiently large and representative sample, it was decided that this paper would make use of the available annual data. This method is in line with that of Denis and Osobov (2008). To ensure that the data remained relevant to the research question, only firms that paid dividends in at least one of the years 2006-2011 were considered in the data set. This paper intends to research the relationship between capital structure and the change in dividend pay outs, therefore the exclusion of firms that choose never to pay out dividends
throughout the timeframe of this research is warranted. Furthermore, firms belonging to the US manufacturing sector have a long history of dividend pay-outs, with a significant majority of firms paying out some level of dividend. For instance, as of 2015, 79% of US manufacturing firms paid out dividends, and as such, a sample of dividend paying firms would be representative of the entire sector.
After the data was collected, it was cleaned and made suitable for testing. The dataset contained many missing data points. Some of these were in the form of unreported
research and development expenses across firms, whilst another common problem was the omission of one or several years’ worth of data for some firms. Furthermore, as in line with the previous research of Fama and French (2001) and Denis and Osobov (2008), this paper remove firms which report a negative equity. After accounting for these problems and cleaning the dataset of all missing data, the final dataset consisted of 138 observations (firms), each consisting of annual data for the period 2006-2011.
Descriptive statistics
Table 3 provides a descriptive overview of the variables which appear in the model. The variables ID and year are used solely in the construction of the panel data regression, and do
21 not appear individually in the regression model. There exists 828 observations for each variable, consisting of 138 annual observations over a 6 year period. It should be noted that it is possible to observe negative values for the dependent variable, DPR, in the case that dividends were paid during a year of negative net income.
Table 3
Correlation matrix
Table 4 provides on overview of the correlations of the variables within the model. From the table there are no obvious cause of concern regarding multicollinearity. The two variables with the highest correlation are Employees and rD. These have a correlation of 0.5489. Whilst this is significantly larger than the correlations of other variables of the model, it is smaller than 0.8, which is the rule of thumb value when considering possible
multicollinearity. Interestingly, there is no obvious cause for the relatively high correlation between these two variables, rather, it would be expected that they would be relatively unrelated.
Variable Std. Dev. Min Max
DPR 2,096,191 -40 15 DE 4,303,209 -7,620,867 2,920,482 ROA .1017613 -1 1,625,642 RETA .7004302 -9,157,294 1,278,521 CashTA .0801282 0 .4858137 Employees 3,780,376 1 297 rD 1,807,033 0 10,991 dprt1 2,094,806 -40 15
22 DPR DE ROA RETA CashTA Employees rD dprt1
DPR 1 DE -0.0113 1 ROA 0.0827 0.0017 1 RETA 0.0313 0.0155 0.0743 1 CashTA 0.0002 -0.0701 0.2235 -0.0881 1 Employees 0.0854 0.0451 0.1187 0.1455 -0.1947 1 rD 0.0653 0.0229 0.1091 0.0777 -0.0131 0.5489 1 dprt1 0.0756 -0.0073 0.0827 0.0464 0.0180 0.1063 0.0659 1 Table 4 5. Results
The results of the regression are described below. Table 1 provides a summary of the regression model, whilst Table 2 provides the output of the regression, including the data of the following statistical t-test on the values of the coefficients.
Table 1 R-squared
The coefficient of determination, or R-squared, as an indicator of the goodness of fit of the regression model. It signifies the proportion of the variance of the dependent variable that can be attributed to the variance of the independent variable(s). In other words, it can give a rough indication on the suitability of the model (Stock and Watson, 2007). Zero is the
Random-effects GLS regression Group variable Number of observations Number of groups Observations per group R-squared (Within, between, overall) Wald chi2(7) Prob > chi2 ID 828 138 6 0.0012 14.32 0.0458 0.2260 0.0172
23 lowest possible value for R-squared, whilst 1 is the highest .The formula for determining the R-squared value is shown below.
𝑅2 = 1 − 𝑆𝑆𝑅 𝑆𝑆𝑇
Where SSR is the sum of squares of the residuals, and SST is the total sum of squares. Table 1 provides information on the regression model. It shows an exceptionally low value of R-squared, suggesting that model only explains a small portion of the variance of DPR due to the explanatory variable. This could be due to the infrequency of the data points. Using quarterly data instead of annual data may yield a stronger model, however, as discussed earlier in the paper, very limited quarterly data made this unfeasible. In the same manner that a large R-squared value doesn’t prove causality, the low value of this model is very possibly due to a suboptimal data set, rather than a lack of causality between the dependent variable and the explanatory variables.
DPR Coefficient Standard Error
t-statistic p-value Confidence Interval 95% (Low/High) DE -.01 .02 -0.42 0.68 -.04 .03 ROA 1.41 .75 1.89 0.06* -.05 2.88 RETA .04 .11 0.37 0.71 -.17 .25 CashTA -.13 .96 -0.13 0.90 -2.02 1.77 Employees 0 0 1.30 0.19 0 .01 rD 0 0 0.53 0.60 0 .01 dptr1 .06 .04 1.77 0.08* -.01 .13 _cons .03 .13 0.26 0.80 -.23 .30 Table 2
24 It can be seen in Table 2 that only two variables proved to be significant at any level within the regression. A company’s return on assets (ROA) and its dividend payment from the previous year (dptr1) are both significantly positive at a significance level of 10%, suggesting that both have a positive effect on the dividend pay-out ratio of firms. These findings are in line with those of previous studies. DeAngelo et al (2006) concluded a positive relationship between return on assets and dividend pay outs, whilst the concept of dividend stickiness, as proposed by Guttman et al (2010) is supported by the significantly positive relationship between dividend pay-put ratio at time (t) and time (t-1).
The main explanatory, debt-to-equity ratio, has, as argued by DeAngelo et al (2006), a negative coefficient. The coefficient is, however, insignificant at all tested values of significance. Furthermore, the remaining four explanatory variables - RETA, CashTA, Employees and rD – all proved to be insignificant.
6. Conclusion
This paper set out to investigate the influence of capital structure on the dividend pay-out policy of US manufacturing firms during the financial crisis. The resulting regression model of this research finds that the debt-to-equity ratio does not have a significant influence on the dividend pay-out ratio. This finding is not in line with that of previous research. More specifically, the study by DeAngelo et al (2006), which provided a strong foundation for this paper, concluded that a larger degree of leverage would constitute a lower dividend pay-out.
Whilst the main explanatory variable did not prove to be significant, the model did indicate two explanatory variables which would appear to influence the dividend pay-out ratio, both of which are significant at the 10% level. These variables of interest are ROA, a company’s return on assets, and dprt1, the dividend pay-out ratio of the previous year.
The first significant variable, ROA, is used as a proxy for profitability. Its significance in this model is in line with that of the paper of DeAngelo et al (2006). It follows from the logical argument that a more profitable firm is more able, and therefore more likely, to pay-out
25 larger dividends. This is explained by the significantly positive coefficient of ROA in the regression.
The second significant variable, dprt1, represents the dividend pay-out ratio for the year prior for each observation. Its positive significance adds weight to the principle of dividend stickiness, first proposed by Guttman et al (2010). It suggests that a firm’s management are hesitant to reduce or suspend existing dividend payments, due to the negative impact it may have on the firm’s share price, as per the signalling theory.
The model’s remaining explanatory variables all proved to be insignificant. This contradicts the findings of existing literature, with all explanatory variables having proved significant in one or more of the aforementioned research papers. One possible cause of this is the quality of the sample used in the research. Due to limited available data, this research considered only annual data. By using more frequent time intervals, such as quarterly data, or by increasing the size of the sample, it is probable that standard errors would decrease and the magnitude of the test statistics would increase, increasing the variables’
significance.
It follows, that this provides an opportunity for further research. This paper was limited by the relatively small sample size; resulting in high standard errors few variables proving to be significant. In the future, should a more complete data set become available on this topic, an opportunity would arise to carry out a similar research, and the results of the two can be compared.
26 7. Bibliography
Acharya, V., Philippon, T., Richardson, M., & Roubini, N. (2009). The Financial Crisis of 2007-2009: Causes and Remedies. Financial Markets, Institutions & Instruments,18(2), 89-137. Almeida, H., Campello, M., Laranjeira, B. A., & Weisbenner, S. J. (n.d.). Corporate Debt Maturity and the Real Effects of the 2007 Credit Crisis. SSRN Electronic Journal.
Brunnermeier, M. (2008). Deciphering the Liquidity and Credit Crunch 2007-08.
Deangelo, H., Deangelo, L., & Stulz, R. (2006). Dividend policy and the earned/contributed capital mix: A test of the life-cycle theory. Journal of Financial Economics, 81(2), 227-254. Denis, D., & Osobov, I. (2008). Why do firms pay dividends? International evidence on the determinants of dividend policy. Journal of Financial Economics,89(1), 62-82.
Easterbrook, F. (1984). Two Agency-Cost Explanations of Dividends. The American Economic
Review,74(4), 650-659.
Elton, E. J., & Gruber, M. J. (1970). Marginal Stockholder Tax Rates and the Clientele Effect. The Review of Economics and Statistics,52(1), 68.
Fama, E. F., & French, K. R. (2001). Disappearing dividends: Changing firm characteristics or lower propensity to pay? Journal of Financial Economics, 60(1), 3-43.
Goodhart, C. (2008). The background to the 2007 financial crisis. International Economics
and Economic Policy,4(4), 331-346.
Gordon, M. J. (1963). Optimal Investment and Financing Policy. The Journal of Finance,18(2), 264.
Guttman, I., O. Kadan, and E. Kandel. (2010), ``Dividend Stickiness and Strategic Pooling,’’ Review of Financial Studies, 23, 4455-4495.
C. Hsiao (2003), Analysis of Panel Data
Ivashina, V., & Scharfstein, D. (2010). Bank lending during the financial crisis of 2008. Journal
27 Jensen, G. R., Solberg, D. P., & Zorn, T. S. (1992). Simultaneous Determination of Insider Ownership, Debt, and Dividend Policies. The Journal of Financial and Quantitative
Analysis,27(2), 247.
Jensen, M. C. (n.d.). Agency Cost Of Free Cash Flow, Corporate Finance, and Takeovers. SSRN Electronic Journal.
Koch, P. D., & Shenoy, C. (1999). The Information Content of Dividend and Capital Structure Policies. Financial Management,28(4), 16.
Lintner, J. (1956). Distribution of Incomes of Corporations Among Dividends, Retained Earnings, and Taxes. The American Economic Review,46(2), 97-113.
Litzenberger, R. H., & Ramaswamy, K. (1982). The Effects of Dividends on Common Stock Prices Tax Effects or Information Effects? The Journal of Finance, 37(2), 429-443.
Maddaloni, A., & Peydró, J. (2011). Bank Risk-taking, Securitization, Supervision, and Low Interest Rates: Evidence from the Euro-area and the U.S. Lending Standards. Review of
Financial Studies,24(6), 2121-2165.
Michaely, R., Thaler, R., & Womack, K. (1994). Price Reactions to Dividend Initiations and Omissions: Overreaction or Drift?
Miller, M. H., & Modigliani, F. (1961). Dividend Policy, Growth, and the Valuation of Shares. The Journal of Business J BUS, 34(4), 411.
Pettit, R. (1977). Taxes, transactions costs and the clientele effect of dividends. Journal of
Financial Economics,5(3), 419-436.
Purnanandam, A. (2010). Originate-to-distribute Model and the Subprime Mortgage Crisis. Review of Financial Studies,24(6), 1881-1915.
Reinhart, C. M., & Rogoff, K. S. (2008). Is the 2007 US Sub-Prime Financial Crisis So
Different? An International Historical Comparison. American Economic Review,98(2), 339-344.
Rozeff, M. S. (1982). Growth, Beta And Agency Costs As Determinants Of Dividend Payout Ratios. Journal of Financial Research,5(3), 249-259.
28 Stock, J. H., & Watson, M. W. (2007). Introduction to econometrics. Boston:
