University of Amsterdam
Economics and Business, specialization Finance and Organization Supervisor: Mr Jan Lemmen
2nd February 2016
The effect of the financial crisis of 2007
on the dividend payout policy of
American public companies from the
materials sector
Bartosz Stankiewicz 10630546
1
Statement of Originality
This document is written by Bartosz Stankiewicz 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.
2
Abstract
Purpose: The purpose of this paper is to investigate whether the dividend policy of American
public companies from materials sector changed following the financial crisis of 2007.
Design/Methodology/Approach: The life-cycle model introduced by DeAngelo et al. (2006) is
used to investigate the relationship between dividend payout ratio and life-cycle variables. The data sample for this research is similar to Fama’s and French’s (2006-2008).
Findings: This study shows evidence that dividend payout ratio does not depend on the
life-cycle variables for the American public companies from materials sector. The dividend policy did not change significantly in the years 2006-2008 and only size variables turned out to be significant, however their effects on DPR were ambiguous.
Originality/value: The outcomes of the study show that the dividend policy did not shift after
the shock of financial crisis. The research provides evidence that firms did not change dividend policy following the worsening of their financial situation.
Keywords: dividend policy, materials sector, financial crisis Paper type: Research paper
3
Table of contents
Abstract ... 2
1. Introduction ... 4
2. Literature review ... 6
3. Research question and hypotheses ... 14
4. Methodology ... 15 5. Results ... 19 6. Conclusions ... 24 7. Discussions ... 26 References ... 30
List of Tables
Table 1. Table showing the expected influence of independent variables on the dependent variable, together with the short explanation
Table 2. Table showing the descriptive statistics of both dividend and non-dividend payers for the years 2007-2009
Table 3. Table showing the correlation coefficients between both dependent and independent variables
Table 4. Table showing the descriptive statistics of dividend payers for the years 2007-2009 Table 5. Table showing the descriptive statistics of non-dividend payers for the years 2007-2009 Table 6. Table showing the model summary for the complete regression
Table 7. Table showing the regression output of the first regression with DPR being the dependent variable
Table 8. Table showing the model summary for the complete regression
Table 9. Table showing the regression output of the complete regression with DPR being the dependent variable
4
1. Introduction
This research takes into account the effect of financial crisis of 2007 on the dividend policy of American public companies from the materials sector. The financial crisis hurt a lot of companies from various sectors financially. However, every sector of the industry has different characteristics, hence the consequences of economic downturn will vary. Therefore it is advisable to look at each sector separately. As the materials sector is characterized by paying high dividends (it paid the highest dividends in 2015, which is irrelevant for the model used in this study, but served as an incentive when narrowing down the scope of research), it would be interesting to see how the patterns changed following the shock, which was the financial crisis of 2007.
The effect of the financial crisis on the dividend policy was investigated extensively by numerous researchers before. As such, Campello et al. (2009) showed that the financial crises influence heavily the financial condition of the companies even years after the event took place. Those negative consequences are reflected in limiting the investment policy or reducing the dividend payout policy. Other researchers worked in the field of explaining the logic behind paying dividends. The studies made by Black (1976) or Fama and French (2001) clearly indicate that the significance of paying dividends differs across various sectors or even firms’ characteristics. However, no research before has focused solely on the materials sector. Hence it should be captivating to find out how the companies that are operating in the highly cyclical industry react to the worsening of economic conditions in the market where they are operating in. Moreover, the situation on the commodities market will play an important role for the materials companies. This aspect will be touched upon in the subsequent sections of the study. The research question guiding this study is as follows: Does the dividend payout policy of the
US public companies from the materials sector change with the occurrence of the financial crisis of 2007?
The effects of financial crisis are extensive. The most apparent are the decreases of sales and earnings, cutting down the level of operation or selling underused assets. In general, the future
5
cash flows are less certain, hence the management is forced to cut down the expenses within the company. A sensible form of decreasing the costs would be cutting down the dividends as many can see it as giving money for free to the shareholders. However, the motivation behind paying out dividends is much more advanced as it acts as a signal to the investors. It means that any cuts of dividend policy will be seen as a bad sign and will cause suspiciousness among the investors, which might reflect in the drop of the share price. Hence the companies might view a dividend cut as the last option when choosing the forms of cutting costs. How is it in reality; does the company retain its dividend policy in the event of the financial crisis or not? Does a decrease in total assets and market value contribute to the decrease of dividends being paid? This study will try to contribute to answering those questions.
The model used in this study builds up on the previous methodology introduced by DeAngelo et al. (2006). This model assumes that the dividend policy can be explained by the life-cycle theory. This theory states that the mature companies have higher profitability and less new investment opportunities and as a result will be willing to use excess cash to pay out dividends. Contrary, young companies lacking funds to realize their projects will be retaining all of the earnings and using them up for investment opportunities that will result in the companies’ growth. As far as the model introduced by DeAngelo et al. (2006) is concerned, it is a logistic regression equation that indicates the probability with which the company will pay the dividends based on all the control variables included in the equation, which are the proxies of firms’ financial characteristics. However, for the purpose of this research the multivariate regression will be used with the dividend payout ratio as dependent variable and the variables introduced by DeAngelo et al. (2006) as independent variables together with few variables suggested by the author of this study. The details will be listed in Section 4. As it is assumed that in the year 2006 the managers and directors of American public companies were not aware of the existence of the subprime mortgage crisis, 2006 will be used as a base year. Subsequently, the regression for the years 2007-2009 will be conducted to test whether the dividend payout policy changed in this post-crisis period.
6
The main regression used in this study turned out to be insignificant. The dividend payout ratio does not depend on the variables suggested by DeAngelo et al. (2006) and it turns out that only PB ratio and logSize variables are significant. It supports the claim that firm size has an influence on the dividend payout policy as both significant variables can be classified as proxies for size. However, the insignificance of the other dependent variables together with dividend dummies is not according to the expectations and will be evaluated in the later stages of the study. The remainder of this study is organized as follows. In section 2 the literature concerning the financial crisis of 2007, the dividend payout policy and materials sector is presented. In section 3 the hypotheses and research question are discussed. Section 4 explains the methodology of the paper; how the results were collected and analyzed. Section 5 entails the actual results of the regressions. Finally, in sections 6 and 7 the obtained results are concluded and discussed.
2. Literature review
The literature review presented in this section consists of three main parts. Firstly, the definition of the financial crisis is presented, together with the links to the financial crisis of 2007. Next, the concept of paying dividends is discussed in detail in order to understand the importance of this mechanism. Finally, the characteristics of the materials sector are provided with necessary links to the aspects of financial crisis and dividend policy.
2.1. Financial crisis – definition
In general, a financial crisis occurs when the financial assets’ nominal value suddenly decreases as a result of some economic shock. As the financial crisis is a relatively broad term to describe all the situations that caused the financial assets to lose the substantial parts of their values, we can distinguish from few different situations. Looking back from the perspective of the past, we can identify different drivers of a financial crisis. Those include stock market crashes, currency crises or sovereign defaults.
7
In his paper, Mishkin (1992) analyzes the phenomenon of financial crisis. In addition, he presents a precise definition of this term, which will be presented in the following sentence. ‘A financial crisis is a disruption of financial markets in which adverse selection and moral hazard problems become much worse, so that financial markets are unable to efficiently channel funds to those who have the most productive investment opportunities. As a result, a financial crisis can drive the economy away from equilibrium with high output in which financial markets perform well to one in which output declines sharply.’ (Mishkin, 1992, p. 115)
There is an ongoing debate among the researchers to determine the exact time frame of the financial crisis of 2007. Although it is called in this paper the financial crisis of 2007, there are numerous researchers trying to prove that 2008 should be selected as the beginning of the financial crisis (Grigor’ev and Salikhov, 2009). However, as the study takes into account financial data, the end of the fiscal year will usually be 31st of December. As a result, any major economic disruptions happening during the year will impact the financial data for this year as well. All the major public American stock exchanges noted declines for the fourth quarter of 2007. Furthermore, consumer confidence decreased at the end of 2007, which had an impact on sales figures of the companies. Although the bank bankruptcies and declines in price levels occurred in 2008, the year 2007 is still used as a beginning of the financial crisis in this paper.
2.2. The credit crisis of 2007
The main cause of the financial crisis was the over usage of the financial instruments called securitized assets as the investors started finding them as an interesting diversification of their portfolio. One of the types of loans that banks give is collateralized loan. Usually the real estate serves as collateral. The low interest rates in the 2000s led to the creation of the housing market bubble, which was characterized by the increased demand for housing and housing investment.
The banks had two models for dealing with the loans: ‘originate and hold’ model and ‘originate to distribute’ model. In the first case, the bank holds the loan until maturity. In case of the failure of the creditor to repay the loan, the bank obtains either collateral (if it was established
8
before setting up the loan) or is left with nothing. Hence, banks were given incentive to carefully monitor and select the customers in order to minimize the risk of default on the loan. Before the 2000s the loans were offered only to the most credit-worthy borrowers. However, the introduction of low interest rates made the loans more accessible and let the banks engage in the more risky investments.
This has led to the second model of handling loans; ‘originate to distribute’. In this model the banks transfer the credit risk to other entities, mostly through the securitization. This is the process of converting illiquid assets, such as mortgages, car loans or credit card loans, into a marketable debt security. The bank does it in order to form a portfolio of such debt securities and then slice it into some tranches. Then those tranches are being transferred to special purpose vehicle, which is an entity that only collects interest and principal from the portfolio of debts and passes those collections to the tranche owners. The tranches have different levels of riskiness and corresponding return. Senior tranches are the first to be paid out; hence they guarantee the lowest return and the lowest risk of default. Junior or toxic waste tranches only pay out after all the previous tranches were paid out, hence they have the highest risk and the investors require the highest return. The last step in the securitization process is selling the tranches by special purpose vehicles to investors in the form of asset-backed securities (ABSs), as indicated by Grigor’ev and Salikhov (2009).
Knowing the mechanism of the process that lead to the financial crisis is essential to understand why banks actually engaged in dealing with debt securities. As suggested by Cabral (2013), the pay structure of the bank managers led to the excessive investment in the debt securities. As the performance of bank managers was determined with respect to their peers, they were inclined to engage in taking excessive risks in order to get a bonus. Hence, the bank ended up possessing more risky tranches under its balance sheet. Furthermore, there was an agency problem at stake as the borrowers were not motivated to disclose all the information about their abilities of repaying the loans. Banks and credit rating agencies had also weak impulse to judge the quality of the securities at stake. As a result, the number of people that were unable to repay the mortgage loans was constantly increasing. The banks were gaining
9
the control over the collaterals established in the loan contracts (houses) and were selling them off. With the greater supply of houses on sale came the drop of their prices. This drop of prices made the borrowers less likely to stick to their loan payments as the value of the underlying asset was less than the value of the loan. As a result, the borrowers stopped paying the interest and were handing in the keys to their houses to the lenders.
The increased number of defaults on mortgages meant that ABSs started to pay out less. Then the rating agencies started to downgrade the ABS tranches, so the overall demand for asset-backed commercial papers dropped significantly, leaving many investors with huge losses and forcing them to conduct fire sales of ABSs. The same happened with banks that were having those securities in their balance sheets. As a result, many of them went bankrupt or had to be saved by the central banks.
2.3. The dividend payout policy
This section concentrates on the concept of paying dividends. The irrelevance theorem of Modigliani and Miller is presented, together with its criticism and the follow up theories, including signaling and clientele theory. At the end, the link with the financial crisis of 2007 is made.
2.3.1. Modigliani and Miller irrelevance theorem and its critique
The logic behind paying dividends and the relation between dividends and the firm value intrigued many people from the world of finance. One of the first researchers that tried to investigate this phenomenon were Modigliani and Miller (1961). In their paper they assume that the markets are perfect, investors are guided by rational decisions and that there is a perfect certainty in the market. Moreover, they assume that the price of the share fully reflects its potential; the rate of return of the share ought to be the same throughout different phases of the market. Hence, in this idealistic world the choice of dividend payout policy does not matter. Modigliani and Miller tried to motivate their findings using the existing valuation theories. As such, the increase in dividend payout policy induces a decrease in terminal value of
10
the share of the same magnitude. Hence, this theory of Modigliani and Miller is often referred to as irrelevance theorem.
Notwithstanding, because of the idealistic assumptions, irrelevance theorem was subject to heated criticism. DeAngelo and DeAngelo (2006) claim that the market imperfections that exist in real world make the irrelevance theorem futile. One of the main arguments that they are using corresponds to the possible ways of distributing the free cash flows (FCF). The irrelevance theorem holds when the managers choose to distribute 100% of the FCF of the company to the shareholders. However, if managers decide to retain some of the FCF or use it for share repurchase instead of paying out dividends, then the share price will be the reflection of the dividend policy and the investment policy, hence the results contradict the irrelevance theorem.
2.3.2. Signaling theory
Signaling theory presents the concept of paying dividends as the agency theory. In the principal-agent model representing the dividend payout, the principal is the group of shareholders and the agent is the management of the company. An agency problem exists when there is asymmetric information between the principal and an agent and the two sides have different interests. As the agent has more information available than the principal, it will need to send a credible signal to the principal in order to convince him that the situation in the company is optimal. Doing so, the manager will provide the shareholders with some inside information available only to the management. This signal could be a dividend payout policy that will reassure the shareholders about the performance of the company.
Multiple researchers touched on the idea of dividend policy as a signaling device. Divecha and Morse (1983) find out that both signaling theory and clientele theory are guiding investors’ decisions. However, the reaction of investors to the change in dividend policy varies as every investor has different preferences and risk aversion. The results indicate that the companies that increased their dividends had abnormal returns. The decrease in dividend payout only strengthened this trend. As a result, investors see the increase in earnings and decrease in the
11
dividend payout as positive signs. Another study conducted by Brav et al. (2005) indicates that managers do not consider dividend payout policy as a costly signaling device. The managers are aware of the information that shareholders receive when seeing the dividend policy of the company, hence they would be reluctant to decrease dividends. Even in the event of liquidity problems, managers will wait until the competitors decrease their dividends in order not to send the wrong signal. The managers know that there is asymmetry in the reaction of the shareholders to changing dividends; investors react more heavily to the decreases of dividends than to the increases.
2.3.3. Clientele theory
When analyzing the clientele effect, the researchers mostly focus on the impact of taxation on the decisions about paying dividends. Feldstein and Green (1983) state that investors are willing to maximize the market value of their assets and that taxation is one of the burdens preventing investors from achieving it. The high level of taxation of dividends results in the shift of investors’ preferences from dividends to capital gains or share repurchases. The conclusion is that the dividend payout policy is dependent on the taxation. Brav et al. (2005) disagree with this conclusion and using their survey, they suggest that managers are not fully guided by the tax considerations when deciding on retention of earnings, dividend payout or share repurchase. The choice of changing the levels of distribution of excess cash is not dependent on the tax rates and even small investors prefer to receive dividends also in the existence of tax burden.
2.3.4. Other theories and determinants
Apart from the signaling and clientele theories, researchers try to find other determinants of the dividend policy. They look into the firm characteristics in order to determine how certain factors inside the company influence the decisions regarding paying dividends. Fama and French (2001) find out that the percentage of companies paying cash dividends decreased from 66.5% in 1978 to 20.8% in 1999. The change is associated with the changing characteristics of the companies, which include the firm size, profitability and investment opportunities. Larger
12
firms with higher profitability are more likely to pay dividends, whereas firms with attractive investment opportunities will be hesitant with distributing the funds to the shareholders. Another important determinant of dividend payout policy is the life-cycle model of the company and the liquidity problem, which were heavily discussed by Redding (1997) in his study. The theory states that the company at the early stages of its life-cycle will need to retain all the earnings it generates in order to grow. At the later stages the firm will mature and the costs of retaining the capital will increase, which will motivate the company to pay dividends. As far as the liquidity is concerned, Redding (1997) found that when liquidity and investment needs are met, investors can expect an increase in dividends. Managers being surveyed by Redding claim that the reduced level of liquidity might have a negative impact on the share price, which will force the manager to solve the liquidity problems first by reducing the amount of shares being repurchased. If this is not enough, the manager might ultimately cut the dividends.
When analyzing the rationale behind paying dividends, the behavioral theory of dividends needs to be described as well. Apart from the signaling and clientele theories, which constitute also part of behavioral theory of dividends, Becker et al. (2011) highlight the different preferences that various investors have regarding the dividend payouts. Dividends are referred as ‘birds in the hand’. They can be classified as tangible capital gains, whereas retained earnings are uncertain as the policy of the company regarding distributing the earnings is not always stable. Hence, risk averse investors should always prefer dividends, even in the instance when the stock price fully reflects the retained and future earnings. Investors simply treat the dividends as a safety net, which is caused by purely psychological reasons. Furthermore, another advantage of dividends over future earnings is reflected in the volatility. Due to the fact that the investors have money in hand, not all of their investment depends on the future performance of the stock. In the case of future earnings, any event that might cause the share price to go down will ultimately decrease the probability of the company paying dividends in the future, which will make the investors worse off. This behavior is explained by the loss aversion of the investors.
13
2.3.5. The link between dividend payout policy and financial crisis
Various researchers tried to find the relationship between the dividend payout policy and financial crisis. The most famous survey was carried out by Campello et al. (2009) who investigated the corporate spending of the companies in the midst of the financial crisis in 2008. The main results state that the financially constrained firms in the US were planning to reduce the dividend payments by 14%, capital investments by 9%, marketing expenditures by 33%, technology spending by 22% and employment by 11%. All the planned cuts are for the year 2009. The levels of reductions for unconstrained firms were much lower and much less significant. The companies that were granted speculative level were planning significant cuts in all categories of expenditures. Investment-grade firms were planning cuts in all expenditures as well, but to a lower degree than speculative-grade firms. With the company’s size in mind, smaller companies were inclined to cut capital spending and marketing expenditures, while larger companies were mostly reducing their spending on technology. The main findings of the survey indicate that in pre-crisis years the planned dividend cuts were estimated at 8%, but during the crisis years this value increased to 28%. Moreover, the financially constrained companies were found to already plan cutting the dividends before the crisis. However, the shock only triggered the cuts and forced the company to sell assets in order to obtain cash.
2.4. The characteristics of materials sector
The Global Industry Classification Standard is one of the standard industry taxonomy developed in 1999 by Standard & Poor’s and MSCI. It consists of 10 sectors into which S&P has categorized all public companies. Materials sector is one of the ten, with the code 15, which was used when the data were collected. Now the short characteristics of the sector will be provided.
As stated by the Sector Overview in Morningstar, materials sector consists of firms that are engaged in the discovery, development and processing of raw materials, which include chemical products, forestry products, mining and refining of metals. Due to strong links with the construction sector, companies from materials sector heavily depend on the strong economy. In addition, as they are involved in the processing of raw materials, the changes in
14
prices of raw materials will heavily impact the profits that the companies will make. Ultimately, the commodity prices are the key determinants of the earnings of the materials’ companies. As the competition among the materials companies is largely driven by the price of the final product, decreasing the costs of production is essential. During the downturns, producers with high fixed costs are the ones more likely to go bankrupt. The performance of the materials sector also relies on the geopolitical situation as the economic growth of emerging markets is heavily dependent on the supply of raw materials. Another factor that needs to be considered is exchange rate risk. Although the prices of all the commodities are denominated in US dollars, the actual costs of extraction or refinement are incurred in foreign currencies, be it South African ran, Brazilian real or Australian dollar.
3. Research question and hypotheses
The literature review helped to recognize the factors that are likely to influence the dividend payout policy of the company. The previous research has found some evidence that firm size, liquidity, some parts of clientele and signaling theories and financial performance all contribute to the attitude of corporate managers towards the dividend policy. This study will aim to incorporate some of those variables in the form of hard financial data and later will try to determine whether or not the event of financial crisis actually influences the change in the dividend policy. Hence, the research question guiding this paper is:
Does the dividend payout policy of the US public companies from the materials sector change with the occurrence of the financial crisis of 2007?
How do managers respond to the economic shock that is the financial crisis? Do they change their dividend policy as the pressure on retaining the earnings builds up during bad times? Are they worried about sending negative signals to investors? Will they be more inclined to cut spending in other departments than dividend payments? The answers to those questions will be given by the multivariate regression model that will be outlined in the next section. However, in order to obtain credible results, the adequate hypotheses must be presented:
15
H0: the dividend payout ratio does not change following the financial crisis
H1: the dividend payout ratio changes following the financial crisis
The linear model that will be used to test the hypotheses will include several coefficients that were proven significant by the previous research, mostly by Fama and French (2001), DeAngelo et al. (2006) and Hauser (2013). However, this will be extensively analyzed in the next section. But building up on the conclusions achieved by the academics in the past, it is expected that the null hypothesis will turn out insignificant, leading to the acceptance of the alternative hypothesis.
4. Methodology
This section of the study outlines the methodology that will be used to answer the research question and test the hypotheses. All the variables used in the regression formula will be explained, followed by the outline of the collection of necessary data, which included eliminating some data as well.
4.1. Model
Through the extensive research done since 2008, it has been proven that the financial crisis influences the financial performances of the companies. This impact has been reflected in various financial data, including size, profitability or earned equity, as indicated by the research done by Denis and Osobov (2008). Then financial ratios that include earnings to total assets, equity to total assets, asset growth rate or return on assets were included by the work of DeAngelo et al. (2006). Then the variable relating to liquidity of the company, Price-to-Book ratio and the new variable for size (market value divided by the personnel) were added as well. As a result, the final regression equation looks as following:
( )
16 Where:
is the dividend payout ratio defined as total dividends divided by retained earnings
is the ratio of retained earnings to the total equity owned by shareholders
is the ratio of total equity to total assets
is the ratio of cash to total assets
is the price-to-book ratio
is the ratio of current assets to current liabilities
is the enterprise value divided by the number of employees, which is taken to logarithm due to the substantial size
is the asset growth rate for the years 2007-2009
is the return on assets measured by dividing the net income by the total assets
is the dummy variable indicating whether the company was paying dividends in the prior year, be it 2006, 2007 or 2008
4.2. Definitions of the variables
As it can be seen in the regression equation, there are a lot of independent variables used for evaluating whether the dividend policy is affected by the financial crisis and certain factors related to the firm characteristic. As it was stated earlier, it builds upon the research conducted by DeAngelo et al. (2006), Denis and Osobov (2008) and Hauser (2013).
The dividend payout ratio is the ratio of total dividends to the retained earnings. Higher values of this variable will indicate that the company is in mature state and hence dividends are a significant part of sharing the earnings. The ratio of retained earnings to total equity owned by the shareholders is the variable indicating the life-cycle stage that the company is in. Low values of RE/TE indicate that the company is in the capital infusion stage, hence they will accumulate capital instead of distributing them. The ratio of total equity to total assets indicates the source of funding that the company is getting; whether the company accumulates the capital by issuing shares or debt. The ratio of cash to total assets is a proxy for size, just as the log of the
17
ratio of market value to the number of employees. The higher the ratio, the more mature a company is. Price-to-Book ratio is the proxy to determine whether the company is overvalued or not. This is a size proxy as well. It will also signal the stage in the life-cycle model in which the company is at the moment. Current ratio signifies the financial condition of the company and whether it will be able to maintain the current dividend policy. It is also a proxy for liquidity. The asset growth rate is a growth proxy. The return on assets is the proxy used for the profitability.
4.3. Sample and data collection process
All of the data utilized in this study were taken from the Compustat North American database. This database contains mainly American and Canadian companies, hence all the Canadian companies had to be eliminated as the study is interested in investigating only American public companies working in the materials sector. After filtering for GICS sectors (selecting industry code 15 for the materials sector) and selecting all the financial data that will be necessary to conduct the research, the first version of data sample was found. In this way, 339 public American companies working in the materials sector were identified. Then, the further selection based on the missing values was carried away. It has led to elimination of exactly 153 companies, so the final number of the companies was equal to 186, both paying and not paying the dividends at all. In order to fully reflect the situation of public American companies working in the materials sector, no selection based on the difference in the life-cycle stage of the companies was made. It means that both the companies listed on S&P500 and the companies hiring 14 employees are contrasted with each other. It follows the same system utilized by the previous researchers that the regression model is based on (Hausner, 2013; DeAngelo et al., 2006).
In order to give a broader picture of the expected effect of each independent variable on the probability of paying the dividends, the table with a general overview will be presented. In the columns, the expected relationship between dependent and independent variable will be presented and the rationale behind this expectation will be outlined as well.
18
Variable Expected
relationship Rationale
Positive
The higher proportion of retained earnings will indicate more mature company, which will make paying dividends easier
Positive
The higher the proportion of shareholders in comparison to debtholders, the more likely the company is to pay dividends in order to maintain good relationship with their source of capital
Positive
The more excess cash is available, the more likely it is that the company will be able to share these extra funds with their
shareholders
Positive If the company is overvalued, it could be inclined to send positive signals to shareholders in order to retain the current PB ratio
Positive
The higher the ratio, the more liquid a company becomes. Hence, it is easier for it to accumulate capital. In this way it technically should omit short term financial problems, which would make
dividend payout more feasible
Positive Following the life-cycle theory, more mature companies should be better equipped to paying dividends
Negative
The lower the growth rate of the company’s assets, the more likely it becomes that it enters a mature state, which makes paying
dividends more likely
Positive
As the return on asset is a proxy for profitability, higher values of net income (the component of ROA) will signal a good financial
condition of the company, hence should make dividends more likely to be paid
Positive The managers will be hesitant to change the dividend policy because of signaling theory
19
Table 1. Table showing the expected influence of independent variables on the dependent variable, together with the short explanation
5. Results
In this section all the results of this study are presented. First, the descriptive statistics is given. The distinction between dividend paying companies and companies that do not pay dividends at all for the period 2006-2008 was used. Apart from that, Pearson correlation table is given as well in order to test for the multicollinearity between independent variables. Later, the actual linear regressions are given, together with the significance levels of each variable. This is the model testing section.
5.1. Descriptive statistics
Firstly, the descriptive statistics of the whole sample of observations will be taken into account.
Variable N Mean Std. Dev. Min Max Year 558 2008 0.82 2007 2009 DPR 288 -0.25 3.50 -42.00 2.49 Current ratio 558 3.42 7.16 0.01 112.91 TE/TA 558 0.16 -2.09 28.11 0.99 RE/TA 558 -4.68 -44.40 964.57 1.22 Cash/TA 558 0.10 0.14 0.00 0.94 LogSize 558 6.16 -1.60 0.65 12.32 PB 558 4.74 -66.19 687.63 1140.00 AGR 558 2.26 -39.72 0.98 921.33 ROA 558 -0.38 -4.03 -86.26 9.74 Div 2006 186 0.49 0.50 0 1 Div 2007 186 0.51 0.50 0 1 Div 2008 186 0.52 0.50 0 1
Table 2. Table showing the descriptive statistics of both dividend and non-dividend payers for the years 2007-2009
As it can be seen, the data are evenly spread among the years. The means and standard deviations of the variables seem reasonable only in certain cases; definitely the values for
20
Current ratio, RE/TA, PB and AGR contain some outliers which is reflected both in the minimum and maximum values. However, the significance of the variables will be tested by the multivariate regression later. Instead, in order to get a better insight into the values obtained so far, the descriptive statistics for dividend and non-dividend payers will be presented. Next, the statistics of dividend payers will be analyzed in greater detail when looking at each year separately.
Current
ratio TE/TA RE/TA Cash/TA LogSize PB AGR ROA Div 2006 Div 2007 Div 2008 DPR Current ratio 1.00 TE/TA 0.31 1.00 RE/TA -0.05 0.37 1.00 Cash/TA 0.44 0.26 0.09 1.00 LogSize 0.44 0.48 -0.01 0.30 1.00 PB 0.00 0.13 0.08 0.06 0.00 1.00 AGR 0.07 0.12 -0.07 0.03 0.28 0.00 1.00 ROA 0.01 0.14 0.37 0.19 0.32 0.01 0.15 1.00 Div 2006 -0.04 0.05 0.07 -0.08 0.13 -0.10 0.19 0.13 1.00 Div 2007 -0.04 -0.10 -0.02 -0.14 -0.20 0.05 -0.11 0.02 -0.47 1.00 Div 2008 0.08 0.05 0.01 0.22 0.06 0.05 -0.11 -0.13 -0.48 -0.49 1.00 DPR 0.01 0.15 0.09 0.05 -0.03 0.78 0.01 0.01 -0.11 0.06 0.05 1.00 Table 3. Table showing the correlation coefficients between both dependent and independent
variables
Looking at the correlation coefficients between variables, it can be seen that there is no multicollinearity problem between predictor variables in this multivariate regression. Instead, the high multicollinearity (>0.6) exists only between DPR and PB variables, which is consistent with the findings of this study as the Price-to-Book value is the only significant variable in multivariate regression (at 5% significance level).
21
Variable N Mean Std. Dev. Min Max Year 288 2008 0.82 2007 2009 DPR 288 -0.25 3.50 -42.00 2.49 Current ratio 288 2.44 3.44 0.46 51.50 TE/TA 288 0.44 0.18 -0.09 0.93 RE/TA 288 0.28 0.29 -0.87 1.22 Cash/TA 288 0.07 0.07 0.00 0.36 LogSize 288 6.12 1.14 2.11 11.51 PB 288 1.29 17.67 -288.33 29.50 AGR 288 0.12 0.50 -0.43 6.54 ROA 288 0.05 0.08 -0.47 0.34 Div 2006 96 0.94 0.24 0.00 1.00 Div 2007 96 0.98 0.14 0.00 1.00 Div 2008 96 1.00 0.00 1.00 1.00
Table 4. Table showing the descriptive statistics of dividend payers for the years 2007-2009
Variable N Mean Std. Dev. Min Max
Year 270 2008 0.82 2007 2009 DPR 288 -0.25 3.50 -42.00 2.49 Current ratio 270 4.47 9.55 0.01 112.91 TE/TA 270 -0.13 2.98 -28.11 0.99 RE/TA 270 -9.96 63.46 -964.57 0.73 Cash/TA 270 0.14 0.19 0.00 0.94 LogSize 270 6.20 1.98 -0.65 12.32 PB 270 8.41 93.34 -687.63 1140.00 AGR 270 4.56 57.07 -0.98 921.33 ROA 270 -0.83 5.77 -86.26 9.74 Div 2006 90 0.00 0.00 0.00 0.00 Div 2007 90 0.00 0.00 0.00 0.00 Div 2008 90 0.00 0.00 0.00 0.00
Table 5. Table showing the descriptive statistics of non-dividend payers for the years 2007-2009 Out of the selected sample, it can be seen that the companies are distributed quite evenly between paying and not paying dividends (96 companies paying dividends and 90 companies not paying dividends). Moreover, it seems that the data for dividend payers have less variance.
22
The standard deviations of all variables except from dummies are higher for non-dividend payers. It is a good indicator as for the purpose of this study only dividend payers will be considered.
5.2. Model testing
The first regression will include all of independent variables except from the dummy variables. The second regression will take into account dummy variables as well. The Pearson correlation tables will be given after each regression output.
Source SS df MS N 288 F (8, 279) 55.68 Model 2156.22 8.00 269.53 Prob > F 0.00 Residual 1350.54 279.00 4.84 R-squared 0.61 Adj R-squared 0.60 Total 3506.76 287.00 12.22 Root MSE 2.20 Table 6. Table showing the model summary for the complete regression
Coefficient S.E. t P>t 95% CI Current ratio 0.03 0.05 0.60 0.55 -0.06 0.12 TE TA 1.46 0.94 1.56 0.12 -0.39 3.31 RE TA 0.00 0.56 0.00 1.00 -1.10 1.11 Cash TA -0.32 2.17 -0.15 0.88 -4.60 3.96 LogSize -0.27 0.16 -1.76 0.08* -0.58 0.03 PB 0.15 0.01 20.45 0.00*** 0.14 0.17 AGR 0.16 0.27 0.60 0.55 -0.37 0.70 ROA 0.69 1.91 0.36 0.72 -3.07 4.46 constant 0.50 0.82 0.61 0.54 -1.11 2.11
Note. Significance at *10%, **5%, ***1% level
Table 7. Table showing the regression output of the first regression with DPR being the dependent variable
23
Looking at the results, despite the fact that R2 value of the complete regression is fairly significant (61% of the variance of the population is explained by the sample of this study), the variables tested in the model are not significant, except from the PB variable and the logSize variable, which is significant only at 10% significance level. The outcome that the rest of variables are insignificant is against the predictions and against the conclusions of previous research. Surprisingly enough, both the Price-to-Book ratio and the logarithmic value of size (which is defined by the market value divided by the number of employees) were the only new variables in this study. It means that they were not investigated by the researchers before and to the knowledge of the author of this study, were not investigated before on the American materials sector public companies. However, the relationship of two significant variables differs; PB variable has a positive influence on the dividend payout ratio, hence it can be concluded that higher market-to-book value increases the probability of companies to pay dividends. On the other hand, the negative coefficient of size is against the theory and expectations indicated in Table 1. This might result from rather unusual proxy for size (the ratio of market value to the number of employees), but further discussion on this issue will be carried out in the subsequent section of this study.
Source SS df MS N 288 F (11, 276) 40.30 Model 2161.28 11 196.48 Prob > F 0.00 Residual 1345.48 276 4.87 R-squared 0.62 Adj R-squared 0.60 Total 3506.76 287 12.22 Root MSE 2.21 Table 8. Table showing the model summary for the complete regression
24 Coefficient Std. Err. t P>t 95% CI Current ratio 0.03 0.05 0.54 0.59 -0.07 0.12 TE TA 1.48 0.94 1.56 0.12 -0.38 3.33 RE TA -0.01 0.57 -0.02 0.99 -1.14 1.12 Cash TA -0.58 2.24 -0.26 0.80 -4.99 3.84 LogSize -0.26 0.16 -1.64 0.10* -0.58 0.05 PB 0.15 0.01 20.18 0.00*** 0.14 0.17 AGR 0.22 0.28 0.77 0.44 -0.33 0.76 ROA 0.83 1.96 0.42 0.67 -3.03 4.68 Divdummy2006 0.11 0.83 0.13 0.89 -1.53 1.75 Divdummy2007 0.37 0.83 0.45 0.66 -1.26 2.01 Divdummy2008 0.41 0.83 0.49 0.62 -1.23 2.05 constant 0.14 1.15 0.12 0.90 -2.12 2.39
Note. Significance at *10%, **5%, ***1% level
Table 9. Table showing the regression output of the complete regression with DPR being the dependent variable
Comparing the two regressions, it can be concluded that the dividend dummies do not increase the power of the model. On one hand, the R2 value of the model increased from 0.61 to 0.62, but this is purely caused by addition of three additional dummy variables. Those dummies for whether the company was paying dividends or not didn’t turn out to be significant. In comparison to the first model which excluded dummy variables, the same variables remain significant (PB and logSize). This result is not surprising as it was proven in the Pearson correlation table that there is hardly any influence of dividend dummies on the rest of independent variables.
6. Conclusions
In the previous section, the outcomes of this study were presented. This part attempts to explain the results by linking the findings to the dividend theories that were outlined in section 2.
25
Both the models with and without dummy dividend variables do not provide the expected results. Only PB and logSize variables have a significant influence on the dividend payout ratio. This disproves alternative hypothesis that the dividend payout ratio changes following the financial crisis as the majority of variables used to describe the conditions of the companies is insignificant. One of the papers used in the build-up of this study, Campello et al. (2009), shows that the financially constrained companies will aim to preserve cash during economic downturn. This study did not find a proof of such a relationship. In the literature review section, plans of reductions in different cost segments of financially constrained companies were listed (dividend payments, capital investments, marketing expenditures, technology spending and employment). The model used in this study included all of those factors except from marketing and R&D spending. Although marketing is not a key element in the budgets of materials companies, R&D spending accounts for a significant part of the yearly expenses. It is possible that significant cutbacks occurred in this area.
However, the study proved that two variables are significant at 10% significance level. Those variables are Price-to-Book ratio and logarithmic value of market value divided by the number of employees. The positive sign of PB variable is justifiable; as investors start to find the company more attractive and allocate their funds in it, the ratio increases. Then the company obtains additional funding, which it can use for growth and development. However, if the company is mature, it might be more inclined to distribute the earnings between its shareholders. Knowing this policy, investors will be more inclined to invest in this company following the logic of dividends being treated as ‘birds in hand’, as it was explained in section 2.3.4.
The significance of the second variable related to size is not surprising. Notwithstanding, the sign of the coefficient of logSize is rather unexpected. Namely, the dividend theory suggests that the increase of size and subsequently firm’s maturity leads to the increase of propensity to pay dividends. On the other hand, the outcome of this study is opposite. This might be attributed to rather novel proxy used for size. The number of employees was used in this variable in order to find out whether companies lay off employees during bad times. However,
26
the materials sector is characterized by rather low employment figures as the vast majority of work is being done by machines. As a result, there were even cases in this study that companies were hiring as low as 14 employees, as in the case of Royal Gold Inc. for the year 2006. Hence, laying off employees in companies with such low employment rate will be dangerous for the well-being of the whole company. Furthermore, increase in market value of the company is the sign of good financial condition, hence it can be connected with additional hiring of staff in order to maintain the growth rate. As a result, if the employment growth is proportional to the market value growth, the logSize variable will not represent any difference. Hence, different proxy should probably be used in order to find out the relationship between firm’s size and the dividend payout ratio.
The insignificant results of dividend dummies are not so surprising. Looking at the Table 4, it can be seen that not a lot of companies changed their dividend policy from not paying dividends at all to paying them. In the year 2008, all companies classified as dividend payers were paying dividends. On the other hand, the mean value of the dummy variable for the year 2006 was 0.94, which means that 90 companies out of 96 were paying dividends before the crisis. This number changed to 96 after the crisis, which is entirely against the expectations as six companies started paying dividends after the economic downturn.
All other ratios that were used in the research by DeAngelo et al. (2006) are insignificant. Again, it is entirely against the expectations and the theory. Such result is worrying and can be attributed to the wrong measure used for total assets, as all variables except from the current ratio and retained earnings to total equity use this value as denominator. However, the results were downloaded straight from Compustat under the variable name ‘Total Assets’, so there was no option of committing an error there.
7. Discussions
In this section the research question is discussed, limitations of the study are provided and the questions for further research are given.
27
The research question asked whether the dividend payout policy changes after the occurrence of the financial crisis. The regression analysis shows no influence of all variables except from PB ratio and logSize on the dividend payout ratio. Both influence the DPR dependent variable, but in various manners due to the opposite signs. The reasons for the lack of significance of the variables used in the study of DeAngelo et al. (2006) and the significance of the two variables mentioned before were outlined in the previous section.
Taking closer look into the dependent variable, it can be concluded that DPR might be a noisy proxy for the dividend payout policy. It is defined as the ratio of total dividends divided by total earnings and the denominator value could be problematic. It is due to the fact that the sample used in this study contains numerous companies at different stages of maturity. Hence, the levels of earnings will change depending on the transitory stage at which the company will be. As far as the hypotheses are concerned, looking at the significance of the dependent variables it can be concluded that the dividend payout policy does not change following the occurrence of financial crisis. The value of R2 suggests that the regression model reflects the population quite accurately (0.62 for the complete regression), hence the insignificant results should be accepted. The relationship between dividend dummies suggests that the dividend policy of the companies does not change drastically, and even if they do, the companies start paying dividends instead of stopping. Hence, the alternative hypothesis should be rejected.
This study had a lot of limitations. First of all, the number of companies was heavily trimmed. The majority of the research done in the field focused on all American public companies, instead of limiting itself to just one sector. Further reductions due to the missing values resulted in the final selection of 96 companies that were paying dividends at some stage in the years 2006-2008. Moreover, the materials sector is heavily linked with the prices of raw materials. Unfortunately, due to the large scope of companies working in this sector it was impossible to account for the fluctuations of prices of commodities. As there were companies focused on extraction and refinement of raw materials ranging from gold to chemicals, the author of this study could not find any feasible way to account for the price changes. This has
28
led to inclusion of more noise in the regression output. Furthermore, the choice of dividend dummies might have been too general. Instead of only checking whether the company was paying dividends or not in the preceding year, the actual change could have been measured; whether the dividends increased or decreased. What is more, the outcome of this study cannot be generalized to all financial crises as it contains only data from financial crisis of 2007. The outcomes cannot be even generalized on the international scale as the data come exclusively from the materials public companies from the American market. Finally, as it was mentioned before, some companies were eliminated due to missing values. Some missing values resulted from the fact that Compustat and other databases could not provide all data (especially for small companies). However, there was a significant number of companies that went bankrupt in the years 2008 and 2009. Those companies were eliminated due to the missing values bias, but the fact that they went bankrupt because of the crisis was important for answering the research question. The exclusion of those values resulted in not reflecting the situation in the materials sector accurately.
Due to the limitations of the study, there are several interesting directions in which the future research can be extended. As proven by Bailly and Elliot (2009), the financial crisis has hit Europe and Asia as well, which leaves a lot of room to investigate whether the reaction of corporate managers to the dividend policies of their companies differed from the reactions of American managers. In addition, the research can be extended to the greater time scale than only 2008. Some researchers claim that the financial condition of the companies was affected by the financial crisis until 2012 or even until now. There is a strong case supporting this line of thought when looking at the performance of European companies. Also the variables that turned out to be significant could be elaborated in more details. Is the ratio of market value to the number of employees an accurate measure of size factor? It will be interesting to investigate the relationship between this proxy and dependent variable of dividend payout ratio for different sectors of economy than materials sector. Moreover, the author was unable to find an accurate proxy for R&D and marketing spending. The inclusion of those variables
29
could increase the significance of regression and will enhance the understanding of effects of the financial crisis on the financial condition of the companies.
30
References
Baily, M.N. & Elliott, D.J. (2009). The US Financial and Economic Crisis: Where does it stand and where do we go from here? Brookings Institution, June 2009.
Becker, B., Ivkociv, Z., Weisbenner, S. (2011). Local dividend clienteles. Journal of Finance, 66, 655–684.
Black, F. (1976). The dividend puzzle. Journal of Portfolio Management, 2, 5-8.
Brav, A., Graham, J.R., Harvey, C.R. & Michaely, R. (2005). Payout policy in the 21st century. Journal of Financial Economics, 77, 483-527.
Cabral, R. (2013). A perspective on the symptoms and causes of the financial crisis. Journal of Banking and Finance, 37(1), 103-117.
Campello, M., Graham, J.R. & Campell, C.R. (2009). The real effects of financial constraints: Evidence from a financial crisis. Journal of Financial Economics, 97, 470-487.
DeAngelo, H., DeAngelo, L., & Stulz, R.M. (2006). Dividend policy and the earned/contributed capital mix: a test of the life-cycle theory. Journal of Financial Economics, 81, 227– 254.
Denis, D.J. & Osobov, I. (2008). Why do firms pay dividends? International evidence on the determinants of dividend policy. Journal of Financial Economics, 89, 62-82.
Divecha, A. & Morse, D. (1983). Market responses to dividend increases and changes in payout ratios. The Journal of Financial and Quantitative Analysis, 18(2), 163-173.
Fama, E.F. & French, K.R. (2001). Disappearing dividends: Changing firm characteristics or lower propensity to pay? Journal of Financial Economics, 60, 3-43.
Grigor’ev, L. & Salikhov, M. (2009). Financial Crisis 2008. Problems of Economic Transition. 51(10), 35-62.
Hauser, R. (2013). Did dividend policy change during the financial crisis? Managerial Finance, 39, 584-606.
Miller, M.H. & Modigliani, F. (1961). Dividend policy, growth, and the valuation of shares. Journal of Business, 34, 411–433.
31
Mishkin, F.S. (1992). Anatomy of a financial crisis. Journal of Evolutionary Economics, 2, 115-130.
Redding, L.S. (1997). Firm size and dividend payouts. Journal of Financial Intermediation, 6, 224-248.
