Effect of a financial crisis on the dividend payout policy of a firm

R.H.J.A. Smits 767593 May 2012

Effect of a financial crisis on the dividend payout policy of a firm

Name:

R.H.J.A. Smits

Date:

May 16th 2012

Program:

Master Accounting

ANR:

767593

1st Reviewer: P.Y.E. Leung MSc

2nd Reviewer: Dr. S. van der Meulen

Accounting Department Faculty of Economics and Business Studies, Tilburg University Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

1

Preface I have spent the last couple of months working on this master thesis, the final product of my study Accounting. It was a real challenge to write this thesis. Aside from some setbacks I am happy to finally finish this thesis. I am also happy to have chosen to continue my study at Tilburg University after finishing my study business economics at Fontys Eindhoven. I have learned a lot of new things in the field of accounting and I am happy to use this knowledge in the future. I would like to thank some people for helping and supporting me during the writing of my thesis. First, I would like to thank my supervisor at Tilburg University, Edith Leung, for her advices and feedback. I would also like to thank my parents, brother and friends for their ongoing support the last couple of months. Remon Smits, Tilburg, May 16th 2012

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

2

Contents Preface ...................................................................................................................................................... 2 Abstract ..................................................................................................................................................... 4 1.

Introduction ...................................................................................................................................... 5

2.

Literature .......................................................................................................................................... 7 2.1

Definition of a financial crisis .................................................................................................... 7

2.2

Understanding securitized assets ............................................................................................. 8

2.3

Pre-credit crisis.......................................................................................................................... 8

2.4

The credit crisis of 2008 ............................................................................................................ 8

2.5

Importance of dividend payout policy ...................................................................................... 9

2.6

Combing dividend payout policy and a financial crisis ........................................................... 15

3.

Research question & hypotheses.................................................................................................... 17

4.

Methodology................................................................................................................................... 19

5.

6.

4.1

Model ...................................................................................................................................... 19

4.2

Definitions ............................................................................................................................... 20

4.3

Sample and data collection ..................................................................................................... 22

Results ............................................................................................................................................. 24 5.1

Descriptive statistics ............................................................................................................... 24

5.2

Simple linear regression .......................................................................................................... 25

5.3

Model testing .......................................................................................................................... 26

5.4

Explanations............................................................................................................................ 30

Discussion........................................................................................................................................ 33 6.1

Discussion................................................................................................................................ 33

6.2

Limitations & future research ................................................................................................. 34

Literature List .......................................................................................................................................... 35 Book .................................................................................................................................................... 36 Thesis .................................................................................................................................................. 36 Appendix A: 2008 Credit crisis ................................................................................................................ 37 Appendix B: Pearson correlation ............................................................................................................ 40 Appendix C: Pearson correlation ............................................................................................................ 41 Appendix D: Full tables regression subsamples ...................................................................................... 42

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

3

Abstract This study investigates the impact of the recent financial crisis on firms’ dividend policy and whether this impact is moderated by firm characteristics. The findings show that overall, the crisis did not affect dividend payout ratios or dividends. However, there is evidence that both firm size and clientele influence the impact of a crisis on dividend policy: dividends increase during the crisis for larger firms and those with a higher percentage of institutional owners. There are several possible explanations for this. First of all, firms might want to communicate to their shareholders that the crisis does not affect the firm (as much) as it does others. This way they want to appear strong, emit trust in the firms’ current and future financial position, attempt to keep shareholders attached to the firm and prevent them from selling their shares in the firm. Another possibility is that firms repay shareholders with excess cash becoming available due to decreasing numbers of good investment opportunities. A third possibility follows up on the agency theory, where shareholders’ distrust towards management forces a bigger dividend payout in times of decreasing numbers of good investment opportunities. Investors could doubt management’s judgment in making good (investment) decisions during the crisis. Extra payment would prevent bad investment decisions on the account of the management.

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

4

1. Introduction This study examines the effect of a financial crisis on the dividend policy of a firm. In 2008 the credit crisis struck, hurting a lot of companies financially. It was clear that especially banks and financial institutions were the first to have difficulties, later followed by other firms. Several studies were conducted to establish what went wrong and what could have been done to prevent or minimize the effect of the financial crisis, for example by Baily & Elliot [2009], Jacobs [2009] and Krugman [2009]. According to Campello, Graham & Harvey [2009+, a financial crisis has consequences for a firm’s financial policy in the years to come, including dividend and investment policy. A number of studies has also examined why firms pay dividends and what the main determinants are (e.g. Black [1976], Feldstein & Green [1983], Fama & French [2001] and Redding [1997]. These studies suggest that the importance of dividends to a firm differ cross-sectionally, making it likely that the impact of the crisis on dividend policy will also vary across firm characteristics. The research question of this study is: “Does a firm change its dividend policy when dealing with a financial crisis?” Where earlier studies established relationships between dividend payout policies and firm value under normal economic conditions, behavior of firms during a financial crisis in relation to dividend policy is relatively unexamined. A crisis affects a firm in a lot of ways like decreasing sales, decreasing earnings, overvalued and unused assets etcetera. The subject of this study will add more to the understanding of how firms behave in times of crisis regarding dividend payout policy. From a practical perspective the exploration of this subject adds knowledge to what shareholders can expect when a financial crisis strikes. Questions arise regarding the effects of a crisis, for example stock prices decline but what happens with the dividend? Does a firm retain it or does it still pay dividend during a crisis. Do all firms adjust their dividend payout or does it depend on firm characteristics like size or investor composition? The main goal of this study is to see how firms behave towards shareholders during a financial crisis. This study uses the model of Denis & Osobov [2008] who used profitability, growth opportunities, earned equity and firm size to establish their relationship with dividend payment. Added to the original model are measures for liquidity, signaling, clientele and a dummy variable to distinguish between the crisis and pre-crisis years. The main expectation is that a financial crisis is associated with a decrease in dividend payout (ratio). Furthermore two hypotheses investigate whether the dividend payout (ratio) of smaller firms decreases more during a crisis compared to larger firms and if that is also the case for low liquidity firms opposed to high liquidity firms. The final two hypotheses research what kind of role signaling and clientele play during a crisis. The hypotheses state that dividend payout (ratio) decrease less when dividends play a signaling role and when firms have more small investors. For this study, hard financial data is used to prove how a crisis influences the dividend payout policy of a firm.

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

5

The analysis consists of several tests where both simple linear regressions and multiple linear regressions are executed. The results are used to establish the relationship of each variable with dividend payout ratio and dividend payout. Two approaches were used to find out whether or not a crisis influences dividend policy of a firm but only one of those approaches turned out to be significant. The main regression using dividend payout ratio turned out to be not significant. Even dividing the sample into groups did not change the outcome of the analysis. This means no evidence could be found to support the hypothesis that variables like firm size, liquidity, signaling and clientele influence the effect of a crisis on dividend payout ratio. However, when using the actual dividend payout, there are some significant results. These results support the hypotheses that firm size and clientele do influence the magnitude of dividend payout of firms during a crisis. However the effect is not a negative one as expected. One of the possibilities is that firms want to appear strong even when experiencing the consequences of a financial crisis and therefore increase dividend payout. Another possibility is that firms pay more dividends because they lack good investment opportunities during a crisis. Excess cash and poor judgment in investment decisions are prevented this way. The other two main variables liquidity and signaling do not influence the magnitude of dividend payment during a crisis. Apart from the influence on the relationship between crisis and dividend payout, both liquidity and signaling are significant in the main model influencing dividend payout in general. Previous studies looked at financial consequences of a crisis on a firms’ performance, why firms pay dividend and what the main determinants are for dividend policy. This study combines these subjects just like Campello, Graham & Harvey [2009] did, but instead of using surveys, this study uses hard financial data which is less biased. This study provides a new perspective on dividend policy during a crisis and contributes more insight on the behavior of firms during a crisis towards shareholders. The remainder of this study is organized as follows. Section 2 describes important literature surrounding the credit crisis and dividend payout policy. Section 3 explains the research question and accompanying hypotheses. The next section, section 4, describes the methodology used in this study including the model and sample. Section 5 presents the data and results. Finally, section 6 will discuss the results and concludes this study.

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

6

2. Literature The literature part of this study consists of two main subjects, the first subject being financial crises and second being dividend payout policies. Each of these subjects will be discussed separately, followed by a discussion of common features. This section will start with useful information regarding defining and discussing a financial crisis, specifically the credit crisis of 2008. 2.1 Definition of a financial crisis Mishkin [1992] describes what a financial crisis is, how it occurs and provides us with a precise definition: “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.” A financial crisis causes panic in the business world and banking sector. When this occurs, the central bank tries to avoid harmful financial disturbances by providing liquidity to the financial system. There are five primary factors which can lead to a substantial worsening of adverse selection and moral hazard in financial markets, which cause a financial crisis. These factors are:  increases in interest rates,  stock market declines,  increases in uncertainty,  bank panics, and  unanticipated declines in the aggregate price level. These factors can also be discovered in the 2008 credit crisis, which is used in this study. It is important to pinpoint the start of the crisis which makes it possible to separate crisis years from pre-crisis years for analysis. Uitdewilligen [2010] looked at the five recognition criteria of a financial crisis and used the years 2004 till 2008 to pinpoint the start of the crisis. US interest rates over the 2004-2008 period show a decline in interest rates as of December 2007 which satisfies the first criterion. Stock markets Dow Jones, S&P 500 and NASDAQ show a decline in growth percentage from the fourth quarter of 2007 and onwards which satisfies the second requirement of Mishkin. Uitdewilligen also provides evidence of increase in uncertainty as costumer confidence decreases at the end of 2007. During 2008 more banks than usual went bankrupt, were taken over or received government support to stay alive which could indicate bank panic. Unanticipated declines in price levels were noticeable at the end of 2008, fulfilling the last requirement of Mishkin’s model. All Mishkin’s criteria were satisfied in 2008, concluding that the credit crisis officially started in 2008. Graphs supporting the results of Uitdewilligen can be found in Appendix A.

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

7

2.2 Understanding securitized assets To understand what happened in the years before the credit crisis, this section will explain some of the backgrounds. Especially during the 90’s, securitized assets became more popular. Securitized assets are very interesting instruments for investors. Financial institutions like banks lend money to customers using for example real estate as collateral for the loan. The banks bundle mortgages with different risk ratios in portfolios and sell these to investors through a special purpose entity (SPE). SPE’s are interesting because external entities securitize the assets before they are sold to the investors. One entity rates the SPE, assesses the risk it carries and provides it with a credit-rating, while the second entity guarantees the SPE (see figure 1). This SPE creates a win-win situation for both the financial institutions and the investors. Investors receive financial instruments which are highly liquid and the financial institutions lower their risk regarding the mortgages and securitized loans [Revsine et al. Figure 1: Structure of a Special Purpose Entity

2008].

2.3 Pre-credit crisis Even before the credit crisis hit in 2008, there already were rumors regarding an upcoming “housing bubble.” Baker *2006+ warned about the dangers of speculation regarding the so-called “housing bubble.” He observed increasing values of real estate but relatively small population and income growth in the US. An increasing number of (second) mortgages could be a problem in the future when the value of real estate suddenly drops. In case of a possible bubble, it is dangerous to do nothing because it endangers the stability of the financial system. Not everyone shared the opinion of Baker. Smith [2005] for example, reported that there was absolutely no evidence of an upcoming bubble in the US. Smith identified a low number of bad mortgages opposed to the total amount of mortgages. He predicted that there would be limited or no change in the near future. M. Smith & G. Smith [2006] also reported that there was no clear evidence of a housing bubble in the near future and there is only suggestive and indirect evidence of an upcoming bubble. Even thought there was a difference in opinion leading up to the credit crisis in 2008, an upcoming crisis is difficult to predict. The next section will take a closer look at the credit crisis. 2.4 The credit crisis of 2008 What some analysts expected happened early 2007. According to Baily & Elliot [2009], investors started doubting the value of the SPE’s and the rating it had received. Homeowners increased their mortgages in the years leading up to the crisis because the value of their real estate increased. The money they Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

8

received was spent on all kinds of things but when the value of real estate decreased, the mortgages became overvalued. This happened at a large scale. Homeowners could not pay their debts anymore and as a result, the value of the financial instruments decreased. Under normal circumstances and on a smaller scale, this would have been a normal and temporary setback, but in this case it was not. In the time period following this development, more and more trading occurred and investors and financial institutions started to sell great numbers of their securitized assets. This sudden event caused a lot of troubles as the pressure on financial firms increased. Some firms started to have serious (liquidity) problems as others went bankrupt like investment firm Bear Stearns and the bank Lehmann Brothers. Firms and investors directly linked to these bad securitized assets were the first to get hit, but by 2008 the whole (financial) market felt the consequences. Not only the US was affected, also Europe and Asia were affected because of import, export and interrelation between banks and investment firms around the globe. The main conclusion of Baily & Elliot is that the worst is over, although this has to be said cautiously. Their expectations are that the US will be the first to recover and Europe and Asia will follow next. Multiple studies concur with Baily & Elliot like Jacobs [2009], Carmassi, Gros & Micossi [2009], Bullard, Neely & Wheelock [2009] en Nobel prizewinner Krugman [2009]. 2.5 Importance of dividend payout policy This section sheds light on different factors relating to dividend payout policy. The first part describes Miller & Modigliani’s irrelevance theorem followed by criticism on this theorem by several researchers. Next, are several theories that explain dividend policy, for example, the signaling theory and the clientele theory are discussed. Some other factors influencing dividend policy will be discussed in short in the last part of this section. 2.5.1 Miller & Modigliani irrelevance theorem Numerous groups, like managers, investors and economists, want to know how dividend policy affects the value of a firm, each group having a different perspective. Miller & Modigliani [1961] were one of the first to study the effect of the dividend policy on the current value of its shares. Miller & Modigliani start with describing effects under the assumption that there are no market imperfections. This can be divided into having “perfect capital markets, rational behavior and perfect certainty.” Furthermore they use the fundamental principal that “the price of each share must be such that the rate of return (dividends plus capital gains per dollar invested) on every share will be the same throughout the market over any given interval of time.” The main conclusion is that given a firm’s investment policy, it does not matter what dividend payout policy is chosen, it will not affect the current price of the shares nor the total return to its shareholders. Searching for additional evidence, Miller & Modigliani assess existing valuation theories, each theory having total firm value in common. When using these theories, they show that when a firm increases dividend payout to shareholders, the terminal value of the shares decreases exactly the same amount. A part of the future dividend stream that would have otherwise have accrued into the future are paid now, resulting in this decrease in terminal value. The fundamental Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

9

conclusion does not need modification when introducing uncertainty in the mix. Current value is unaffected by differences in dividend payments in any future period and thus dividend policy is irrelevant for the determination of market prices, if investment policy is regarded separable from dividend policy. The paper is widely used in later research and referred to as the Miller & Modigliani’s irrelevance theorem. 2.5.2 Critique on Miller & Modigliani’s irrelevance theorem Not everyone agrees with Miller & Modigliani’s irrelevance theorem. Walter [1963] analyzed the influence of the dividend policy of a firm and the changes in value just like Miller & Modigliani. Shareholders share in the operating cash flows of each period to the degree that cash dividends are declared and paid, and in future cash flows insofar as they are reflected in the market price of the stock. The market price of the shares along with their own subjective calculation of anticipated dividend streams and terminal value determines the decision of the investor to buy, hold or sell the shares. Walter [1963] concludes that we do live in a world with imperfections and those imperfections lead to differences in firm value, which contrasts with Miller & Modigliani’s irrelevance theorem. DeAngelo & DeAngelo [2006] heavily criticize Miller & Modigliani’s theorem and describe the irrelevance of the irrelevance theorem in their study, as Miller & Modigliani restrict outputs which in real life cannot be influenced. For example, according to Miller & Modigliani, all capital structures and dividend policies are optimal and provide identical stockholder value, and so the choice a manager makes is irrelevant. This is not true because irrelevance fails when not 100% of free cash flow (FCF) is paid out to shareholders, allowing retention of FCF. When retention is allowed, investment policy is not the sole determinant of value, but dividend policy is as well. Another argument is that more than one payout policy could satisfy optimal distribution of free cash flow to currently outstanding shares. The choice a manager makes in this case is irrelevant but the manager is also confronted with possibilities which do not meet the requirement of distributing 100% of FCF. It is assumed that managers always choose valuemaximizing dividend payout options, but in fact managers are confronted with a certain kind of opportunity set to choose from. This makes the choice of the manager indeterminate and not irrelevant. 2.5.3 Dividend through the years A firm has to make a decision on whether or not to pay dividend, decide on the level of the dividend and analyze what factors are important to consider in the decision-making process. This results in a decision regarding dividend policy, investment policy and sometimes share-repurchase policy. “The dividend puzzle” as Black [1976] called it, starts with the question: “Why do corporations pay dividends?” Black discussed a lot of factors which could possibly influence dividend policy. He finds numerous factors influencing policy decisions but eventually he still cannot answer the question why firms pay dividend. He simply answers with: “We don’t know.” Black discusses all main factors that might influence policy decisions like taxes (part of clientele theory), transaction costs (part of clientele theory), what kind of Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

10

information dividend communicates to shareholders (signaling theory), capital structure and investors’ demand for dividend (part of clientele theory). With signaling and clientele theory being discussed in the next sections, one remains unexplained: capital structure. Black includes capital structure because of the conflict of interest between equity holders and debt holders. A company cannot pay dividend as it pleases, because when dividend payout increases, fewer assets are available for debt holders should the firm (suddenly) go bankrupt. Evidence found to support this assumption is less significant than factors relating to either clientele theory or signaling theory. Regarding dividend policy, Fama & French [2001] discovered that the percentage of firms which paid cash dividend decreased from 66,5% in 1978 to 20,8% in 1999. This was a result of changing firm characteristics and newly introduced listings. Fama & French analyzed variables like profitability, investment opportunities and firm size. They conclude that larger firms and firms with a high profitability are more likely to pay dividends. Financially distressed firms stop dividend payment while smaller firms usually do not pay dividend. Firms with higher investment opportunities are less likely to pay dividends. Finally, payers of dividends are larger in size than non-payers. Just like Fama & French [2001], Denis & Osobov [2008] find that there is a substantial decline in the proportion of firms paying dividends in the US. They also find evidence of determinants of the likelihood of paying dividends like firm size, growth opportunities and profitability. Additionally, Denis & Osobov provide evidence on signaling, clientele and life-cycle explanations by examining the concentration of dividends and earnings. The life-cycle theory of dividend policy was proven and according to this theory as firms matures the expected cost of retention increases. Consequently the propensity to pay dividends increases. Firms which already pay dividend are more likely to continue doing so and firms which do not pay dividend are not likely to initiate dividend payment. Additionally, the payment of dividend is concentrated among the large firms. In line with Denis & Osobov [2008], Redding [1997] determined that firm size and liquidity affect the decision whether to pay dividend or not. It is argued that larger firms tend to be more liquid because of the higher demand for shares and the tendency to pay more dividends. From the life-cycle theory point of view, firms pay out assets to shareholders upon reaching an efficient firm size, which is preferred by institutional investors. As the firm matures, the costs of retention increase resulting in a higher propensity of paying dividends. Liquidity is also discussed in the study of Brav, Graham, Harvey & Michaely [2005], along with the decision-making process of financial executives. They find that the relationship between earnings and dividend weakened and that after maintaining the level of dividends per share, payout policy is a second-order concern. When investment and liquidity needs are met, increases in dividend can be considered. Survey evidence suggests that dividend choices are made simultaneously (or a bit sooner) than investment decisions and that financial executives anticipate good Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

11

investment decisions. Repurchase decisions are made later. Regarding liquidity, executives think that reduced liquidity can hurt their stock price. Managers would reduce share repurchases first to try and solve the liquidity problem, because of the flexibility it offers. When that is not enough and the manager absolutely has to he might cut dividends. Brav et. al. [2005] describe some basic rules concluding the study: “Expect a severe penalty for cutting dividends, do not deviate far from competitors, maintain a good credit rating, have a broad and diverse investor base, maintain flexibility, and given that an important portion of investors price stocks using earnings multiples, so do not take actions that reduce EPS.” 2.5.4 Signaling theory The agency theory plays a big part in the signaling theory. The agency theory states that there is a conflict of interest between the principal and the agent, in this case shareholders being the principal (owner(s) of the company) and management being the agent (who controls the company on a daily basis). An agency problem exists between the principal and agent because of asymmetric information provision and both parties having different interests. Dividend can be seen as a communicative mechanism (possibly costly) to provide shareholders with additional (inside) information. Management selects a certain dividend policy to communicate how well a firm is performing and how management expects to perform in the future in terms of sustaining income or growth. This view is supported by multiple studies. Feldstein & Green [1983] describe that there is a natural distrust between principal and agent. Shareholders prefer to receive dividends because they fear that management will make poor investment decisions, increase management compensation and/or make other wrong decisions resulting in lower earnings for shareholders. The market will try to adjust to this by specialization of ownership and dividend policies. Combining two methodologies, Divecha & Morse *1983+ attempt to explain the market reaction to firms’ dividend decisions. The first method is based on market returns surrounding announcements of increases in dividend and the second method is based on portfolios that are selected on dividend increases influencing the dividend payout ratio (dividends/earnings). They look at the response to a dividend announcement, conditional on dividend payout ratios. Results confirm that there is both a signaling effect and a clientele effect (taxation aspect) present. The signaling effect is harder to explain because of the complexity. Investors all have their own interpretation of the dividend announcement and announced earnings. The main conclusion is that firms which announced an increase in dividends generally had positive abnormal returns and that this is more pronounced when dividend payouts decrease. This implies that shareholders might think of the increase in earnings and decrease in payout ratio as good news signaling as a result of increased investment opportunities. This proves that earnings are important to investors when interpreting a dividend change. Divecha & Morse use the dividend payout ratio as a proxy for the dividend strategy of a firm but this contains noise because earnings changes might be only a short period of time. A survey Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

12

study performed by Brav, Graham, Harvey & Michaely [2005] found that managers do not think signaling through dividend is a costly signal. They do feel that dividend policy contains information valuable to shareholders and that the information communicated should be the same as information communicated through other communication mechanisms. Managers do not want to cut dividend because it might send a negative signal to shareholders and is only done when the manager absolutely has to. When liquidity problems for example are short term and dividend is cut, shareholders might interpret this signal wrong. It is possible that managers wait with the reduction of dividend until air cover is provided by competitors lowering dividends. Managers feel that there is an asymmetry between reactions to increases and decreases of dividend payment, with shareholders reacting more heavily to decreases in dividend payment. Results of Denis & Osobov [2008] cast doubts on signaling being a firstorder determinant of dividends. Firms which do not necessarily need to signal their profitability to shareholders (because they already have high earnings) appear to pay dividends. The question remains whether or not it is necessary to signal inside information to shareholders when a firm’s financial performance is great. 2.5.5 Clientele theory Questions exist about how much you should pay your shareholders and investors and what the valuemaximizing options are, but it is difficult to find a right answer to that. Shareholders can be roughly divided in two main groups: the first group consists of small individual investors combined with nonprofit organizations, and the second group of financial institutions. Smaller investors deal with large transaction costs and as a result do not trade so often as a result. Taxation is also an important determinant because capital gains are taxed differently from dividend. Usually dividend is taxed more heavily than capital gains. Asquith & Mullins [1983] regard dividend payments to be highly visible compared to other communications or announcements of the firm. Researching the impact of dividend initiation on shareholder wealth they find both a negative and positive wealth impact. The negative wealth impact results from costs associated with dividend payments, like transaction costs, administration costs, costs of initiating a dividend program and possibly issuing new equity to pay dividend. Another negative impact might be an enlarged tax burden on shareholders associated with receiving dividends as described by the clientele theory. A positive wealth impact occurs from signaling of superior information known to management towards shareholders through dividends. According to Asquith & Mullins [1983], this reduces constraints with institutionalized investors, provides a heterogeneous shareholder group and supplies benefits for investors who prefer periodical cash dividend. The main conclusion is that the negative wealth impact is more than offset by the positive wealth impacts investors place on being paid dividend.

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

13

Feldstein & Green [1983] find that investors want to achieve the highest market value possible and that this depends (partly) on taxation. When dividend is highly taxed, investors prefer capital gains and viceversa. The main conclusion is that the dividend policy is dependent on the tax situation and has to be adjusted to that. Brav, Graham, Harvey & Michaely [2005] do not agree with Feldstein & Green [1983]. This survey study gains insight in what managers think when deciding on dividend payout, share repurchases and retention of earnings. Managers say that tax consideration is not a dominant factor in the decision process. The choice of whether or not to pay dividend, increase or decrease in dividend and the choice between dividend payment and share repurchase is not dependent on the existing tax rates. Common thought is that smaller investors prefer dividend in spite of the tax burden. Managers also think that dividend payment is a significant factor for institutions in their decision-making process. Taxation and transaction cost are only a part of the clientele theory. This study uses a proxy to measure the relation between a crisis and the dividend payout ratio on the highest level. Transaction costs and taxation are not included as an individual variable. 2.5.6 Other determinants There are many more factors influencing dividend payout, some of which are discussed in short this paragraph. Gugler [2001] establishes that the ownership and control structure of a firm is a significant determinant of the dividend payout policy. State controlled firms differ from family controlled firms. The state controlled firms have higher dividend payout ratios, are less reluctant to cut dividends and are more likely to smooth dividend over time. Family controlled firms have significantly lower dividend payout targets, do not smooth dividend and are least reluctant to cut dividend. Owner-managers react to investment opportunities and adjust its dividend policy accordingly. Firm with less investment possibilities will smooth dividends more and will have higher payout ratios to get rid of cash. Dividends significantly negatively influence capital investment and (insignificantly) also R&D investments. Firms invest more in R&D when investment opportunities are good and will have lower dividend payout ratios as a result. The decision to pay dividend is considered part of the firm’s investment decisions. The results of Gugler are consistent with the “outcome model” of La Porta et. al. [2000], where corporate outsiders can force insiders to pay dividends (within their legal rights and where outsiders are protected by law). Dividends are in this case are an outcome of legal protection of shareholders. La Porta et. al. *2000+ tested two agency models of dividend. The second model was the “substitute model” where insiders are interested in establishing a reputation for the treatment of minority shareholders and dividends are a substitute for legal protection of shareholders. Conclusions are that minority shareholders who are protected by law receive higher dividends and are, on the other hand, more willing to wait for their dividend when investment opportunities are good, which is usually the case with fast growing firms.

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

14

Dickerson, Gibson & Tsakalotos [1998] investigated whether dividend payout is used as a defense mechanism against takeovers. There is evidence that a strategy to pay high dividends reduces the probability of a takeover. Firms deal with several investment decisions one of which is dividend payout. Firms can also decide to retain the earnings or to invest it in new projects. Investment in new projects is also significantly associated with lower takeover risk. The question is then, which decision is more effective when they both reduce the risk of a takeover. Results found (ceteris paribus) suggest that paying more dividends is more effective against the possibility of a takeover. A secondary explanation given is that firms use dividend to maintain shareholder loyalty to decrease the possibility of a takeover considering the active market for corporate control. Managers within a firm have influence on the dividend policy of a company. Lambert, Lanen & Larcker [1989] found that introducing an executive stock option plan causes an incentive for managers to decrease the level of corporate dividends as the value of the stock option would be less when more dividends are paid to shareholders. In this case managers can influence the dividend payout policy to work in its own advantage and increase the value of their own stock options. 2.6 Combing dividend payout policy and a financial crisis Fewer studies discuss the relationship between a crisis and the dividend policy of firms. Campello, Graham & Harvey [2009] wrote a survey paper on the corporate spending of firms having to deal with financial constraints during the crisis of 2008. The survey was taken in the midst of the financial crisis in the fall of 2008. The main conclusions of the study were that financially constrained firms in the US, on average, planned to dramatically reduce employment (by 11%), technology spending (by 22%), capital investment (by 9%), marketing expenditures (by 33%) and dividend payments (by 14%) in 2009. Unconstrained firms also planned cuts but they were, on average, significantly smaller. For Europe and Asia similar patterns are found. Evidence is found on the view that financially constrained firms build cash reserves as a buffer to credit shocks. They also pass up attractive investment opportunities due to difficulty in obtaining external financing. Tight credit markets also force some firms to sell assets to obtain cash. When looking at the variable size, larger firms planned bigger cuts in technological expenditures, while smaller firms cut more capital spending cuts, marketing expenditures and try to preserve cash. Consequently this means cuts in dividend payout, among other things. When comparing differences between speculative and investment grades (based on Standard & Poors credit ratings), speculative firms planned significant cuts in all expenditure categories. Also smaller cash reserves and greater dividend cuts were expected for 2009. Investment-grade firms planned cuts as well across all variables but smaller. Based on planned firm policies, financially constrained firms systematically planned to pay fewer dividends in upcoming years. In pre-crisis years planned dividend cuts were around 8%, but in crisis years the dividend is cut 28% showing the impact of the crisis on dividend payout. Point of criticism is that some measures for the crisis period are noisy, for example, small firms’ Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

15

dividend payouts increase during the crisis. The main points of attention in the model are that financially constrained firms already plan to reduce dividend payout, but a crisis enhances the cuts, the cash position of the firm worsens and that the firm sells assets to obtain cash. Campello, Graham & Harvey [2009] use the cash to asset ratio to assess the financial position of a firm. This study differs from Campello, Graham & Harvey’s by using hard financial data to support or reject hypotheses, where Campello, Graham & Harvey used survey data gathered from CFO’s which is more sensitive to bias. This study adopts Denis and Osobov’s model as a basis and adds several variables to be tested relating to financial constraints, signaling, clientele and liquidity.

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

16

3. Research question & hypotheses From the literature we learn that a lot of factors influence dividend policy decisions like firm size, financial performance, liquidity, aspects relating to signaling and clientele. This study will focus on hard data (financial performance and liquidity) and tries to determine whether or not firms use a financial crisis to adjust dividend policy. The research question is: “Does a firm change its dividend policy when dealing with a financial crisis?” How do managers respond to a financial crisis and do they use this event to change the dividend policy of the firm? Do they adjust dividend payments downwards as earnings are under pressure, or do managers continue to pay shareholders dividend despite the fact that earnings are down. What kind of role does liquidity play in the choice to change dividend policy? Financial data is used to establish a relationship between a financial crisis (2008 credit crisis) and dividend policy. An explanation of definitions of each variable is discussed in the next section. The research question can be decomposed in several hypotheses. The first hypothesis explores the effect of a crisis on the dividend payout (ratio). As prior research has shown (Campello et al. 2009), I expect that the financial crisis will lead firms to decrease their dividend payouts due to financial constraints. The first hypothesis is: H1: A financial crisis is associated with a decrease in dividend payout (ratio) As discussed earlier, Redding [1997] discovered that firm size and liquidity affects the decision whether to pay dividends or not. In this study the firm size and liquidity will also be used to determine whether the dividend payout (ratio) is affected by a financial crisis. When earnings decline and the payout stays the same, the ratio would increase. It is argued that larger firms tend to be more liquid because of the higher demand for shares and the tendency to pay more dividends. From the life-cycle theory point of view, firms pay out assets to shareholders upon reaching an efficient firm size, which is preferred by institutional investors. As the firm matures, the costs of retention increase resulting in a higher propensity of paying dividends. Based on Redding’s findings, the ratio of smaller firms is more easily affected by changes in earnings. I expect there to be cross-sectional differences resulting from a financial crisis on the dividend payout ratio. Based on Campello et. al. (2009), I expect firms which already have a low liquidity ratio to reduce dividend earlier than firms which have a better liquidity ratio. This results in the next hypotheses: H2: The dividend payout (ratio) of smaller firms decreases more during a crisis compared to larger firms H3: The dividend payout (ratio) decreases more during a crisis at firms with a low liquidity opposed to firms with a high liquidity

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

17

Both signaling and clientele could play an important part during a crisis. If we include Mishkins definition of a financial crisis that information asymmetry problems become much worse during a crisis, we could argue that managers face difficult decisions relating to dividend payout. Signaling theory posits that dividend can be considered an information source for shareholders. In this case a manager could consider lowering dividend payment as too costly and would try to avoid this. This could mean that a financial crisis would not have a strong effect on dividend payout (ratio) for firms where dividends serve a signaling role. This assumption is translated into hypothesis four. Also, clientele theory posits that a certain type of investor base (small investors) desires a steady stream of dividends. It is expected that a manager is reluctant to cut back on dividend payout as a result. He/she will explore other options first before reducing dividend payment to shareholders. Therefore, when dividends are more important to investors of a firm, one might observe a smaller impact of the crisis on dividend payout (ratios) than for firms where dividends are not as important to the investor base. This is summarized in the fifth hypothesis. The fourth and fifth hypotheses are: H4: Dividend payout (ratios) decrease less during a financial crisis when dividends serve a signaling role H5: Dividend payout (ratios) decrease less during a financial crisis when firms have more small investors

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

18

4. Methodology This section describes the methodology used in this study. First the model is described followed by the definitions of the different variables. Then the collection process of the sample is discussed including elimination of unusable data. 4.1 Model It has been shown that a financial crisis has an effect on the financial performance of a firm. Literature also suggests several relationships between (first-order) determinants like signaling, clientele, firm size and liquidity with dividend policy. What has not been shown is what kind of changes occur in dividend policy when a firm deals with a financial crisis. This study focuses on the dividend side of shareholder gains/losses and does not include capital gains/losses resulting from the credit crisis. I examine whether the financial crisis of 2008 affects dividend policies, and whether the relationship is moderated by determinants of dividend policy found in prior research. The usage of hard financial data is considered to be less biased opposed to usage of soft data like surveys as conducted by Campello, Graham & Harvey [2009]. Figure 2 represents the model used in this study.

Research model 2008 Credit Crisis

Dividend payout (ratio)

Moderator

Size Liquidity Signaling Clientele

Moderators measured by: - Size: Total assets - Liquidity (quick ratio, current ratio) - Signaling: Bid/Ask spread - Clientele: Investor composition

Figure 2: Research model

A simple linear regression analysis is used to establish the relationship between each individual independent variable and the dependent variables dividend payout ratio (DPR) and dividend payment of firms. A multiple linear regression analysis is used to establish the validity of the whole model using these independent variables. Possible correlations between independent variables will be analyzed using Pearson’s correlation coefficient. The model of Denis & Osobov [2008] on dividend payments is used as a basis as important determinants of dividend payments (like profitability, growth opportunities, firm size and earned equity) are already included. Earlier research concluded that during a financial crisis these aspects are affected negatively. Added to this model are measures of liquidity, signaling and clientele as well as a dummy for crisis years. Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

19

The main regression formula used in this study is: DPR =

0+ 1(P)+ 2(GO)+ 3(EE)+ 4(FS)+ 5(L)+ 6(S)+ 7(C)+

8(CRI)+εi

Where: DPR

= Dividend payout ratio

P

= Profitability

GO

= Growth opportunities

EE

= Earned equity

L

= Liquidity

S

= Signaling

C

= Clientele

CRI

= Dummy variable crisis (Dummy is 1 for years 2008-2010)

ε

= Error term

To examine hypotheses 2-5, the main regression is run separately for subsamples partitioned at the median of the variables of interest. For instance, to test hypothesis 2, I separate the sample into smaller and larger firms at the median and examine whether the coefficient on crisis differs between the two groups. The same is done for the other three hypotheses, using liquidity, bid-ask spreads and investor base to form the subsamples. DPR includes total earnings in the ratio which could be noisy. The third regression checks the influence of the variables on dividend payout (DP) instead of DPR. This regression is also run for the different subsamples as explained above. The additional regression formula used in this study is: DP =

0+ 1(P)+ 2(GO)+ 3(EE)+ 4(FS)+ 5(L)+ 6(S)+ 7(C)+

8(CRI)+εi

4.2 Definitions The dependent variables in this model are the dividend payout ratio (DPR) and dividend payout (DP), both proxies for dividend policy. The dividend payout ratio divides dividend paid out to shareholders by net earnings of a firm. The dividend payout ratio consists of a nominator (total dividend) and a denominator (total earnings). That could introduce some noise into the model as large earnings changes can be of a transitory nature. To be more complete, this study includes total dividend payout as well. The formula of DPR is:

The model of Denis & Osobov [2008] included profitability, growth opportunities, firm size and earned equity. This study uses the same composition and definitions of the variables of Denis & Osobov. Profitability (Et/At) is defined as the ratio of earnings before interest but after tax to the book value of Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

20

total assets. Growth opportunities (Vt/At) is measured as the ratio of the market value of the firm to the book value of total assets. Size is a variable which can be measured in more than one way. Size can be measured using sales, number of employees and total assets. Denis & Osobov used total assets in millions (At) in their model and so will this study. Finally the last variable of Denis & Osobov’s model is earned equity (REt/BEt). It measures the ratio of retained earnings to total book value of equity. The liquidity aspect can be explained using different formulas. Lower liquidity means that there is less money available for operational purposes. Current ratio, quick ratio and operating cash flow ratio can be used to calculate liquidity. Current ratio is calculated by dividing total current assets by total current liabilities. Quick ratio differs from that by decreasing total current assets with inventory which makes quick ratio a more conservative measure. Operating cash flow ratio uses the cash flow from operations divided by total current liabilities because cash is used to pay debt. The choice between the three options is a difficult one. Literature identifies a lot of effects on a firm due to a financial crisis. One of which is decreasing sales. Decreasing sales could also mean an excess in inventory. Changes in inventory go hand in hand with a crisis, meaning quick ratio could weaken in the model compared to current ratio which includes inventory. Operating cash flow consists of several components and is more difficult to gather. For one, the sample lacks several important figures to calculate operating cash flow ratio. An initial analyses of quick ratio and current ratio shows that both variables are highly correlated (Pearson correlation is close to 1). Therefore both can be used in the model and it will not matter which one is used when doing the regression. In this study quick ratio is used. The signaling theory of dividends state that dividends can be seen as a communicative device to provide shareholders with inside information regarding current and expected future performance. There are two main possibilities to measure signaling. The first one is described by Maury & Pajuste [2002] and uses cash flow and control rights to measure signaling. Maury & Pajuste use voting rights and cash flow decision rights to determine how much shareholders control the firm. When a group of shareholders have a lot of influence, they can also control dividend payout policy resulting in a weakened separation of ownership and control. The dataset collected lacks this specific data to analyze the variable using this proxy. The second possibility is to use the bid-ask spread as a proxy. The bid-ask spread indicates the difference between what shareholders wish to pay and what they want to receive for shares of firms. The smaller the difference, the better (inside) information is being relayed to the market. A bigger difference indicates that less information is available and leaves room for speculation. Venkatesh & Chiang [1986] use earnings and dividend announcements to determine how the information asymmetry is affected when made public. The bid-ask spread is used as a proxy to measure the information asymmetry. On average there is a normal information asymmetry before earnings and dividend announcements. When traders and investors suspect an upcoming and irregular announcement the information asymmetry increases, resulting in a widening of the bid-ask spread as investors try to offset Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

21

expected losses. Howe & Lin [1992] also explore the information asymmetry and conclude that there is evidence supporting the hypothesis that payment of dividend conveys information to the market and influences the bid-ask spread as there is a positive relationship between the level of information asymmetry and the magnitude of the spread. For the variable signaling, the bid-ask spread is used as a measure. The variable clientele identifies what kind of investors a firm has. This is divided into two groups. The first group contains smaller investors like individual investors and non-profit investors. The second group contains large financial institutions. In this study, the percentage of ownership by large institutions is used. Previous literature suggests several differences between smaller investors and financial institutions. Smaller investors deal with larger transaction costs and are more long term focused. Taxation of dividend and capital gains changes the preferences of the investors. As these factors influence shareholder preferences, Brav, Graham, Harvey & Michaely [2005] suggest that in spite of a higher tax burden, small investors prefer periodical dividend. Managers also do not want to changes dividends during a crisis. Assuming that smaller investors prefer dividend, the DPR during a time of crisis should decrease less when a firm has more small investors. It is important to answer the main question when did the crisis start? According to Campello, Graham & Harvey [2009] the financial crisis reached its peak in the fall of 2008. Financial performance showed that 2008 and the upcoming year 2009 would be very difficult years for firms. As Uitdewilligen [2010] described, all of Mishkin’s requirements were met in 2008. This will also be the starting point of this study. The basic idea is to compare data from pre-crisis years (2005-2007 and dummy value 0) with data in the years of crisis (2008-2010 and dummy value 1). For a better analysis, groups are created to examine hypotheses 2-5. The regression is run separately for each subsample partitioned at the median of the variables of interest. In total, four main variables are divided into groups: firm size, liquidity, signaling and clientele. 4.3 Sample and data collection The data set will be collected using Compustat North America database, which contains mainly American and Canadian firms. Only US firms are used in the analysis process. All firms use US GAAP as reporting standard and US dollar as reporting currency. This takes care of the geographical region, accounting standard and currency selection. For the collection of investor information the Thomson – Reuters investor database is used.

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

22

The data will be subject to elimination of a few SIC-codes like utility firms (SIC codes 4900-4949) and financial firms (SIC codes 6000-6999). Also negative or zero values which are illogical (like zero assets or negative sales). Additionally lack of data which in effect prevents a calculation of a proxy, will be eliminated. Table 1 shows the sample selection and eliminations. Without any elimination, the North American sample contains 10.465 unique firms and the total size of the sample is 65.697. The sample contains 8.801 (84,1%) American firms and 1.664 (15,9%) Canadian firms. The sample also contains 3.152 (30,1%) financial firms and 314 (3%) utility firms and which will be eliminated before the initial analysis of descriptive statistics. Table 1: Sample selection Total observations North America (2005-2010) Less: Missing year or missing CUSIP

65.697 (16)

Less: Canadian firms

(10.149)

Less: Utility firms SIC 4900-4949

(1.785)

Less: Financial firms SIC 6000-6999

(16.850)

Observations after elimination

36.897

After eliminating missing years, missing CUSIP number (used for matching), Canadian firms, utility firms and financial firms the total sample size is 36.897. The distribution over the years has changed due to the elimination. An analysis of the sample shows an almost equal distribution of data over the years 2005-2010 (table 2). Table 2: Distribution of sample across years Year 2005

Frequency 6.969

Percent 18,9

2006

6.731

18,2

2007

6.382

17,3

2008

5.991

16,2

2009

5.662

15,3

2010

5.162

14,0

Total

36.897

100,0

Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

23

5. Results The previous section described what kind of research method is used in this study, defined the variables and described the collection process of the sample. This section takes a look at the descriptive statistics of the refined sample, discusses the relation of individual variables to DPR and DP and the results of the main regressions. Alternatively, another approach is used by running the regression again with the subsamples. 5.1 Descriptive statistics After the initial elimination, the descriptive statistics show large standard deviations opposed to the mean of each variable. The spread of most variables is quite big, especially profitability, both liquidity ratios and the bid-ask spread. Refining the sample should provide better regression results. In this study, winsorising is preferred over trimming the results as it does not destroy possible valuable information but simply replaces it with values which are more logical and present in the sample. The extreme values have been adjusted to values at the 2% and 98% percentiles. Table 3 shows the new descriptive statistics and there have been some changes regarding decreasing means of firm size and liquidity. Of course the spread of each variable decreased resulting in a decrease in standard deviation. Table 3: Winsorized descriptive statistics Variables

N

Minimum

Maximum

Median

Mean

Std. Deviation

DPR

34.785

-344,05

2.218,71

0,00

0,19

14,15

DP

34.742

0,00

955,00

0,00

87,83

547,30

Profitability

35.319

-5,08

0,31

0,05

-0,21

0,89

Growth opportunities

27.777

0,00

11,36

0,84

1,46

2,10

Earned equity

34.606

-35,38

35,83

0,30

-0,14

9,17

Firm size

35.664

0,14

40.424,14

245,27

2.918,06

7.521,90

Liquidity (quick ratio)

34.724

0,02

15,24

1,32

2,25

2,82

Signaling

32.082

0,06

50,70

6,58

10,19

11,00

Clientele

27.828

0,00

1,00

0,32

0,40

0,37

Crisis (dummy)

12.581

0,00

1,00

0,00

0,45

0,50

Note: Data has been winsorized at the 2% and 98% percentiles.

To gain more insight into the sample, descriptive statistics of crisis years and non-crisis years are compared in table 4. For the variables taken from the model of Denis & Osobov, profitability, growth opportunities and earned equity, little to no change appears to be present between crisis and pre-crisis years. As for firm size, there is a difference between non-crisis years and crisis years. Firms seem to have more assets available, but the standard deviation of the sample is also higher. Firms also seem to be slightly less liquid in times of crisis and the bid-ask spreads increase a bit during a crisis. As for the variable clientele, the shares owned by large investors increases a bit from 37% to 43% owned. DPR seems to be larger for pre-crisis years as the mean in crisis years is almost two times smaller compared to pre-crisis years. Remon Smits | Effect of a financial crisis on the dividend payout policy of a firm

24

Table 4: Comparison non-crisis and crisis descriptive statistics 2005-2007

2008-2010

N

Minimum

Maximum

Mean

Std. Deviation

DPR

18.481

-344,05

2.218,71

0,26

19,27

16.304

DP

18.584

0,00

955,00

78,30

545,05

16.158

0,00

955,00

98,80

549,68

Profitability

18.801

-5,08

0,31

-0,21

0,88

16.518

-5,08

0,31

-0,20

0,90

Growth opportunities

15.208

0,00

11,36

1,56

2,19

12.569

0,00

11,36

1,34

1,96

Earned equity

18.462

-35,38

35,83

-0,15

8,98

16.144

-35,38

35,83

-0,12

9,38

Firm size

19.028

0,14

40.424,14

2.612,84

7.022,18

16.636

0,14

40.424,14

3.267,16

8.041,58

Liquidity

18.511

0,02

15,24

2,30

2,92

16.213

0,02

15,24

2,18

2,70

Signaling (bid-ask spread)

17.176

0,06

50,70

9,33

9,95

14.906

0,06

50,70

11,17

12,02

Clientele (% institutional investor ownership)

15.247

0,00

1,00

0,37

0,37

12.581

0,00

1,00

0,43

0,36

Variables

N

Minimum

Maximum

Mean

Std. Deviation

-89,73

167,24

0,11

2,52

5.2 Simple linear regression This section will show how each variable separately relates to dividend payout ratio and dividend payout (table 5). This will provide a first look at how significant the independent variables are. Only the variable signaling (t=2,039, sign. 0,041) is significant (using α=0,05). The remaining variables are not significant at α=0,05. Even if α=0,10 is used, the variables remain not significant. Using dividend payout, every single one of the variables is significant, even if α=0,01 is used. The variable crisis is significant but also positive which is somewhat odd considering a crisis normally has a negative influence on a firms' financial performance. A overview including an analysis of simple regression using subsamples available in Appendix B. Table 5: Simple linear regression results Dependent Variable DPR Independent Variable

Dependent Variable DP

t-statistic

Significance

t-statistic

Significance

Profitability

,817

,414

11,184

,000*

Growth opportunities

,070

,944

6,788

,000*

Earned equity

,191

,848

2,695

,007*

Firm size

1,131

,258

124,597

,000*

Liquidity

-,970

,332

-12,081

,000*

Signaling

2,039

,041**

25,430

,000*

Clientele

-,910

,363

-5,305

,000*

Crisis

-,968

,333

3,482

,000*

Note: Test taken individually. This represents a simple linear regression with DPR and DP being the dependent variables. Significance at * p