The influence of a financial crisis on the dividend policy of firms

The influence of a financial crisis on the

dividend policy of firms

06/2013, Amsterdam

Nikki Paes

6152937

The influence of a financial crisis on the dividend

policy of firms

Name:

NJM Paes

Studentnr.:

6152937

Supervisor:

T. Homar

Date:

08/04/2013

Bachelor:

Business Economics, Finance and Organization

University of Amsterdam.

Abstract

This study is about the influence of a financial crisis on the dividend policy of firms,

specific the financial crisis of 2008. The model measures the influence of a financial

crisis on the dividend payout ratio, controlling for important firm characteristics.

Results of the model show that overall the financial crisis does not affect the payout

ratio. They do show that when the signaling effect is high, the dividend payout ratio

increases during the crisis. This shows that managers consider their shareholders when

deciding on dividend payouts. An increase in the dividend payout ratio signals to the

shareholders that the firm is able to overcome the financial crisis. Therefore they might

want to prevent the shareholders from selling their shares in the firm. The results also

show the importance of profitability and size in determining the dividend policy of firms.

Preface

This bachelor thesis is part of my study in Business Economics, Finance and

Organization at the University of Amsterdam. The last few months I worked on this

thesis about the influence of a crisis on the dividend policy of a firm. By writing this

thesis, I have learned a lot about the crisis and the behaviour of firms. I experienced

some trouble when building the research model, but in the end it worked out well.

Therefore I would like to thank my supervisor Timotej Homar at the University of

Amsterdam, for his help and advices during the writing of this thesis.

Nikki Paes

Table of contents

Abstract ... 3

Preface ... 4

1. Introduction ... 6

2. Literature Review ... 7

2.1.1 Why do firms pay dividends? ... 7

2.1.2 Signalling Effect ... 9

2.2.1 What is a financial crisis? ... 10

2.2.2 Cause of the financial crisis? ... 10

2.3 Dividend policy in a financial crisis ... 12

3. Research question and hypotheses ... 13

4. Methodology ... 15 4.1 Model ... 15 4.2 Data ... 18 5. Results ... 18 5.1 Descriptive statistics ... 18 5.2 Regression results ... 20

5.2.1 Dividend payout ratio ... 20

5.2.2 Dividend Growth ... 23

6. Results Analysis ... 25

6.1 Dividend payout ratio ... 25

6.2 Dividend Growth... 26

7. Discussion ... 27

Reference list ... 29

1 Introduction.

This study measures the effect of a financial crisis on the dividend policy of a firm. The most recent financial crisis started in 2007. According to a study by Shiller (2008), this financial crisis was caused by the housing market in the US. Because of increasing house prices, homeowners started to take more mortgages. But when house prices began to fall, these mortgages became overvalued. Financial institutions had to absorb these losses, resulting in a widespread bank distress. The crisis first started in the US, but eventually Asia and Europe were also hit. Financial institutions and other firms around the world became financially constraint. To deal with the effects of a financial crisis, firms have to adjust their financial policy. There are several studies that describe the causes of the financial crisis, and what should have been done to prevent the effects of a financial crisis. This study will take a closer look at the effect of a financial crisis on the payout policy of firms, specific about dividends. The research question is:

To what extent does a firm changes its dividend policy when facing a financial crisis? There is an abundance of studies about the payout policy of firms. These studies examine why firms pay dividends, how they should combine dividends and repurchases or how shareholders react to certain payout policies. But there is not a lot written about the influence of a financial crisis on these findings. There are some studies that show the impact of a crisis on the firm’s financial policy. Campello, Graham &Harvey (2009) show how firms planned to cut

expenditures and how they adjusted their financial policy in financial crisis. Floyd, Li and Skinner (2011) show evidence on financial payout policies of firms during crisis. They show the reluctance of managers to cut dividends during crisis, and the consistency of the dividend payout ratio.

This study focusses on the dividend policy of firms in crisis years. Where earlier research described the most important determinants for the payout policy of a firm, this study tries to understand what determines the dividend policy changes during a financial crisis. Expected is that firms decrease their dividends during a financial crisis, and thereby lowering the dividend payout ratio. The characteristics studied are taken from an article by Mitton(2004), investment opportunities, profitability and size. Hypothesized is that due to a decreasing number of

investment opportunities, the firm’s dividend payout ratio increases. And the dividend payout ratio of less profitable and smaller firms will decrease more compared to more profitable and bigger firms. The signalling effect will also be included, which is expected to be an important factor in determining the dividend policy. Hypothesized is that the dividend payout ratio decreases less, when dividends are clear signals.

To test these effects, several multiple regressions are done on the dividend payout ratio. The regressions either use fixed effects estimation or random effects estimation. To exclude the effect of earnings, these regressions are also done on dividend growth. Both regressions provide only a few significant results. Most results are unexpected. The variable crisis is not significant when testing the dividend payout ratio. Therefore no clear evidence is found for the effect of a crisis on the dividend payout ratio of a firm. The model does show the importance of

profitability and size in determining the dividend policy. When controlling for these variables in testing the relationship of crisis and dividend payout ratio, only the signaling effect has a

significant effect. The regression on dividend growth eliminates the effect of earnings, but still does not show an effect of crisis on dividend policy.

This study starts with describing literature related to dividend policies and financial crisis in chapter 2. Chapter 3 describes the research question and the related hypotheses. Chapter 4 describes the model and the used methodology. And in chapter 5 the results are presented and analysed. Chapter 6 analyses the results of chapter 5. And at last, chapter 7 discusses the findings of this study and gives the conclusion.

2 Literature review

2.1.1 Why do firms pay dividends?

An important question is why firms pay dividends at all? What influences the dividend policy of a firm? One of the most important theories is the irrelevance theorem of Miller and Modigliani (1961). They started with the assumption that there are no market imperfections and concluded that the dividend policy a firm chooses does not change the value of the firm. An important proposition is: In perfect capital markets, holding fixed the investment policy of a firm, the firm’s

choice of dividend policy is irrelevant and does not affect the initial share price. And therefore it

does not matter whether they payout in dividends or share repurchases, because when a firm increases its dividends, the terminal value of the shares decreases with that amount. This theory is heavily criticized by many studies, because most of them assume that there are market

imperfections and that these imperfections do affect the firm value.

The findings of a well-known study by Black, Fischer and Scholes (1974) are in line with the irrelevance theorem. They tried to identify the effect of dividend policy on share prices by examining the relationship between dividend yield and stock returns. They showed that share prices were not influenced by the dividend yield or the proportion of earnings paid out in the

form of dividends. This suggests that market imperfections have no effect. However, they did test the tax effect hypothesis, and assumed that there could be a clientele effect when the imperfections generated different dividend preferences.

Miller and Modigliani (1961) also assume that the free cash flow of a firm determines the level of payout that can be made to its shareholders. An important related proposition is: In perfect

capital markets, if a firm invests excess cash flows in financial securities, the firm’s choice of payout versus retention is irrelevant and does not affect the initial value of the firm. They assume that the

choice of the manager about payout is irrelevant, because all capital structures and dividend policies provide the same stockholder value. DeAngelo and DeAngelo (2006) disagreed with this assumption. When retention is allowed, a firm can pay out less than 100% of free cash flow, and thereby reducing its value. This shows that dividend policy does matter, and investment policy is not the sole determinant of value. Also a manager has more than one payout policy to choose from to optimize the distribution of free cash flow. With each policy a different amount of free cash flow can be paid out. Therefore the choice the manager makes is relevant. This does not meet the Miller and Modigliani assumption about the irrelevance of the choice of payout policy. DeAngelo and DeAngelo (2006) introduced a new view on payout policy. Their main conclusion is that the optimal payout policy is driven by the need to distribute the firm’s free cash flow. The dividends paid depend on the life-cycle theory, which states that in the early year’s firms have more investment opportunities than internally generated capital. As the firm matures, the firm’s internal capital exceeds the investment opportunities, and pays out more dividends in order to spend their excess cash well. Denis & Osobov (2008) findings to the propensity to pay dividends are consistent with this life cycle-based theory of dividend policy. They found a strong positive association between earned/contributed equity mix and the propensity to pay dividends.

Their

results show that a firm’s propensity to pay dividends is high when earned/contributed equity

is high and that it is low when earned/contributed equity is negative.

dividends are concentrated among firms with the greatest earnings, which shows that the distribution of free cash flows is the primary determinant of dividend policies.

Black’s (1976) findings were also in line with the leverage analysis of Miller and Modigliani. Black took a closer look to all the factors influencing policy decisions and tried to find an answer to the question why dividends are paid at all. This he called the ‘Dividend Puzzle’. It is important that firms act out the optimal payout policy. They have to consider dividends, retention and share repurchases. These might be affected by factors as transaction costs, investor’s demand and most important, taxes.

Black argued that when there are taxes on payouts, firms should distribute little or no cash to shareholders. This is in line with Miller and Modigliani, in a perfect capital market, with no taxes, all leverage decisions lead to the same shareholder wealth and thus the same firm value. And when corporate income is taxed, the optimal decision is to have maximum leverage. This leads to Black’s argument that in a world where dividends are taxed more heavily (for most investors) than capital gains, and where capital gains are not taxed until realized, a corporation that pays no dividends will be more attractive to taxable individual investors than a similar corporation that pays dividends (Black, 1976). This might increase the price of the non-dividend-paying corporation’s stock. Many corporations will be tempted to eliminate dividend payments.

2.1.2 Signaling effect.

Considering the Dividend Puzzle, an important effect on dividend policy will be discussed, the signaling effect.

An important market imperfection is the existence of asymmetric information. This is based on the agency theory, which states that there is a conflict of interest between the agent and the principal. The management of the firm being the agent, and the shareholders being the principal. The agent knows more than the principal, and they both have different interests. Therefore the principal is not sure if the agent is acting in the principal’s best interest or in his own best interests. In this case, dividends can provide shareholders with additional information and act as an indicator of how well the firm is performing. According to an article by DeAngelo,

DeAngelo, and Skinner’s (2004) dividends are concentrated among the largest, most profitable firms. This shows that signaling might not be a first-order determinant, because precisely those firms that don’t really need to signal their profitability appear to be paying dividends. Small, less profitable firms, who are most in need of signaling, were paying fewer dividends (Denis and Osobov, 2008).

A survey done by Brav, Gram,Harvey & Michaely (2005) showed that executives feel that dividend policy contains information that is valuable to shareholders and that the policy should be consistent with other forms of communication. The survey also showed that executives don’t want to reduce dividends, because that might send a negative signal. Also managers assume that there is an asymmetry between dividend increases and decreases, and shareholders react more heavily to dividend decreases. Therefore, some firms wait with decreasing their dividend, until competitors reduce their dividends.

2.2.2 What is a financial crisis?

But what exactly is a crisis, and what factors are important in causing the crisis? According to a study by Mishkin (1996) a financial crisis is a nonlinear disruption to 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. This means that financial crises lead to inefficient functioning markets and economic activity. In his study he outlines the role of asymmetric information in the financial system, and why this is so important to the economy. He also developed a theory for financial crisis using the asymmetric information framework. According to Mishkin (1996) the factors leading to financial crises are: increasing interest rates, increasing uncertainty, asset market effects on balance sheets and bank panics.

Increasing interest rates lead to more adverse selection, because those firms with the most risky investment projects are more willing to pay high interest rates. And as with increase in

uncertainty, adverse selection becomes even greater, because it is harder to see what the bad and what the good credit risks are. This results in lenders becoming less willing to lend, leading to a decline in lending, investment and economic activity. The asset market effects on the balance sheet result in a decline of net worth and worse adverse selection and moral hazard problems. Bank panics reduce the financial transactions of banks. Because when depositors do not know the safety of their deposits, they withdraw them from the bank. This leads to a decline in lending. Both the asset market effects and bank panics decrease investments and economic activity (Mishkin, 1996).

2.2.1 Cause of the financial crisis

This study measures the effect of a financial crisis on the dividend policy of a firm. A crisis affects the behaviour of firms and people. For firms to react in the best way, it is important to know what exactly went wrong. And what were important factors leading to the financial crisis. In the years before the start of the crisis, there were some interesting developments in the housing sector. In the period of 1995 until 2005 the value of real estate and homeownership in the US were increasing (Figure 1). The increasing prices and homeownership rate created high expectations when investing in real estate. And because of the increase in the value of their real estate, homeowners started to increase their mortgages. In this same period there was an interesting development. The securitization of assets became a very often used instrument by investors. Securitization is the process where a financial institution combines different financial assets and sells them to investors.

Homeowners increased their mortgages by borrowing more money from banks with their home as collateral for the loan. The bank

combines mortgages in one large pool, and divides this large pool into smaller pieces with

different risk profiles. These smaller pieces are then sold to investors through a SPE, a special purpose entity. According to Gorton and Souleles (2005), a SPE is a legal entity created by a firm by transferring assets to the SPE, to carry out some specific purpose, circumscribed activity or a series of such transactions. SPEs have no purpose other than the transaction for which they were created. The investors buying financial instruments from the SPE receive very liquid instruments. And because of these SPEs the financial institutions reduce the risk of the mortgages and are able to make more loans.

Due to this rapid increase in securitization of mortgage loans, some rumours came up about an upcoming housing bubble. The combination of increasing value of real estate, and the relatively small population and income growth in the US leaded to these rumours. Kindleberger(1987) defined a bubble as a sharp rise in price of an asset or a range of assets in a continuous process, with the initial rise generating expectations of further rises and attracting new buyers—

generally speculators interested in profits from trading in the asset rather than its use or

earning capacity. The rise is usually followed by a reversal of expectations and a sharp decline in price often resulting in financial crisis. This definition consists of many of the features of the crisis in 2008: rapidly rising prices, focus in future price increases, and eventually decreasing prices. Because the market price of real estate had risen a lot in the years before the bubble, homeowners increased their mortgages. But eventually the market prices of real estate had risen too far above the fundamental value of the real estate, and the market prices rapidly dropped. The mortgages became overvalued, and homeowners could not pay their debts anymore, because they already spend all the money received. The value of the securitized mortgages decreased, and banks and investors tried to sell them. In the following period, financial institutions got into trouble due to liquidity problems as a result of the falling house prices. Some financial institutions even went bankrupt (Baily and Elliot, 2009). According to Shiller (2008), ‘The housing bubble was a major cause, if not the cause, of the subprime crisis and of the broader economic crisis...’ First financial institutions and investors got into trouble that had directly traded in the securitized mortgage bonds. But eventually investors around the world were hit by this housing bubble. This was caused by the rapid increase in securitized assets. Before, mortgages were the domain of the traditional banking system. But due to the

securitization, the mortgages could be traded in open markets within the US and outside the US borders.

2.3 Dividend policy in a financial crisis.

Floyd, Li and Skinner (2011) wrote a study about the payout policy of firms through the

financial crisis. They show how share repurchases and dividends were affected and the effect of taxes on payout policy. They show evidence on the payout policy of firms in the past 30 years, including the financial crisis of 2008. The conclusion of their study was that dividends paid by industrial firms were strongly persistent through the crisis, because managers are very reluctant to cut or stop paying dividends. In the years leading up to the crisis firms strongly increased their cash payouts, driven by an increase in earnings and the extent to which those earnings are paid out. During the crisis they continued to pay dividends. Evidence was found by the number of dividend payers from 2007 to 2009, this number declined only for around 1% of all US firms. And aggregate dividends declined by only 2% over the same period (Figure 2). There are even some companies who continued to increase their dividends.

They show that the dividend payout ratio for industrial firms was very consistent over the last 30 years (Figure 2). The ratio varies around 30%, even during the crisis. This stability shows evidence of the reluctance of industrial firms to cut dividends. Because the dividend payout ratio varies around 30% before the crisis, the large increase in payouts through 2007 is caused by a large increase in repurchases. During the crisis industrials cut repurchases sharply, but this

Figure2: Dividend payout ratio(blue) and total payout ratio(red) for US Industrials 1980-2010 (Floyd et al.2011)

Note: Percentage of dividend payout ratio is common dividend divided by income before extraordinary items. Percentage of total payout ratio is total payout divided by income before extraordinay items.

did not affect the payout ratio, because they continued to pay dividends. Both dividend and repurchases rebound strongly in 2010.

To study the corporate spending of firms during the financial crisis, Campello, Graham and Harvey (2009) surveyed CFO’s of firms in the US, Europe and Asia. They studied the corporate spending of financially constrained firms during the financial crisis. Their main conclusion is that due to the financial constraints, firms planned to cut in technological spending, capital spending, marketing spending, employment and dividend payments. However, most of these plans were enhanced during the crisis. Planned cuts in dividend payments before the start of the crisis were around 14%, but during the crisis dividend payments were cut by 28%. In crisis years it was difficult to gain some extra cash for firms, because it was more difficult to borrow externally. Therefore they were restricted to invest in attractive investment opportunities. And due to these credit constraints during the crisis, firms also had to sell more assets, in order to fund their operations. In their study they matched firms based on size, profitability and growth prospects. Concluding that smaller firms planned bigger cuts in dividend payments and that the bypassing of positive NPV projects reduces the strength of future economic growth (Campello, Graham and Harvey, 2009).

3 Research question and hypothesis.

This study will look at the effects of a financial crisis on dividend policy. Some important factors influencing the dividend policy will be included in the model; investment opportunity, size and profitability. The research question is:

To what extent does a firm changes its dividend policy when facing a financial crisis?

A crisis affects the behavior of firms and people. Firms have to react on the changing markets by revising their financial policy. But do they actually change the dividend policy of the firm? Expected is that there will be a decrease in the dividend payout ratio, due to the fact that there is less cash available for firms to payout dividends. An earlier study by Floyd, Li and Skinner (2011) showed that not all kind of firms decreased their dividends during a financial crisis, because a lot of managers do not want to cut or drop dividends. However, most firms cannot overcome the financial constraints, as stated by Campello, Graham and Harvey (2010), and do change their dividend payouts during a financial crisis.

- First Hypothesis, H1: Firms decrease the dividend payout ratio during a financial crisis.

According to Mitton (2004) the main factors influencing the dividend payout ratio are investment opportunities, profitability and firm size. Because of the decreasing number of positive investment opportunities available, firms could be better off by increasing their dividend payout instead of making risky investments. This might lead to a small increase in the dividend payout ratio.

- Second Hypothesis, H2: Dividend payout ratio increases a little, due to a decreasing

number of investment opportunities.

According to Aivazian, Booth and Clearly (2006) the probability of a firm paying a

dividend increases with the firm’s profitability. Therefore I expect firms that are less profitable to decrease their dividend payout more.

- Third Hypothesis, H3: Dividend payout ratio decreases more, due to a decrease in the

profitability of firms.

And due to the assumption that the dividend payout ratio is more affected by changes in earnings for smaller firms, I expect the dividend payout ratio of smaller firms to decrease more compared to larger firms. Also because larger firms tend to pay more dividends because of a higher liquidity rate (Denis and Osobov, 2008).

- Fourth Hypothesis, H4: Dividend payout ratio of smaller firms decreases more compared

to larger firms.

Some market imperfections, like taxes and asymmetric information, appear to have a lot of effect on dividends during a crisis. I expect the asymmetric information to become worse during the crisis, causing an increase in the effect of signaling. The role of dividends will become more important, as managers have to consider the signaling effect of dividend announcements. When dividends are clear signals, managers are more reluctant to reduce the payments even during a financial crisis.

Fifth Hypothesis, H5: Dividend payout ratio decreases less when dividends are clear signals.

4 Methodology

4.1 Model

The literature showed that financial crises are related to the financial policy of firms. This study will measure the effects of the financial crisis of 2008 on dividend policy of firms. Prior research showed some important determinants of dividend policies, which will be included in the model. A multiple regression model will be used to test the effect of the crisis. The Dividend Payout Ratio (Dpr) will be the dependent variable. The independent variables of the regression will be the important determinants, resulting from a study by Mitton (2004). He included the factors firm size, firm growth and profitability as the basis in determining dividend payouts. He tested the effect of corporate governance on the dividend policy. The other independent variable will be a proxy for the signaling effect.

The dependent variable Dividend payout ratio will be tested and regressed on these

independent variables. The difference between the outcome of the model before and during the crisis will be showed by using a dummy for crisis. The model will also include a coefficient for time trend, which enables to find a pattern over time. This time trend is linear. With t = 8, including the years 2003 – 2010. The model will be tested using fixed effect and random effect estimation, using a panel of firms over multiple years. The fixed effect estimation controls the model for firm specific effects. The random effect estimation assumes this variation across firms is random and allows these time-invariant variables to play a role as independent variables.

The regression model used for Dividend Payout Ratio:

Dprit= β₀ + β₁IOit + β₂Prit + β₃ lnSzit + β₄ lnSEit + β₅Criit + ηi + γ₁t + εit (1)

Because the Dividend Payout Ratio includes the effects of income, the model will also be tested on the dependent variable Dividend Growth. The effects of income might result in the wrong conclusions about dividends in crisis years. Because when income decreases during a crisis, it may appear that the dividend payout ratio has increased, even though dividends have fallen.

The regression model used for Dividend Growth:

Dividend Growth Rateit = β₀ + β₁IOit + β₂Prit + β₃ lnSzit + β₄ lnSEit + β₅Criit + ηi + γ₁t + εit (2)

The models use the following variables: Dependent:

- Dividend Payout Ratio (Dpr) - Dividend Growth

Independent:

- Investment Opportunity (IO) - Profitability (Pr)

- Size ln(Sz)

- Signaling Effect ln(SE) - Crisis (Cri)

- Firm Fixed Effects (η)

- Time Trend, with time = t: ϒ₁t - Error Term (ε)

The dependent variable Dividend Payout Ratio is measured as the ratio of (Common Dividends/ Income Before Extraordinary Items-Adjusted for Common Stock Equivalents). The ratio shows the percentage of income paid to shareholders.

The dependent variable Dividend Growth is measured by the difference between dividends in the current year and dividends in the previous year, divided by the dividends in the previous year ((Dividend paid this year – Dividend paid previous year)/Dividend paid previous year). Three important variables in the model by Mitton (2004) are investment opportunity, profitability and size. For the definition of the variables, a more recent study is used of Denis and Osobov(2008). They used data from the Compustat database merged with the CRSP database, to measure the likelihood of paying dividends.

Investment opportunity is measured by the market-to-book ratio (Market Value of the Firm/

Book Value of the Firm). Market-to-book ratio measures the value of a firm by its expected future growth opportunities.

Profitability is measured by the Operating Profit Margin (Earnings Before Interest and Tax to

net Sales). This margin looks at EBIT as a percentage of sales. Operating profit margin ratio is a measure of overall operating efficiency.

Size is measured by the Total Assets of the firm. Total Assets is the combined value of all assets owned by a company.

Signaling effect ln(SE), measured by the Bid –Ask Spread by Abarbanell & Kim (2008). Spread is

the difference between closing daily asks and closing daily bids, divided by the average of closing daily asks and bids on the day of the dividend announcements.(Ask-Bid)/((Ask – Bid)/2). The asking price is the price the seller wants to receive for their stock, the bid price is the price that the buyer is willing to pay for the stock. The bid-ask spread is used to measure information asymmetry. A bigger spread means that less inside information is available for investors. On the day of the dividend announcement, this information asymmetry can lead to positive or negative signals to the investors.

Crisis (Cri), dummy for crisis which is equal to 1 for the years 2007-2010, and equal to 0 for the

years 2003-2006. This dummy compares data of years before the crisis with data of years during the crisis. According to Mishkin (2008) the first signs of the start of the crisis were in 2007.

The model is controlled for Firm Specific Effects (η), by using Fixed Effect Regression. This method controls for omitted variables in panel data, which differ between firms but are constant over time.

The model will also be tested using Random Effects Regression. With the Random Effects Regression, the variation across firms is assumed to be random and uncorrelated with the independent variables included in the model.

The model also contains Time Trend (γ₁t), with t = 1,2,3,4,5,6,7,8 for the years 2003 – 2010. This enables to compare one period to another time period, and find patterns of change in an indicator over time. If the coefficient turns out to be significant, this will mean that Dpr does change over time given the values of the other independent variables.

4.2. Data

The data on yearly fundamentals will be gathered using Compustat and data on stock prices using CRSP. Data of the financial crisis starting in 2007 and of previous years will be used (2003-2010), and it will consist only of firms in North America. Both U.S. dollars (USD) and Canadian dollars (CAD) will be the used currency.

The data will be eliminated for two SIC – codes, financial firms (SIC code 6000 – 6999) and utility firms (SIC code 4900 – 4949). Also data with missing CUSIP will be eliminated, because this financial security code is used for merging the data files. And data with missing values, values equal to zero (of Assets, Liabilities, Shares Outstanding and Stock Price) and illogical

values (of market-to-book ratio and profitability), will also be eliminated. After merging the Compustat database with CRSP database, the nonmatched data will be eliminated. After elimination of all these data, the total sample size is 5.137. This sample contains 925 Canadian firms (18%) and 4.212 American firms (82%). The following table shows the sample selection.

Sample selection.

Observations Compustat 2003 -2010 87.836

- Elimination SIC Financial 6000 – 6999 - Elimination SIC Utility 4900 – 4949 - Missing CUSIP

-Missing values or zero values

24.868 2.605 56 5.781 Total Observations after elimination 54.526

Observations CRSP 2003 -2010 13.798.993

- Elimination SIC Financial 6000 – 6999 - Elimination SIC Utility 4900 – 4949 - Missing CUSIP

-Missing values or zero values

4.643.196 304.711 0 8.840.952 Total Observations after elimination 10.134 Observations Compustat/CRSP Merged

Not Matched From Master From Using 50.948 47.666 3.282 Matched - Duplicates Total Observations 6.860 - 8 6.852 Total Observations 6.852 -Illogical values

(Negative Market to Book ratio and 1<Profitability<-1)

1715

Total Observations 5.137

Number of firms 1.487

5 Results.

In this part the results of the model will be described. Starting with the descriptive statistics of the sample, followed by some regressions on the dividend payout ratio and some regressions for the crisis variable within different groups. At last some regressions on Dividend Growth will be included. The data is collected and eliminated, resulting in a total sample size of 5137.

5.1 Descriptive statistics

The descriptive statistics show that the data contains very large outliers. The standard deviations are very large compared to the mean of the variable. Especially investment

opportunity and profitability contain some extreme values. In order to prevent these values to have a bad influence on the regressions, the data will be winsorized. The outliers will not be excluded from the data, but transformed to values at the 1% and 99% percentiles. This is done for the Dividend Payout Ratio, Investment Opportunity,Profitability and Dividend Growth. The following table (1) shows the descriptive statistics after winsorization.

Table 1: Descriptive Statistics after winsorization.

Variable Obs Mean Std. Dev. Min Max

Dividend 5137 86.42793 524.9952 0 9985

Dividend Payout Ratio 5137 .1540614 .4827981 -.8751686 2.996236

Investment Opportunities 4801 2.963229 3.529789 .1623838 25.4316

Profitability (in %) 5137 .0494203 .2027712 -.7814615 .4893206

Size (in original values) 5137 3991.086 15869.79 .028 275644

Crisis 5137 .4580494 .4982855 0 1

Signaling Effect 5137 -5.459327 .348386 -5.904609 -4.870008

This study measures the effect of a financial crisis on the dividend policy of firms. To show how the variables differ between the years before the crisis and the years during the crisis, table(2) contains the descriptive statistics for these different periods.

Table 2: Decriptive statistics on pre – crisis years and crisis years.

Variable Mean Mean

Pre- Crisis Crisis T- Test Significance 2003 - 2006 2007 - 2010

Dividend 72.38057 103.0484 -2.0290 0.0425**

Dividend Payout Ratio .1585193 .1487869 0.7215 0.4706

Investment Opportunities 3.146363 2.751738 3.8918 0.0001*

Profitability (in %) .046125 .0533192 -1.2649 0.2060

Size (in original values) 3538.326 4526.778 -2.1829 0.0291**

Crisis 0 1

Signaling Effect -5.406251 -5.522126 12.1608 0.0000*

Note: The dependent variable Dividend Payout Ratio is measured by Common Dividends/Income Before Extraordinary Items-Adjusted for Common Stock Equivalents, the independent variable Investment Opportunities is measured by the Market-to-Book ratio, Profitability is measured by Operating Profit Margin, Size is measured by Log Total Assets, Crisis is a dummy for crisisyears and Signaling Effect is measured by the Bid-Ask Spread.

Note: The dependent variable Dividend Payout Ratio is measured by Common Dividends/Income Before Extraordinary Items-Adjusted for Common Stock Equivalents, the independent variable Investment Opportunities is measured by the Market-to-Book ratio, Profitability is measured by Operating Profit Margin, Size is measured by Log Total Assets, Crisis is a dummy for crisisyears and Signaling Effect is measured by the Bid-Ask Spread. Significant at P<0.01* P<0.05**

There appear to be no big differences between the two periods. The mean of the dependent variable Dpr almost stays the same, which is quite relevant for this study. Firms seem to have less investment opportunities during the crisis. And, they seem to have more assets available during crisis years. The signalling effect decreases a little, while profitability increases.

Differences between the means of the pre-crisis period and the crisis period are tested using a t-test. The t-test shows significant differences between the means of Dividend (at α=5%),

Investment Opportunities (at α=1%), Size (at α=5%) and Signaling Effect (at α=1%).

5.2 Regression results. 5.2.1 Dividend payout ratio.

Table 3 shows the regression on dividend payout ratio including all variables and time trend, and controlled for fixed effects. The model is valid because the F statistic is 2.63, which is higher than the critical value at a significance of 5%. This means that the variables are jointly

significant. The overall R-squared is 1,21%, indicating that the variables explain 1,21% of the sample variance of dividend payout ratio. Errors are negatively correlated (-0.3496) with the regressors in the fixed effects model. The rho shows the percentage of the variance due to differences across panels, which is 47,92%.

Table 3: Regression on dividend payout ratio using fixed effects.

Dividend Payout Ratio Coef. Std. Err. t P>|t| [95% Conf. Interval]

Constant -15.76993 15.55817 -1.01 0.311 -46.27455 14.73469 Investment Opportunities .0003728 .0029648 0.13 0.900 -.0054402 .0061858 Profitability .1265172 .0624034 2.03 0.043** .0041641 .2488704 Size -.0344601 .0196816 -1.75 0.080*** -.0730495 .0041292 Signaling Effect -.0458146 .0248237 -1.85 0.065*** -.0944859 .0028567 Crisis -.0234706 .0299126 -0.78 0.433 -.0821197 .0351785 Time Trend .0079157 .0078096 1.01 0.311 -.0073965 .0232279

Fixed-effects (within) regression Number of obs = 4801

Group variable: cusip1 Number of groups = 1487

R-sq: within = 0.0048 Obs per group: min = 1

between = 0.0204 avg = 3.2

overall = 0.0121 max = 8

Corr(u, Xb) = -0.3496 F(6,3308) = 2.63

Rho = .47916633 Prob > F = 0.0151

Looking at the independent variables, there are three significant variables. Of the variables used by Mitton(2008) only profitability is significant. The variables Investment Opportunity, Crisis and Year are insignificant. The signaling effect is also significant, its t-statistic is above the critical value at α=0.10. The signaling effect negatively influences the dividend payout ratio with a coefficient of -0.0458146.

The next regression (Table 4) tests all variables on dividend payout ratio, controlled for random effects and linear time trend. This model is also valid according to the test on the joint

significance of the variables. Random effects model is tested by the wald chi square test, with a value of 130.50, which is higher than the critical value at α=0.05.

The overall R-squared is 4,12%, indicating that the variables explain 4,12% of the sample variance of the dividend payout ratio. The differences across firms are uncorrelated with the regressors in the random effects model. This allows for time-invariant variables to play a role as explanatory variables.

Note: The dependent variable Dividend Payout Ratio is measured by Common Dividends/Income Before Extraordinary Items-Adjusted for Common Stock Equivalents, the independent variable Investment Opportunities is measured by the Market-to-Book ratio, Profitability is measured by Operating Profit Margin, Size is measured by Log Total Assets, Crisis is a dummy for crisisyears and Signaling Effect is measured by the Bid-Ask Spread. Significant at P<0.01* P<0.05** P<0.1***.

Table 4: Regression on dividend payout ratio using random effects.

Dividend Payout Ratio Coef. Std. Err. t P>|t| [95% Conf. Interval]

Constant -6.519709 14.4139 -0.45 0.651 -34.77043 21.73102 Investment Opportunities .000303 .0019952 0.15 0.879 -.0036076 .0042135 Profitability .2235712 .0387782 5.77 0.000*** .1475674 .2995751 Size .0268357 .0039859 6.73 0.000*** .0190236 .0346479 Signaling Effect -.0417896 .0236115 -1.77 0.077* -.0880674 .0044881 Crisis -.0364129 .0287804 -1.27 0.206 -.0928214 .0199957 Time Trend .003137 .0072319 0.43 0.664 -.0110372 .0173113

Random-effects GLS regression Number of obs = 4801

Group variable: cusip1 Number of groups = 1487

R-sq: within = 0.0022 Obs per group: min = 1

between = 0.0691 avg = 3.2

overall = 0.0412 max = 8

Corr(u, Xb) = 0 (assumed) Wald chi2(6) = 130.51

Rho = .2036134 Prob > F = 0.0000

To decide whether to use the fixed effects estimation or the random effects estimation, the Hausman test is used. The Hausman test tests for the efficiency and consistency of the random effects estimation. Using the chi-squared distribution, the test shows a P-value of 0.0037, which is smaller than the significance level of 5%. The null hypothesis that the random effects would be consistent and efficient is rejected at α=0.05, and the fixed effect estimation is preferred for testing the model.

The crisis dummy is not significant in the regression model used above. To study the effect of each variable related to crisis, the following table(5) shows the coefficient of crisis within the specific groups. These groups are separated by the mean of each variable and added to the model. The regression on Dividend Payout Ratio is done for each group separately, to hypothesize the effect of crisis for each variable.

Note: The dependent variable Dividend Payout Ratio is measured by Common Dividends/Income Before Extraordinary Items-Adjusted for Common Stock Equivalents, the independent variable Investment Opportunities is measured by the Market-to-Book ratio, Profitability is measured by Operating Profit Margin, Size is measured by Log Total Assets, Crisis is a dummy for crisisyears and Signaling Effect is measured by the Bid-Ask Spread. Significant at P<0.01* P<0.05** P<0.1***.

Table 5: Regression coefficients of crisis on Dpr for separate groups, using fixed effects.

Dummy Crisis Coefficient P>|t|

More profitable firms -.0155795 0.691

Less profitable firms -.0397666 0.427

More investment opportunities -.0292074 0.602

Less investment opportunities -.033549 0.399

Larger firm size -.0015446 0.975

Lower firm size -.0368802 0.300

Higher signaling effect 1.355879 0.002*

Lower signaling effect -.0365452 0.309

With α=0.10 the coefficient of crisis is significant only for higher signaling effect. With higher signaling effect, crisis has a positive influence on Dividend Payout Ratio.

5.2.2 Dividend Growth

The normal regression model on dividend growth is valid because the F statistic is 32.69 which is much higher than the critical value at a significance of 5% (Table 6). This means that the variables are jointly significant. The R-squared is 11,20%, indicating that the variables explain 11,20% of the sample variance of dividend growth. The sample of dividend growth is smaller than the sample of dividend payout ratio due to the values of dividend equal to zero.

Note: The dependent variable Dividend Payout Ratio is measured by Common Dividends/Income Before Extraordinary Items-Adjusted for Common Stock Equivalents, the independent variable Investment Opportunities is measured by the Market-to-Book ratio, Profitability is measured by Operating Profit Margin, Size is measured by Log Total Assets, Crisis is a dummy for crisisyears and Signaling Effect is measured by the Bid-Ask Spread. Significant at P<0.01* P<0.05** P<0.1***.

Table 6: Regression coefficients on Dividend Growth

Dividend Growth Rate Coef. Std. Err. t P>|t| [95% Conf. Interval]

Constant 188.5143 2430.361 0.08 0.938 -4578.616 4955.644 Investment Opportunities .4925185 .3070843 1.60 0.109 -.1098246 1.094862 Profitability 1.839066 5.585326 0.33 0.742 -9.1165 12.79463 Size 6.574088 .5168818 12.72 0.000*** 5.560229 7.587947 Signaling Effect 3.98842 4.047655 0.99 0.325 -3.951018 11.92786 Crisis -.0110671 4.849959 -0.00 0.998 -9.524216 9.502082 Time Trend -.0987606 1.219297 -0.08 0.935 -2.4904 2.292879

Source SS df MS Number of obs = 1562

Model 348174.577 6 58029.0962 F(6, 1584) = 32.69

Residual 2760228.11 1555 1775.06631 Prob> F = 0.000

Total 3108402.69 1561 1991.28936 R – squared = 0.1120

Root MSE = 42.132

Taking a look at the variables, there is only one significant variable. The other variables used by Mitton(2008), investment opportunities and profitability are both insignificant at α=0.05 and α=0.10. Also the signaling effect, crisis and the time effect seem to be insignificant. Size is the only variable that is significant at α=1%, it’s t-statistic is above the critical value at α=1%. Size positively influences the dividend growth (6.574088).

6 Results Analysis.

The descriptive statistics on pre-crisis years and crisis years separately show significant results for dividend, investment opportunities, size and the signaling effect. In crisis years dividends were significantly higher than in the years before the crisis. This might be because of

significantly lower investment opportunities available during crisis or because of the stronger negative effect of signaling. This relation is showed by the following regressions on dividend payout ratio and dividend growth.

Note: The dependent variable Dividend Growth is measured by the difference between dividends in the current year and dividends in the previous year, divided by the dividends in the previous year, the independent variable Investment Opportunities is measured by the Market-to-Book ratio, Profitability is measured by Operating Profit Margin, Size is measured by Log Total Assets, Crisis is a dummy for crisisyears and Signaling Effect is measured by the Bid-Ask Spread. Significant at P<0.01* P<0.05** P<0.1***.

6.1 Dividend Payout Ratio.

The regression on the dividend payout ratio does not show significant results for crisis. It does show a significant effect of profitability, size and the signaling effect on the dividend payout ratio.

The first hypothesis of this study expected a decrease in the dividend payout ratio caused by the financial crisis. But the variable crisis is not significant in the regression on Dpr, and therefore provides no evidence of a negative effect of a crisis on the dividend payout ratio. This might be due to the reluctance of managers to change dividends during the crisis, as concluded by Floyd, Li and Skinner (2011). To keep the dividend payout stable they might lend more cash or save cash by spending less in other segments, like marketing expenditures as showed by Campello, Graham and Harvey (2009).

The results show a significant positive effect of profitability on the dividend payout ratio (0.1265172). Firms with higher profits will have a higher dividend payout ratio. This result concurs with the conclusion of Aivazian, Booth and Clearly (2006), who state that the

probability of a firm paying dividends increases with the firm’s profitability. Size has a negative effect on the dividend payout ratio (-0.0234706). This result is not in line with the findings of Denis and Osobov (2008) that larger firms tend to pay more dividends. This unexpected result is possibly due to the effect of earnings in the model.

The signaling effect shows a negative effect on the dividend payout ratio. It reduces the dividend payout ratio, by - 0.0425808. This result shows that the larger the information asymmetry between the firm and its shareholders is, the more the dividend payout ratio decreases. This indicates that managers do not consider shareholders behaviour when determining the dividend. When information asymmetry is large and the dividend payout ratio decreases, the firm signals are negative and shareholders might believe that the managers of the firm gave up hope that earnings will rebound in the near future, and that they had to reduce the dividend payment in order to save cash. This negative result of the signalling effect might be because of firms not considering the signaling effect to be a first-order determinant (DeAngelo, DeAngelo and Skinner, 2004).

Investment opportunities is not significant. This might be due to the fixed effect estimation. When using fixed effect estimation, the model is controlled for influences of firm specific effects. These time-invariant characteristics are removed from the regression. This means that the constant is different for each firm, because it includes these firm fixed effects. The test showed that 47,91% of the variance is due to the differences across firms, this means that the fixed

effects are heavily correlated to the variables. Controlling for them lowers the significance of the variables. According to the Hausman test, fixed effect estimation should be used instead of the random effects model. With the random effects model the variation across firms is assumed to be random and uncorrelated with the independent variables included in the models.

Because the coefficient of crisis is not significant in the regression on Dpr, the same regression is done within different groups for profitability, investment opportunities, size and the signaling effect to test the dummy for crisis. The group tests for profitability, investment opportunities and size do not show significant effects on the relationship between the crisis and dividend payout ratio. These insignificant results seem inconclusive regarding the related hypotheses about their effects on the dividend payout ratio during a crisis. Still profitability and size are important determinants for the dividend policy of a firm, as showed by the main model.

The group higher signaling effect does show a significant effect of crisis on the dividend payout ratio at α=0.10. This shows that when the effect of signaling is high, the crisis has a positive effect on the dividend payout ratio. When the signaling effect is high, the role of dividend becomes more important. As showed by the results, when dividend payments are clear signals during a crisis, the managers are more reluctant to cut dividends and even increase them. This could signal that the managers believe that the firm can overcome the financial constraints of the crisis. This result is line with the expectation that when the role of dividends becomes more important, managers are more reluctant to reduce the payments during a crisis and even increase them.

6.2 Dividend Growth.

The regression on the dividend payout ratio includes the effect of earnings, and therefore the same regression is done on dividend growth. This resulted in only one significant variable, size. The size effect shows a coefficient of 6.574088. Larger firms tend to pay more dividends than smaller firms. Because the effect of earnings is excluded in this regression, the positive effect of size shows that larger firms pay higher dividends and are able to keep it high even during crisis years. It might be due to their overall liquidity rate, which is higher than that of smaller firms, enabling them to spend more cash on dividend payments (Denis and Osobov, 2008). It might also be that larger firms have more cash available to spend on dividends, because of a small number of positive investment opportunities. And therefore they might prefer to pay dividends instead of investing the cash in risky investments. Smaller firms are more discrete in spending their available cash.

7 Discussion.

This study measures the effect of a financial crisis on the dividend policy of a firm. The financial crisis of 2008 is used to test whether there are differences in pre-crisis years and crisis years. A crisis does influence the firms financially, forcing them to adjust their financial policy. This study focuses on the dividend policy of firms during a crisis, the main question is; to what extent does a firm changes its dividend policy when facing a financial crisis? To answer this question this study uses a model by Todd Mitton (2004). This model tests the influence of several factors on the dividend payout ratio of firms. Included in the model are investment opportunities, profitability and firm size. Added to the model is the signaling effect and time trend.

The results of the regressions on the dividend payout ratio show only a few significant variables. Expected was that the dividend payout ratio would decrease due to the financial crisis. But the crisis variable does not show a significant influence on the dividend payout ratio. This assumes that there is no effect of crisis on the dividend policy of a firm. The only significant effects on the dividend payout ratio were caused by profitability, size and the signaling effect. These results confirm that these variables are important when determining the dividend policy of a firm. By splitting the sample into groups, the results show a significant effect of crisis on the dividend payout ratio when the signaling effect is high. During the crisis, firms increased their dividend payout ratio when the effect of signaling was high. This is in line with the hypothesis.

The regression on dividend growth resulted in significance of firm size. Size has a positive effect on dividend growth, assuming that larger firms keep the dividend payment level high. The crisis variable showed no significant effect on the dividend growth.

The regressions on the dividend pay-out ratio and dividend growth both provide no significant results for investment opportunities. Expected was that the payout ratio of firms with a lower profitability would decrease more compared to more profitable firms. This study therefore does not concur with earlier research, which show that investment opportunities are important in determining the dividend policy of a firm.

As a result of these unexpected findings of the effects of crisis on dividend policy, it might be useful to consider different methods of measurement for the used variables. Future research could possible use these different measurement methods to provide more significant results in testing the effect of crisis on the dividend policy of a firm. Future research could also use a different sample. It could be interesting to include data of after crisis years or data on other financial crises.

To conclude this study, whether firms change their dividend policy during a financial crisis depends on whether dividends serve a signaling role. Profitability, investment opportunities and size are important determinants of dividend policy in general, but the used model did not show a relationship between the financial crisis of 2008 and the dividend policy of firms when controlling for these variables.

Reference list

Abarbanell, J. and Kim, S. (2008). Why Returns around Earnings Announcement Days are More Informative than Other Days. University of North Carolina at Chapel Hill.

Acharya,V.V., Gujral, I. and Song Shin, H. (2009). Dividends and Bank Capital in the Financial Crisis of 2007-2009. New York University , London Business School (MBA 2007) and Princeton

University.

Aivazian, V., Clearly, S. (2006). Dividend Smoothing and Debt Ratings. Journal of

Financial and Quantitative Analysis, vol. 41, no.2, p. 439-453.

Allen, F. and Michaely, R. (2002). Payout Policy. Cornell University.

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.

Black, F. (1976). The dividend puzzle. Journal of portfolio management 2. pp. 5-8. Black, Fischer, and Myron S. Scholes, 1974, The Effects of Dividend Yield and Dividend Policy on Common Stock Prices and Returns. Journal of Financial Economics 1, 1-22. Brav, A., Graham, J.R., Harvey, C.R. & Michaely, R. (2005). Payout policy in the 21st century.

Journal of financial economics 77. 2005. pp. 483-527.

Campello, M., Graham, J.R. and Harvey, C.R. (2009). The real effects of financial constraints: Evidence from a financial crisis. Journal of financial economics. Volume 97 No. 3 (2010). pp. 470-487.

DeAngelo, H. & DeAngelo, L. (2006). The irrelevance of the MM dividend irrelevance theorem.

Journal of financial economics 79. 2006. pp. 293-315.

DeAngelo H. , DeAngelo L and Skinner D (2004), Are Dividends Disappearing? Dividend Concentration and the Consolidation of Earnings. Journal of Financial Economics, 72, 425-456. 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.

Floyd, E., Li, N., and Skinner, D.J. (2011). Payout Policy Through the Financial Crisis: The Growth of Repurchases and the Resilience of Dividends. Chicago Booth Research Paper, No. 12-01 Gorton, G and N Souleles (2005). Special purpose vehicles and securitization. NBER Working

Papers, no 11190.

Kindleberger, C., (1987), Bubbles. In J. Eatwell, M. Milgate, P. Newman (eds.), The New

Palgrave:A Dictionary of Economics 1, A to D, pp. 281-282.

Miller, M.H. and Modigliani, F. (1961). Dividend policy, growth, and the valuation of shares.

Journal of Business 34, 411–433.

Mishkin, F. S. (1996). Understanding Financial Crises: A Developing Country Perspective. Annual

World Bank Conference on Development Economics, pp. 29-62.

Mitton, T. (2004). Corporate Governance and Dividend Policy in Emerging Markets. Emerging

Market Review. Volume 5. pp. 409-427

Redding, L.S. (1997). Firm size and dividend payouts. Journal of financial intermediation 6. pp. 224-248. (1997).

Vankatesh, P.C. & Chiang, R. (1986). Information asymmetry and the dealer’s bid-ask spread: A case study of earning and dividend announcements. Journal of finance. Volume XLI nr. 5.

Books

Shiller, R. J. (2008). The Subprime Solution: How Today’s Global Financial Crisis

Happened and What to Do about It. New Jersey: Princeton.University Press.