Miller and Modigliani (1961) posit that under the assumption of perfect capital markets, it is a firms' investment decision that determine its value and not dividend policy; thus, dividend decision is irrelevant. This has often been described in finance and business literature as firm having residual dividend policy, and in effect makes dividend policy irrelevant to firm value.
Many researchers [1] have questioned the dividend irrelevance posture taking by Miller and Modigliani in determining a firm's value. These writers have argued that firm's dividend decisions are not residual ones as was suggested by Miller and Modigliani. With a lot of market imperfections [2] and the of risk corporate default, dividend policy is not independent but rather a simultaneous consideration is given to investment and financing decisions as management make a decision on dividend payment and dividend sustainability.
The concept of dividends and its payments have been quite challenging and controversial in the literature and practise of contemporary finance. Apart from many theories and models, quite a number of empirical work have been done in explaining the factors management should consider in deciding dividend payment. Thus, setting corporate dividend policy remains a challenge for managers and involves judgment by decision makers. According to Brook et al. (1998) corporate dividend policy is not driven by a single goal. For Black (1976), as the study of dividend policy increases, it becomes like a puzzle whose parts seem not to fit perfectly. What determines dividend pay-out?; has been a question most business and finance researchers have been answering over the past decades.
According to Berk J. and DeMarzo P. (2011), a firm's dividend pay-out is shaped by proï¬tability, institutional holdings, cash flow, taxes, agency cost, market to book value and asymmetry information. Whiles some of these determinants are perhaps country specific, others to a larger extent are industry specific. As said by Baker and Powell (2000), some determinants of dividend pay-out are industry specific though future earnings are the major determinant.
1.1 Objectives and Research Questions
Though there has been a considerable volume of work (Megginson 1991; Leither and Zimmerman 1993; Farinha2002; Dhanani 2005 and Ferris et al 2006) on corporate dividend policy in the UK, empirical evidence have been on the broad economy with much emphasis on the industrial and retail sectors. Ferris et al (2006) [3] , particular excluded financial institutions from their study of UK dividend policy due to regulatory issues. The thrust of this study is to investigate the determinants of dividend policy using financial institutions not only Banks but the entire financial sector of the UK economy. These financial institutions include Banks, Investment Companies, Speciality and other finance and non-index investment companies. The general purpose is to examine the determinants of dividend pay-out policy of financial companies and firms in the UK market. Do these determinants have same relationship with dividend pay-out as shown by studies of other sectors in the UK economy? The factors shall include; profitability, tax, growth opportunities, firm size and risk (leverage). The choice of these variables is consistent with available literature and similar work on this topic. In addition to the general purpose, we will also find out which of these determinants has much influence on deciding dividend payment.
The study will therefore answer and/or test five questions/hypotheses with respect to the proposed determinants by similar works in other countries or sectors of the economy.
Highly levered or risky firms pay fewer dividends to their shareholders.
Is there a positive relationship between size of a firm and dividend payout?
Increase (decrease) in corporate tax will lead to a decrease (increase) in dividend payouts.
Matured firms pay more dividends than growth firms.
What is the relationship between firm profitability and dividend payouts?
1.2 Motivation of the Study
As a student of Finance and Economics with special interest in Corporate Finance, I have come across and discuss many articles and seminal papers on corporate dividend policies for both developed and under developed economies. I had much interest in this area during the teaching and seminars of the Corporate Finance module. In other to get a better understanding and appreciation of this area of finance, I have decided to have a practical and empirical study of dividend pay-out policy of corporations. The financial sector of UK was chosen because of the scarce literature that exists. Besides financial companies play a major role (funding) in the growth of other firms-perhaps a reason why some critics blame the occurrence of the recent financial crisis on financial institutions.
In effect, this study will contribute significantly to the existing literature on dividend pay-out in a quite restrictive and crucial area (financial sector) of the UK economy.
1.3 Structure of the Study
The rest of the study is organized as follows. The next chapter gives a review of the existing related literatures, both theoretical and empirical studies on the subject. Chapter three describes data and sample as well as the various statistical tools and methodology use in the analyses. Chapters four and five present and discuss the results of the empirical analysis and conclude the study respectively.
CHAPTER TWO
LITERATURE REVIEW
Theoretical Review
We begin this chapter by first reviewing financial theories, hypothesis and models that support and/or explain dividends payout policies of firms. These include tax-adjusted theories; tax preference or trade-off theories; agency cost; cash flow hypothesis, investment opportunities hypothesis and dividend signalling hypothesis. These theories and hypothesis to a larger extent influence and guide the payment of dividends by corporations.
The dividend signalling theory suggest that corporate dividend policy is used by managers as means of putting quality message across in terms of the firm's performance level, current and future profitability. Proponents of this theory include Miller and Rock (1985); Ofer and Thakor (1987); and Rodriguez (1992). They argued that information asymmetry between managers and outside shareholders make the former to use dividends as a tool to signal private information about a firm's performance [4] and future prospects to outsiders. Thus, payment of high dividend send information to investors and the capital market about the firm's high profit level or future earnings. Rock's (1985) developed a model based on the asymmetric information principle and indicated that dividends announcement provides the missing information about the sources or uses of funds, and also allows the market to estimate the firm's current earnings and perhaps prospective earnings. According to Lasher (2000), a decrease in dividend, is normally taken as terrible news. He posit that dividend reduction is as a result of sustained reduction in earnings, and sends information to the market and analyst that management does not expect the company to have the cash or profits it had in the previous years.
Another theory that seeks to explain corporate dividend policy is the Agency theory [5] . This theory attempts to explain the conflict of interest that arises between shareholders and managers-with proponents suggesting dividend payment to reduce it. The thrust of the agency theory is that managers may take actions that only serve their own interest and contradicts that of shareholders thereby not benefiting them (shareholders). This has been a major issue in corporate governance. According to Jensen and Meckling (1976), Agency costs are lower in firms with high managerial ownership stakes because of better alignment of shareholders' interest and managerial control. Shleifer and Vishney (1986) also posit that firms with large block of shareholders have low agency cost because they are able to monitor the activities of managers. Another area of Agency problem is the ownership and control of the firm either by bondholders or shareholders. Thus, there is a threat of wealth transfer from bondholders to shareholders.
One possible way of getting around the Agency cost issue is to pay high dividends. As explained by Fama and Jensen (1983), the best way to solve this conflicting issue is to pay dividend to shareholders. Jensen (1986) in his free cash flow hypothesis posit that funds remaining after financing all positive Net Present Value projects cause conflicts of interest between managers and shareholders. Therefore, payments of dividends and debt interest decrease the free cash flow available. In effect, dividend payout should be used by management to reduce the problem of agency cost as suggested by the above theorists.
The tax preference or trade-off theory suggests that dividends are subject to a higher tax cut than capital gains. According to the theory, capital gain tax is charged when stocks are sold, whiles dividends are directly taxed. This obviously will make investors prefer retention of firms' profit to distributing them as dividends. Tax-adjusted models which assumes that investors mostly maximizes their after tax income, explain that investors of firms expect higher returns on shares of dividend paying stocks. There are different investors group with respect to the tax bracket they belong, and therefore varying dividend preferences. The effect of this theory plus differences in tax preferences is what is termed clientele effects. While one group of investors want the firm to pay out a higher percentage of its earnings, another may prefer otherwise. Thus, individuals in high tax bracket will prefer stocks that pay less dividends as oppose to tax free investors who will want high dividend paying-stocks. Modigliani (1982) argues that this clientele effect is responsible for the rebalancing and alteration in portfolio composition by fund and portfolio managers. As the tax obligation of the investor changes, so will his demands for dividends and returns on the investment.
As a rule of the thumb for corporate tax, most corporations will pay dividends after corporate taxes have been paid; the effect is less dividend pay-out if high taxes are charged.
According to the Classic theories of pay-out policy, dividends should be paid out of a firm's free cash flow; thus that part of the cash flow which cannot be invested in any project with positive NPV. This theory is similar to the free cash flow hypothesis advanced by Jensen (1986). By implication, dividend payment has a negative relationship with investment opportunities. Higgins (1972) in his dividend decision model indicated that pay-out ratio is negatively related to a firm's need for funds to finance growth opportunities and investment.
Another theory that also explains corporate dividend policy is the transaction cost and residual theory. This theory explains that firms incurring large transaction costs will be required to reduce dividend pay-outs to prevent perhaps the huge cost associated with external financing [6] . Proponents of this theory include; Higgins (1972) and Alli et al. (1993)
More recently, DeAngelo H. and DeAngelo L. (2006) proposed a life-cycle theory which draws ideas from Jensen and Meckling (1976) agency cost theory and the investment opportunities idea from Fama and French (2001). According to this theory, firms continually touch their dividend policy over time when their investment opportunities changes. They argue that, firms pay fewer dividends in the early years of operation when investment needs exceed that of the internally generated funds. However, they pay high dividends when their internally generated funds exceed the growth and investment opportunities in the latter years. In effect, as firms go through the various stages of growth and development (life-cycle), their dividend policies also change to suit each stage of the firm's life.
Variable measurement and Empirical Literature Review
Dividend payout ratio can best be described as the fraction of net income a firm pays to its stockholders or shareholders in dividends. This study defines the dividend payout ratio as the percentage of profits paid as dividend. Mathematically, we calculate it as the ratio of dividend per share to earnings per share. Pruitt and Gitman (1991) in their study indicated that, dividend payment is affected by the company's past and current profits.
Fama and French (2001) stated that the likelihood of firms to pay dividends depends on size, profitability, and growth opportunities. This section will review the framework of dividend policy and discuss several studies that tested determinants of dividend policy in different types of economies. This will include five hypotheses and/or factors base on empirical literature and the factors considered by this study.
2.1.1 Taxes
Tax in this study specifically refers to corporate tax. It is measured as the ratio of corporate tax to profits before tax. According to Short et al. (2002) taxation policy is considered a key determinant of dividend pay-out in the developed countries (like UK). Wu, C. and Hsu, J. (1996) on his part posit that firms continue to change their dividend policy when government position on corporate and income tax changes. Masulis and Trueman (1988) also argued that taxes have a greater influence on corporate dividend policy. Their model showed investors with different tax liabilities (due to income level, investment horizon, and tax jurisdiction) will prefer different dividend policy. Their conclusion was that as tax liability increases the dividend payment decreases and vice-versa. This is inconsonance with the tax preference or trade-off theory which says that favourable dividend tax should lead to higher pay-outs.
Singhania, (2006), studied the payment of dividends of manufacturing, non-government, non-financial, and non-banking companies listed on Bombay Stock Exchange (BSE) and found that investors prefer profits retention to receiving dividends; which supports the tax preferences theory. His analysis showed that a favourable change in the Indian tax regime saw average dividend pay-out percentage increased from 31.33% to 47.05%.
In a similar vein, Brandon and Ikenberry (2004) in their study among US firms documented using the Bush tax cut [7] as an example that a tax reduction will see an increase in dividend paying firms and an increase in the percentage of payment among already existing dividend payers. From the literature, the conventional hypothesis is that, there exists a negative relationship between tax and dividend pay-out ratio. Specifically in the United Kingdom, companies pay UK corporation tax on their profits and the remainder can be paid to shareholders as dividends. The dividend is seen to have been received net of 10% and so shareholders in the basic tax rate bracket do not pay further tax. In effect, whiles corporate tax has a negative relation on dividend pay-out; dividend or income tax could have either relation depending on the tax bracket the individual shareholder belongs.
Contrary to this general assertion, Abor and Amidu (2006) recorded a positive relationship between tax and dividend pay-out ratio among listed Ghanaian firms. This means that increasing tax is associated with increasing dividend payment.
2.1.2 Investment Opportunities (Growth)
The study uses growth as a proxy for investment opportunities. Though there are several measures of firm growth [8] , the firm's market-to-book ratio is used as a measure of growth which is consistent with other literatures. The general hypothesis and literature state that there exist a negative relationship between firm's investment needs and dividend pay-out rate. Higgins (1981) documented a direct link between growth and financing needs of firms. His argument was that rapidly growing firms will have greater financing needs; the effect is the payment of fewer dividends to shareholders.
Indeed, Jensen (1986) argued that, in order to maximize value, older and matured companies with limited growth and investment opportunities should commit to pay out all their "free cash flow"-that is, all (non-operating) cash that cannot be reinvested in positive NPV projects. Ferris et al (2006) used annual percentage change in the firm's total assets as proxy for investment opportunities and find that U.K. dividend nonpayers do not consistently demonstrate higher asset growth rates. They continued that firms that do not pay dividends have high asset growth. They further stated that firms that have never paid dividend have higher median asset growth rates than dividend payers. This means that the investment opportunities of dividend payers are less as compared to their counterparts on the other side. By implication, the investment opportunities of a firm reduce the likelihood of paying high dividends. Firms with attractive internal investments are less likely to have funds available for distribution as dividends [9]
Collins et al. (1996) had their findings and conclusions on investment opportunities and dividend payments pointing to one direction; Market-to-book value and growth in sales were used as proxies for future investment opportunities, a negative and statistically significance association was found between these two variables and dividend pay-out. What this means is that, growing firms require much funds in other to finance future investment, which makes them retain much funds instead of paying dividends. Brandon and Ikenberry (2004) used growth and takeover activities as proxies for investment. They indicate that, periods of high takeovers and growth saw a considerable cut in dividend payment among US firm. Their conclusion was that, firms only pay dividend in spirit to the maturity hypothesis. Thus, dividend policy is seen as the "residual" outcome of a company's capital budgeting process- "Pay out whatever cannot be reinvested to earn the cost of capital".
David and Igor (2008) find that the relationship between dividend payments and growth opportunities is not uniform across countries in their country analysis of dividend payers and non-payers. They documented that, in the US, Canada, and the UK, dividend payers tend to have less valuable growth and investment opportunities than nonpayers. However, in Germany, Japan and France, the above relationship is different and mostly that of a mix one.
The findings of D'Souza (1999) was quite inconsistence with other literature. She found positive statistically insignificant relationship in and negative but insignificant relationship of growth and market-to-book value respectively and dividend pay-out.
2.1.3. Debt/Risk
The study used the ratio of total liabilities to shareholders' equity (leverage) as a measure of risk or the amount of debt in the firm. The effect of financial constraint on dividend decisions has been documented by a lot of researchers. The hypothesis is that firms with relatively high debt ratio are unable to pay many dividends as expected.
Previous models and theories have looked at relationship between dividend pay-out and a firms' beta co-efficient; where the beta measures the firms' business or market related risk
Higgins (1972) developed a dividend decision model in which he posits that taking away the assumption of no corporate tax and default risk on corporate bonds, the optimal capital structure of a firm is an integral factor to making dividend policy. Again, Higgins' (1981) sustainable growth model shows that the absence of financial constraint of firms will make them pay out every surplus fund. From his model, he argued that a company might want to make an annual financing decisions without regard to dividend policy, (simply distribute anything remaining as a dividend), however such a view of the dividend decision overlooks deficit financing problem which might later arise. Rozeff (1982) in a related model to that of Higgins established that firms lower their dividend pay-out ratios when they have higher beta co-efficient. He concluded with his model that "dividend payments are quasi-fixed charges which are substitutes for other fixed charges". Fama and French (1997) made a similar study of debt decisions among U.S firms and conclude that debt is the residual financing decisions made by U.S corporations. In effect, debt principal and interest payments reduce a firm's residual income to guarantee high dividend payments.
Aivazian et al (2003) found an inverse relationship between debt/risk and dividend pay-out. Likewise, D'Souza (1999) finds a negative relationship between risk and dividend pay-out. Brandon and Ikeberry (2004) in their paper indicted that US companies with low debt ratio had increased dividend pay-out. Amidu and Abor (2006) also found a negative but insignificant relationship between risk and dividend pay-out ratios; this means that firms with high debt ratio pay lower dividends to their shareholders. They stated that, "firms experiencing earning volatility find it difficult to pay dividend, such firms would therefore pay less or no dividend". Many other researchers including Faccio et al. (2001); Al-Malkawi (2005); Al Kuwari (2009) showed in their studies that highly levered firms look forward to maintaining their internal cash flow for financing decisions and end up paying less dividend to their shareholders. .
2.1.4. Size
Firm size (SIZE) is measured as logarithm of total assets. An alternative way of measuring size is the log of net income or the market capitalisation of firms. The existing literature has been that, firms with large size should be able to pay higher dividends; hence one expects a positive relationship between dividend payments and size. Some theorists have suggested that large firms pay high dividends to shareholders in other to reduce agency cost [10] . Earlier studies have tested the impact of firm size on the dividend payment.
Vogt (1994) in his studies indicates that firm size is a crucial variable in explaining the dividend pay-out ratio of firms. His argument was that larger firms have easier access to the financial and capital markets (when size is a proxy for access to financial market); this reduces their dependence on internally generated funds for funding decisions thereby allowing them to pay higher dividend pay-out ratios. Fama and French (2000) documented that large firms distribute a higher amount of their net profits as cash dividends, than small firms. Eriotis (2005) in his study among Greek firms indicated that, Greek firms distribute dividend each year according to their target pay-out ratio, which is determined by distributed earnings and size of these firms. Ferris et al (2006), in their study of UK dividend paying and non-paying firms, documented dividend payers being larger than nonpayers. They used total assets as a measure of firm size and argued that, "the assets of the dividend-paying firms are 11% more than the non-paying firms".
2.1.5. Profitability
The level of a firm's profit has been a major determinant of dividend policy. Thus, profits and/or anticipated future earnings have been used as the primary and most fundamental indicator(s) of a firm's capacity to pay dividends to its shareholders. Profitability can be measured as the logarithm of net income; the ratio of earnings before interest (EBIT) to the book value of total assets and/or ratio of after-tax earnings to the book value of equity. This study in particular, used log of net income as a profitability measure. The concept of profitability and dividend payment can be best explained by the dividend signalling and the agency cost theories. Whiles dividend signalling describes the perception investors and the market have about a firm due to its dividend policy, agency cost theory seeks to explain how management dividend decisions can enhance shareholders wealth maximisation or otherwise. According to the "information content" explanation, dividends convey important financial and profitability information about the firm in question. All other things being equal, one should therefore expect a positive relationship between dividend pay-out and profitability. High dividend payment policy mostly make investors and analyst to think that the firm has achieved higher profit or expects high profit in future-a situation often referred to as the dividend signalling hypothesis. Thus, dividends are mostly used as a credible source of signalling to investors in capital markets. Conversely, dividend cut will send a signal of low profitability currently or in the future all other things being equal.
The argument of agency cost is that after tax profit will induce management to engage in wastage spending and empire building at the expense of the shareholders' wealth maximization. To eliminate this conflict of interest, payment of dividends will be the only form of communication to show management commitment to shareholder wealth maximisation. High profit levels will result in a high dividend pay-out ratio. According to Ferris et al (2006) "firm profitability appears to uniformly increase the likelihood that a firm will pay dividends". Abor and Amidu (2006) on their part identified a positive and statistically significant relationship between dividend pay-out and profitability. Fama and French (2001) indicated in their study that, the percentage of U.S. firms paying dividends has fallen from 67% of listed firms in 1978 to 21% in 1999. This they attributed to the fact that the character of exchange new lists has tilted toward firms with lower profitability.
CHAPTER THREE
Data Description and Methodology
After a review of theoretical and empirical literature in chapter two, this chapter will develop the main hypotheses for this study and provide detailed description of data and the methodology use in the analyses.
3.0 Data
The study made use of data extracted from the Extel Financials database [11] produced by Thomson Financial. The data extract was narrowed to country and industry specific (UK Financial Sector) as appropriate for the interest of the study. The industry is made up of four financial firms, comprising Banks, Investment companies, Investment & other Speciality and non-index financial companies. In effect, a resulting sample size of 4,020 firm-year observations over a period of 16 years (1993-2009) was used for the study.
3.1 Methodology
All statistical analyses and regressions were done using STATA 11.2. With the STATA commands (shown at the Appendix) and the data, the results shown in chapter four (4) can be replicated by readers. The raw data was converted (calculated) into variables as appropriate for the study.
Dividend pay-out ratio, which is the dependent variable, is defined as the dividend per share divided by earnings per share. The explanatory variables include profitability (PROF), risk (RISK), log of total assets (SIZE), corporate tax (TAX), and log of net income (GROWTH). Table 3.1.1 shows the proxy variables for each of the determinants, their definitions and the hypothesised or predicted signs.
Table 3.1.1: Proxy Variables Definition and Predicted Relationship
Proxy variables definitions predicted sign
PROF log of net income +
GROW market to book value -
TAX tax/net profit before tax -
RISK total liabilities/shareholders' equity -
SIZE log of total assets +
The panel character of the data allows for the use of panel data methodology. The term panel data refers to a multi-dimensional data which contains observations on multiple phenomena observed over multiple time periods for the same company or entity. The nature of the research questions and/or hypothesis requires statistical regressions to answer them.
The dividend pay-out (PAYOUT) can be modelled as follows:
PAYOUTi,t = β0 + β1RISKi,t + β2GROWTHi,t + β3TAXi,t + β4SIZEi,t + β5PROFi,t+μi,t (2)
Where:
PAYOUTi;t = dividend per share/earnings per share for firm i in period t,
RISKi;t = total liabilities/shareholder's equity for firm i in period t,
GROWTHi;t = market-to-book value for firm i in period t,
TAXi;t = corporate tax/net profit before tax for firm i in period t,
SIZEi;t = log of total asset for firm i in period t,
PROFi;t = log of net income for firm i in period t,
Panel data regression was employed as opposed to any form of regression because;
dynamics can be explored in the data unlike cross-sectional data,
one can control for factors that vary across entities but do not vary over time,
I can also control factors that cause omitted variable bias if they are omitted.
About six (6) different types of panel regression were done to get estimates for the above regression model. These include OLS, fixed-effect (RE), random-effect (RE), TOBIT, instrumental variables (IV), and robust regression (Rreg).
The first regression technique considered is ordinary least square estimations (OLS). It is used to estimate unobserved coefficients in a linear regression model. The many assumptions of OLS such as homoscedastic and serially uncorrelated errors, normally distributed errors, no perfect multicollinearity makes this method quite simple. However, if these assumptions do not hold (which is true in most cases), the estimates will be inconsistent and/or biased. These issues with OLS requires a more robust and statistically powerful regression models
A fixed effect model assumes differences in intercepts across states or time periods. Thus, with fixed effect regression, you can control for unobserved variables that vary across entities but not over time, and/or vary over time but not across entities. This model however, needs variations in the independent variables within states. It is also possible that the standard errors might be correlated overtime
Random effect model explores differences in error variances. With random effect, the firm-specific effect is a random variable that is uncorrelated with the explanatory variables. A typical random parameter model has an advantage of assuming neither heteroskedasticity nor autocorrelation within the panels to avoid complicating the covariance matrix.
Robust regression is a form of analysis designed to get around certain limitations of traditional parametric and non-parametric methods of estimation. In most situations, it is used when the data contain many outliers. In the presence of extreme outliers, ordinary least squares (OLS) estimations are inefficient, can be biased and are mostly non-robust to outliers.
TOBIT regression/model, also called a censored regression model, is used to estimate relationships between variables when there is either left-censoring or right-censoring in the dependent variable. Depending on the distribution of the dependent variable, an above or below censor can be used in the estimations to correct for observations that are above or below a certain threshold respectively.
In disciplines such as econometrics, where controlled experiments are not feasible, instrumental variables (IV) are used to estimate causal relationship. Instrumental variable regression gives consistent estimations when the explanatory variables are correlated with the error terms of the regression model. The existence of this correlation can lead to inconsistent and biased estimates if OLS is used.
To get desirable and perhaps statistically significant results of my estimations, the varying characteristics of the data requires not just a single regression technique but several other techniques as explained above. The study therefore, employed all the above regression types in estimating the coefficients of the model unlike other studies that used only one or two types of regressions.
Using STATA 11.2, I first conducted a descriptive statistics of all variables both at the firm and industry level. Secondly, we test three econometric problems to check if there is violation of assumptions of OLS regression. The tests include;
Heteroskedasticity
Autocorrelation
Multicollinearity
The results of these tests are shown and discussed in chapter four.
I also conduct a simple regression for each explanatory variable against pay-out ratio to see which one has a strong influence on the dependent variable. Finally, general regressions were done for the estimations and the relationship between the dependent and explanatory variables. All the regressions are made robust to correct for any assumption problems during estimations.
CHAPTER FOUR
RESULTS AND DISCUSSION
This chapter discusses the results of the regressions including summary statistics of the variables as well as the dividend pay-out rate (percentages) for each of the financial firms. The regression results are made up of different types of regression depending on which one suits the data and the occurrence of the variables statistically.
For instance, the study used IV1 to solve the econometric problem of endogeneity and/or omitted variable bias which might lead to inconsistent estimates; TOBIT regression (lower bound censored) is used and subsequently censored at zero due to the distribution of the dependent variable (payout). In effect, one cannot use only OLS in this analysis because the statistical assumptions of OLS do not hold for this data set, which means we may not have consistent and efficient estimates of the coefficients. Subsequently, six different regression results are discussed below.
Figure 4.0 below shows a frequency distribution of payout for all the financial firms.
Fig 4.0: Frequency distribution of payout
It can be seen that a greater part of the firm-year observation do not pay dividends to their shareholders, hence recorded zero (0) payouts on many times. The high ratios of payout (from 0.5 to 2.0) have lower frequencies. This implies that the financial institution in most occasions pay fewer or no dividends at all.
Summary Statistics
Table 4.1.1 Descriptive statistics of dependent and independent variables
Variables Obsv. Mean Std. Dev. Min. Max.
PAYOUT 7576 .678 .511 0 1.985
RISK 7874 2.255 4.301 0 48.288
GROWTH 7874 1.255 4.373 -232.5 95.927
SIZE 7874 17.969 2.078 7.603 28.558
TAX 6252 .00062 .0126 -.117 .5128
PROF 6285 14.320 2.131 6.555 23.674
Table 4.1.1 presents the descriptive statistics for all the variables including the dependent variable payout.
The dependent variable dividend payout ratio is calculated as the ratio of dividend per share to earnings per. The study uses the ratio of market-to-book value for growth and investment opportunities. Also log of total assets and log of net profits are used as firm size and profitability respectively, whiles taxes are measured as the ratio of corporate tax to income before tax.
From the table, the average dividend pay-out ratio (measured as dividend per share/ earnings per share) is 0.6788, which implies that all firms in the UK financial sector (per the sample of this study) pay about 67% of profits as dividends and the average risk (firm debt or leverage) is 2.255. Growth which is measured by the ratio of market to book value has an average value of about 1.255. The average values of size and profitability are 17.96 and 14. 32, which are quite high, compared to the other variables. Apart from corporate tax having the least observation (6252), the mean (0.00062), minimum and maximum (-0.1169 and 0.51282 respectively) values of look quite different and might raise few questions. A possible explanation is that though companies pay high corporation taxes, most of these payments are deferred to the preceding years, thereby rendering the average payment of tax by the firms small and even the minimum being negative.
4.2 Summary Statistics of Pay-Out Levels for Each Firm
Table 4.2.1: Pay-Out Levels for Each Firm.
Firms Obsv. Mean Std. Dev. Min Max
Investment Comp. 220 .56 .28 0 1.98
Speciality& other Fin. 4677 .78 .49 0 1.98
Invest. comp. (non-index) 1088 .72 .55 0 1.99
Banks 1591 .37 .44 0 1.94
Table 4.2.1 shows dividend payout levels for each of the firms that constitute the UK financial sector. Though all the firms do not pay dividends at all during certain periods (as shown by the minimum values), the highest average payout is about 78%, recorded by Speciality and Other Finance. Among all the firms, Banks paid the least profits as dividends (36%) during the study period. In general, UK financial institutions (as per the sample) pay about 67% of profits as dividends to their respective shareholders during the period under study.
4.3 Tests to check conditions of Regression
4.3.1 Autocorrelation:
The first assumption of regression model to test is to see if the covariance between error terms is zero. Thus, the error term should be random and it should not exhibit any pattern.
We run an auxiliary regression by estimating the residual on the regressors and lagged of the residuals (Breusch-Godfrey). The R-squared of the auxiliary regression is then multiplied by the number of observation to get the test statistic and compared it to the critical value on the chi square table. The result is shown below;
R-squared = 0.4314
Number of Observation = 3680
Test statistics = 0.4314*3680 = 1587.552
Chi2 (1, 5%) = 3.84, using lagged 1
Since 1587.552>3.84, we reject the null hypothesis of no autocorrelation.
4.3.2 Multicollinearity:
The problem of multicollinearity arises, when there is a linear relationship between the independent variables of regression. This will give biased and inconsistent estimates using OLS. The result of the multicollinearity test is shown on the table below.
Table 4.3.2.1: Multicollinearity Test
Variable VIF 1/VIF
size_ta 4.62 0.216292
prof 4.03 0.248188
risk 1.51 0.661333
growth 1.07 0.931445
tax 1.00 0.998506
Mean VIF 2.45
From the result, both VIF and 1/VIF do not look worrisome to guarantee further investigation of multicollinearity in the occurrence of the data. Thus, none of the variance inflator factor is greater than 10 to raise the issue of multicollinearity. However, the inclusion of time (year) in the panel regression might raise the issue of multicollinearity
4.3.3 Heteroskedasticity
The condition implies that the data should be homoskedastic. This means that the variance should be constant and same otherwise the data would be heteroskedastic. The result of a Breusch-Pagan test for the null of no heteroskedasticity is shown below
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of payout
chi2(1) = 54.12
Prob > chi2 = 0.0000
From the result, the p-value is small (actually 0), meaning we reject the null that the variance of the residuals is homogenous. Clearly, two assumptions of the classical linear (OLS) have been violated, therefore using only OLS will not be BLUE and the estimation method might give as inconsistent estimates-a good reason to use other types of regression.
Regression Results
This section first discusses the results of the individual independent variable on payout ratio, followed by a discussion of the general regression results
Table 4.4.1 shows a simple regression of the dependent variable against each of the independent variables one after the other. The R2 and P-values (significance) of each coefficient are first estimated to know which of the variables has greater influence and high explanatory power on the dependent variable.
Table: 4.4.1. Simple Regression Model
Dependent Variable Independent Variables R2 P>|t|
Risk 0.0534 0.000
Pay-out Tax 0.0207 0.082
Prof 0.0355 0.000
Size_ta 0.0926 0.000
Growth 0.0482 0.115
After observing the values of R2 (Coefficient of determination) and P values in the above table, we can say that firm size (measured by log of total asset) has significant impact on dividend pay-out in the UK financial sector during the sixteen years of study (1993-2009)
Among the five variables, firm size has the highest value for R2 (9.3%) and is statistically significant. Though some other variables are significant, they could not explain much of the variations in dividend payout (as firm size did) because they have relatively small R2 values.
Table 4.4.2 shows the various types of regressions of the dependent variable against the independent variables. The standard errors are those in parenthesis, whiles their respective coefficients are directly below them.
Table 4.4.2: Regression model results
OLS [12] RE [13] FE [14] Rreg [15] TOBIT [16] IV1 [17]
Risk -0.029*** -0.015*** -0.007* -0.020*** -0.031*** -0.013*
(0.002) (0.004) (0.004) (0.001) (0.003) (0.008)
Growth -0.009 -0.003 -0.001 -0.115*** -0.011 0.024
(0.008) (0.002) (0.001) (0.003) (0.009) (0.024)
SIZE 0.011 -0.001 -0.015 -0.068*** 0.003 0.007
(0.010) (0.013) (0.018) (0.005) (0.018) (0.048)
Tax -1.680** -0.668* -0.167 -2.120*** -1.868 -0.919
(0.652) (0.354) (0.181) (0.437) (1.368) (1.008)
PROF 0.044*** 0.065*** 0.073*** 0.098*** 0.063*** 0.074*
(0.009) (0.011) (0.014) (0.005) (0.016) (0.039)
Firm-years 4,020 4,020 4,020 4,019 4,020 3,323
R2 0.10 0.04 0.44 -0.10
A firms' level of profitability is an important influence on dividend payment. For profitability as a determinant of pay-out, all types of regressions show a positive relationship between pay-out and profitability. The coefficients are statistically significant at both the 1% and 5% level of significance. This means that high profitable firms pay more dividends and those with less profits will pay fewer dividends. Similar to the findings of Ferris et al (2006) [18] , this result suggest that firms in the UK financial sector are not different. Thus, highly profitable financial institutions tend to declare and pay more dividends to their shareholders. This result supports the initial hypothesis that there is a positive relationship between profitability and dividend payment of firms. Fama and French (2001) found in their research that dividend payment among US firms has fallen from 67% to 21% from 1978 to 1999 due to low levels of profitability recorded by firms. The result is also consistent with the findings of Pruitt and Gitman (1991), Amin and Abor (2006). They all documented positive and statistically significant relation between pay-outs and firm profitability.
Higgins (1981) argued that in the absence of financial constraints, firms will pay every surplus of funds as dividends. As stated by the null hypothesis, one expects a negative relationship between dividend pay-out and debt or leverage in the firm (defined here as risk). All the six types of regression results recorded negative coefficients of risk, indicating a negative relationship between firm or business risk and dividend payout. The results of the study show a negative and very significant (both at the 1% and 5% levels of significance) association between risk and dividend pay-out ratios, which suggest that, high-risked financial firms pay few dividends to their shareholders. Thus, firms with high amounts of debt in their capital structure and experiencing earning volatility usually pay fewer dividends as compared to less risky firms.
This results support the works of previous researchers and available literature on dividend pay-out for different countries and sectors. Kowalski et al (2007) indicated that more indebted firms prefer to pay fewer dividends. Brandon and Ikeberry (2004) in their study of US firms found that low debt and less risky firms pay higher dividends. Similar works by D'Souza (1999), Aivazian et al. (2003), Kowalski (2007) and many others pointed to one direction; a negative relationship between pay-out ratio and debt or beta coefficient of firms. The above regression results are therefore consistent with the findings of previous related works and study.
Generally, growing firms have a lot of investment opportunities, and will mostly pay fewer or no dividends, so one expects a negative association between growth and dividend payments. The study found a negative but statistically insignificant association between growth and pay-out using OLS, FE RE and TOBIT regressions. However, the ROBUST regression gave a negative and statistically significance association. This negative association is supported by the works of Ferris et al. (2006), who in their study of UK firms (excluding financial companies), found a regression analysis that suggest an inverse relation between dividend pay-out and corporate investment opportunities. Also similar works by Brandon and Ikeberry (2004), suggested a negative association between these two variables. A possible explanation is that growing financial firms have less access to funds (particularly equity) and their overall financial statues are much poorer than matured firms.
Quite inconsistent and strange is the direct (positive) relation shown between pay-out and growth by the IV1 regression. This contradicts the existing knowledge of pay-out and a firm's investment opportunities. Nevertheless, the mixed result and findings could be due to the different firms that make up the UK financial sector, as each of them may adopt firm specific pay-out polices. Besides, David and Igor (2008) in their cross country analysis found that the relationship between dividend payments and growth opportunities is not uniform; using countries like Germany, France, UK, Japan, US and Canada.
The next determinant is firm size as measured by total assets or market capitalisation of the firm. The general hypothesis and documented literature has been a positive association between these two variables. However most of the regressions of this study found an inverse relation between firm size and dividend pay-out ratios in the UK financial sector. Though IV1, TOBIT and OLS regressions showed positive coefficients, they were not significant as may be expected. Very inconsistent is the result of the ROBUST regression which has a negative and statistically significant association with dividend pay-out. The findings of Ferris et al. (2006), Eriotis (2005), Fama and French (2000) all suggest a positive relation, which implies the larger the size of a firm, the more likelihood that the firm will pay high dividends to its shareholders. The robust regression (which is significant at the 1% level) obviously, does not support the above findings. Though from the individual factor analysis, firm size explains much of the dependent variable, the relationship between them rejects the general null hypothesis. In effect, the UK financial sector (specifically to this data and period) does not follow the general pattern of dividend pay-out with respect to firm size as a determinant.
As expected and suggested by the null hypothesis, all the regression results indicated negative relationship between dividend pay-out and taxes. Except IV1, FE and TOBIT regressions, the remaining three regressions reported statistically significant coefficients. This means that increasing tax is associated in a decrease in dividend pay-out. This negative association of tax and the dependent variable is consistent with similar work and available literature. For instance Brandon and Ikeberry (2004) in their study of dividend policy among US firms recorded that, the introduction of the "Bush tax cut" saw corporations pay more tax to their shareholders. Other works by Singhania (2006) and Short et al (2006) indicate that corporations pay fewer dividends when they have more taxes to pay. The financial sector of UK is therefore not different from this dividend policy and corporate tax pattern.
CHAPTER FIVE
CONCLUSION AND RECOMMENDATIONS
5.0 Conclusion
The purpose of the study was to examine the determinants of dividend pay-out ratios of firms in the UK financial sector. We investigate to see if financial institutions in the UK have similar dividend payout policy pattern like other industries. The analyses are performed using data derived from the financial statements of firms for a 16-year period. Different types of regression analyses such as OLS fixed effect, random effect, tobit regression, robust regression and many others were used to estimate the relationship between pay-out and the independent variables. In general the analyses support the assertion that "dividend-paying companies are on average larger, more profitable, and less levered than non-dividend-paying firms". (Talavera O. et al, 2007)
The results show a positive and statistically significant relationship between dividend pay-out and profitability for all the regression equations estimated, which suggest that less profitable firms will pay fewer dividends. The relation between firm size and dividend payment is a mixed one. Whiles some of the estimations recorded a positive association, other estimation results were negative. As stated earlier, the negative associations are not consistent with existing literatures. The results also show negative association between dividend pay-out and risk (leverage). Thus, highly levered or risky firms usually pay less or do not pay dividends at all as they normally experience earnings volatility and the UK financial institutions are not exceptions to this pattern of dividend policy. Growing firms (financial companies) retain a greater percentage of their earnings for investment purposes unlike matured firms. In effect, firms with many investment opportunities may not pay dividends to its shareholders. This explains why the results showed an inverse relation between growth and dividend pay-out.
The results of taxes and dividend payment were not quite different from previous related studies. The study found a negative and significant association between corporate tax and dividend pay-out for all the estimation equations except instrumental variable regression. UK financial companies usually pay dividends after corporate taxes have been paid from their earnings. High tax rate will obviously reduce the profits-after tax which in turn, lowers the amount dividends pay to shareholders. The results of this study generally support previous empirical related studies. However, there are few inconsistent results and findings as illustrated by the estimations.
This study has added some findings to dividend policy with respect to financial institutions. Like any other industry, the dividend policy of UK financial institutions is influenced by profitability, firm size, taxes, growth or investment opportunities and risk or the amount of debt in the company. Managers and shareholders should be mindful that Changes in any of these determinants will obviously affect dividend policy and decision.
5.1 Recommendations
The study could be further extended in many directions for further research. Perhaps, much more important area of further research is the impact of corporate governance and institutional holdings on dividend policy of these financial institutions. Secondly, dividend policy in other firms like insurance and real estates can be explored.
