VARIABLE MEASUREMENT IN LIGHT OF EMPIRICAL LITERATURE This section will review the framework of dividend policy and discuss

several studies that tested dividend policy in different markets. This section formulates seven expected signs (hypotheses) to further examine the factors, which may affect corporate dividend policy. This section also explains the appropriate proxy variables used to measure the factors affecting dividend payouts. As regards the empirical literature the roots of the literature on determinants of dividend policy is related to Lintner (1956) seminal work. After this work the model is extended by the Fama and Babiak (1968). A number of factors have been identified in previous empirical studies to influence the dividend payout ratios of firms including profitability, risk, cash flow, agency cost, and growth (see Higgins, 1981; Rozeff, 1982; Lloyd et al., 1985; Pruitt and Gitman, 1991; Jensen et al., 1992; Alli et al., 1993; Collins et al., 1996; DSouza, 1999). DSouza (1999) finds negatively relationship between agency cost and market risk with dividends payout. However, the result does not support the negative relationship between dividend payout policies and investment opportunities. DeAngelo et al. (2004) document highly significant association between the decision to pay dividends and the ratio of earned equity to total equity controlling for size of the firm, profitability, growth, leverage, cash balance and history of dividends. Baker et al. (2007) reports that Canadian dividend paying firms are significantly larger and more profitable, having greater cash flows, ownership structure and some growth opportunities Dependent Variable Dividend Payout Ratio Payout ratio is calculated by dividing the total dividend to net profit of every stock. We have calculated net profit and dividends of each company individually for every year in order to control the problem of extreme values in individual year that lead the results to low or negative net income. Most of the previous studies employed dividend payout ratios as a determinant of dividend in lieu of dividend per share and dividend yield ( Rozeff, 1982; Lloyd,1985; Jensen et al., 1992; Dempsey and Laber, 1992; Alli et

al., 1993; Mohd et al., 1995; Holder et al., 1998; Chen et al., 1999; Saxena, 1999; Mollah et al., 2002; Manos, 2002; Travlos, 2002). The dividend payout ratio is also used in this research, rather than dividend per share and dividend yield, for two reasons. Firstly, the dividend payout ratio takes into consideration both dividend payout and dividend retention. Secondly, dividend per share and dividend yield was considered unsuitable, because neither takes into account the dividend paid in relation to the income level. Independent variables Current or Anticipated Earnings The empirical analysis by Adaoglu (2000) shows that the firms listed on Istanbul Stock Exchange follow unstable cash dividend policy and the main factor for determining the amount of dividend is earning of the firms. Eriotis (2005) reports that the Greek firms distribute dividend each year according to their target payout ratio, which is determined by distributed earnings and size of these firms. In investigating the determinants of dividend policy of Tunisian stock Exchange Naceur et al. (2006) find that the high profitable firms with more stable earnings can manage the larger cash flows and because of this they pay larger dividends. Moreover, the firms with fast growth distribute the larger dividends so as attract to investors. Baker et al. (1985) also find that a major determinant of dividend payment was the anticipated level of future earnings. Alli et al. (1993) reveal that dividend payments depend more on cash flows, which reflect the companys ability to pay dividends, than on current earnings, which are less heavily influenced by accounting practices. They claim current earnings do not really reflect the firms ability to pay dividends. The proxy used for earnings is the ratio of companys operating earnings before interest and tax (EBIT) to total assets. Hypothesis 1 Dividend Payout is positively associated with current earnings. i. Corporate Tax Omet (2004) comes to the same conclusion in case of firms listed on Amman Securities Market and further the tax imposition on dividend does not have the significant impact on the dividend behaviour of the listed firms. In Indian case Reddy (2006) show that the dividends paying firms are more profitable, large in size, and growing. The corporate tax or tax preference theory doesnt appear to hold true in Indian context. Corporate tax has been taken as an explanatory variable with the expected negative association with dividend payout by Anil and Kapoor (2008) in their study on IT sector of India and found it to be insignificant, consistent with Reddys conclusion. It would be interesting to note the effect of corporate tax in Pakistan . Hypothesis 2 Dividend Payout is negatively associated with corporate tax. ii. Growth Opportunities Higgins (1972) shows that payout ratio is negatively related to a firms need for funds to finance growth opportunities. DSouza (1999) however shows a positive but insignificant relationship in the case of growth. Rozeff (1982), Lloyd et al. (1985), and Collins et al. (1996) all show a significantly negative relationship between historical sales growth and dividend payout. Higgins (1981) indicates a direct link between growth and financing needs: rapidly growing firms have external financing needs because working capital needs normally exceed the incremental cash flows from new sales. Growth rate is measured as the growth rate of sales ( Rozeff, 1982; Lloyd et al., 1985; Jensen et al., 1992; Alli et al., 1993; Mohd et al., 1995; Holder et al., 1998; Chen etal., 1999; Sexsena, 1999; Manos, 2002; Travlos, 2002). Thus, growth rate has been identified in this study by Annual Sales Growth. Overall literature portrays a negative as well as a positive relationship between the dependent variable

and sales growth. Hypothesis 3 Dividend Payout is positively/negatively associated with growth opportunities.

iii. Firm Size Eriotis (2005) reports that the Greek firms distribute dividend each year according to their target payout ratio, which is determined by distributed earnings and size of these firms. Research by Lloyd, Jahera, and Page (1985), and Vogt (1994) indicates that firm size plays a role in explaining the dividend-payout ratio of firms. They find that larger firms tend to be more mature and thus have easier access to the capital markets, which reduces their dependence on internally generated funding and allows for higher dividend-payout ratios. The hypothesized relationship between firm size and dividend-payout ratios is positive. Firm size (SIZE) is measured as a natural logarithm of total assets. This is due to the fact that large firms will pay large dividends to reduce agency costs (Ghosh and Woolridge, 1988; Eddy and Seifert, 1988; Redding, 1997). Eddy and Seifert (1988), Jensen et al. (1992), Redding (1997), and Fama and French (2000) indicated that large firms distribute a higher amount of their net profits as cash dividends, than do small firms. Several studies have tested the impact of firm size on the dividend-agency relationship. Lloyd et al. (1985) were among the first to modify Rozeff's model by adding firm size as an additional variable. They considered it an important explanatory variable, as large companies are more likely to increase their dividend payouts to decrease agency costs. Their findings support Jensen and Mecklings (1976) argument, that agency costs are associated with firm size. Holder et al. (1998) revealed that larger firms have better access to capital markets and find it easier to raise funds at lower costs, allowing them to pay higher dividends to shareholders. This demonstrates a positive association between dividend payouts and firm size.The positive relationship between dividend payout policy and firm size is also supported by a growing number of other studies (, Eddy and Seifert, 1988; Jensen et al., 1992; Redding, 1997; Holder et al., 1998; Fama and French, 2000; Manos, 2002; Mollah 2002; Travlos et al., 2002; Al-Malkawi, 2007). Al Kuwari (2009) too found a significantly positive relationship between the two. Hypothesis 4 Dividend Payout is positively associated with firm size. iv. Financial Leverage A growing number of studies have found that the level of financial leverage negatively affects dividend policy (Jensen et al., 1992; Agrawal and Jayaraman, 1994; Crutchley and Hansen, 1989; Faccio et al., 2001; Gugler and Yurtoglu, 2003; Al-Malkawi, 2005). Their studies inferred that highly levered firms look forward to maintaining their internal cash flow to fulfil duties, instead of distributing available cash to shareholders and protect their creditors. However, Mollah et al. (2001) examined an emerging market and found a direct relationship between financial leverage and debt-burden level that increases transaction costs. Thus, firms with high leverage ratios have high transaction costs, and are in a weak position to pay higher dividends to avoid the cost of external financing. To analyze the extent to which debt can affect dividend payouts, this study employed the financial leverage ratio, or ratio of liabilities (total short-term and longterm debt) to total shareholders equity. Al Kuwari (2009) too found a significantly negative relationship between the two. The proxy used for financial leverage is Debt to Equity Ratio as used in all these studies. Hypothesis 5 Dividend Payout is negatively associated with financial leverage. v. Profitability The financial literature documents that a firms profitability is a significant and explanatory variable of dividend policy (Jensen et al., 1992; Han et al., 1999; Fama and French, 2000). However, there is a significant difference between dividend policies in developed and developing countries. This difference has been reported by Glen et al. (1995), showing that

dividend payout rates in developing countries are approximately two-thirds of those in developed countries. Moreover, emerging market corporations do not follow a stable dividend policy; dividend payment for a given year is based on firm profitability for the same year. Profitability (PROF) is the ratio of net profits to the amount of money that shareholders

have put into the company. ROE has been used in several studies as a proxy for firm profitability (Aivazian et al., 2003, ap Gwilym et al., 2004.) and is calculated as follows: PROF = (Net profit/shareholders equity)*100 This creates the assumption that the dividend ratio per year is based on firm earnings for the same year. Amidu and Abor (2006) find dividend payout policy decision of listed firms in Ghana Stock Exchange is influenced by profitability, cash flow position, and growth scenario and investment opportunities of the firms. Profits have long been regarded as the primary indicator of a firms capacity to pay dividends. Pruitt and Gitman (1991), in their study report that, current and past years profits are important factors in influencing dividend payments. Al Kuwari (2009) too found a significantly positive relationship between the two. Hypothesis 6 Dividend Payout is positively associated with profitability. THEORETICAL FRAMEWORK To investigate the six variables identified in this study associated with the impact on dividend payout ratio of KSE listed companies, this study undertook an empirical testing of a model with the following framework: DIVIDEND PAYOUT = f (EARNINGS, TAX, GROWTH, SIZE, LEV, PROF) From the empirical literature, the following expected relationship of dividend payout ratio with the independent variables has been concluded: Insert table 1 & figure 1 here DATA COLLECTION The Data of 53 non financial listed companies was selected for 6 years (2005-2010). The reason for selection for 6 years time span was that one business cycle is completed in 5-7 years. The reason to select these 53 companies from 11 sectors was consistent dividend payment in the past 6 years. The source of data was Balance Sheet Analysis by SBP for 2004-2009 & 2005-2010. Hence the data is totally transparent in context of authenticity. Softwares of Eviews 3.1 and Microsoft Excel 2007 were used for data analysis. A detail of these companies is given in the appendix later on. Insert table 2 here ECONOMETRIC METHODOLGY & EMPIRICAL RESULTS For the analysis of panel data for six years i.e. 2002 to 2011, the following steps were taken on the data: Tests to check conditions of Regression Multivariate Regression

Tests to check conditions of Regression Three econometric problems are being tested here to check for violation of assumptions of regression, namely:

B.I. Heterokedasticity: The condition of classic linear regression model implies that there should be homoskedasticity between variables. This means that the spread should be constant and same. Variance of residuals should be constant otherwise, the condition for existence of regression, homoskedasticity, would be violated and the data would be heteroskedastic. To check for this, white heteroskedasticity test was run for the residuals. As the probability is more than 5%, hence it implies absence of heteroskedasticity and the data has uniform spread. Insert table 4 here B.II. Autocorrelation: Another basic assumption of regression model says that the covariance between error terms should be zero. This means that error term should be random and it should not exhibit any kind of pattern. Cov (t , s) = 0 If there exists covariance between the residuals and it is non zero, this phenomenon is called autocorrelation. To test for autocorrelation, three methods can be used. We apply all three here. B.II.i. Graphical Method The following graph clearly implies that the residuals show a random behavior. Thus implying absence of covariance and hence autocorrelation. Insert figure 2 here B.II.ii. Durbin Watson Stat Durbin Watson Statistics if found closer to 2, implies absence of autocorrelation. The calculated value is 1.8, which is closer to 2. Hence again proving absence of autocorrelation and the residuals to be random. Insert table 5 here B.II.iii. ARCH LM Test Finally, the ARCH LM test was run and the found value of probability of autocorrelation existence was 0.89, which is way beyond the significance level of 5%. Hence the covariance between residuals is zero, data is normal and absence of autocorrelation problem was found conclusively from all three tests. Insert table 6 here B.III. Multicollinearity: The problem of multicollinearity arises, when there exists a linear relationship between the explanatory variables of regression. This results in biased and inconsistent estimates of OLS as well as Beta. T-statistics are also affected and biased that leads to biased hypothesis testing. Multicollinearity can be econometrically explained as:

Multicollinearity can be detected through 2 methods. One is checking the level of correlation among the variables in correlation matrix. While the other is checking for the value to be in tolerable limit through variance inflationary function test. Both tests have been done here and interesting and mixed results were found.

B.III.i. Correlation Matrix The correlation matrix below shows that there is reasonable relationship among all the variables. The variance-covariance among variables is not too high except for profitability and earnings which is on the

higher side (76.66%). This indicates that both these variables represent the same economic phenomenon. Both these commove. To avoid the problem of multicollinearity in regression, we can run the regression with the two variables to be mutually exclusive. This means testing regression in isolation once with profitability and once with earnings so as to get unbiased and consistent results. Insert table 7 here B.III.ii. Variance Inflationary Function The benchmark for tolerable limit is 5. Thus value higher than 5 falls in intolerable limit and hints at the existence of multicollinearity. But the VIF for the given variables is well within the benchmark value of 5 hence implying absence of multicollinearity in the overall model.

A) Multivariate Regression = This study employs multivariate regression analysis of panel data to measure the impact of the different factors on the companys capital structure. The panel data analysis helps to explore cross-sectional and time series data simultaneously. Since the conditions of regression are being completely fulfilled, so it is an appropriate tool to work with. Thus the results that would be achieved through regression can be safely said to be efficient, unbiased and consistent. In this regard, an attempt was made to develop a multiple regression equation using identified key variables. The dividend payout (Y) was used as dependent variable and other variables (x1 ,x2, x3, x4, x5, x6) were used as independent variables. On this basis, under mentioned multiple linear regression equation was developed. Where: Dependent Variable Y = Dividend Payout Ratio Independent Variables X1 = Current Earnings X2 = Sales Growth X3 = Firm Size X4 = Financial Leverage X5 = Profitability X6 = Corporate Tax = Intercept or Regression Constant = Regression Coefficient for x1 = Regression Coefficient for x2 = Regression Coefficient for x3 = Regression Coefficient for x4 = Regression Coefficient for x5 = Regression Coefficient for x6 The general form of model is: = TL = 0 + 1 PR + 2 SZ + 3 GR + 4 UN + 5 OR + 6 AGE

The above model tests the following null hypothesis there is no significant impact of the factors upon the capital structure choice. The dependent variable is the leverage (total leverage) and the independent variables are the different factors previously determined (Asset Structure, Profitability, Size, Expected Growth, Uniqueness, Operating Risk and the Age of the company).

The regression coefficients indicate the amount of change in the value of dependent variable for a unit change in independent variable. Since mixed results were derived for multicollinearity, correlation matrix implying a chance of existence of multicollinearity between profitability and earnings and VIF implying absence of any multicollinearity in the variables under study. Hence, to be on the safe side, regression was run thrice; once with profitability and earnings, both being part of the explanatory variables data set simultaneously, next time taking only earnings and excluding profitability from the explanatory variables and lastly including profitability in the absence of earnings. All 3 regression results were found to be consistent with the same results of significance. These are reported as under: Insert table 8, 9 &10 here From the regression results above, it can be clearly observed, that out of the 6 explanatory variables under study, only Firms Size and Corporate Tax were found to be variables having significant impact on dividend payout. The probability of both these variables is within 5% benchmark probability level. Hence

it can be said with 95% confidence level that these variables play a significant role in determining the dividend payout in Pakistan. The observed values of T-Statistics are also more than the tabulated t-stat for both firms size and corporate tax, reinforcing the level of significance of probability. Thus if there is 1% change in firm size, it would determine up to approximately 5% change in dividend payout whereas one unit change in corporate tax would determine up to 87-90 units change in the dividend payout policy in Pakistan. Interestingly both explanatory variables are found to have positive relationship with the dependent variable of dividend payout. The results report insignificant relationship of earnings, sales growth, financial leverage and profitability with dividend payout decision in Pakistan. The hovers around 4.5% mark whereas adjusted around 3% mark for all 3 regression results. This shows low explanatory power of the variables in determination of dividend payout in this model. This implies addition of more variables to improve the explanatory strength of the model in Pakistani context. It is noteworthy to report here that these study was also carried out in panel data analysis under all three models of common effect, fixed effect (using industrial dummy) and random effect models and their results were found to be consistent with the above mentioned regression results.

S. No.

Variable

Description

Expected Relationship/Hypothese s

DEPENDENT VARIABLE Dividend / Net Dividend Payout Ratio Profit INDEPENDENT VARIABLES Earnings before Current 1 Earnings Total Assets Corporate Tax / 2 Corporate Tax Profit before Tax New Sales - Old 3 Sales Growth Sales / Old Sales Natural log (Total 4 Firm Size Assets) 5 Financial Leverage Total Debt / Equity NEGATIVE POSITIVE POSITIVE / NEGATIVE NEGATIVE Interest & Tax / POSITIVE

Net Profit / Equity * 6 Profitability 100 POSITIVE

Earnings Corporate Tax Sales Growth Firm Size Financial Leverage Profitability

Dividend Payout

Independent Variables

Dependent Variable

Business Implication of Dividend Pay-Out Study The study finds a negative relationship between dividend pay-out and profitability which generally predicts that profitable telecom companies dole out lesser profit to shareholders. This supports the findings of Baker and Powells study in which they concluded that dividend determinants are industry specific. There is a positive as well as significant relationship between cash flow and dividend pay-out which confirms higher cash flow results in more dividend for shareholders. There is a positive relationship between dividend pay-out and debt-to-equity which is significant at 10%. If the proportion of debt goes higher in the capital structure, then the company can efficiently take care of the interest of shareholders since shareholders contribution to the overall capital remains fixed.