What determines dividend policy: A comprehensive test"article"

What determines dividend policy: A comprehensive test
Tao Zeng. Journal of American Academy of Business, Cambridge. Hollywood: Mar 2003. Vol. 2, Edisi 2; pg. 304, 6 pgs
Abstrak (Ringkasan)
This paper designs an empirical work to investigate the determinants of corporate dividend policy under the Canadian situation. It shows that firms pay dividend as a signal and to reduce agency costs. It also shows that liquidity and tax clientele effect are related to dividend policy.
» Lompat ke pengindeksan (detil dokumen)
Teks Lengkap (3792 kata)
Copyright Journal of American Academy of Business Mar 2003
[Headnote]
ABSTRACT



[Headnote]
This paper designs an empirical work to investigate the determinants of corporate dividend policy under the Canadian situation. It shows that firms pay dividend as a signal and to reduce agency costs. It also shows that liquidity and tax clientele effect are related to dividend policy.



1. INTRODUCTION
There has been a great deal of financial, economic, as well as accounting literature analysing why firms pay dividends, given the fact that effective tax rate on capital gains lower than the effective tax rate on dividends, i.e., the Adividend puzzle@ (Holder et al 1998, Dhaliwal et al 1995, Lamoureux 1990, Chaplinsky and Seyhun 1990, Abrutyn and Turner 1990, Mann 1989, Crockett and Friend 1988, Kose and Williams 1985, Feldstein and Green 1983, Litzenberger and Ramaswamy 1982, 1979, Miller and Scholes 1982, Feldstein 1970, and so on). To shed light on this puzzle, researchers try to figure out the benefits from paying dividends, which may offset the tax disadvantages. Some survey studies find that CEOs choose to pay dividends because they believe dividend can server as a signal to shareholders (Baker and Powell 1999, Abrutyn and Turner 1990, Baker et al 1985); because dividend can reduce agency costs and enforce manager to act in the interest of shareholders (Abrutyn and Turner 1990); because clientele effects exist (Baker et al 1985). Some event studies on signalling effect attempt to test whether a positive equity price response is associated with unexpected dividend increase, or vice visa. Several studies present evidence consistent with this argument (Dielman and Oppenheimer 1984, Aharony and Swary 1980, Charest 1978, Pettit 1977, 1972). Some studies, however, do not find the evidence indicate that dividend changes reflect no more information than that reflected in earnings (Gonedes 1978). Tax clientele studies (Dhaliwal et al 1995) argue that the ownership by tax-exempt or tax deferred investors will increase when firms begin to pay dividends. This study differs from prior researches on 3 important ways. First, this paper examines the relationship between firm-specific characteristics and dividend policy. Mann (1989) argued that studies should go beyond event studies around dividend announcement days since there may exist underlying factors other than dividend itself that drive the change of return around dividend announcement. Second, this study designs the tests using corporate financial data, rather than taking a survey study. The argument from survey may provide reasons justifying the managers' behaviour afterwards, rather than a motive beforehand. In addition, survey research involves non-- response bias, and the bias is severe when the response rate is low. Third, this study makes a comprehensive test of the determinants of dividend policy. Prior researches usually focus on only one factor, e.g., signalling (Dyl and Weigand 1998, Brook et al 1998, Bernheim and Wantz 1985), agency cost (Born and Rimbey 1993, Crutchley and Hansen 1989, Easterbrook 1984), tax clientele (Dhaliwal et al 1995, Scholz 1992), or investment opportunity (Gaver and Gaver 1993, Smith and Watts 1992). It is argued that dividend policy may be a decision based on the combination of many factors inside and outside. The remainder of this paper is organized as follows. In section 2, I review the literature relevant to the determinants of firms dividend policies. In section 3, I provide the empirical test method, data collection and variable measurement. In section 4, the test results are presented. Finally, I summarize and conclude in section 5.
2. THE FACTORS EFFECTING DIVIDEND POLICY
Given the effective tax rate on capital gains lower than that on dividends, researches have been taken to figure out the benefits from paying dividends, which may compensate the tax disadvantages. The benefits from pay dividends or the reasons justifying dividend payoff are summarized as follows:
2.1. Tax clientele
One argument of why firms pay dividends involves tax effect. Shareholders receive and are taxed the returns to shares either as dividends or capital gains. Dividends and capital gains are taxed differently among various types of investors, individual investors, corporate investors, and tax-exempt or tax-deferred investors. Tax clientele hypothesis argues that tax clienteles prefer different dividend policies, and investors may attach to firms that have dividend policies appropriate to their particular tax circumstance. For example, corporate investors, whose dividends are taxed at a lower rate than capital gains, may prefer high dividend payout ratio; on the other hand, high-income individual investors, whose dividends are taxed at a higher rate than capital gains, may prefer low dividend payout ratio. The tests in this study for tax clientele hypothesis are that: positive relationship between firm dividend payoff and the ownership of corporate investors and tax-exempt or tax-deferred investors; a negative relationship between firm dividend payoff and the ownership of high-income individual investors.
2.2. Agency cost
Another argument of why firms pay dividends is that dividends provide a mechanism for restricting managerial discretion. It reduces the agency costs of free cash flow by cutting down the cash available for spending at the discretion of management, and hence provide some protections to the firm against management that might benefit itself at the shareholders= expenses. If firm size is positively related to diversification and decentralization, then the lager the firm, the less observable the actions of management and the higher agency costs may be incurred. Hence I may expect large firms pay dividends. Also, Gaver and Gaver (1993) argue that when the firm growth opportunities increase, the observability of managerial action decrease. It is difficult for outsiders like shareholders, without inside information and specified knowledge of managers, to ascertain the growth opportunities available to the firm. In contrast, maintenance and supervision of existing assets are more readily observable. If the managerial actions are less observable, the agency costs may be higher, and we expect the firm may rely on dividends to restrict managerial discretion. Hence we expect growth firms pursue a high dividend payout policy.
2.3. Signalling
Signalling hypothesis argues that dividends are paid to communicate information to investors about firms future prospects. Under information asymmetry, insiders are better informed than outsider investors. If managers have information that outside investors do not have, they may use dividends as a way to signal this private information and reduce information asymmetry. Dividends may affect firm market value because they may signal favourable inside information. Hence insider will pay dividends to communicate information and achieve a higher market price for the firm-s stocks than would otherwise prevail. To serve as a signal, dividend payout must provide the market with the useful information which can not be conveyed by alternative forms of communication, e.g., current earnings and cash flow. The test in this study is to examine whether dividend payoff provide an improved estimate of firm's stock price.
2.4. Corporate liquidity
Liquidity requirement may also affect firms= dividend decisions. A high degree of liquidity might be expected to encourage dividends by enabling high dividends to be paid without resort to external finance. Firms' free cash flows may be used as an indicator of firms' liquidity, and the test is that high degree of free cash flow encourage high dividend payments. If leverage is used as one indicator of the future default, and positively related to the costs of financial costs, when the firm's leverage ratio is high, and dividend payoff may increase financial distress, firms may reluctant to pay dividends. To increase liquidity, firms might lower dividend payments. The test in this study is the inverse relationship between firm leverage and dividend payoff. Also, if firms have significant opportunities, they may undertake the opportunities rather than paying dividends. Hence growth firms are expected to pursue a low dividend payout policy (Gaver and Gaver 1993).
3. TEST METHOD, DATA COLLECTION AND VARIABLES
3.1. Data collection and variables
I collect data on the "Canadian Financial Post Card" for the years of 1984-88. The sample consists of all firms that meet the following criteria: (1) availability of accounting data on the Financial Post Card for the time period of 6 years from 1984 to 1988; (2) not in the financial, insurance, or real estate industries. There are 313 companies, i.e., 1565 observations. The variables used to test the propositions are as follows: To test tax clientele hypothesis, I have to specify the corporate investor, tax-exempt or tax-deferred investors, and high-income individual investors. The institutional investors may represent the corporate investor and tax-exempt or tax-deferred investors. The institutional investors include companies, banks, insurance, pension plans, and trust investors. I look up the major shareholders for each firm around 1986 on the financial post card, and calculate the percentage of major institutional shareholders to all shareholders. I also calculate the percentage of major individual shareholders to all shareholders. It is argued that, if an individual is a major shareholder, he may be expected to be in a high-- income bracket. The hypothesis is that dividend payout is positively related to the percentage of major institutional shareholders, and negatively related to the percentage of major individual shareholders. There are 3 variables to measure dividend policy: (1). dividend paid or declared per share, (2). dividend payout ratio (dividend per share divided by after-tax earnings per share), and (3). dividend yield (dividend per share divided by share price).
1. is the dividend paid or declared per share, calculated by the total common stock dividend payment divided by total common share outstanding. To make it comparable over the five year time periods, the dividend per share calculated above will be multiplied by the split number, when firms split shares. For example, for a 2-for-1 split, (1) is calculated by the total common stock dividend payout divided by total common shares outstanding (after split), times 2.
2. is calculated by total common stock dividend payment divided by total common share outstanding, divided by earnings after tax and before extraordinary items divided by total common share outstanding. It is the same as the total common stock dividend payment divided by earnings after tax and before extraordinary items. To use this measure, negative earnings should be deleted.
3. is calculated by total common stock dividend payout divided by total common shares outstanding, divided by share price. Share price is measured as the average high price and low price in the year. The high and low prices are obtained from the historical data statistics on the financial post card.
In this study, I use (1) and (3) measures since using (2) has to delete the observations with negative earnings and hence lose some observations. Firm size is measured as the log of operating revenue. The agency cost hypothesis is that firm size is positively related to dividend payments. One liquidity indicator is firm free cash flow, which is measured as the cash flow from operation. It is deflated by total asset value to avoid size effect. The liquidity hypothesis suggests that free cash flow is positively related to dividend payout. Another liquidity indicator is investment/cash flow ratio: the change of capital properties (land, building, and equipments), divided by cash flow from operation. If a firm with low cash flow level makes a great deal of investment in fixed assets, it may incur liquidity problem. The liquidity hypothesis suggests that high investment/cash flow ratio disencourages dividend payout. Leverage is measured as the sum of short term and long term debt, divided by total asset value. The liquidity hypothesis suggests that leverage is negatively related to dividend payments. Firm growth has a mixed prediction. On the one hand, firms with investment opportunities may incurred higher agency costs since monitoring managerial actions of choosing new investment projects is more difficult for outside shareholders than maintaining existing assets. Agency hypothesis suggests that growth firms may have higher dividends. On the other hand, firms with promising investment opportunities may undertake these opportunities rather than paying dividends. Given the assumption of investment and dividends are linked through the firm=s cash flow identity, the liquidity hypothesis expects lower dividend payments with growth firms. I following Smith and Watt (1992) and measure growth opportunity as the market-to-book ratio: firm market value divided by total asset value. They argue that the lower the market-to-book ratio, the higher the ratio of asset in place to firm value, and the lower the ratio of growth opportunities to firm value.
3.2. Test method
First, I classify firms into two groups: those paying dividends (242 firms), and those not paying dividends since 1980 (71 firms). I compare the means for theses two groups. The significance of the difference in the means for these two groups is assessed using a t-test. For each group, we regress share price on current after-tax earnings and cash flow, and compare the R-- squared values. For firms paying dividend, I also regress share price on current after-tax earnings and cash flow as well as current dividend payout, and compare the R-square with that obtained from the regression without current dividend payout. The null hypothesis is that the coefficient on the variable of current dividend payout is zero. Then I design a following linear regression model, where I regress dividend per share (or dividend yield) on institutional ownership, individual ownership, firm size, growth, leverage, and liquidity. To test tax clientele hypothesis, I use two variables: institutional ownership and individual ownership. To test agency theory, I use one variable: firm size. To test liquidity hypothesis, I use two variables: cash flow and leverage. I do not use the investment/cash flow ratio, because to use this variable, I have to delete the observations with negative cash flow, and hence lose some observations (106 observations). The growth variable can be used to test either the agency theory or the liquidity hypothesis, and its predicted sign is mixed. When I use dividend yield as dependent variable, I drop the market-- to-asset ratio variable. This is because the market-to-asset ratio relies on share price, and the inverse relationship between dividend yield and share price make it sensitive to the dividend policy of the firm.
4. RESULTS (The results are shown in table 1-6)
4.1. Tax clientele
Table 1 shows some evidence consistent with the tax clientele hypothesis: firms paying dividends have significantly higher institutional ownership than firms not paying dividends, and the individual ownership for pay-- dividend firms is lower but not significantly than non-paying-dividend firms.
4.2. Agency cost
Table 1 shows that the firms paying dividends is significantly larger than the firms not paying dividends. Table 3 and table 4 show that firm size is significantly positively related to dividend per share or dividend yield. If firm size is positively related to agency cost, then larger firms will have larger agency cost and more likely to resort to paying dividends to reduce this cost.

Perbesar 200%
Perbesar 400%

Table 1.
Table 2.
Table 3 and 4



4.3. Signalling
Table 1 shows that current earnings and cash flow can estimate share price much better for firms not paying dividends than firms paying dividends. The R-squared value for paying-dividends firms is 0.1419, while the R-- squared value for not-paying-dividend firms is 0.3086, about 2.2 times as large as paying-dividend firms. I also regress share price on current earnings, cash flow, and dividend per share for the firms paying dividends. The R-- squared value is 0.2836, which is two times as large as the R-squared value from the regression not using dividend variable. The null hypothesis that the coefficient on the current dividend per share variable is zero, is rejected at 0.001 level (F-test value=79.4772). I argue that the current accounting data i.e., earnings and cash flow for paying-- dividend firms do not predict share price as well as non-paying-dividend firms. This may justify the firms using dividend to communicate some relevant information that is not conveyed by current earnings and cash flow. When I add dividend payoff as an independent variable for the pay-dividend firms, the R-squared value is significantly improved, and dividend payoff can explain share price as well as the combination of earnings and cash flow.
4.4. Liquidity
Table 1 shows the evidence that strongly supports the liquidity hypothesis. It shows that pay-dividend firms have significantly lower investment/cash flow ratio and leverage (i.e., less liquidity problem) than non-pay-- dividend firms. Pay-dividend firms also have significantly higher level of free cash flow than non-paying-dividend firms. The regression results from Table 3 and 4 also support the liquidity hypothesis. It is shown that leverage is significantly negatively related to dividend per share and dividend yield; cash flow is significantly positively related to dividend payout measured either as dividend per share or dividend yield. A caveat is that the free cash flow variable may measure some agency problems since management with more free cash flow may has higher probability of discretionary use of this cash flow. The positive relationship between the free cash flow and dividends may confirm the agency theory, i.e., firms with high agency costs use dividends to bond managerial behaviour. Therefore, this positive relationship is consistent with either the liquidity hypothesis or the agency theory.
4.5. Other results
The growth variable can either be used to test the agency theory or be used to test the liquidity hypothesis. Tables 1 shows that, growth firms, measured as market-to-book ratio, do not pay dividends. It is consistent with the liquidity hypothesis. However, Table 2 shows the coefficient on the market-to-book ratio is not significant. Table 5 presents the correlation matrix of the independent variables. The smallest correlation is -0.5590 between institutional ownership and individual ownership, and the largest correlation is 0.1777 between liquidity and firm size. Only one correction have an absolute value larger than 0.3. This suggests that multi-collinearity is not a problem.
I also classify the time periods into two periods: 1984-1986, and 1987-1988, and test whether the 1987 tax reform changes the relationship between the independent variables and dividend policy. The regression results, which are not presented in this paper, show that the results do not change.
5. CONCLUSION
This paper examines the determinants of dividend policy in the Canadian situation. It is argued that there may exist benefits for paying dividends, in order to compensate the tax disadvantage. Paying dividends may communicate additional information about firms= future profitability - signalling hypothesis. Dividends may be paid to tax-exempt or tax-deferred investors, or investors with lower effective tax rate - tax clientele hypothesis. Dividends may also be used to bond managerial discretionary behaviour - agency theory. Finally, firms may pay dividends if no investment opportunities available - liquidity hypothesis. The empirical tests in this studies support the signalling hypothesis, the agency theory, and liquidity hypothesis. It also shows some evidence supporting the tax-clientele hypothesis, but the evidence is not strong. One improved test is to collect the institutional and individual ownership data year by year, and look at all shareholders rather than the major shareholders. Another extension of the study is to use a different time periods. This study uses the data for five years from 1984-1988. There are two reasons to choose this time period: (1) Cross-sectional test using only year data may not be appropriate because dividend payout policies may take time and be costly to achieve. (2) we seek to test whether the 1987 tax reform changes the hypothesises on the determinants dividend policy. Future study can be taken to collect data for the more recent years, and examine whether those hypothesises on the determinants of dividend policy are changed with the rapidly developed technology.

Perbesar 200%
Perbesar 400%

Table 5.



[Sidebar]
The Author gratefully acknowledges that financial support for this research was received from a grant funded by WLU Operating funds, the SSHRC Institutional Grant awarded to WLU, and CMA Canada - Ontario.



[Referensi]
REFERENCE



[Referensi]
Abrutyn, S. and R.W. Turner, Taxes and Firms Dividend Policies: Survey Results, National Tax Journal 43, 491-497 (1990).
Aharony, J. and 1. Swary, Quarterly Dividend and Earnings Announcements and Stockholders= Returns: An Empirical Analysis, Journal of Finance, 1-12 (1980).
Ambarish, R., Efficient Signalling with Dividends and Investment, Journal ofFinance 42, 321-343 (1987).
Baker, H.K., G.Farrelly and R.Edelman, A Survey of Management Views on Dividend Policy, Financial Management 14, 78-84 (1985).
Baker, H.K. and G.E. Powell, How Corporate Managers View Dividend Policy, Quantitative Journal of Business and Economics 38(2), 17-35 (Spring 1999).
Bernheim, B.D. and A.Wantz, A Tax-based Test of the Dividend Signalling Hypothesis, The American Economics Review 85(3), 532-551 (1985).
Born, J.A. and J.N. Rimbey, A Test of Easterbrook Hypothesis Regarding Dividend Payments and Agency Costs, The Journal of Financial Research 16(3), 251-260 (Fall 1993).
Chaplinsky, S. and H.N. Seyhun, Dividends and Taxes: Evidence on Tax-Reduction Strategies, Journal of Business 63(2), 239-258 (1990).
Charest, G., Dividend Information, Stock Returns and Market EfficiencyB II, Journal ofFinancial Economics, 297-330 (1978).
Crockett, J. and I. Friend, Dividend Policy in Perspective: Can Theory Explain Behaviour?, The Review of Economics and Statistics, 603-613 (1999).



[Referensi]
Crutchley, C.E. and R.S. Hansen, A Test of the Agency Theory of Managerial Ownership, Corporate Leverage, and Corporate Dividends, Financial Management 18(4), 36-46 (Winter 1989).
Dhaliwal, D.S., M. Erickson, and R. Trezevant, A Test of the Theory of Tax Clienteles for Dividend Policies, National Tax Journal 47, 179-194 (1995).
Dielman, T. and H. Oppenheimer, An Estimation of Investor Behaviour During Periods of Large Dividend Changes, Journal of Financial and Quantitative Analysis, 197-216 (1984).
Easterbrook, F., Two Agency Cost Explanations of Dividends, American Economics Review 74, 650-659 (1984). Feldstein, M and J. Green, Why do Companies pay Dividends? American Economics Review 73, 17-30 (1983).
Gaver, J.J. and K.M Gaver, Additional Evidence on the Association between the Investment Opportunity Set and Corporate Financing, Dividend, and Compensation Policies, Journal of Accounting and Economics 16, 125-160 (1993).
Gonedes, N., Corporate Signalling, External Accounting, and Capital Market Equilibrium: Evidence on Dividends, Income, and Extraordinary Items, Journal of Accounting Research, 26-79 (Spring 1978).
Gottardi, P., An analysis of the Conditions for the Validity of Modigliani-Miller Theorem with Incomplete Markets, Economic Theory 5, 191207 (Spring 1995).
Grundy, B.D., Preferred and Taxes: the Relative Price of Dividends and Coupons, Working Paper (September 1992).
Holder, M.E., Langrehr, F.W. and J.L. Hexter, Dividend Policy Determinants: An Investigation of the Influences of Stockholder Theory, Financial Management 27(3), 73-82 (Fall 1998).
Jensen, M.C. and W.H. Meckling, Theory of the Firm: Managerial Behaviour, Agency Costs and Ownership Structure, Journal of Financial Economics 3, 305-360 (1976).
Kose, J. and J. Williams, Dividends, Dilution, and Taxes: A Signalling Equilibrium, Journal ofFinance, 1053-1070 (1985). Lamoureux, C.G., Dividends, Taxes, and Normative Portfolio Theory, Journal of Economics and Business, 121-131 (1992).
Mackie-Mason, J.K., Do Taxes Affect Corporate Financing Decision? The Journal ofFinance, VoI.XLV, No.5, 1471-1493 (December 1990). Mann, S.V., The Dividend Puzzle: A Progress Report, Quarterly Journal of Business and Economics 28(3), 3-35 (1989).
Pettit, R., Taxes, Transaction Costs and Clientele Effect of Dividends, Journal ofFinancial Economics, 419-436 (1977).
Scholz, K.J., A Direct Examination of the Dividend Clientele Hypothesis, Journal ofPublic Economics 49, 261-285 (1992).
Wolfson, M.A., The Effects of Ownership and Control on Tax and Financial Reporting Policy, Economic Notes 22, 318-332 (1993).



[Afiliasi Pengarang]
Dr. Tao Zeng, Wilfrid Laurier University, Waterloo, Ontario, Canada