Download Dividends and Dividend Theories: International Research and Signalling Effects and more Lecture notes Accounting in PDF only on Docsity!
5 8
L T A 1 / 0 1 • P. 5 8 – 9 7
S E P P O K I N K K I
Dividend Puzzle – A Review of
Dividend Theories
ABSTRACT
Dividend policy has been one of the areas of corporate finance to be analyzed with a rigorous model, and it has since been one of the most thoroughtly researched issues in modern finance. There are a number of theories of dividend behaviour, and empirical studies provide little evidence for one over the other. Also the conceptions concerning corporate dividend theories are different. The main part of the discussion is related to the evaluation of financial research, because at all times researchers have tried to solve the dividend puzzle by using new theories and insights. Key words: Dividends, value of the share, agency theory, information content, signalling, clientele effects, ex-date effects
1. INTRODUCTION
In finance, there are some areas, which have puzzled researchers. One of them is the dividend behaviour of firms. Along with capital structure, dividend policy has been one of the first areas of corporate finance to be analyzed with a rigorous model, and it has since been one of the most thoroughly researched issues in modern finance^1. In spite of this, much remains unex-
- The author wish to thank Harri Seppänen, Pasi Sorjonen, Leo Virtaneva and an anonymous referee for very useful comments on earlier drafts of this paper. Financial support for the research from the Liikesivistysrahasto is gratefully acknowledged. 1 According to Frankfurter–Philippatos (1992) the subject of dividend policy has created a voluminous literature of its own and a handful of theories have been offered to resolve the inherent paradox of dividend payments to
SEPPO KINKKI, Lic.Sc. (Econ), Principal Lecturer Helsingin liiketalouden ammattikorkeakoulu • e-mail: seppo.kinkki@helia.fi
5 9
D I V I D E N D P U Z Z L E – A R E V I E W O F D I V I D E N D T H E O R I E S
plained concerning the role of dividends^2. Black (1976 p. 5) epitomizes the lack of consensus by stating ”The harder we look at the dividend picture, the more it seems like a puzzle, with pieces that just don’t fit together.” Dividend policy determines the division of earnings between payments to stockholders and reinvestments in the firm. Managers’ task is to allocate the earnings to dividends or re- tained earnings. Retained earnings are one of the most significant sources of funds for financ- ing corporate growth. Corporate growth makes it eventually possibly to get more dividends. The goal of this study is to describe the discussion on dividends and dividend theories. The main part of that discussion is related to the evolution of financial research, because at all times people have tried to solve the dividend puzzle by using new theories and insights. There are a number of theories of dividend behaviour, and empirical studies provide little evidence for one over the other. The discussion has lasted over half a century, and as a whole, these studies are interested in the following questions: (1) Why do firms pay dividends? (2) How do they set their dividend policy? (3) Does dividend policy affect share value? 3
2. DIVIDENDS AND FIRM VALUE
This chapter reviews theories concerning dividends and firm value. This question is one of the unsolved problems in the financial theory and has resulted in a great number of studies during the decades. To a great degree, the accepted theory tells us that dividends are irrelevant when firms are financing their actions. That is both in the absence of taxes (Miller–Modigliani (1961)) and in their presence (Miller–Scholes (1978)). At the same time researchers have documented a statistically significant relation between dividend yields and stock prices^4. The basic ques-
shareholders. Cooley–Heck (1981) studied the opinions of finance professors concerning articles with significant contributions to the finance literature. In ranking among others were mentioned Walter (1956), Miller–Mod- igliani (1961) and Brigham–Gordon (1968) although they represent different theoretical views on the dividend decision. 2 Weston (1981) and Megginson (1997). 3 In international research on dividends and dividend announcements are mentioned at least in the following connections: relationship between dividends and firm value (Williams 1938), clientele effects of dividends (Miller– Modigliani 1961), signalling effects of dividend announcements (Miller–Modigliani 1961), behavioural models of dividend policy (Lintner 1962), ex date effects of dividends (Elton–Gruber 1970), dividends and CAP (Brennan 1970), consumer preference theory (Bar–Yosef – Kolodny 1976), dividends and taxes (Miller–Scholes 1978), dividends and investment decisions (Fama 1974), stable dividend hypothesis (Matripragada 1976), asymmetric information (Bhattacharya 1979), dividends and Wealth Transfer Hypothesis (Kalay 1980), agency costs and div- idends (Rozeff 1982), the effect of regulation on dividends (Choi 1989), executive stock option plans and divi- dend policy (Lambert–Lanen–Larcher 1989), information content of stock repurchases (Karanjia 1990), various ways to divide dividends (Löyttyniemi 1991), option valuation (Adams–Wyatt–Walker 1994), corporate cross- ownership (Salin 1995), foreign influences on dividends (Hines 1996), major shareholder’s influence on divi- dend decision (Kinkki 1998), dividends as intra-industy information transfers (Laux–Starks–Yoon 1998), the ef- fects on legal regimes in different countries (LaPorta et al 1999). 4 Hess (1982) and the studies cited there.
6 1
D I V I D E N D P U Z Z L E – A R E V I E W O F D I V I D E N D T H E O R I E S
edge. The model of Whitbeck–Kisord (1963) is not based on discounted dividends but also in their model dividends is one of the illustrative factors. Eades (1982) developed a dividend- signalling model of the dissipative signalling cost type. Hagen (1973) determined the market value of the stochastic process representing the company’s dividend policy. Ohlson (1990) reviewed and synthesized the theory of security valuation for multiple- date settings with uncertainty. The theory results in a formula that determines security value as a function of expected dividends adjusted for their risk and discounted by the term structure of risk-free rates. Models such as CAPM is only seen in special cases. Earnings are seen as an information variable that suffices to determine a security’s payoff, price plus dividends^10. Ohl- son postulates that only (anticipated) dividends can serve as a generally valid capitalization (present value) attribute of a security. Goetzmann–Jorion (1995) re-examined the ability of dividend yields to predict long-hori- zon stock returns. They used two series beginning in 1871 (up to 1993), a monthly series for the United States, and an annual series for the United Kingdom. As a result, dividend yields only display marginal ability to predict stock market returns in either country. Dempsey (1996) advanced a discounted dividend model of share prices in the context of personal taxation. In terms of the model, consistent costs of capital expressions are advanced relating investor and firm perspectives. Rees (1997) analysed a sample of 8,287 firms/years drawn from UK industrial and commercial sectors during the years 1987–1995. The evidence strongly suggests that earnings distributed as dividends have a bigger impact on value than do earnings retained within the firm. In Finland Torkko (1974) tested the application of Gordon’s model. The sample was 23 firms from the years 1963–1971 but the results were not very encouraging. Suvas (1994) test- ed, in his dissertation, Gordon’s new model, as well as the models of Malkiel–Cragg (1970) and Bower–Bower (1970). Martikainen (1990) studied 28 Finnish companies listed on the Hel- sinki Stock Exchange during 1975–1986 and found significant positive correlation between the dividend growth rate and stock market returns
2.2. Earnings theories
Many researchers are critical of dividend theories. In traditional earning theories, the market price of a share depends on the company’s profits. Dividends have no effect on the shareprice. Shareholders are presumed to be so traditional that, when the company keeps the profits and does not pay dividends, they expect the firm to invest capital so that it gives at least their rate of return. Dividend policy then does not affect the market price of the share.
10 Later Ohlson (1992) developed a model that relates earnings and unexpected earnings to market returns.
6 2
L T A 1 / 0 1 • S. K I N K K I
According to the theory of financial economics, the value of the company can be regard- ed as the present value of its cash flows.^11 Earnings theoreticians include, for instance, Miller– Modigliani (1961, 1966)^12 , Baumol (1963), Friend–Puckett (1964), Watts (1973), Fama (1974), Black–Scholes (1974), Black (1976), Rubinstein (1976), Ross (1977), Miller–Scholes (1982) and Copeland–Weston (1988)^13. The classic work of Miller–Modigliani’s demonstrated that the firm’s investment decisions and dividend decisions do not depend on one another^14. They found that a firm’s taxes, growth and capital structure do not affect dividends. Thus dividend policy ”does not matter”^15. Pilotte (1986) studied, in his dissertation, the impact of stockholder wealth on external financing by non-dividend-paying firms. The sample was limited to firms having an established ”no cash dividend” policy. The results were generally consistent with the hypothesis that the decision to issue new securities conveys information about the managers’ assessments of the value of the firm’s assets in place and investment opportunities. The sample also indicates that at least some firms obey cash dividend policy, as it would not affect the value of their shares. A different view was also adopted by Bar–Yosef–Kolodny (1976) who explained dividend poli- cy by using the consumer preference theory. Yli-Olli (1979) and Suvas (1994) tested the propositions of Modigliani–Miller in Finland. Yli-Olli presented the link between the cost of capital and the market valuation of the firm based on the theory of Modigliani–Miller and made attempts to modify the assumptions of that theory to be comparable with capital markets in Finland. The results showed that in Finland the dividend policy has no effect on the market value of the firm.^16 Suvas (1994) studied the propositions of Modigliani–Miller from different angles. Accord- ing to model, the value of the firm’s equity becomes zero when expected cashflows to the shareholders are clearly positive. Suvas presented an alternative definition of the cost of equi-
11 The value of the firm in the absence of taxes may be seen as (1) the discounted cash flows, or (2) the current earnings plus future investment opportunities, or (3) the stream of dividends, or (4) the stream of earnings (Mill- er–Modigliani (1961). 12 Miller–Modigliani’s model assumes that dividend policy does not affect the firm’s share price but is deter- mined by certain characteristics of the firm. 13 Black–Scholes (1974) tested the relationship between security returns and dividend yield by forming well- diversified portfolios and ranking them on the basis of their systematic risk (their ”beta”) and then divided yields within each risk class. They found out that dividend yield had no effect on security returns. 14 Fama’s (1974) results are consistent with this view, even though, because of possible sampling errors, his study did not reject the hypothesis that there could be period-to-period association between the dividend and investment decisions of the firm. Brennan (1971) was critical against Gordon’s valuation models on dividends. His conclusion was that dividends do not affect the share prices through discounted dividends but, instead, in another way as determined in CAPM. 15 There are also researchers who explain the market value of the share and do not explicitly argue that divi- dends do not matter. Such researchers include, for instance, Steward (1973). 16 Significant dependent variables were however earnings (expected earnings before interest), leverage (value of the outstanding debt of the firm), growth of the firm and size of the firm (the book value of total assets).
6 4
L T A 1 / 0 1 • S. K I N K K I
security returns, (3) research concentrating on semi-efficiency tests using accounting data, (4) research on the connection between corporate finance and the capital asset pricing model. Dividends are included in (1) share price valuation models, (3) models concerning semi- efficiency tests and (4) models concerning corporate finance and the capital asset pricing model. Dividends are not directly included in studies on (2) unexpected earnings and their contempo- raneous association with security returns. Indirectly, researchers have however been interest- ed in the connection between dividends as signals of future earnings. Modigliani–Miller (1958) and Miller–Modigliani (1961) already presented this connection. As a conclusion, the discus- sion concerning dividends was divided into many branches but still researchers were puzzled by dividends. Stock market efficiency tests in Finland are briefly described next, as they relate to divi- dends, and then the results of the Capital Asset Pricing Model (CAPM), which includes divi- dends, are explained. The studies concerning stock market efficiency in Finland have produced mixed results. According Korhonen (1977) the Helsinki Stock Exchange seems to be in the category of infor- mational week-form efficiency by nature. Berglund (1986) and Virtanen–Yli-Olli (1987), tested Korhonen’s results^18. Korhonen (1977) and Berglund–Liljeblom–Wahlroos (1987) have stud- ied semi strong-form efficiency in Finland. In testing strong-form efficiency, Korhonen also tested how the market reacted to various dividend informations such as stock dividends, divi- dends and new issues of stock^19. Berglund–Liljeblom–Wahlroos (1987) extended the results obtained by Korhonen by us- ing daily data. They found significant positive excess returns on the announcement day for stock dividends and mixed announcements of stock dividends and new issues. For new issues, the announcement day return was insignificant, although a slightly positive preannouncement price development was detected. The results from these studies show that the efficiency of the Finnish stock market is not especially high compared with other stock markets in the world. According to Martikainen (1989), there exists evidence on anomalies from market evidence even in the weak-form sense^20. According to the results presented in the previous chapter, corporations’ market value has a strong relation to dividend cashflows. If the markets have a concrete view on future dividend deci- sions, then the market value of the share should beforehand react to dividend announcements.
18 They determined that the UNITAS index does not follow the random walk model. Both theoretically and statistically satisfactory models were found. 19 Other public sources of information tested by Korhonen included accounting income figures (three alterna- tive definitions of income and two models of expected income were used), mergers and divestitures. 20 Martikainen (1989) p. 8.
6 5
D I V I D E N D P U Z Z L E – A R E V I E W O F D I V I D E N D T H E O R I E S
Stock market efficiency tests have mainly concentrated on earnings and economic char- acteristics of the firm rather than dividends. Informational characteristics of earnings mean that earnings convey information about future earnings^21. As a result, accounting earnings an- nouncements and stock price fluctuations are closely related to one another. Numerous researchers^22 showed that accounting earnings reflect factors that affect stock prices. The evidence also indicates that annual and quarterly earnings convey information to the market. The stock price change on the day of the quarterly earnings announcement is smaller than the total stock price change associated with the unexpected quarterly earnings. Therefore alternative sources of information exist that allow the market to anticipate the accounting earn- ings. It was also found that abnormal returns are more highly associated with earnings than with operating cashflow. According Charitou–Vareas (1998) the relationship between cash flows and dividend changes substantially depends (a) on the magnitude of total accruals and (b) on growth oppor- tunities as proxies by the firm’s market-to-book ratio. Assuming stock markets to be informa- tionally efficient, they should rather lead than lag accounting earnings. Fama (1981) discov- ered that real stock market returns lead economic variables and are not led by them. As a conclusion, the studies concerning stock market efficiency tests in Finland have found that stock prices react to announcements on the financial characteristics of a firm. Among these characteristics are: profitability (Koskela 1984, Martikainen 1990, Laitinen 1991), growth (Kos- kela 1984, Martikainen 1990, Laitinen 1991), financial status (Koskela 1984), risk (Koskela 1984), accrued earnings (Niskanen 1990, Martikainen–Puttonen 1991), cash flow (Niskanen 1990, Martikainen–Puttonen 1991), financial leverage (Martikainen 1990, Laitinen 1991), op- erating leverage (Martikainen 1990), capital investment (Ikäheimo–Lumijärvi 1990), rate of in- terest for debt (Laitinen 1991).
2.4. Dividends and macroeconomic factors
There are a number of studies, which try to explain the market price of a share by using vari- ous kinds of information. That information can be divided into two groups: (1) information under the control of the managers and (2) information out of management control. The second group includes, for instance, macroeconomic factors. According to U.S. results, macroeco-
21 Ball-Brown (1968) tested the hypotheses according to which unexpected increases in earnings are accompa- nied by positive abnormal rates of return and unexpected decreases are accompanied by abnormal rates of re- turn. They found that the prices reacted positively to ”good” news and negatively to ”bad” news. The average movement to ”good” news was a positive change of 7.5 % and the average movement to ”bad” news was a negative change of 10%. They also reported that 85–90% of the relevant accounting earnings figures leaked to the investors before their normal release. 22 For instance Watts–Zimmermann (1986), chapter three and the studies mentioned there.
6 7
D I V I D E N D P U Z Z L E – A R E V I E W O F D I V I D E N D T H E O R I E S
The relation between expected equity returns and anticipated dividend yields has led to the hypothesis that they are positively related. Studies testing the before-tax return differentials across dividend yields are based on Brennan’s (1970) after-tax formulation of the capital asset pricing model. The relation between a firm’s before-tax equilibrium return and anticipated div- idend yield 26 has been found to be significantly positive in many studies, including Litzen- berger–Ramaswamy (1980, 1982), Morgan (1982), Elton–Gruber–Rentzler (1983) and Christie (1990). Dividend studies explaining the value of the firm are divided into two groups: (1) stud- ies based on discounted dividends and (2) studies based on the CAPM model.^27 According to Bar–Yosef–Kolodny (1976) ”implicit in the wide practical use of the CAPM is the assumption of dividend irrelevance”.
Least effective public information concerning stock prices
Weak form earlier prices efficiency macroeconomic factors dividend announcements Semi-strong –no changes efficiency –cash dividends –stock dividends –dividend changes
economic characteristics Strong-form earnings efficiency all possible information
Most effective public information concerning stock prices
FIGURE 1. Impact of public information on stock prices
significant contribution to explaining the return received on the firm’s security. As a conclusion, Bar-Yosef & Kolodny argued that investors have a net preference for receiving their return in the form of dividends to receiv- ing it in the form of capital gains. Contrary to the assumptions made in the use of CAPM, security market imperfec- tions and/or institutional factors exist in the market place to the extent that they have a significant influence on investor behavior. 26 The model is derived under the assumptions of unlimited borrowing and lending at the risk free rate of inter- est and unrestricted short sales. The dividends paid by corporations are assumed to be certain and known to investors. 27 It must noticed that, even though we are studing dividend models, dividends are not the only factor affecting the value of the firm. They are here called dividend models because according to them dividends, among other factors, affect the value of the firm.
6 8
L T A 1 / 0 1 • S. K I N K K I
2.6. Clientele effects and ex-date effects of dividends
2.6.1. Clientele effects
When firms make their dividend decisions, one important question is how stock prices are affected by dividends. The effects of firms’ dividend policies on stock returns have been wide- ly studied^28. For the most part, researchers have documented a statistically significant relation between dividend yields and stock returns; however, the explanation of this common empiri- cal finding has been controversial. This problem can be divided into two various questions: (1) Do firms with higher dividend yields have higher stock prices and vice versa? And, on the other hand, (2) how does stock markets react to dividend announcements? The first question is called the clientele problem and the second is called ex-date effects on dividends. Miller–Modigliani (1961) originally suggested clientele effects. They argued that investors choose the corporations whose payout ratio they prefer. Each payout ratio tends to attract a class of investors, a clientele. From the firm’s point of view, any clientele is as good as any other. If the firm changes its payout ratio, the result would be a change in the clientele, but that will not affect the value of the firm because any clientele is, from the firm’s perspective, as good as any other. In clientele effects, the studies assume that some classes of investors may prefer different levels of dividends due to their different levels of taxation. There may be a hypothesis that low-dividend firms attract investors with a high tax rate and that high-dividend firms attract investors with a low tax rate. Lease–Lewellen–Schlarbaum (1976) used panel data collected at Purdue to analyse the demographic attributes and portfolio compositions of a wide variety of individual investors. According to the results of Lease–Lewellen–Schlarbaum, private investors preferred long-term capital gains, followed by dividend income and then short-term capital gains. The optimal dividend yield, for example, for private investors, corporations and capital funds may be different. In studies, this phenomenon has been widely demonstrated^29. Oppo- site views are presented by Hess (1982) and Barclay (1987), whose empirical evidence does not support clientele effects on asset prices. Booth–Johnson (1984) examined the ex-dividend day behaviour of Canadian stock prices, but the ex-dividend day price ratios do not provide much evidence in support of dividend tax clienteles.
28 A partial listing includes Friend–Puckett (1964), Brennan (1970), Elton–Gruber (1970), Black–Scholes (1974), Litzenberger–Ramaswamy (1980, 1982), Hess (1982), Booth–Johnston (1984), Kaplanis (1986), Barclay (1987), Davidson–Mallin (1989), Crossland–Dempsey–Moizer (1991), Michaely (1991). 29 Elton–Gruber (1970), Eades–Hess–Kim (1984), Handjiinicolaou–Kalay (1984), Grinblatt–Masulis–Titman (1984), Lakonishok–Vermaelen (1986), Backlay (1987) and Crossland–Dempsey–Moizer (1991) (UK-data). See also Davidson–Mallin (1989) and Davidson (1989).
7 0
L T A 1 / 0 1 • S. K I N K K I
There are two main competing hypotheses about the determination of the fall-off.^33 The so-called ”clientele hypothesis” (Elton–Gruber 1970) assumes that shareholder clienteles do not switch around the ex-dividend day, whereas the ”short-term trading hypothesis” (Kalay
- claims that the normal clientele of a share changes around the ex-dividend day, so that the fall-off is determined by the activities of short-term traders who are taxed equally on divi- dends and capital gains. Miller–Scholes (1982) represent opposite views in the ”short-term traders” hypothesis. They argue that, if the stock price drop on the ex-dividend day is different from the dividend amount, short-term traders who face no differential taxes on dividends versus capital gains could make arbitrage profits. According to McInish–Puglisi (1980), when transaction costs are considered, the market for preferred stocks is efficient with respect to ex-dividend related price behaviour. Booth–Johnson (1984) examined ex-dividend day behaviour of Canadian stock prices (when Canada first began to tax capital gains) and found that ex-dividend day price was significantly different from zero or one. Kaplanis (1986) studied option price movements around ex-dividend, using UK Traded Options Market data from 1979 to 1984 and found that the average expected fall-off implicit in option prices is around 55 to 60% of the dividend and significantly different from it. Also the fall-off varied inversely with the dividend yield, which is consistent with the prediction of the ”tax clientele hypothesis”. Barclay (1987) researched ex-dividend day behaviour of common stock prices during the pre-tax period before the year 1910. Grammatikos (1989) studied the Tax Reform Act of 1986^34. On the other hand, Crossland–Dempsey–Moizer (1991) provides evidence of the clientele ef- fect in the UK stock market^35. Han (1994) examined whether the Tax Reform Act (TRA) of 1986 had an effect on ex-date stock behaviour for the National Association of Security Deal- ers Automatic Quotation (NASDAQ), the New York Stock Exchange (NYSE) and the American Stock Exchange (AMEX). Baker–Farrelly–Edelmaqn (1985) made a survey of management views on dividend policy. In Finland,^36 Hietala (1987, 1990) and Sorjonen (1988, 1995,2000) have studied ex-date effects. Sorjonen’s findings were consistent with the findings of Hietala. Stock prices fell from 78% to 92% of the amount of dividend from 1960 to1985 and from1989 to1990 and from 1993 to 1997 stock prices fell on average 70–75% of the amount of dividend. Thus, according
33 Kaplanis (1986) 34 The Tax Reform Act of 1986 eliminated the preferential tax treatment of capital gains relative to dividends. 35 Davidson–Mallin used a sample of 172 UK Stock Exchange ex-dividend events when Crossland–Dempsey– Moizer had 1.020 ex-dividend events. 36 Ex-dividend day behavior of stocks made by various data Hietala (1987) pp 50–53 and the studies men- tioned there.
7 1
D I V I D E N D P U Z Z L E – A R E V I E W O F D I V I D E N D T H E O R I E S
to Sorjonen, this is consistent with the view that taxes affect the valuation of dividends^37. Hie- tala–Keloharju (1995) investigated the ex-dividend behaviour of two classes of shares, whose trading is potentially dominated by investors under different tax regimes. The groups are re- stricted stocks (on Finnnish investors) and unrestricted stoks. The results support the hypothe- sis that long-term investors are the marginal investors and that the unrestricted shares face much higher taxation in dividends. As a conclusion of studies concerning dividends’ clientele and ex-date effects, it seems that the theories and empirical results are internationally mixed in nature. Modigliani–Miller claimed that empirically noticed stock price changes in connection with dividends are caused by a so-called information effect. Elton–Gruber explained that they are caused by different taxation between clienteles. Kalay argued that short-term traders cause them. In Finland, the empirical results support the tax effect theory.
2.7. Signalling effects of dividend announcement
The signalling effect of dividends assumes that dividends convey information about future earn- ings.^38 Changes of dividends give messages to investors about the firm’s future cash flows. Modigliani–Miller (1959) and Miller–Modigliani (1961) hypothesized that dividend reductions convey information that future earnings prospects are poor. The basic hypothesis includes that dividends and future earnings are in relation to each other. The studies then examine fundamentally how dividends affect future earnings. Such stud- ies are, for instance, Lintner’s (1956) and Watt’s (1973) propositions. Under the title of signal- ling or information content of dividends, a number of studies have been made to examine the reaction of stock markets to dividend announcements. These studies have, in fact, examined stock markets’ semi strong-form efficiency. Empirical results have found the signalling effect of dividends especially on U.S. data 39.
37 See also Kasanen’s (1988) comment on Sorjonen where he postulates that, even if all the calculations and estimates are correct, the true shareholder preferences also depend on other factors besides taxation. Kasanen mentions the company position with the labor unions and goverment and various creative ways to give divi- dend-like payments. The ex-dividend price drop can reflect other things than the personal tax rates of typical investors. Actually, according to Kasanen, we may be witnessing an estimate of an overall tax preference of a biased group of investors. 38 According to Karanjia (1990), there is no universal answer to the question: What especially do dividends signal? Asquith–Mullins (1986) comment that announcement dividend increases signal ”improved prospects” for the firm’s real earnings and the shares are priced on these improved prospects rather than the dividend increase per se. Miller (1986) argues that a steady dividend flow signals that the firm’s finances are under control. Also Easterbrook (1984) claims that it is unclear what dividends signal, or, do they do so, why dividends are better signals than apparently cheaper methods. Easterbrook argues that dividend increases are ambiguous. They can pretend either future growth or a lack of investment opportunities. Shefrin–Statman (1984) claim that raising and lowering of dividends provide information that is not otherwise available. 39 Watts (1973), Aharony–Swary (1980) and Kwan (1981). Dann (1981) analyzed the returns of various security classes around announcement of common stock repurchases. He hypothesizes that security value around stock
7 3
D I V I D E N D P U Z Z L E – A R E V I E W O F D I V I D E N D T H E O R I E S
Karanjia (1990) found similar results in his doctoral thesis. The abnormal stock price re- action to a stock repurchase reaction was found to be almost twice as large as that corre- sponding to a dividend increase. Liljeblom (1989) revealed a significant difference between the price reactions to announcements of proxy statements and the price reactions to equally good announcements (stock dividends and/or stock split).^41 John–Lang (1991) placed insider trading on a level with dividend announcement and found out that insider trading immediately prior to the announcement of dividend initiations has ex- planatory power. Schatzberg–Datta (1992) studied corporate dividend announcements and the weekend effect. According to them, stock returns vary systematically across days of the week and average returns are actually negative on Mondays^42. Frankfurter–Lane (1992) criticize event studies, in which positive (negative) abnormal returns are associated with announcements of dividend increases (decreases). Employing random samples has mainly made studies concerning the signalling effect of dividends. DeAngelo–DeAngelo–Skinner (1994) used selected samples when they researched firms with losses and firms without losses. They found significant differences between these two groups in their dividend decisions^43. DeAngelo–DeAngelo–Skinner (1996) studied the sig- nalling content of managers’ dividend decisions for 145 NYSE firms whose annual earnings decline after nine or more consecutive years of growth. They found no support for the notion that dividend decisions help identify firms with superior future earnings. The increasing diver- sity of corporate control, and the emergence of more varied types of ownership structure sug- gest that this is an appropriate time to reflect on recent developments and their implications for understanding of the underlying economics of the processes involved. Brooks (1996) empirically investigated the change in asymmetric information at earnings and dividend announcement. According to Brooks, the dividend announcement may be im- portant for reasons other reduction of information asymmetry, but the actual announcement does not reduce information asymmetry among traders. Akhigbe–Madura (1996) measured the change in corporate long-term performance following dividend adjustments. They focused spe- cially on dividend initiations and omissions. They found that firms experience favourable long-
41 For a five-day event window around the announcement, the average excess return was +3.7% for stock splits and +5.2% for stock dividends and stock splits. For stock dividends the average five-day return was +1.2%. For announcements of convertible debt issues by means of rights issue, an excess return of –1.2% at the event time t=+1 was detected. 42 In a sample of 138,824 dividend announcements over 26 years, 3,484 firms were investigated to find possi- ble seasonalities. Tests provided no support for the information hypothesis and suggest that the anomalous pat- tern of returns is driven by some factor unrelated to information arrivals. 43 Their sample consisted of 167 NYSE firms with losses and 440 NYSE firms without losses from the years 1980–1985. As a result they found out that 50.9% of loss firms reduced their dividends versus 1.0% of firms without losses.
7 4
L T A 1 / 0 1 • S. K I N K K I
term share price performance following dividend initiations. Conversely, firms omitting divi- dends experience unfavourable long-term share price performance. According Laux–Starks–Yoon (1998) same dividend announcement can have diverse ef- fects within the same industry. Although managers signal information about the announcing firm, the dividend change can also disclose information relevant for the announcer’s rivals. Their results suggest that for rivals without extensive market power or growth options relative to the announcer, dividend increases elicit a negative reaction. Conversely, rivals with rela- tively more market power and growth options experience positive reactions to dividend in- creases and no reaction to dividend decreases. According Howe–Shen (1998) the stock prices of industry competitors do not react to dividend initiations. Thus, the information conveyed to the market by the decision to initiate dividends contains no industry-wide component. As a conclusion based on international signalling studies concerning dividends: cash div- idends also signal alone, and dividend and earning announcements are in relation to each other. Price reaction begins very quickly, unchanged dividends have no effect, stock dividends and/or stock splits signal, and also stock repurchases signal. The signalling effect is stronger in selected samples. Finnish data is not as encouraging. Finnish results are, however, based on random sam- ples. On the other hand, share repurchases, which are used in the U.S., and according to em- pirical results include some kind of information, are seldom used in Finland owing to legal restrictions. Korhonen (1977) and Wahlroos (1979) found poor results on Finnish data concerning the applicability of Lintner’s model (1956)^44. Yli-Olli (1980) tested models based on propositions of Lintner and Watt and found out that the causality between dividends may be reversed and that in Finland the firms ”show” net income only so much as they will pay dividends. The models based on the information content of the dividend hypothesis did not gain any empiri- cal support in the Finnish stock market. Yli-Olli (1982) compared informational content on Japanese, Swedish and Finnish firms and found this effect especially in some Japanese and Swedish firms. In Finnish firms this conclusion only gained weak support in a small sample of firms. Based on Finnish data, it also seems that, if dividends only include cash dividends, the results are not very encouraging 45. When stock dividends and right issues are also studied, the results are much better. Berglund–Liljeblom–Wahlroos (1987) found high returns connect-
44 According to Wahlroos (p. 234) reasonable doubt concerning the ”information content of dividends” hypothesis may be expressed. 45 Finnish tests on the signaling hypothesis have been made by testing Lintner’s applications. Lintner connects dividends, however, to cash dividends. ”The dividend theory” states that prices are equal to the present value of the cash flow to the investor (that is, cash dividends).” Lintner (1962) p. 268.
7 6
L T A 1 / 0 1 • S. K I N K K I
3. DIVIDEND POLICY
3.1. Behavioural models of dividend policy
Behavioural models of dividend policy assume that the change in dividends can be explained by the last period’s dividends and the target dividends, which can be expressed as a fraction of this period’s earnings. Lintner (1956) first published the basic model for that kind of dividend policy. His model is based on a set of interviews with managers about their dividend policies. From Lintner’s interviews, it was apparent that dividend policies across firms were hardly uniform. Some common characteristics were however identified: managers tend to change divi- dends when unanticipated and non-transitory changes have occurred in their firm’s earnings. Using an econometric model based on these perceived patterns, Lintner found that he could explain a significant portion of annual dividend changes for a sample of companies over the period 1918–1941. According to Lintner, the key determinant of dividend changes is the firm’s bottom line net income.^49 Lintner’s original findings were supported in later empirical works by Pettit (1972), Watts (1973) and Fama (1974), who used the same types of models as Lintner. Healy–Palepu (1988) researched earnings information conveyed by dividend initiations and omissions, and the findings reported provided strong support for Lintner’s description of managers’ decision- making processes. As a conclusion, Copeland–Weston (1988 p. 663) propose that most cor- porations desire to avoid reducing dividends, because a dividend cut signals a future perma- nent decline in earnings. Thus dividends only increase with a lag after earnings rise, and dividends increase only after an increase in earnings appears clearly sustainable and rela- tively permanent. The Lintner model has fared well relative to its competitors in tests on aggregate data by Brittain (1966), who in his studies isolated the major determinants of corporate dividend poli-
49 Lintner (1956 p. 112) also argued that there is no evidence that the normal or target equilibrium ratio of dividends to profit for corporations as a whole would be any different during the postwar years than during the preceding quarter century. Lintner found in interviews that: (1) Managers believe that firms should have some long-term dividend payout ratios, (2) Large earnings changes not in line with current dividend payouts were the chief factor behind the dividend decision, (3) Managers avoided making changes in dividend payout rates that might have to be reversed within the next year, (4) Managers focused on the changes in dividend payout rates rather than the amount of the payout itself. (5) Investment funding requirements had little effect on changing the pattern of dividend behavior. Copeland–Weston (1988) pospones that paying stable dollar dividends is not the only dividend policy. They described three major types of dividend payout schemes: (1) Stable dollar amount per share is also the policy implied by the words ’stable dividend policy’. That dividend policy is followed by most firms. (2) Constant payout ratio is followed by very few firms. Since earnings fluctuate, following that poli- cy means that also dividends will fluctuate and it results in unreliable signals to the market about the future prospects of the firm. (3) Low regular dividend plus extras is a compromise between the first two and gives flexibility to the firm but leaves investors a bit uncertain about what their income will be.
7 7
D I V I D E N D P U Z Z L E – A R E V I E W O F D I V I D E N D T H E O R I E S
cy. Theobald (1978) tested intertemporal dividend models by Lintner by using UK data but the results were quite poor^50. Mantripragada (1976) tested the question of stable dividend policy and its relation to share prices. The stable dividend hypothesis argues that the market price of a share with stable divi- dend payments should be higher than the market price of a similar share with payments, which fluctuate, on average, by approximately an equal amount. Little empirical support was how- ever found in support of the stable dividend hypothesis. Mantripragada’s empirical study was extended by Schnabel’s (1981) theoretical work where he pointed out that, in the presence of cash demands and quadratic liquidation costs, the investor will attempt to hold a portfolio that is efficient along the three dimensions of expected return, variance of return, and the variance of dividend yield. Kolb (1981) developed, by using the discriminant analysis, a model based on economic and institutional factors to determine the payment of dividends and to predict changes in the annual cash dividend of a firm. The most significant factors were earnings, liquidity and profit- ability. Variables designed to measure managerial attitudes were of relatively minor impor- tance. Kim (1985) in his dissertation studied if Lintner’s model could explain repatriation of earn- ings from U.S.-owned foreign subsidiaries to their U.S. parent. The results showed that the U.S.-owned foreign subsidiaries had more stable dividend payment records than their U.S. par- ent companies and they did not follow their U.S. parent companies’ dividend payout policy. Also Hines (1996) used Lintner’s models when comparing U.S. corporations’ dividend payout from domestic and foreign profits. Baker–Farrelly–Edelman (1985) carried out a questionnaire survey just like Lintner and found that most of his observations were still valid. According to Marsh–Merton (1987 p. 3), ”except for certain debt-indenture restrictions and accumulated-earnings tax penalties, there do not appear to be any significant legal, ac- counting-convention, or corporate-tax factors to exert pressures on managers of publicly trad- ed and widely held corporations to follow any particular dividend policy”. In that situation, there are factors influencing the dividend decision and dividend policy^51. They criticized Lint- ner’s model on the basis that it does not take into account the cross-sectional dependencies among firms’ dividend policies. According to them, it is reasonable to expect that, in addition
50 Theobald’s data consists of 41 UK firms from the period 1964 to 1975. The Advanced Corporation Tax in March 1973 effectively reduced the cost of dividends to the firm, which perhaps partly explains the poor results. 51 Weston–Copeland (1991) pp. 658–661. Choi (1989) isolated, in his dissertation, common factors which de- termined the dividend policy of unregulated firms. These factors were tax factor, transaction cost factor, agency costs factor, information signaling factor, risk factor and profitability factor. Within the signaling framework, he found that dividends are related positively to the profitability factor and the agency costs factor, and related negatively to the risk factor and transaction cost factor.
