Cash dividends and share repurchases
as value enhancing activities
Evidence from the United Kingdom
Author:
Mark Wolkotte1
Date: June 2010
Supervisor:
Dr. H. Von Eije
Faculty of Economics and Business
Master Thesis, MSc Business Administration Specialization: Corporate Financial Management
1
Cash dividends and share repurchases
as value enhancing activities
Evidence from the United Kingdom By: Mark Wolkotte
ABSTRACT
By using panel data analyses and the methodology of Fama and French (1998), this paper examines if the payments of cash dividends and share repurchases increases firm value. The sample consists of 717 U.K. listed firms in the period 2002 to 2006. From theoretical perspective there are several arguments discussed that explain why payouts may affect firm value. The results reveal a clear relationship between cash dividends and firm value, while similar conclusions about share repurchases were not found. Thereby it excludes the belief that the surge in share repurchases is explained by the objective to maximize firm value. The level of cash dividends and the future changes in cash dividends are positively related to firm value, in line with Fama and French (1998). In contrast to their results, I find mixed results in the relationship between the change in past cash dividends and firm value.
JEL classification: G35, G39
TABLE OF CONTENTS
1 INTRODUCTION ... 4
2 PREVIOUS LITERATURE ... 7
2.1 The time trend ... 7
2.2 The irrelevance theory of payout policy ... 9
2.3 Motives to pay cash dividends ... 10
2.3.1 Investors perceive the level and changes in dividend payout as a credible signal... 10
2.3.2 Investors perceive dividend payouts as effective ways to oppose agency costs... 10
2.3.3 Managers adjust their payout policy in line with investor’s expectations... 11
2.4 Motives to use share repurchases... 12
2.4.1 Investors perceive the level and changes in share repurchases as a credible signal ... 12
2.4.2 Investors perceive payouts as effective ways to opposite agency costs ... 13
2.5 Differences between payments of cash dividends and share repurchases ... 15
2.5.1 Investors perceive cash dividends as less flexible than share repurchases... 15
2.5.2 Cash dividends are higher taxed than share repurchases ... 15
2.5.3 Risk-averse investors prefer cash dividends above share repurchases ... 16
3 HYPOTHESES AND METHODOLOGY ... 18
3.1 Hypotheses... 18 3.2 Base Models... 19 3.3 Variable construction ... 21 3.3.1 Dependent variable... 21 3.3.2 Variables of interest... 21 3.3.3 Control variables... 22
3.4 Measurement technique and issues ... 24
3.4.1 The Fama and MacBeth methodology ... 25
3.4.2 Panel data methodology... 25
3.4.3 Comparative analysis... 25
4 DATA AND DESCRIPTIVE STATISTICS ... 28
4.1 Data ... 28
4.2 Descriptive statistics ... 28
5 RESULTS... 30
5.1 Level and changes in level of payouts ... 30
5.2 Level of payouts and changes in payout policy ... 34
5.3 Comparative analysis ... 37
5.4 Robustness check ... 38
5.5 Discussion of the results ... 40
6 CONCLUSION & LIMITATIONS ... 42
6.1 Conclusion ... 42
6.2 Limitations and suggestions for future research ... 43
REFERENCES ... 44
1 INTRODUCTION
The purpose of this paper is to unveil if cash dividends and share repurchases enhance firm value. Moreover this paper examines which form of payout has a bigger impact on firm value. According to Koller (2005) three types of financial decisions influencing the value of the firm are investment decision, financing decision, and payout decision. These are interdependent. However, the investment decision has an impact on future earnings and thereby it also affects the size of future payouts. Furthermore, the payout decision influences the amount of equity capital in a firm’s capital structure and thereby it affects the cost of capital. Thus interrelations exist through time among the investment, financing and payout decision. This implies that the decision which determines the amount and form of payout becomes relevant in maximizing firm value and shareholder’s wealth.
Historically Miller and Modigliani (1961) were the first to challenge the hypothesis that higher payouts translates into higher value. They conclude that the mix of retained earnings and payout do not affect firm value. This finding implies that the form of payout does not affect firm value. However, the authors assume this under the restrictive condition of perfect and frictionless markets, which is unlikely. Moreover, DeAngelo & DeAngelo (2006) show that the Miller and Modigliani (1961) assumption only hold under the assumption of full payout. Subsequent literature has extensively examined the influence of market frictions (e.g. agency costs, information asymmetries, and tax differential)2 that all affect the Miller and Modigliani (1961) propositions. Imperfections may imply that the size of the distribution matters, the form of distribution matters, or that they both matter. Several theoretical justifications are developed and analysed. Until now there is a lack of consensus in the literature regarding the empirical validity of the different theories. This paper examines the different payout forms and their impact on firm value.
The selected sample examines firms in the United Kingdom (U.K.) over the bullish period 2002-2006. All in the context to clarify the main question: Are payments of cash dividends
and share repurchases value enhancing activities? According to Von Eije and Megginson
(2008), U.K. listed firms are responsible for 50 percent of the total share repurchases in the European Union (E.U.) in the period 1989 to 2005. Moreover, the study of Khan (2006) shows us that investors in the U.K. are poor monitors which increase agency costs. Payouts mitigate agency costs, which may explain why investors place a premium on firms that make
payments through cash dividends or share repurchases. Finally, U.K. listed firms are often used in prior studies, which makes it easier to compare results.
This paper complements the existing literature in several ways. Firstly, by not restricting the analyses to one form of payout or total payout, but by examining whether different forms of payout (cash dividends versus share repurchases, or a combination of both payout forms) have an effect on firm value. Furthermore most prior studies examine payout effects by using an event study methodology which measures abnormal returns. This methodology is often restricted to a relatively short time frame. This paper rather focuses on long-term changes in firm value and thereby uses a repeated cross-sectional analysis as an alternative approach in line with Fama and French (1998). However, a cross-sectional analysis is static and does not control for unobserved variables. So secondly, to overcome this drawback, this paper uses a panel data approach in addition. In contrast to repeated cross-sectional analyses, the change is explicitly incorporated into the design so that individual changes in a set of variables are directly measured repeatedly over time. Thirdly, this paper recognizes that share repurchases in the U.K., even as other member states of the E.U. have the option to account the amount of share repurchases at book value or market prices. In order to recognize the way a firm reports, this study demonstrates an alternative way to measure share repurchases. As a result the actual amount of share repurchases is measured more accurate. Finally, this paper contributes to the scarce empirical literature about share repurchases in the E.U.
value, as Faulkender & Wang (2006) illustrated that the results of Fama and French (1998) were potentially influenced by a measurement error of the dependent variable3.
The remainder of this paper is organised as follows. Section 2, provides a literature overview that relates the cash dividends and share repurchases effects on firm value. Section 3, discusses the methodology and constructs the hypotheses. Section 4, elaborates on the sample selected and the descriptive statistics. Section 5, provides the results and compares it with prior studies. Finally, section 6 concludes.
3
2 PREVIOUS LITERATURE
This chapter describes the underlying theories that explain why cash dividends and share repurchases have an effect on firm value. First, subsection 2.1 starts with a description how the two different forms of payout evolve through time. Subsection 2.2, describes the classical payout irrelevance theory of Miller and Modigliani (1961). Subsection 2.3, highlights prior studies with the motives for dividend payments. Subsection 2.4 highlights the motives for share repurchases. And finally, section 2.5 highlights three important differences in how investors perceive cash dividends and share repurchases.
2.1 The time trend
The payout policy landscape has changed drastically in the last decades (Brav, Graham, Harvey, and Michaely; 2002). First, the share of public firms that payout through dividends decreased drastically. For example, Fama and French (2001) found in the United States (U.S.) that in 1978 roughly 67 percent of public firms’ payout consists of dividends and in 1999 this percentage decreased to an all time low of 21 percent. The subsequent study of Baker and Wurgler (2004a) likewise find a decline in the number of dividend payers. Evidence from the U.K. (Ferris, Sen, and Yui, 2006) also reports a declining propensity to pay dividends, although, the decline is smaller in magnitude than in the U.S. and appears only recently. For example, in 1999 the corresponding number of dividend paying firms is 63 percent compared to 21 percent found by Fama and French (2001). Moreover the number of U.K. firms paying dividends declines from 76 percent in 1998 to 55 percent in 2002. Despite the decline in the number of firms paying dividends, the aggregate volume of dividends paid has not decreased, as shown in DeAngelo, DeAngelo, and Skinner (2004). This suggests that a few large firms are responsible for the level of dividend payouts.
Bagwell and Shoven (1989) argue that the sustained increase in share repurchases indicates that firms have learned to substitute dividends for share repurchases in order to generate lower-taxed capital gains for investors. This suggestion is examined by Grullon and Michaely (2002), and they find evidence from U.S. listed firms which suggest that large-established firms partially finance their repurchase programs with money that could also be used for cash dividend increases. This result suggests that firm’s substitute share repurchases for dividends. However, the results of other researchers dispute this conclusion. For example Brown and O’day (2006), who apply the same methodology as Grullon and Michaely (2002), fail to conclude that firm’s substitute share repurchases for dividends in Australia.
The empirical literature is thus mixed about the question whether share repurchases are replacing dividends as the primary form of distributing profits. The study of Fama and French (2001) shows us that firms who pay dividends increasingly payout through share repurchases, but the cash dividend payout policy remains unchanged. In addition they report that newly listed firms often make their first payouts through share repurchases. Existing firms who only pay dividends do not give up their dividend payments for share repurchases (Grullon and Michaely, 2002). Finally DeAngelo, DeAngelo, and Skinner (2004) show us that in the U.S. the total level of dividends increased by 16 percent between 1978 and 2000. This increase is driven by substantial increases in dividends by a few large firms. This is striking since comparable firms are also responsible for the surge in share repurchases (Fama and French, 2001).
2.2 The irrelevance theory of payout policy
The adherents of this theory state that dividends or changes in the dividend policy have no effect on firm value. The essence of this theory is that firms will only pay dividends from residual earnings. In the time this theory was developed share repurchases were no topic of interest, but it is fair to assume that similar comments also apply for share repurchases. The theory of dividend irrelevancy was first coined by Miller and Modigliani (1961). They argue that firm value is driven by operating and investment decisions. They state that the firms’ investment policy is fixed and independent of its dividend policy, as dividends are only paid from residual earnings. In addition, they state that an optimal dividend policy does not exist. However, all their assumptions are based on perfect and frictionless capital markets. Hence, there are no taxes, no transaction costs, investors are rational, and all investors have access to the same information and share the same expectations about the firm’s future as its managers. In reality, however, financial markets suffer from several imperfections. Moreover, the model of Miller and Modigliani (1961) implicitly assumes full payouts, since their assumptions prohibit retention. De Angelo and De Angelo (2006) show that the payout policy matters in exactly the same sense that the investment decision does, if retention is allowed. Thereby the authors show that investment policy is not the sole determinant of firm value, even in frictionless markets.
2.3 Motives to pay cash dividends
In general, there are two research lines on dividends and dividend policy, one is based on the motives for dividend payments and the other is on the market reaction to dividend payments. In the end this research only focuses on the perception of the market. However, I have no grounds to hypothesize a dividend effect on firm value without knowing the underlying motives. This section surveys theories concerning dividend payments, and seeks to reconcile them under a common set of assumptions.
2.3.1 Investors perceive the level and changes in dividend payout as a credible signal
Signaling theories state that dividends are a costly signal available to managers to convey information about a firm’s future prospects (Bhattacharya, 1979; John and Williams 1985 Aharony and Swary, 1980; and Miller and Rock, 1985). Investors examine every action a firm takes for implications for future cash flows and firm value. When firms announce changes in dividend policy, they are conveying information to markets, whether their managers intend to or not. Investors tend to view announcements made by firms about their future prospects with a great deal of skepticism, since firms routinely make exaggerated claims. According to the signaling theory such firms may take actions that cannot be easily imitated by firms without good projects. The level of dividends and dividend changes are viewed as such actions. In cases where dividends become more expensive (by dividend taxation), dividend payments become even more informative. This theory assumes that by signaling with higher level of dividends or dividend increases, firms enhance firm value. Numerous empirical studies support the signaling theory (Grullon, Michaely, and
Swaminathan, 2002; Healy and Palepu, 1987; Aharony and Swary, 1980).
2.3.2 Investors perceive dividend payouts as effective ways to oppose agency costs
institutions. Institutions have a relative advantage in monitoring firms and thereby alleviating agency problems (Allen, Bernando and Welch, 2000; Jensen, 1986). Moreover higher dividend payouts may force firms to raise external capital thereby increasing monitoring and transparency, which also mitigate agency conflicts. Numerous studies provide empirical support for the agency theory (Agrawal and Jayaraman, 1994; Jensen, Solberg, and Zorn, 1992; Lang and Litzenberger, 1989; Rozeff, 1982). According to this theory, investors will recognize and reward cash dividend payments by placing a premium on firm value.
2.3.3 Managers adjust their payout policy in line with investor’s preferences
The theory is also known as the catering theory introduced by Baker and Wurgler (2004a, b). Catering implies that managers will tend to initiate or continue paying dividends when investors put a higher price on payers, and omit dividends or avoid initiating them when investors favour nonpayers. If managers are indeed able to adjust their payout policy to investors needs, then in equilibrium, no firm is able to influence firm value by the level of dividend or changes in dividend policy. However, the catering theory acknowledges that markets are affected by sentiment, and thereby it relaxes the market efficiency assumption of Miller and Modigliani (1961).
The catering theory assumes that investors have an uninformed, time-varying demand for dividend paying firms4. By the costs and limits on arbitrage, as indicated by Shleifer and Vishny (1997) and D’Avolio (2002), payments of dividend become relevant for firm value. Firms managers cater to investor demand in response, but the market inefficiency cannot be completely resolved. The main issue of inefficient markets is that firms’ managers have to decide which of two prices to maximize: a short-term price affected by uninformed demand, or a fundamental long-term value determined by investment policy. Catering maximizes the short-term price (Baker and Wurgler, 2004a), while the traditional finance literature emphasizes fundamental value. Several empirical studies provide support for the catering theory such as Ferris, Sen, and Yui (2006), Baker and Wurgler (2004a, b) and Li and Lie (2006) though other studies like Von Eije and Megginson (2008) and Denis and Osobov (2008) do not corroborate this theory. If the theory holds, it implies that payouts of cash dividends may have a short-term positive effect on firm value.
4
2.4 Motives to use share repurchases
This section surveys the motives and theories concerning share repurchases, and seeks again to reconcile them under a common set of assumptions. According to the finance theory and prior empirical findings, the decision to use share repurchases may be related to distribution, investment, capital structure, corporate control, or compensation policies. Prior literature cites that share repurchases are followed by an increase in share price and thus firm value (Kahle, 2002; McNally, 1999; Ikenberry, Lakonishok, and Vermaelen, 1995; Comment and Jarrell, 1991; Asquith and Mullins, 1986).
2.4.1 Investors perceive the level and changes in share repurchases as a credible signal
When share prices are temporarily below their true fundamental value, managers use share repurchases to signal that their shares are undervalued. Whereas investors may discount bullish statements and favourable predictions, concrete actions such as share repurchases are likely to be taken more seriously. If the market recognizes that share repurchases are being undertaken because shares are undervalued, share prices are likely to rise quickly. Assuming they rise to their intrinsic values, the real wealth of investors that continue to hold the shares will not be increased, but it now will be reflected in the current market value. The studies outside the U.K. such as Lie (2005), Ikenberry, Lakonishok and Vermaelen (1995), Bartov (1991), Comment and Jarrell (1991), Hertzel and Jain (1991), Vermaelen (1981) and Dann (1981) show us that share repurchase announcements increase firm value as the markets receive manager’s signal as credible. To illustrate, the results of Ikenberry, Lakonishok and Vermaelen (1995) show on average a 3.5 percent abnormal return after share repurchase announcements in the U.S.. There is also ample of evidence in the U.K. that is in line with this theory (Oswald and Young, 2004; Rau and Vermaelen, 2002; Rees, 1996). The more recent study by Oswald and Young (2004), who examine the period 1995 until 2000, report a 1.95 percent 11-day abnormal return, and a significant positive one year abnormal return of 7.53 percent following the announcement. Overall these results are consistent with the information-signaling hypothesis of Bhattacharya (1979) and Miller and Rock (1985). Finally, the empirical results of Stephens and Weisbach (1998) show that share repurchases are negatively correlated with prior stock returns. This indicates that firm’s use share repurchases when their share price is perceived as undervalued.
originally announced. Lie (2005) suggests that the announcement of the repurchase intention alone may not be a good indicator of superior future financial performance and cash flow. Previous studies in the U.K. by Rees (1996) and in Hong Kong by Zhang (2005) find that investors also react positively to actual share repurchases, even after it has shown a positive response to initial announcements of intent to repurchase. Thus actual share repurchases appear to provide value relevant information in addition to that conveyed by the initial repurchase program announcements. These findings are in line with the findings of Ikenberry et al. (2000) that information conveyed by share repurchases announcements are underestimated by investors. Assuming that managers have inside information, they can better judge if the firm’s shares are truly undervalued. If managers think that the current share price is too low, and the market ignores it, managers can repurchase shares at bargain prices until the shares reach their intrinsic value. Based on this line of reasoning, the share repurchases effect can be positive for firm value.
2.4.2 Investors perceive payouts as effective ways to oppose agency costs
Share repurchases may also mitigate agency costs. For example, share repurchases effectively reduce excess cash levels (1) and it increases firm leverage (2). On the other hand agency costs might surge when share repurchases are used for managerial option plans or increases in earnings per share (3). According to Li and McNally (2007), Grullon and Michaely (2004), and Jagannathan and Stephens (2003), mitigating agency costs is the main reason why investors value share repurchasing firms at a premium. As share repurchases
create firm value by reducing the resources for negative net present value projects (Jensen, 1986). In this section I will discuss the three underlying motives in the sequence order as
mentioned above.
reduce excess cash, unlike cash dividend payments which are relatively constant over time. This implies that share repurchases have a positive effect on firm value.
Second, prior research indicates that firms undertake share repurchases in order to manage their capital structure (Opler and Titman, 1996; Bagwell and Shoven, 1988). Managers of firms with additional debt capacity may use share repurchases to move closer to a more desirable capital structure. However, according to Dann (1981) it is questionable if there is an identifiable optimal capital structure. Even if there is such, the question remains whether share repurchases are the most effective and least costly way of moving towards this target. Alternative ways such as direct debt for equity exchanges or new debt issues are perhaps more effective. The only thing that share repurchases distinguish from the alternatives is the reduction in a firm’s size (total assets decreases). However, this may not be in the long term interests of a firm. This is probably also the reason why studies that favor this theory are limited (Dittmar, 2000).
Third, in the last two decades we have seen a trend of an increasing use of stock options to compensate managers (Fenn and Liang, 2001; Jolls, 1996). In that way managers’ compensation is stronger linked to the goals of investors and thereby, hopefully alleviates agency costs. The underlying issue is that share repurchases are expected to increase firms’ share price (Vermaelen, 1984), whereas dividends reduce share prices and thereby also the value of options (Fenn and Liang, 2001; Lambert, Lanen, and Larcker, 1989). In other words share repurchases can be used to offset dilution effects when options are exercised, whereas dividends cannot. In the most positive scenario it might even help to increase the earnings per share. This is at least positive for existing shareholders. However, as Dann (1981) argues, this motive is simply one of financial illusion. Investors will not be fooled by these financial adjustments.
2.5 Differences between payments of cash dividends and share repurchases
Previous subsections show several motives why investors place a premium on firms that make payments in the form of cash dividends or share repurchases. But so far, a clear preference for cash dividends or share repurchases has been insufficiently addressed. This subsection addresses three reasons in cases where either cash dividends or share repurchases are preferred by investors.
2.5.1 Investors perceive cash dividends as less flexible than share repurchases
The first reason is that share repurchases increase flexibility (Jagannathan et al., 2000). Prior empirical findings of Guay and Harford (2000) note that share repurchases are no substitutes for cash dividends, as they fulfill different functions. This is confirmed by the survey conducted by Brav et al. (2005), in which managers report the flexibility argument as an important advantage of share repurchases. Figure 1 visualizes the conceptual framework of this concept. Several studies have shown that managers usually feel committed to maintaining a sustainable level of dividend payments, contrary to share repurchases. This means that it is unlikely that transient cash flows are paid out by increases in cash dividends. Share repurchases seem to be a more appealing option in this case, because unlike cash dividends, there is no expectation that the distribution through share repurchases occur on a regular basis. In addition, investors react negative to dividend decreases unlike reductions in share repurchases (Denis, Denis and Sarin, 1994; Bajaj and Vijh, 1990; and Kaplan and Reishus, 1990). So, investors reward firms with an increase in cash dividend higher than firms who increase share repurchases. The opposite reasoning applies for dividend decreases or cancellation of share repurchases.
Excess cash flows Permanent cash flows Dividends
Transient cash flows Share repurchases
Figure 1: conceptual framework for distributing cash to shareholders
2.5.2 Cash dividends are higher taxed than share repurchases
some control over the timing of capital gains by choosing when to sell their shares, unlike dividend payments which normally rest in the hands of a firm’s manager. From this line of reasoning, the rise in share repurchases can at least partly be explained by tax effects. The results of the study from Grullon and Michaely (2002) confirm this result as they show that the reaction to share repurchases in the U.S. becomes more positive when tax gains are larger. Jagannathan et al. (2000) find that that tax reasons are an important argument in favour of substitution, but is not sufficient alone to explain the surge in share repurchases. They find that after the 1986 Tax Reform Act elimination (with previous preferential tax treatment for capital gains) the surge in share repurchases in the U.S. continues. According to Miller and Scholes (1978), the influence of taxes is irrelevant in the firms’ payout decision. In a follow-up study the authors argue that investors can avoid taxes by investing in shares via pension plans or by offsetting deductions in other tax payments on a personal level. In a subsequent paper Miller and Scholes (1982) state that pricing of shares are dominated by investors subject to symmetric taxation of ordinary income and capital gains. Although, this argument sounds plausible this does not hold for U.K. listed firms in the period of interest as taxes are not symmetric during this period, which is illustrated next.
2.5.3 Risk-averse investors prefer cash dividends above share repurchases
3 HYPOTHESES AND METHODOLOGY
In this paragraph, I will start in subsection 3.1 with the hypotheses construction, derived from previous literature as described in section 2. The lines of thought are based on the assumption that managers maximize firm value and shareholders’ wealth. In subsection 3.2 the base models are explained. Subsequently, in subsection 3.3 the variable construction is explained. Finally, in subsection 3.4, the measurement technique is justified and potential econometric issues are addressed.
3.1 Hypotheses
Fama and French (1998) use value regressions to examine the effect of dividend payments on firm value with the expectation that a tax rate differential for capital gains and dividends will lead to an inverse relationship between value and dividends. However, their results show the opposite relation between firm value and dividends, suggesting dividends capture information regarding future cash flows missed out in the control variables used to evaluate firm value. This paper aims to re-examine this relationship. In addition I examine if share repurchases have an effect on firm value. Section 2 discussed why investors’ most likely place a premium on firms who make payouts through cash dividends and/or share repurchases. From here, I expect to see that both cash dividends and share repurchases have a positive effect on firm value. However, it is not clear if the cash dividends or the share repurchases have a bigger effect on firm value. As is shown by the current literature on the subject, the answer is by no means self-evident. In order to compare the relation between cash dividends and firm value and similarly share repurchases and firm value, I hypothesize: H10: Cash dividends have no effect on firm value
H11: There is a positive effect between cash dividends and firm value H20: Share repurchases have no effect on firm value
H21: There is a positive effect between share repurchases and firm value
3.2 Base Models
Having concluded from the literature section that the payout decision may influence firm value, this section develops a model which examines both the cash dividends and share repurchases effect on firm value. I use the value model from Fama and French (1998) as a starting point and the model will be extended by the share repurchases variable5. Previous literature shows us that both forms of payouts signal information, mitigate agency costs, but differ from tax perspective. To test whether the coefficients from cash dividends and share
repurchases are positively significant, I estimate the following regression:
(MVAi,t - Ai,t) / Ai,t = α0 + β1 DIVi,t / Ai,t + β2 dDIVi,t / Ai,t + β 3 dDIVi, t+2 / Ai,t + β4 REPi,t / Ai,t + β5 dREPi,t / Ai,t + β6 dREPi,t+2 / Ai,t + β7 Ei,t / Ai,t + β8 dEi,t / Ai,t + β9 dEi,t+2 / Ai,t +
β10 INTi,t / Ai,t + β11 dINTi,t / Ai,t + β 12 dINTi,t+2 / Ai,t +
β13 dAi,t / Ai,t + β14 dAi,t+2 / Ai,t + β15 dMVAi,t+2 / Ai,t + εi,t (1)
Where Xi,t represents variable X for firm i at time t. Also, dXi,t represents the change in X from t-2 to t, (Xi,t – Xi,t-2), while dXi,t+2 represents the change in Xi from period t to t+2. The dependent variable (MVAi,t - Ai,t) is the spread of a firm’s market value over cost. Cash dividends (DIVi,t) and share repurchases (REPi,t), account for payouts, earnings (Ei,t) controls for profit, interest expenses (INTi,t) controls for financing decisions, and assets (Ai,t) represents growth (or net investments). The past and future two year changes in the independent variables are assumed to capture investors’ expectations concerning variables growth. The intercept (α0) indicates the point where the regression line intercepts the y axis, representing the amount of the dependent variable firm value when all the independent variables are zero. The β1, β2…… β t form the calculated coefficients, representing the amount of firm value changes when the corresponding independent changes one unit. And finally, εi,t is the residual term. The dependent and the independent variables are scaled by the book value of total assets (Ai,t), to prevent that the results are dominated by the largest firms. The model provides a test of the relevance of cash dividends and share repurchases. If investors only care about profitability, then the coefficients of the profitability proxies (Ei,t) are positive and significant while the payout coefficients (DIVi,t) and (REPi,t) will be statistically not significant. However, recalling my hypotheses, I expect to see a positive and significant sign from all payout coefficients.
5
Equation (2) re-estimates equation (1) by using changes in policies related to payout and leverage instead of the changes in payout and finance levels. This results in the following regression:
(MVAi,t - Ai,t) / Ai,t = α0 + β4 DIVi,t / Ai,t + β2 d(DIVi,t / Ai,t) + β3 d(DIVi,t+2 / A i,t+2) + β4 REPi,t / Ai,t + β5 d(REPi,t / Ai,t) + β6 d(REPi,t+2 / A i,t+2)+ β7 Ei,t / Ai,t + β8 dEi,t / Ai,t + β9 dEi,t+2 / Ai,t +
β10 INTi,t / Ai,t + β11 d(INTi,t / Ai,t) + β 12 d(INTi,t+2 / Ai,t) +
β13 dAi,t / Ai,t + β14 dAi,t+2 / Ai,t + β15 dMVAi,t+2 / Ai,t + εi,t (2)
Where again Xi,t represents variable X for firm i at time t. However, dXi,t from equation (1) is replaced for d(Xi,t / Ai,t) to represent the change in (Xi / Ai) from t-2 to t, (Xi,t / Ai,t) – (Xi,t-2 / Ai,t-2). Similar reasoning applies for d(Xi,t+2 / Ai,t+2), to represents the change in (Xi / Ai) from period t to t+2. This holds for the cash dividends (DIVi,t), share repurchases
(REPi,t), and the leverage (INTi,t) variables, all in order to measure the changes in payout- and leverage policy. The main differences with equation (1) are the brackets around the payout and interest variables. As a result, the changes in the ratio of payouts to assets, and interest to assets are measured. The dependent variable (MVAi,t - Ai,t) remains the same as in equation (1) and is therefore defined as the spread of a firm’s market value over cost. Also, the control variables earnings (Ei,t), and growth (or net investments) representsed by assets
3.3 Variable construction 3.3.1 Dependent variable
Firm value (MVAi,t - Ai,t): the firm value is proxied by the sum of the market capitalization and net debt minus the book value of total assets. Similarly as Fama and French (1998) who defined this variable as the spread of value over cost. This variable is deflated by total assets to correct for size effects. This variable is a proxy for firm value. The data about the market capitalization, total net debt and total assets is retrieved from DataStream.
3.3.2 Variables of interest
Cash dividends (DIVi,t): the base models include five cash dividend variables, to test whether cash dividends affect firm value. Regression (1) measures if the level of cash dividends in year t and the change in level of cash dividends between year t-2 to year t and year t to year
t+2 affects firm value in year t. The results of Fama and French (1998) show us that the
proxy for firms’ dividend policy; cash dividends to total assets, also affects firm value. Therefore regression (2) replaces the change in level of cash dividends for the change in cash to total assets ratios, as indicated by the brackets in equation (2). This accounts between year
t-2 to year t and year t to year t+2. Again all variables are deflated by total assets. The usual
proxy for a firm’s dividend policy is cash dividends to earnings (Lintner, 1956). However cash dividends to earnings become meaningless when earnings are negative and the ratio explodes when earnings are close to zero. Another option is the cash dividend to price ratio, but this seems also inappropriate since price forms a large part of the dependent variable firm value. Therefore the cash dividends to total assets ratios are used as a proxy for the change in firms’ dividend policy. The cash dividends variable is retrieved from DataStream. According to the theory and prior results positive signs of the coefficients might be expected.
(2008) and Fama and French (2001). This paper uses balance sheet items since it nets out those cases where share repurchases are used for other purposes than payouts to investors. More specifically it excludes the influence of stocks reissued for employee stock ownership plans and repurchased stocks which are used to finance a merger6. Another issue is that firms in Europe have the option to account treasury stock by either using book values or market prices. By making an assumption it becomes possible to revalue treasury stock of firms that most likely report treasury stock at book values7. The data used to calculate the amount of share repurchases is retrieved from Worldscope. In line with the theory, positive signs of the coefficients might be expected
3.3.3 Control variables
The regressions controls for earnings, investment and financing effects on firm value. In addition the model incorporates a variable that offsets the error in the realized change in all payout - and control variables as a measure of the expected change in those variables. The control variables are scaled by total assets to put them in the same units as the dependent variable and the variables of interest.
Earnings (Ei,t): the level and changes in level of earnings are included in the models to proxy for firms’ profitability. Regression (1) and (2) measures if the level of earnings in year t and the change in level of earnings between year t-2 to year t and year t to year t+2 affects firm value in year t. The proxy variable is defined as: earnings before extraordinary items and preferred dividends after tax. For the earning variables positive coefficients are expected in line with Fama and French (1998).
Debt/Leverage (INTi,t): the level and changes in level of debt are included in regression (1) by using the total interest costs as a proxy. Regression (2) measures in addition to the level of debt the changes in leverage police, whereas the leverage ratio is estimated by total
6
Basically I follow the method of Skinner (2008) by using the change in common treasury stock between year t and year
t-1. However, the method is inappropriate in cases where the outcome delivers a result below zero. Skinner (2008) suggests
in those cases to use the retirement method as an alternative. The retirement method is calculated by taking the difference between purchases and sales of common stock. Share repurchases are set at zero when the outcome is negative and also when the purchase of common stock is unavailable.
7
Assume that the firms who use share repurchases will not outperform the market on average. When the change in treasury stock divided by treasury shares is lower (higher) valued than the years’ lowest (highest) stock price, share repurchases are assumed to be valued at book value. This holds for firms with an equity market-to-book ratio larger (smaller) than one. In those situations the following re-calculation method is applied: the numbers of shares repurchased are multiplied with the 12 month average share price during the fiscal year divided by the reported price.
Total amount of share repurchases i,t = number of shares repurchased i,t X 12 month average share price i,t
interest costs to total assets. This proxy is a direct measure of firms’s book leverage. In cases where the agency costs for intangible assets are high the target leverage may be closely related to the book leverage (Myers, 1977). Alternative proxies who measures market values (e.g. total debt to total equity) are inappropriate in this case since they are closely related to firm value. In the financial theory debt is often assumed to lower the cost of capital (due to the tax shield), and thereby creating firm value. However, the tax advantage on corporate level can be offset by taxes on investor level. The earnings will also be negatively affected by the high fixed interest costs, although operational income stays unaffected. Investors will take this into account. According to Miller (1977) negative slopes are expected, assuming that the model uses after tax earnings. The study of Fama and French (1998) confirms this expectation, by reporting a negative relation between firm value and the debt/leverage variables.
Investment opportunity (Growth): the investment opportunity is proxied by the change in
assets divided by current assets, following Fama and French (1998). Regression (1) and (2) measures if change in level of total assets between year t-2 to year t and year t to year t+2 affects firm value in year t. The growth of a firm is seen in the finance literature as one of the most important value drivers since it is expected to deliver positive future cash flows (Koller, 2005). Firms with strong growth prospects are expected to create more firm value than firms with weak growth prospects. Firm value is more likely to be maximized by financing good potential investments, than by paying out cash to shareholders. Thus positive coefficients are expected from the growth variables.
Change in future firm value (dMVAi,t+2): the research of Kothari and Shanken (1992) reports that investors have expectations about future market value that affects current market value. The change in the market value of the assets from year t to t+2 is used a proxy for future firm value. The slope of this variable offsets the error in the realized change in all payout - and control variables as a measure of the expected change in those variables. This implies the opposite sign which results in a negative impact on firm value.
Table 1: Variable definition and overview of the expected signs
Variables Definition Sign
DIVi,t / Ai,t Total cash dividends divided by total assets in year t +
dDIVi,t / Ai,t The change in cash dividends from year t-2 to t divided by total assets in year t +
dDIVi,t+2 / Ai,t The change in cash dividends from year t+2 to t divided by total assets in year t +
d(DIVi,t / Ai,t) The dividend assets ratio in year t-2 divided by the dividend assets ratio in year t +
d(DIVi,t+2 / A i,t+2) The dividend assets ratio in year t+2 divided by the dividend assets ratio in year t +
REPi,t / Ai,t Total share repurchases divided by total assets in year t +
dREPi,t / Ai,t The change in share repurchases from year t-2 to t divided by total assets in year t +
dREPi,t+2 / Ai,t The change in share repurchases from year t+2 to t divided by total assets in year t +
d(REPi,t / Ai,t) The share repurchase assets ratio in year t-2 divided by the share repurchases assets ratio in year t +
d(REPi,t+2 / A i,t+2) The share repurchases assets ratio in year t+2 divided by the share repurchases assets ratio in year t +
Control variables
Ei,t / Ai,t Total earnings before tax after depreciation divided by total assets in year t +
dEi,t / Ai,t The change in earnings from year t-2 to t divided by total assets in year t +
dEi,t+2 / Ai,t The change in earnings from year t+2 to t divided by total assets in year t +
INTi,t / Ai,t Total interest divided by total assets in year t
-dINTi,t / Ai,t The change in interest from year t-2 to t divided by total assets in year t
-dINTi,t+2 / Ai,t The change in interest from year t+2 to t divided by total assets in year t
-d(INTi,t / Ai,t) The interest assets ratio in year t-2 divided by the dividend assets ratio in year t
-dINTi,t+2 / Ai,t The interest assets ratio in year t+2 divided by the dividend assets ratio in year t
-dAi,t / Ai,t The change in assets from year t-2 to t divided by total assets in year t +
dAi,t+2 / Ai,t The change in assets from year t+2 to t divided by total assets in year t +
dMVAi,t+2 / Ai,t The change in market value assets from year t+2 to t divided by total assets in year t
-3.4 Measurement technique and issues
3.4.1 The Fama and MacBeth methodology
One concern is cross-sectional correlation among firms in the pooled sample. To address this I use average coefficients from annual cross-sectional regressions. This idea originates from the Fama and MacBeth two-step method which was developed for datasets with long time series. However, it turns out that this procedure can easily be modified in such a way that is applicable for panel data with large number of observations in a short time frame. The procedure is as follows. In the first step, for each single time period I perform a cross-sectional regression. Then, in the second step, I obtain the final coefficient estimates by taking the average of the first step coefficient estimates. The t-statistics can be derived by dividing the average coefficients through the standard error. Cochrane (2001) demonstrates that this methodology is essentially equivalent to panel data approach, under the assumption that the explanatory variables are constant through time. However, this last assumption is extremely restrictive and rather unrealistic as shown in previous research (Bates, Kahle, and Stulz, 2009; Von Eije and Megginson, 2008; Grullon and Michaely, 2002).
3.4.2 Panel data methodology
According to Hsiao (1985) the main advantage of panel analysis in comparison to a cross-sectional or time serie analysis is that it permits to study the dynamics of change in short time series. The combination of time series with cross-sections enhances the quality and quantity of data in ways that would be impossible using only one of these two dimensions. Moreover panel data usually contains more degrees of freedom and thereby it increases the power of the test. To run a panel data regression by using ordinary least squares there are three possibilities: a common regression which uses the same intercept for all pool members, fixed effects which uses separate intercepts estimated for each pool member, and random effects which treats intercepts as random variables across pool members. The three approaches all have their supporting and opposing arguments which are explained next.
Pooled regression: the pooled regression has constant coefficients, referring to both
intercepts and slopes, as shown in equation (3). This implicitly assumes that the average value of all variables and their relationship between them are constant over time and across all of the cross-sectional units (Brooks, 2008). This is a very restrictive model, and to overcome the limitations of this pooled regression, I will first test whether less restrictive approaches are more appropriate.
Fixed effects model: in the fixed effects model the error term is decomposed in two
time. Equation (5) includes an individual effect, which differs over time, but is constant for each cross-sectional unit. Finally equation (6) allows for both cross-sectional and time fixed effects in the same model. These alternative models allow having more than one intercept α, which differs across firms and / or over time. In that way the model is appropriate to identify and control for omitted variables that are constant by firm or over time. To examine whether individual firm effects or time effects are present, I conduct a redundant fixed effects test. This test assigns a dummy for each firm or time variable and examines if there are significant differences between each firm and / or time variable. If the null hypothesis is rejected it would be inappropriate to pool all data together (assuming one intercept α) as explained in the pooled regression. I note that the intercept term (α) from the fixed effects model is often removed to avoid the ‘dummy variable trap’, which arises in case there is perfect multicollinearity between the dummy variables and the intercept. One drawback of the fixed effects model is that each cross-sectional or time variable requires a dummy variable. As a result the number of degrees of freedom is reduced, which makes the test less powerful. This drawback can be avoided through the use of the random effects models.
Random effects models: Similar as the fixed effects model, this approach recognizes
differences in the intercepts α. More importantly it has the advantage that it does not use dummy variables to capture the heterogeneity in cross-section and/or time. This implicates that the number of degrees of freedom are saved. However, this approach may be inappropriate if the individual effects represent omitted variables which are likely to be correlated with the explanatory variables in the model. Moreover, the model requires additional restrictions. For example the model assumes that the error term (εi,t)has a zero mean, at constant variance, and it is independent of all explanatory variables (Xi,t) and the cross-section specific error term (vt). Equation (7) shows the model. By using a Hausman test, I will test whether the requirement of no correlation between the error terms and the explanatory variables is valid. If the null hypothesis is rejected, it is not appropriate to use this random effects model.
Pooled regression: Yi,t = α0 + β1 Xi,t + εi,t (3)
Cross-sectional fixed effects: Yi,t = α0 + β1 Xi,t + ui + εi,t (4)
Time-fixed effects: Yi,t = α0 + β1 Xi,t + vt + εi,t (5)
Time and cross-sectional fixed effects: Yi,t = α0 + β1 Xi,t + ui + vt + εi,t (6)
Random effects: Yi,t = α0 + β1 Xi,t + ωi,t where ωi,t = εi,t + vt (7)
Where (Yi,t)is the dependent variable, (α) is the intercept term, (β) are the k x 1 vector of parameters to be
estimated, (Xi,t) are the k x 1 vector of observations on the explanatory variables, (εi,t) is the error term, (ui)is
the firm specific error term, (vt) is the time specific error term, and (ωi,t) is the error term of the random effects
3.4.3 Comparative analysis
This test compares if each cash dividend variable statistically differs from each share repurchase variable. For example, the coefficient of the current level of cash dividends is compared to the coefficient of the current level of share repurchases. By comparing the coefficients of the different payout variables, it becomes possible to test hypothesis (3), which states that the cash dividends effect differs from the share repurchases effect on firm value. Regression (1) and regression (2) both estimate the coefficients of three payout variables for both forms of payout (cash dividends and share repurchases). For each cash dividend coefficient in the model there will be an associated share repurchase coefficient. Recalling that the nullhypothesis assumes an equal effect, I will examine in both regressions the following hypotheses:
H30: β1 = β4, β2 = β5 and β3 = β6 H31: β1 ≠ β4, β2 ≠ β5 and β3 ≠ β6
The Wald test is a way of testing the significance of the assumed equality in the coefficients of both forms of payout variables8. If the Wald test is significant for a pair of payout coefficients, then I will conclude that the coefficients are not equal. In other words separate variable coefficients should be included in the model, as shown in regression (1) and (2) in section 3.2. If the Wald test is not significant, then one of the coefficients can be omitted from the model and the two explanatory variables can be added up. To illustrate, regression (3) assumes that all nullhypotheses cannot be rejected, as shown below.
(MVAi,t - Ai,t) / Ai,t = α0 + β1 (DIVi,t / Ai,t + REPi,t / Ai,t) + β2 (dDIVi,t / Ai,t + dREPi,t / Ai,t) + β3 (dDIVi, t+2 / Ai,t + dREPi,t+2 / Ai,t) + β4 Ei,t / Ai,t + β5 dEi,t / Ai,t + β6 dEi,t+2 / Ai,t + β7 INTi,t / Ai,t + β8 dINTi,t / Ai,t + β 9 dINTi,t+2 / Ai,t + β10 dAi,t / Ai,t + β11 dAi,t+2 / Ai,t + β12 dMVAi,t+2 / Ai,t + εi,t (3)
8
4 DATA AND DESCRIPTIVE STATISTICS
4.1 Data
This paper examines whether cash dividends and share repurchases are firm enhancing activities. In order to answer this question I use recent data from the United Kingdom (U.K.). The timeframe is selected from 2002 until 2006. This period is the only upturn period in the 21st century. According to the findings of Grullon and Michaely (2002), share repurchases are often paid in addition to dividend payments, which will more likely occur in an upturn period. This is induced by the increasing cash levels of firms during this period. The sample selected is based on available ISIN codes from U.K. listed firms. However, data requirements limit the size of my sample. For example, the unrestricted sample has 4365 unique firms, but only 1446 of those firms have complete data over the complete sample period. Furthermore I exclude financials and utilities in line with Trojanowski and Renneboog (2005), which results in a sample of 898 firms and a total of 4490 observations. Financials are excluded due to different reporting standards. Utilities are excluded due to regulation that impacts their payout policies and access to external financing. As all variables are scaled by total assets, I trim the total assets variables on a 0.5% level. This is in order to prevent that the results are influenced by high ratios, caused by unrealistically small denominators. As a result the total sample consists of 717 firms and a total of 3585 observations. The constructed variables are found in the DataStream database, except the variable share repurchases which is derived from the Worldscope. This variable is altered by taking into account different reporting options of firms9. An overview of all variables and their specific codes are shown in appendix A.
4.2 Descriptive statistics
Table 1 with the summary statistics reveals that the mean ratio of the dependent variable (MVAi,t - Ai,t) / Ai,t is positive with 0.615; however the median ratio is negative with 0.026. This ratio can be interpreted as the difference between the market value and the book value of the assets divided by the book value of the assets. The high standard deviation (3.428) of the dependent variable shows that the data are spread out over a large range of values. This high standard deviation also appears in all share repurchases and investment ratios. This also holds for the change in the earnings and market value ratios. Another point of interest is that the mean of the share repurchase to assets ratio is higher than the dividend
9
to assets ratio. This implicates that firms which actually pay out through share repurchases will do this with a relatively high amount in comparison to their total size proxied by total assets.
Table 2: Descriptive statistics
Variable Observations Mean Median Std. Dev. Regression
(MVAi,t - Ai,t) / Ai,t 1017 0.615 -0.026 3.428 1.2
DIVi,t / Ai,t 1017 0.022 0.013 0.083 1.2 dDIVi,t / Ai,t 1017 0.003 0.000 0.084 1 dDIVi,t+2 / Ai,t 1017 0.001 0.001 0.082 1 d(DIVi,t / Ai,t) 1017 0.002 0.000 0.084 2 d(DIVi,t+2 / A i,t+2) 1017 -0.003 0.000 0.081 2 REPi,t / Ai,t 1017 0.096 0.000 2.365 1.2 dREPi,t / Ai,t 1017 0.089 0.000 2.367 1 dREPi,t+2 / Ai,t 1017 0.116 0.000 4.494 1 d(REPi,t / Ai,t) 1017 0.089 0.000 2.367 2 d(REPi,t+2 / A i,t+2) 1017 0.086 0.000 3.882 2 Ei,t / Ai,t 1017 0.059 0.036 0.528 1.2 dEi,t / Ai,t 1017 0.141 0.018 1.523 1.2 dEi,t+2 / Ai,t 1017 0.062 0.022 0.600 1.2 INTi,t / Ai,t 1017 0.012 0.008 0.018 1.2 dINTi,t / Ai,t 1017 -0.002 0.000 0.031 1 dINTi,t+2 / Ai,t 1017 0.005 0.000 0.025 1 d(INTi,t / Ai,t) 1017 -0.001 0.000 0.015 2 dINTi,t+2 / Ai,t 1017 0.004 0.000 0.072 2 dAi,t / Ai,t 1017 -0.108 0.078 1.515 1.2 dAi,t+2 / Ai,t 1017 0.513 0.186 2.685 1.2 dMVAi,t+2 / Ai,t 1017 0.978 0.117 12.313 1.2
Descriptive statistics of all variables used in regression 1 and 2. The period estimate is from 31/12/2002 untill 31/12/2006. Where Xi,t represents variable X for firm i at time t. Also, dXi,t and represents the change in X from t-2 to t, Xi,t – Xi,t-2, while dXi,t+2 represents the change in X from period t to t+2. This accounts for the market value (MV), total assets (A), cash dividends (DIV), Share repurchases (REP), Earnings before interest after depreciation (E) and Interest (INT). The constructed variables are explained in table 1.
5 RESULTS
This section describes the results of the regressions, all in order to answer the main research question: ‘‘are payments of cash dividends and share repurchases value enhancing
activities? ’’ The results start with the estimates of regression (1) which examines the level
and changes in the level of cash dividends and share repurchases variables. The results reveal first the Fama and MacBeth approach and second the panel data approach. Then in a similar way I will estimate regression (2) which examines the level of payouts and the changes in payout policy. From these results I provide the answer on the first two hypotheses, whether there is a cash dividends and / or share repurchases effect on firm value. In order to answer the third hypothesis, whether there is a significant difference between the cash dividends and share repurchases effect on firm value, I conduct a comparative analysis. The estimates of the profitability, financing, investment, and future expectations about the firm market value are incorporated as control variables. Finally, a robustness check is performed by using and alternative measure of firm value.
5.1 Level and changes in level of payouts
The results obtained from the estimation of regression model (1) according to the Fama and Macbeth methodology are presented in table 3. The estimates are calculated by taking the average of the coefficients of the annual cross-sectional regressions. The t-statistics are calculated, by dividing the average coefficients by the standard error.
Table 3: Results regression (1) by using the Fama and MacBeth method, to explain the level of (MVAt-At)/At
Variable Coëfficiënt T-statistic
DIVi,t / Ai,t 9.243 6.316 ** dDIVi,t / Ai,t 6.667 1.396 dDIVi,t+2 / Ai,t 8.849 10.772 *** REPi,t / Ai,t 7.853 1.695 * dREPi,t / Ai,t -6.722 -1.389 dREPi,t+2 / Ai,t 0.575 1.042 Ei,t / Ai,t -2.340 -7.766 *** dEi,t / Ai,t 0.402 2.075 *** dEi,t+2 / Ai,t -0.686 -2.085 ** INTi,t / Ai,t 0.726 0.285 dINTi,t / Ai,t -4.514 -1.132 dINTi,t+2 / Ai,t -1.122 -0.496 dAi,t / Ai,t 0.187 2.024 *** dAi,t+2 / Ai,t 0.477 2.134 *** dMVAi,t+2 / Ai,t -0.353 -1.807 *** Constant -0.108 -0.859 Significance levels: * = 10%, ** = 5%, *** = 1%
Table 3 shows that the current level of cash dividends and share repurchases are both positively related to firm value. The level of cash dividends (DIVi,t / Ai,t) shows a value of 9.243, which is interpreted as a nine units increase in market value of a firm if the independent variable level of cash dividends increases by one unit, holding all other independent variables constant. The level of share repurchases (REPi,t / Ai,t) has a slope of 7.853. The size of the cash dividends slope with 9.243 is higher in relation to prior results of Fama and French (1998), who find a average slope of 4.220. Statistically, the cash dividend is significant on a five percent significance level and the share repurchases is significant on a ten percent significance level. These results are consistent with hypotheses (1) and (2). Recalling that hypotheses (1) and (2) predict that there is a positive relation between cash dividends and firm value, and a positive relation between share repurchases and firm value. The results also show us that the level of cash dividends variable is stronger related to firm value than the level of share repurchases to firm value, as stated in hypothesis (3). However, it is invalid to draw any conclusion about hypothesis (3) without using a comparative analysis, which is discussed later on in section 5.3.
The slope of the change in the past cash dividends variable (dDIVi,t / Ai,t) does not confirm hypothesis (1), as the positive slope of 6.667 is not significant (corresponding
t-statistic of 1.369). On the other hand the future cash dividends variable (dDIVi,t+2 / Ai,t) shows a highly significant positive relation to firm value, with a slope of 8.849 (corresponding t-statistic of 10.772), which again is in line with hypothesis (1). The slopes of the change in the past and future share repurchases variables (dREPi,t / Ai,t) and (dREPi,t+2 /
Ai,t) are unable to support hypothesis (2), that predicts a positive relation between share repurchases and firm value, as both slopes are not significant. I note that the former variable shows a negative slope of -6.722.
In the control variables there is an inconsistency between the findings of this study and the study of Fama and French (1998). The current level of earnings (Ei,t / Ai,t), the change in future earnings (dEi,t+2 / Ai,t) and the current level of interest (INTi,t / Ai,t) show the opposite signs of what might be expected, although the latter is not significant. The sign of all other control variables are in line with the findings of Fama and French (1998). Finally, the future change in market value (dMVAi,t+2 / Ai,t), which is meant to purge other future changes of their unexpected components, shows a negative slope (-0.353), as expected10.
10
The results of the panel data regression (1) are shown in table 4. The redundant fixed effects test shows that the period p-values are significantly different from zero. So, it is not valid to pool all data together (assuming one intercept) across years. This implies that the intercepts are allowed to vary over time but are constant cross-sectionally. The Hausman test which assumes error terms to be uncorrelated with the explanatory variables shows us that the p-values are significantly different from zero. As a result, I estimate regression (1) by using period fixed effects, indicated in table 4 by means of year dummies. To prevent the ‘dummy variable trap’ as described by Brook (2008), I exclude the first year dummy.
Table 4: Results regression (1) by using period fixed panel data, to explain the level of (MVAt-At)/At
Variable Coëfficiënt T-statistic
DIVi,t / Ai,t 11.972 3.215 *** dDIVi,t / Ai,t 0.024 0.008 dDIVi,t+2 / Ai,t 7.467 2.029 ** REPi,t / Ai,t -0.268 -0.214 dREPi,t / Ai,t 0.313 0.251 dREPi,t+2 / Ai,t 0.004 0.151 Ei,t / Ai,t -3.086 -8.378 *** dEi,t / Ai,t 0.535 4.487 *** dEi,t+2 / Ai,t -1.233 -4.567 *** INTi,t / Ai,t -5.728 -0.976 dINTi,t / Ai,t -2.587 -0.609 dINTi,t+2 / Ai,t -1.233 -0.239 dAi,t / Ai,t 0.489 3.442 *** dAi,t+2 / Ai,t 0.098 1.844 * dMVAi,t+2 / Ai,t -0.073 -8.102 *** Constant 0.145 0.651 DUM2003 0.148 0.564 DUM2004 0.246 0.685 DUM2005 0.033 0.119 DUM2006 1.500 3.075 *** Adjusted R-Squared 0.142 Observations (N) 1017 Significance levels: * = 10%, ** = 5%, *** = 1%
This table reports the coefficients and their t-statistics from the estimates of regression (1). Where Xi,t represents variable X for firm i at time t. Also, dXi,t and represents the change in X from t-2 to t, Xi,t – Xi,t-2, while dXi,t+2 represents the change in X from period t to t+2. This accounts for cash dividends (DIV), Share repurchases (REP), Earnings before interest after depreciation (E), Interest (INT), total assets (A), the market value of the assets (MVA). The constructed variables are explained in table 1. The time period for t is from 2002 to 2006.
The slope of the change in the past cash dividends variable (dDIVi,t / Ai,t) is still positive with a value of 0.024, but not significant with a t-statistic of 0.008. The slope of the future cash dividends variable (dDIVi,t+2 / Ai,t) is significantly positive with a value of 7.467 (corresponding t-statistic of 2.029), in line with hypothesis (1). The slope of the change in the past share repurchases variable (dREPi,t / Ai,t) is small with a coefficient of 0.313 and a t-statistic of 0.251. Although, the positive relation of the past share repurchases variable to firm value is in line with hypothesis (2), the results are not significant. Also, the future share repurchases variable (dREPi,t+2 / Ai,t) show a not significant positive relation to firm value, with a small slope of 0.004 and a t-statistic of 0.151.
From the control variables I note that the current level of earnings (Ei,t / Ai,t) and the change in future earnings variable (dEi,t+2 / Ai,t) are again both highly significant negatively related to firm value. On the other hand the change in the past earnings variable (dEi,t / Ai,t) is highly significant positive. Moreover the interest variables (INTi,t / Ai,t), (dINTi,t / Ai,t) and
(dINTi,t+2 / Ai,t) are all negative but not significant. Furthermore the investment variables
5.2 Level of payouts and changes in payout policy
Regression (2) is almost the same as the former regression examined in section 5.1, except that the changes in cash dividends and changes in share repurchases are replaced for changes in the cash dividends to assets ratio and the share repurchases to assets ratio. In line with Fama and French (1998), the new specification addresses the change in policies related to dividend and repurchases. In order to be consistent the change in the financing control variables are adapted in a similar way. The differences are visible by the brackets around the change variables, as shown in the table 5 and 6. All other variables are similar defined as in section 5.1.
Table 5 shows us that the current level of cash dividends and share repurchases are strongly related to firm value by using the Fama and MacBeth approach. For example, the slope of the level of cash dividends (DIVi,t / Ai,t) shows a coefficient of 11.602 and a t-statistic of 11.029. The size of the cash dividends slope with 11.602 is close to the results in table 4, but higher than the results in table 3. Furthermore the current level of share repurchases (REPi,t /
Ai,t) shows a slope of 7.660 and a t-statistic of 1.967. This slope is higher than the results in table 4, but close to the slope reported in table 3. Statistically the postive relation is significant on a five percent significance level which is higher compared to the ten percent significance level in table 3. Again, these results are consistent with hypotheses (1) and (2).
Table 5: Results regression (2) by using the Fama and MacBeth method, to explain the level of (MVAt-At)/At
Variable Coëfficiënt T-statistic
DIVi,t / Ai,t 11.602 11.029 *** d(DIVi,t / Ai,t) 6.168 1.664 * d(DIVi,t+2 / A i,t+2) 7.028 2.923 *** REPi,t / Ai,t 7.660 1.967 ** d(REPi,t / Ai,t) -6.600 -1.520 d(REPi,t+2 / A i,t+2) 0.801 1.029 Ei,t / Ai,t -2.721 -9.387 *** dEi,t / Ai,t 0.349 1.012 dEi,t+2 / Ai,t -0.681 -2.002 ** INTi,t / Ai,t 1.286 0.478 d(INTi,t / Ai,t) -12.143 -5.917 *** dINTi,t+2 / Ai,t 0.522 0.158 dAi,t / Ai,t 0.207 3.435 *** dAi,t+2 / Ai,t 0.802 3.064 *** dMVAi,t+2 / Ai,t -0.540 -2.716 *** Constant -0.181 -1.350 Significance levels: * = 10%, ** = 5%, *** = 1%
The slope of the change in the past cash dividends ratio d(DIVi,t / Ai,t) is positive with a slope of 6.168, but only marginally significant with a t-statistic of 1.664. The slope of change in the future cash dividends d(DIVi,t+2 / Ai,t+2) is also positive with a slope of 7.028 and highly significant (corresponding t-statistic of 2.923). Thus both change variables confirm hypothesis (1). On the other hand the past and future change in share repurchases policies,
d(REPi,t / Ai,t) and d(REPi,t+2 / Ai,t+2), are not doing such a good job in explaining firm value, as both slopes are not significant. The sign of the former variable is even negative with a slope of -6.600, similarly as shown in table 3. The result of the change in policies related to the share repurchase do not confirm hypothesis (2).
The results of the panel data regression (2) are shown in table 6. The redundant fixed effects and the Hausman test show us that the regressions can be best estimated by using period fixed effects, indicated in table 6 by means of year dummies. Again, to prevent the ‘dummy variable trap’ as described by Brook (2008), I exclude the first year dummy.
Table 6: Results regression (2) by using period fixed panel data, to explain the level of (MVAt-At)/At
Variable Coëfficiënt T-statistic
DIVi,t / Ai,t 10.437 2.562 *** d(DIVi,t / Ai,t) -0.128 -0.041 d(DIVi,t+2 / A i,t+2) 5.893 1.355 REPi,t / Ai,t -0.377 -0.286 d(REPi,t / Ai,t) 0.423 0.320 d(REPi,t+2 / A i,t+2) 0.003 0.106 Ei,t / Ai,t -2.926 -8.193 *** dEi,t / Ai,t 0.485 4.476 *** dEi,t+2 / Ai,t -1.150 -4.303 *** INTi,t / Ai,t -9.314 -1.233 d(INTi,t / Ai,t) -6.452 -0.825 d(INTi,t+2 / Ai,t) 1.937 1.083 dAi,t / Ai,t 0.418 3.471 *** dAi,t+2 / Ai,t 0.108 2.058 ** dMVAi,t+2 / Ai,t -0.073 -8.078 *** Constant 0.235 1.025 DUM2003 0.142 0.547 DUM2004 0.239 0.670 DUM2005 0.043 0.157 DUM2006 1.512 3.089 *** Adjusted R-Squared 0.141 Observations (N) 1017 Significance levels: * = 10%, ** = 5%, *** = 1%
