Firm size, the earned/contributed capital mix, and payout policy
of UK firms
Dilyan Markov1
Abstract: This paper tests the sensitivity of small and large firms to the RETE ratio as a driver of the
probability to pay dividends or repurchase shares. The main result is that small firms are more sensitive to the RETE ratio, as a determinant of dividend policy. This is attributed to signaling theory, because small firms with high earned capital would be the best candidates to pay, as they need to overcome information asymmetry, and this is thus an indication that small firms signal more than large firms. There is not enough evidence to conclude that the sensitivity to the RETE ratio is different for small and large repurchasing firms. This might be due to the fact that repurchases are not as binding as dividends. A new result for repurchases is that the RETE ratio is positively related with the probability to repurchase shares, when controlling for size, growth, profitability and cash holdings, so lifecycle theory holds for repurchases as well.
Keywords: Payout policy, Dividends, Repurchases, Earned/contributed capital mix, Firm size,
Signaling, Agency, Lifecycle
JEL Classification: G32; G35
Supervisor: Dr. J.H. von Eije
1
1. INTRODUCTION
On a regular basis, firms face the decisions of how much cash to return to shareholders and what form the payment should take, in terms of dividends or repurchases. Because these decisions change over time they are labeled as payout policy, implying that they do not arise in a random manner (Allen and Michaely, 2003). The importance of corporate payout policy is subject to a debate over the years. Lintner (1956) concludes that dividend policy is influenced by current net earnings, investment decisions are residual of dividend policy, and companies smooth their dividends. Modigliani and Miller (1961) claim that in a perfect world with perfect capital markets, characterized by rational behavior and perfect certainty, and where FCF is paid out for 100%, the payout policy of a company, given its investment policy, does not influence the price of its shares and is thus irrelevant. After Modigliani and Miller’s (1961) irrelevance theory, researchers have tried to relax the strict assumptions under which it holds and show that indeed, in the presence of market imperfections, the payout policy of companies is highly relevant. Two of the main market imperfections that have been proposed to have influence on payout policy are: asymmetric information and agency problems (Dhanani, 2005). The former is believed to influence payout policy, because payouts can be costly tools which convey information to the market for the future prospects of a company (Bhattacharya, 1979; Miller and Rock, 1985) and the latter has effect on dividends and repurchases because payments can be used to remove excess funds from the hands of managers and thus minimize overinvesting (Jensen, 1986).
A relatively new advancement in the literature is the lifecycle theory, which suggests that a tradeoff between retention and distribution evolves over time as profits accumulate and investment opportunities decline, so that paying dividends becomes more desirable as firms mature (DeAngelo, DeAngelo and Stulz, 2006). Empirical research so far has confirmed the lifecycle hypothesis, by using the earned/ contributed capital mix (also known as the retained earnings to total equity ratio - RETE) as a measure for a firm’s maturity (DeAngelo, DeAngelo and Stulz, 2006; Denis and Osobov, 2008). The RETE ratio is assumed to capture firm maturity, but nothing is yet known for the impact of the RETE ratio on payouts of small or large firms. This paper exploits this gap in the literature and the main research question is:
Is there a difference between the sensitivity to the earned/contributed capital mix, as a determinant of payout policy, in small and large UK firms?
2
pressure by investors to pay than small firms because of agency problems. This is because typically large firms are associated with free cash flow behavior (Vogt, 1994; Jensen, 1986) and small firms are characterized with bigger information asymmetry (Mougoue and Rao, 2003). Therefore, there are two possible directions of the impact of the earned/ contributed capital mix: (1) large firms might be more sensitive than small firms to the RETE ratio as a determinant of their payout policies, because its effect might be amplified by bigger agency costs for large firms, so that stock holders put more pressure on managers to pay out the excess cash; (2) small firms might be more sensitive than large firms, to the earned/ contributed capital mix, because its effect might be augmented by the desire to signal and thus overcome the bigger information asymmetry and difficulty in attracting outside funding. In order to test these propositions, a multivariate logit model is utilized as in Fama and French (2001). The model takes the payment/ non-payment of dividends or repurchase/ non-repurchase of shares as a dependent variable and measures for size, profitability, growth, cash holdings and the RETE ratio as control variables. The hypotheses are tested by constructing an interaction variable (SIZExRETE), which makes it possible to analyze whether the effect of RETE increases or decreases with size. The data used for the regressions are retrieved from the Datastream (Worldscope) database, and only non-financial and non-utility UK firms, listed on the London Stock Exchange, during the period 1989-2012, are used for the analysis.
The main result of this paper is that small firms are more sensitive to the earned/ contributed capital mix as a determinant of their decision to pay dividends. The coefficient estimates of the interaction variable are significant for both specifications of the size measure for dividends. This result does not reject that there is influence of the earned/ contributed capital mix on dividend policies of large companies - it simply states that a small company with a given RETE ratio is more likely to pay dividends than a large company with the same RETE ratio. This can be attributed to signaling, and it indicates that small companies use dividends more as a signaling mechanism than large companies. Typically when a firm is small it is characterized by lower amount of earned capital, bigger information asymmetry, and hence more difficulty in raising financing. Therefore, if a small firm has amassed a substantial amount of slack (so it has high RETE ratio) it can choose to pay dividends in order to signal its promising future and attract funds for its projects.
3
the results are insignificant for repurchases, because this type of payout is less binding than dividends (Jagannathan, Stephent and Weisbach, 2000), and thus dividends are preferred for signaling. A new result for the literature for repurchases, however, is the fact that the RETE ratio is a significant driver of the probability of firms to repurchase their shares even after controlling for size, growth, profitability and cash holdings. Therefore, lifecycle theory is confirmed for repurchases, too, although the effect is weaker than for dividends, which also might be a possible explanation for the insignificant interaction coefficient for repurchases.
Furthermore, consistent with prior empirical research (Fama and French, 2001; Von Eije and Megginson, 2008; Denis and Osobov, 2008), larger, more profitable, and less growing companies are more likely to pay dividends or repurchase shares in the UK. In addition, more cash holdings do not necessarily mean that a company is more likely to pay out funds to shareholders, since the results show an inverse relationship between cash holdings and the probability to pay dividends. Companies with more cash are, however, more likely to repurchase shares.
Finally, this paper also shows several patterns of payout policy in UK firms. First, consistent with Fama and French (2001) the proportion of UK firms that pay dividends decreases with almost 50% during the period 1989-2012. This decrease is strongly influenced by small companies as the decrease in the proportion of small firms that pay dividends is twice the decrease in the proportion of large companies that pay. Second, the fraction of UK firms that repurchase shares increases during the years, which is in line with Skinner (2008). This increase is mainly driven by large companies. Third, the aggregate nominal amount of dividends and repurchases increases, despite the decreasing proportion of firms that pay dividends or repurchase shares. More than 85% of this amount is paid out by large firms for each of the 24 years in the sample period.
This paper continues as follows: section 2 outlines the relevant literature, section 3 states the hypotheses and the research method used to test them, section 4 describes the data in detail and gives patterns of payout policy of UK firms, section 5 presents the empirical results, and finally, section 6 concludes.
2. LITERATURE REVIEW
4
2.1 Asymmetric information, signaling theory, and payout policy
If the assumptions of Modigliani and Miller (1961) are relaxed, dividends and repurchases may have “informational content”. If managers know more about their company than the outside world (there is a high level of information asymmetries), dividends can be used to convey information not previously known to the market, or they can be used as a costly signal to change market opinion about future prospects (Allen and Michaely, 2003). Similarly, a repurchase announcement is one way for a firm to convey its profitability to the market. When the stock of a repurchasing firm is undervalued, the firm may have an incentive to announce a share buyback, because it will be seen as a good self-investment (Liang, 2012).
Miller and Rock (1985), argue that in a world of rational expectations and information asymmetry, firms’ dividend payout announcements serve as a costly signaling tool to the outside world for the current earnings situation. The intuition is that a high dividend payment signals high earnings, which in turn increases the share price, thus current investors can sell their shares to earn a profit. The cost of signalling that attribute to the market by increasing (net) dividends is the foregone use of the funds in productive investment (Miller & Rock, 1985). Bhattacharya (1979) develops a signalling model where firms commit to dividends to inform investors about the quality of their projects. If the payoff of the projects is not sufficient to pay the dividends, the firm has to resort to capital markets to raise funds in order to pay to stock holders, which is a high (transaction) cost to the firm. Thus, investors see firms with higher payout as ones that have more profitable projects. The two models above can reasonably be applied to repurchases as well as for dividends (Allen and Michaely, 2003).
5
repurchase tend to perform abnormally well in the long run. Furthermore, they also find that value stocks (high B/M ratio) of firms that announce share repurchase have the highest abnormal buy and hold returns. The results are consistent with the notion that undervalued firms can use their payout policies as a signal to the market. Liang (2012) analyzes the outperformance of firms with high B/M ratios that repurchase shares compared to ones with low B/M. He decomposes the B/M ratio into tangible and intangible components and finds out that it is the intangible part that drives the higher short term and long term abnormal returns for firms with higher B/M ratios. Thus, high B/M firms exhibit positive abnormal returns when announcing repurchases, because investors see their intangible value (future operating performance) as undervalued.
On the other hand, the empirical literature does not find a positive relationship between increase in dividends and repurchases, and future earnings performance. For example, Grullon, Michaely and Swaminathan (2002) conclude that dividend changes are negatively related to future ROA. Grullon and Michaely (2004) do not find evidence that shows a positive operating performance following announcements of share repurchases. If the positive prospects are not a result of increase in earnings they might come from decrease in discount rates. Grullon, Michaely and Swaminathan (2002) report that dividend increasing firms experience a decline in the factor loadings of the market portfolio, SMB and HML factors of the Fama-French (1993) three factor model. This decline translates to a decrease in risk premium of about 1% per year for dividend increasing firms, for the 3-year post announcement period. Grullon and Michaely (2004) confirm the results for repurchases.
In general, empirical research has not reached a consensus on what exactly dividends or repurchases signal, however, it is widely recognized that dividend/ repurchase initiations and increases are followed by a positive market reaction.
2.2 Agency problems and corporate payout policy
6
(shareholders). For example, managers can spend company resources on perquisites or make suboptimal investments, or in other words overinvest. Because of this, principals spend resources to monitor managers and incur monitoring costs. Agent would also spend resources to show their good intentions and will in turn incur bonding costs. Even after these actions, the interests are not perfectly aligned, which results in a residual loss. The sum of the bonding costs, monitoring costs, and the residual loss, forms the so called agency costs, which reduce the value of the firm. If contracts are incomplete or not fully enforceable, shareholders might payout cash to reduce agency costs between them and the managers (Allen and Michaely, 2003).
Conflicts of interest between shareholders and managers over payout policies are especially severe when the organization generates substantial free cash flow (Jensen, 1986). Jensen (1986) defines free cash flow as the excess cash that remains within the firm after all positive NPV projects are financed. The problem therefore becomes how to remove this excess cash from the managers’ discretion and prevent it from being wasted. One way to do this is to increase dividends or repurchase shares in order to reduce the slack, which can otherwise be misused (Jensen, 1986). Investors may interpret a decrease in dividends as bad news because they are not sure managers would use the additional resources wisely (Grullon, Michaely and Swaminathan, 2002). Furthermore, Easterbrook (1984) claims that paying out excess cash either by dividends or repurchases will make firms return to capital markets, which in turn makes monitoring less expensive and reduces agency costs.
7
positive impact on the stock price of firms with Q less than one but no impact on firms with Q larger than one. This evidence is consistent with the conjecture that dividend changes for firms with high agency costs signal information about investment policies (Lang and Litzenberger, 1989).
La Porta et al. (2000) develop two hypotheses about the relationship between the payout ratio of firms and agency costs. The first is that the payout ratios are higher for companies in countries with high legal protection (and hence lower agency costs) since shareholders are in a better position to make insiders to pay out excess cash - the outcome hypothesis. The second is that the payout ratios are higher, in countries with weak legal protection (high agency costs), since companies would want to establish a good reputation for not expropriating shareholders by paying out the excess cash flows - the substitution hypothesis. Their results show that firms operating in countries where shareholders are better protected pay more dividends. Furthermore, in such countries high growth firms are found to pay less than low growth firms – this is a result which is consistent with agency theory and supporting the outcome hypothesis (La Porta et al., 2000).
Whereas, La Porta et al. (2000) confirm only the outcome hypothesis, Brockman and Unlu (2011) find empirical evidence for both, by using disclosure quality instead of shareholder protection as a proxy for agency costs for a global sample of firms from 31 countries. They find a u-shaped relationship between the likelihood to pay and disclosure quality. Firms in transparent disclosure environments are more likely to pay dividends because shareholders can more accurately measure excess cash and in turn make managers to pay it out, whereas firms in opaque disclosure environments are more likely to pay because they want to signal their intentions to minimize agency costs (Brockman and Unlu, 2011).
Lie (2000) finds positive price effects for firms that pay special dividends and repurchase shares and have excess cash above the industry median. Furthermore, there is a positive relation between stock price, and special dividends or repurchases for firms with high cash levels and low Tobin’s Q, which is consistent with the agency theory explanation of payout policy (Lie, 2000). More recently, Gangopadhyay and Yook (2010) confirm these results for repurchases.
2.3 Lifecycle theory
8
opportunities decline, so that paying dividends becomes more desirable as firms mature (DeAngelo, DeAngelo and Stulz, 2006).
The process in which a firm moves from a growth phase to a more mature phase is characterized by declining investment opportunities set, declining reinvestment rates, declining return on investment and declining risk (Grullon, Michaely and Swaminathan, 2002). Generally, a company which is in its growth phase has many investment opportunities, earns large economic profits, has high cash outflows, and is subject to an increase in its earnings. As the firm continues to develop and more firms enter the industry, some of the market share is lost to new competitors, which reduces economic profits. In this period the investment opportunities of the company diminish and it grows at a slower rate. This leads to a decline in capital expenditure, which in turn results in accumulation of free cash flows. After this transition stage the firm enters a mature phase where return on investment is equal to the cost of capital and the free cash flows are substantial (Grullon, Michaely and Swaminathan, 2002). The firm is likely to pay the excess cash out by dividends. Furthermore, Grullon, Michaely and Swaminathan (2002) point out that even though in the maturity phase the growth opportunities for a firm decrease, the risk of overinvestment is still high because managers will continue to invest in less profitable projects to increase the resources that they control. Thus an increase in dividends at this stage is desirable as it would indicate a lower probability that the management will waste resources. Therefore the free cash flow hypothesis and agency costs relationship with dividend policy is an intricate part of the life cycle hypothesis (Grullon, Michaely and Swaminathan, 2002). In a later article, Grullon and Michaely (2004) suggest that the lifecycle hypothesis applies for repurchases as well.
9
the large, mature and profitable firms that were found to be most likely to pay dividends, now are substituting the dividend payout for repurchases, which is implicit evidence that the lifecycle theory can apply to repurchases as well.
Size, profitability, and growth opportunities are not always useful to make a clear dichotomy between mature and young firms. DeAngelo, DeAngelo and Stulz (2006) propose that the RETE ratio is a better proxy for firm lifecycle because it measures the extent to which the firm is self-financing or reliant to external capital. Firms with a low ratio tend to be in the capital infusion stage, whereas companies with high ratios are more mature with high profits and substantial amount of internal resources which makes them good candidates to pay dividends (DeAngelo, DeAngelo and Stulz, 2006). Furthermore, they claim that the RETE ratio is a better proxy for maturity than the amount of cash flow available to the company, since less mature firms can also hold high cash balances in given circumstances: for example, a recent equity offering or funds that are to be used for future investments. By, conducting a logit regression analysis, DeAngelo, DeAngelo and Stulz (2006), find empirical proof for the lifecycle hypothesis by observing a positive relationship between the earned/ contributed capital mix and the probability of US firms to pay dividends, controlling for short term profitability, size and investment opportunities. Furthermore, they conclude that the RETE variable is empirically distinct from other variables which relate firm characteristics to the likelihood to pay. Moreover, the earned/ contributed capital mix has a greater impact on the likelihood to pay than profitability and growth opportunities (DeAngelo, DeAngelo and Stulz, 2006).
10
2.4 Agency theory, signalling, and firm size
Some authors offer insights into how agency theory can influence firms with different characteristics. For example, larger firms with more diverse ownership structures are more likely to suffer agency problems (Vogt, 1994). Managers have incentives to cause their firms to grow beyond the optimal size. Growth increases managers' power by increasing the resources under their control (Jensen, 1986). Therefore big companies can reduce their agency costs by paying out the excess cash flow by either repurchasing shares or paying out dividends. In addition larger firms have easier access to the capital markets, which in turn reduces their dependence on internally generated financing and allows for higher payout (Holder, Langrehr and Hexter, 1998). Vogt (1994) analyzes the relationship between cash flow and investment spending and hypothesizes that it can be driven by either the free cash flow hypothesis or the pecking order hypothesis. Thus the importance of free cash flow in the firm’s investment decision might be because firms waste free cash flow or because they face excessive cost of external financing. If firms exhibit free cash flow behavior the relationship between cash and investment will be higher in low Q (overinvesting) firms. If firms finance their operations with primarily internal funds, because of costly external financing, this relationship will be higher for high Q (underinvesting) firms. Vogt’s (1994) findings suggest that larger, low-dividend firms with low Q values, and small, non-dividend firms with high Q values show positive relation between cash flow and investment. Those results are indicative of the fact that larger firms are prone to free cash flow behavior and small firms use internal finds for investment because of costly external financing, but not as a source for unnecessary investments.
11
typically characterized by lower info asymmetry (Mougoue & Rao, 2003). When they split their sample on firms whose dividend policies exhibit patterns, explained by signaling and on ones whose policies cannot, Mougoue and Rao (2003) report that on average signaling firms are twice as small as signaling ones and the difference is statistically significant for non-utility firms. For non-utility firms, again, signaling firms are smaller however the difference is statistically significant only on one out of three variables used to proxy size. Therefore, Mougoue and Rao (2003) find evidence that smaller firms exhibit greater signaling strength than larger ones. Furthermore, Ikenberry, Lakonishok and Vermaelen (1995) report that the market reacts more favourably to repurchase announcements made by smaller firms. The smallest firms in their sample exhibit 8.19% abnormal returns compared to the 2.09% for the largest ones. If firm size is viewed as a proxy for information asymmetries, the observed relationship between size and abnormal returns is consistent with the signalling hypothesis (Ikenberry, Lakonishok and Vermaelen, 1995).
The empirical literature is unanimous to the effect of size to the likelihood to pay. The empirical studies of Fama and French (2001), DeAngelo, DeAngelo and Stulz (2006), Denis and Osobov (2008), and Brockamn and Unlu (2011) report a positive relationship between firm size and the probability to pay dividends for the US, UK, France, Canada, Germany, and Japan. Von Eije and Megginson (2008) find similar results for dividends as well as for repurchases for firms, operating in the European Union, and Coulton and Ruddock’s (2011) results are consistent for repurchases by Australian firms. In general, those studies are in a consensus about what firm characteristics drive payout policy: larger, more profitable firms, with less investment opportunities are found to be most likely to pay dividends or repurchase shares. On the other hand, smaller, less profitable firms with a lot of investment opportunities are least likely to pay funds either as dividends or repurchases.
3. HYPOTHESES AND RESEARCH METHOD 3.1 Hypotheses
12
firms can also have excessive internal funds and consequently high RETE ratios. There are two potential directions of this sensitivity, the first of which can be explained by agency theory and the second with signaling. Therefore, it would be interesting to examine the sensitivity of differently sized firms on the earned/contributed capital mix, which can indicate whether indeed small firms are characterized by more signaling behavior than large firms, or alternatively, bigger firms are put under bigger pressure by investors to pay, because of higher agency costs.
First, in common law countries with good disclosure quality, such as the UK, investors are in a better position to demand the excess cash from managers, in order to minimize agency costs. Larger firm size is related to free cash flow behavior and small size is typically associated with more concentrated ownership and hence lower agency costs. Therefore, investors’ demand for paying out the excess funds of small firms might be less than for large firms. Even though some small companies have large RETE ratios, the benefits of retention might outweigh the costs, since, as the above discussion suggests, smaller firms are typically associated with lower agency costs and more difficulty in attracting financing, and they might choose to build up slack for future investments. Therefore, because the impact of agency problems is less pronounced in small firms, they might be less sensitive to the earned/ contributed capital mix:
Hypothesis 1
H0: The sensitivity to the earned/ contributed capital mix on the probability to pay dividends is equal for small and large firms.
H1: The sensitivity to the earned/ contributed capital mix on the probability to pay dividends is higher for large firms, compared to small firms.
Since the implications of agency theory apply for dividends as well as for repurchases, a similar hypothesis can be made for firms of different sizes that repurchase shares:
Hypothesis 2
H0: The sensitivity to the earned/ contributed capital mix on the probability to repurchase shares is equal for small and large firms.
H1: The sensitivity to the earned/ contributed capital mix on the probability to repurchase shares is higher for large firms, compared to small firms.
13
can be used to remove excess cash from the hands of managers, they are not as strong commitment as dividends according to Jagannathan, Stephens and Weisbach (2000). So if agency costs prevail, dividends might be more suitable to minimize them.
Second, and alternatively, signaling theory proposes that firms try to overcome information asymmetry via dividend payout or share repurchases. Small firms typically suffer more from information asymmetry than large firms and thus have more difficulty in attracting outside financing. This implies that they will resort to dividends (or repurchases) as means of signaling more than large firms. Even though the effect of RETE on the probability to pay, as suggested by lifecycle theory, will still hold for small as well as large firms, large firms with high RETE ratios won’t have the need to use dividends to convey information to the market as extensively as smaller firms, because as DeAngelo, DeAngelo and Stulz (2006) suggest they will be the most mature firms in the least need to establish reputations. Small firms with high RETE ratios will be the best candidates to pay, because they need to signal, and the higher retained earnings will make this possible. Thus, it is expected that smaller companies are more sensitive to the RETE ratio than large firms because the effect of the earned/ contributed capital mix on the probability to pay dividends (or repurchase shares) is amplified by bigger information asymmetry for small companies and their desire to signal positive news to the market.
Hypothesis 3
H0: The sensitivity to the earned/ contributed capital mix on the probability to pay dividends is equal for small and large firms.
H1: The sensitivity to the earned/ contributed capital mix on the probability to pay dividends is higher for small firms, compared to large firms.
Signaling models apply not only to dividends but also to repurchases and since smaller firms are characterized by higher degree of information asymmetry than large firms, the last hypothesis is:
Hypothesis 4
H0: The sensitivity to the earned/ contributed capital mix on the probability to repurchase shares is equal for small and large firms.
14
Again, for hypothesis 4, it is possible that there is no difference in the effect of RETE for small and large firms (so that H0 is not rejected), because although, repurchases can be used for signaling, they are more flexible than dividends and do not show as strong commitment for future payouts as dividends.
3.2 Research method
To test the hypotheses, first a univariate analysis is performed by examining sample statistics. Then, consistent with Fama and French (2001), a multivariate logit model is utilized for the regression analysis. This model overcomes the problems of the linear probability model by using a function that transforms the regression model so that the fitted values are bounded within the (0, 1) interval (Brooks, 2008). The logit procedure used, takes the payment/ non-payment of dividends or the repurchase/ non-repurchase of shares of firm i in each year as dependent variables and RETE, and measures for size, growth, profitability and also cash holdings as independent variables. To avoid endogeneity problems, the explanatory variables in all specifications of the model are lagged one year. The logit model takes the form:
( )
where zi equals:
( )
15
estimates ( ̂ ) from each year. The significance of the mean coefficient estimates for the entire period is determined by using the following t-statistic:
( ̂̅ ) ̂̅ ( ̂) √
where ̂̅ is the mean coefficient estimate for the respective variable j over the whole period, ( ̂) is the standard deviation of the same coefficient estimates during the period, and n is the number of years in the period. The coefficient of interest for this research is ̂̅ and it will determine the validity of the hypotheses. After the methodology has been described, the testable implications can be more explicitly put. A summary can be seen in Table 1.
Table 1. Summary of testable hypotheses. Null and alternative hypotheses for each testable implication are presented. The hypotheses are separated according to the theory used to make them: agency theory (2nd and 3rd column) and signalling theory (4th and 5th column). Furthermore, per theory, the hypotheses are formulated for both dividends and repurchases.
Agency Theory Signalling Theory
Payout type Dividends Repurchases Dividends Repurchases
Hypothesis Hypothesis 1 Hypothesis 2 Hypothesis 3 Hypothesis 4
Null hypothesis H
0: ̂̅ = 0 H0: ̂̅ = 0 H0: ̂̅ = 0 H0: ̂̅ = 0 Alternative hypothesis H
1: ̂̅ > 0 H1: ̂̅ > 0 H1: ̂̅ < 0 H1: ̂̅ < 0
4. DATA 4.1 Sample selection and variables
The data sample used in this paper contains all UK companies, listed on the London Stock Exchange (LSE), for the period 1989-2012. The starting period is chosen to be 1989, because prior this year the Datastream (Worldscope) database, which is used to retrieve all the data for the empirical research, has a limited coverage of accounting company data (Denis and Osobov, 2008; Von Eije and Megginson, 2008). In addition, repurchases were not a factor until the mid-1980s (Allen and Michaely, 2003), which means that 1989 is a good starting point if one wants to analyze repurchases. Year 2012 is the most recent period for which there is accounting firm information available.
16
by deleting companies operating in the utility and financial sectors. The dividend decisions of those companies are a subject to regulation, and including them in the sample might give biased results (Fama and French, 2001). In order to trim the data for utilities and financials, the primary SIC codes for all firms were downloaded, and companies having a SIC code between the ranges 4900-4949 (utilities) and 6000-6999 (financials) were deleted. The final sample contains 3225 companies, for which for the first sample year 981 companies have available information in Datastream (Worldscope) on cash dividends, of which 912 are payers. During the years, the Worldscope coverage gradually increases and for 2005 the companies with available data on total cash dividends are 1770, from which 740 are dividend payers. Through the crisis, the number of firms with available data on dividend payments decreases, probably due to suspended companies2, and in 2012 the number of companies that have cash dividend data is 1137, of which 517 pay dividends. The observations on repurchases follow a similar pattern and for 1989 there are 973 companies with repurchase data. In this year, however, only 77 buy back their shares. This number grows gradually until 2005 and during the crisis decreases. For 2012 there are 1102 companies with repurchase data available of which 230 repurchase their shares.
For the regression analysis only companies with available data on cash dividends paid, repurchases, market capitalization, total assets, sales, net income, cash, and retained earnings are used. In addition, a company must have a positive book value of common equity in order to be included in the regression sample. Furthermore, data form 1989 and 1990 is used only for construction of lags. This yields 21,813 firm-year observations for dividends and 21,654 firm-year observations for repurchases. All variables are winsorized at 1% to mitigate the impact of outliers. This means that all values for a given variable, in a given year, above the 99th percentile or below the 1st percentile, are made equal to the 99th percentile and the 1st percentile, respectively. A summary of the variables used for the regression analysis can be found in Table 2.
Table 2. Description of the variables collected for UK companies and used in a multivariate logit regression analysis. The first column contains the name of the variable and the second column contains an explanation for the calculation and the source for the variables.
Variable Description and Source
DIV Dummy variable: equals one if a firm pays dividends in a given year and zero otherwise. Dividends are defined as the total common and preferred dividends paid to shareholders of a company. Source: WorldScope Database.
17
Table 2 Continued...
REP Dummy variable: equals one if a firm repurchases shares in a given year and zero otherwise. Repurchases are measured as funds used to decrease the outstanding shares of common and/or preferred stock. Source: WorldScope Database.
RETE Retained earnings as a percentage of book value of total common equity. The variable is calculated only for companies with positive book value of common equity, in order to be economically meaningful. Retained earnings represent accumulated after tax earnings of the company which have not been distributed as dividends to shareholders or allocated to a reserve account. Total common equity represents common shareholders’ investment in a company. Source: WorldScope Database.
ROA Return on assets: (net income + interest expense on debt (if available))/ book value of total assets. Net income represents income before extraordinary items and preferred and common dividends, but after operating and non-operating income and expense, reserves, income taxes, minority interest and equity in earnings. Interest expense on debt represents the service charge for the use of capital before the reduction for interest capitalized. Total assets represent the sum of total current assets, long term receivables, investment in unconsolidated subsidiaries, other investments, net property plant and equipment and other assets. Source: WorldScope Database.
SGR Sales growth rate: (sales in current year - sales in year t-1)/ sales in year t-1. Sales represent gross sales and other operating revenue less discounts, returns and allowances. Source: WorldScope Database.
SIZE Size of a firm, measured by the natural logarithm* of its market capitalization. Market capitalization represents market price year-end, multiplied by common shares outstanding. Source: WorldScope Database.
SIZED Dummy variable: equals one if a firm’s market capitalization is bigger than the median market capitalization for the year and zero otherwise. Market capitalization represents market price year-end, multiplied by common shares outstanding. Source: WorldScope Database.
CASHTA Cash as a percentage of the book value of total assets. Cash represents money available for use in the normal operations of the company. Total assets represent the sum of total current assets, long term receivables, investment in unconsolidated subsidiaries, other investments, net property plant and equipment and other assets. Source: WorldScope Database.
SIZExRETE Interaction variable that is used to measure the sensitivity of firms of different sizes to RETE. It is constructed by multiplying SIZE by RETE**.
*Size is used in log form to account for the wide dispersion of total assets and market capitalization over firms. ** The variable is constructed once by using SIZE and once by using SIZED as a measure of size.
4.2 Payout patterns for small and large UK firms
18
Furthermore, it can be seen in Figure 1 that both type of firms: small (dashed line) and large (dotted line) are driving the decrease in the proportion of dividend payers. However, the proportion of small firms that are paying dividends is declining much more steadily than the proportion of large firms: the proportion of large payers decreases by 30%, from 97% in 1989 to 67% in 2012, whereas the proportion of small firms goes down by 63.5% (more than twice the decrease for dividends).
Figure 1. Percentage of dividend payers and companies that repurchase their shares by year. The bars represent the proportion of dividend payers (solid fill) or firms that repurchase shares (grid fill) for each sample year, calculated from the total sample. The lines represent the proportion of payers for each year from two sub-samples: small firms and large firms. The proportion of small dividend payers for every year is represented by the dashed line and the proportion of large dividend payers by the dotted line. The proportions of small firms that repurchase shares are indicated by the solid line and the percentages of large firms that repurchase are represented by the dash-and-dots line. The sample includes non-utility and non-financial companies, listed on the LSE, with a primary SIC code out of the ranges 4900-4949 and 6000-6999. The sample is split on the median market capitalization each year.
This is consistent with empirical literature that small firms are less willing to pay dividends, one explanation for which is bigger asymmetric information and difficulty to attract financing. On the other hand it can be observed that the increase in repurchases during the period 1989-2012 comes mainly from large firms (dash-and-dots line) as the proportion of small repurchasing firms (solid line) stays fairly constant. A possible explanation for this is higher agency costs for larger firms, which are mitigated by paying out resources that otherwise might be wasted. The pattern of the decrease of the proportion of large dividend payers is mirroring the pattern of increase in the proportion of large firms that repurchase
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 19 89 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Dividends Repurchases Small (div.)
19
shares. This is an indication, although, only descriptive in nature, that large UK firms are substituting dividends for repurchases as a means of payout, similar to US companies as reported by Skinner (2008).
Despite the fact that the number of firms paying dividends decreases, Figure 2 shows that the aggregate amount of dividends paid is increasing throughout the sample period. While in 1989, the nominal amount paid out as dividends is 13.5 billion British Pounds, it grows to 60.6 billion in 2012. The pattern of the nominal amount for repurchases is also interesting. Prior research reports an increase in the aggregate amount of funds used for repurchases until the start of the ongoing crisis and no information is available afterwards. As it can be seen in Figure 2, the nominal aggregate amount of repurchases in UK increases steadily until 2006 when it is 85% of the nominal amount of cash dividends paid. After this period, however, a plummet in the aggregate amount of repurchases is observed, which confirms the findings of Skinner (2008) and Jagannathan, Stephens and Weisbach (2000), that repurchases are much more sensitive to earnings variability than dividends. The lines show the nominal amount of dividends (solid line) or nominal amount of repurchases (dashed line) that large firms are paying out. It can be seen that almost all of the payout (more than 85%), no matter in the form of dividends or repurchases, comes from big companies, and that is the reason the graph does not report the payouts of small firms separately. This is also consistent with Fama and French (2001) and Skinner (2008) who find that dividend payouts are concentrated in large firms.
Figure 2. Nominal, aggregate amount of cash dividends paid (solid bar) and nominal, aggregate amount of repurchases (grid bar) for UK listed companies, on the LSE, with a primary SIC code outside the ranges 4900-4949 and 6000-6999. The solid line represents nominal, aggregate amount of cash dividends paid by large firms and the dashed line represents nominal, aggregate amount of repurchases by large firms. The sample is split on the median market capitalization each year. The values are in British Pounds.
0 10 20 30 40 50 60 70 1 9 8 9 1 9 9 0 1 9 9 1 1 9 9 2 1 9 9 3 1 9 9 4 1 9 9 5 1 9 9 6 1 9 9 7 1 9 9 8 1 9 9 9 2 0 0 0 2 0 0 1 2 0 0 2 2 0 0 3 2 0 0 4 2 0 0 5 2 0 0 6 2 0 0 7 2 0 0 8 2 0 0 9 2 0 1 0 2 0 1 1 2 0 1 2 B illi on B rit is h P oun ds ( £)
20
5. RESULTS 5.1 Descriptive statistics and univariate analysis
Table 3 presents summary statistics of the explanatory variables used in the multivariate logit regressions in the next sub-section of this paper. Firm size (SIZE) is measured by the natural logarithm of market capitalization. Sales growth (SGR) is used as a measure of growth and ROA measures profitability. Cash as a percentage of total assets (CASHTA) is also used to analyze whether there is a relationship between the cash amount a firm holds and the decision to pay. The RETE3 variable represents the earned/ contributed capital mix, which is a proxy for a firm’s lifecycle. The numbers presented are calculated for payers and non-payers once for the total sample and then for two sub-samples containing only either small or large firms. Each year the firms are split by the median natural logarithm of market capitalization and the ones above the median are considered large and the ones below small. For each year the median value of each variable is calculated for the respective (sub)sample and then the median of the medians for the 24 years is taken to calculate the values which are presented in the table. The median number of firms in each (sub)sample is calculated in a similar manner.
Panel A of Table 3 shows the descriptive statistics for dividends. Overall, according to the results for the total sample, the median dividend paying firm holds more earned equity (0.483) than the median non-paying firm (-0.369). The difference between the median RETE ratio for the two groups is 0.852. This result holds also for both sub-samples of small and large firms. However, the difference between the median payer and non-payer is much larger for smaller firms. Consistent with Fama and French (2001) and subsequent empirical research, more profitable firms are better candidates to pay dividends as the ROA of payers is higher than the ROA of non-payers by 10.5%. Again this difference is more pronounced for small firms. The relationship between growth and payers and non-payers is as expected – the median values of SGR indicate that paying firms grow less than non-paying firms. This result is driven mainly by large firms since the difference between the two groups is much more pronounced for bigger firms. A possible explanation is the fact that small firms do not have easy access to capital, because of bigger information asymmetry, and growing is thus more difficult than for larger firms. Furthermore, on median, payers are larger than non-payers and the median dividend payer has less cash than the dividend non-payer. For dividend payers the cash as a percentage of total assets is 5% and for non-payers it is 10.6%. The figures are
3
21
similar for small, as well as for large companies. This last result is an indication that cash is not a good proxy for the lifecycle of a UK firm, as proposed by DeAngelo, DeAngelo and Stulz (2006), and firms that hold a lot of cash are not necessarily good candidates to pay dividends.
Table 3. Descriptive statistics for UK industrial firms with SIC codes outside the ranges 4900-4949 and 6000-6999, listed on the LSE for the period 1989-2012. For each year the median of each variable for payers and non-payers is calculated for the total sample. In addition the sample is split into two groups containing either small or large firms and the medians are calculated for both subsamples. Then the median of the medians for the 24-year period is taken to calculate the values which are presented in the table. The calculations are made for dividends (Panel A) and are also repeated for repurchases (Panel B). For the split of the sample, in each year, a company is considered as small if the natural logarithm of its market capitalization is below the median and as large if it is above. RETE is retained earnings to total book value of equity; ROA is return on assets; SGR is sales growth rate; SIZE is the natural logarithm of market capitalization; and CASHTA is total cash divided by the book value of total assets. All descriptive statistics are calculated only for those firms that have positive book value of common equity and all variables are winsorized at 1%.
Panel A: Dividends
Total sample SIZE SGR ROA CASHTA RETE # Firms
Payers 11.224 0.082 0.072 0.050 0.483 827
Non-payers 9.713 0.103 -0.033 0.106 -0.369 620
Small firms SIZE SGR ROA CASHTA RETE # Firms
Payers 9.621 0.057 0.056 0.046 0.432 284
Non-payers 9.147 0.078 -0.064 0.107 -0.594 374
Large firms SIZE SGR ROA CASHTA RETE # Firms
Payers 12.270 0.095 0.079 0.056 0.498 485
Non-payers 11.516 0.254 -0.008 0.111 -0.117 142
Panel B: Repurchases
Total sample SIZE SGR ROA CASHTA RETE # Firms
Payers 11.462 0.052 0.073 0.060 0.483 148
Non-payers 10.340 0.089 0.029 0.065 0.136 1157
Small firms SIZE SGR ROA CASHTA RETE # Firms
Payers 9.569 0.021 0.050 0.085 0.366 45
Non-payers 9.302 0.064 -0.008 0.081 -0.114 515
Large firms SIZE SGR ROA CASHTA RETE # Firms
Payers 12.573 0.068 0.077 0.063 0.495 87
Non-payers 11.941 0.110 0.058 0.062 0.381 516
22
that do not is larger for smaller firms. According to the descriptive statistics growth is a determinant of whether a firm repurchases or not, with payers exhibiting less growth than non-payers, no matter small or large. Similarly to dividends, small firms as a whole have much lower growth rates than large firms. Furthermore, the median firm that buys back shares is larger than the typical non-repurchasing firms. Last, there is no substantial difference between the CASHTA ratios of paying and non-paying firms, so conclusions on whether or not cash influences the decision to repurchase cannot be drawn from the descriptives.
By simply comparing the descriptive statistics, strong conclusions about the sensitivity of differently sized firms on RETE cannot be made, though it can be noted that the difference between the RETE ratios for payers and non-payers is considerably larger for small firms than for large firms, for both dividends and repurchases. This is in line with H1 for hypotheses 3 and 4 that smaller and younger firms have difficulty in attracting external financing because of asymmetric information and can afford to signal that they have good future opportunities only when they gather substantial amount of internal capital which will ensure that they can support future payments. On median, however, the RETE ratio is smaller for small payers in comparison to large payers.
Table 4 shows a relationship between the earned/ contributed capital mix and the proportion of firms that pay dividends (Panel A) or repurchase shares (Panel B). For every year the sample of firms is split in small and large companies and then companies from each subsample with available payout information are ranked on RETE4. Then the firms are put into 10 equally sized groups and for each group the percent of payers in the group is calculated. Then the averages of the percent of payers across the 24 years are taken for each decile to calculate the numbers presented in the table. The number of firms is the average of the size of each decile over the 24 year period.
Panel A of Table 4 shows that the percent of dividend payers grows monotonically from decile 0 to 10, for small and large firms. Overall the difference between the percent of payers from decile 0 and decile 10 for small firms is 0.685 and for large firms is 0.483. These results show a more pronounced increase in the proportion of payers for small firms than for large companies, with increase in RETE, indicating that smaller firms with high RETE ratios are more willing to pay than large firms with high earned/ contributed capital mix. This is also shown by the fact that after decile 5 the proportion of small firms that pay dividends increases by more than the proportion of large firms that pay. Although this evidence is only
23
descriptive, it indicates that smaller firms are more sensitive to RETE, which can be explained by asymmetric information and difficulty in attracting financing. The fact that for large firms the proportions are still increasing in RETE but the increase is more moderate fits with the agency/ lifecycle theory explanation of company payout policy.
Table 4. Proportion of dividend payers (Panel A) and firms that repurchase shares (Panel B) for 10 RETE (retained earnings to total assets) deciles. The sample consists of UK firms with primary SIC codes out of the ranges 4900-4949 and 6000-6999, listed on the LSE for the period 1989-2012. The sample is split into two groups containing either small or large firms, by using the median natural logarithm of the market capitalization in each year. The RETE ranked deciles contain only firms with positive book value of common equity. For each year, and in each small/large subsample, companies with available cash dividend data (Panel A), repurchases data (Panel B), book equity data and retained earnings data are sorted on their RETE ratios. Then the companies are put into 10 equally sized deciles. The proportion of dividend payers (Panel A) or proportion of firms that repurchase shares (Panel B) are calculated for each decile and the average proportion of payers for each decile across 24 years is calculated. The resulting numbers are reported in the table. The number of firms is the average number of firms for each decile for 24 years.
Panel A: Dividends
RETE Deciles: Small firms 1 2 3 4 5 6 7 8 9 10
Proportion of payers 0.168 0.235 0.314 0.388 0.441 0.522 0.637 0.779 0.810 0.853
Number of firms 61 61 61 61 61 61 61 61 61 59
RETE Deciles: Large firms 1 2 3 4 5 6 7 8 9 10
Proportion of payers 0.454 0.509 0.641 0.804 0.879 0.907 0.921 0.942 0.944 0.937
Number of firms 61 61 61 61 61 61 61 61 61 60
Panel B: Repurchases
RETE Deciles: Small firms 1 2 3 4 5 6 7 8 9 10
Proportion of payers 0.032 0.035 0.045 0.053 0.038 0.067 0.072 0.106 0.121 0.150
Number of firms 57 57 57 56 56 56 58 59 60 57
RETE Deciles: Large firms 1 2 3 4 5 6 7 8 9 10
Proportion of payers 0.113 0.103 0.102 0.130 0.177 0.182 0.201 0.204 0.210 0.241
Number of firms 60 59 59 60 60 60 60 61 61 59
Panel B of Table 4 presents results from the same analysis for repurchases. Although the proportion of repurchasing firms is growing with every other decile, the proportion of firms that repurchase shares are not as high as the proportion of dividend payers. The differences between decile 1 and 10 are similar for both subsamples and a conclusion about the sensitivity of differently sized firms on RETE and their decision to buy back shares cannot be drawn from this analysis. At least there is evidence that the decision of UK firms to repurchase shares is driven by RETE and this result is new to the literature. More formal evidence for this can be found in section 5.2 and in the Appendix.
5.2 Regression analysis
Dividends. Table 5 presents the results from a logit regression analysis for dividends. For each of the 22 years of the sample period5 a logit model is run with a dependent variable
5
24
being a dummy that takes the value of one if a firms is a payer and zero otherwise. The independent variables are proxies for size (SIZE), growth (SGR), profitability (ROA), earned/ contributed capital mix (RETE), cash holdings (CASHTA) and an interaction term (SIZExRETE) that indicates how the sensitivity on RETE changes with size. All variables are lagged by one year to avoid endogeneity problems. The t-statistics for the explanatory variables are calculated form the fitted coefficients, obtained from the 22 logit models. The coefficient estimates are on the top row and the t-statistics are in brackets, immediately below. The table presents the results of two models.
Table 5. Logit analysis of the decision to pay dividends as a function of firms size (SIZE), growth (SGR), profitability (ROA), earned/ contributed capital mix (RETE), cash holdings (CAHTA) and an interaction term (SIZExRETE) which measures how the sensitivity of firms on RETE changes with size. The dependent variable (DIV) is a dummy equal to one if a firm pays dividends in a given year and zero otherwise. All independent variables are lagged by one year. Model 1 uses the natural logarithm of market capitalization as a measure of size (SIZE) and model 2 uses a dummy (SIZED) which is equal to one if the market capitalization of a firm is bigger than the median market capitalization for a given year and zero otherwise. Only UK firms, listed on the LSE with a primary SIC code out of the ranges 4900-4949 and 6000-6999, available data on all variables, and positive book value of common equity, are included in the analysis. The t-statistics are calculated by dividing the average coefficients for 22 periods by their standard deviation and multiplying this output by the square root of the number of periods. The average of the McFadden R-Squared is presented on the right.
Model 1 SIZE ( ̂̅ ) SGR ( ̂̅ ) ROA ( ̂̅ ) CASHTA ( ̂̅ ) RETE ( ̂̅ ) SIZExRETE ( ̂̅ ) Int. ( ̂̅ ) ̅2
coefficient 0.448 -0.481 10.734 -1.669 2.731 -0.189 -3.525 42.3%
t-statistic (18.472)*** -(7.852)*** (10.483)*** -(5.940)*** (5.800)*** -(4.882)*** -(10.573)*** Model 2 SIZED ( ̂̅ ) SGR ( ̂̅ ) ROA ( ̂̅ ) CASHTA ( ̂̅ ) RETE ( ̂̅ ) SIZEDxRETE ( ̂̅ ) Int. ( ̂̅ ) ̅2
coefficient 1.244 -0.472 11.853 -1.815 1.107 -0.722 0.542 40.2%
t-statistic (15.510)*** -(6.652)*** (11.060)*** -(5.900)*** (6.811)*** -(3.967)*** (2.870)***
***Significant at 1% (two-sided)
25
The main result for this paper is the fact that the interaction coefficient ( ̂̅ ) is negative (-0.189) and highly significant (at 1%). This leads to the rejection of the null and alternative hypotheses of hypothesis 1, as the coefficient estimate is negative. Also, H0 of hypothesis 3 can be rejected. The alternative hypothesis of hypothesis 3 is not rejected and therefore, there is statistically significant empirical evidence that small firms are more sensitive to the RETE ratio than large firms. This result can be explained by the fact that the effect of the earned/ contributed capital mix on the dividend payout of small firms is amplified by the desire to signal positive news to the market.
Model 2 uses a dummy variable (SIZED) which is equal to one if a firm is large and zero otherwise. A company is defined as large if its market capitalization for a given year is above the median. The rest is the same as in model 1. The results stay similar: larger firms are more willing to pay than small ones, growing companies are less likely to distribute cash to shareholders in the form of dividends, and more profitable firms are better candidates to pay. In addition the earned/ contributed capital mix remains a highly significant driver of the decision to pay. The fact that the coefficient estimate on CASHTA is negative for both specifications of the model is consistent with DeAngelo, DeAngelo and Stulz’s (2006) results that cash holdings are not a good proxy to measure the lifecycle of a company, because high cash holdings might indicate either excess funds6 or a slack that is going to be used for planned investments7. Most importantly, the interaction coefficient of the SIZEDxRETE variable is -0.722 and is highly significant. This means that small firms with high RETE ratio are more likely to pay than large firms with the same amount of earned capital as a fraction of total equity, which again gives support to the alternative hypothesis of hypothesis 3.
Repurchases. Table 6 presents the results from a logit regression analysis for repurchases. In this case the model takes the repurchasing/ non-repurchasing of shares as a dependent variable. All independent variables are as in table 5. The model is run for 22 periods and the fitted coefficients for each period are used to calculate the t-statistics which determine the significance of the variables.
Model 1 uses the natural logarithm of market capitalization as a measure of size. It can be seen that, as for dividends, the probability of a company to repurchase its shares increases in size. Furthermore, growth is negatively associated with the probability to
6 Then positive relationship is expected since dividends will be used to decrease the funds at discretion of managers.
7
26
repurchase shares as the coefficient on SIZE is negative and significant. In addition, the significantly positive coefficient on ROA is indicative for the fact that the more profitable a firm is, the more likely it is to repurchase its shares. All those results are consistent with Von Eije and Megginson (2008), who find similar results for their sample of firms operating in the European Union. In addition, in contrast to dividends, cash holdings have a positive relationship with the probability to repurchase. A new finding for the literature is the fact that RETE is a significant driver of the probability to buy back shares for UK firms, as the coefficient is positive and significant, after controlling for size, growth, profitability, and cash holdings. It should be noted that the effect is much smaller than for dividends, since the coefficient drops substantially (0.414 versus 2.731 for dividends), and it is significant only at the 10% level. A more extensive analysis on the impact of the RETE ratio on the probability to repurchase shares can be found in the Appendix.
Table 6 Logit analysis of the decision to repurchase shares as a function of firms size (SIZE), growth (SGR), profitability (ROA), earned/ contributed capital mix (RETE), cash holdings (CAHTA) and an interaction term (SIZExRETE) which measures how the sensitivity of firms on RETE changes with size. The dependent variable (REP) takes the value of one if a firm repurchases shares in a given year and is zero otherwise. The independent variables are lagged one year. Model 1 uses the natural logarithm of market capitalization as a measure of size (SIZE) and model 2 uses a dummy (SIZED) which is equal to one if the market capitalization of a firm is bigger than the median market capitalization for a given year and zero otherwise. Only UK firms, listed on the LSE with a primary SIC code out of the ranges 4900-4949 and 6000-6999, and available data on all variables are included in the analysis. Firms with negative values of book common equity are excluded. The t-statistics are computed by averaging the coefficient for a given variable for the whole sample period, dividing it by its standard deviation and multiplying the result by the square root of the number of periods. The average of the McFadden R-Squared is presented on the right.
Model 1 SIZE ( ̂̅ ) SGR ( ̂̅ ) ROA ( ̂̅ ) CASHTA ( ̂̅ ) RETE ( ̂̅ ) SIZExRETE ( ̂̅ ) Int. ( ̂̅ ) ̅2
coefficient 0.215 -0.539 2.010 0.607 0.414 -0.029 -4.602 10.2%
t-statistic (5.649)*** -(4.525)*** (4.684)*** (3.208)*** (1.892)* -(1.545) -(15.229)*** Model 2 SIZED ( ̂̅ ) SGR ( ̂̅ ) ROA ( ̂̅ ) CASHTA ( ̂̅ ) RETE ( ̂̅ ) SIZEDxRETE ( ̂̅ ) Int. ( ̂̅ ) ̅2
coefficient 0.501 -0.558 2.628 0.248 0.163 -0.013 -2.451 7.8%
t-statistic (3.526)*** -(4.606)*** (5.399)*** (1.178) (2.931)*** -(0.140) -(23.875)***
***Significant at 1%; *Significant at 10% (two-sided)
The interaction coefficient ( ̂̅ ), presented in Table 6, is negative (-0.029). This would suggest that, as for dividends, the sensitivity of firms to the impact of the RETE ratio, on the probability that they will pay dividends, decreases with size. The coefficient estimate is, however, insignificant, leading to the non-rejection of H0 and rejection of H1 for hypotheses 2 and 4. Therefore, although the results are economically meaningful they are not statistically robust and the only conclusion which can be made is that the effect of the earned/ contributed capital mix on the probability of a company to repurchase its shares does not depend on its size.
27
size, growth and profitability is as expected and consistent with prior empirical research. The impact of lifecycle remains significant and the probability of companies to repurchase shares increases with RETE. Cash holdings are insignificant driver of the probability of firms to buy back stocks, according to this specification of the model, as the coefficient is insignificant. The interaction coefficient ( ̂̅ ) is again negative and insignificant. This means that (as for model 1) there is not enough evidence to reject H0 for hypotheses 2 and 4, and none of the alternative hypotheses are confirmed. In general, the regression results for repurchases show that small firms do not have a different sensitivity to the earned/ contributed capital mix, as a determinant of their decision to repurchase shares, compared to large companies.
To sum up, the regression results for both specifications of the model for dividends show negative and significant coefficient estimates on the interaction variables. This indicates that the sensitivity on RETE to pay dividends decreases with size. This finding can be explained with signaling theory. Since smaller firms suffer more information asymmetry and have difficulty in raising outside capital, they want to signal to investors that they have promising future investment opportunities. Because of the fact that they are small, they will typically have lower amount of earned equity. The ones with higher amounts, however, will be willing to payout some of the funds to be able to signal the market at a cost that they are good and thus continue to grow. So a small firm with high RETE will be more willing to pay dividends than a large firm because the impact of RETE is amplified by the large information asymmetry and the desire to signal. Overall, this suggests that small companies use dividends more to convey information to the market, than large companies. This result does not reject the fact that agency theory explains payout policy. In fact the positive and significant effect of RETE for large firms is an indication about agency costs explanation to dividends because agency theory is an integrated part of the lifecycle hypothesis (Grullon, Michaely and Swaminathan, 2002). The results simply state that small firms are more sensitive to RETE than large firms because they signal more. The fact that the earned/ contributed capital mix is a determinant to pay for large firms gives empirical credibility to agency theory and its impact on payout policy and confirms DeAngelo, DeAngelo and Stulz’s (2006) suggestion that larger, more mature firms do not need to signal that much.
28
shares, is different. A possible explanation is the fact that dividends are more binding than repurchases and thus small companies choose to signal via dividends as they show stronger commitment. Another possible explanation for the insignificant results for repurchases is the fact that the effect of the RETE ratio, on the probability to repurchase shares, is less strong compared to dividends, although it is still present.
6. CONCLUSION
This paper analyzes whether there is a difference between the sensitivity to the earned/ contributed capital mix as a determinant of payout policy in small and large UK firms. Two main hypotheses are developed for dividends and two for repurchases. First, large firms are typically characterized with more dispersed ownership and bigger agency problems. A high RETE ratio means that firms are profitable and have probably amassed a large amount of slack. Therefore the possibility of managers to use the excess funds for their own benefits increases. In this case payment of dividends or repurchase of shares is necessary to mitigate overinvestment and therefore the sensitivity of large firms on the RETE is expected to be higher than for small firms. Second, signaling is a costly way for a company to overcome asymmetric information problems and get easier access to the market. Small firms are characterized by more information asymmetry than large firms. Thus a profitable small firm, that have high RETE ratio can afford to signal in order to get financing for its future projects and thus grow. Therefore, in this case, small firms are expected to be more sensitive to the RETE ratio as a determinant of the probability to pay dividends or repurchase shares.
