Stock repurchases and loyalty
A study into the effect of the presence of blockholders on the payout mix in US companies in 2003-2012Julia van Haren
10524398
Economie en Bedrijfskunde
Finance and Organization
Dr. I.J. Naaborg
June 2016
Abstract: Research in the 21st century suggests that dividends are a diminishing form of payout and stock repurchases are now more dominant. However, the financial meltdown of 2008 made the shortcoming of repurchases as opposed to dividends clear. The stock repurchases made before the crisis benefited the selling non-loyal stockholders at the expense of loyal stockholders. Loyalty of stockholders can be measured by the presence of blockholders in a company. The overall effect of stockholder concentration on the payout mix has not yet been examined in earlier research. Results based on US companies (2003-2012) show that companies with a bigger concentration of stockholders tend to pay relatively more dividend.1. Introduction In the 21st century, the decline of dividends and the growth of stock repurchases have been investigated intensively. However there has not yet been any research on the effect of loyalty, measured by the concentration of blockholders, after the crisis. Research has suggested that stock repurchases will continue to grow. However, after the crisis this growth could be tempered or even reversed. Thus far, this has not been examined and it is therefore interesting to contribute to these studies with research about the effect that both the crisis and loyal stockholders have on the payout mix. First of all, Fama and French (2000) conclude that there were two reasons for disappearing dividends; the increasing number of companies that had the characteristics of companies that never paid dividends and companies became less likely to pay dividends given that they had the characteristics of companies who normally did pay dividends. Second of all, Brav et al (2005) argue that one of the main reasons of the increase in stock repurchases had been the flexibility of this kind of payout. After the crisis however, the disadvantages of stock repurchases also became clear. During the pre-crisis stock market bubble, a lot of companies repurchased stock at relatively high prices compared to the prices after the collapse of the market. Therefore, looking back, it can be argued that these repurchases happened at the expense of the loyal stockholders (DeAngelo et al. 2008). This disadvantage is now widely known and it is therefore relevant to investigate whether it has affected the growth of stock repurchases. Did managers pay more attention to this shortcoming or did the growth of stock repurchases continued? Overall managers do not like to change their dividend policies (Brav et al. 2005). They surely do not want to lower dividends because this can substantially lower the stock price and the prospect of future equity. In addition, stockholders can also be against the decline in dividends and will try to replace managers who were responsible for the decline. This will cause managers to be reluctant to increase dividends as well. Dividends are therefore commonly used only to distribute permanent streams of free cash flows. In contrast, stock repurchases are perfect to distribute transitory positive cash flow shocks. Since flexibility and timing of payouts will always be an important
factor in corporate payout policies, companies will want to maintain the flexible element in their policy. However, as suggested by DeAngelo et al (2008), the disadvantage of stock repurchases could blunt the growth of this form of payout. Managers might want to take the interests of loyal stockholders into account. Even though loyalty is a difficult variable to measure, it can be argued that blockholders have many features of this loyal stockholder. Blockholders are stockholders who own at least 5% of the total stock of a company. Blockholders often desire to earn profits at the expense of liquidity investors, which are investors who can leave a company easily and trade their stock of a company more regularly. These investors are thus less loyal to a company than blockholders are. Since they are loyal to a company, blockholders have the incentive to gather costly information. Because of this gathering of information, blockholders cause the stock prices to reflect the fundamental value of a company instead of the current earnings of the company. This leads to managers investing in the long-term growth of their company rather than the short-term profits the company can make. The blockholder, however, does not only encourage the manager to invest in the long-term by being a long-term investor who never sells. The possibility that the blockholder sells in the short-run will encourage the manager to think long-term. Therefore, a blockholder will add value to a company, with good long-term prospects but weak short-term earnings, by remaining a stockholder. Whereas other stockholders would have sold their stock, the blockholder thinks in the long run and will keep his stock, as he knows the fundamental value of a company. Skinner (2008) performed an OLS regression of the payout mix on hypothesized determinants. So, to examine whether loyal stockholders do temper the growth of stock repurchases after the crisis, this regression can be used as a building block. Skinner uses six control variables to capture the variation in the level of payout and three control variables, which should explain the incentives to repurchase stock. In particular, the first six control variables account for the size, profitability and growth of a company. Skinner examines these cross-sectional determinants and shows the significance of these variable per year from 1980 until 2005. This regression is also performed, but a few additions are made. First, an index variable for loyalty will be added (Herfindahl-Herschman index of the three biggest stockholders of a company). Second, a dummy variable for crisis (0 in the years 2003-2007 and, 1 in the years 2008-2012) will be
added to this regression. Third, since the crisis variable is included, data is collected for many years for many companies. Thus panel data is used in this study instead of cross-sectional data. For this analysis the data on the control variables is gathered from Compustat. Unfortunately data on blockholders is less easily found and therefore has to be manually gathered from Thomson ONE. A random selection is made in Excel to choose 200 companies of which all the data necessary is collected. The regression is thus performed on these 200 companies within 2003-2012. The amount of years companies either repurchase stock or pay dividends to their stockholders differs. Thus for some companies more data is collected than for other companies, which accumulates to exactly 1421 observations. This research will contribute to the existing research on the developments of the payout mix of companies. Until now the research on the effect of the presence of blockholders has been limited. It is relevant to contribute to this by measuring the effect of the concentration of blockholders in more recent years. This will lead to an understanding of the payout decisions managers make. In addition, it is easier to understand the meaning of the payouts decision. Normally, when a company chooses to repurchase stock it can be seen as a signal of undervaluation. However, the company also makes the payout decisions based on the interest of its blockholders. Considering this it can be argued that the findings of this study contribute to less information asymmetry between companies and its stockholders. This paper will continue as follows: in section 2, related literature will be reviewed and a hypothesis is developed about the relationship before and after the crisis between loyalty and stock repurchases. In section 3, the methodology and the data will be described and descriptive statistics will be provided. In section 4, the main results will be presented and a robustness check will be carried out. In section 5, a conclusion about stock repurchases and loyalty in the post-crisis period will be given.
2. Literature review 2.1 Theories regarding payout structure In 1961 Miller and Modigliani introduced their dividend irrelevance theorem. They started with basic assumptions about perfect capital markets, rational behaviour and perfect certainty for the validity of their theorem. With these basic assumptions in mind, they conclude that payout policies do not have impact on stockholders wealth. However, these basic assumptions did not hold entirely when more recent research introduced agency costs as discussed by Jensen in 1986 and security valuation problems as discussed by Myers and Majluf in 1984. These two problems suggest that, because the basic assumptions do not entirely hold payout policy might have an impact on stockholders wealth. A simple asymmetric information framework needs to take these problems into account when examining the need to distribute free cash flows. Even though the decline in dividends is concluded in a lot of studies, dividends do still play a big part in payout policy. Lintner (1956) finds that markets put a premium on the stability or gradual growth of dividends. The level of current dividends is therefore an important determinant of the company’s dividend decision. From this level, management considers whether or not dividends should be changed. If the level of dividend is changed, current earnings are the most important factor determining this amount of change. With this in mind, Brav et al (2005) argue that flexibility is one of the main reasons stock repurchases have increased. They conclude that in the 21st century the findings of Lintner still hold true: the dividend policy of companies is still very conservative. Only in extreme cases dividends are changed by managers and the current level of dividend is primarily taken as given. The flexibility of repurchases allows managers to alter their payout policy in response to good investment opportunities. In contrast, managers first want to raise external funds to pay for investments before they want to cut in their dividend level. Research suggests that dividends start to diminish as a form of payout and stock repurchases take over an important part of the payout policy of companies. However, DeAngelo et al. (2008) raise an important question whether this growth of repurchases will continue. They describe the shortcoming of stock repurchases for loyal stockholders. Before the financial meltdown of fall 2008, many companies repurchased
a large amount of stock at high prices, which eminently exceeded the prices after the collapse of the market. The decision of managers on its payout policy thereby distributed a disproportionate fraction of value to their non-loyal stockholders. Since managers now know this disadvantage, they know that they face the danger to impair the interest of loyal stockholders by repurchasing shares from other owners at high prices. The danger of damaging the interests of loyal stockholders is especially high in times of a financial meltdown and crisis. DeAngelo et al. (2008) point out that this, now widely known fact, might slow down or even reverse the growth of repurchases in the post-crisis period. They therefore suggest that it is indeed interesting to research the question whether loyalty of stockholders is influencing the payout policy of a company and whether managers now incorporate this knowledge in their payout decisions. However, loyalty is a difficult variable to measure. Yet, the paper of Edmans (2009) indicates the important characteristics of blockholders. Like described in the first section, blockholders significantly enhance company value in the long run, even if they do not have many controlling rights. This is caused by the incentive blockholders have to gather costly information: they desire to earn profits at the expense of liquidity investors. Because of this gathering of information the fundamental value of a company will be reflected rather than its current earnings. When a blockholder knows this fundamental value of a company, it will remain a stockholder when the company has good long-term prospects, even if the company faces weak short-term earnings at the moment. On the contrary, stockholders with less information might sell its stock in this situation. These characteristics of blockholders are therefore a good indication of loyalty. A manager, who is aware of the loyalty of its blockholders, will also think long term when making decisions. He is encouraged to invest in the long-term prospects of a firm, not only because a blockholder will remain long-term investor, but also because of the possibility that the blockholder can indeed sell in the short-run. This additionally enhances the value of a company. 2.2 Empirical findings in the literature As it is now investigated intensively, the amount of dividend payouts has declined as opposed to stock repurchases. Fama & French (2001) argue that there is a decline in companies who pay dividends. This decline is seen after the post-1972 peak of 66,5% in 1978. After this year, the amount of non-financial non-utility companies that pay
dividends decline to 20,8% in 1999. This is partly due to the growth of new companies with the characteristics of companies that never paid dividends. These companies have almost the same characteristics, which are as follows: small size, low earnings, and large investments relative to earnings. The decline in the amounts of dividends paid is also partly due to the fact that given their characteristics companies become less likely to pay dividends. Fewer companies initiate dividends and they become slightly more likely to stop paying them. Fama & French perform a logit regression for each year of the 1963-1998 period to explain which firms pay dividends. The explanatory variables they use are profitability, growth, market-to-book ratio and the percent of NYSE companies with the same or lower market capitalization. The regression shows that the coefficients are all very significant in most years. Overall it is mainly mature companies who pay dividends. In contrast, young companies face relatively abundant investment opportunities, while having limited recourses. Because of their higher profitability and fewer attractive investment opportunities, mature companies are more likely to pay dividends. Empirically there is only a rough idea of the characteristics of companies that pay dividends (DeAngelo et al. 2006). Companies that have high-profitability and low-growth rates tend to pay dividends. Conversely, companies with low-profitability and high-growth rates tend to retain their profits (Fama and French 2001). DeAngelo et al. perform a logit analysis on the decision of a company to pay dividends as a function of a couple of variables within the years 1973-2002. Likewise, they use profitability, growth and size to explain this decision. However, they add the variable retained earnings to total common equity to explain the dividend decision. Their analysis shows that this variable has a high t-test value. It is therefore statistically significant and explains why a company pays dividend. They also include the variable cash to total assets and the variable dividends in the prior year. Especially the last variable is a very significant in explaining the decision on dividends. Skinner (2008) indicates the growth of stock repurchases as opposed to dividends. He finds that stock repurchases are replacing dividend in the payout policy of companies. This is proven by the fact that repurchases are increasingly linked to earnings. With an empirical test he provides the evidence on this link between earnings and overall payout policy. In addition, he presents evidence that managers more regularly use repurchases instead of dividends to absorb variation in earnings. Large,
mature, and profitable companies continue to pay dividends, mostly because of their dividend history. In addition, these mature firms also start making repurchases on a regular basis. Skinner finds that companies that only make repurchases have grown over time and other dividend payers are now less important. He even suggests that there may come a time when dividends completely disappear. In his analysis he uses the variables size, profitability, growth, market-to-book ratio, retained earnings to equity and cash to total assets to explain changes in total payouts. In addition, he adds the three variables past return, dividend history and ESO to explain changes in the payout mix. Within the timeframe 1980-2005 these variables are significant in some years. Renneboog and Trojanowski (2011) consider the effect of blockholders on the payout mix. They argue that the dominant type of blockholder in a company can have a substantial influence on its payout decisions. A certain type may prefer dividends due to taxation for instance. Therefore, they control for the effect that certain groups of blockholders might have on the payout mix. They examine the likelihood that a company pays its fund out to stockholders and the likelihood that company chooses for a particular form of payout for UK companies over the period 1992-2004. The variables used to analyse this decision are size, profitability, leverage, investment opportunities and the ownership variables. With their analysis they find that certain groups of blockholders have different preferences for payouts. For example, control power of tax-exempt financial institutions relates to a higher probability that a company makes payouts. 2.3 Conclusions of the literature All in all, both dividends and stock repurchases play an important part in the payout policy of a company. Managers are reluctant to change dividends, because markets put a premium on the stability or gradual growth of dividends. The main advantage of stock repurchases is their flexibility. Previous research suggests that stock repurchases are increasingly important and that dividends are a diminishing part of companies’ payout policy. These papers did not take into account the now well-known disadvantage of stock repurchases. The growth of stock repurchases might actually be reversed, because managers do not want to alienate their loyal stockholders by distributing too much cash to non-loyal stockholders. It is therefore interesting to examine the effect of the presence of loyal blockholders on the payout mix of a company.
In conclusion, empirical work suggests that profitability, size and growth are important determinants of a payout policy. Overall, companies with low-profitability and high-growth rates tend to retain their profits. In contrast, mature companies usually distribute their earnings to stockholders. In previous research these characteristics are included in the regression to explain the payout decisions of a company. DeAngelo et al. (2006) add a few variables to such a regression to explain differences in the payout mix. Skinner (2008) also use the basis of profitability, size and growth to explain total payouts and adds a few variables to explain the payout mix. Renneboog and Trojanowski (2011) already point out the importance of blockholders on the payout mix. However their analysis only contains information until 2004 and does not examine overall the effect of the concentration of blockholders. All in all, research concludes on the relevant determinants of total payouts and the payout mix. Nevertheless, it is still interesting to determine whether loyalty, as measured by the overall blockholder concentration, influences the decisions a company makes on its payout mix. Altogether, it is expected that stock repurchases will still be used widely, but that the scope of stock repurchases programs will be tempered. Therefore, the suggestion that there may come a time when dividends completely disappear is probably not the case. Managers are now aware of the disadvantage that stock repurchases have for loyal stockholders and might want to take this into consideration when constructing their payout policy. The variable loyalty, which is measured by stockholder concentration, is expected to have a positive effect on the ratio dividends to total payout. This variable will be measured using the Herfindahl-Herschman index of the three biggest stockholders of a company. This index will be used in order to measure the effect that large stockholders of a company have on the payout ratio. A company, which has many stockholders with a substantial percentage of the total stock, will repurchase less stock than a company without any of these loyal stockholders. This index variable will have a positive effect on the ratio dividends to total payouts. In addition, the dummy variable crisis is expected to have a positive effect on the ratio dividends to total payout. In the years 2003-2007 (dummy variable will be zero) this disadvantage was not well know, so it is expected that the ratio is relatively lower for these years.
3. Methodology and Data 3.1 Methodology Skinner investigates the evolving relationship between earnings, dividends, and stock repurchases in 2008. He examines the fact that dividends are increasingly replaced by repurchases. Skinner concludes that repurchases have become the dominant form of payout for corporations. In his analysis he performs an OLS regression of the payout mix on the hypothesized determinants for the period 1980-2005. In this paper the evolving relation in the payout mix is also examined and so almost the same model will be used. However, the effect of the crisis and the effect of loyalty of stockholders should now be included. Holderness (2009) found that 96% of U.S. public firms have at least one blockholder. It is therefore not convenient to take the dummy variable whether a company has or does not have a blockholder. His study offers a conclusion on ownership concentration in the USA: namely that ownership concentration is similar in the USA compared to other countries. This contradicts the belief that ownership in the United States is diffused. A way to calculate the concentration of ownership is to take a Herfindahl-Herschman of the three biggest stockholders of a company. This index is therefore useful to measure loyalty. The regressions, which follows from Skinner and incorporates both loyalty and crisis, is the following: 1) ! !! =∝!+∝! 𝑆𝑖𝑧𝑒!+∝! 𝑅𝑂𝐴 +∝! 𝐺𝑟𝑜𝑤𝑡ℎ!+∝! !"#$%& !""# !+∝! !"#! !" !+∝! !" !"!+ ∝! 𝑃𝑎𝑠𝑡 𝑅𝑒𝑡𝑢𝑟𝑛! +∝! 𝐸𝑆𝑂!+∝!𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑝𝑟𝑖𝑜𝑟 𝑦𝑒𝑎𝑟!+ ∝!" 𝐿𝑜𝑦𝑎𝑙!+ 𝑢! 2) ! !! =∝!+∝! 𝑆𝑖𝑧𝑒!+∝! 𝑅𝑂𝐴 +∝! 𝐺𝑟𝑜𝑤𝑡ℎ!+∝! !"#$%& !""# !+∝! !"#! !" !+∝! !" !"!+ ∝! 𝑃𝑎𝑠𝑡 𝑅𝑒𝑡𝑢𝑟𝑛! +∝! 𝐸𝑆𝑂!+∝!𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑝𝑟𝑖𝑜𝑟 𝑦𝑒𝑎𝑟!+ ∝!" 𝐿𝑜𝑦𝑎𝑙!+∝!! 𝐶𝑟𝑖𝑠𝑖𝑠!+ 𝑢! 3) ! !! =∝!+∝! 𝑆𝑖𝑧𝑒!+∝! 𝑅𝑂𝐴 +∝! 𝐺𝑟𝑜𝑤𝑡ℎ!+∝! !"#$%& !""# !+∝! !"#! !" !+∝! !" !"!+ ∝! 𝑃𝑎𝑠𝑡 𝑅𝑒𝑡𝑢𝑟𝑛! +∝! 𝐸𝑆𝑂!+∝!𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑝𝑟𝑖𝑜𝑟 𝑦𝑒𝑎𝑟!+ ∝!" 𝐿𝑜𝑦𝑎𝑙!+∝!! 𝐶𝑟𝑖𝑠𝑖𝑠!+ ∝!" 𝐿𝑜𝑦𝑎𝑙!∗ 𝐶𝑟𝑖𝑠𝑖𝑠!+ 𝑢!
Each of these regressions includes nine control variables. Skinner tabulates the results of these control variables over the years 1980-2005. Some of these have a significant result in his regression, but this also varies other this time interval. The first six control variables Size, ROA, growth, market-to-book, CashTA and RE/SE are designed to capture variation in the level of payout and not the variation in the mix of payouts. They are therefore not expected to have a significant effect on the ratio of dividends to total payouts. The variable size should have a positive effect on the total amount of payouts. A more mature firm, which does not have that many interesting investment opportunities, will make more payouts and retain less cash. The variable Return on Assets incorporates the profitability of a company into the model. High profitability characterises companies, which make high payouts. Therefore, this variable will also have a positive effect on the overall level of payouts. The control variable growth is measured by the percentage change in total assets from the prior year. If a company has a high growth rate, it is expected to make fewer payouts. Companies with a high growth rate have a lot of investment opportunities and therefore want to retain their FCF’s. Growth therefore negatively influences total payouts. A higher market-to-book value also reflects greater expected future gains because of perceived growth opportunities. With the same arguments as for growth, it can be argued that this also negatively effect total payouts. The fifth variable in this section is cash to total assets. If a company has a high amount of cash it can choose to divide this among its stockholders. Higher cash to assets ratio therefore positively affects the total amount of payouts. DeAngelo et al. (2006) examine which characteristics influence companies to make payouts. They concluded that companies tend to pay dividends when retained earnings are a large fraction of total equity. They argue that a higher ratio is consistent with the characteristics of firms that make payouts. Their findings are consistent with the life-cycle theory of dividends. Therefore retained earning to equity is expected to have a positive effect on total payouts. A variable, which is expected to have a significant effect on the dependent variable, is called past stock return. When a companies’ stock price is low, managers will be more likely to repurchase than in the case of high stock prices. This variable is therefore expected to have a positive effect on the ratio. The control variable Employee stock option is also expected to have a significant effect. ESO plans will dilute stockholders’ claim. Thus, when there are large ESO holdings, there is a strong tendency
to favour repurchases over dividend. This variable is expected to have a significant negative effect on dividends. Dividend prior year also has an effect on the ratio dividends to total payouts. Managers are reluctant to cut dividends or to stop paying them. This could substantially lower the stock price and the prospects of future equity. The current level of dividends is primarily taken as given and almost never changed, since markets put a premium on the stability of dividends. Thus, this variable will probably have a positive effect on the ratio dividends to total payouts. The index variable loyal and the dummy variable crisis are the most important variables in this regression. DeAngelo et al. (2008) suggest that these variables will have an effect on the decision managers will make about the payout mix. With the benefit of hindsight, managers now know that stock repurchases can benefit the selling stockholders at the expense of the loyal stockholders. The most important variable loyal measures whether the magnitude of the holdings of the three biggest stockholders changes the payout mix. If there are more stockholders with a significant amount of stock, this ratio will be higher. If there are no blockholders in a company, this ratio will be very low. It is therefore argued that this variable has a positive effect on the amount of dividends a company will pay. The second variable crisis is hypothesised to have a positive impact on the dependent variable since the described disadvantage only became clear after the crisis. Three regressions will examine the effect of these two variables. The first regression shows the overall effect of loyalty, measured by stockholder concentration. The second regression measures both the effect of the variable loyalty and the dummy variable crisis on the payout mix. The third regression uses an interaction variable, which examines whether loyalty has a different effect in either timeframe (2003-2007 and 2008-2012). 3.2. Data and descriptive statistics To perform this regression the information on the variables is collected in Compustat for North-American companies. This annual data is collected after inserting two conditional variables in the search request, namely: common dividends had to be bigger than zero and/or the purchase of common and preferred stock had to be bigger than zero. Within this North-American data, only the USA companies are used. In the end a big dataset remained with annual information on USA companies, which either paid dividends or repurchased stocks, within the years 2003-2012.
This information is available for over 3000 companies. Nonetheless, the information on blockholders is not that easily obtained. Even though this data can be found, it has to be done so manually with the database Thomson ONE. However for the years, which have to be examined, this is very time consuming considering that each company has to be looked up individually and each year has to be looked up in a different search request. Therefore a random selection of 200 companies is made. For these selected companies data on dividend and net repurchases also needs to be collected for the years 2001 and 2002, because the dependent variable will include both repurchases and dividends of the current and the previous two years. Lagged assets are also used in the model, which is why data on assets also needs to be included for the year 2002. For the same year data on the closing price needs to be collected to calculate the return in 2003. The rest of the data only needs to include the years 2003-2012. If the company uses the treasury stock method for repurchases, the net stock repurchases are measured by taking the increase in common treasury stock. When the treasury stock is zero in the current and prior year, it is inferred that a company uses the retirement method. Net repurchases are then measured as the difference between the purchases of common and preferred stock and stock issuances from the statement of cash flows. If the amount of net repurchases is negative, repurchases are set to zero. Each of these 200 companies will have information about the variables on several years that they paid either dividend or repurchased stock. Thus, the dataset will now have 1421 observations, where each company includes its own number observations within the timeframe 2003-2012. Each of these 1421 observations is manually looked up in Thomson ONE and the information on stockholders is added to the data we derived from Compustat on the other variables. The data, which is imported into STATA, will be used as a panel data set. However, before importing the data into STATA, some outliers were deleted. Four companies had to be deleted in total, because they included numbers, which were at least ten times the standard deviations. Table 1 tabulates the summary statistics for both dependent and independent variables. The mean of the dependent variable is 0.628. This means that on average companies still use around 63% of their payouts to pay dividends. There are some companies who only use repurchases and some companies who only pay dividends. From this dataset it can be argued that companies still mainly use dividends
in their payout mix. The table also points out that there are almost as many observations before the crisis as there are after the crisis, as the mean is almost .5000. The most important fact, which can be derived from the correlation matrix in Appendix 1, is that the variable loyalty indeed has a positive relationship with the dependent variable. If the concentration of blockholders increases, there tend to be less stock repurchases as opposed to dividends. Crisis was also expected to have a positive relationship with the dependent variable. However this is not the case, crisis is negatively correlated with the dependent variable. Past return of stock does not have the expected sign and is negatively correlated with the ratio. The expectation that the variable for employee stock option expense would be negatively correlated with the dependent variable is also not the case. The table also shows that there are not highly correlated variables. 4. Analysis 4.1. Empirical Results To perform the regression a panel data set is used. To make sure whether fixed or random effects should be used, a Hausman test is performed. After running this test it can be concluded with a probability of 0.0000, that fixed effects should be used in the regression. This table is included in the Appendix 2. A modified Wald test and a Wooldridge test in STATA point out that there is both heterosekasticity in the fixed effects model as there is serial correlation in the panel data. Therefore, the regressions will be made with robust standard errors to correct for this model robustness. Table 1: Summary statistics for both dependent and independent variables
Variable Mean Std. Dev. Min Max Observations
DP .6281532 .3597072 0 1 N=1400 n=191 Size 6.203205 2.358625 0.4999619 11.07347 N=1419 n=192 ROA .115259 .1306763 -.8701137 1.137969 N=1379 n=188 Growth .1279159 .5347275 -1 13.93178 N=1417 n=192 MB 2.276729 3.216132 -18.83437 49.75108 N=1416 n=191 RESE .6111996 2.031298 -33.43898 26.68421 N=1419 n=192 CashTA .1013443 .1214453 0 .918637 N=1419 n=192 PR .074082 .450155 -1 3.800943 N=1336 n=184 ESO .0069738 .0581489 -.0536197 1.045343 N=1418 n=192 DPY .8408451 .3659494 0 1 N=1420 n=192 Index .0354484 .075799 0 0.6848808 N=1415 n=192 Crisis .4711268 .4993415 0 1 N=1420 n=192
Table 4 presents the main results from a random effect regression. The two main variables in the regression are the loyalty index and the dummy variable crisis. Both were expected to have a positive impact on the ratio dividends to total payout. From the first regression it can be concluded that the variable loyalty does have the expected effect on the payout ratio. When the index of loyalty is increasing, the ratio of dividends to total payout is increasing. This means that loyalty indeed decreases the amount of purchases a managers wants to make. This variable is significant with a p-value of 0.024. In the second regression the effect of the crisis on the payout mix is also added. This variable does not have this expected effect on the ratio. This dummy variable has a small negative impact instead. Which means that after the Crisis the amount of stock repurchases is actually bigger than in the years before the crisis (2003-2007). The growth of stock repurchases as described in the literature review, is still confirmed by this analysis. However, this variable is not significant with a p-value of 0.274. In this regression loyalty also has a positive effect on the payout ratio with a significant p-value of 0.051. The third regression also examine if the effect of loyalty changes in the years after the financial meltdown. From the regression it can be concluded that this variable does not have a significant impact on the payout mix. The corresponding p-value of 0.426 shows that this variable is not slightly significant. In conclusion, the thought of DeAngelo et al. (2008) can be confirmed. Loyalty of stockholders does have a positive effect on the payout ratio of a company. The crisis is however not positively correlated to the payout ratio and has a negative effect. This is however not influenced by the crisis. This can be concluded from the third regression, which includes an interaction variable between the index of Loyalty and the dummy crisis. Size, return on assets, growth, market-to-book ratio, retained earnings to common equity and cash to total assets are all incorporated to capture variation in level of payout, as opposed to the mix of payouts (Skinner 2008). The three regressions in table 4 shows that there are no significant values for the variables Size, retained earnings to book value of equity, and cash to total assets. It was however not expected that ROA, growth, and market-to-book would have a significant value in these regressions. These variables do however have a significant effect on the payout mix. ROA was expected to have a positive effect on total payout. From the table it can be derived that it also has a negative
effect on the payout ratio. The higher the profitability of a firm, the more stock repurchases a company will make as opposed to dividends. Growth also has a significant positive effect on the payout ratio. The variable growth is measured by the percentage change of assets. Most of the time a lower growth rate will mean that a company is in the mature phase of its lifecycle. The other way around, a high growth rate will suggest that it is a new firm with a lot of investment opportunities. From the table it follows that a higher growth rate causes a company to pay more dividends as opposed to repurchases. Combining the variables ROA and growth rate, it can be concluded that more mature firms repurchase relatively more than new firms with low growth rates and low profitability. This is in contrast with earlier research, which pointed out that new firms did not start paying dividends and only wanted to use stock repurchases, whereas a lot of mature firms did still continued paying dividends. Market-to-book contradicts the expectation that it would have the same effect as the variable growth. From the analysis Table 4: Main results: analysis of the hypothesized determinants on Dividends to Total Payouts for 150 USA companies in the years 2003-2012 The dependent variable ratio dividends to total payouts is total dividend in the current and previous two years divided by the sum of total payouts over the same interval. Size is the natural logarithm of total assets. ROA is measured by dividing operating income before depreciation by lagged total assets. Growth is the percentage change in total assets from the prior year. The market-to-book variable is measured by dividing the market value of equity by the book value of common equity. Cash/TA is measured by dividing cash by total assets. RE/SE is measured by dividing the retained earnings by the book value of equity. Past Return is the raw stock return. ESO is the pro forma employee stock option expense divided by sales. History is a dummy variable, which is one if the company paid dividends in the prior year. Loyal is a Herfindahl-Herschman of the three biggest stockholders of a company. Crisis is a dummy variable: 0 for the period 2003-2007 and 1 for the period 2008-2012. _cons is the constant in the regression. *,** and *** indicate significance at 10%, 5% and 1% respectively Dependent variable: Dividend/Total Payouts (1) (2) (3) Size .0161905 (0.46) .0341008 (0.95) .0035016 (0.93) ROA -.2388785 (-1.65)* -.274429 (-1.89)* -.2792364 (-1.89)* Growth .0673521 (4.74)*** .0661635 (4.99)*** (4.96)*** .066344 MB -.0095016 (-3.35)*** -.0098797 (-3.38)*** -.0098566 (-3.38)*** RESE .0013166 (0.18) .0019739 (0.25) .0019103 (0.25) CashTA -.0253185 (-0.17) -.00365 (-0.02) -.0029513 (-0.02) PastReturn .0160488 (1.24) .0156859 (1.22) .0165597 (1.28) ESO -.0752419 (-1.57) -.0661217 (-1.38) -.0666787 (-1.38) Dividend Prior Year .1828397 (4.13)*** .1908152 (4.32)*** .1911419 (4.34)*** Loyal .3412096 (2.28)** .301766 (1.96)* .3022638 (1.97)* Crisis -.0250258 (-1.10) -.0211664 (-0.87) Index*Crisis -.1737604 (-0.79) _cons .3917095 (1.70)* .2895727 (1.26) .2938412 (1,27) R-square 0.2606 0.2084 0.2087 F-test F(11,176)=7,05
Prob>F=0.0000*** Prob>chi2 =0.0000*** F(11,176) = 6.87 Prob>chi2=0.0000*** F(12,176) = 6.29 N n 1288 177 1288 177 1288 177
it can be concluded that this variable has a negative effect, which is very significant. This means that a higher market-to-book does have a negative impact on the dividends a company pays. This is in line with earlier research, which suggests that high growth companies pay less dividends than mature companies do. New companies are suggested to initiate dividends less often and in particular use stock repurchases as a form of payouts. The three control variables, which should explain a relative higher amount of repurchases, are the variables PastReturn, ESO and Dividend Prior Year. The variable PastReturn is not significant at a 10% significance level, though it does have a positive sign. The intuition behind this is, that a manager will be more likely to repurchase shares in the case of low stock prices. When stock prices are high, managers are less likely to repurchase stock. The other variable, which should explain repurchases, is ESO. The sign of this variable is indeed negative, though it is not significant. The argument behind this variable would be that when a company has ESO plans, that this will dilute stockholders’ claim. Thus when there are large ESO holdings, there is a strong tendency to favour repurchases over dividend. The model does not confirm this tendency. The last variable Dividend Prior Year is indeed very significant and has a positive effect on the ratio dividend to total payout. Managers are reluctant to change their dividend policy. Therefore, they will more often continue paying dividends if they also paid dividend in the prior year. 4.2. Robustness check In the panel data regression the effect of loyalty on the payout mix became clear. It is however interesting to look at the results per year to see the separate effects of the coefficients. Therefore, an annual OLS regression is performed to determine the effects of the control variables and the variable loyalty. A separate data set is generated for each year from the existing panel data set. This new data set is therefore about 10% of the total data set, and it contains on average information on a 130 companies per year. From these OLS regressions it can be argued that most of the variables do not change much over time. The only variable, which does substantially change is ESO. Nevertheless, it is only a relatively small number of companies (about 5% of the companies) that do actually have employee stock option expenses. Thus it can be argued in the OLS regressions this variable does not play an important part in explaining the payout mix.
The significance of the coefficients does however vary over time. The only variable, which is in every year very significant, is the variable dividend prior year. This was also considered very significant in table 4. In table 4 the variable Growth was also considered to be very significant. In the separate OLS regression this is however in almost every year not the case. The variable ROA and Market-to-book do indeed show a significant coefficient in a substantial number of years. In the panel regression, main variable was the index variable, which was significant with a p-value of 0.052. From this regression we concluded that loyalty does have an effect on the payout mix. It also had the hypothesized effect and caused companies to pay out more dividends as opposed to repurchases. However, in the annual OLS regression this variable is almost never significant. From the OLS regression it can be argued that in individual years it did not matter if the concentration of stockholders became higher. This means that the null hypothesis is not rejected and from this analysis it can be concluded that loyalty does not play a part in the decision on the payout mix. It was hypothesized that the crisis would play a part in the significance of the index. However, there are not more significant values of index after the crisis than that there are before the crisis. It is however interesting, that in the actual years of the crisis (2007-2008) this variable became significant. In a financial meltdown, companies actually paid more dividends as opposed to stock repurchases. All in all, this analysis shows opposite results from our normal panel data regression. However, the panel data regression made use of all the 1421 observations. With only a Table 5: Annual OLS regression of payout mix for 200 Compustat companies that either paid dividend or make repurchases in the period 2003-2012. The dependent variable ratio dividends to total payouts is total dividend in the current and previous two years divided by the sum of total payouts over the same interval. Size is the natural logarithm of total assets. ROA is measured by dividing operating income before depreciation by lagged total assets. Growth is the percentage change in total assets from the prior year. The market-to-book variable is measured by dividing the market value of equity by the book value of common equity. Cash/TA is measured by dividing cash by total assets. RE/SE is measured by dividing the retained earnings by the book value of equity. Past Return is the raw stock return. ESO is the pro forma employee stock option expense divided by sales. History is a dummy variable, which is one if the company paid dividends in the prior year. Loyal is a Herfindahl-Herschman of the three biggest stockholders of a company. Crisis is a dummy variable: 0 for the period 2003-2007 and 1 for the period 2008-2012. *,** and *** indicate significance at 10%, 5% and 1% respectively
Year Int. Size ROA Growt
h MB RESE CashT A PastRe turn ESO Dividen d prior year Index 𝑅! 2003 0.24** 0.02 -0.91** 0.15 -0.01 0.05 -0.34 -0.05 -0.14 0.45*** 0.46 0.44 2004 0.18* 0.01 -0.22 0.01 -0.02*** -0.02 -0.34 0.12** -0.14 0.62*** 0.15 0.51 2005 0.32*** 0.01 -0.68** 0.17 -0.01** -0.05 0.08 -0.02 -2.43** 0.49*** 0.53* 0.42 2006 0.17* 0.01 -0.73*** 0.17 -0.02 -0.03 -0.02 -0.03 -30.07* 0.61*** 0.69** 0.44 2007 0.48*** -0.00 -1.07*** -0.05 -0.00 -0.04 -0.21 0.05 0.54 0.29*** 0.64* 0.27 2008 0.20* 0.01 -0.64*** 0.08 -0.02* -0.02 -0.40* 0.13 0.03 0.47*** 1.02** 0.37 2009 0.31*** 0.01 0.14 0.17 -0.05*** -0.01 -0.35 -0.05 0.14 0.34*** 0.50 0.27 2010 0.13 0.01 -0.52* 0.18* -0.02 -0.02 -0.33 0.03 0.27 0.66*** 0.57 0.45 2011 0.35*** -0.01 -0.30 0.07 -0.02* -0.02 -0.25 0.00 -28.00** 0.53*** 0.25 0.31 2012 0.23* 0.00 -0.29 0.13 -0.02** 0.01 0.06 -0.17 -3.40 0.51*** 0.69 0.24
Therefore, the panel data could be seen as a better predictor of the determinants of the payout mix. 5. Conclusion and discussion In this paper the effect of loyalty, as measured by stockholder concentration, on the ratio of dividends to total payouts has been examined. The data was collected for 200 randomly selected companies within the time frame 2003-2012. Information about the three main stockholders was added to these observations and from this information a loyalty variable was created. This index variable contained a Herfindahl-Herschman index of the three main stockholders of a company. It had been suggested that loyalty has a negative effect on the amount of repurchases a company wants to make. The model has significantly confirmed this suggestion. In addition a dummy variable was included to examine whether the crisis had a negative effect on the growth of stock repurchases. From a robust regression it was be concluded that the crisis did not have a significant effect on the ratio dividends to total payouts. In addition, crisis also had a negative effect on this ratio instead of a positive one. This means that the growth of stock repurchases indeed continues. This is in line with previous research, which all argued that dividends as a form of payout are declining. The effect of loyalty is however not dependent on the crisis. The effect of loyalty is not bigger after the crisis than it is before the crisis. This is confirmed by the not significant interaction variable between the index and the dummy variable crisis. Thus, this does not confirm the suggestions of DeAngelo et al (2008) that loyalty would have a negative impact on the growth of stock repurchases after the crisis. It can now be considered that loyal stockholders indeed prefer relatively less repurchases. Managers have to take this into account when making up their payout mix. They have to take into consideration how many loyal stockholders there are in a company and whether these blockholders just have a 5% share or possess a bigger block of shares. A manager always needs to take the interest of its stockholders into account. Thus, if a company has lots of loyal stockholders it is easier to make the decision on paying more dividends as opposed to stock repurchases. Therefore, these findings can help mangers to understand the interest of its stockholders and can help to reflect these interests in its payout policy. For an investor it might also be interesting to know that
loyal stockholders prefer stock repurchases. If an investor wants to relatively receive more dividends than payout in the form of stock repurchases, it can choose to invest in a company with more loyal stockholders. This study can therefore help investors making their investment choices. There possibly are some implications for this study. For example, it must be considered that this data only contains 200 companies. Therefore, it would perhaps be interesting to increase this sample further. Secondly, it must be considered that blockholders also have other influences on companies. Most of the time, blockholders play part as an active stockholder in a company. Such a stockholder often achieves better corporate governance for a company. This can lead to higher payouts and it might also lead to higher net stock repurchases. Because of these limitations, further research could look into the effect loyalty has within a bigger data set of companies. It might also be interesting to examine the effect of loyalty in different countries in the world. Some loyal investor might still prefer repurchases to dividends and these preferences might differ all over the world. Another option for further research might be to investigate whether it is only the improved corporate governance, which causes stock repurchases to relatively increase or that is indeed only the loyalty which causes repurchases to increase.
Appendix 1 Appe ndix 2 Correlation Matrix
DP Size ROA Growth MB RESE CashTA PR ESO History Index Crisis
DP 1.0000 Size 0.1384 1.0000 ROA -0.2274 0.0138 1.0000 Growth 0.0277 -0.0092 0.2355 1.0000 MB -0.1682 0.0200 0.2977 0.0219 1.0000 RESE 0.0274 0.1104 0.1299 -0.0247 -0.1409 1.0000 CashTA -0.2031 -0.2481 0.2026 -0.0185 0.1303 -0.0440 1.0000 PR -0.0451 -0.0489 0.1544 0.1645 0.1309 -0.0576 0.0921 1.0000 ESO 0.0095 -0.0405 -0.0772 -0.0094 -0.0085 -0.0642 -0.0000 -0.0098 1.0000 DPY 0.4899 0.2170 -0.0990 -0.0403 0.0092 0.1172 -0.1313 -0.1016 -0.0172 1.000 Index 0.0512 -0.1712 0.1407 0.0065 0.0115 0.0287 0.1221 0.0917 -0.0183 -0.0185 1.0000 Crisis -0.0162 0.0570 -0.0643 -0.0699 -0.1548 0.1013 0.0564 -0.0712 0.0104 0.0692 -0.0685 1.0000 Hausman test Null hypothesis: the preferred model uses random effects. Test whether the unique errors are correlated with the repressors. If the null hypothesis is not rejected, they are not correlated. ---Coefficients --- (b) Fixed (B) Random (b-B) Difference Sqrt(diag(V_b-V_B)) S.E. Size .0341549 .0137938 .0203611 .0241981 ROA -.2745917 -.3062374 .0316457 .0658097 Growth .0661583 .0658131 .0003452 .0035347 MB -0.0098831 -.0112842 .0014011 .0009438 RESE .0020222 .006376 .0013846 .0017682 CashTA -.0048939 -.1705087 .1656148 .0661274 PastReturn .0160599 .0188311 -.0027712 .0022762 ESO -.0660433 -.0636618 -.0023815 .014254 History .1909516 .2594666 -.068515 .0126918
References Banyi, M. E.A. Dyl, and K.M. Kahle (2008), ‘Measuring share repurchases’. Working paper, Oregon State University and University of Arizona. Black, F. (1976), ‘The dividend puzle’. Journal of Portfolio Management 2, 5-8. Brav, A., J.R. Graham, C.R. Harvey and R. Michaely (2005), ‘Payout policy in the 21st century’. Journal of Financial Economics 77, 483-528. De Angelo, H., L. DeAngelo, and R.M. Stulz (2006). ‘Dividend policy and the earned/contributed capital mix: a test of the lifecycle theory’. Journal of Financial Economics, 81, 227-254. DeAngelo, H., L. DeAngelo, and D.J. Skinner (2008), ‘ Corporate Payout Policy’. Foundations and Trends in Finance 3, No. 2-3 Edmans, A. (2009), ‘Blockholder trading, Market Efficience, and Managerial Myopia.’ The Journal of Finance 64, No. 6, 2481-2513. Fama, E.F. and K.R. French (2001). ‘Disappearing dividends: changing firm characteristics or lower propensity to pay?’ Journal of Financial Economics 60, 3-40 Holderness, C.G. (2009), ‘The myth of diffuse ownership in the United States’. The Review of Financial Studies 22, 1377-1408. Index .300872 .3105937 -.0097217 .075086 Crisis -.0253197 -.0273368 .0020171 .0068422 Chi2(11) = 54.66 Prob>chi2= 0.0000 Prob>chi<0.05 -> use fixed effects
Lintner, J. (1956), ‘Distribution of incomes of corporations among dividends, retained earning, and taxes’. Journal of Accounting and Economics 39, 97-113 Miller, M. and F. Modigliani (1961), ‘Dividend policy, growth, an the valuation of shares’. Journal of Business 34, 411-1051. Renneboog, L. and G. Trojanowski (2011), ‘Patterns in payout policy and payout channel choice’. Journal of Banking & Finance 35, 1447-1490 Skinner, D.J. (2008), ‘The evolving relation between earnings, dividends, and stock repurchases’. Journal of Financial Economics 87, 582-609
