“Dividend policy on controlling and non-controlling shareholders: case in Indonesia”

The purpose of this study is to examine dividend policy on both the controlling and non-controlling shareholders based on assumptions according to theories of life cycle, and free cash flow. The sample for this study is 241 listed firm in Indonesia Stock Exchange during the period from 2010 to 2015. This study divides the sample based on quartiles and analyzes it by conducting logistic regression with significant rate at 0.05. This study provides the evidences that: (1) firms as dividend payers tend not distribute their dividend for controlling shareholders and non-controlling shareholders while the composition for both shareholders are almost equal; (2) firms as dividend payers also have tendency not to distribute dividend on controlling shareholders when this shareholders have largest percentage of ownership; and (3) firms as dividend payers tend not distribute dividend on non-controlling shareholders while they have lowest retained earnings. The findings imply that life cycle theory and free cash flow theory can explain the behavior of dividending policy on controlling shareholders and non-controlling share-holders depend on their circumstances. The study uses alternative measurement for non-controlling shareholders as this variable together with controlling shareholders are moderating the other independent variables for testing the model of dividend policy.


INTRODUCTION
The "dividend puzzle" by Black (1996) is still recent issue for most studies in the fields of economics and finance.Black (1996) finds that insiders use the dividend policy as a tool to convey the information in order to align the interests between them and shareholders or investors.Asquith and Mullins (1986) confirm this circumstance as a signal as the firm is viewed as "black box" by the outsiders.Black (1996) explains that if the insiders give bad news to market, then the reduction on dividends will trigger the sentiment of investors, which causes the decrease of the firm's share market price, but the opposite impact will happen on share market price if the insiders give the good news.These facts are reasonable in condition when the investors believe that dividend shall increase their wealth (Asquith & Mullins, 1986).
Most of experts find that dividend policy can be viewed from many aspects.H. DeAngelo, L. DeAngelo, and Stulz (2006) as confirmed by Fairchild, Guney, and Thanatawee (2014) find that firm shall distrib-ute dividends for shareholders while firms are at mature level.Grullon, Michaely, and Swaminathan (2002) also explain that at mature level, firms tend have less investment opportunities, but available for much free cash, which is available to distribute as dividends for shareholders.On the other hand, Grullon, Michaely, and Swaminathan (2002), as confirmed by Fairchild, Guney, and Thanatawee (2014), show that as the firms are at mature level, they have tendency to face internal conflict or agency problem between insiders and outsiders in the context to decide whether the free cash should be allocated on unprofitable investments, which benefit for insiders or to be distributed as dividends for shareholders in order to maximize their wealth.
The motivation of this study starts from the pros and cons of prior findings about dividend policy being still around, since many studies have different arguments from the perspective of some theories such as life cycle, and free cash flow or agency problem.The studies of Grullon, Michaely, and Swaminathan (2002), H. DeAngelo, L. DeAngelo, and Stulz (2006), and Fairchild, Guney, and Thanatawee (2014) provide evidence that firm dividend policy has close relationship with their maturity.On the other hand, the finding of Bøhren, Josefsen, and Steen (2012) as confirmed by Fairchild, Guney, and Thanatawee (2014), and Lin, Chen, and Tsai (2017) also show that most dividend policies have relationship with internal conflict or agency problem between shareholders as principal and insiders as agent in case of usage on firm's free cash.Furthermore, there are many studies that find that there is a relationship between firm dividend policy and its ownership, such as state ownership (Gugler, 2003; Lin, Chen, & Tsai, 2017), institutional ownership (Wei, Wu, Li, & Chen, 2011; Reyna, 2017), individual ownership (Bøhren, Josefsen, & Steen, 2012), and family ownership (La Porta, De Silanes, Shleifer, & Vishny, 2000; Gugler, 2003; Wei, Wu, Li, & Chen, 2011).Those findings show interesting point that minority shareholders by their small bargaining power together with majority shareholders have a role in determining dividend distribution.The similar circumstance is also confirmed by Gugler and Yurtoglu (2003).
This study notices the basic assumptions by Miller and Modigliani (1961) that dividend policy is mostly affected by: (1) perfect markets; (2) rational behavior; and (3) perfect certainty.These assumptions imply that as the emerging market, most of outside investors in Indonesia are still uninformed, which brings them have difficulty to set their portfolios in an objective to get dividends at optimum return.At normally, the informational content by dividend shall lead the Indonesian investors to react for overvaluing the share prices.But, on the other hand, the investors also shall behave irrationally in preferences between dividend or capital gains.
The purpose of this study is to examine dividend policy on each firm's ownerships.To meet this objective, the study sets the assumptions according to theories of life cycle, and free cash flow or agency problem.On these settings, this study uses retained earnings ratio and price to book ratio as the proxies of maturity in the context of life cycle theory, and long-term debt ratio as the proxy to detect agency problem in the context of free cash flow theory.In order to confirm dividend policy on each firm's ownerships, the study uses variables of each ownership to moderate each proxy for each theory.This study divides the ownership into two categories, which are: (1) controlling shareholders; and (2) non-controlling shareholders.This study identifies that the controlling shareholders for most of Indonesian public firms can be a combination of state, institutional, and individual, while non-controlling shareholders are normally called public ownership.
This study provides the evidence that dividend payers tend not to distribute dividend for controlling shareholders and non-controlling shareholders when they have almost equal composition of ownership (such as 25% of sample or area below Q1).Also, the evidence shows that firms as dividend payers also have tendency not to distribute dividend on controlling shareholders when these shareholders have largest percentage of ownership (such as area above Q3).The other evidence shows that firms as dividend payers only distribute dividend on controlling shareholders when they have the lowest mean of retained earnings ratio (such as area between Q1 and Q2).Overall, the findings imply that life cycle theory and free cash flow theory can explain the behavior of dividend policy on controlling shareholders and noncontrolling shareholders depending on their circumstances.The remainder of this study proceeds as follows: section 1 reviews relevant literature and develops the hypotheses, section 2 describes research method of this study, section 3 provides the result of analysis and discussion, final section concludes the study with its implications.

Life cycle theory
The study of H1: The more mature the firm, the higher the probability to pay dividend.

Free cash flow theory
Myers (2001) says that "free cash flow theory is designed for mature firms that are prone to overin-vest".Myers (2001) clarifies that despite the financial distress threat, the risky debt shall increase the firm value when firm have excess cash flow over their profitable investment opportunities.On prior studies, Modigliani and Miller (1958) as confirmed by Miller and Modigliani (1961) explain that the interrelation between dividend policy and debt shall arise when dividend distributions will not reflect the future earning opportunities under assumption of uncertainty, unstable dividend, and the presence of debt used for financing.
The study of Easterbrook (1984) as confirmed by Jensen (1986) shows the circumstances as manifestation of the conflict of interest between principals and agents, which give the result for principals deciding to use debt and demand for dividend for controlling the agents.
The study of Grullon, Michaely, and Swaminathan (2002) shows that firms should distribute the earnings as dividends to reduce the excess cash and prevent the overinvestment by agents.The study of Agrawal and Jayaraman (1994) during the period from 1979 to 1983 on 71 firm shows that interaction between debt policy and dividend policy with the aim to reduce the agency problem and at once supports the finding of Jensen (1986).They show that increasing leverage shall increase the demand for dividend by shareholders.As the implication for agency problem, Ince and Owers (2012) suggest that under tax regimes, in order to maintain or to increase the firm value, then obtaining more debt should be followed simultaneously by increasing the dividend payment.
The study of Lin, Chen, and Tsai (2017) on Chinese firms listed on Shanghai and Shenzhen exchanges during the period from 2003 to 2012 uses the information asymmetry for representing the agency problem.Based on the results, they confirm that firms with higher information asymmetry shall pay lower dividends to their shareholders.But, they also give the evidence that relative to non state-controlled firms, the state-controlled firms with higher information asymmetry shall distribute higher dividends in order to satisfy the controlling shareholders.Based on these review, this study proposes the hypothesis 2 as follows: H2: More higher the firm long term debt, higher probability to pay dividend.In order to extend the result of analysis, this study explores the sample by dividing it into quartile based on the percentage ownership of controlling shareholders (CS).According to calculation, the study finds that the points for quartile 1 (Q1) is 58.15%, quartile 2 (Q2) is 75.13%, and quartile 3 (Q3) is 84.99%.After analyzing all the sample, this study compares the results by analyzing 75% of the sample (the area below point Q3) and 50% of sample (the area below point Q2).To confirm those results, this study then analyzes each area of sample, which was divided into 25% of sample or area below Q1, the area between Q1 and Q2, the area between Q2 and Q3, and area above Q3.

Variables
This study uses dividend policy as dependent variable and measures it with dummy, where 1 is for firms as dividend payers, and 0 is for firm as non-dividend payers.To be justified as dividend payers, then the firms must at least pay their dividend more than 0 in average during the observed period.The independent variables for this study are as follows: • • following Jensen (1986), this study uses the longterm debt ratio (LTDAR) to detect the conflict of interest or agency problem between insiders and outsiders.This study calculates LTDAR as ratio of long-term debt over total assets; • this study uses ownership as additional independent variable and also as amplifier for other independent variables.The ownership variables are categorized as follows: (1) controlling shareholders (CS); and (2) non-controlling shareholders (NCS).This study defines the controlling shareholders (CS) as percentage ownership of controlling shareholders.In order to avoid the redundant data of analysis, this study controls the measurement for non-controlling shareholders (NCS) variable.This study defines the non-controlling shareholders (NCS) as logarithm of difference of ownership percentage between controlling shareholders and non-controlling shareholders.

Regression model
This study conducts the logistic regression in order to test the hypotheses.To justify the results, this study uses the standard error (significance rate) at 0.05.The regression model for this study is as follows:  The study finds the same results when confirming 25% of the sample, area between Q1 and Q2, area between Q2 and Q3, and area above Q3.

Descriptive statistics
The mean on all sample of long-term debt ratio (LTDAR) shows that dividend payers have lower long-term debt relative to non-dividend payers, which indicates that these firms have possibilities to pay dividend, because they bear less debt interest expense.Other possibility of the firms is that they have tendencies to increase dividend when they face the internal conflict or reduce dividend payment in order to cover additional interest in condition of higher debt.This study finds similar results when analyzing 75% and 50% of sample and also when confirming those results on 25% of sample, area between Q1 and Q2, area between Q2 and Q3, and area above Q3.
Furthermore, the result of price to book value (PBV) on all sample and 75% of sample shows that firms as dividend payers have higher mean relative to firms as non-dividend payers, which indicates the possibilities to reduce dividend if they still have more investment requirements.This study has different result when analyzing 50% of sample.The mean shows that firms as dividend payers have lower price to book value (PBV) relative to firms as non-dividend payers, which indicates that firms tend to distribute less dividend in order to fund additional investments.Confirming each area, this study finds that 25% of sample or area below Q1 have similar characteristics when analyzing 50% of the sample.But, the area between Q1 and Q2, area between Q2 and Q3, and area above Q3 show similar characteristics as results on all sample and 75% of the sample.
The mean on all sample and 75% of the sample show that firms as dividend payers have lower Q3, the mean of percentage of controlling shareholders (CS) is lower for firms as dividend payers relative to firms as non-dividend payers.The results of non-controlling shareholders (NCS) are reverse of controlling shareholders (CS) for dividend payers relative to firms as non-dividend payers.

Mean difference test
Table 2 presents mean difference test between firms as dividend payers and firms as non-dividend payers to confirm the results of descriptive statistics.The result of retained earnings ratio (RETA) shows that dividend payers are more mature when analyzing all sample and 75% of the sample, but it is insignificant when this study analyzes 50% of the sample.Confirming those results, this study finds that retained earnings ratio (RETA) is significant on 25% of the sample or area below Q1, area between Q2 and Q3, and area above Q3, while area between Q1 and Q2 is insignificant.The results indicate that most of dividend payers are more mature and able to pay dividend than non-dividend payers, except the circumstance in area between Q1 and Q2.
The result of long-term debt ratio (LTDAR) shows that mean difference between dividend payers and non-dividend payers is insignificant only on 25% of the sample or area below Q1, while other results show significant difference.Commonly, the results of long-term debt ratio (LTDAR) confirm the descriptive statistics, which indicates that these firms have possibilities to increase or to decrease dividend when facing agency problem or to cover the interest expense.The result of price to book value ratio (PBV) shows that mean dif-ference is insignificant on all sample, 75% of the sample, and 50% of the sample.But when analyzing each area, the study finds that the mean differences of price to book value ratio (PBV) are significant.The results of controlling shareholders (CS) and non-controlling shareholders (NCS) show that mean difference is significant on 50% of sample, 25% of sample, and area between Q2 and Q3, which indicates that controlling shareholders (CS) on those areas plays the role in affecting the firm policy.Table 3 shows that the analysis on all sample, 75% of the sample, and 50% of the sample has similar results, which indicates that dividend policy on controlling shareholders (CS) and non-controlling shareholders (NCS) is different to each other.

Logistic regression test
The case for controlling shareholders (CS) shows that dividend policy on this shareholders has tendency to follow the setting of life cycle theory, as the result of PBV CS ⋅ have supported for it.This finding is consistent with Lang and Litzenberger (1989), Fama and French (2001), Grullon, Michaely, and Swaminathan (2002), and Fairchild, Guney, and Thanatawee (2014), which make the study accept H1.Whereas the result of LTDAR CS ⋅ shows that dividend policy on controlling shareholders (CS) is inconsistent with setting of free cash flow theory, which means that the study rejects H2.
Furthermore, the case for non-controlling shareholders (NCS) shows that dividend policy on these shareholders is inconsistent with the setting of life cycle theory as supported by the result of .

PBV NCS ⋅
Based on this result, the study then rejects H1 in case for the shareholders.The study assumes this condition is most possibly point of the consequence of non-controlling shareholders (NCS) when dividend has moved to controlling shareholders (CS), since descriptive statistics shows that controlling shareholders (CS) have larger ownership relative to non-controlling shareholders (NCS).Reversely, the result of LTDAR NCS ⋅ shows that dividend policy on this shareholder is consistent with the setting of free cash flow theory in case of agency problem, which means that the study accepts H2.This finding supports the studies by Easterbrook (1984), Jensen (1986), Agrawal and Jayaraman (1994), and Ince and Owers (2012) especially when non-controlling shareholders (NCS) considers debt as control mechanism on insiders.
Till this point, the study proves that dividend policy on controlling shareholders (CS) and non-controlling shareholders (NCS) is based on different reasons.This study provides the evidence that firms as dividend payers tend to distribute dividend for their controlling shareholders (CS) in case when the firms consider at mature level, while dividend policy on non-controlling shareholders (NCS) is most dominated by conflict of interest or agency problem.The study then continues the analysis on 25% of the sample or area below Q1, the area between Q1 and Q2, the area between Q2 and Q3, and area above Q3 in term to confirm the results on all sample, 75% of the sample, and 50% of the sample.

Reviews on 25% of the sample
First, this study normalizes the retained earnings ratio (RETA) and price to book value ratio (PBV) by Based on those results, the study then rejects H1 and H2.As additional, this study notes that circumstance of dividend policy in area between Q1 and Q2 possibly has its own complexity since dividend payers along this area have the lowest mean of retained earnings ratio (RETA) among dividend payers on other area.

Reviews on area between Q2 and Q3
As in previous procedures, the study normalizes the retained earnings ratio (RETA), long-term debt ratio (LTDAR), and price to book value ratio (PBV) by logarithm 10 and finds that goodness of fit (GOF) on model is favorable.This study finds similar results with the results on all sample, 75% of the sample, and 50% of the sample.On this area, the study accepts H1, while dividend policy for shareholders (CS) tends to follow the setting of life cycle theory, whereas dividend policy on non-controlling shareholders (NCS) is consistent with the setting of free cash flow theory in case of agency problem which makes this study accept H2.

Reviews on area above Q3
This study normalizes retained earnings ratio (RETA), long-term debt ratio (LTDAR), and price to book value ratio (PBV) by logarithm 10 and finds that goodness of fit (GOF) on model is favorable.The regression results show that only price to book value ratio moderated by controlling shareholders ( )

PBV CS ⋅
and non-controlling shareholders ( ) PBV NCS ⋅ is significant, although they have different signs.Those results indicate that dividend policy on controlling shareholders (CS) is inconsistent with life cycle theory, but consistent on non-controlling shareholders (NCS).In those cases, this study rejects H1 on controlling shareholders (CS) but accepts it on case for non-controlling shareholders (NCS).Whereas the result of long-term debt suggest the study rejects H2 for both shareholders.
Those results imply that circumstances of dividend policy on this area have its own uniqueness, because it seems that any investment opportu- nities as suggested by Fama and French (2001) now has become a consequence of controlling shareholders (CS).Under this circumstance, the case on controlling shareholders (CS) may follow the finding of Wei, Wu, Li, and Chen (2011) in assumption when this ownership normally has heterogeneous preferences.Under this assumption, controlling shareholders (CS) tend to "willingly" switch their dividend for non-controlling shareholders (NCS).In this case, dividend policy of non-controlling shareholders (NCS) could be an alternative explanation on free cash flow theory or agency problem.Note that the finding of the work of Bøhren, Josefsen, and Steen (2012) imply that distributing dividend for non-controlling shareholders (NCS) somehow is a strategy to mitigate the conflict of interest between both shareholders.Descriptive statistics in Table 1 shows that dividend payers in this area have the largest for controlling shareholders (CS) and the lowest for non-controlling shareholders (NCS) of 90.99% and 9.01%, respectively.

Robustness checks
The last step to analyze the model of dividend policy for this study is robustness checks.The study chooses the firm size as the additional variable to check the robustness of model and measures it by logarithm of total assets.Table 4 presents the results for the model after putting the firm size as the additional variable.The results show that after putting the firm size as additional variable, the results show that most of variables have similar signs as previous results.Based on these results, this study finds that the models for dividend policy on controlling shareholders (CS) and non-controlling shareholders (NCS) are robust.

CONCLUSION
This study provides some evidence relate to dividend policy on controlling shareholders and non-controlling shareholders.First, the evidence shows that firms as dividend payers tend not to distribute their dividend for controlling shareholders and non-controlling shareholders while the composition for both shareholders are almost equal.Second, the evidence shows that firms as dividend payers also have tendency not to distribute dividend on controlling shareholders when these shareholders have the largest Notes: This table reports the results of logistic regression for robustness checks on model of dividend policy on controlling shareholders and non-controlling shareholders.Dependent variable is dividend policy (Div), which was measured by dummy, where 1 for firm as dividend payers, and 0 for firm as non-dividend payers.RETA is ratio of retained earnings over total assets.LTDAR is ratio of long-term debt over total assets.PBV is ratio of share market price over share book value.Size is logarithm natural of total assets.Controlling shareholders (CS) is percentage ownership of controlling shareholders.Non-controlling shareholders (NCS) is logarithm of difference of ownership percentage between controlling shareholders and non-controlling shareholders.The figures of * indicates statistical significance at 0.05.
Investment Management and Financial Innovations, Volume 16, Issue 2, 2019 (2)p://dx.doi.org/10.21511/imfi.16(2).2019.28 Data for this study are drawn from Indonesia Stock Exchange (www.idx.co.id) during the period from 2010 to 2015.This study excludes the new listed firms or firms with incomplete data, which are needed for analysis.This study follows Denis and Osobov (2008), Wei, Wu, Li, and Chen (2011), and Fairchild, Guney, and Thanatawee (2014) for excluding financial firms and property, real estate, and construction firms.The final sample for this study is 241 listed firm with 1,446 of total observed data. 2.1.Sample

Table 1 .
Investment Management and Financial Innovations, Volume 16, Issue 2, 2019 Descriptive statistics This table reports descriptive statistics between payers and non-payers.RETA is ratio of retained earnings over total assets.LTDAR is ratio of long-term debt over total assets.PBV is ratio of share market price over share book value.Controlling shareholders (CS) is percentage ownership of controlling shareholders.Non-controlling shareholders (NCS) is logarithm of difference of ownership percentage between controlling shareholders and non-controlling shareholders.

Table 3
model.Whereas on 50% of the sample, the study normalizes only for long-term debt ratio (LTDAR) and price to book value ratio (PBV) by logarithm 10 to get better goodness of fit (GOF) on model.

Table 2 .
Mean difference test This table reports the mean difference test between payers and non-payers.RETA is ratio of retained earnings over total assets.LTDAR is ratio of long-term debt over total assets.PBV is ratio of share market price over share book value.Controlling shareholders (CS) is percentage ownership of controlling shareholders.Non-controlling shareholders (NCS) is logarithm of difference of ownership percentage between controlling shareholders and non-controlling shareholders.The figures of * indicates statistical significance at 0.05.

Table 3 .
Dividend policy on controlling and non-controlling shareholdersNotes: This table reports the results of logistic regression about dividend policy on controlling shareholders and non-controlling shareholders.Dependent variable is dividend policy (Div), which was measured by dummy, where 1 is for firm as dividend payers, and 0 for firm as non dividend payers.RETA is ratio of retained earnings over total assets.LTDAR is ratio of long-term debt over total assets.PBV is ratio of share market price over share book value.Controlling shareholders (CS) is percentage ownership of controlling shareholders.Non-controlling shareholders (NCS) is logarithm of difference of ownership percentage between controlling shareholders and non-controlling shareholders.The figures of * indicates statistical significance at 0.05.

Table 4 .
Robustness checks on model of dividend policy on controlling and non-controlling shareholders