University of Groningen
Faculty of Economics and Business
Master Thesis
The effects of dividend policy and ownership concentration on firm
value: The case of China
Abstract: This paper intends to find the relationship among dividends, ownership
concentration, their interactions and firm value in Chinese stock markets. 2376 listed and delisted firms from Shanghai and Shenzhen Stock Exchanges make up our sample. A panel data analysis for the period 2006-2015 is used. Results with interaction effects on Tobin’s q show significant positive effects of cash dividend amounts and ownership concentration which is consistent with the signaling theory and agency cost theory. We also find a significant negative interaction effect of cash dividend magnitude and ownership concentration which implies that higher cash dividend in highly concentrated firms may case extra damage on firm value. However, a robustness test based on the market to book value of equity shows that the findings cannot be corroborated. Therefore, further research on Chinese dividend policies and ownership concentration is needed.
JEL Classification: G35, G32
Keywords: Dividend, Firm Value, Ownership Concentration, China, Interaction
Effect
Author: Shurong Ren
1. Introduction
One of the central questions regarding corporate finance is how to maximize firm value and this has drawn more and more attention from scholars in recent decades. Considering the influence factors of firm value, it is widely believed that dividend policies and ownership structure may have effects on firm value, such as through mitigating or exacerbating the agency problems (Bebchunk and Hamdani, 2009).
In the early 1960s, Miller and Modigliani (1961) proposed the irrelevance of dividend view which indicates that dividend distribution has no effect on firm value. Their theory is based on the perfect market assumptions of information symmetry, no tax, no agency costs and rational investors. Since then scholars introduced imperfections to study the effect of dividends. They introduced a wide range of imperfections, such as the effects of tax, transaction costs, asymmetric information and agency costs (Farrar and Selwyn, 1967; Pettit, 1972; Jensen and Meckling, 1976). Besides, the conflict of interest within the companies is another factor that influences the performance of companies. Conflicts may exist among managers, minority shareholders and blockholders caused by different ownership structures. The relationship between ownership concentration and firm value is getting more and more important since the conflicts between different shareholders are arising. As a result, both dividends and ownership concentration can have effects on firm value.
Most of the research on dividends has concentrated on mature markets, such as the U.S. market (Grinblatt et al. 1984). Research focused on developing markets is relatively rare though more knowledge of these markets might be relevant. With its unique legal systems, institutional background, and its state-owned economy, China’s security market can be used as an interesting example regarding this issue. As one of the fastest developing economies in the world, China opened the stock markets in 1990. As a civil law country, Chinese listed firms used to have a low level of dividend distribution which supports the view of La Porta et al. (2000) that firms in civil law countries often pay fewer dividends.
Shareholder Rights in Public Shares was released by the China Security Regulatory Commission (CSRC). According to the regulations, three years of cash dividends are required if a firm wants to make a seasoned equity offering (Deng et al., 2015). As the regulations are widely implemented in 2006, the phenomenon of the low level of dividend payments has been reduced. Also, Chinese firms have their unique characteristics, such as large shares held by blockholders and state-owned shares, which also makes the Chinese listed firms worth studying.
Ownership concentration is another important mechanism of corporate governance and need to be taken into account in Chinese stock markets. It is well known that the effects of ownership concentration may be magnified in emerging markets because such markets lack the protection of minority shareholders’ interests. This phenomenon may be more severe in Chinese stock markets considering its high concentrated ownership (Anderson et al., 2011), inadequate legislation, fraud and corruption and shares owned by the state (Ma et al., 2010). In addition, majority shareholders may have more power than expected which enlarges the effects of ownership concentration as well.
Therefore, this paper attempts to deliver insights on China by addressing three research questions: (1) Do dividends affect Chinese firm value? (2) Does ownership concentration affect Chinese firm value? (3) Does the interaction between dividends and ownership concentration affect Chinese firm value?
when we include the interaction between dividends and ownership concentration, it seems that both dividend amount and ownership concentration have significantly positive effects while their interaction effect is significantly negative. These findings imply that dividends work as a to transfer information about the potential future earnings and lower risks of the firms and that they can be used to mitigate agency costs. In the meantime, the positive effects of ownership concentration indicate that its positive monitoring effects dominate the entrenchment effects which would harm the shareholders. The negative interaction effect, however, supports the idea that high cash dividends are used as a mean to entrench firms’ resources and to shift minority shareholders’ wealth to concentrated owners. However, these results are not relatively robust and this may be caused by the unpredictability of Chinese managers and investors.
The previous studies investigating the effect of dividend or ownership concentration on firm value in China focus on the period before the regulations were announced or during the period when the regulations were just applied (Deng et al., 2015; Anderson et al., 2011; Zeng, 2010). However, few studies have investigated the effect after these regulations have been fully implemented. Our paper is the first that examines the effects in the period from the start of the regulation till recent years. This is the first contribution to science that this paper makes. The second contribution of this study is the involvement of an interaction term between cash dividends and ownership concentration which examines the unique characteristics of the Chinese stock markets. Lastly, the results found in this paper can be regarded as a guideline for Chinese managers and investors concerning dividend policies and ownership concentration.
2. Literature review
In this section, we introduce theories on dividend, ownership concentration and the interaction between them. Then, studies aiming at the Chinese equity markets and its special characteristics are also discussed.
2.1 Dividend theories
2.1.1 Traditional theories
One of the central questions regarding corporate finance is whether dividend can influence firm value. The first relevant theory on dividend is the theory of Lintner (1956) who has interviewed managers from 28 firms about the most important determinants of firm’s dividend policies. The payout policies are found not to be only based on the demand of cash of financing projects in the short-term, but also on the assessment of managers of the sustainability of earnings in the long-term. Lintner indicated that managers are cautious about changing dividend distribution, they would only initiate or increase dividend when they believe that the future earnings will reach and persist at a certain level. Therefore managers try to maintain a long-term target payout ratio corresponding to their profit level.
In the early 1960s, Miller and Modigliani (1961) showed that under the perfect capital market assumption (perfect with no information asymmetry, no taxes and agency costs but rational investors), dividend policy is irrelevant. Although this famous Miller-Modigliani theorem is not relevant in practice since imperfections exist in the real world market, it has become the starting point of modern dividend policy theories. However, as the real market does not satisfy the perfect capital market assumptions, dividends relevant theories have been proposed considering the imperfection of real world.
the losses caused by selling shares in case of a high need for cash. Therefore, firms have to pay high dividends to maximize the share price which will lead to the increasing of firm value.
In the perspective of the possible influence of taxes contrary to the perfect market assumptions of the Miller-Modigliani theorem, Farrar and Selwyn (1967) have proposed the tax preference theory which advocates that investors prefer a low level or no dividend based on the situation that tax on dividends is higher than on capital gains. Additionally, investors can control the time when capital gains are realized since it is not paid until the investment is sold. By contrast, the time of dividend payments is controlled by the managers. Lee (2005) has done a comprehensive study to analysis the effect of tax and found the highly taxed investors tend to hold stocks with low or zero dividends in line with the tax-effect hypothesis.
2.1.2 Modern theories
2.1.2.1 Signaling Hypothesis
Considering that information asymmetry exists in the real stock markets which is against the assumption of Miller-Modigliani theorem, Pettit (1972) has been the first to present the market reaction to the informational content of dividends as a signal for the expected earnings of the firm and by linking the changes in dividends to information on long-term cash flows. Ross (1977) has extended the asymmetric information theory into the analysis of dividend policy by illustrating that outsider investors who lack internal information on e.g. investment risk and expected revenue, can only use information that is transferred by dividend policies to assess firm performance and risks.
signal the inside information they possess about the future prospects of the firms. Benartzi et al. (1997) try to examine the application of signaling theory using firms listed in the U.S. markets as a sample to investigate whether dividend changes have information content about future performance. They found strong lagged and contemporaneous correlation between dividend changes and earnings. Focusing on the information of risks that may be signaled by dividends, scholars find that announcement and change in dividend may convey information about the risks in the future. It is indicated that lower dividends contain the information of lower risks in the future which may cause a lower cost of capital and then result in a better firm performance (Dyl and Weigand, 1998; Grullon et al., 2002).
Kaestner and Liu (1998) examine the information content of dividend announcements. They find a significantly positive relationship between the average stock price and the amount of dividend payment. They further infer that the capital market regards dividend payment as an important source of information about firm’s performance.
2.1.2.2 Agency costs hypothesis
The Miller-Modigliani theorem is based on perfect markets, which includes the assumption that no conflict exists between shareholders and managers, which means that managers will work to maximize shareholder’s value. However, this assumption may not always be true in the real world due to the separation of enterprise’s management and ownership leading to the agent-principal problem (Jensen and Meckling, 1976). Corporations beyond a certain size, in general, cannot be financed by a few owner-managers, therefore these firms are owned by multiple shareholders and operated by managers. Boards are supposed to select and monitor the managers to make sure the decisions made by the managers are in line with shareholders’ interest. However monitoring managers and verifying whether their decisions are good in the long run are difficult when managers have more inside information. Then agency costs may arise as the decisions are made by managers while most of the outcomes are received by shareholders (Lease et al., 1999).
shareholders will reduce any profits gained from ownership. However, if the shareholders work as a group, they can eliminate part of the agency costs which is larger than the costs of monitoring they have to incur. Easterbrook holds the view that dividend payments force the managers to raise funds more frequently in financial markets instead of relying on internal funds. In this way, the managers are forced to be exposed to the scrutiny of outside professionals, such as regulatory authorities, new shareholders, security institutions, investment bankers, lawyers and public accountants. Unlike individual investors, these institutions play a stronger role in monitoring managers as their ability of monitoring is tied to their reputation. As a result, being monitored by outside professionals is more efficient in reducing managers’ self-interested behavior with lower monitoring cost, and it helps in improving shareholder’s returns and firm value.
In line with Easterbrook’s point of view, Jensen (1986) puts forward the free cash flow hypothesis which presents the risk of letting managers control a large amount of free cash flow and regard cash dividend as a tool to reduce the disposable cash of managers. If excess cash (free cash flow) is under the control of managers who cannot be supervised, they will serve their own interests rather than the shareholders’ interests by using cash in ways that deviate from the interests of shareholders, such as value-harming investment projects. Therefore, the best way to serve shareholders’ interests is to minimize the cash needed for investment by using dividend to extract surplus cash from managers’ control.
European countries and Asian countries, but this effect showed more significant in Europe as countries there have the stronger legal protection.
2.1.3 Chinese dividend policy
Dividend policies are important channels to distribute corporate earnings to investors. However, different from firms in developed countries, fewer firms choose to distribute dividends in developing countries (Reddy and Rath, 2005). China as the largest economic entity in developing countries started its stock market in the early nineties as the first national stock exchange, the Shanghai Stock Exchange was founded in December 1990 (Li et al.,.2014). According to Chen et al. (2009), the Chinese listed companies’ averaged dividend payout ratio is 19.98% in the period 1990-2004, while the median is 0% in the same period. This payout ratio was among the lowest in the world. Furthermore, around 53% of the listed companies even distribute no dividend at all. Some Chinese listed firms even seldom paid any dividend to shareholders in a period longer than ten years since they went public. These firms are called “Grandet1” by Chinese media for their being unfair to their
shareholders. For instance, a famous listed firm from Shenzhen Stock Exchange, Wuliangye Spirit, which earned over 2 billion Yuan in 2008 with a growth rate of 10.14% from 2007, only paid a 0.1 yuan dividend per share for over ten years (Tao et al., 2016).
To ensure that companies distribute returns to their investors, many developing countries, such as Chile, Turkey, Venezuela, Brazil and other countries use legislated policies to force firms to distribute dividends. In other words, mandatory dividend payments. For instance, Chile stipulated that firms have to distribute no less than 30% of their profits as dividend. The Turkish government asked for half of the firms’ profit as cash dividend to be given to shareholders (Adaoglu,2000). The biggest benefit of this mandatory dividend policy is it supplies guarantee of the level of cash dividend payment (La Porta et al.,2000). However, this mandatory dividend policy remains controversial in the academic community. Some scholars put forward that mandatory dividend policy cannot reach an ideal result since the level of dividend distribution
1
didn’t increase significantly when Turkish government installed this policy in 2003 (Adaoglu,2000). Also, mandatory dividend policy can intervene in companies’ decision-making which is the number one reason why few countries adopt this policy (Martins and Novaes, 2012).
China also intended to solve the problem of its imperfect capital markets, imperfect shareholder protection mechanisms, and the opaqueness and lack of constraints of listed firm dividend policies by requiring dividend payouts (Chen et al., 2015). However, unlike other developing countries, the Chinese government chose a semi-mandatory dividend policy, which regards dividend distributions as requirements of gaining qualifications of refinancing to give incentives to firms to pay cash dividends. Instead of obligatory, this semi-mandatory dividend policy is more like a soft constraint which imposes payouts on firms with refinancing needs. To regulate the distribution of cash dividend, the China Security Regulatory Commission (CSRC) issued a series of regulations to support the listed firms to pay more cash dividend and improve information transparency. These series of policies started from 2001 when the China Security Regulatory Commission (CSRC) started to require the listed firm which did not pay any dividends in the last three years to give reasonable explanations for it. In 2004, the Regulations Protecting Shareholder Rights in Public Shares was issued that publicly listed firms that did not make any cash dividend payments in the last three years are forbidden to apply for seasoned equity offerings to the public, which connects cash dividends to the opportunities of refinancing. The following policies further detailed the minimum requirement of cash dividends used as securities issuance. According to the regulation pronounced in 2006, the cash or stock dividends should be more than 20% of the average annual disposable profits of the last three years. Then in 2008, this required ratio of cash dividend became 30%. These policies had remained in force until 2013 when the CSRC published The Third Regulatory Guidelines for Listed Firms On Cash Dividend Payments. This regulation requires all publicly listed firms in China to distribute cash dividends in the amount of more than 20 percent of the firms’ earning each year (Deng et al., 2015).
Figure 1 Time trend combo of Dividends and Dividend payments
Note: The figure is based on data of Chinese listed firms observed for 11 consecutive years from 2006 to 2015. The data are retrieved from WIND database (see Appendix 1). It shows the time trend curve of percentages of firms paying dividends, paying cash dividends and paying stock dividends2. The total amount of cash dividends is also
presented by the area diagram.
Figure 1 shows the time trend of Chinese publicly listed firms (from the Shenzhen and Shanghai stock exchanges) that pay dividends, pay cash dividends and pay stock dividends for the years 2006-2015. Although there many kinds of dividends distributed by Chinese listed firms (including allotment, stock repurchase and others), cash dividends and stock dividends play the dominate roles in the Chinese stock markets. As can be seen from the figure, from 2005 to 2015, the proportion of firms that pay cash dividends kept increasing and reached the peak in 2013. This trend shows strong evidence that the policies pronounced by the China Security Regulatory Commission (CSRC) achieve the design target of encouraging listed firms to distribute cash dividends more effectively. The result shows that the situations of low-level cash dividends in Chinese stock markets and low level of protection for minority shareholders are improved. Comparing the trends of dividends, cash
2
There are several other type of dividends exist in Chinese stock markets which takes up the rest part of firms paying dividends. This data is not showed in the figure because it cannot be retrieved.
100 200 300 400 500 600 700 800 900 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 M ag n itu d e b y B illi o n Yu an
Chinese Dividends
dividends and stock dividends, it is obvious that most of firms that pay dividends choose cash dividends required by regulations. Because the distribution of stock dividends is not constrained, stock dividends account for a small part of firms which pay dividends (lower than 10% in the research period) and this small proportion maintained in the whole period without major overall changes.
2.2 Ownership concentration
2.2.1 Ownership concentration theory
As an important mechanism of corporate governance, ownership structure has been the focus of the study by scholars (Connelly et al.,2010. Many empirical studies showed that in both developing and developed countries, a majority of ownership is concentrated in the hands of few shareholders (Barca and Becht, 2001; Maury and Pajuste, 2005). Berle and Means (1932) was the first that study in how ownership concentration may influence firm value and pointed out that the diffuse ownership can create the free-riding problem because minority shareholders have difficulty monitoring managers. As a result, a negative relationship between diffuse ownership and firm value was found as conflicts between the interests of managers and shareholders exist and managers may not work with the aim of maximizing shareholders’ value. Berle and Means regard ownership concentration as a disciplining device to mitigate such agency problems.
but no one will share the cost of monitoring with the largest blockholder. Hence, managers have more discretion if the ownership is not concentrated (Hart, 2001). Hyung (2011) stated that the inefficient monitoring will allow managers to use up inside cash for their own gains and fail to achieve the aim of maximizing firm value. Because managers do not need to pay full costs of the decisions they make, the possibility that managers try to choose projects that maximize their private benefits increases and this may not necessarily be in line with shareholders’ interests. Pedersen and Thomsen (1999) did research on 435 firms from 12 countries in Europe and presented that ownership concentration has a significant positive effect on return on equity. Claessens, Djankoy and Lang (2000) also came to the same conclusion based on a study of firms from East Asia. This is the main reason why the optimal ownership structure should be the one with at least one active dominant shareholder especially in countries with weak institutional systems and limited contract enforcement. This active blockholder will ensure active supervision of the managers (Slovin and Sushka, 1993).
diversification of stock which increases the idiosyncratic risk and also poses a negative effect on firm value (Maug 1998).
In addition, a non-linear relation between ownership concentration and firm value has come up based on the circumstance that large shareholders play different roles for different levels of concentration. In other words, there is a combination of positive and negative effects because of the improved monitoring of managers and blockholders expropriation of minority shareholders (Iturriaga and Crisóstomo, 2010). This point of view is suggested by Morck et al. (1988) who found an up-down-up trend relationship, pointing to an inverted U-shape relation. This special type of relationship depends on the trade-off between incentives and entrenchment related to controlling shareholders (Selarka, 2005; Balsmeier and Czarnitzki, 2015).
2.2.2 The case of China
Compared with developed countries, developing countries are quite different in governance structures. As a result, existing models may not explain behavior in emerging countries well, and there are few papers that focus on the influence of governance system on firm value in emerging markets (Bekaert and Harvey, 2002). It is well known that ownership structure, especially ownership concentration plays an important role as a governance device in emerging markets since such markets often lack strong legal protection of minority shareholders. Ownership concentration in markets of developing countries has persisted through years of privatization and economic reform (Claessens et al., 2000). For instance, for China’s stock market, which is famous for its inadequate legislation, fraud and corruption, highly concentrated ownership is not only a result of economic reform but also a necessary strategy to protect investors (Ma et al., 2010). Stocks are divided into tradable shares and non-tradable shares in China’s stock market, most of the latter ones are owned by state or legal persons who cannot be traded freely on secondary markets (Hovey et al., 2003). This unique structure leads to highly concentrated firms and distorted markets. In order to change this situation, the Chinese government has carried out the so-called “split-share reform3” to transfer non-tradable shares to tradable shares continuously
3
(Liu et al., 2014). Even so, the level of ownership concentration in China is still higher than the international average level. This highly concentrated ownership structure is called “one stockholder, biggest shareholder4” which presents that the
blockholders own extremely strong power inside the companies. According to Anderson et al. (2011), the largest shareholders owns 42.65% of the average of the listed firms in China during 2000-2008. Shares owned by the state is another characteristic of China’s stock market. The state shareholders represent the Chinese central or local governments and are sometimes the controlling shareholders in public firms (Sun and Tong, 2003; Wei et al., 2004). This unique type of shareholder makes controlling shareholders not to be motivated by private gains, but by political and government wishes (Ma et al., 2010). Since part of the control shareholders of Chinese listed firms are non-tradable shareholders and state-owned shareholders, the effects of ownership concentration may be no in line with or smaller than effects proposed by international literature. As a result, ownership concentration may be less effective in China regarding the special background and the rule of law (Ma et al., 2010).
Quite some studies have been attempted to discover the relationship between ownership concentration and firm performance based on the Chinese market. Three views result : a) ownership concentration has positive effects on firm performance as the scale of biggest shareholdings positively affect firm performance in non-protective industry (Chen and Xu, 2001; Lipinga et al., 2006); b) negative effects exist between concentrated ownership and firm performance (Gunasekarage et al., 2007); c) ownership concentration and firm performance are non-linear relations, such as an inverted U-shape relationship (Li and Fang, 2006).
2.3 Interaction between dividend and ownership concentration
In principle, the ownership concentration has a negative effect on dividend distribution. The excess private benefits created from controlling may lead
4
blockholders to be not interested in a high level of dividend distribution. They could influence the managers to make financial decisions of low level of dividend distribution in order to consume private benefits. As a result, large shareholders may have the tendency to retain earnings which can create private value for themselves instead of dividends (Tang et al., 2007). Rozeff (1982) pointed out that the more dispersed the ownership is, the higher the percentage of dividend payment will be. Gugler and Yurtoglu (2003) also found that the biggest shareholders prefer to reduce the dividend payment in Germany. However, this view does not fit the situation of China as the dividend distribution is connected with refinancing qualification and seasoned equity offering by regulation is imposed by the Chinese government. Therefore, provisions of law and supervisions from the government become the most important drivers of dividend distribution. Under these circumstances, instead of considering how ownership concentration may influence the decision of dividend, the interaction effect of ownership concentration and dividend distribution is perhaps as important since the special ownership structure features for Chinese listed firms can lead to extra effects on firm value of dividend policies (Tang et al., 2007).
support since the weak institutional environment, and the lack of minority shareholders protection mechanisms exacerbates the situation of expropriation. Minority shareholders may not be that eager to receive cash dividends since controlling shareholders will receive the majority of cash dividends and small shareholders are also subject to the income tax (Guo, 2016). This entrenchment phenomenon in Chinese stock market is mostly accomplished through cash dividend distribution because the regulations proposed by the Chinese government have proposed requirements both on the distribution and scale of cash dividends which gives guarantee for the opportunities of entrenchment behavior. In this case, cash dividends have became the optimal choice for removing resources from firms and expropriate minority shareholders’ wealth.
2.4 Hypotheses
In line with the aforementioned literature, hypotheses are proposed in this section.
2.4.1 Dividend and firm value
stock markets. Considering the theories mentioned above and dividing dividends into dividend payment and dividend amounts, the first hypothesis can be developed as follows:
H10: Cash dividends have no effect on firm value.
H1A: Cash dividends have a positive effect on firm value.
H1Aa: The paying of cash dividends is positively related to firm value. H1Ab: The amount of cash dividends is positively related to firm value.
2.4.2 Ownership concentration and firm value
The second aim is to discuss how ownership concentration may influence firm value. Considering ownership concentration, scholars hold two different views that ownership concentration can mitigate conflicts between shareholders and managers, and that ownership concentration can exacerbate conflicts between small shareholders and controlling shareholders (Jensen and Meckling, 1976; Shleifer and Vishny, 1986; Bennedsen and Wolfenzon, 2000; Balsmeier and Czarnitzki, 2015). The true result may be the mix of both views which indicates an absence or a nonlinear relationship. The absence of relationship happens when the effects cancel out. An up-down-up trend relationship may, however, also exist between ownership concentration and firm value which points to an inverted U-shape relation caused by the combination of first positive and later negative effects (Morck et al., 1988). However, we examine the linear relationship because the true relationship may depend on the which side of the inverted U-shaped that different countries or firms within countries find their position5. We assume that the effect of monitoring overwhelms the effect of entrenchment by blockholders. Therefore the second hypothesis can be developed as follows:
H20: Ownership concentration has no effect on firm value
H2A: Ownership concentration has a positive effect to firm value.
5
2.4.3 Interaction effect between dividend and ownership concentration
The last hypothesis is to examine how the interaction between dividend and ownership concentration will effect firm value. According to the agency costs theory, conflicts also exist between minority shareholders and controlling shareholders, which can result in the entrenchment and expropriation problems. Blockholders may use their influential power on cash dividend policy to expropriate minority shareholders’ wealth (Gugler&Yurtoglu, 2003; Cheung et al., 2006). This phenomenon may be more widespread in China because of its semi-mandatory dividend distribution policy and the weak legal protection for minority shareholders. The last hypothesis can be developed as follows:
H30: Dividends and ownership concentration have no interaction effect on firm value. H3A: Dividends and ownership concentration have a negative interaction effect on firm value.
H3Aa: Cash dividend announcements and ownership concentration have a negative interaction effect on firm value.
H3Ab: The amount of cash dividends and ownership concentration have a negative interaction effect on firm value.
3. Empirical design
Based on the empirical methodology developed by Iturriaga and Crisóstomo (2010), we build on the long-term study method to study on the relationship between firm value and these items.
3.1 Variables
This section introduces the variables used in this thesis. The symbol and definition of all variables are shown in Appendix 2.
3.1.1 Dependent Variables
Tobin’s Q was first proposed by Kaldor (1966) and reintroduced by Tobin and Brainard (1968). Tobin’s q is defined as the ratio of equity market value to the replacement value of firm’s assets. A higher Tobin’s Q means that investors are more optimistic about this company’s future development. Therefore, Tobin’s Q can represent market value premium and offers an attractive alternative measure of firm value (Stevens and Jose, 1992). Our study uses Tobin’s q as a proxy for firm value as many previous studies did (Cho, 1998; Land and Stulz, 1994; Morck et al., 1988). Since there is no clear definition of the measurement of replacement value of firm’s assets, scholars measure Tobin’s q in different ways. In this thesis, we use the measure in line with Setia-Atmaja (2009) which uses the ratio of the sum of market value of equity and the book value of all liabilities to total assets.
3.1.2 Independent Variables
We also use a dummy variable to represent the announcement of cash dividends. This dummy variable is used as a proxy of the cash dividend payment event, which equals to 1 when cash dividends are announced among the research period and equals to zero otherwise. Since the announcement of dividends (or changes in dividends) convey information to the stock market, this dummy variable is used by many scholars to examine whether the cash dividend announcement can influence firm value (Ehikioya,2015; Guo, 2016; Pettit, 1972; Aharony and Swary, 1980).
Considering the variable of ownership concentration, two kinds of measurement are mostly used in studies. Some scholars use the Herfindahl Index, which adds up the squares of shareholding ratios of first five shareholders to represent ownership concentration (Maury and Pajuste, 2005; Li et al., 2008). The top three of top five largest shareholders are mostly used to calculate the Herfindahl Index. A higher value of this index relates to a higher level of ownership concentration. Some other scholars (see e.g. Erickson et al.,2005) use the percentage of ownership or sum of percent ownership of the largest shareholder, the five largest shareholders or the ten largest shareholders. Some scholars even use the sum of the percentage ownership of all shareholders who hold shares more then a certain threshold (Khan, 2006; Denis et al., 1997; Balsmeier and Czarnitzki, 2015). In this paper, we use the sum of percent ownership that is owned by the five largest shareholders in line with the study of Prommin et al., (2016).
3.1.3 Control Variables
state-ownership into account. The following is the description of the seven control variables used in this paper.
Many previous studies suggest that firm size has a positive influence on firm value since larger firms are more capable of raising funds in financial markets, create competitive advantages and distribute dividends (Ben-Nasr, 2015). The natural logarithm of total assets is used to represent the firm size variable as many studies do see e.g. Ehikioya (2015). Fama and French (2001) used market-to-book ratio to represent firms’ growth. In this thesis, we cannot use the market to book value as it is closely (if not exactly equal to) the dependent variable, therefore we use the change of total assets throughout the whole year over the value at the beginning of the year to represent firm (asset) growth. Leverage is another control that needs to be considered since ratio of debt to assets can force managers to maximize firm’s profit because meeting the required debt payment helps to alleviate agency problems (Jensen, 1986). Following Ehikioya (2015), we control for leverage using total debt divided by total assets. It is reasonable to expect that earning ability has a positive effect on firm value (Beiner et al., 2004). The earning ability can be measured by return on assets (Setia-Atmaja, 2009). However, in this paper, we use the net income before extra/preferred dividends which is in line with Ben-Nasr (2015). We also control for cash holdings and interest expenses using the ratio of Cash (including cash equivalents) to total assets and the ratio of interest expense to total debt respectively.
3.2 Sample, Data and Summary Statistics
In this study, we use a panel sample of 2376 listed Chinese firms from Shanghai and Shenzhen Stock Exchanges in a period between 2006 and 2015. As the regulations proposed by the China Security Regulatory Commission (CSRC) started at 2004 and obtained remarkable achievements at 2005, we choose the sample starting from 2006 because we want to be able to use variables one year lagged to control for endogeneity caused by simultaneously. In this way, we examine the period after the reform on dividend and eliminate the effects caused by the reform in the early implementation. We obtain firm-level data from Datastream (equity market data) and WIND database (shareholder data). The initial sample consists of all Chinese listed firms and the delisted firms from Shanghai and Shenzhen stock Exchange between 2006 and 2015. To screen the sample objectives, first we exclude firms without ISIN (International Securities Identification Number) or SIC (Standard Industrial Classification) code. Then, we exclude financial firms (Standard Industrial Classification (SIC) code from 6000 to 6999) and firms from utility industry (Standard Industrial Classification (SIC) code from 4900 to 4999). These types of firms are subject to regulatory supervision from the whole sample as well (Opler et al., 1999). The firms that delisted in this period are included in the sample to avoid survival bias. The Tobin’s q is calculated as the sum of market capitalization (Worldscope item 08001) and total debt (Worldscope item 03255) divided by total assets (Worldscope item 02999). The magnitude of dividend is calculated as the product of dividend per share (Datastream item DPS) and the number of shares (Datastream item NOSH) divided by total assets (Worldscope item 02999). Size is defined as the natural logarithm of total assets (Worldscope item 02999). Growth is calculated as the change in total assets throughout the whole year, total assetsi,t – total
assetsi,t-1, over the value at the beginning of the year, total assetsi,t-1. Leverage is
the potential bias due to outliers. After all the screening process, our sample contains 2376 firms.
3.3 Regressions
We define multivariate regression models to investigate the influence of dividends and ownership concentration on firm value. Moreover, we use two variables to represent the two aspects of cash dividend in two regressions. We use a dummy variable, which represents whether firm distributes cash dividend in the underlying year, and the amount of cash dividend to examine the effect of dividend.
Using panel data allows us to control the unexpected effects of any unobserved heterogeneity (Arellano, 2003) and the fixed effects. Cross-section fixed effects and time fixed effects are both used in all regressions indicated by the Hausman Test to eliminate the effect of unobserved firm factors and time factors which allow intercepts to differ across different firms and different years. Moreover, cash dividends, ownership concentration and other control variables can also be influenced by firm value which leads to the endogeneity problem (Cho, 1998). To solve this problem, we use lagged independent and control variables in all regressions as instrumental variables. We build on long-term event study method to measure the effect of dividends and ownership concentration based on Iturriaga and Crisóstomo (2010):
Regression 1: t i t i t i t i t i t i t i t i t i t i D H S G C L I E D TQ, 01 1,12 ,13 ,14 ,15 ,16 ,17 ,18 ,19 2. Regression 2: t i t i t i t i t i t i t i t i t i t i DIV H S G C L I E D TQ,
0
1 ,1
2 ,1
3 ,1
4 ,1
5 ,1
6 ,1
7 ,1
8 ,1
9 2
.We use D1i,t-1 equals to 1 for firms i that distribute cash dividends at year t-1,
otherwise D1i,t-1 will equal to 0; Divi,t-1 represents the amount of cash dividends
distributed by firm i at year t-1; Hi,t-1 is the ratio held by the first five shareholders in
the total shares of company i at year t-1; Si,t-1 is the natural logarithm of total assets of
the value at the beginning of t-1 year; Ci,t-1 is the ratio of cash and cash equivalent to
total assets of firm i at year t-1; Li,t-1 indicates the ratio of total debt to total assets of
firm i at year t-1; Ii,t-1 represents the interest expense of company i at year t-1; Ei,t-1 is
the ratio of net income before extra/preferred dividends to total debt of firm i at year t-1; D2i,t-1 equals to 1 when the firm i is state-owned at year t-1, otherwise D2i,t-1 will
equal to 0. Therefore, the first hypothesis is tested by the sign of coefficients α1 in
both of two regressions (H1Aa and H1Ab respectively), and the second hypothesis is
tested by the sign of α2.
In order to test the interaction effects of cash dividends and ownership concentration, we involve the interaction variables into the two regressions to develop the third and fourth regressions as follows:
Regression 3: t i t i t i t i t i t i t i t i t i t i t i D H HD S G C L I E D TQ, 01 1,12 ,13 1,14 ,15 ,16 ,17 ,18 ,19 ,110 2. Regression 4: t i t i t i t i t i t i t i t i t i t i t i Div H HDIV S G C L I E D TQ, 01 ,12 ,13 ,14 ,15 ,16 ,17 ,18 ,19 ,110 2.
Where HD1 represents the interaction term between dividend announcement dummy and the variable of ownership concentration and HDIV is the interaction between the amount of cash dividends and the variable of ownership concentration. Therefore, the third hypothesis is tested by the direction of coefficients α3 in both regressions.
4. Empirical results
The descriptive statistics and the results of tests before running the regressions will be first displayed in this section. Then, we provide the results of regressions and the discussion of the results. The effects of cash dividend and ownership concentration on firm value will be examined in Section 5.2.1. The interaction terms between cash dividend and ownership concentration will be involved in Section 5.2.2. Finally, we discuss the robustness of the results in Section 5.3.
4.1 Descriptive statistics
Table 1 shows the descriptive statistics (after winsorizing at 5% and 95% tails) of the variables for the sample of Chinese listed firms. Table 1 shows that Tobin’s q has a mean and standard deviation of 2.408 and 1.565 respectively. The mean and standard deviation of cash dividend amount are 0.006 and 0.0261 respectively. In addition, ownership concentration has a mean and standard deviation of 53.024 and 14.931 respectively which indicates a large diversity. This can be seen from the values that range from 26.640 to 77.970 which means that highly concentrated ownership exists in Chinese firms. The mean and standard deviation of the interaction term between ownership concentration and dividend are 0.002 and 0.052 respectively, and it takes value from 0 to 129.300 illustrating that there is a large diversity.
Note: Table 1 contains summary statistics (number of observations (N), mean, standard deviation (Std.Dev), minimum and maximum) of the sample used in this paper from 2006 to 2015. All the variables were winsorized at 5% and 95% tails. TQ is the ratio of the market value of equity to book value of total assets; Div the ratio of the amount of dividend (which is dividend per share multiplied by the number of shares) to total assets; H represents ownership concentration which is the sum of the proportions of the biggest five shareholders; HD1 is the product of H and D1 ; HDIV is the product of H and DIV ; S is firm size measured as natural logarithm of total assets; G represents firm growth which is the change
Table 1. Descriptive Statistics for the 2006-2015 Sample Period
Variable N Mean Std. Dev Minimum Maximum
TQ 18498 2.408 1.566 0.769 6.533 DIV 17423 0.006 0.026 0.000 3.085 H 19253 53.024 14.931 26.640 77.970 HD1 17414 30.207 28.367 0.000 77.970 HDIV 17414 0.002 0.052 0.000 129.300 S 20659 6.314 0.507 5.389 7.304 G 19658 0.222 0.341 -0.148 1.258 C 20716 0.202 0.142 0.032 0.547 L 20720 0.221 0.172 0.000 0.557 I 18129 0.093 1.244 0.000 137.960 E 20717 0. 047 0.053 -0.056 0.170 i,t i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 of total assets throughout the whole year over the value at the beginning of the year; C is cash holdings which is defined as cash and cash equivalents divided by total assets; L is measured as total debt divided by total assets; I is defined as interest expense divided by total debt; E is earning net income before extra/preferred dividends divided by total assets.
i,t i,t-1
i,t-1
i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1
i,t-1 i,t-1
Appendix 3 summaries the results of the stationary test. Using three common used test: LLC test (Common root- Levin, Lin, Chu), IPS test (Individual root- Im, Pesaran, Shin) and ADF test (Individual root- Fisher- ADF) to test on the whole panel data, the results show that the data series are generally stable (according to two out of the three tests), and we therefore conclude that they are suitable for panel analysis.
Appendix 4 summaries the correlations for all variables in the panel data. The correlation between Tobin’s q and dividend (cash dividend magnitude) is positive and significant. Furthermore, except for the coefficients of correlations between dividend announcement dummy and the interaction term (between dividend announcement dummy and ownership concentration), dividend amount and the interaction term (between dividend amount and ownership concentration) are larger than 0.5, all the other correlation coefficients are smaller than 0.5. We therefore conclude that there is no strong evidence for multicollinearity.
4.2 Results
4.2.1 Dividend and ownership concentration
The first hypothesis is developed to test the effect of cash dividend on firm value based on two different aspects of cash dividend announcement and cash dividend amount respectively. Hypothesis H1Aa which illustrates the effect of cash dividend
announcement is tested by model 1 in Column 1 in Table 2. The coefficient of cash
Note: This table presents results for cash dividends and ownership concentration in 2006-2015. All the variables are winsorized at 5% and 95%. TQ is the ratio of the market value of equity to book value of total assets; D1 equals to 1 when the firm is state-owned and equals to 0 otherwise; Div the ratio of af the amount of dividend (which is dividend per share multiplied by the number of shares) to total assets; H represents ownership concentration which is the sum of the proportions of the biggest five shareholders; S is firm size measured as natural logarithm of total assets; G represents firm growth which is the change of total assets throughout the whole year over the value at the beginning of the year;
Table 2. Regression Results for Cash Dividends and Ownership Concentration
Model 1 Model 2 Variables TQ TQ D1 0.029 [1.202] Div 0.231 [0.831] H 0.003* 0.003* [1.900] [1.823] S -2.159*** -2.152*** [-35.252] [-35.354] G 0.273*** 0.274*** [6.929] [6.970] C -0.358*** -0.368*** [-3.1248] [-3.219] L 0.195* 0.196* [1.744] [1.748] I -0.004 -0.004 [-0.580] [-0.580] E 2.294*** 2.293*** [7.815] [7.809] D2 0.009 0.01 [0.146] [0.164] Constant 15.930*** 15.907*** [43.384] [43.398]
Cross-fixed effect Yes Yes
Time-fixed effect Yes Yes
Adjusted R 0.723 0.723 F-statistic 14.709*** 14.707*** Observations 10803 10803 i,t i,t i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 2
C is cash holdings which is defined as cash and cash equivalents divided by total assets; L is measured as total debt divided by total assets; I is defined as interest expense divided by total debt; E is earning net income before extra/preferred dividends divided by total assets; D2 equals to 1 when firm i is state-owned in year t-1 and equalts to 0 otherwise. Standard errors are in parentheses. *, ** and *** indicate statistical significant level at 10%, 5% and 1% respectively
dividend distribution is 0.029 and it is not significant if we stick to a 5% significance level which means no significant relation can be found between cash dividend announcement and firm value based on our sample. This result refutes hypothesis H1Aa, which assumes that a positive relationship exists supported by signaling theory
and agency cost theory, but the findings cannot reject the null hypothesis H10. This
result of no significant relationship may be caused by the unpredictable of Chinese firms’ managers. On one hand, the unpredictability of Chinese managers may be caused by the opaque information of the managers’ decisions. On the other hand, it may also be caused by the situation of Chinese firms that are partly owned by the state. State-owned shareholders may consider other issues prior to the interests of the remaining shareholders and may have a strong influence on the managers’ financial decisions and this may also make the managers unpredictable (Chen et al., 2004).
The coefficients of firm size, growth, cash holding, leverage, interest expense, earning and state-owned dummy are -2.159, 0.273, -0.358, 0.195, -0.004, 2.294 and 0.009 respectively. The coefficients of firm size, growth, cash holding and earning are significant at 1% significant level, while the coefficient of leverage is marginally significant at 10% significant level. The effects of interest expense and state-owned dummy variable are insignificant for Tobin’s q if we stick to a 5% significance level. These results indicate that a decrease in firm size and cash holding can lead to higher firm value, while higher growth, leverage and earnings may cause an increase in firm value. We used cross-fixed effects and time-fixed effects, and the goodness of fit (R2) is 72.3%.
Hypothesis H1Ab that illustrates the effect of the amount of cash dividend is tested by
the second model in Column 2 in Table 2. The coefficient of cash dividend amount is 0.231 and it is not significant at 5% significance level which means no significant relation can be found between the amount of cash dividend and firm value based on our sample. This result refutes hypothesis H1Ab which assumes that the amount of
cash dividend will have a positive effect on firm value in line with the signaling theory and agency cost theory (Pettit, 1972; Easterbrook, 1984; Jensen 1986) but it cannot reject the null hypothesis H20. The absence of a relationship is the same as for
The coefficients of firm size, growth, cash holding, leverage, interest expense, earning and state-owned dummy are -2.152, 0.274, -0.368, 0.196, -0.004, 2.293 and 0.010 respectively. The coefficients of firm size, growth, cash holding and earning are significant at 1% significant level, while the coefficient of leverage is marginally significant at 10% significant level. The coefficients of interest expense and state-owned dummy variable are not significant if we stick to a 5% significance level. It is indicated decrease in firm size and cash holding can lead to higher firm value, while higher growth, leverage and earning may cause an increase in firm value. We used cross-fixed effect here according to the Hausman test, and the value of goodness (R2) is 72.3%.
The second hypothesis which illustrates the impact of ownership concentration on firm value is examined by both model 1 and model 2 in the two columns of Table 2. The coefficients of ownership concentration are 0.003 in both regressions and are both marginally significant at the level of 10%. Although the magnitude of coefficients is small and the level of significance is low, it still indicates that a marginally significantly positive relationship between ownership concentration and firm value exists which supports hypothesis H2A. This positive relationship supports
4.2.2 Interactions
Table 3 reports the results of third and fourth regression models. The cash dividend is still measured in two aspects: cash dividend announcement and cash dividend amount. Model 3 and model 4 involve interaction terms between cash dividend and ownership concentration added to the first and second regression models. Column 1 represents the results of interaction effect between the cash dividend announcement and ownership concentration, while Column 2 represents the results of interaction effects between cash dividend amounts and ownership concentration.
The third hypothesis is developed to test the interaction effect between cash dividend and ownership concentration on firm value. The interaction term HD1 is added in Column 1 to test hypothesis H3Aa which illustrates the effect of interaction between
cash dividend announcement and ownership concentration. The coefficient of the interaction is 0.0004 but insignificant at 5% significance level. The coefficients of cash dividend announcement and ownership concentration in model 3 are 0.008 and 0.003 respectively. They do not change much from model 1, but none of them is significant if we stick to a 5% significance level. The result shows an insignificant relationship between the interaction term and firm value which cannot support hypothesis H3Aa.
Note: This table presents results for cash dividends and ownership concentration in 2006-2015. All the variables are winsorized at 5% and 95%. TQ is the ratio of the market value of equity to book value of total assets; D1 equals to 1 when the firm is state-owned and equals to 0 otherwise; Div the ratio of af the amount of dividend (which is dividend per share multiplied by the number of shares) to total assets; H represents ownership concentration which is the sum of the proportions of the biggest five shareholders; HD1i,t-1 is the product of H and D1 ; HDIV is the product of H and DIV ; S is firm size measured as natural logarithm of total assets; G represents firm growth which is
Table 3. Regression Results for Interaction between Cash Dividends and Ownership Concentration Variables Model 3 Model 4 TQ TQ D1 0.008 [0.095] Div 5.053*** [2.715] H 0.003 0.003** [1.634] [2.090] HD1 0.0004 [0.254] HDiv -0.111*** [-2.620] S -2.157*** -2.150*** [-35.025] [-35.336] G 0.272*** 0.269*** [6.926] [6.835] C -0.357*** -0.373*** [-3.106] [-3.269] L 0.194* 0.191* [1.734] [1.709] I -0.004 -0.004 [-0.580] [-0.582] E 2.294*** 2.309*** [7.815] [7.865] D2 0.009 0.01 [0.144] [0.169] Constant 15.928*** 15.885*** [43.369] [43.340]
Cross-fixed effect Yes Yes
Time-fixed effect Yes Yes
Adjusted R 0.723 0.723 F-statistic 14.701** 14.714*** Observations 10803 10803 i,t i,t i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 2
the change of total assets throughout the whole year over the value at the beginning of the year; C is cash holdings which is defined as cash and cash equivalents divided by total assets; L is measured as total debt divided by total assets; I is defined as interest expense divided by total debt; E is earning net income before extra/preferred dividends divided by total assets; D2 equals to 1 when firm i is state-owned in year t-1 and equalts to 0 otherwise. Standard errors are in parentheses. *, ** and *** indicate statistical significant level at 10%, 5% and 1% respectively.
i,t i,t-1
i,t-1
i,t-1
The interaction term HDiv is added in Column 2 to test Hypothesis H3Ab, whether the
more cash dividend distributed in firms with more concentrated ownership can cause extra damage or benefit on firm value. The coefficient of HDiv is -0.111 and significant at the 1% significant level. This significantly negative interaction effect of cash dividend amount and ownership concentration indicates the entrenchment and expropriation problems based on the conflicts between minority shareholders and blockholders. In other words, the bigger share that blockholders’ hold, the higher possibility that they will use excess cash dividend distribution to entrench firms’ resources and pursue their own interests (Shleifer and Vishny, 1997). Considering that blockholders often have a high proportion of shares in Chinese listed firms, they can use their power to influence managers’ decision of dividend policy. In addition, since cash dividend distribution has been encouraged and regulated by the Chinese government on its scale, it has became the best way for entrenching and expropriating minority shareholders’ wealth. The coefficient of cash dividend amount becomes 5.503 and also significant at the 1% significant level which indicates that a significant positive relationship is represented between cash dividend amount and firm value which supports hypothesis H3Ab. On the one hand, signaling theory illustrates that
dividend transfers information of good future expectation to outside investors, the more cash dividends are paid, the stronger confidence in the future profitability will be showed (Pettit, 1972). On the other hand, agency cost theory gives the view that cash dividends can reduce the disposable cash controlled by managers. In this way, the managers are forced to raise funds more from the outside which leads to better monitoring by putting managers under the scrutiny of outside institutions and investors. In addition, the reduced disposable cash eliminate the possibility that managers serve their own interests in front of shareholders’ interests which also reduces the agency costs. The result is in line with many previous studies such as Easterbrook (1984) and Jensen (1986). The coefficients of ownership concentration is the same with that of model 2 but significant at the 5% significant level. This result supports the second hypothesis which proposes that a higher ownership concentration plays a role of better monitoring rather than entrenchment.
We use cross section-fixed effect and time-fixed effect according to the result of Hausman Test and the value of R2 is 72.3% which is the same as model 2.
4.3 Robustness Test
We use Tobin’s q to represent firm value in the original regressions in this thesis, and our measurement of Tobin’s q is defined as the ratio of the sum of market value of equity and the book value of all liabilities to total assets following Setia-Atmaja (2009). However, book value of debt is used in this measurement instead of market value of debt since the data of the latter one is unavailable. This replacement adds a proxy number (book value of debt instead of market value of debt) into Tobin’s q. Last years’ cash dividend distribution may cause the firm to take more debt which increases both the numerator and denominator of Tobin’s q. With the proxy case, we do not know how it will change in Tobin’s q. To eliminate the effects of this unknown change, we focus on the real market value by researching on the market value of equity only. In this case, we do the robustness by removing the effects of debt. In the robustness test, we therefore use the market to book value of equity (market capitalization divided by the difference between total assets and total debt) to represent firm value.
Note: This table presents results for cash dividends and ownership concentration in 2006-2015. All the variables are winsorized at 5% and 95%. MB is the ratio of the market to book value of equity; D1 equals to 1 when the firm is state-owned and equals to 0 otherwise; Div the ratio of af the amount of dividend (which is dividend per share multiplied by the number of shares) to total assets; H represents ownership concentration which is the sum of the proportions of the biggest five shareholders; S is firm size measured as natural logarithm of total assets; G represents firm growth which is the change of total assets throughout the whole year over the value at the beginning of the year; C is cash holdings
Table 4. Regression Results for the Robustness Test of Cash Dividends and Ownership Concentration Robust 1 Robust 2 MB MB D1 -0.093* [-1.916] Div 0.539 [1.027] H 0.002 0.003 [0.707] [0.896] S 0.185 0.157 [1.541] [1.316] G 0.071 0.07 [0.900] [0.886] C -0.206 -0.167 [-0.916] [-0.745] L -0.450** -0.440** [-2.062] [-2.012] I -0.007 -0.006 [-0.571] [-0.568] E -0.274 -0.297 [-0.479] [-0.520] D2 -0.029 -0.028 [-0.247] [-0.242] Constant 1.602** 1.685** [2.242] [2.361]
Cross-fixed effect Yes Yes
Time-fixed effect Yes Yes
Adjusted R 0.256 0.256
F-statistic 2.829*** 2.826***
Observations 9375 9375
Variables i,t i,t
i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 2
Note: This table presents results for cash dividends and ownership concentration in 2006-2015. All the variables are winsorized at 5% and 95%. MB is the ratio of the market to book value of equity; D1 equals to 1 when the firm is state-owned and equals to 0 otherwise; Div the ratio of af the amount of dividend (which is dividend per share multiplied by the number of shares) to total assets; H represents ownership concentration which is the sum of the proportions of the biggest five shareholders; HD1 is the product of H and D1 ; HDIV is the product of H and DIV ; S is firm size measured as natural logarithm of total assets; G represents firm growth which is the change of total assets
Table 5. Regression Results for the Robustness Test including Interaction between Cash Dividends and Ownership Concentration
Variables Robust 3 Robust 4 MB MB D1 -0.005 [-0.030] Div 4.793 [1.311] H 0.003 0.003 [0.859] [1.005] HD1 -0.002 [-0.531] HDiv -0.098 [-1.175] S 0.177 0.158 [1.465] [1.328] G 0.072 0.066 [0.908] [0.832] C -0.212 -0.17 [-0.939] [-0.758] L -0.446 -0.442** [-2.041] [-2.024] I -0.007 -0.007 [-0.572] [-0.569] E -0.276 -0.281 [-0.483] [-0.491] D2 -0.029 -0.028 [-0.243] [-0.240] Constant 1.616** 1.666** [2.260] [2.334]
Cross-fixed effect Yes Yes
Time-fixed effect Yes Yes
Adjusted R 0.256 0.256 F-statistic 2.827*** 2.825*** Observations 9375 9375 i,t i,t i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 2
throughout the whole year over the value at the beginning of the year; C is cash holdings which is defined as cash and cash equivalents divided by total assets; L is measured as total debt divided by total assets; I is defined as interest expense divided by total debt; E is earning net income before extra/preferred dividends divided by total assets; D2 equals to 1 when firm i is state-owned in year t-1 and equalts to 0 otherwise. Standard errors are in parentheses. *, ** and *** indicate statistical significant level at 10%, 5% and 1% respectively.
i,t
i,t-1 i,t-1
i,t-1
i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1 i,t-1
i,t-1
i,t-1
i,t-1 i,t-1
5. Conclusion
In this section, the conclusion based on the results of the regression models is summarized first. Then, the contribution and limitation of this thesis and suggestions for further research are provided.
