University of Groningen Faculty of Economics and Business Msc Business Administration June 2009
Ownership and Firm Performance:
Evidence from China
Project ID: EWM066A20 Supervisor: Lammertjan Dam Student: Ying Chen
Student ID: S1655833
Email Address: s1655833@student.rug.nl
- 2 -
Content
I Introduction ---4
II Literature Review ---7
2.1 Agency theory and Corporate Governance ---7
2.2 Ownership Concentration---10
2.3 Ownership type ---13
2.4 Industry ---16
2.5 Hypothesis ---16
III Methodology and Data ---19
3.1 Methodology ---19
3.2 Data ---20
3.3 Dependent variable ---20
3.4 Independent variables ---21
3.5 Control variables ---22
3.6 Summary ---23
IV Results ---23
4.1 Data description ---23
4.2 Regression results ---27
4.3 Robustness tests ---31
V Conclusion ---33
5.1 Implications ---33
5.2 Limitations ---35
5.3 Future researches ---36
References ---37
Appendix ---41
- 3 -
Abstract
This study investigates the relationship between ownership and performance in China.
Ownership is classified into two dimensions: concentration of owners and types of owners. The sample includes 402 listed firms in the spinning industry and the electronic industry from 2004 to 2007 in China. The results are threefold. First, we find a bell-shaped relationship between ownership concentration ratio and firm performance. Second, state owners have a negative impact on the firm performance, but they also generate a bell-shaped relationship. Third, we find that there are differences between industries regarding these relations.
Key Words
Firm performance, ownership structure, ownership concentration ratio, ownership type, China
- 4 -
Ownership and Firm Performance: Evidence from China
I Introduction
The relationship between ownership and firm performance is a hotly debated topic in the corporate governance field, as well as in the corporate finance field. Berle and Means (1932) addressed that in modern corporations, ownership and control rights are separated. The principle-agency theory predicts that this separation may cause agency problems. On the one hand, it is argued that the more professional manager can manage the firm more efficiently and ideally, which will profit both the firm and the shareholders. On the other hand, it is argued that the manager may pursue their own private benefits. Managers may not act in the best interest of the shareholders of the firm. The latter issue is called the principle-agent problem (Jensen and Meckling, 1976; Jensen 1986). Good corporate governance is needed to help the owners to resolve the agency problem, enhance their monitoring of the manager and thus protect their stakes. However, corporate governance is not only attracting attention from scholars, but also policy makers. For instance, the British South Sea bubble in the 18th century resulted in a systematic reform in British corporate law. Comparable systematic crises also happened in the U.S., for example the savings and loan debacle in 1980 and in Asia with the famous Asian financial crisis. Individual corporate crises have played an important role in the rethinking of corporate governance structures, such as the bankruptcy of Barings Bank and the scandal of Enron, WorldCom, and Marconi, etc. All these individual corporate affairs happened due to the misbehavior of the manager. Therefore, the system of corporate governance is always being pressured.
There are many economists who want to enhance the system of corporate governance so that the value of shareholders can be protected. Related discussions have been done in the field of law before. Many researchers studied the law system of the country and its relationship to corporate governance (LLSV1, 1998). These kinds of studies try to
1 LLSV is the abbreviation of the name of four authors. They are La Porta R., Florencio Lopez-de-Silanes, Andreo Shleifer and Robert W. Vishny.
- 5 -
show how origin law system and company laws help to monitor managers. Tirole (2006) also discusses the corporate governance issues. He focuses on the optimal contract between the investors and the manager to ensure the manager to work efficiently. This paper focuses on the role of ownership. To provide a comprehensive study on ownership, we classify ownership into two dimensions. They are concentration of ownership and types of owners.
The specific relationship between ownership concentration ratio and firm performance is receiving high attention, and there is ample empirical work, for example Becht et al.
(2002). Shleifer and Vishny (1997) show their attention on the importance of ownership concentration in the system of corporate governance. However, former research failed to reach a consensus on this relationship. The conclusion regarding the relationship between ownership concentration ratio and firm performance varies between different countries. Some found that there is not a clear relationship between the economic performance and shareholder concentration (Zhuang and Zhang, 2008).
Some found a positive relationship between them (Cubbin and Leech, 1983).
Thomsen and Pedersen (2000) found a positive relationship, but the effect leveled off for high ownership concentration. Recently, there is also growing attention to indicate the effects on firm performance generated by shareholder types (Thomsen and Pedersen, 2000). Whitley (2000) argues that different shareholder types will have different effect on corporate governance, hence the performance. Useem (1998) also drew a similar conclusion on the effect of foreign institutional investors on shareholder value. Thus, ownership structure and the type of the owner are interesting and a hot topic. In this paper, we also want to compare a traditional with a new industry. The former one, such as spinning industry, is usually stable and familiar to investors, while the latter one, such as electronic industry, is new, thus the monitoring cost can be different. Zechouser and Pound (1990) and Chen (2004) both find that relationship between ownership and performance is different within industries.
Therefore, the research question of the paper is: “What is the relationship between ownership and firm performance in China in the spinning and electronic industry?”
From 402 data collected in two stock exchanges in China from 2004 to 2007, we
- 6 -
would like to answer two subquestions. First, we would like to see the relationship between ownership concentration ratio and its firm performance. Second, we want to find the effect of the largest owner’s identity on the firm performance.
This paper is developed based on these prior researches, but there are some main contributions of this paper. First, we classify the ownership by two dimensions: types of owners and concentration of ownership. We aim to identify the relationship between ownership concentration ratio and firm performance. Also, we want to find what the role of shareholder type in this relationship is. Second, the research is done in China. Historically, studies are done in some developed countries (US, EU) or region (Taiwan, Hong Kong). The mainland of China does not attract a lot of research due to the difficulty of collecting data and the language difference. However, China is one of the dramatically grown economies, and its role in the world economy and politics is becoming increasingly important. Thus, it is attractive to do the research on China. Third, other researches are done in one specific industry, or use many industry dummies in the regression, to find the industry-specific effects. However, this paper will only focus on two industries, traditional spinning industry and modern electronic industry. Since the spinning industry is a traditional industry and the electronic industry is a modern industry, we suppose the monitoring structure and monitoring costs is different in two industries. As the spinning industry is traditional, shareholders can monitor the firm in accordance with previous experience or a certain pattern, the monitoring costs can be low. On the other hand, as a new industry, the way to monitor the firm is a need to explore, and hence the shareholders may put more time and money on their governance. This is why we suppose the relationship between ownership and performance can be different between the two industries. Chen (2004) did a similar study in Taiwan and found that the relationship is not same between two industries.
The remainder of this paper consists of four sections. We first discuss prior literature in more detail and give some theoretical background. In the second part, based on the reviewed literature, the main hypothesis is formed. Then in section three, the paper
- 7 -
will introduce the methodology and the measurement of the variables. Also, we will summarize the data based on the models in this part. In the fourth section, the results and implications will be given. In the last part, the paper will give the final conclusion, as well as some limitations and further researching recommendations.
II Literature Review
This section gives an overview of ownership and its relationship to economic performance of the firm. First, we would like to give the basic theoretical background of the agency theory and the concept of corporate governance. Then, we discuss the relationship between the ownership concentration ratio and firm performance. Next, we review how shareholder type influences firm performance. Finally, some hypothesis will be developed based on the theoretical background and these prior literatures.
2.1 Agency theory and Corporate Governance
As mentioned in the first section, “corporate governance” is a hot topic both for practitioners and scholars. The basic idea behind good “corporate governance” is to resolve the agency problem. We first discuss possible causes of agency problems and some of its solution, and next, we give a framework of corporate governance.
2.1.1 Agency Problems
Agency problems happen in many situations. Basically, a principal assigns a person, the agent, to represent him to undertake a certain action. However, the principal is not sure whether the agent is really acting in the best interest of the principal. The main reason for agency problems is asymmetric information. When the managers run and organize the firm everyday, the owners may not have enough time to know how everything is going. Thus, the information between the two parties is asymmetric.
Then, adverse selection, moral hazard and shirking problem may occur, which cause the so-called agency cost to increase.
- 8 -
Agency problems can also be derived from the inconsistent objectives between the two parties. The owners are interested in maximizing the firm value; while the managers are interest in their compensation. Denis (2001) introduced three kinds of interest conflicts. The first conflict deals with the managers’ desire to remain in control. Managers always want to remain their position in the firm. Hence, when the shareholders think that the performance can increase under another manager; they will face the threat of being replaced. This will result in a serious conflict between shareholders and managers. The second conflict is managerial risk aversion. For a shareholder, he can diversify his risk by holding many different shares. Thus, the shares he owned in one firm are relatively small in his portfolio. However, the managers’ compensation, power and even human resource are related to the performance of the firm. Therefore, for a single project, the level of risks the managers can bear is much lower than the shareholder. The managers bear more in a negative project. Thus, the manager will refuse some projects which are favored by shareholders. This last issue is related to free cash flow, especially when the compensation of the manager will increase with the scale of the firm. Managers can use free cash flow to invest, even in some negative value projects, aiming to increase the firm’s size. On the other hand, owners want the managers to do the best investments and may thus require some dividends to decrease the level of free cash flow. When owners lack monitoring power, this conflict will be severe.
Denis (2001) also gives three solutions for the agency problem. The first solution he mentions is bonding; managers and shareholders make a contract beforehand to ensure the manager to take actions which will maximize shareholder value. However, this is virtually impossible to accomplish in the real world. The second solution is incentive alignment. Watts and Zimmerman (1976) suppose that the compensation can be a tool to align interests. The compensation should be related to the performance of the firm. However, in reality, this is also hard to accomplish. This paper will focus on the third solution, monitoring. Denis (2001) shows that shareholders can monitor the managers on the decisions they made. However, the author also announces that the lack of professional knowledge and monitoring incentive will be two drawbacks. This
- 9 -
paper tries to find how much they own and what kind of shareholders will have the monitoring power and hence deal the agency problem.
Figure 1 A corporate governance framework: the internal and external architecture
- 10 -
2.1.2 Corporate Governance
Corporate governance is aimed at diminishing the agency problem. Shleifer and Vishny (1997) and Tirole (2001, 2006) give a narrow definition of corporate governance. They define it as “ways in which the suppliers of finance to corporations assure themselves of getting a return on their investment.” The shareholders are a kind of finance supplier and the solutions given by Denis (2001) are the ways to assure their rights. Cadbury (1999) addressed that only those firms, which disclose their information to all investors, have a good corporate governance system. The World Bank (1999) gives the framework for corporate governance (Figure 1).
It gives two kind of governance, internal and external. The internal corporate governance includes ownership structure, composition of board of directors, duration and compensation of managers, etc. The external corporate governance relates to regulation, capital market, financial market and accounting rules, etc. This paper focuses on the internal corporate governance, the shareholders in the top left part of the figure. In the following section, the paper will introduce some background on the ownership concentration and shareholder identity. LLSV (1999, 2000) divide the level of corporate governance into state level and company level. The right column in Figure 1 is related to state governance, while the left column is refer to company level governance. The middle column is mixed.
2.2 Ownership Concentration
Ownership concentration is directly reflecting the power distribution or the control rights of the firm, therefore we may expect that a different ownership concentration will have a different impact on the actions and performance of the firm. In the following part, we will first introduce the effect of ownership concentration ratio on firm performance, and then the causal relationship between them.
2.2.1 Three effects
The paper will show three effects. The first effect, entrenchment effect, is introduce by Demsetz (1983) and Jensen and Ruback (1983). They have an entrenchment
- 11 -
hypothesis. They think when the ownership concentration ratio increases to a certaion proportion, more power and more voting rights will be concentrated on the shareholder’s hand. In that case, the shareholders may control the firm and benefit from self-dealing. He can transfer the profit of the firm to his own pocket, which is called “tunneling” by Johnson et al. (2002). He aims to remain his power regardless profit-maximizing actions. For instance, large shareholders may oppose takeover, even though the takeover can increase the firm value. Therefore, they think the larger the level of ownership concentration, the higher the probability the shareholders seize the firm, the worse the performance of the firm. There is a negative relationship between ownership concentration and firm performance according to this effect.
The second effect is convergence of interest effect, which is in conflict with the entrenchment effect. Jensen and Meckling (1976) announce that when the shares owned by the shareholder increase, the benefits or the losses of the firm are more correlated to the wealth of the shareholders. Then, the shareholder will have the incentive to monitor the manager and maximize the economic performance of the firm.
The active monitoring hypothesis (Agrawal et al., 1996) and the signaling effect (Leland and Pyle, 1977) both support this idea. They show that the shareholders with larger shares will send a positive signal, which will make the outsiders regard the firm as a growth firm and worthy of investments. There is a positive relationship between ownership concentration and firm performance according to this effect.
The third effect is called the collateralize effect. This effect shows the relationship between collateralized shares of shareholders and firm performance. Chiou et al.
(2002) found that the higher the shareholders collateralized their shares, the poorer the firm performance. However, litter researches are done in this field. There is a negative relationship between ownership concentration and firm performance according to this effect.
The two main effects show definitely different effects of ownership concentration on firm performance. Hence the mixed results on ownership concentration can be
- 12 -
explained by the different domain effect in different sample. Begin with Shleifer and Vishny (1986), prior studies hypotheses that there is a linear relationship (Bebchuk, 1999 and Lichtenberg and Pushner, 1994). For example, Amihud and Lev (1981) observes that firms controlled by large block shareholders are less likely to take inefficient diversification. Most studies support the view that a positive exist between ownership structure and performance, such as Lloyd, Hand and Modani (1987); Leech and Leahy (1991). They support the convergence of interest effect. Recently, more and more researchers think that there is a nonlinear relationship, which can be seen a combination of the above effects. Some find there is a U-shape relationship (Chen, 2004; Lin et al., 2008), while most of the studies find there is an inverse U-shape, or so-called Bell-shape (Morck et al., 1988; Hu and Wezumida, 2008). A Bell-shape means “convergence of interest effect” domains at the lower level of ownership concentration, while “entrenchment effect” domains at the higher level.
2.2.2 Causal relationship
Another hot topic in discussing this relationship is the causal relationship. On the one hand, Pederson and Thomsen (2000) argue that the ownership structure, as an exogenous variable, is quite stable. They suggest that if ownership can influence firm performance, why not shareholders change their portfolio and maximize the profit of the firm. Loderer and Martin (1997) estimate a simultaneous equation and found that Tobin’s Q is not predicted by insider shareholding, but a negative predictor of insider ownership. Cho (1998) used the two stage least square (2SLS) model and concluded that Tobin’s Q affect insider ownership, but ownership’s impact on Tobin’s Q is insignificant.
In contrast, Demsetz (1983) contends that the ownership structure is an endogenous outcome of the firm, balancing the monitoring benefits of concentrated structure and owner’s portfolio risk. He announces that “an outcome of competitive selection in which various cost advantages and disadvantages are balanced to arrive equilibrium organization of the firm.” According to his argument, ownership structure is not influenced by corporate performance, since the free market will achieve the optimal
- 13 -
structure which can maximum the firm value. Thus, since each company has different firm-specific risks, nature and complexity characteristics, the optimal ownership structure for each firm is different. Demsetz and Lehn (1985) show insignificant relationship between ownership structure and performance by empirical tests, when controlling other variables. Hu and Wezumida (2008) also show a U shape relation of ownership concentration to performance. They used grange causality tests and found that the changes in performance will follow the changes in concentration, but they did not find a reverse relationship.
Both causal relationships are reasonable and many empirical studies support either of them. Thus, in this paper, we ignore the causal relationship between them, just aim to find that kind of relationship between ownership concentration and firm performance.
2.3 Ownership type
In this paper, we will also consider the type of the largest owner of the firm. Short (1994) demonstrates that the shareholder’s identity will have crucial effects on firm’s performance. Thomsen and Pedersen (2000) further argue that different types of shareholder have different objectives, which will lead to different firm strategy and economic performance. According to their demonstration, ownership structure is the equilibrium outcome when all shareholders are pursuing their goals. They argue that firms can be modeled as a nexus of contracts with its stakeholders and ownership, which can be assigned to each stakeholder. If a stakeholder becomes the owner of the company, their transactions can be internalized and contracting costs will be relieved.
However, at the same time, the stakeholder must incur ownership costs. Since different types of stakeholders have different contracting and ownership costs, different types of dominate shareholder will show different costs and advantages. As a consequence, the objective of the firm reflects the particular costs and benefits of the dominate shareholder.
2.3.1 Background of ownership type in China
Shares in China are divided to domestic shares (A-shares) and foreign shares (B-, H-,
- 14 -
N-shares). We only consider A-shares, which are held widely by individuals. There are four types of A-shares in China. Wang and Xu (1999) give a short summary on it.
The state shares are shares which hold directly by the central or local governments, or the shares hold by the enterprise which is 100% owned by the government. The ultimate owner of state shares is the State Council. Those state shares are forbidden from public trading, but can be transfer to domestic institutions on the agreement of CSRC2. The legal person shares are the shares owned by domestic entities, which can be companies, financial institutions and even can be a person. The ultimate owners of those domestic entities are either a person or a group of person, including family and union. The ultimate owners can also either be Chinese or foreigner. The shares owned by them can be tradable and nontradable. The annual report of the firm will distinguish these two parts and inform the outsiders. The third type is the least popular one in China, employee shares. These shares are owned by the managers and other employees of the firm. Usually these shares are working as an incentive scheme, which aim to align the interest between shareholders and managers. In China, many listed firms do not issue employee shares. Thus, we do not consider this type in this paper. The forth type is the tradable A-shares. Company is required to issue tradable A-shares, which are no less than 25% of the total outstanding shares in its IPO. Such shares can be traded without any restriction, and hence they are widely held by the public. Therefore, in this paper, we focus on the first two types: state shares and legal person shares. We consider the ultimate owner of the largest shareholders and the total shares they owned including tradable and nontradable shares.
2.3.2 State shares
The state owners have some other objectives besides pursuing largest profit of the firm. Usually, they aim to support some new or technology industries. Thus, they can bear the bad performance in shout run in order to have long run profit (Brickely et al., 1988). Chen (2004) summarizes three characteristics of state stocks. First, they can not be traded freely. Alchaiin and Demsertz (1972) showed that the untradable state’s ownership right may cause insufficient and inefficient monitoring. Since they are
2 CSRC is the abbreviation of “China Securities Regulatory Commission”
- 15 -
untradable, people can not judge the firm performance in stock market. Second, they are lack of economic incentive. Since it is owned by the government, managers can turn to the government if the firm is really in trouble, which is called soft budget constraint (Kornai, 1980). Then the firm can avoid bankruptcy. Therefore, managers in state owned company are more likely to misbehave and embezzle (De Alessi, 1980).
Third, as mentioned before, state owners may have mutliobjectives. It is hard for them to accomplish all the objectives at one time. Since other objectives, rather than make profit, is more important, the state owners often want to accomplish other purpose in the expense of profit. On the one hand, the behavior of the manager is even harder to recognize (Pryke, 1982). On the other hand, these firms are manipulated by politicians as their political advantage (Milward and Parker, 1983). Thus, the firm owned by state may have worse performance than others.
2.3.3 Legal person shares
Here, we divide the legal person shares into three categories: individual person, group cwners and foreign owners. We first analyze individual person, which means that only one person is the largest ultimate owner of the firm. The way they own the firm is just like state. They have shares of firms directly or solely owned enterprises. In our sample, none of these persons are the manager or worker of the firm, thus we do not have managerial ownership in this paper. In other words, the ultimate owners are the outsiders. Weisbach (1988) and Huson et al. (2001) suppose that although the owners do not organize the firm, they have sufficient professional knowledge. Also, they can monitor the firm in a fairer and more objective way. Thus, the more shares the outside individual person owned, the better the firm performance.
The second one is group owners. Most of them are family, and few are some unions, such as working union and village3. They have the shares directly and through solely owned enterprises. Since the groups often invest in many stocks, they diversify the risks through their portfolio. Therefore, although they may have a large proportion of the shares, they do not monitor the management sufficiently and effectively. Moreover,
3 People, in a village, form an investment club. The club does the investment on behalf of all the members.
- 16 -
they may make private profit at the cost of other small shareholders (La Porta et al., 1998). Hence, the more shares the group owned, the worse the performance.
The last one is foreign investors, which can be a foreigner or a foreign group.
However, the foreigner here is not really a foreigner. In some cases, the foreign owner is a Chinese person with a foreign nationality. Although they are Chinese, they have foreign nationality and hence the policy and regulation is different for them. We do not consider this group since the foreign sample is too small.
2.4 Industry
Zechouser and Pound (1990) indicate that price/earnings ratio increases with ownership concentration in industries which have lower monitoring costs. This finding implies that the monitoring cost in different industry can be different. We think that the monitoring costs will be different in traditional industry and modern industry. The traditional firms are much stable and transparent to investors. The long history will help the shareholders to monitor the firm, so the monitoring cost will be low. The modern firms are new to investor. Not only the corporate governance system, but also the production technology and process will be unfamiliar to shareholders.
Thus, the monitoring costs may be high. Therefore, the industry will have impact on the relationship between ownership structure and performance. In this paper, we choose two typical industries. Spinning industry is a typical traditional industry, while the electronic industry is a typical modern industry. Chen (2004) found the relationship between ownership and performance is different in spinning and electronic industry.
2.5 Hypothesis
Based on the prior researches, the paper made some hypotheses. First, we think the relationship between ownership concentration ratio and firm economic performance is nonlinear. When the concentration ratio is low, which means the firm is almost widely held, and thus the monitoring incentive is low and the free rider problem may occur frequently. On the other side, when the concentration ratio is very high, the largest
- 17 -
owners control the firm, and hence they can pursuit their own interests. The cost will be relatively low, when they search for their private benefit. Thus, we agree with Thomsen and Pedersen (2001) that the relationship is an inverse U-shape, in other words, Bell-shape.
H1: Company performance is a bell-shaped function of the ownership concentration H1a: When the concentration level is low, there is a positive relationship between ownership concentration ratio and firm performance
H1b: When the concentration level is high, there is a negative relationship between ownership concentration ratio and firm performance
Second, we consider the ultimate owner of the largest shareholders. As defined by La Porta et al. (1999), there are several types of shareholders within the company.
Different type will lead to different strategy and performance. The paper tries to find their different impact on the firm performance. According to the classification of China, the paper divides shareholders into three main categories. They are state owner, individual owner, and group owner. Thus, according to prior researches, the paper has following hypothesis:
H2: The relationship between the different ultimate type of the largest shareholder and firm performance will be different
H2a: There is a negative relationship between state owner and firm performance H2b: There is a positive relationship between individual owner and firm performance H2c: There is a negative relationship between group owner and firm performance
Third, the paper combines the ownership concentration and owner identity. The paper wants to find whether the type of the largest shareholder will have impact on the relationship between ownership concentration ratio and firm performance. We think the impact on the relationship is the same as their impact on firm performance directly.
We propose that:
- 18 -
H3: Different type of the largest shareholder’s ultimate owner will have different impact on the relationship between ownership concentration ratio and firm performance.
H3a: State owner will have a negative effect on the relationship H3b: Individual owner will have a negative effect on the relationship H3c: Group owner will have a negative effect on the relationship
Forth, the paper aims to find whether the relationship differs in different industries.
We choose spinning industry and electronic industry. The reason we choose these two industries is that the spinning industry is a traditional industry, while the electronic industry is a modern and relatively new industry. We think the monitoring structure and monitoring costs may be different in two industries, thus the relationship between ownership concentration ratio and performance can be different:
H4: The relationship between ownership concentration ratio and firm performance will be different in spinning industry and electronic industry.
To sum up, we want to find the relationship showed in figure 2.
Figure 2 Research Relationship
Ownership
Ultimate Largest Owner
Firm Economic Performance H1
H3 H2 Industry
H4
Control Variables - Growth of profit - Financial leverage ratio - Firm size
- 19 -
III Methodology and Data
The model we used in this paper is derived from Thomsen and Pedersen (2000). After introducing the model, this section will continue to introduce how the data is collected and how the variables are measured.
3.1 Methodology
The dependent variable in this model, ROA, will be the measurement of firm performance, and the independent variables will be the proxy of ownership and the control variables. The regression model is shown below:
Model 1 (testing hypothesis 1):
General model (M1a):
ROAi,t = a1 + b1*CONi,t + b2*CONi,t2
+ b3*INDi,t + b4*GROWTHi,t + b5*LEVi,t + b6*SIZEi,t + εi,t
For each industry (M1b, testing hypothesis 4):
ROAi,t = a1 + b1*CONi,t + b2*CONi,t 2
+ b3*GROWTHi,t + b4*LEVi,t + b5*SIZEi,t + εi,t
Model 2 (testing hypothesis 2 and 3):
General model (M2a):
ROAi,t = a1 + ∑cm*ID*CONi,t + ∑cn*ID*CONi,t
2 + ∑cq*IDi,t + b3*INDi,t
+ b4*GROWTHi,t + b5*LEVi,t + b6*SIZEi,t + εi,t
where cq is used to test hypothesis 2;
cm and cn are used to testing hypothesis 3 For each identity (M2b, testing hypothesis 4):
ROAi,t = a1 + b1*CONi,t + b2*CONi,t
2 + b3*IDi,t + b4*INDi,t
+ b5*GROWTHi,t + b6*LEVi,t + b7*SIZEi,t + εi,t
Where
i and t is the equations denoted the firm and the time subscripts, Y = ROA, CON
- 20 -
= Ownership concentration ratio,CON2 = the square of CON, IND = Industry Dummy
= 1, if the industry is spinning industry; and 0 otherwise, GROWTH = growth of profits, SIZE = Natural logarithm of total asset of the firm in the end of the year, LEV
= leverage = Value of all liabilities in the end of year / Book value of total assets in the end of year, ID = the identity of owners, ε = error term
3.2 Database
Since the study is focused on China, We collected the data from two stock exchanges:
Shanghai and Shenzhen, in China. They are the only two stock exchanges in China.
They have provided complete and detailed information of the firm, including their annual report. There are 42 spinning firms and 28 electronic firms listed in Shanghai Stock Exchange (SSE), while there are 36 spinning firms and 44 electronic firms listed in Shenzhen Stock Exchange (SZSE). Thus, in total, the paper will study 78 spinning firms and 72 electronic firms. First, we want to collect the data from 2000 to 2007; however, the annual report listed in SZSE is mostly only available from 2004.
Therefore, in order to keep the observations in each year balanced; the duration is reduced to four years, from 2004 to 2007. Also, not all firms are available. Some firms were bankrupted before 2004, and thus, we exclude those firms from our database. In the end, we have 402 observations in four years with an increasing trend. We have 64 observations in 2004, 97 observations in 2005, 111 observations in 2006 and 130 observations in 2007. There are 219 observations and 68 firms in the spinning industry and 183 observations and 69 firms in the electronic industry. All the variables can be obtained in the annual report of the firm.
3.3 Dependent variable
The dependent variable used in this paper is to reflect the firm’s economic performance. We have the measurement, return on asset (ROA), which is the ratio of earnings before interest and income tax (EBIT) to total assets. ROA is a popular measure of firm performance, which has been used by many authors (Lin et al., 2008;
Lskavyan and Spatareanu, 2006). It reflects the short term probability of the firm, and does not influenced by tax effects. Many researchers use market measures as proxies
- 21 -
of firm performance, such as return on equity (ROE) and stock price, but this paper does not. One reason is that the market measures are not directly provide in the firm annual reports, self calculation may cause some errors. The other reason is we try to follow the idea of Demsetz and Villalonga (2001). They announced that it is more sensible to test the relationship between ownership and performance based on what
“has been” done, not based on what “will be” done. This idea is also supported by Prowse (1992). He found that the market returns will be influenced by the divergences between owners and managers. Furthermore, he pointed out that ROA, which is accounting-based, is preferred in this relationship study.
3.4 Independent variables
There are three major independent variables used in this paper. They are ownership concentration ratio, industry and identity.
The first variable is the ownership concentration ratio (CON), which is the measurement of ownership structure used by Thomsen and Pedersen (2000). It is the percentage of shares owned by the first largest shareholders.
The second variable is industry. Industry is measured by the dummy variable. We have two industries in the study. They are the spinning industry and the electronic industry. According to “The Guide of the industry classification of the listed firms” in China, the classification of industry is according to Global Industry Classification Standard (GICS), which is based on the potential of investment. Although both industries are in “C”, which represented the manufacturing industry, the spinning industry has a subcode C1 and the electronic industry is C5. The industry dummy, IND, is 1 if a firm belongs to the spinning industry and 0 otherwise.
The third independent variable used in the model is the ownership identity. Just like the measurement of industry, the types are also measured by dummy variables.
However, the owner types mentioned here is the ultimate owner of the firm’s largest shareholder. In the sample, we have four ultimate owners. They are state owners,
- 22 -
individual owners, group owners and foreign owners, however, the companies belong to the last type is too few. Therefore, when analyzing the effect of the owner’s type, we exclude the last group. We then have three dummies, named STA, PER and GRO.
If the firm’s largest shareholder is state, then STA will be 1, otherwise 0. If the firm’s largest shareholder is a person, then PER will be 1 and otherwise 0. If the firm’s largest shareholder is a group, then GRO will be 1 and otherwise 0.
Thus, model 2 is developed to estimate:
ROAi,t = a1 + c1*GROi,t*CONi,t + c2*GROi,t*CONi,t
2 + c3*GROi,t
+ c4*PERi,t*CONi,t + c5*PERi,t*CONi,t
2 + c6*PERi,t
+ c7*STAi,t*CONi,t + c8*STAi,t*CONi,t
2 + c9*STAi,t
+ b4*INDi,t + b5*GROWTHi,t + b6*LEVi,t + b7*SIZEi,t + εi,t
3.5 Control Variables
In line with Thomsen and Pedersen (2001) and Y-C Lin et al. (2008), we have three control variables.
The first control variable is GROWTH, which is the profit growth rate of the firm in each year. It is defined as the annual incremental value of profit divided by the lagged profits. It is obvious that growth rate is directly related to the economic performance.
High growth rate often means the firm is enjoying good performance and having growth opportunities, and vice verse.
The second control variable is leverage (LEV). The financial leverage ratio of the firm is defined as debt-to-assets ratio. It is calculated as the book value of all the liabilities in the firm in the end of the year, divided by the firm’s total assets in the end of the year. It can be relevant for two reasons. First, many literatures addressed that firm’s leverage ratio will influence its firm performance (Harris et al. 1991; Y-C Lin, 2008).
Second, leverage always works as a mechanism, which can limit the manager’s misbehavior. Usually it help avoiding the abuse of free cash flow in the firm and reducing the opportunities of overinvestment (Jensen, 1989; Williamson 1985).
- 23 -
The last control variable used is to control the size of the firm. SIZE is measured of the natural logarithm transformation of the total assets of the firm in the end of the year. The size of the firm is often a proxy of the potential economies of scale and scope of the firm (Yoshikawa et al., 2005).
3.6 Summary
To sum up, Table 1 gives the list of the variables and its definition in the model.
Table 1 List of variable definitions
Variables Definition
Return on assets (ROA) EBIT/Total assets
Ownership Concentration (CON) Shares owned by the largest owner (%) Categorical variable (dummy for each):
STA = state owner
PER = individual person owner Ownership type (ID)
(the type of the ultimate owner of the largest owner)
GRO = group owner
Categorical variable (dummy):
C1 = spinning industry Industry (IND)
C5 = electronic industry Growth of profit (%) (GROWTH) incremental sales/lagged profits Leverage (LEV) all liabilities/all assets
Firm size (SIZE) natural logarithm of the total assets
IV Results
4.1 Data description
Table 2 shows the descriptive statistics of the variables. First, we turn to the ownership concentration ratio. The smallest percentage the largest shareholder own is less than 10%, while the largest one is more than 70%. Its mean and median values are both around 35%. However, expect concentration ratio, other variables varies a lot since they have relative large standard deviation, for example, Growth rate. The maximum value of firm’s leverage ratio is 1, which means that debt is the main source of financing of the firm, it hardly issue any shares. In the table, we also show the statistics of firm size. From both normal size, which is in billion and the natural
- 24 -
logarithm of the size, we can find that firm size varies a lot.
Table 2 Descriptive statistics for variables
ROA CON GROWTH LEV Normal
Size*
Logarithm SIZE mean 0.93% 36.19% -15.22% 0.47 2462.05 3.13 median 4.68% 34.61% 10.57% 0.48 1230.31 3.09 maximum 252.75% 72.70% 697.24% 1.00 30041.04 4.48 minimum -665.32% 8.02% -3567.33% 0.04 2.77 0.44 Sta. Dev 40.60% 14.40% 244.03% 0.21 3691.72 0.47 ROA = return on assets, CON = ownership concentration ratio, GROWTH = profit growth rate of the firm, LEV = leverage of the firm and SIZE = the natural logarithm of total assets,
*number in normal size is in million.
Table 3 Correlation matrix of all independent variables used in the model Correlation CON CONSQ GROWTH LEV SIZE
CON 1
CONSQ 0.979 1
GROWTH -0.053 -0.054 1
LEV 0.02 0.038 -0.054 1
SIZE 0.049 0.063 -0.044 0.285 1
CON = ownership concentration ratio, CONSQ = the square of CON, Growth = profit growth rate of the firm, Lev = leverage of the firm, Size = the natural logarithm of total assets.
Table 4 Mean value of each variable in five CON groups
Mean CON ROA GROWTH LEV Normal
Size*
Logarithm SIZE
LL 17.11% -7.86% -4.56% 0.48 2615.89 3.15
LH 26.57% 0.63% -8.93% 0.44 2112.60 2.96
MEDIAN 34.42% 6.21% -14.64% 0.48 2610.31 3.17
HL 45.01% 7.08% 32.47% 0.42 2654.52 3.17
HH 58.50% -3.31% -80.63% 0.52 2263.56 3.18
CON = ownership concentration ratio, ROA = return on assets, GROWTH = profit growth rate of the firm, LEV = leverage of the firm, and SIZE = the natural logarithm of total assets,
*number in normal size is in million.
LL = LowLow group, LH = LowHigh group, MEDIAN = middle group, HL = HighLow group, and HH = HighHigh group.
Table 3 shows the correlation between the independent variables. From the Table, we can figure out that none of the variable is highly related to each other, then, we can
- 25 -
conclude that the model will not face a multilinear problem.
The paper divides the total data into five groups, according to the number of ownership concentration ratio. The 75 observations which have the lowest concentration ratio will be grouped into “LowLow (LL)”. Then, the next 75 observations will form the group “LowHigh (LH)”. On the other extreme of the data, the 75 observations which have the highest concentration ratio will be in the group of
“HighHigh (HH)”, and the next 75 observations, which has lower concentration ratio than them, will be the group of “HighLow (HH)”. At last, the remaining 102 observations will in the group of “Median”. If the hypothesis holds, than the average ROA will be lowest in either “LL” or “HH” group, and the highest one will be in either “LH”, “HL” or “MD” group. The Table 4 shows the statistics in each group.
From this table, we can find that ROA is lowest in group LL, and highest in group HL.
Thus, we can preliminary view that the ROA is bell-shaped related to concentration ratio, although it is a bit skewed to the right. However, we can not find a clear trend in the other three variables, especially the leverage of the firm. Except in the HL group, the growth rate has an adverse trend to the ROA, which is a bit surprise. The trend in normal size and natural logarithm size of the firm are different. The reason is that we calculate the mean of natural logarithm in each group, not the natural logarithm of the mean in each group. We think the normal one show a more precious picture. However, its relationship to concentration ratio is unclear. Therefore, from these five groups, we only have a preliminary impact on the relationship.
Then, we would like to turn to the type of the shareholders. The table 5 below will show the mean value, given each industry and each type, and the number of observations in each category. The observations in each industry do not vary a lot;
however, the state owners have the largest observations. The foreign group contains only 11 observations. The mean values of industries only have a slight difference.
Within each type group, this difference is large in the person owner group and foreign group, where the mean value in the electronic industry is much larger than the one in
- 26 -
another industry. Other two groups, state owners and group owners, they have similar mean values in the two industries, where the value is a bit higher in the spinning industry.
Table 5 Statistics Summary based on industry and type
Number PER STA GRO FOR Total
Spinning Industry 45 95 36 7 183
Electronic Industry 42 110 63 0 219
Total 87 205 99 11 402
Mean
Spinning Industry 28.54% 38.84% 37.62% 25.05% 35.54%
Electronic Industry 36.65% 37.65% 35.17% 37.43% 36.74%
Spinning industry = C1, electronic industry = C5, PER = individual person owner, STA = state owner, GRO = group owner, FOR = foreign investors.
Table 6 Distribution of ownership based on different ownership types Number of observations in each type
Distribution
PER STA GRO FOR Total
<10% 2 2 2 8 14
10.1%-20% 16 13 12 3 44
20.1%-30% 23 56 21 0 100
30.1%-40% 16 46 26 0 88
40.1%-50% 18 37 23 0 78
50.1%-60% 9 34 11 0 54
>60% 3 17 4 0 24
Total 87 205 99 11 402
Average 32.45% 38.20% 36.06% 28.48% 36.19%
PER = individual person owner, STA = state owner, GRO = group owner, FPER = foreign investors.
At last, we would like to see how the ownership concentration distributes between different ownership types. As mentioned before, there are four identities in the sample.
The ownership concentration ratio will be divided into seven groups. In each group, we will find the number of observations in that group and finally, we will have the average concentration ratio in each identity group. Table 6 is shows the results.
- 27 -
According to the table above, we can see that not only foreign shareholders are few, but also the number of shares they holds is relatively smaller than other shareholders.
There foreign background may be the main reason. It is state which has the largest percentage of shareholding than other two. However, the average percentage holding in different identities do not vary a lot. Also, we can see that largest shareholders hold shares from 20.1% to 30%, especially individual person and state owners. Therefore, according to the table, we conclude that distribution of ownership in our sample is a bell-shape.
4.2 Regression results
In this part, we show the regression results of the models. Table 7 shows the OLS results of model 1. We first turn to the whole sample. The coefficient for CONSQ is negative and significant, while the coefficient for CON is positive and significant.
These two coefficients show that there is a bell-shape relationship between firm performance and ownership concentration ratio. A Bell-shape means “convergence of interest effect” domains at the lower level of ownership concentration, while
“entrenchment effect” domains at the higher level. Thus, we cannot reject hypothesis 1. The concentration ratio at the peak point is 41.67%. According to Table 6, only 78 firms distribute in the range from 40.1% to 50%. Thus, we can see that not many firms have the optimal concentration ratio. The coefficient of industry dummy in the whole sample is not significant, so we separate the whole sample into two industry samples to have more detailed pictures in different industries. From column 2 and column 3, except GROWTH and LEV, none of the variables are significant. Only constant term has different sign in electronic industry, all other signs of coefficients are same with the whole sample. Therefore, we find an insignificant bell-shape in the two industries. According to the table, we reject hypothesis 4 and indicate that there is no industry effect and the monitoring structure and costs do not vary a lot between the spinning and the electronic industries.
- 28 -
Table 7 Relationship between firm performance and ownership concentration (n =402)
Total Electronic Spinning
C -0.35* 0.017 -0.641
(0.181) (0.114) (0.309)
CON 1.55** 0.779 2.153
(0.657) (0.413) (1.149)
CONSQ -1.804** -1.124 -2.322
(0.838) (0.536) (1.446)
GROWTH 0.04*** 0.020*** 0.106***
(0.008) (0.004) (0.021)
LEV -0.321*** -0.276*** -0.316*
(0.098) (0.065) (0.163)
SIZE 0.075* 0.006 0.118
(0.043) (0.027) (-0.077)
INDUSTRY -0.015
(0.039)
R-squared 10.24% 24.53% 14.68%
Adjusted R-squared 8.89% 22.39% 12.68%
F-statistic 7.514 11.503 7.329
Prob (F-statistic) 0 0 0
The dependent variable is firm performance, which is represented by ROA. C = constant term, CON = ownership concentration ratio, CONSQ = the square of ownership concentration ratio, GROWTH = profit growth rate of the firm, LEV = leverage of the firm, SIZE = the natural logarithm of total assets, INDUSTRY = industry dummy.
*Significant at 0.10 confidence level using a two-tailed test, ** Significant at 0.05 confidence level using a two-tailed test, *** Significant at 0.01 confidence level using a two-tailed test
Table 7 shows that GROWTH is positively correlated with firm performance at the 1% significance level in all three cases. This suggests that the higher the growth opportunities of the firm, the better its performance. SIZE is positively related to firm performance; however, it is only significant at the 10% level in the whole sample. It is insignificant in both spinning and electronic industry. Generally speaking, table shows that the larger the firm, the better the company’s performance. LEV is negatively correlated with firm performance. It is significant at the 1% significance level in the whole sample and spinning industry and at 10% significance level in electronic industry. This suggests that the higher the level of leverage, the worse the performance of the firm. It is reasonable because the higher the leverage level, the less shares issued in the firm, the higher the probability that the concentration level is low for largest shareholders, and hence, the worse the firm performance.
- 29 -
Table 8 Relationship between firm performance and ownership types (n=391)
Total State Person Group
C -0.107 -0.818** -0.173 0.075
(0.179) (0.408) (0.120) (0.142)
CONGRO -0.467 -0.670
(1.281) (0.498)
CONGROSQ 0.620 0.856
(1.665) (0.645)
GRO 0.095
(0.259)
CONPER 0.195 0.210
(1.399) (0.496)
CONPERSQ -0.138 -0.164
(1.990) (0.706)
PER -0.014
(0.251)
CONSTA 3.906*** 3.860***
(0.992) (1.348) CONSTASQ -4.461*** -4.409***
(1.206) (1.638)
STA -0.765***
(0.221)
GROWTH 0.040*** 0.039*** 0.047*** 0.074***
(0.008) (0.012) (0.010) (0.012)
LEV -0.315*** -0.366** -0.245*** -0.277***
(0.099) (0.180) (0.087) (0.076)
SIZE 0.087** 0.082 0.093*** 0.067***
(0.044) (0.088) (0.034) (0.033)
INDUSTRY -0.011 -0.013 -0.007 -0.019
(0.040) (0.079) (0.032) (0.030)
R-squared 12.86% 10.86% 35.57% 34.55%
Adjusted R-squared 9.94% 8.16% 30.74% 14.68%
F-statistic 4.406 4.021 7.361 8.093
Prob(F-statistic) 0 0 0 0
The dependent variable is firm performance, which is represented by ROA. C = constant term, CON = ownership concentration ratio, CONSQ = the square of ownership concentration ratio, PER = individual person owner, CONPER = CON*PER, CONPERSQ = CONSQ*PER, STA = state owner, CONSTA = CON*STA, CONSTASQ = CONSQ*STA, GRO = group owner, CONGRO = CON*GRO, CONGROSQ = CONSQ*GRO, GROWTH = profit growth rate of the firm, LEV = leverage of the firm, SIZE = the natural logarithm of total assets, INDUSTRY = industry dummy.
*Significant at 0.10 confidence level using a two-tailed test, ** Significant at 0.05 confidence level using a two-tailed test, *** Significant at 0.01 confidence level using a two-tailed test
- 30 -
Now we want to test the type effects. We only consider three types of owners: state owners, individual person owners and group owners. Thus, the data base will reduce to 391 observations according to Table 5 and Table 6. The OLS regression results of model 2 are in Table 8. Table 8 shows the impact of different owners on firm performance and on the relationship between firm performance and its ownership concentration ratio. Column 1 gives general results. Coefficient on GRO is positive and insignificant, on PER is negative and insignificant, and on STA is negative and positive. Therefore, generally we cannot reject hypothesis 2. Although the signs in GRO and PER are opposite to what we proposed, we cannot reject hypothesis 2b and 2c, due to the insignificant coefficient. Also, we cannot reject hypothesis 2a, since the coefficient is negative and significant. We also separate the whole sample into three type groups. The results on each group are shown in column 2 to 4. The sign and the significance in each group are constant with those in column 1. Therefore, we confirm our conclusion that hypothesis 2 cannot be rejected.
We then turn to the coefficients on the interaction term in column 1. Coefficients on group owners show a U-shape relationship, however, it is insignificant. Coefficients on individual person owners indicate that there is a bell-shape relationship; however, it is also insignificant. Only the coefficients on state owners are significant, which address the relationship between firm performance and state owners’ concentration ratio is bell-shaped. Conclusions on three owner types are supported in column 2 to 4, when each type is separated. Therefore, we cannot reject hypothesis 3 according to the regression results. We cannot reject hypothesis 3b and 3c due to insignificancy, and hypothesis 3a due to the supporting results. The conclusion that state owners have negative impact supports the prior literature reviews, which illustrate that the untradable state’s ownership right, lack of economic incentive and multiobejectives of state shares will generate poor performance. The control variables, GROWTH and LEV show similar results related to Table 7, while SIZE is much more significant in Table 8. GROWTH is significant at 1% significant level in all four cases. LEV is significant at either 1% or 5% significant level. Compared to Table 7, SIZE is significant at 1% level in group owners and individual person owners and significant
- 31 -
at 5% level in the whole sample. However, only the results on state owners are meaningful and SIZE in that model turns to be insignificant, just like Table 7.
4.3 Robustness tests
4.3.1 Other regression models
We have used the OLS regression model to test the relationship. However, it is insufficient. The reason is that the database is panel data, which cannot be simply tested by OLS. Thus, we introduce some other models here. OLS is called independently pooled panels, which ignore the unique characteristic of the dependent variables. To have a more precise picture for panel data, we should use fixed effect models and random effect models on panel data. The difference between the two models is that whether the unique and time constant characteristics results from random variation or not. Generally, a random effect model is more intense than a fixed effect model. However, we can use the Hausman test to decide which model should be used. Both cross-sectional and period will have fixed and random effects. We first do panel regression on the whole sample. Table A and Table B in the Appendix show the panel regression results on Model 1 and Model 2, respectively.
We can only do cross-sectional random effect model in the sample. The reason is that the model contains industry dummy, which is constant over the times. This will make the fixed effect model impossible. Also, we can only do period fixed effect model, since year effect is an obvious effect, which do not vary across the time. We also put the outcomes of simple OLS regression in the tables. Compared with the prior results, the results do not change if we use different models, and hence we can conclude that the results are robust in both simple OLS and panel regressions. Hypothesis 1, which illustrate that there is a bell-shaped relationship between ownership concentration ratio and firm performance, cannot be rejected.
We also separate the database into four sub samples, based on years. Data in the same year will form a group, and hence we have four sub samples: 2004, 2005, 2006 and 2007. We are aiming to find out whether the bell-shaped relationship is constant in
