Internal Control, Double Agency Costs and Corporate Social Responsibility

With the development of economy, corporate social responsibility has been paid more and more attention. This paper examines the relationship between internal control and corporate social responsibility, and examines whether double agency costs act as intermediary roles. The research results show that effective internal control can improve the performance of corporate social responsibility and reduce the double agency costs. The first type of agency cost plays a part of intermediary role in the relationship between internal control and corporate social responsibility, while the second type of agency cost has no intermediary role.


Introduction
Since the 21st century, China's economy has experienced rapid growth, but various social problems have also erupted. "Sanlu milk powder", "ShenZhen Foxconn employee suicides", "the Changsheng vaccine" and the recent chemical plant explosion incident in XiangShui county are microcosm of China's frequent incidents of corporate social responsibility. Environmental pollution, production accidents, food safety and employee exploitation show that some enterprises in China have serious defects in social responsibility. At the same time, some enterprises are doing well in social responsibility: Cosco group, through its internal charity foundation, has made generous donations and actively rescued in the snow disaster in south China and Wenchuan Earthquake. It has continuously devoted itself to social welfare and designated assistance to backward areas and has been awarded commendations and awards by the Chinese ministry of civil affairs for many times. Its charity foundation also has been rated as a 4A social organization by the Chinese ministry of civil affairs. What factors lead to such a huge gap in corporate social responsibility? Studies have shown that many factors, such as external environment, internal governance and senior executives' values, have an impact on corporate social responsibility. Internal control, as the key and basic system of modern enterprise governance, infiltrates into all aspects of the enterprise, which is bound to have an important impact on the implementation of corporate social responsibility. In addition, the separation of the two rights is a common feature of modern enterprises. The principal-agent problem is getting more and more attention, and its impact on corporate social responsibility cannot be ignored.
Most of the existing researches are based on internal control or agency costs alone, and examine the relationship between it and corporate social responsibility. In fact, many studies have proved that internal control, corporate social responsibility and agency cost are related to each other. Internal control plays an active role in reducing agency costs and improving corporate social responsibility performance. There are few studies that include the three in the same analytical framework. This leads to the effective path of internal control to corporate social responsibility which has not been fully explored. Besides, there is an obvious phenomenon of equity concentration in Chinese companies and the incidents of major shareholders abusing their power to extract illegal interests from minority shareholders occur frequently. The relationship between the second type of agency problem (major shareholders and minority shareholders) and internal control and corporate social responsibility is also worth exploring. Therefore, based on the existing research and empirical conclusion, this paper analyzes the relationship and mechanism between internal control, dual agency cost and corporate social responsibility, which deepens the understanding of the effective path of internal control. According to the empirical conclusions, this paper puts forward relevant suggestions for the development of internal control and the fulfillment of corporate social responsibility in China.

Internal Control and Corporate Social Responsibility
According to the traditional theory, shareholder profit or corporate profit maximization is the sole responsibility of an enterprise. However, with the continuous development of economy and theory, more and more scholars believe that the corporate responsibility is more than that. Davis (1960) proposed the "iron law of responsibility" [1]: an enterprise does not exist in a vacuum, but in social relations. The production and expansion of an enterprise cannot do without the support of other stakeholders in the society. When an enterprise enjoys its power, it should assume corresponding social responsibilities. Corporate Social Responsibility (CSR) was first formally proposed by British scholar Oliver Sheldon [2] [4]. In addition, the relationship between internal control and CSR is not a simple one-way action, but a two-way interaction, they promote each other [5]. Internal control promotes the fulfillment of CSR by improving business performance and strengthening CSR supervision [6]. Based on the perspective of spillover effect, Li (2014) empirically tested that internal control has a significant positive effect on the fulfillment of CSR, and this effect is stronger in state-owned controlled enterprises [7]. From the perspective of "meta-definition", Li and Xiao define corporate social responsibility as follows: under specific institutional arrangements, enterprises effectively manage the impact of their own operations on society, stakeholders and the natural environment, and seek to maximize social welfare within the expected duration [8]. Internal control, as the basis and focus of internal governance, is the supervision mechanism generated by enterprises in order to max- which is used to guide internal control to ensure the fulfillment of corporate social responsibility. It puts "safe production, product quality, environmental protection and resource conservation, promotion of employment and protection of employees' rights and interests" into the internal control construction, and clearly regards internal control as a means to promote CSR. Therefore, hypothesis 1 is proposed in this paper.
H1: When other conditions are the same, good internal control can promote the implementation of CSR.

Internal Control and Dual Agency Costs
The separation of control and ownership is a typical feature of modern enterprises. This relationship conforms to the trend of modern enterprise development and also creates agency problems. Jensen and Meckling formally proposed the concept of agency cost in 1976 [9], analyzing the first type of agency problem between managers and shareholders. They believe that the agency cost consists shareholders. They pointed out that the shareholding structure of enterprises in many countries is not highly dispersed but there are several major shareholders.
At this point, the major shareholder becomes an agent, who exercises the power of the entire company's shares to influence the company's decision-making in order to encroach on the interests of minority shareholders [10]. As the internal rights protection mechanism, internal control strengthens the information transmission within the enterprise, effectively guarantees the performance of the contract, reduces the misconduct of the enterprise, and thus produce protection effect of the minority shareholders [11] [12]. And the integrated management of internal control and external media supervision can reduce management agency costs and protect the interests of stakeholders more effectively [13]. Yang et al. (2009) found improving the quality of internal control helps to curb the two types of agent costs [14]. Lin et al. (2012) found the internal control of different property rights has different effects on different agency problems [15]. The root cause of the two types of agency problems is the inconsistency of the objective function between the principal and agent caused by information asymmetry and non-convergence of interests. Lv et al. (2008) pointed out that although the implementation subjects and specific methods of benefit encroachment may have different forms, the essence is that the interest groups who hold the residual control rights of the company seek to realize their own interests by exercising the residual control rights. In this sense, the core of restricting interest encroachment is to limit the abuse of residual control through a series of internal and external institutional arrangements [16]. Internal control is a kind of control arrangement and control mechanism, which can effectively guarantee the performance of the enterprise contract, make up for the incompleteness of the contract and lock the residual control right in the "cage". In addition, the internal control optimizes the interests of the principal and the agent by designing an effective incentive mechanism. It also alleviates the degree of information asymmetry and reduces the possibility of opportunistic accounting choices by supervising the reliability of financial accounting information. Therefore, this paper proposes hypotheses 2a, 2b.
H2a: when other conditions are the same, good internal controls can reduce the cost of the first type of agency.
H2b: when other conditions are the same, good internal controls can reduce the cost of the two types of agency.

Internal Control, Dual Agency Cost and Corporate Social Responsibility
Enterprises need to invest resources in fulfilling their social responsibilities. These inputs tend to be rewarded in the long run, and may even have a negative return on corporate performance in the short term [17], while the agent's interest interval is often short-term. The compensation of enterprise managers is often linked to the short-term performance of the company. When there is a serious agency conflict with the shareholders, the manager is likely to reduce the corresponding social responsibility expenditures to maintain the market competitiveness and business performance, even manipulate the accounting surplus and reduce information disclosure. When there is a serious agency conflict between the major shareholder and the minority shareholders, it will intensify the behavior of the major shareholder about invading the company's assets and harming the interests of the minority shareholders to pursue their own interests.
At this time, the major shareholders will not pay attention to the overall needs of other stakeholders, resulting in low levels of CSR. They even cover up their behavior by reducing the disclosure of relevant information. Although the market supervision environment is becoming more stringent, the behavior of controlling shareholders may not be effectively constrained [18]. Internal control, as the most basic institutional arrangement in internal corporate governance and daily operations, infiltrates into every aspect of corporate life. Good internal control can constrain the behavior of managers and major shareholders, ease information asymmetry, strengthen the supervision of the principal on the agent, alleviate the two types of agency problems, and improve the fulfillment level of CSR.
Li et al. (2019) concluded that internal control can alleviate the first type of agency conflict to promote corporate social responsibility [19]. Therefore, this paper proposes the hypotheses 3a, 3b.
H3a: When other conditions are the same, internal control promotes corporate social responsibility performance by reducing the cost of the first type of agency, that is, the first type of agency costs play a mediating role.
H3b: When other conditions are the same, internal control promotes corporate social responsibility performance by reducing the cost of the second type of agency, that is, the second type of agency costs play a mediating role.

1) Interpreted variables-CSR
There are two main methods for measuring CSR in the existing research: First, analyze the dimensions of CSR and use the relevant data to build a CSR rating through calculation formulas [20]. Second, use the CSR rating of third-party professional organizations to measure the level of fulfillment of social responsibility. Based on the research method of Jia [21], this paper takes the CSR rating score published by HeXun.com as an index to evaluate the performance of CSR.
The higher the overall score, the better the CSR performance.

2) Explanatory variables-internal control
Shenzhen Dibo's internal control and risk management database integrates the five elements of internal control to objectively evaluate the internal control of China's listed companies. In recent years, Dibo's internal control index of Chinese listed companies has been widely used in related research. Therefore, by referring to the research methods of Quan et al. [22], this paper takes Dibo's internal control index of Chinese listed companies as an index to evaluate the quality of internal control. The larger the value, the better the internal control quality.

3) Intermediary variables-dual agency costs
For the first type of agency cost, according to the calculation method of Ang [23] et al., the agency cost between managers and shareholders is measured by the ratio of management fees to operating income. The higher the ratio, the more serious the first type of proxy conflict.
For the second type of agency cost, the most common way for major shareholders to occupy the interests of minority shareholders is to occupy the funds of listed companies. Therefore, this paper draws on the practice of Wei et al. [24] to measure the agency cost between major shareholders and minority shareholders by the ratio of other receivables to total assets. The larger the ratio, the more serious the second type of proxy conflict.

4) Control variables
Based on the existing research, this paper selects the enterprise size, profitability, executive compensation, independent director ratio, financial leverage, enterprise growth, equity concentration, and nature of property as control variables. At the same time, this paper also controls the annual effect and industry effect. The specific definition and calculation method of all variables are shown in Table 1.

Model Design
In order to verify the hypothesis of this paper, the following model is con-

Descriptive Statistics
Descriptive statistics of the study variables are shown in Table 2 Table 3 shows the results of correlation analysis of variables. The correlation coefficient of the enterprise size and financial leverage is 0.525, indicating that there may be multiple collinearity. After calculating the variance inflation factor, it is found that the mean value of variance inflation factor is 1.38 and the maximum value is 2.36, which is much lower than 10. Therefore, the possibility of severe multicollinearity among variables is basically ruled out. In addition, internal control and CSR are positively correlated at a significant level of 1%, which preliminarily confirms hypothesis 1. Internal control and two types of agency costs are both negatively correlated at a significant level of 1%, preliminarily confirming hypothesis 2a and 2b. Two types of agency costs are negatively correlated with CSR at a significant level of 1%, which preliminarily verifies the hypothesis of mediating effect in this paper.

Regression Results and Discussion
In this paper, the F-test, the likelihood ratio test and the Hausmann test are used to finally adopt the fixed effect model for regression. Moreover, considering the heteroscedasticity of long panel data, all regressions in this paper use robustness estimates and Table 4  [25] to test the mediating effects of two types of agency costs. The first step is to test the coefficient 1 α of model 1, which is significantly not zero. The second step is to test the coefficient 1 β of the model 2 and the coefficient 2 η of the model 3, both of which are significantly not zero. In the third step, the coefficient 1 η of the model 3 is tested, which is also significantly not zero, and 1 2 β η * is the same symbol as 1 η , indicating that the first type of agency cost plays a partial intermediary role in the relationship between internal control and CSR. The ratio of the mediating effect to the total effect is ( ) , which is about 4%, hypothesis 3a is verified. We use the same method to verify whether the second type of agency costs have a mediating effect.
Step one, 1 α is not significantly zero.
Step two, the coefficient 1 β ′ of the model 4 is not significantly zero while the coefficient 2 η′ of the model 5 does not pass the significance test.
Step three, the mediation effect of AC2 was tested with Bootstrap method, referring to Wen's original steps. The results are shown in Table 5. The mediating effect failed to pass the significance test and hypothesis 3b failed to be verified.

Robustness Test
The robust estimation method is used for regression in this paper, which guarantees the reliability of the conclusion to a certain extent. In addition, this paper also performs logarithmization on the interpreted variable CSR and explanatory variable IC and uses the ROE and the first major shareholding ratio TOP1 as the surrogate indicators of profitability and equity concentration in the control variables. The results of regression (shown in Table 6) and mediating effect test are not substantially different from the above.   Table 3 shows Pearson correlation analysis. *, ** and *** indicate significant at the levels of 10%, 5% and 1%.

Research Conclusions and Recommendations
This

Conflicts of Interest
The author declares no conflicts of interest regarding the publication of this paper.