An Investigation of the Transmission Mechanism of Executive Compensation Control to the Operating Performance of State-Owned Listed Companies

Abstract

Salary control is an effective mechanism used to modulate executive incentives and behaviors in many state-owned listed companies (SOLCs). This is especially true in China considering the vast number and scale of SOLCs. To gain a deeper understanding of the effect of salary control on operating performance of SOLCS, this study investigated salary control policy introduced by the Chinese government in 2015 using data acquired from Shanghai-Shenzhen A-share listed companies from 2010–2017. It was identified that primarily, executive compensation regulation leads to the decline of the operating performance of state-owned listed companies, and the impact of salary control is mainly transmitted through executive behavior. Second, there has been no significant change in the level of OJC of executives before and after pay control and thirdly, salary control reduces the actual effort of state-owned listed company executives, which is reflected in the high level of investment behavior such as the reduction of investment level and the reduction of the frequency of mergers and acquisitions. The above conclusions show that salary control is more likely to lead to a decline in business performance through the “lazy politics” of executives.

1. Introduction

The Chinese government has always implemented controls on the remuneration of senior executives in state-owned listed companies (SOLCs). The establishment of the State-Owned Assets Supervision and Administration Commission of the State Council in 2003 and the 2008 financial crisis made people pay more attention to the issue of excessively high executive salaries. The first salary limit order (SLO) was issued in 2009, and in 2014, the second SLO, the “Remuneration System Reform Plan for Heads of Centrally Managed Enterprises” was reviewed and approved. This was aimed at limiting the remuneration and benefits of state-owned enterprises (SOEs), especially those of the heads and senior managers of enterprises directly under the central government. This policy has aroused people’s discussion on executive compensation, thus triggering scientific inquiry [1,2,3]. The second SLO has been implemented for more than four years, and thus it can be asked if this way of controlling executive compensation through administrative means has affected the operating performance of SOEs? If there is an impact, how does the regulation of executive compensation affect company performance by influencing executive behavior? These issues are worth exploring.

Regarding the relationship between on-the-job-compensation (OJC) and operating performance, there are two main views. One is that OJC will increase the company’s costs and therefore, adversely affect operating performance. This kind of literature includes Jenson and Meckling’s belief that OJC is one of the principal-agent conflicts that leads to the reduction of corporate value [4]. Liu et al. [5], Zhou et al. [6] and Hou [7] have all found that OJC is negatively correlated with company operating performance. Under China’s special SOE executive compensation control system, Chen et al. found that the lower the relative salary, the more motivated the executives of SOEs to seek OJC instead of salary [8]. Huang also confirmed the existence of a substitute relationship between the two. Another view is that OJC can effectively motivate executives, and the improvement of work efficiency can improve the company’s operating performance [9]. This type of literature includes Chen et al. who believed that OJC was positively related to corporate value, and that OJC may be the second-best choice for motivating executives [10]. According to the above research, the 2015 SLO will lead to an increase in the on-job consumption of SOEs. However, at the same time as salary control, the central government’s anti-corruption efforts are increasing. Mei and Yang found that the introduction of the “eight regulations” effectively curbed the growth of OJC [11], therefore, due to political factors, the ways for SOE executives to obtain implicit incentives through OJC was restricted.

Kang et al. suggested that acquisitions have led to significant increases in executive compensation but additionally, although the effect of acquisitions on privately-owned publicly listed companies (PLCs) was statistically insignificant, they nevertheless had a significant effect on state-owned PLCs [2]. Jiang and Zhang investigated the effect of regulatory restriction on executive compensation in Chinese SOEs and found that restrictions were negatively associated with a firm’s accounting performance [12]. More recently, however, based on the Generalized Method of Moments (GMM) estimation approach of 860 non-financial firms on Chinese Stock Exchanges from 2004–2018, Rehman et al. found a positive and significant association between corporate profitability and executive pay [13]. Shao et al. found that regulation has had a significant effect on compensation gaps in central SOEs and a contagion effect on local SOES, but not for non-SOEs [14]. Additionally, central, and local SOEs experience reduced firm performance after the compensation regulations, but not the non-SOES, which then indicates that compensation regulation does not have favorable economic consequences for directly affected central SOEs and indirectly affected local SOEs. Flynn suggested, based on a descriptive review of executive compensation law and shareholder protections in China (alongside France and the United States), that if a country regulates the use of real options as compensation, then that country would also be more likely to have strong shareholder protection laws [15]. Additionally, as it is likely that these countries are relying on the Crowding Out Theory, a balance should be made between low pay but encourages executives to work harder, and high pay which disincentives executives from pursuing alternative forms of compensation which may ultimately harm shareholders.

Based on the available literature, it can be understood that there is a relationship between executive compensation, corporate investment, and mergers and acquisitions (M and A), in addition to the relationship between investment and M and A. For example, Ning and Qiu focused on listed companies from 2001–2010 and found that executive currency compensation can improve investment efficiency and thus contribute to the improvement of company performance. However, they also identified that senior executives’ OJC is significantly negatively correlated with inefficient investment. Xin et al. found that in listed companies from 2000–2004, excessively low executive salaries led to excessive investment [16]. Similarly, Xu used listed companies from 2004–2010 and found that the intensity of company investment is inversely proportional to the level of executive compensation [17]. Liu and Feng found that investment behavior plays an intermediary role between the salary gap within executives and operating performance, and the widening of the salary gap within executives can increase the level of investment of the company, thereby improving the operating performance of the company [18]. Mergers and acquisitions are also major long-term investment behaviors that affect corporate value, and executive compensation is one of the influencing factors driving corporate mergers and acquisitions. According to managerialism, the scale of a company can be expanded rapidly after mergers and acquisitions, and executives can gain a greater reputation and can seek higher salaries for themselves. Research by Zhang et al. also show that the frequency of corporate mergers and acquisitions is positively correlated with the salary levels of corporate executives [19]. Li, Mao, and Zhao discussed the relationship between mergers and acquisitions and executive income [20]. The study found that executives can obtain higher monetary compensation and OJC when they initiate mergers and acquisitions, and executives initiate mergers and acquisitions more for seeking OJC. Research on listed companies that have successfully undergone mergers and acquisitions conducted by Shi et al. showed that the higher the monetary salary of the management, the easier it was to frequently initiate M and As [21]. Within the hospitality sector, Camilleri [22] showed from 462 research participants that employer stakeholders could trigger their businesses to engage in ethical behaviors, responsible human resources management, and to invest in environmentally friendly initiatives. In an earlier study, Camilleri [23] showed that corporations and large entities disclose financial and non-financial information in integrated reports, and as a result, organizational stewardship could be improved and reinforcing their legitimacy with other institutions and stakeholders. Regarding the research on pay control, most scholars base their work on the frictionless management labor market theoretical model analysis, and some scholars are concerned about the effectiveness of the “say on pay” mechanism in the United States and the United Kingdom. However, these are not exactly the same as the special salary control of Chinese SOEs. Most of the domestic scholars’ research on salary control has focused on the direct impact on executive compensation, business performance, and OJC, but there are few studies on the ways to affect business performance. This article not only analyzes how compensation control affects business performance, but also studies the transmission mechanism of how compensation control affects business performance by changing the behavior of executives. This study examines the behavior of executives to test whether the enthusiasm of executives to work hard has changed, and whether the investment level and frequency of mergers and acquisitions of state-owned listed companies have changed before and after compensation controls. This research has two primary contributions. First, in terms of research content, existing studies on the economic consequences of salary control mainly focus on the aspects of executive incentives. Generally, there is less literature on executive behavior. This article discusses the impact of compensation control on the impact of executive behavior. Second, in the research sense, paying attention to the behavior of executives that can reflect the actual effort and effort of executives to a certain extent, which provides a new way for the government to formulate performance evaluation indicators. Several questions are posed and answered: does the 2015 SLO have an impact on business performance of SOEs? If there was an impact, was it caused by executives’ in-service consumption, or by inaction by senior management? The remainder of this paper is structured as follows. Section 2 describes the hypotheses to be tested in this study. Section 3 provides the data and methodology. Section 4 describes the model and main findings of this study. Section 5 discusses the primary mechanisms of compensation control, and Section 6 gives the conclusion.

2. Hypotheses

2.1. The Impact of Executive Compensation on the Operating Performance of SOLCs

To date, the 2015 SLO has been implemented for over five years and a reasonable question to ask is what has been the economic effect of the policy? Using a complete set of financial indicators from 2010–2017, this question can be answered in conjunction with others: does the SLO have an impact on incentives for executives and is there any effect on executive behavior? The executive incentive explored in this paper includes two types of incentive: explicit incentive represented by salary performance sensitivity and implicit incentive represented by OJC; executive behavior selects financial decision-making behavior that can reflect the efforts of executives to a certain extent, including investment level and acquisition rate.

First, it puts forward assumptions about the impact of SLO on the operating performance of SOEs. According to the optimal contract theory, enterprise executives will not automatically seek to maximize the interests of shareholders. Therefore, it is necessary to provide enterprise executives with necessary incentives to encourage enterprise executives to take actions to improve business performance, to maximize the interests of shareholders. One of the incentives to solve the agency problem is to pay according to performance. However, the salary control directly limits the salary of senior executives of SOEs. The 2015-implemented SLO clearly stipulates that the salary of the head of central enterprises shall not exceed 7–8 times the average salary of employees. To some extent, this may reduce the original salary incentive, affect the work enthusiasm of senior executives, and eventually lead to the decline of business performance. Therefore, this paper puts forward the following assumptions:

Hypothesis 1.

SLO has an adverse impact on the operating performance of SOEs.

2.2. The Transmission Mechanism of Executive Compensation Regulation Affecting SOLC Operating Performance

The impact of SLO on OJC. Executives of SOEs will seek implicit incentive mechanism as an alternative when their salary is regulated. The most common implicit incentive is the executives’ OJC. Then, does SLO lead to the change of executives’ OJC, and does OJC play an intermediary effect in the path of SLO affecting the company’s operating performance? According to the relevant literature, the assumptions are as follows:

Hypothesis 2.

SLO does not affect OJC, and SLO does not affect the operating performance of state-owned listed companies through OJC.

The impact of compensation regulation on executive behavior. The salary analysis combining salary income and performance evaluation whether the incentive received by the executives of Chinese listed companies changes before and after the salary control, and whether the salary control affects the business performance of the enterprise through the intermediary effect of OJC. However, this salary incentive or alternative incentive is not completely equivalent to the work enthusiasm of the executives themselves. This is because the efforts of senior executives may not be immediately reflected in the performance indicators, or the accounting performance indicators used for performance appraisal contain more noise [16]. Therefore, this study attempts to test whether these executive behaviors that can reflect the degree of executive effort have changed before and after the policy from the perspective of investment level and acquisition rate and analyze whether these executive behaviors play an intermediary effect between SLO and the operating performance of state-owned listed companies. When executive compensation is regulated, executives may reduce their workload and reduce their efforts. As investment and mergers and acquisitions often involve significant workloads, executives need to spend a lot of time learning new knowledge and bear greater responsibility and pressure, so it can better measure the efforts of executives. According to the synergy theory of M and A value creation, M and A is conducive to the complementarity of resources and other aspects of both sides, creating value. According to the above theories, the remaining hypotheses of this paper are as follows:

Hypothesis 3.

SLO leads to the decline of enterprise investment level, and SLO affects the operating performance of state-owned listed companies through the intermediary effect of investment level.

Hypothesis 4.

SLO leads to the decline of M and A frequency, and SLO affects the operating performance of state-owned listed companies through the intermediary effect of M and A frequency.

3. Data and Methodology

3.1. Data

The purpose of this paper was to explore the impact of the 2015 SLO on the operating performance of state-owned listed companies. To avoid the possible interference of the 2009 SLO, all A-share state-owned listed companies in Shanghai and Shenzhen from 2010 to 2017 were taken as research samples. Whether a listed company was a SOE was judged according to the actual controller of the listed company, and combined with the attribute of the actual controller, it was divided into central enterprises and local SOEs. In total, 7514 samples were finally obtained (excluding ST, ST companies that do not meet the requirements and some companies with missing financial data), including 2562 observations in central enterprises and 4952 observations in local SOEs.

Tobin-Q data were sourced from the WIND database while the purchasing frequency and other variables were from the China Stock Market & Accounting Research (CSMAR) Database. Statistics concerning GDP growth and other data were acquired from the National Statistical Yearbook.

3.2. Variable Description

Pay control: This paper used virtual indicators to measure SLO (LIM). As the second “salary restriction order” was officially implemented on 1 January 2015, this paper set 2010–2014 as 0, indicating that the salary control had not been implemented; the salary control index from 2015–2017 was set to 1, indicating that the salary of senior executives of SOEs was restricted by salary control at that stage.

Business performance: business performance indicators can be divided into accounting performance indicators and market performance indicators. Accounting performance indicators commonly include total asset net interest rate (ROA), return on net assets (ROE), etc., while market performance indicators are often dominated by stock return. Because the market performance index depends on the market performance, it is more disturbed by macro factors; in addition, when the SASAC assesses the business performance of the heads of SOEs, the performance benchmark set is more reference to the accounting performance indicators such as the total annual profit. Therefore, this study finally selected the accounting performance indicator of total asset net interest rate (ROA) to measure the business performance of enterprises.

Other variables. (1) In order to study the impact of SLO on executive OJC, this paper defined OJC (LNPERK) as the implicit welfare obtained by executives by taking advantage of their position. Due to the hidden characteristics of OJC, there is no clearly recognized measurement index in the academic circles. At present, two methods are widely used: one is the method of Chen et al. that uses the total of eight expenses, such as office expenses and travel expenses under the notes of “paying other cash related to operating activities” in the financial statements [8]. However, this method is limited by the quality of information disclosure of listed companies; second, Luo et al. and Wang et al. used the remaining part of the “management expense” subject after deducting bad debt provision, inventory depreciation provision, executive compensation, and amortization of intangible assets as “enterprise OJC” [24,25], then use “enterprise OJC” to regress the relevant variables and obtain the “expected OJC” from the regression equation. The difference between “enterprise OJC” and “expected OJC” is expressed as “abnormal OJC”. This paper mainly draws lessons from the second method and further considers the changes of accounting standards. Bad debt reserves, inventory falling price reserves and intangible assets transferred to production products and assets are no longer included in management expenses. Therefore, the OJC (LNPERK) calculated in this paper was the management expenses minus the total annual remuneration of the management. (2) To study the impact of SLO on investment level, this paper extracted the cash to be paid for the construction of fixed assets, intangible assets, and other long-term assets from the cash flow of investment activities, deducting the net cash received from the disposal of fixed assets, intangible assets, and other long-term assets, and then divided it by the opening balance of total assets as an index to measure investment expenditure (INV). (3) To study the impact of SLO on M and A frequency, acquisition rate (y) refers to the number of M and A announced and finally completed by the sample company in 2010–2017. Data were sourced from the CSMAR database. Mergers and acquisitions included asset acquisition, merger absorption, tender offer, and share transfer under the “restructuring type” of the CSMAR database (CSMAR). At the same time, the transaction status of the sample company was denoted as the “buyer”.

Control variables. The control variables selected in this paper include company size (LNSIZE), asset liability ratio (LEV) and other financial indicators; corporate governance indicators such as dual, the shareholding ratio of the largest shareholder (LSH) and the proportion of independent directors (IDD); executive compensation (LNAPAY) indicator; GDP growth rate at the level of economic environment, etc. The specific variables are defined in

Table 1. Variable Definitions.

Variable	Definition
ROA	Return on total assets, net profit × 2/(opening balance of total assets + ending balance of total assets) × 100%
LIM	Salary control: 0 in 2010–2014 and 1 in 2015–2017
LNAPAY	For executive compensation, the per capita annual salary of the management was taken as the natural logarithm
LNSIZE	Enterprise scale, total assets, ending balance, natural logarithm
LEV	Asset liability ratio, total liabilities at the end of the period divided by the ending balance of total assets
GDP	GDP growth rate
DUAL	If the chairman is also the general manager, the value was 1, otherwise it was 0
LSH	Shareholding ratio of the largest shareholder
IDD	Proportion of independent directors in the number of directors
GROWS	Enterprise growth, total operating revenue, year-on-year growth rate
LNPERK	For OJC, the management expenses shall be deducted from the annual salary of the management
INV	The balance of investment expenditure, cash paid for the purchase and construction of fixed, intangible, and other long-term assets minus the net cash received from the disposal of fixed, intangible, and other long-term assets, and then divided by the opening balance of total assets
TQ	Investment opportunity Tobin Q, the sum of stock market value and liability market value, divided by total assets
CFO	Enterprise cash flow, net cash flow from operating activities/opening balance of total assets
Y	Acquisition Rate, the number of finalized acquisitions made and reported from 2010–2017

Note: From the statistical description of these indicators, the operating performance (ROA) of the sample state-owned listed companies varies greatly, with the minimum value of −76.89%, the maximum value of 47.70% and the standard deviation of 5.65. There are also large differences in asset liability ratio (Lev), and there are insolvent enterprises. There is a greater difference in the concentration of ownership structure, with the largest shareholding proportion of the largest shareholder reaching 93.67%.

. It should be noted that within China, only listed companies have an obligation to publish their financial data, and as a result, only listed companies or SOEs were used as sample data. Additionally, because there are stringent requirements for Chinese listed companies, only the most outstanding companies in an industry were thus most typical of the dataset.

4. Model Construction and Empirical Analysis Results

4.1. Model Construction

The following model is built to test how executive compensation regulation effects the operating performance of SOLCs:

$$\begin{gathered} {\mathrm{ROA}}_{\mathrm{it}}=\mathsf{\alpha }+{\mathsf{\beta }}_1{\mathrm{LIM}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{LNAPAY}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{LNSIZE}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{LEV}}_{\mathrm{it}}+{\mathsf{\beta }}_5{\mathrm{GDP}}_{\mathrm{it}} \\ +{\mathsf{\beta }}_6{\mathrm{DUAL}}_{\mathrm{it}}+{\mathsf{\beta }}_7{\mathrm{LSH}}_{\mathrm{it}}+{\mathsf{\beta }}_8{\mathrm{IDD}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}} \end{gathered}$$

(1)

where the explanatory variables are operating performance (total assets net interest rate; ROA), and the explanatory variables are salary control dummy variable and executive compensation. The SLO dummy variable (LIM) was used to study the impact of SLO on the operating performance of SOEs. If the regression coefficient was significantly negative, it indicated that SLO had an adverse impact on the operating performance of SOEs and Hypothesis 1 was verified. To explore whether SLO affected executive incentive and executive behavior, which ultimately affects the company’s operating performance, these represented both the individual effect and residual term.

4.2. Empirical Analysis Results

In this section, the Petersen fixed effect regression model of Equation (1) was used with a robust clustering standard error to reduce the impact of possible heteroscedasticity [26]. The regression results of the impact of SLO on the operating performance of SOEs are given in

Table 2. Regression Results of the Impact of Executive Compensation Regulation on the Operating Performance of SOLCs.

Explanatory Variable	Full Sample	Central Enterprises	Local Enterprises
LIM	−0.804 *** (0.142)	−0.642 ** (0.282)	−0.873 *** (0.158)
LNAPAY	2.271 *** (0.227)	2.564 *** (0.428)	2.080 *** (0.273)
LNSIZE	1.596 ***	1.378 ***	1.767 ***
	(0.249)	(0.358)	(0.354)
LEV	−0.172 ***	−0.176 ***	−0.168 ***
	(0.015)	(0.021)	(0.021)
GDP	1.038 ***	1.025 ***	1.075 ***
	(0.065)	(0.105)	(0.085)
DUAL	0.115	0.971	−0.124
	(0.300)	(0.656)	(0.334)
LSH	0.025 *	−0.003	0.042 **
	(0.015)	(0.028)	(0.018)
IDD	−3.463 **	−4.370 **	−3.499
	(1.579)	(2.196)	(2.248)
_cons	−61.829 ***	−59.747 ***	−63.827 ***
	(5.747)	(7.763)	(8.366)
Obs.	7514	2562	4952
R-squared	0.226	0.210	0.237

Note: Values within parentheses are the robust standard deviation of a company’s clustering, *** represents the 1% significance level, ** represents the 5% significance level, and * represents the 10% significance level.

. The first column is the result of the regression of all SOEs. The second and third columns are the results of grouping regression for central enterprises and local SOEs, respectively.

In Table 2, it can be observed that that the coefficient of LIM is significantly negatively correlated in both the full sample regression and the grouping regression equation between central enterprises and local SOEs at the 5% significance level. This shows that the 2015 SLO has had an adverse impact on the operating performance of China’s SOEs, and Hypothesis 1 has been verified. Comparing the coefficient of LIM in the grouping of central enterprises and local SOEs, the operating performance of local SOEs was more adversely affected by SLO than that of central enterprises. The regression coefficient of executive compensation (LNAPAY) was significantly positive at the level of 1% in the three results, which indicates that there was a strong positive correlation between executive compensation and performance of state-owned listed companies in China. In terms of control variables, the company size of state-owned listed companies was significantly positively correlated with operating performance, and the asset liability ratio was significantly negatively correlated with operating performance. The GDP growth rate and the operating performance of state-owned listed companies were significantly positive at the level of 1%, which is in line with expectations, indicating that the macroeconomic cycle can indeed affect the operating performance of enterprises.

5. An Investigation on the Primary Mechanisms

5.1. The Impact of SLO on OJC

Based on the research on OJC such as those conducted by Chen [8] and Chen [10], a model to study the impact of SLO on OJC was built as follows:

$$\begin{gathered} {\mathrm{LNPERK}}_{\mathrm{it}}=\mathsf{\alpha }+{\mathsf{\beta }}_1{\mathrm{LNAPAY}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{LIM}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{LNSIZE}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{LEV}}_{\mathrm{it}}+{\mathsf{\beta }}_5{\mathrm{GROWS}}_{\mathrm{it}} \\ +{\mathsf{\beta }}_6{\mathrm{DUAL}}_{\mathrm{it}}+{\mathsf{\beta }}_7{\mathrm{LSH}}_{\mathrm{it}}+{\mathsf{\beta }}_8{\mathrm{IDD}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}} \end{gathered}$$

(2)

If the regression coefficient is not significant, it shows that the salary regulation has no impact on the on-the-job consumption of executives. The first half of Hypothesis 2 was verified. The fixed effect regression model was used on Model 2, with results shown in

Table 3. Regression Results of the Impact of Executive SLO on OJC.

Explanatory Variable	Full Sample	Central Enterprises	Local Enterprises
LNAPAY	0.091 ***	0.065 **	0.091 ***
	(0.016)	(0.026)	(0.022)
LIM	0.010	0.021	0.006
	(0.013)	(0.021)	(0.016)
LNSIZE	0.676 ***	0.761 ***	0.633 ***
	(0.029)	(0.047)	(0.036)
LEV	0.001	−0.001	0.002 *
	(0.001)	(0.001)	(0.001)
GROWS	−0.000 **	−0.000	−0.000 *
	(0.000)	(0.000)	(0.000)
DUAL	−0.033 *	0.032	−0.045 **
	(0.018)	(0.037)	(0.020)
LSH	−0.001	−0.003	−0.001
	(0.002)	(0.002)	(0.002)
IDD	−0.137	0.044	−0.213
	(0.133)	(0.191)	(0.188)
_cons	2.743 ***	1.409	3.635 ***
	(0.678)	(0.977)	(0.877)
Obs.	7514	2562	4952
R-squared	0.617	0.714	0.557

Note: Values within parentheses are the robust standard deviation of a company’s clustering, *** represents the 1% significance level, ** represents the 5% significance level, and * represents the 10% significance level.

. The first column is the result of the regression for all SOEs in the sample company, and the second and third columns are the results of grouping regressions for central enterprises and local SOEs, respectively. As shown in Table 3, the coefficient of LIM was not significant in the full sample regression equation and the grouping regression equation of central enterprises and local SOEs, which indicates that the SLO policy in 2015 did not change the OJC of executives, and the first half of Hypothesis 2 was confirmed. The possible reason is that during this period, the central anti-corruption efforts increased, so that executives did not obtain implicit incentives through OJC.

It can also be observed that LNAPAY was significantly positively correlated with OJC, which is consistent with the findings of Chen et al. [10]. This may be because the explicit incentive of executive compensation and the implicit incentive of OJC reflect the position level of executives. Executives with higher salary can enjoy more OJC.

Analysis of the intermediary effect of executive OJC. According to the research of Wen and Ye [27], there are three commonly used methods for mediating effect test, namely step-by-step, Sobel, and bootstrap methods. Among these, Baron and Kenny’s stepwise method is the most commonly used, and its test result is stronger than the Sobel test result [28]. However, as the Sobel method relies on the assumption of normal distribution of test coefficients, the bootstrap method is instead used here. Given that the model has been verified, here the impact of SLO on OJC, and the impact of OJC on business performance was examined. If the impact is significant, the indirect effect is established, otherwise it needs to be tested by bootstrap method. However, the results of Model 2 (Table 3) showed that the coefficient of SLO was not significant. Verified by bootstrap method, this paper finds that the indirect effect of OJC does not exist.

Table 4. Bootstrap Test Results of Intermediary Effects on OJC.

	Observed Coef.	Bias	Bootstrap Std. Err.	95% Confidence Level after Bias Correction
r (ind_eff)	0.0090	0.0003	0.0069	−0.0009	0.0283	(BC)
r (dir_eff)	−0.3248	−0.0002	0.1497	−0.6376	−0.0348	(BC)

shows the results of bootstrap 1000 times and that the confidence interval of indirect effect R (ind_eff) includes 0, and the indirect effect are not significant. Therefore, Hypothesis 2 can be confirmed.

5.2. Analysis on the Influence of SLO on Executive Behavior and Its Intermediary Effect

Referring to the research models set by Chen et al. [29], Li et al. [30] the model constructed in this paper to study the impact of SLO on the investment efficiency of SOEs was as follows:

$${\mathrm{INV}}_{\mathrm{it}}=\mathsf{\alpha }+{\mathsf{\beta }}_1{\mathrm{TQ}}_{\mathrm{i},\mathrm{t}-1}+{\mathsf{\beta }}_2{\mathrm{LIM}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{LNSIZE}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{LEV}}_{\mathrm{it}}+{\mathsf{\beta }}_5{\mathrm{CFO}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(3)

Model 3 was used to study the impact of SLO on investment level. If the regression coefficient of is significantly negative, it indicates that SLO leads to the decline of investment level, the first half of Hypothesis 3 is verified. Model 3 selects fixed effect regression, and the regression results of the impact of SLO on investment level are shown in

Table 5. Regression Results of the Impact of SLO on Investment Level.

Variable	Full Sample	Central Enterprises	Local Enterprises
TQ	0.049 ***	0.010 ***	0.055 ***
	(0.012)	(0.002)	(0.012)
LIM	−0.056 ***	−0.035 ***	−0.054 ***
	(0.007)	(0.004)	(0.006)
LNSIZE	0.047 ***	0.022 ***	0.051 ***
	(0.011)	(0.009)	(0.010)
LEV	0.000	−0.000	0.000 *
	(0.000)	(0.000)	(0.000)
CFO	0.086	0.018	0.112
	(0.076)	(0.013)	(0.091)
_cons	−1.119 ***	−0.464 **	−1.223 ***
	(0.271)	(0.187)	(0.244)
Obs.	7173	2435	4738
R-squared	0.520	0.073	0.622

Note: Values within parentheses are the robust standard deviation of a company’s clustering, *** represents the 1% significance level, ** represents the 5% significance level, and * represents the 10% significance level.

. The first column is the result of regression for all SOEs in the sample company, and the second and third columns are the results of grouping regression for central enterprises and local SOEs, respectively.

As can be seen in Table 5, the coefficient of LIM was significantly negative in the regression of the whole sample, the grouping of central enterprises and local SOEs, which means that under the condition that other factors remain unchanged, the SLO policy led to the reduction of investment expenditure of central enterprises and local SOEs. The first half of Hypothesis 3 was confirmed.

Referring to the setting of the Zhang et al. [19], the model constructed to study the impact of SLO on M and A frequency was as follows:

$${\mathrm{Y}}_{\mathrm{it}}=\mathsf{\alpha }+{\mathsf{\beta }}_1{\mathrm{LNAPAY}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{LIM}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{LNSIZE}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{LEV}}_{\mathrm{it}}+{\mathsf{\beta }}_5{\mathrm{LSH}}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(4)

where, if the regression coefficient is significantly negative, it indicates that SLO leads to the decline of M and A frequency, so the first half of Hypothesis 4 was verified. The explanatory variable in Model 4 is the M and A frequency, which is a discrete non-negative integer in a finite interval. The least square method is not suitable for this model. When making descriptive statistics on the variables, it is found that the variance of M and A frequency (Y) is significantly greater than its mean value, that is, there is “excessive divergence” in the M and A frequency of the explained variable, and negative binomial regression is adopted.

Table 6. Regression Results of the Impact of SLO on M and A Frequency.

Variable	Full Sample	Central Enterprises	Local Enterprises
LNAPAY	0.160 ***	0.249 **	0.118 *
	(0.058)	(0.127)	(0.062)
LIM	−1.221 ***	−1.421 ***	−1.138 ***
	(0.087)	(0.171)	(0.100)
LNSIZE	0.112 ***	0.051	0.183 ***
	(0.033)	(0.053)	(0.042)
LEV	0.001	−0.000	0.002
	(0.002)	(0.003)	(0.002)
LSH	−0.004 *	−0.006	−0.004
	(0.002)	(0.005)	(0.003)
_cons	−5.284 ***	−4.869 ***	−6.415 ***
	(0.820)	(1.700)	(0.875)
lnalpha: cons	1.335 ***	1.655 ***	1.129 ***
	(0.063)	(0.105)	(0.076)
Obs.	7514	2562	4952
Wald chi2	230.10	80.26	162.43

Note: Values within parentheses are the robust standard deviation of a company’s clustering, *** represents the 1% significance level, ** represents the 5% significance level, and * represents the 10% significance level.

shows the regression results of the impact of SLO on M and A frequency.

As can be seen in Table 6, the coefficients of LIM were significantly negative, indicating that the frequency of mergers and acquisitions of central enterprises and local SOEs had decreased significantly after the promulgation of SLO policies, which verifies the first half of Hypothesis 4 in this paper. The regression results show that the coefficient of executive compensation (LNAPAY) was significantly positive, which is consistent with the regression results of Zhang et al. [19]. This shows that the higher the salary level of senior executives, the more motivated they are to initiate mergers and acquisitions and expand the scale of enterprises.

Analysis of the intermediary effect of executive behavior. According to the test process proposed by Wen and Ye [25], to verify the second half of Hypotheses 3 and 4 according to Baron and Kenny’s step-by-step method [26], we first tested the impact of SLO on business performance, which has been experienced in the previous article. The second step was to test the impact of SLO on investment level and acquisition rate, which has also been confirmed above. The third step was to test the impact of investment level and M and A frequency on operating performance. Therefore, based on Model 1, add two indicators: investment level (INV) and M and A frequency (y), and build the model as follows:

$$\begin{gathered} {\mathrm{ROA}}_{\mathrm{it}}=\mathsf{\alpha }+{\mathsf{\beta }}_1{\mathrm{LIM}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{INV}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{Y}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{LNAPAY}}_{\mathrm{it}}+{\mathsf{\beta }}_5{\mathrm{LNSIZE}}_{\mathrm{it}}+ \\ {\mathsf{\beta }}_6{\mathrm{LEV}}_{\mathrm{it}}+{\mathsf{\beta }}_7{\mathrm{GDP}}_{\mathrm{it}}+{\mathsf{\beta }}_8{\mathrm{DUAL}}_{\mathrm{it}}+{\mathsf{\beta }}_9{\mathrm{LSH}}_{\mathrm{it}}+{\mathsf{\beta }}_{10}{\mathrm{IDD}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}} \end{gathered}$$

(5)

If the regression coefficient of sum is significant, according to the stepwise method of intermediary effect test, this shows that the investment level and M and A frequency have intermediary effect between SLO and the operating performance of SOLCs. Hypotheses 3 and 4 have been fully verified.

The fixed effect regression model was selected for Model 5, and the regression results are shown in

Table 7. Validating Regression Results of Executive Behavior Mediating Effects.

Variable	Full Sample	Central Enterprises	Local Enterprises
LIM	−0.736 ***	−0.499 *	−0.814 ***
	(0.140)	(0.276)	(0.157)
INV	0.053 **	4.319 ***	0.049 ***
	(0.021)	(1.240)	(0.017)
Y	0.144 ***	0.147 **	0.141 ***
	(0.037)	(0.059)	(0.048)
LNAPAY	2.267 ***	2.599 ***	2.073 ***
	(0.227)	(0.428)	(0.273)
LNSIZE	1.623 ***	1.281 ***	1.798 ***
	(0.250)	(0.365)	(0.355)
LEV	−0.172 ***	−0.175 ***	−0.169 ***
	(0.015)	(0.021)	(0.021)
GDP	1.055 ***	1.003 ***	1.089 ***
	(0.066)	(0.106)	(0.086)
DUAL	0.107	1.012	−0.135
	(0.299)	(0.650)	(0.334)
LSH	0.025 *	−0.007	0.042 **
	(0.015)	(0.028)	(0.018)
IDD	−3.403 **	−4.178 *	−3.487
	(1.581)	(2.211)	(2.252)
_cons	−62.578 ***	−58.110 ***	−64.620 ***
	(5.771)	(7.780)	(8.395)
Obs.	7514	2562	4952
R-squared	0.227	0.216	0.239

Note: Values within parentheses are the robust standard deviation of a company’s clustering, *** represents the 1% significance level, ** represents the 5% significance level, and * represents the 10% significance level.

. The first column is the regression results of all SOEs in the sample company, and the second and third columns are the results of grouping regression for central enterprises and local SOEs, respectively.

For clarification, ROA was used as a performance indicator because it reflects the rate of return on assets, which fully reflects the extent to which SOEs’ assets are fully utilized, and it is also the evaluation indicator of SOEs by the state-owned assets supervision and administration commission of China. Additionally, management fees after the deduction of the annual total compensation management are used as OJC indicators. This is because usually management includes management personnel salary, welfare enterprises, management department for daily produce and other chores. As state-owned assets supervision and administration of SOEs have led to the worsening of state-owned assets management behavior, so the management approach is more greatly negative than M and As. Moreover, as the purchase of assets leads to a deduction of corporate cash flow, this will be able to partly reflect risk management and enthusiasm.

The control variables selected in this paper include company size (LNSIZE), asset-liability ratio (LEV) and other financial indicators. At the corporate governance level, dual (DUAL) corporate governance indicators, such as the shareholding ratio of the largest shareholder (LSH) and the proportion of independent directors (IDD); executive compensation (LNAPAY) indicators; at the level of economic environment, GDP growth rate has an impact on a company’s business performance (ROA). For example, GDP growth rate reflects the economic cycle, and asset-liability ratio reflects the robustness of an enterprise. In addition, these indicators are also commonly used in domestic literature to study business performance.

It can be seen from Table 7 that in both full sample and group regression, there is a significant positive correlation between investment level and operating performance, which shows that investment activities for enterprise fixed assets, intangible assets and other long-term assets can improve the company’s performance. Acquisition rate is also significantly positively correlated with business performance, which verifies the synergy theory of M and A value creation, that is, M and A makes the resources of both sides complement each other and create value.

In the intermediary effect test, from Table 7, the regression results of Model 5 confirm that the investment level and acquisition rate can have a significant positive impact on business performance. The regression results of Model 1 show that for the whole sample of SOLCs, SLO will lead to the decline of operating performance. The results of Models 3 and 4 show that SLO will lead to a significant decline in the investment level and Acquisition rate of SOEs, respectively. The results of Model 5 show that when considering the impact of SLO, investment level and acquisition rate on business performance, the investment level and M and A frequency are significantly positively correlated with business performance, so the indirect effect of investment level and Acquisition rate is significant. The SLO coefficient of Model 5 is still significantly negative, and its absolute value is smaller than that of Model 1, indicating that the indirect effect of financial behavior characterized by investment level and acquisition rate belongs to partial intermediary effect, and Hypotheses 3 and 4 are confirmed.

In addition, the grouping test is also a kind of robustness test. It can be seen from Table 2, Table 5, Table 6 and Table 7 that some intermediary effects of investment level and acquisition rate also exist in the two sub samples of central enterprises and local SOEs and are consistent with the whole sample of SOEs. Therefore, the conclusion of this paper is robust.

6. Conclusions

This study takes A-share state-owned listed companies in Shanghai and Shenzhen from 2010 to 2017 as the research sample to investigate the impact of salary control policies in 2015 on the operating performance, executive incentive, and executive behavior of SOLCs. The results show that: first, SLO leads to the decline of operating performance of SOLCs. After controlling macro factors, corporate factors and other variables, the implementation of salary control policy in 2015 led to a significant decline in the operating performance of central enterprises and local SOEs. In addition, the decline in the operating performance of SOLCs was not caused by the action mechanism of executives seeking implicit benefits such as OJC, to increase the operating cost of enterprises, but by executives’ behaviors such as investment and M and A, which reflect the hard work of executives. Second, the OJC level of senior executives was not significantly affected by SLO. Third, SLO reduced the actual work enthusiasm of executives of SOLCs, which was reflected in the reduction of investment level and the frequency of mergers and acquisitions. The impact of SLO on the investment level of central enterprises and local SOEs was significantly negative. In terms of M and A frequency, both central enterprises and local SOEs decreased significantly after SLO. In a word, compared with before SLO, senior executives of SOLCs tend to be inactive to a certain extent.

The conclusions of this paper have the following policy implications: first, improve the incentive mechanism of SOE executives and establish a diversified performance evaluation index system. At present, SOEs basically link executive compensation with accounting data such as profits and sales revenue for performance assessment. According to the assessment methods issued by SASAC, the assessment indicators linked to performance are the total annual profit and economic added value of the enterprise. However, this ignores at least two problems: (1), this performance appraisal system is easy to lead to executives’ short-sighted behavior and lack of attention to the long-term value of the enterprise; (2), this performance appraisal system does not pay attention to investment, mergers and acquisitions and other financial activities closely related to the behavior of executives, which can provide relevant information about the hard work of executives, to effectively alleviate the principal-agent problem and the information asymmetry between principals and agents.

Second, this paper finds that SLO will have an impact on the financial behavior of senior executives of SOEs. At present, the salary system of senior executives of SOEs in China still lacks sufficient attention. Therefore, it is suggested that SOEs establish a diversified performance evaluation index system and pay appropriate attention to the evaluation index that can reflect the efforts of senior executives, and comprehensively evaluate the contribution of senior executives to the enterprise. Potentially, this may have the effect of enhancing enterprise performance if the hypotheses of this study can be validated. This leads to the major limitations of the current work. This study was hampered by the reliability and validity of the measures used. This in part was due to the data available to use in our study and the appropriate statistical tools of which to study them as provided in the literature. To ensure that the results and conclusions presented here are robust, future research should use longer intervals of data on either side of 2015 (i.e., before and after the imposition of the SLO) and more advanced statistical techniques. Moreover, because only Chinese listed companies have an obligation to publish their financial data, this study was biased towards them or SOES as sample data. Additionally, only the most outstanding companies in an industry were represented in the data and, as such, less profitable companies were necessarily neglected, potentially skewing the present results and conclusions. An expanded dataset may change the conclusions of this study, and when available, future research should use it to test this study’s hypotheses. If a new SLO is issued by Chinese or other governments, it would be illuminating to re-test the hypotheses suggested by this study in the current economic environment both within China and other jurisdictions.