Sales Maximization or Profit Maximization? How State Shareholders Discipline their CEOs in China*
Abstract
This study examines the determinants of Chief Executive Officer (CEO) turnover in Chinese state-owned firms. Based on a sample of 1 555 turnover cases among listed firms in China during the period 1999–2003, we obtain three main results. First, CEO turnover is negatively related to the sales performance but not the profitability of the core business. Second, the negative relationship between CEO turnover and sales is stronger for firms with excessive employment and higher organizational slack. Third, there is a significant post-turnover increase in sales but a decline in profitability of the core business. Overall, our evidence suggests that state shareholders put a greater emphasis on sales generation than on profitability when they monitor their CEOs.
1. Introduction
The way state owners run their companies has been a focal point of theoretical and empirical research. The general notion is that state owners have weak profit incentives due to the fact that they enjoy control rights without cash-flow rights. Moreover, state-owned enterprises (SOEs) typically benefit from generous transfers of government resources (Firth et al., 2009) and do not run the risk of bankruptcy (e.g. Shleifer and Vishny, 1994; Shleifer, 1998). A direct implication of shareholders’ weak profit incentive is that they are unlikely to discipline their managers on the basis of firm financial performance.
Notwithstanding, recent studies exploring management turnover in China’s state-controlled listed firms find a higher turnover risk in instances where financial performance is weak. Specifically, management turnover is negatively related to sales-related measures (Aviazian et al., 2005; Firth et al., 2006; Kato and Long, 2006), and the accounting profitability measure of return on assets (Firth et al., 2006; Kato and Long, 2006; Chang and Wong, 2009). Evidently, these findings pose a serious challenge to property rights theory as they suggest that state shareholders of corporatized firms, similar to private shareholders, are more likely to discipline their managers when firm performance is poor. This would suggest no systematic difference between state-controlled and private companies when it comes to management monitoring and turnover.
A salient characteristic of the existing studies on determinants of CEO turnover is that they use the above two types of performance measures as interchangeable measures of firm financial performance. The relative importance of these two performance measures in determining CEO turnover has remained unexplored. The lack of a more precise distinction, however, may obscure underlying shareholder motives in CEO monitoring. Theories on state ownership have emphasized that state owners tend to have a weaker incentive than private owners to maximize profits because of the lack of real cash flow rights on the part of state owners (Alchian, 1965; Demsetz, 1988; Vickers and Yarrow, 1991; Shleifer, 1998). On the other hand, state shareholders will have an inherent incentive to maximize sales revenue and output, as firms with a higher sales volume can generate more cash inflows which in turn increase the resources available to pursue political and personal objectives (Millward and Parker, 1983; Kornai, 1992; Matsumura, 1998; Xu and Birch, 1999). The distinction between profit and sales motives is therefore far from trivial as it helps to provide additional evidence on the question to what extent shareholders of state-controlled and private firms converge or differ in their efforts to monitor their CEOs.
This study seeks to explore underlying shareholder objectives and firm strategies guiding CEO-monitoring. Specifically, our study is – to our knowledge – the first to distinguish between profit and sales performance measures and examine their relative impact on forced turnover of CEOs in China’s listed firms. China’s listed companies offer a valuable testing ground to discern ownership differences because the stock market offers a mixture of state-owned and private firms operating under similar institutional conditions.
Our investigation proceeds in three major steps. First, we examine the relative importance of profit margin (PM, defined as profit over sales) and asset turnover (AT, defined as sales over assets) in guiding CEO monitoring. Whereas a company’s PM reflects shareholders’ emphasis on cost-saving strategies, AT captures how much cash inflow a firm can generate with a given asset size (Revsine et al., 1999). Unlike other studies in which various performance measures are included separately and independently, we include both PM and AT simultaneously to capture their independent effects and estimate their relative importance. Second, we explore whether reliance on sales performance as a tool for management monitoring is driven either by profit or non-profit objectives. This is important because cash inflows are not only used for enhancing profits, but also serve multiple non-profit purposes, including the realization of political and personal objectives (Millward and Parker, 1983; Kornai, 1992). To this end, we use excessive employment and managerial agency costs as proxies for the existence of political and personal objectives to estimate their influence on the sensitivity of management turnover to sales performance. Finally, we examine changes in post-turnover performance. Performance changes after a change in CEO can provide further information on the monitoring incentive of state shareholders because post-turnover performance should reflect how new CEOs are selected and monitored.
Our study is based on a sample of 1 555 turnover cases recorded in China’s listed firms between 1999 and 2003, a period offering relatively ideal conditions to compare the monitoring activities of state and private shareholders. First, the overall sample of China’s listed firms is sufficiently diversified to allow a comparison of shareholder behavior among private and state-controlled listed firms. Before 1999, the number of privately controlled listed firms was simply too small for comparative studies as privatization of China’s listed firms had taken place only in the late 1990s (Chen et al., 2008). In addition, the sample period is characterized by a relatively stable institutional framework regulating corporate governance mechanisms. After 2003, the government initiated a series of institutional reforms affecting the various corporate governance mechanisms of listed firms. Reforms included the introduction of independent directors, strengthening of disclosure requirements and implementation of administrative procedures to punish corporate misdeeds and securities misfeasance (Lin et al., 2009). Given the uneven and often gradual introduction of these mechanisms over the sample population, the identification of shareholder incentives would necessarily be complicated and not allow clear-cut empirical inferences.
Our analysis yields three main results. First, CEO turnover in state-controlled firms is negatively related to sales performance but not to the profitability of the core business. In contrast, turnover in private firms is negatively related to both sales performance and the profitability of the core business. These results suggest that state shareholders place greater weight on asset generation than on cost saving strategies when assessing CEO performance. Second, the negative relationship between CEO turnover and sales performance in state-controlled firms is stronger when firms have excessive employment and higher managerial expenses, whereas no such relationship is observed for private firms. These results are consistent with the hypothesis that state shareholders’ emphasis on sales performance is closely associated with their desire to generate cash flows for the pursuit of political and personal objectives rather than as a means to enhance profits. Third, there is a significant post-turnover increase in sales performance but a decline in core business profitability. Overall, our study supports the hypothesis that state shareholders have different strategic goals to private shareholders. In particular, state shareholders still place significantly less weight on profitability as the basis of their CEO monitoring, while non-profit motives seem to continue to influence firm strategies and management monitoring in SOEs.
Our study connects with two related lines of research. First, it contributes to the field of comparative corporate governance research and provides a first attempt at distinguishing between sales and profit motives of state shareholders in CEO monitoring in China. Despite massive waves of privatization in recent decades, state ownership remains globally important in many vital industries, such as telecommunications, energy, public utilities, and banking,1 and the recent global financial crisis has led to the emergence of even more state-controlled firms. The quality of corporate control as exercised through CEO monitoring in wholly or partially SOEs is therefore of continued, and possibly increasing, relevance for policymakers and international investors.
Second, our study relates to the long-standing debate as to whether the privatization of SOEs is necessary to improve their profitability. The property rights school regards privatization as a prerequisite to improving performance in traditional SOEs (Shleifer, 1998; Narjess et al., 2009), whereas the market approach claims that increasing competition and organizational changes in governance structures will be sufficient to improve performance (Yarrow, 1986; Vickers and Yarrow, 1991; Fang et al., 2004). Although listed firms in China operate under corporate governance structures that closely resemble the rules of the game in mature market economies, our evidence suggests that the inherently weak profit motive of state shareholders continues to influence company management through CEO monitoring – the key component of corporate control.
The remainder of this paper is structured as follows. Section 2 provides a discussion of corporate governance mechanisms and the incentive structures of state shareholders in listed firms in China. Section 3 introduces the data and research method while Section 4 presents the empirical results and robustness checks. Section 5 concludes the paper.
2. State-owned Firms in China: Constraints on Profit-making
By the end of 2003 (the end of our investigation period), the Chinese state still controlled about two-thirds of the total equity of the majority of listed companies through holdings of non-tradable state shares (Sun and Tong, 2003). The political claim is that state-owned and state-controlled companies are not necessarily less profitable than private firms. Market forces, as is generally argued, should provide SOEs with the necessary competitive pressure to force them to catch up with private firms’ productivity. Although the government invested heavily in economic reforms to increase the efficiency of SOEs, there are still major constraints, making purely profit-oriented management and management supervision highly unlikely.
To begin with, SOEs are not exposed to the risk of bankruptcy in instances of poor performance or loss-making (World Bank, 1997). It is therefore likely that the profit motive will carry less weight in the objective function of state shareholders. Second, state shareholders are not allowed to sell their firm’s stock, and are therefore unable to capitalize on potential gains from increases in firm profitability and stock prices. Furthermore, entrusted local state asset administrative agents do not receive dividends, as these are instead directly transferred to the state budget. Overall, these constraints imply that incentives to maximize firm profitability are bound to be weak.2
Moreover, institutional arrangements guiding the corporate governance of SOEs further reinforce the weak profit motive of state shareholders. Since control rights of state-owned shares are administered by an asset administration system allocated under the State Council, agents remain closely linked with the government, inviting persistent government interference with an emphasis on political and social objectives rather than the firm’s financial performance.3
The fact that firm profitability is not the overarching principle of the state asset administration is also reflected in the formal regulations governing the administration of state ownership in corporatized state-controlled firms. According to the “Preliminary Method for the Administration of State Shares in Listed Companies,” state shareholders are merely requested to guarantee, protect, and further increase the controlling position of state-dominated companies in line with industrial policy guidelines (He, 1999). To this end, local asset administrators are required to oversee company activities and managerial performance based on the management’s ability to maintain and increase the value of state assets (Art. 17). Local agents, in turn, are reviewed by NABSOP based on the same guiding principle that they should “supervise and administer the preservation of, and increase in the value of state-owned assets.”4 Although these regulations do not explicitly specify how the value of state-owned assets is to be measured, the lack of explicit reference to profitability signals that the profit motive is unlikely to play a dominant role when it comes to the assessment of state asset management.
Last but not least, state influence is not limited to the formal authority of state shareholders to vote at shareholder meetings. Equally important, the state retains the formal right to approve the decisions of the Board of Directors on appointments and dismissals of CEOs and other key personnel (Qian, 1995). This de facto monopoly power over management recruitment also provides a valuable opportunity to align management priorities with political preferences.
3. Data and Research Methods
3.1. Data
A sample of firms listed on the Shenzhen and Shanghai Stock Exchanges covering the period from 1999 to 2006 is used to empirically estimate the extent to which profit and sales objectives influence CEO turnover. The data on CEO turnover comes from the China Corporate Governance Research Database (CCGRD) provided by the GTA Information Technology Co. For our analysis of CEO turnover, we focus on people who hold the formal title of General Manager or Chief Executive. We limit the turnover cases to those occurring during the period of 1999–2003 to retain three additional years of financial data for the investigation of post-turnover performance changes.
Table 1 provides an overview of recorded changes in CEO during our observation period. During this time 1 555 such changes were made, with at least one change in 879 out of the 1 255 firms listed on the stock exchanges at the end of 2003. Turnover activity declined slightly over time, with 29.74% of firms experiencing a CEO turnover in 1999 compared with 25.98% in 2003. With an average turnover rate of 27.99%, our sample clearly surpasses the figures reported in earlier studies that focus on the US and Japanese stock markets.5 Of the 879 firms for which changes in the CEO were made, 105 report multiple changes in a given year. In following the standard in the literature of consolidating multiple CEO changes and reporting only the final turnover, the number of valid CEO changes for our sample is reduced from 1 555 to 1 438. Correspondingly, the adjusted turnover rate falls from 27.99% to 25.89%.
| 1999 | 2000 | 2001 | 2002 | 2003 | 1995–2003 | |
|---|---|---|---|---|---|---|
| Number of listed companies | 918 | 1054 | 1136 | 1192 | 1255 | 5555 |
| Total number of CEO changes | 273 | 332 | 314 | 310 | 326 | 1555 |
| Annual turnover rate (%) | 29.74 | 31.5 | 27.64 | 26.01 | 25.98 | 27.99 |
| Number of CEO changes after consolidation | 254 | 303 | 284 | 293 | 304 | 1438 |
| Annual turnover rate after consolidation (%) | 27.67 | 28.75 | 25 | 24.58 | 24.22 | 25.89 |
Detailed reasons for officially recorded changes in CEO are provided by the CCGRD and include: (i) change of job, (ii) retirement, (iii) contract expiration, (iv) change in controlling shareholder, (v) resignation, (vi) dismissal, (vii) health, (viii) personal reasons, (ix) corporate governance reform, (x) legal disputes, (xi) no reason given, and (xii) completion of acting duties. Table 2 summarizes the distribution of the reasons for turnover for the total and consolidated samples. Change of job is the most common reason given for CEO turnover, occurring 427 times (or 29.69%) in the consolidated sample. This is followed by contract expiration with 319 turnover instances (22.18%) and resignation with 265 (18.43%).
| Full sample | Consolidated sample | |||
|---|---|---|---|---|
| Number | Percentage of sample (%) | Number | Percentage of sample (%) | |
| Change of job | 464 | 29.84 | 427 | 29.69 |
| Retirement | 31 | 1.99 | 30 | 2.09 |
| Contract expiration | 327 | 21.03 | 319 | 22.18 |
| Change in controlling shareholder | 43 | 2.77 | 43 | 2.99 |
| Resignation | 298 | 19.16 | 265 | 18.43 |
| Dismissal | 65 | 4.18 | 53 | 3.69 |
| Health | 49 | 3.15 | 45 | 3.13 |
| Personal reasons | 11 | 0.71 | 9 | 0.63 |
| Corporate governance reform | 146 | 9.39 | 137 | 9.53 |
| Legal disputes | 5 | 0.32 | 5 | 0.35 |
| No reason given | 103 | 6.62 | 93 | 6.47 |
| Completion of acting duties | 13 | 0.84 | 12 | 0.83 |
| Total number of observations | 1555 | 100 | 1438 | 100 |
An assessment of the motives underlying shareholders’ decisions to replace a CEO hinges on the adequate identification of instances of forced CEO turnover. The ability to distinguish forced turnover from non-forced turnover, however, is inherently difficult due to pronounced information asymmetry (Denis and Denis, 1995; Huson et al., 2004). The classification system provided by CCGRD suffers from a similar problem, as a change of job, for instance, may actually be the result of a forced turnover. Inferences on the true nature of a job change therefore need to build on additional information. A reliable indicator is the change in position experienced by the outgoing CEO. If the post-turnover position held by a CEO is less attractive in terms of salary, status, and authority than the previously held position, then the job change turnover is likely to have been involuntary.
Our identification strategy for forced turnover proceeds as follows. First, we exclude all turnover cases that are due to retirement, health (including death), corporate governance reform, and a change in controlling shareholder.6 As the focal point of our study is the corporate monitoring of state shareholders, we also exclude those cases of turnover resulting from legal lawsuits since such suits are not initiated by state shareholders as part of their normal monitoring activities. These exclusions leave us with 1 178 turnover cases with an unclear motivation. For these cases, additional information is retrieved on the post-turnover position of the outgoing CEO to partition them into the categories of “voluntary” and “forced.” We use multiple data sources to maximize the data availability and reliability, employing information from the annual reports of the firms, Infobank’s China Economic News Database, Infobank’s China Listed Firms Database, China’s Listed Firms Database7, and online material retrieved through the Internet search engine Baidu8. We define as “voluntary” all instances of turnover where the status of the CEO’s post-turnover position is comparable or higher than the original position held, and label as “forced” turnover all cases where the status of the post-turnover position is significantly lower than the previously held position.
Table 3 summarizes the results of our sorting strategy. Out of 1178 unclear turnover cases we identify 665 that can be classified as voluntary. These include 17 cases where the CEO took over a high-ranking government position at the municipal or provincial leadership level; 225 cases where the CEO retained the position of board chair or vice chair; 197 cases where the CEO was promoted to the position of board chair or vice chair; and 150 cases where the post-turnover position was a comparable management position in another listed company or within the firm’s parent company. In 10 cases turnover was due to health reasons (these cases are in addition to those officially registered by CCGRD in this category); in 34 cases turnover was associated with a change in controlling shareholder; in 24 cases it was due to legal investigations or criminal conviction; and in eight cases the CEO left the position to enroll in an educational program outside China.
| Destination | No. observations | Percentage of sample (%) |
|---|---|---|
| Voluntary turnover | ||
| CEO position taken up at another listed company or within the parent company | 150 | 13.11 |
| Promoted to board chair or vice chair | 197 | 17.22 |
| Important government position taken up | 17 | 1.49 |
| Health problems | 10 | 0.87 |
| Remaining as board chair or vice chair | 225 | 19.67 |
| Arrested or under investigation | 24 | 2.10 |
| Going abroad to study | 8 | 0.70 |
| Change in controlling shareholder | 34 | 2.90 |
| Non-voluntary turnover | ||
| Information unavailable | 288 | 25.17 |
| New position lower than CEO position | 198 | 17.31 |
| CEO position taken up at another unlisted and small company | 27 | 2.36 |
| Total | 1178 | 100 |
The remaining 513 cases are categorized as forced turnover. In 198 cases, the post-turnover positions of the CEOs carried a weaker authority and status; in 27 cases the CEOs were employed by small-scale, non-listed firms; and in 288 cases no post-turnover position could be traced before the end of the observation period. Given our comprehensive search strategy, this lack of traceable information indicates that a CEO’s post-turnover career ceases to be of interest to the business media, and therefore the new position is likely to be connected with a decline in post-turnover status and authority.
From the 513 cases classified as forced changes, we exclude 62 cases with less than 1 year of CEO tenure, as it is unlikely that such turnover decisions would be due to performance assessments. Further, we transfer from voluntary to forced turnover 19 cases involving retirement as the official turnover reason because the CEO’s age was less than China’s official retirement age of 55 years. In total, our sample of 1 555 instances of CEO turnover includes 470 instances (30.23%) of forced turnover.9
3.2. Estimation Models
(1)Our dependent variable is a binary variable that equals 1 if there is an instance of forced turnover in a given period. Rather than using the common measure of ROA as a measure of profitability, we use two performance measures to capture whether CEO turnover is sensitive to the profit motive or the sales motive. The first is the industry-adjusted profit margin (PM) of the core business, which is defined as profit over sales minus the corresponding ratio of the industry, and is used to capture the profit motive. PM is a measure of the effectiveness of cost control, and is particularly useful for capturing the profit motive because state-controlled firms tend to operate at a higher cost.
We also use the industry-adjusted asset turnover (AT), which is defined as sales over assets minus the corresponding ratio of the industry, to capture the sales motive. AT is a traditional financial ratio that measures the ability of management to efficiently employ assets to generate sales (Singh and Davidson, 2003). A decline in asset turnover or a generally low asset turnover rate compared with the industry average indicates that the management of the firm is not generating sufficient sales to justify its asset size. Given the Chinese government’s emphasis on the increase and preservation of the value of state assets, state shareholders are likely to benchmark the sales performance of their CEOs on the basis of assets they can utilize. As a result, AT appears as a relevant measure used by state shareholders in China to evaluate their CEOs’ sales performance. Both PM and AT are incorporated simultaneously into regressions in order to determine the relative importance of PM and AT to state shareholders.
In order to isolate possible confounding influences, a set of control variables is incorporated into the model. First, three variables are included to capture individual features of the departing CEO. The CEO’s age (Age) and tenure (Tenure) are controlled for, as older CEOs and CEOs with a longer tenure seem to be less frequently subject to forced turnover (Kang and Shivdasani, 1995). We also control for whether CEOs are concurrently holding the position of board chair (Duality), as a more powerful CEO is naturally in a better position to resist the threat of dismissal. Further, several firm characteristics are controlled for. We control for the number of years a firm has been listed on the stock exchange (Years), as a longer listing may be correlated with greater shareholder monitoring. Firm size, as measured by the natural logarithm of the book value of total firm assets (Size), is included because managers seem to be more entrenched in larger firms (Dalton and Kesner, 1983) and also because larger firms seem to enjoy greater bargaining power vis-à-vis the government. As debtors may attract additional management monitoring (Jensen, 1986), we also include the capital structure as measured by the book value of debt over the total book value of assets (DAR). Finally, we include a set of dummy variables to indicate the year of CEO turnover to control for business cycle effects.
As there is naturally a certain time lag in turnover decisions, we use the previous year’s measures for all variables (except those associated with the personal qualities of a CEO) in instances where a CEO change occurs in the first 6 months of a given year. Where the turnover date falls in the second half of the year, we include measures for the current year. This procedure is in line with Huson et al. (2001) and aims to alleviate potential endogeneity problems. Moreover, as the average CEO held their position for only 2.8 years over the observed period, the use of half-year lags also appears suitable for the analysis.
4. Empirical Results
4.1. Sample Selection and Descriptive Statistics
There are a total of 5 555 firm-year observations in the period from 1999 to 2003 after excluding those that involve firms listed only in the B-share market and firms in the finance industry.10 We also exclude firm-year observations that involve firms with negative equity and those that involve firms listed for less than 6 months. As our ROA, DAR, PM, and AT data have some extreme values, we winsorize these variables at the 1% level. After further eliminating observations with missing values in the variables included in our regression analysis, our final sample includes 3 815 firm-year observations.
Data on the ownership identity of the controlling shareholder is obtained from the Ultimate Ownership of Listed Chinese Firms Dataset provided by Sinofin. To ensure that our information is accurate, we also crosscheck these data with information provided by the WIND Information Co.
Panel A of Table 4 shows the summary statistics for the variables used in the subsequent analyses. Our sample firms have, on average, been listed for 4.97 years, with the average age and length of tenure of the managers being 46.43 and 2.75 years, respectively. Overall, we have 3 110 firm-year observations (81.5%) where the state is the controlling shareholder and 705 firm-year observations (18.5%) where private shareholders are in a controlling position. Duality is not a common feature, with only 15.4% of CEOs also serving as board chair.
| Panel A: Summary statistics for the key variables | ||||||
|---|---|---|---|---|---|---|
| Variables | Number | Mean | Median | SD | Minimum | Maximum |
| Years | 3815 | 4.973 | 5 | 2.324 | 1 | 12 |
| Age | 3815 | 46.434 | 46 | 7.280 | 26 | 70 |
| Tenure | 3815 | 2.754 | 2.5 | 1.615 | 0 | 12 |
| Duality | 3815 | 0.154 | 0 | 0.361 | 0 | 1 |
| DAR | 3815 | −0.003 | −0.001 | 0.159 | −0.516 | 0.535 |
| Size | 3815 | 21.033 | 20.977 | 0.841 | 17.917 | 26.632 |
| Private | 3815 | 0.185 | 0 | 0.388 | 0 | 1 |
| ROA | 3815 | 0.105 | 0.098 | 0.057 | −0.065 | 0.329 |
| PM | 3815 | 0.013 | 0.000 | 0.126 | −0.545 | 0.569 |
| AT | 3815 | 0.072 | 0.001 | 0.302 | −0.761 | 1.689 |
| Panel B: Univariate tests of financial performance for state-owned and private firms | ||||||
|---|---|---|---|---|---|---|
| Mean test | p-value | Median test | p-value | |||
| State-owned sample | Private sample | State-owned sample | Private sample | |||
| Performance measures without industry adjustment | ||||||
| Asset turnover | 0.532 | 0.442 | 0 | 0.442 | 0.370 | 0 |
| Profit margin | 0.237 | 0.267 | 0 | 0.212 | 0.247 | 0 |
| Performance measures with industry adjustment | ||||||
| Asset turnover | 0.087 | 0.006 | 0 | 0.014 | −0.049 | 0 |
| Profit margin | 0.007 | 0.043 | 0 | −0.004 | 0.027 | 0 |
- This table reports the univariate tests of financial performance for state-owned and private firms in China. The sample period is from 1999 to 2003. Asset Turnover is industry-adjusted asset turnover, defined as sales over assets minus the corresponding ratio of the industry. Profit Margin is the industry-adjusted profit margin of the core business, defined as profit over sales minus the corresponding ratio of the industry.
Panel B of Table 4 presents some univariate analyses of state and private firms for our two performance measures. Consistent with our hypothesis on the weak profit motive of state shareholders, the PM of state-controlled firms (mean = 0.007) is significantly lower than that of private firms (mean = 0.043). Additionally, the AT of state-controlled firms (mean = 0.087) is substantially higher than that of private firms (mean = 0.006), which indicates a relatively strong sales motive among state shareholders. Similar results are obtained when median tests are used.
4.2. Regression Results for the Baseline Models
Two estimation issues are worth noting before we discuss our results. First, there is a potential lack of independence across observations for a given CEO because of the existence of certain unobservable person-specific factors. We therefore estimate the model using the Huber/White/sandwich robust method with adjustment for within-cluster correlation for each CEO (Wooldridge, 2002).11 Second, through conducting a Pearson correlation test it is found that all of the correlations among the variables included in our models are lower than 0.5. In also calculating the variance inflation factors (VIF) for each independent variable, it is observed that the VIFs never exceed 3, suggesting that our models are not plagued by serious multicollinearity problems.
Table 5 reports estimates of the sensitivity of turnover to PM and AT for state-controlled and private firms, respectively. To illustrate the importance of decomposing the profit and sales motives, we also report the results using the ratio of profit from the core business over total assets (ROA) as an alternative explanatory variable. Although the coefficient for ROA is negative and significant at 5% for SOEs, the breakdown of ROA into PM and AT indicates that the negative coefficient for PM is not statistically significant at conventional levels, while only the coefficient for AT is significant at 10%. The results show that state shareholders rely more on sales performance than on firm profitability for monitoring CEOs and that the negative relationship between ROA and turnover is actually driven by the sales component (AT) rather than the profit component (PM). For the private firm sample, the coefficients for PM, and AT are significantly negative at 5%, which suggests that CEO turnover in private firms is sensitive to both profitability and sales performance.
| State-owned sample | Private sample | Interaction Effect | |||
|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | |
| Years | 0.022 (1.367) | 0.021 (1.336) | 0.036 (1.270) | 0.037 (1.281) | 0.03** (2.432) |
| Age | 0.019** (3.692) | 0.019*** (3.769) | 0.029*** (3.171) | 0.029*** (3.226) | 0.02*** (4.824) |
| Tenure | −0.237*** (7.910) | −0.236*** (7.879) | −0.190*** (3.300) | −0.193*** (3.363) | −0.217*** (8.318) |
| Duality | −0.446** (3.388) | −0.455*** (3.474) | −0.466** (2.182) | −0.465** (2.193) | −0.462*** (4.341) |
| DAR | −0.040 (0.182) | 0.017 (0.075) | 0.605 (1.495) | 0.580 (1.409) | 0.126 (0.663) |
| Size | −0.145** (3.212) | −0.141*** (3.112) | −0.147* (1.778) | −0.147* (1.763) | −0.131*** (3.589) |
| ROA | −1.890** (2.723) | −2.361* (1.662) | |||
| Private | 0.175** (2.321) | ||||
| PM | −0.301 (0.915) | −1.080** (2.267) | −0.224 (0.718) | ||
| AT | −0.264* (1.868) | −0.498** (2.100) | −0.261* (1.807) | ||
| Pri_PM | −0.951* (1.680) | ||||
| Pri_AT | −0.217 (0.827) | ||||
| Constant | 0.938 (1.005) | 0.869 (0.927) | 0.64 (0.373) | 0.649 (0.376) | 0.797 (1.054) |
| Observations | 3110 | 3110 | 705 | 705 | 3815 |
| Pseudo R-squared | 0.078 | 0.075 | 0.093 | 0.098 | 0.072 |
Similar to Kato and Long (2006), we use the estimated coefficients from our benchmark profit models to calculate the predicted change in the probability of CEO turnover if PM (AT) improves from the 25th percentile to the 75th percentile while all other variables are kept at their mean level. The increase in PM will decrease the CEO turnover rate from 7.4% to 6.9% for private firms. It will, however, slightly increase the turnover rate from 5.4% to 5.5% for SOEs. On the other hand, the corresponding increase in AT will decrease the CEO turnover rate from 6.3% to 5.7% for SOEs and from 9.3% to 8.8% for private firms. The percentage change in turnover rate is greater for SOEs (about 10% decline) than private firms (about 5% decline).
An additional estimation confirms that state and private shareholders employ different performance measure to monitor their CEOs. Creating a new dummy variable, Private, which is set to 1 if a firm is privately-controlled, we interact this variable with our two performance measures and re-estimate the baseline model for the full sample of the listed firms. The last column of Table 5 reports the results. The coefficient on the interaction term of Private and PM is significantly negative, which indicates that private shareholders place a greater emphasis on profit margin than state shareholders.12 On the other hand, the coefficient on the interaction term of Private and AT is statistically insignificant, suggesting that private and state shareholders respond similarly to the performance measure of AT.
Our control variables behave broadly as expected. Similar to the results obtained in previous studies, the coefficients of Age are significantly positive and the coefficients of Tenure are significantly negative, indicating that the probability of forced turnover is lower for younger CEOs and for those with a longer tenure. The coefficient for an individual who is both a CEO and holding a board chair position (Duality) is significantly negative, suggesting that the duality structure undermines CEO monitoring and reduces the possibility of forced turnover within state-controlled firms.13
4.3. What Drives Sensitivity to Sales Performance?
Our baseline results suggest that CEO turnover in state-owned listed firms are more sensitive to sales than profit performance. As discussed in the introduction, the emphasis on sales performance can serve either profit or non-profit motives. To tease out the underlying motivation, we examine whether the sensitivity of turnover to sales performance is stronger among firms that have a greater need to generate cash flows to support non-profit motives. Specifically, we use over-employment and organization slack to capture the typical use of cash flow for either political or personal considerations. If we can confirm that the emphasis on sales performance as a monitoring device is significantly associated with any of these two indicators, this would suggest that non-profit motives are an important driver of sales sensitivity. Conversely, failure to identify such a statistical association would instead suggest the dominance of profit motives.
We include a dummy variable to indicate whether firms maintain over-employment (Over). Over-employment typically reflects political motives, as governments tend to respond to voter and interest group pressure by offering excess labor (Shleifer and Vishny, 1994). During our sample period the problem was apparently pervasive as a 2002 survey indicated that 50.6% of SOEs had no control over hiring and firing decisions. Similarly, 24% lacked control over promotions and demotions (Li, 2009). To construct a measure of over-employment, we use three outcome variables that capture a firm’s employment situation: the total number of employees, the ratio between the total number of employees and the book value of total assets, and the ratio of the number of employees to the main operating income. Further, we assume that a firm’s normal labor demand is determined by firm size, capital intensity, firm growth, industry (controlling for different technology and production characteristics), and year (controlling for macroeconomic variations). We then employ three median regressions to estimate the supposed normal employment conditions for our three outcome variables and compare them with real employment conditions.14 We regard over-employment as robustly confirmed for observations where all three median regressions produce positive residuals. We then construct the binary variable Over, which equals 1 if a firm has over-employment, and 0 otherwise.
Second, we introduce a binary variable to indicate whether a firm’s administration fee over the book value of total sales is larger than the year–industry adjusted median value (Fee). This variable is used to capture bureaucratic slack, which often signals the extent of firm consumption (Ang et al., 2000; Singh and Davidson, 2003).
We include these two dummy variables and their interaction terms with AT to capture the effects of political and personal interests on the sensitivity of turnover to sales performance. If the emphasis of state shareholders on AT is closely related to political and personal interests, a stronger turnover sensitivity can be observed for firms with higher level over-employment and organizational slack. Table 6 reports the results.
| State-owned sample | Private sample | |||||
|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | |
| Years | 0.027* (1.667) | 0.028* (1.715) | 0.028* (1.726) | 0.042 (1.455) | 0.042 (1.447) | 0.042 (1.446) |
| Age | 0.019*** (3.797) | 0.019*** (3.757) | 0.019** (3.727) | 0.029*** (3.185) | 0.029*** (3.165) | 0.029*** (3.159) |
| Tenure | −0.238*** (7.951) | −0.240*** (7.940) | −0.239** (7.958) | −0.190*** (3.280) | −0.189*** (3.259) | −0.189*** (3.258) |
| Duality | −0.448*** (3.413) | −0.450*** (3.421) | −0.449** (3.415) | −0.476** (2.244) | −0.477** (2.248) | −0.477** (2.248) |
| DAR | 0.101 (0.437) | 0.106 (0.457) | 0.096 (0.416) | 0.590 (1.443) | 0.589 (1.437) | 0.588 (1.438) |
| Size | −0.155*** (3.305) | −0.155*** (3.331) | −0.153** (3.262) | −0.186** (2.154) | −0.186** (2.151) | −0.185** (2.151) |
| PM | −0.235 (0.688) | −0.203 (0.602) | −0.197 (0.583) | −1.034** (2.095) | −1.041** (2.122) | −1.040** (2.120) |
| AT | −0.088 (0.540) | 0.092 (0.500) | 0.186 (0.959) | −0.383 (1.418) | −0.458 (1.484) | −0.455 (1.391) |
| Over | −0.047 (0.620) | −0.082 (1.105) | −0.062 (0.804) | 0.021 (0.146) | 0.020 (0.137) | 0.020 (0.136) |
| Fee | −0.123 (1.580) | −0.084 (1.081) | −0.090 (1.154) | −0.223 (1.532) | −0.224 (1.536) | −0.223 (1.526) |
| Over * AT | −0.630** (2.077) | −0.559* (1.772) | 0.002 (0.003) | −0.014 (0.030) | ||
| Fee * AT | −0.629** (2.419) | −0.579** (2.231) | 0.165 (0.382) | 0.166 (0.382) | ||
| Constant | 1.191 (1.220) | 1.220 (1.258) | 1.169 (1.197) | 1.525 (0.856) | 1.524 (0.856) | 1.518 (0.853) |
| Observations | 3110 | 3110 | 3110 | 705 | 705 | 705 |
| Pseudo R-squared | 0.080 | 0.081 | 0.083 | 0.103 | 0.103 | 0.103 |
Consistent with our expectation, the negative relationship between turnover and AT is stronger for firms with excessive employment and higher organizational slack. When interaction effects are included, AT also loses its independent effect on forced turnover.15 Overall, the findings are consistent with the hypothesis that the emphasis of state shareholders on AT is driven by political and personal interests.
4.4. Post-turnover Performance Change
The manner in which performance changes follow a change in CEO can further shed light on the monitoring incentive of state shareholders because post-turnover performance will be affected by how the new CEO is selected and monitored. If sales performance rather than profitability is the major cause of managerial turnover, then the new manager is more likely to be selected and monitored on the basis of AT rather than PM. This, in turn, means that a post-turnover increase in AT is more likely than an increase in PM.
Following Huson et al. (2004) for the post-turnover analysis, we use a control group to isolate the component of performance change that is attributable to the mean reversion of accounting performance. The timing of performance comparisons is a crucial issue, as both outgoing and incoming CEOs have greater incentive to manage company accounts. Outgoing managers tend to over-report company performance in an effort to secure their jobs, whereas incoming managers may under-report company performance to facilitate chances to “realize” performance increases in the following years. To mitigate problems stemming from account management, we construct two control groups. For one group we use the performance reported in the turnover year (Year 0) and for the other group we use the performance reported in the year before the turnover occurs (Year −1).
We construct the control groups as follows. We first match each firm that experienced a change of CEO in a given year with a firm that is both in the same industry and which had a similar recorded firm performance (in both PM and AT; ±20% of the sample firm’s performance) in the corresponding year while not undergoing a change in CEO in either the event year or the three preceding years. Where multiple firms fulfill these conditions, we elect to choose the firm with an asset size that is closest to that of the sample firm. If there are no intra-industry firms with a performance level within the specified band, we loosen our restrictions and match our sample firm with a firm exhibiting similar performance but in a different industry. In total, our control group includes 325 (319) firms that match in terms of both industry and performance in Year 0 (and Year −1) and an additional 80 (78) firms that match the sample firms in terms of only performance in Year 0 (Year −1). Finally, we exclude 31 (39) firms from our sample because we are not able to identify any firms with matching performance.
Table 7 presents the median post-turnover performance changes for the samples of state-owned and private firms. As consistent results are obtained when using Year 0 and −1 as the comparison benchmark, the table reports only the results when Year 0 is used as the reference. Panel A reports the performance changes for SOEs. There is a significant decline in the unadjusted profit margin of the core business in all years, but no significant change in the industry adjusted profit margin, control group adjusted profit margin (measured as PM minus the median of the corresponding ratio in the control group), or the industry and control group adjusted profit margin. However, there are significant increases in the unadjusted asset turnover and industry adjusted asset turnover in the 3 years following a CEO change. Significant positive changes in control group adjusted asset turnover and industry and control group adjusted asset turnover can also be observed in all years except for Year 3, in which the changes are still positive but not statistically significant. Overall, the results indicate a significant improvement in asset turnover but no significant improvement in profit margin among state-controlled firms in the post-turnover years. The results are consistent with our regression results, suggesting that sales performance carries greater weight than profitability when state shareholders monitor their CEOs.
| Panel A: State-owned sample | Panel B: Private sample | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Unadjusted PM | Industry adjusted PM | Control group adjusted PM | Both industry and control group adjusted PM | Unadjusted PM | Industry adjusted PM | Control group adjusted PM | Both industry and control group adjusted PM | ||
| (+1, 0) | −0.011 | 0.004 | 0.000 | 0.000 | (+1, 0) | 0.009 | 0.022 | 0.010 | 0.014 |
| p-value | 0.005 | 0.962 | 0.945 | 0.987 | p-value | 0.927 | 0.117 | 0.189 | 0.201 |
| (+2, 0) | −0.025 | −0.007 | 0.000 | −0.003 | (+2, 0) | −0.012 | 0.008 | 0.029 | 0.028 |
| p-value | 0.000 | 0.373 | 0.640 | 0.711 | p-value | 0.148 | 0.504 | 0.096 | 0.090 |
| (+3, 0) | −0.025 | −0.010 | 0.004 | 0.002 | (+3, 0) | −0.014 | 0.018 | 0.009 | 0.016 |
| p-value | 0.000 | 0.935 | 0.425 | 0.350 | p-value | 0.262 | 0.098 | 0.088 | 0.090 |
| Unadjusted AT | Industry adjusted AT | Control group adjusted AT | Both industry and control group adjusted AT | Unadjusted AT | Industry adjusted AT | Control group adjusted AT | Both industry and control group adjusted AT | ||
| (+1, 0) | 0.045 | 0.032 | 0.014 | 0.024 | (+1, 0) | 0.015 | −0.007 | 0.017 | 0.012 |
| p-value | 0.000 | 0.003 | 0.040 | 0.039 | p-value | 0.225 | 0.923 | 0.192 | 0.141 |
| (+2, 0) | 0.071 | 0.036 | 0.018 | 0.023 | (+2, 0) | 0.021 | 0.001 | 0.000 | 0.003 |
| p-value | 0.000 | 0.000 | 0.099 | 0.071 | p-value | 0.036 | 0.632 | 0.305 | 0.256 |
| (+3, 0) | 0.095 | 0.035 | 0.014 | 0.019 | (+3, 0) | 0.032 | −0.012 | −0.010 | −0.027 |
| p-value | 0.000 | 0.001 | 0.477 | 0.386 | p-value | 0.011 | 0.748 | 0.486 | 0.299 |
Panel B reports the performance changes for private firms. The control group adjusted profit margin and the industry and control group adjusted profit margin are positive and statistically significant at the 10% level except for Year 1, which indicates that the control group adjusted profit margin improved for these firms. The changes in unadjusted asset turnover, moreover, are positive and statistically significant at the 5% level except for Year 1. There are no significant changes in the industry adjusted, control group adjusted, and industry and control group adjusted asset turnover.
4.5. Robustness Checks
We explore the robustness of our findings across several dimensions. In our baseline models, we have chosen AT as a measure of sales performance. To ensure that evidence of the sales motive is robust to alternative measures of sales performance we also use sales growth (GROW) to capture the sales motive.16 Compared with AT, however, annual sales growth has two limitations. First, it is closely related to changes in asset size in the corresponding year. We therefore also include the change in asset size as an additional control variable (Asset_change). Second, annual sales growth often displays a great deal of variability. We employ two measures to deal with this problem. First, we restrict our sample by excluding the upper and lower 5% of observations. The resulting new sample for the regression using annual sales growth as a performance measure has a total of 3 312 observations. We construct the 3-year moving average sales growth rate (MGROW) as an alternative measure of sales growth performance, as this smoothes out annual fluctuations in sales growth. The measure also allows us to explore whether CEO turnover is more sensitive to average or annual sales performance.
The results from using the two sales growth performance measures are consistent with those obtained from our baseline model. As shown in Table 8, the relationship between CEO turnover and annual sales growth rate is significant at the 10% level, while the relationship between turnover and average sales growth rate is significant at the 5% level. This indicates that turnover is more sensitive to the average growth rate than the annual growth rate. When the sales growth variables are included, PM is insignificant in all cases. Overall, our results provide robust support for the emphasis on sales performance among state shareholders.
| Annual sales growth | Average sales growth | |||
|---|---|---|---|---|
| State-owned sample | Private sample | State-owned sample | Private sample | |
| (1) | (2) | (3) | (4) | |
| Years | 0.003 (0.177) | 0.032 (0.881) | 0.006 (0.350) | 0.020 (0.545) |
| Age | 0.022*** (3.857) | 0.020* (1.913) | 0.022*** (3.857) | 0.024** (2.322) |
| Tenure | −0.256*** (8.409) | −0.187*** (2.723) | −0.260*** (8.610) | −0.213*** (3.135) |
| Duality | −0.588*** (3.817) | −0.659** (2.429) | −0.506*** (3.444) | −0.556** (2.173) |
| DAR | 0.239 (1.020) | 0.157 (0.346) | 0.232 (1.012) | 0.260 (0.604) |
| Size | −0.156*** (3.147) | −0.333*** (3.344) | −0.158*** (3.196) | −0.290*** (3.037) |
| Asset_change | −0.235 (0.949) | 0.490 (1.196) | −0.151 (0.621) | 0.209 (0.523) |
| PM | −0.475 (1.410) | −1.512*** (2.813) | −0.227 (0.681) | −1.256** (2.471) |
| GROW | −0.301* (1.726) | 0.204 (0.759) | −0.511** (2.317) | 0.505 (1.361) |
| Constant | 1.178 (1.161) | 4.910** (2.346) | 1.261 (1.231) | 3.944** (1.986) |
| Observations | 2721 | 591 | 2721 | 591 |
| Pseudo R-squared | 0.096 | 0.122 | 0.095 | 0.114 |
Another concern arises if CEOs have, for instance, only limited managerial tools to actually affect firm profitability (e.g. due to persisting price regulations in state-controlled industries). In this case, reliance on AT rather than PM would not necessarily signal a stronger sales motive but rather reflect a convenience-driven choice for CEO monitoring in state-controlled industries. To address possible confounding effects resulting from government regulations, we partition our sample firms into those that operate in a state-controlled industry and those that operate in a liberalized industry. In line with the transition literature (Brada, 1996), we use the percentage of private employment in different industries as the basis for the classification of liberalized industries. Accordingly, an industry is classified as liberalized if the share of private employment in that industry is above the countrywide median value for private employment. By following this classification, we place 720 firm-year observations into a state-controlled industry sub-sample and 3 095 firm-year observations into a liberalized industry sub-sample. We re-estimate our baseline model for the two types of firms and report the results in Table 9.
| Liberalized industries | State-controlled industries | |||
|---|---|---|---|---|
| State-owned sample | Private sample | State-owned sample | Private sample | |
| (1) | (2) | (3) | (4) | |
| Years | 0.019 (1.051) | 0.059* (1.846) | 0.026 (0.689) | −0.071 (1.030) |
| Age | 0.016*** (2.990) | 0.029*** (2.915) | 0.039*** (2.823) | 0.028 (1.213) |
| Tenure | −0.212*** (6.532) | −0.196*** (3.164) | −0.391*** (6.221) | −0.223 (1.437) |
| Duality | −0.478*** (3.313) | −0.490** (2.078) | −0.352 (1.059) | −0.384 (0.902) |
| DAR | −0.041 (0.167) | 0.304 (0.650) | 0.214 (0.385) | 1.696* (1.674) |
| Size | −0.114** (2.286) | −0.201** (2.047) | −0.304** (2.352) | 0.245 (1.491) |
| PM | −0.351 (0.939) | −1.037* (1.929) | −0.291 (0.378) | −0.428 (0.317) |
| AT | −0.273* (1.865) | −0.511* (1.935) | −0.491 (1.018) | −0.168 (0.302) |
| Constant | 0.397 (0.386) | 1.479 (0.735) | 3.582 (1.335) | −6.292* (1.716) |
| Observations | 2538 | 557 | 572 | 148 |
| Pseudo R-squared | 0.065 | 0.107 | 0.157 | 0.167 |
For the SOEs operating in liberalized industries, there is still a significant negative relationship between turnover and AT but no relationship between turnover and PM. This suggests that the insensitivity of turnover to PM is not simply a response to weakly liberalized firm operations. However, there is neither a significant relationship between turnover and AT nor between turnover and PM in state-controlled industries, which is consistent with our expectation that CEOs are unlikely to be evaluated on the basis of financial performance in weakly liberalized industries.
Critics could also question the reliability of our chosen performance measures. Evaluations of CEOs may be based on their average performance rather than fluctuations in annual performance. In order to account for this, we therefore employ two alternative measures of CEO performance including a 3-year moving average of PM (MPM) and a 3-year moving average of AT (MAT) over a CEO’s tenure. Consistently, we find a negative relationship between turnover and MAT but no such relationship between turnover and MPM.
We also respond to concerns that our results could be caused by heterogeneity in the broader institutional environment. Clearly, China’s transition economy is characterized by pronounced variability in the extent and scope of its marketization. To rule out unobserved variable bias due to this institutional heterogeneity, we control for the degree of provincial-level market development by using the National Economic Research Institute of China (NERI) marketization index (Fan and Wang, 2006). This composite index covers the fields of government and market relations, development of the non-state economy, development of the product market, development of factor markets, and the legal environment. The index values range from 1 to 10, with 10 indicating the highest level of marketization. Our results are not only confirmed under inclusion of the comprehensive marketization index, but also hold under inclusion of specialized sub-indices that focus on quality of private sector development, legal system, and firm-government relations.17
Different industries tend to have different PM and AT values due to the specific conditions within which they operate. Although our baseline models use industry-adjusted performance measures to filter out some of these industry effects, we nevertheless incorporate a set of industry dummy variables into our models to further ensure that our results are not driven by industry effects. Again, the findings remain unaltered.
Finally, we check the sensitivity of the results to our method of classifying instances of CEO turnover. In doing so, both 60 and then 65 years of age are used as the benchmark for the classification of forced retirement. We also include turnover instances that are associated with legal disputes in the forced turnover category. Finally, turnover cases in which we fail to identify the destinations of the departing CEOs are excluded from our forced turnover sample. We continue to obtain consistent results when employing these alternative classification schemes.18
5. Conclusion
Previous studies have confirmed that managerial turnover in state-controlled firms is responsive to profitability and different sales-related measures. These findings seemingly contradict the common notion that a weak profit motive guides and shapes state shareholder behavior. The problem with these previous studies, however, is that none explicitly distinguishes between sales and profit motives to examine their relative importance. To fill this gap in the literature, this study distinguishes between both objectives and estimates their relative impact on CEO turnover in a sample of listed firms in China.
We obtain three main results. First, CEO turnover is negatively related to the sales performance but not the profitability of the core business. Second, further tests on the interaction effects confirm that the sensitivity of CEO turnover to sales is stronger for firms with over-employment and excessive administrative expenses. This suggests that state shareholders do not treat sales maximization as a means of enhancing profits but rather as a tool to generate cash flows for use in realizing social and political objectives such as personal rent-seeking behavior and the provision of excess employment. Third, there is a significant post-turnover increase in sales performance but a decline in profitability of the core business. Overall, our study suggests that CEO monitoring is guided by the interest of state shareholders in maximizing sales rather than profitability.
From a broader perspective, our study contributes to the literature on the corporate governance of partially privatized firms. Our results indicate that CEO turnover, as one of the key mechanisms of corporate control, differs significantly between state-controlled and privately-controlled firms. In particular, shareholders of the former place far less importance on the profit motive when monitoring CEOs and are instead strongly inclined to use sales as the crucial determinant (Buchanan et al., 1980; Kornai, 1992). Our evidence from listed firms in China therefore suggests that corporatization and public listing may alone be insufficient to turn SOEs into profit-oriented entities.
In an era of a global revitalization of state shareholdings, these findings may have critical implications for private investors in corporatized SOEs. Notwithstanding, before prematurely generalizing our results, we advocate caution on two counts. First, our evidence is obtained from publicly listed firms in China only, and further evidence both from developing and developed economies is needed to rule out the impact of country-specific cultural or political effects that might influence state shareholder behavior. Second, as the specific social welfare functions are unknown our results do not invite inferences on overall welfare effects. Theoretically, the dominance of the sales motive could yield positive short-term welfare effects if local employment rates and wage levels were to be taken into account. Nevertheless, such potentially positive social effects would be partly financed by lower firm profitability.
Footnotes
Appendices
| Variable | Definition |
|---|---|
| Force | A binary variable that equals 1 if there is an instance of forced turnover in a given period |
| Years | The number of years a firm has been listed on the stock exchange |
| Age | The CEO’s age |
| Tenure | The CEO’s tenure |
| Duality | A dummy variable that equals 1 if a CEO is concurrently also holding the position of board chair |
| DAR | Capital structure as measured by the book value of debt over the total book value of assets, adjusted by industry–year median |
| Size | Firm size, as measured by the natural logarithm of the book value of total firm assets |
| Private | A dummy variable that equals 1 if the private shareholder is in the position of the controlling shareholder |
| ROA | The ratio of profit from the core business over total assets |
| PM | The industry-adjusted profit margin of the core business, which is defined as profit over sales minus the corresponding ratio of the industry |
| AT | Industry-adjusted asset turnover, which is defined as sales over assets minus the corresponding ratio of the industry |
| Over | A dummy variable to indicate whether there is any over-employment in a firm, which equals 1 if a firm has over-employment and 0 otherwise |
| Fee | A binary variable to indicate whether a firm’s administration fee over the book value of total sales is larger than the year–industry adjusted median value |
| Over * AT | The interaction term of variable OVER and AT |
| Fee * AT | The interaction term of variable FEE and AT |
| Asset_change | Changes in asset size in the corresponding year |
| GROW | A measure of annual sales growth, defined as ln(sales/sales(t − 1)) as the measure of annual sales growth |
| MGROW | Three-year moving average sales growth rate |
Appendix B Interaction effects between financial performance and private ownership
The two horizon lines in the left figure around zero illustrate the critical values of z-statistics at the 5% level.
Appendix C Interaction effects between over-employment (organizational slack) and sales performance
The two horizon lines in the left figure around zero illustrate the critical values of z-statistics at the 5% level.
