Can Governments Effectively Regulate Levels and Growth Rates of CEO s Compensation? Some Evidence from the Chinese 2009 Regulation

Transcription

1 Can Governments Effectively Regulate Levels and Growth Rates of CEO s Compensation? Some Evidence from the Chinese 2009 Regulation Ling Mei Cong School of Accounting Curtin University l.cong@curtin.edu.au Zoltan P. Matolcsy School of Accounting University of Technology, Sydney Zoltan.Matolcsy@uts.edu.au Fifth Early Draft; 5 March 2014 Abstract Currently, there is an ongoing debate in the literature on whether CEO compensation should be regulated. Empirical evidence based on western countries questions the effectiveness of the regulations, whilst little research has been done in emerging economies. This paper provides evidence in this area by examining the effectiveness of the 2009 Regulation issued by the Chinese government to cap CEOs compensation. Specifically, we investigate if the 2009 Regulation changed the level and growth rate of CEO compensation in Chinese SOEs. We also test if the pay-performance relation changed after the 2009 Regulation. Our statistical results show that the levels of CEO cash compensation and CEO-worker pay ratio in Chinese SOEs did not decrease. The growth rates of the CEO cash compensation and CEO-worker pay ratio were not affected by the 2009 Regulation. Meanwhile, there is no evidence that the 2009 Regulation changed the pay-performance relation in Chinese SOEs. Findings from our study can help regulators to reconsider how to further improve the effectiveness of these regulations and/or whether regulating executive compensation may only have political rather than economic benefits. Key words: regulation, CEO compensation, Chinese SOE Page 1 of 35

2 1. Introduction Over the last decades, policy makers and commentators all around the world have expressed concerns about the level and growth rate of executive compensation, which has been reinforced by the global financial crisis (GFC) during the late 2000s. These concerns have lead, all around the world, including the US, UK, EU, Australia and China, the introduction of legislation and/or regulations to curb increases in executive compensations. These legislations/regulations either explicitly reduce different parts of executive compensations, such as the base salaries (e.g. Spain, France) or the size of the bonus components of executive compensations (e.g. EU) 1. Further, some of these regulations (e.g. the US, UK, Australia) enhance shareholders right to explicitly vote on executives and directors compensation proposals at the annual general meetings (hereafter say on pay or SOP) 2. Similar to Western governments, the Chinese government has released several regulations to limit executive pay of State Owned Enterprises (hereafter SOEs) in response to societal concerns. The most important of these regulations is The Guideline for Regulating Compensation of Executives in Charge of Central Enterprises (hereafter 2009 Regulation) issued by six central government departments in 2009 (Xinhua Net 2009). The 2009 Regulation requires that the basic salary of executives in SOEs must not exceed five times the average pay of the workers and the upper limit for bonus is three times their basic salary (Xinhua Net 2009). Following the central government, various local governments in China also issued similar regulations in the same spirit of the 2009 Regulation to regulate local SOEs. As the results of the above regulatory changes, there is now emerging academic literature, which provides evidence on the effectiveness of these regulations. The primary focus of this research is based on the SOP regulations. For example, Armstrong, Gow and Larcker (2013) provides some US evidence on shareholders voting on future CEOs equity compensations, Ferri and Maber (2013) provide some of the UK evidence, whilst Monem and Ng (2013) provides the Australian evidence on the same issue. However, we are not aware of any 1 In early 2012, the Spain government announced basic annual salaries at state-owned firms would be limited to 105,000 (The Guardian, 2012). Recently, the French government announced plans to limit the salaries of French state firms to 450,000 per year (France 24, 2012). In February 2013, the E.U. agreed to cap bankers bonuses at twice their salaries (Time 2013). 2 In 2002, the U.K. was the first country mandated an annual non-binding shareholder vote on executive pay (Ferri and Maber 2013). In 2010, the U.S. SEC issued Dodd Frank Act requiring firms subject to the federal proxy rules to provide shareholders with an advisory vote on executive compensation (Armstrong et al. 2013). In 2011, Australia introduced two-strikes rule, which states if 25% of shareholders vote against a firm s remuneration report at two consecutive annual general meetings, the entire board may have to stand for re-election within three months (Monem and Ng 2013). Page 2 of 35

3 studies, which would explicitly provide evidence on regulators ability to reduce the levels and the growth rates of executive compensation by nominating fixed targets. Accordingly, the objective of our study is threefold. First, we investigate whether the 2009 Regulation reduces the absolute levels of CEO compensation and the levels of CEO compensation relative to average workers pay in Chinese SOEs. Second, we examine if the 2009 Regulation curbs the absolute and relative growth rates of CEO compensation in Chinese SOEs. Third, we test if the 2009 Regulation enhances the pay-performance relations in Chinese SOEs. The motivation of our paper is twofold. First, China offers a unique setting to examine the regulations on CEOs compensation for a number of reasons. First, unlike Western countries which experienced low economic growth after the global financial crisis, China achieved a steady and high economic growth since This offers a unique opportunity to examine if the restriction on CEOs compensations by a government in an economy that leads the world s economic growth rate as opposed to economies where the economic growth rate has been low. Hence the observed changes in executive compensation in low growth economies may be driven by poor macroeconomic performances rather than restrained by the firms. Second, unlike many western regulations which focus on SOP, the Chinese pay regulation focuses on predetermined target levels and the ratios of CEO compensation to average employee pay, hence we can provide a different insight into the effectiveness of government regulations on CEOs compensation. Third, we do not have an identification problem as the 2009 Chinese government regulation is imposed on SOEs only, whilst the executive compensation of non-soes is not regulated. This unique setting enables us to have an experimental design where we have a natural control for macroeconomic and institutional factors such as regulatory changes, macroeconomic conditions, and/or changes in the labour market. The second motivation for this paper is to build on a strand of literature that examines the Chinese executive compensation. Extant studies on executive pay in China focus on the determinants of executive compensation (e.g. Kato and Long 2006; Firth et al. 2007; Conyon 3 According to the World Bank, China's GDP growth rates were 9.2%, 10.4% and 9.3% for 2009, 2010 and 2011 respectively, while the average GDP growth rate of the western countries (the US, UK, EU) was below 3% during the same period of time. Page 3 of 35

4 and He 2011; Wang and Xiao 2011; Lam et al 2013; Hu et al. 2013), pay-performance sensitivity (e.g. Firth et al. 2010; Conyon and He 2012), non-pecuniary compensation (e.g. Matolcsy et al. 2006; Chen et al. 2010; Adithipyangkul et al. 2011; Gul et al. 2011), and executive compensation contracts (e.g. Li et al. 2013). Whilst say on pay has been an emerging hot topic in western literature (e.g. Conyon and Sadler 2010; Ferri and Maber 2013; Monem and Ng 2013), little research has paid attention to the regulation on pay issues in China. We are only aware of a study by Hu and Monem (2012), who find that a higher worker-ceo pay ratio is associated with unintended consequences such as lower profitability and lower Tobin s Q. However, Hu and Monem (2012) use the data, hence does not directly examine if the 2009 Regulation is effective or its economic consequences. Our study differs to Hu and Monem (2012) by using data before and after 2009 to directly investigate the effects and economic consequences of the 2009 government regulation. Therefore, this research fills the void in the literature and generates most comprehensive evidence on the effect of regulation on CEOs compensation in the Chinese setting. The evidence of our study is based on 3,182 Chinese SOE firms for the period Two sets of analysis were conducted using the pre-2009 SOEs as the non-soes as the control samples respectively. The SOEs and non-soes are matched within year, industry and firm size. Our results show the 2009 Regulation does not reduce the levels of total cash compensation or the CEO-worker pay ratio of SOEs. The 2009 Regulation does not decrease the growth rates of CEO total compensation or the CEO-worker pay ratio, either. Fixed effects regressions suggest the 2009 Regulation fails to affect pay and performance relation in the SOEs. Overall, our findings offer little support for the effectiveness of government regulation of CEOs compensation in the Chinese setting. Our research makes several significant contributions to the literature. First, it is one of the first to provide direct, large sample evidence on whether compensation limitations imposed by the Chinese government really work or not. Due to the socialist nature of the Chinese government, the Chinese authorities face the pressure to ease the enlarging income gap. Findings from our research supply important empirical evidence to check if social media s doubts about the effectiveness of the regulations hold or not. Second, it builds on the say on pay literature using the unique setting of China. Our evidence is consistent with that strand literature on SOP from all around the world, which questions the effectiveness of governments regulations to curb executive pay by enhancing shareholders voting rights. Page 4 of 35

5 Our evidence suggests that setting fix targets for executive pay levels and/or growth rate may also be ineffective. Third, our study has important implications for policy makers and regulators with respect to the pay/performance relation between executive pay and firm performance as well. Results from this research show that the capping regulation does not improve the pay-performance link. Findings from our study, thus, can help regulators to reconsider how to further improve the effectiveness of these regulations and/or whether regulating executive compensation may only have political rather than economic benefits. The rest of our paper proceeds as follows. Section 2 provides a literature review and draws hypotheses. Section 3 presents the research design of the paper. This is followed by results reporting in Section 4 and robustness checks in Section 5. Finally, Section 6 concludes the study. 2. Literature Review and Hypothesis Development 2.1 Literature review of regulation and executive compensation Currently there are conflicting theoretical perspectives on the effect of, and the need for regulating executive compensation. The capture theory, represented by Bebchuk et al. (2002), advocates the government regulation on executive pay. Arguments based on capture theory suggest that as managers become more entrenched they extract rent from companies by paying themselves excessively (Bebchuk et al. 2002). Bebchuk and Spamann s (2010) view is that the government has a legitimate interest in the compensation structures of firms such as financial firms due to its responsibility to safeguard the economy. Thus, the pay-setting process should not be left to unconstrained choices of informed players inside firms and government intervention can be an important tool for regulation. Scholars holding this perspective generally believe that the strengthened monitoring on the CEOs by imposing CEO pay capping or shareholder voting can increase management s responsibility, strengthen pay-performance link and reduce the executive pay level (e.g. Bebchuk et al. 2007; Davis 2007; Clarkson et al. 2009). In contrast, arguments based on agency theory questions the need for regulation on executive pay and suggest that government regulation may lead to sub-optimal pay practice (Gordon 2009; Bainbridge 2011). Proponents of the agency theory (Jensen and Murphy 1990; Murphy and Zábojník 2008; Murphy 2011) maintain market forces are able to determine efficient Page 5 of 35

6 contracts for executives. Thus, firms should leave the market to adjust the CEO compensation and the consequence of pay regulations could be both unintended and costly (Murphy 2011). Murphy (2011) explains that a large part of the regulatory failure is that the regulation is often mis-intended. The emerging empirical academic evidence is generally consistent with the arguments of agency theory. For example, there is evidence that political intervention such as SOP laws have a negative effect on the valuations of some firms (Larcker et al. 2011; Larcker et al. 2013), and that pension funds actions on executive compensation may be driven by political agendas, potentially destroying shareholder value (Bainbridge 2011; Larcker and Tayan, 2012). Larcker et al. (2011) note an alternative explanation when the regulations are less restrictive than expected and their laxness surprises the market, they could lead to unintended consequences such as ineffectiveness of the regulations or firm value decrease. Recent evidence based on the say on pay legislation also casts doubts on the effect of the government regulation. For instance, Conyon and Sadler (2010) find little evidence that the UK say on pay materially alters the subsequent level of CEO compensation. Armstrong et al. (2013) find shareholder votes for equity pay plans have little substantive impact on firms incentive compensation policies. Meanwhile, Ertimur et al. (2013) conclude that though SOP votes have an effect on compensation practices in firms with high SOP voting dissent, they do not have a detectable impact on their quality. Cuñat et al. (2013) and Iliev and Vitanova (2013) find no evidence that the US say on pay affect the level or composition of the executive compensation 4. Little research has been conducted in the emerging market context. We only find two papers in the literature 5. Based on South Korean firms, Garner and Kim (2010) find that the capping system only reduces cap-salary differences in firms with an efficient external monitoring system (such as foreign shareholders). To our knowledge, Hu and Monem (2012) is the only paper on Chinese firms. They find that a lower CEO-worker pay ratio in Chinese SOEs leads to unintended consequences such as lower Tobin s q, lower accounting profitability, more frequent CEO promotion and more increases in average worker s pay in poor performance 4 However, other studies find political interventions do take effect. For example, Bebchuk and Spamann (2010) show that compensation structure coupled with capital structure induces U.S. bank executives to take excessive risk, and regulation on pay is necessary to remedy the problems. Using a large cross-country sample of observations from 39 countries, Correa and Lel (2013) document that compared to control group of firms, say on pay laws are associated with a lower level of CEO compensation. Page 6 of 35

7 firms. Their paper, however, does not provide direct evidence if government regulation affects the compensation level or growth rate. 2.2 The 2009 Regulation and levels and growth rates of CEOs compensation On September 16 th, 2009, six administrative departments of China s central government 6 issued The Guideline for Regulating Compensation of Executives in Charge of Central Enterprises (2009 Regulation). The 2009 Regulation was cited as the Chinese version of regulation on executive pay and thus became a milestone in say on pay history in China (Xinhua Net, 2009). This regulation applies to Chinese central SOEs 7 in each industry and supersedes prior industry specific regulations such as those released by Ministry of Finance to limit the top executive compensation in state owned finance firms 8 (SCMP 2009). The issuance of the 2009 Regulation is in response to growing resentment at the excessive pay to CEOs in Chinese SOEs and the enlarging CEO-worker pay gap (ifeng, 2009). Its objective is to achieve an executive compensation system with a reasonable structure, an appropriate level and standardized monitoring for Chinese SOEs (Xinhua Net, 2009). The guideline raised five principles to be upheld in designing the compensation contract. First, the principle of combining market adjustment and government monitoring; second, the principle of unifying stimulation and restriction; third, the principle of considering both short-term and long-term incentives; fourth, the principle of coordinating executive pay rise with worker pay growth; fifth, the principle of complementing the compensation system improvement with regulating the supplementary insurance and occupational consumption (Xinhua Net, 2009). The 2009 Regulation prescribes that the CEO compensation should comprise three parts: basic salary, bonus and long-term incentive pay. The document states that as the long-term incentive schemes are still in a trial period in China, the guideline focuses on the first two types of compensation. But SOEs are encouraged to explore the incentive schemes cautiously (Xinhua Net, 2009). The 2009 Regulation requires that the basic salary of executives of SOEs should not exceed five times the average pay of employees in the previous year and the bonus 6 The six central government departments are: Ministry of Human Resource and Social Security, Organisation Department, Ministry of Supervision, Ministry of Finance, State-owned Assets Supervision and Administration Commission of the State Council (SASAC) and China National Audit Office. 7 Central SOEs are SOEs controlled by the central government represented by the SASAC. 8 It also supersededs the 2002 regulation issued by the State Commission on Asset Control requiring that the top executives annual compensation of central SOEs should not exceed 12 times the average employees pay. Page 7 of 35

8 pay should not exceed three times the basic salary. The 2009 Regulation is applicable to 2009 executive compensation in Chinese central SOEs (ifeng, 2009). The regulation is then replicated at the lower provincial level and applies to local SOEs 9 (Xinhua Net, 2009). Since its implementation, doubts have been cast on the effect of the 2009 Regulation, as anecdotal evidence seems to show that CEO compensation levels in SOEs have not dropped (Xinhua Net 2012). Further anecdotal evidence also seems to show that the specific capping targets of CEOs compensation has not been tight enough (ifeng 2009). Given the socialist nature of the Chinese government and the above anecdotal evidence, we agree with Murphy s (2011) argument, that regulations of CEOs compensation are driven by politicians and their political agenda, rather than concern with creating shareholder value. Accordingly, our hypotheses are: H1: The 2009 Regulation does not change the level of CEO compensation in Chinese SOEs. H2: The 2009 Regulation does not change the growth rate of CEO compensation in Chinese SOEs Regulation and the pay-performance relation Compared to the effect of regulation on compensation contract, fewer studies have directly examined the effect of regulation on pay-performance relation. For instance, Ferri and Maber (2013) report that UK firms react to negative say on pay voting outcomes by removing controversial CEO pay practices criticised as rewarding for failure and increasing the sensitivity of pay to poor performance. Monem and Ng (2013) investigate the effect of 2011 say on pay legislation in Australia and show that the pay-performance link did not change in 2011, but improved significantly in A prior study by Clarkson et al. (2009) also find that increased shareholder oversight (through no votes on the remuneration report) in Australia strengthens the pay-performance link and makes the pay setting process more accountable. Meanwhile, the international study by Correa and Lel (2013) concludes that say on pay laws lead to a higher pay for performance sensitivity. 9 Various provinces issued similar rules to constrain local SOEs CEO compensation. For instance, six local government departments in Jiangxi Province issued a guideline that the basic salary of executives of local SOEs should not exceed three times the average pay of employees in the previous year and the bonus pay should not exceed three times the basic salary (Jiangxi Government, 2013). Xiangfan City in Hunan Province issued a guideline stating the the basic salary of executives of local SOEs should not exceed four times the average pay of employees in the previous year and the bonus pay should not exceed three times the basic salary (Xiangfan City, 2009). Some local governments (e.g. Suzhou City, Yancheng City) use a formula to calculate the bonus. Generally speaking, all local SASACs require the bonus should not be more than three times the basic salary no matter using a fixed target or a formula. Page 8 of 35

9 Prior studies on Chinese pay-performance sensitivity (e.g. Firth et al. 2010; Conyon and He 2012) generally find a positive relationship between CEO compensation and firm performance. However, whether the relation will be influenced by the government capping regulation remains an open question. Proponents of capture theory may argue that the government regulation strengthens the external governance environment and the CEOs may be more accountable. Hence under the regulation pressure, the compensation contract is more closely related to their performance and consequently increases the pay-performance link. On the other hand advocates of agency theory may claim that 2009 Regulation capping the CEOworker pay ratio, provide less incentives for CEOs to achieve better performance, hence the pay-performance sensitivity may even decrease. However, as prior studies also show executives of Chinese SOEs receive pecuniary and non-pecuniary compensations (Chen et al. 2010), hence the CEOs may still be motivated to maintain the performance. Consequently, we expect that the compensation capping regulation does not change the pay-performance relation in Chinese SOEs. Accordingly, we predict: H3: The 2009 Regulation does not change the pay-performance relation in Chinese SOEs. 3. Research design 3.1 Sample and data Table 1 illustrates our sample selection process. This paper collects data of all Chinese SOEs with an A-share listing on Shanghai or Shenzhen Stock Exchanges 10. Following Kato and Long (2006) and Firth et al. (2010), a listed firm is identified as an SOE if its ultimate controlling shareholder is the government or government entity. Consequently, 5,127 observations are identified as SOEs between 2006 and New listing firms (387 observations) were excluded from the sample since these firms do not have continuous compensation information during Firms with nil compensation were also removed from the sample (258 observations), as these are usually due to data errors or unusual circumstances. In line with Ferri and Maber (2013), to ensure consistency in the sample composition between pre ( ) and post ( ) periods, the analysis is restricted to firms with at least one year in the pre-period and one year in the post-period. Firms with missing data were excluded from the sample. 884 observations were thus eliminated. We also eliminated the top and bottom 1% of the observations from the sample. 10 Chinese firms can issue A-Shares or B-Shares on the two domestic stock exchanges. As B-shares have not been actively traded in the recent years (Chen et al. 2010), firms only issuing B-Shares have been excluded from our sample. Page 9 of 35

10 The final data set is an unbalanced panel of 3,182 SOE observations over the period of The CEO compensation data, governance and other financial data are collected from the China Stock Market and Accounting Research (CSMAR) database. To form a control sample to compare with the SOEs, data for non-soes are also collected from the CSMAR. Following a similar sample selection process, the final dataset of non-soes include 2,893 firm year observations. The matched pairs of SOEs and non-soes are drawn from the two respective samples. 3.2 Regression models The model to test H1 To test the effect of 2009 Regulation on CEO compensation level, two sets of regressions are conducted. The following four regression models are hence estimated. Ln(Comp) i = α 0 + α 1 Post09dum i + α 2 CentralSOE i + Governance characteristics i + CEO characteristics i + Economic controls i +β i Ind_dum i + β i Reg_dum i + β i Year_dum i + ε i (1) CWP i = α 0 + α 1 Post09dum i + α 2 CentralSOE i + Governance characteristics i + CEO characteristics i + Economic controls i +β i Ind_dum i + β i Reg_dum i + β i Year_dum i + ε i (2) Ln(Comp) I = α 0 + α 1 SOE i + Governance characteristics i + CEO characteristics i + Economic controls i +β i Ind_dum i + β i Reg_dum i + β i Year_dum i + ε i (3) CWP i = α 0 + α 1 SOE i + Governance characteristics i + CEO characteristics i + Economic controls i +β i Ind_dum i + β i Reg_dum i + β i Year_dum i + ε i (4) The first set of regressions (Models 1 and 2) is based on the SOE full sample. Specifically, it tests if the post-2009 cash compensation and CWP ratios in SOEs have changed relative to the pre-2009 period. The second set of regressions (Models 3 and 4) uses matched non-soes as the control sample to see if the post-2009 compensation and CWP ratio in SOEs are Page 10 of 35

11 significantly different to the non-soe counterparts. A three-way matching method 11 is used to match SOEs and non-soes. We match the SOEs and non-soes firstly on industry, secondly on financial year and thirdly on sales revenue. Consequently, 1,156 matched pairs are generated from the original samples. In the above four models, we use two proxies to measure the absolute and relative levels of compensation respectively, namely, CEO cash compensation (denoted Comp) and CEOworker pay ratio (denoted CWP). Consistent with prior studies (Firth et al. 2010; Conyon and He 2011; Wang and Xiao 2011), given the equity-based compensation is still rare in Chinese companies, CEO cash compensation (includes salary, bonus and other cash payments) is used to measure CEO compensation 12. The CEO-worker pay ratio is calculated as the CEO total cash compensation divided by the average employee cash pay, whilst the average employee cash pay is total cash paid to employees divided by number of employees. This is consistent with Chen et al. (2010) and Firth et al. (2010). In the first set of analysis, a dummy variable (denoted Post09dum) is created to indicate the pre and post periods. The Post09dum equals to one [1] when the data is from and zero [0] when the data is from A dummy variable is also created to indicate central and local SOEs in the first set of analysis. A central SOE is defined as a firm controlled by the SASAC representing central government and a local SOE is an SOE firm controlled by local government and government agencies. CentralSOE is coded one [1] if a firm is a central SOE and zero [0] otherwise. In the second set of regressions, a dummy variable (denoted SOE) is included to distinguish SOEs from non-soes. SOE is coded one [1] if its ultimate controller is the government or government entity and zero [0] otherwise. Prior literature (e.g. Armstrong et al. 2010) finds that corporate governance characteristics affect the level of CEO compensation. For example, Core et al. (1999) report CEO pay is higher when the board is larger and more outside members are appointed by the CEO. Independent directors tend to provide more effective monitoring over the compensation contract (Fama and Jensen 1983; Coles et al. 2008). Meanwhile, Cyert et al. (2002) report that CEO pay is percent higher when the CEO is the chairman of the board. 11 An alternative to the matched-pair design is to use matched treatment and control observations based on propensity scores of the treatment variable. Propensity score-based matching has been done in recent compensation studies (e.g., Armstrong et al., 2010; Armstrong et al. 2012). However, we do not employ this design because there is no established literature on the determinants of the SOE and non-soe firms. 12 In the main analysis, the nominal compensation is used. The robustness checks use CPI-deflated 2006 RMB to re-test the hypotheses. All results remain statistically the same. Page 11 of 35

12 Consistent with prior literature, this paper includes three variables to control for corporate governance characteristics: board size (denoted Bdsize), board independence (denoted Inddirs%) and duality (denoted Duality). Bdsize represents the number of directors on the board. Inddirs% is the fraction of directors that are independent. Duality is an indicator variable equal to one [1] if the CEO is also the chair of the board, zero [0] otherwise. CEO characteristics are frequently found to impact CEO compensation. Hill and Phan (1991) report a positive association between CEO tenure and CEO compensation. Armstrong et al. (2012) shows CEO tenure affects the compensation level. CEO turnover often affects the compensation level as well. When a new CEO is appointed, there is usually a change in the compensation contract. Hence CEO tenure (denoted CEO_tenure) and CEO turnover (denoted CEO_turnover) are controlled for in the regression. CEO_tenure is measured as the number of years of service of the current CEO. CEO_turnover is an indicator variable equal to one [1] if a CEO changes during the year, zero [0] otherwise. A number of economic characteristics, including firm characteristics and firm performance are found to affect the level of CEO compensation. Prior literature generally finds a positive association between firm size and CEO compensation (e.g. Core and Guay 1999; Garbaix and Landier 2008). We use the natural logarithm of sales to control for firm size (denoted LnSales). Financial risks were found to influence the CEO compensation level as well (Core et al. 1999). Therefore, leverage (denoted Lev), computed as year-end total liabilities divided by total assets, is included in the regression. Consistent with prior literature (Matolcsy et al. 2006; Firth et al. 2010), a dummy variable Foreign is included to indicate if a firm issues B- Share, H-Share or other foreign shares. A firm will be coded one [1] if it issues foreign shares and zero [0] otherwise. Meanwhile, the percentage of shares controlled by the ultimate shareholder (denoted ConShPer) is also included in the model to control for the ownership structure. To control for investment opportunities, book-to-market value (denoted BMV) is included in the model. Firm performance has been well documented to influence compensation levels (Lambert and Larcker 1987; Sloan 1993; Core et al. 2008; Armstrong et al. 2012). Accounting returns and stock returns are common proxies for performance. In line with prior literature (Core et al. 1999, 2008), we use the return of assets (denoted ROA) and annual stock returns (denoted RET) to measure performance. To control for risk factors, the standard deviations of Page 12 of 35

13 accounting and stock performance (Std3ROA and Std3RET) are also included in the model. Given cash compensation is usually based on the previous year s performance (Perry and Zenner, 2001; Core et al. 2008), all performance measures included in the regressions are the lagged numbers. To control for the industry specific fixed effects, 13 dummy variables for industries are included 13. Due to different levels of economic development, firms from different regions vary in compensation levels (Firth et al. 2007). Consistent with prior literature (Chen et al. 2010), to control for the regional fixed effects, 3 dummy variables are created: East, Central and West. A firm is coded one [1] if it is from the region and zero [0] otherwise. Six year dummies are also constructed to control for year fixed effects where the variable is equal to one [1] if a firm is from a year and zero [0] otherwise The model to test H2 To test the effect of the 2009 Regulation on the growth rates of CEO cash compensation and CWP ratios, the following two sets of regression models are estimated: ΔComp i = α 0 + α 1 Post09dum i + α 2 CentralSOE i + ΔGovernance characteristics i + characteristics i + ΔEconomic controls i + β i Ind_dum i + β i Reg_dum i + β i Year_dum i + ε i (5) ΔCWP = α 0 + α 1 Post09dum i + α 2 CentralSOE i + ΔGovernance characteristics i + characteristics i + ΔEconomic controls i + β i Ind_dum i + β i Reg_dum i + β i Year_dum i + ε i (6) ΔComp i = α 0 + α 1 SOE i + ΔGovernance characteristics i + characteristics i + ΔEconomic controls i + β i Ind_dum i + β i Reg_dum i + β i Year_dum i + ε i (7) ΔCWP = α 0 + α 1 SOE i + ΔGovernance characteristics i + characteristics i + ΔEconomic controls i + β i Ind_dum i + β i Reg_dum i + β i Year_dum i + ε i (8) Models 5 and 6 use the SOE full sample, and Models 7 and 8 are based on the matched sample. As shown in the regressions, cash compensation growth rate (denoted ΔComp) and change in the CWP ratio (denoted ΔCWP) are used as dependent variables. ΔComp and 13 The CSRC classifies Chinese firms to 13 industries: A: Agriculture and fishery, B: Mining, C: Manufacturing; D: Electricity, water and other energy manufacturing and supply; E: Construction; F: Transportation and logistics; G: Information technology; H: Wholesales and retails; I: Finance and insurance; J: Real estate; K: Service; L: Communication; M: Others. Following this classification, 13 indicator variables are created where the variable is equal to one [1] if a firm is from an industry and zero [0] otherwise. Page 13 of 35

14 ΔCWP are measured as an annual percentage of change in CEO cash compensation and CEOworker pay ratio respectively. The independent variables are the dummy variables of Post2009dum and SOE. Control variables are measured as the changes in continuous variables of governance characteristics and economic factors The model to test H3 To further test the association between CEO compensation and subsequent firm performance, the following OLS regressions are estimated based on the SOE full sample: ROA t+1,i = + Ln(Comp) t,i + Ln(Comp) t,i *Post09dum i + Post09dum i + LnSales t,i + Lev t,i + Std3ROA t,i + iind_dum i + ireg_dum i + iyear_dum i + ε i (9) RET t+1,i = + Ln(Comp) t,i + Ln(Comp) t,i *Post09dum i + Post09dum i + LnMkv t,i + BMV t-1,i + Std3RET t,i + iind_dum i + ireg_dum i + iyear_dum i + ε i (10) ROA t+2,i = + CWP t,i + CWP t,i *Post09dum i + Post09dum i + LnSales t,i + Lev t,i + Std3ROA t,i + iind_dum i + ireg_dum i + iyear_dum i + ε i (11) RET t+2,i = + CWP t,i + CWP t,i *Post09dum i + Post09dum i + LnMkv t,i + BMV t,i + Std3RET t,i + iind_dum i + ireg_dum i + iyear_dum i + ε i (12) Using matched non-soes as the control sample, the second set of regression models testing H3 are as follows: ROA t+1,i = + Ln(Comp) t,i + Ln(Comp) t,i *SOE i + SOE i + LnSales t,i + Lev t,i + Std3ROA t,i + iind_dum i + ireg_dum i + iyear_dum i + ε i (13) RET t+1,i = + Ln(Comp) t,i + Ln(Comp) t,i *SOE i + SOE i + LnMkv t,i + BMV t-1,i + Std3RET t,i + iind_dum i + ireg_dum i + iyear_dum i + ε i (14) ROA t+2,i = + CWP t,i + CWP t,i *SOE i + SOE + LnSales t,i + Lev t,i + Std3ROA t,i + iind_dum i + ireg_dum i + iyear_dum i + ε i (15) RET t+2,i = + CWP t,i + CWP t,i *SOE i + SOE i + LnMkv t,i + BMV t,i + Std3RET t,i + iind_dum i + ireg_dum i + iyear_dum i + ε i (16) We examine both operating and stock performance in line with Core et al. (1999). The accounting measure is ROA and the market-based measure is RET. Both measures are calculated for the subsequent one and two years after compensation is awarded. Page 14 of 35

15 To test if the pay-performance relation changes after 2009, an interaction of compensation and Post09dum is created (denoted Ln(Comp)*Post09dum and CWP*Post09dum). The original interacting variables of Ln(Comp), CWP and Post09dum are also included in the respective regressions. Consistent with the ideology in Core et al. (1999), it is expected that if the 2009 Regulation increases the pay-performance link in the Chinese SOEs, the signs on the Ln(Comp)*Post09dum (or CWP*Post09dum) interaction term will be positive and significant. If the 2009 Regulation fails, the coefficients on interaction terms will be insignificant. Similarly, when matched non-soes are used as the control sample, an interaction of compensation and SOE is constructed (denoted Ln(Comp) *SOE and CWP*SOE). It is expected that if the pay-performance link in the Chinese SOEs is different to non-soes, the signs on the interaction of CEO cash compensation (or CEO-worker pay ratio) and SOE is positive and significant. If the 2009 Regulation fails, the coefficients on interaction terms will be insignificant. Consistent with prior studies (e.g. Core et al. 1999), control variables include size (LnSales or LnMkv), leverage (Lev), book-to-market value (BMV), and risk (Std3ROA or Std3RET) as well as industry, region and year indicators. 4. Results reporting 4.1 Main results Table 2 reports the descriptive statistics of CEO compensation in Chinese SOEs for the full sample and subsamples partitioned by year, industry and region. Over 2006 to 2011, the average (median) cash compensation of Chinese SOEs is 496,070 RMB (352,300 RMB) 14, which is much lower than the U.S. companies in the similar period (Murphy 2011). The average CEO-worker pay ratio is This is higher than that reported by Firth et al. (2010) (3.50 for ) and Hu and Monem (2012) (6.59 for ) 15. The average cash compensation growth rate over is 25.16% for the SOE full sample. On average the CWP increased by 13.00% from Using an average exchange rate of 1US$ = 6.9 RMB during , the mean and median CEO compensation in Chinese SOEs is about 73,480US$ and 52,170US$. 15 Frydman and Saks (2007) report that in the U.S. the ratio of CEO pay relative to average worker pay rose from 30 in 1970s to approximately 120 by Page 15 of 35

16 Table 2, Panel A reports the mean and median values of CEO compensation split by year. Overall, the data in Panel A suggest a rising trend of total cash compensation. The regulatory changes seem to not have curbed the overall CEO compensation level. In contrast, the CEOworker pay ratios dropped in 2008 and 2009, but then went up in 2010 and The growth rate of total cash compensation or CWP seems to not to have a specific trend. The highest growth rate of total cash compensation (CWP ratio) appeared in 2006 and the lowest appeared in Table 2, Panel B shows industry distribution of mean and median values of CEO compensation in Chinese SOEs. Not surprisingly, on average CEOs in the finance industry receive the highest pay of over 2 million RMB. The second highest paid industry is real estate (930,937 RMB), which could be due to the real estate boom in China in recent years. Meanwhile, CEOs in the agriculture and utilities industries are paid the lowest (315,859 and 406,757 RMB respectively). As expected, on average the CEO-worker pay disparity for the finance industry is the highest (10.559) among all industries, which is followed by the services industry (10.344). Of all sectors, the utilities industry has the lowest average CEOworker disparity (5.232). The IT industry has a slightly higher ratio (5.352). Importantly, none of the industry has exceeded the CEO-worker pay ratio limit set by the 2009 Regulation. This could mean that the 20-times cap is actually a rather generous limit. Regarding the cash compensation growth rate, CEOs in the agriculture and communication industries had the highest average pay increase during 2006 to 2011, whilst the pay rises for finance and utilities sectors are the lowest. In terms of changes in CWP ratio, agriculture and communication are again the two with the quickest increase in the CWP gap, whilst transportation and mining industries are the two slowest. Panel C of Table 2 demonstrates CEO compensation in Chinese SOEs partitioning by the level of government supervision. About 20% of the SOEs are supervised by the central government and the rest by local governments. On average, CEOs in these central SOEs are paid a higher level of cash compensation than local SOEs. The CWP ratios in central SOEs are also higher than their peer local SOEs. Meanwhile, the cash compensation growth rates and CWP growth rates are at a similar level. Table 3 presents the comparison of descriptive statistics for the SOE full sample between pre and post-regulation periods. The average total cash compensation and CEO-worker pay ratio Page 16 of 35

17 have both increased after The t-statistics show the difference between mean values is only significant for total cash compensation (p<0.01). The Wilcoxon-Mann-Whitney tests, however, indicate that the median differences in total cash compensation and CEO-worker pay ratio are both statistically significant. The growth rates of cash compensation and CWP ratios dropped after 2009, but neither of the difference is significant. SOE firms tend to have a significantly smaller board after 2009 compared to the pre-2009 period (p<0.05). The average independent director percentage in the post-period is significantly higher than the pre-period, whilst CEO turnover is significantly higher as well (p<0.01). Meanwhile, firms tend to have significantly larger sales, higher leverage, and lower book-to-market ratio in the post-2009 period than the pre-period (p<0.01). Table 3 also shows the average ROA levels were not significantly changed after the 2009 Regulation. The average and median stock returns are, however, significantly lower in the post-2009 period (p<0.01). Table 4 reports the contrast of post-2009 descriptive statistics of SOEs and matched non- SOEs. The average total cash compensation for SOEs (546,831RMB) is moderately lower than non-soes (595,330RMB) (p<0.10). The Wilcoxon-test result shows the medians of CEO cash compensation are not significantly different. The CEO-worker pay ratios for non- SOEs are significantly higher than SOEs in both the pre- and post-2009 periods (p<0.01). The growth rates of cash compensation and CWP of SOEs are lower than non-soes though neither difference is significant. Regarding CEO and firm characteristics, on average, SOEs tend to have a bigger board and a lower percentage of CEO duality arrangement. SOEs also have a bigger size, a higher leverage level, a higher percentage of firms issuing foreign shares, a more dominant controlling shareholder and a higher book-to-market ratio. Meanwhile, the accounting returns and stock returns of SOEs are significantly lower than non-soes. 4.2 Effect of 2009 Regulation on levels of compensation Table 5 reports the multivariate results testing H1 based on two benchmarks, the pre-2009 SOEs (Panel A) and non-soes (Panel B). In panel A, the F-statistics of both two models are highly significant. The adjusted R-square values are moderate. The coefficient on Post09dum is positive and significant in Model 1 (p<0.01), whilst insignificant in Model 2 though negative. This indicates that the absolute level of CEO cash compensation in Chinese SOEs increased rather than decreased after the 2009 Regulation, whilst the CWP ratio does not change significantly after the 2009 Regulation. This finding is consistent with prior western Page 17 of 35

18 studies such as Perry and Zenner (2001) and Conyon and Sadler (2010). The results, therefore, support the H1. Regarding control variables, the coefficients on the CentralSOE is positive and highly significant in the Model 1 (p<0.01), which shows that central SOEs have significantly higher CEO total cash compensation than local SOEs. In line with expectation, firms with CEOs chairing the board, longer CEO tenure, lower CEO turnover and larger size tend to have higher CEO cash compensation and CWP ratios. Higher leverage firms pay significantly lower CEO total compensation. Firms issuing foreign shares tend to have higher total CEO compensation and CWP ratio. This is likely because these firms use foreign firms as benchmarks for their CEO compensation. There is a negative relationship between shares controlled by the ultimate shareholder and the total cash compensation (CWP ratio). During , firms with a lower book-to-market ratio pay CEOs more cash compensation. The performance measures are significant in some models. Specifically, ROA is positive and significant in both Models 1 and 2, which means the total CEO cash compensation and CWP ratio are both driven by them. Meanwhile, stock returns and Std3ROA are positively associated with CEO total cash compensation. Table 5, Panel B shows regression results testing H1 using non-soes as the control sample. The F-statistics of Models 3 and 4 are highly significant (p<0.01). The R-Square values of the two models show reasonable explanatory power. The coefficient on the SOE is negative and significant in the Model 2 only (p<0.01), meaning that the CWP ratio of SOEs is significantly lower than non-soes. This seems to show that the regulated SOEs have significantly lower CWP ratios relative to non-soes. In terms of control variables, the directional signs of these variables are all consistent with those in Panel A and the majority of these control variables remain significant in Panel B. Overall, results show that the 2009 Regulation did not decrease the cash compensation level or the CWP in Chinese SOEs compared to the pre-2009 period. However, relative to the nonregulated firms (non-soes), the post-2009 CWP ratios of regulated firms (SOEs) are significantly lower. Page 18 of 35

19 4.3 Effect of 2009 Regulation on growth rates of compensation Results testing H2 are reported in Table 6. Again Panels A and B use two alternative control samples to examine the effect of 2009 Regulation on the growth rates of cash compensation and CWP ratios. Table 6, Panel A presents regression results based on the SOE full sample., The F-value of the Models 1 and 2 are significant (p<0.01). The adjusted R-Squares of both models are rather low (below 1%). None of coefficient on Post09dum is significant, this means the growth rates of cash compensation and CWP ratio in Chinese SOEs have not changed after Most of the control variables are insignificant with some exceptions. Specifically, in Model 1 ΔLn(sales) is positive and significant, whilst ΔBMV is negative and significant. ΔConShPer and ΔBMV are negative and significant in Model 2. Results based on the matched sample are shown in Table 6, Panel B. Again, both models are significant, though the adjusted R-Squares are low. The coefficients on SOE are insignificant in both models. This means after the 2009 Regulation, the growth rates of CEO cash compensation or CWP ratio in Chinese SOEs are not significantly different to those in non- SOEs. Regarding control variables, ΔLn(sales) and ΔRET are significant in Model 3 and ΔBdsize is significant in Model 4. In summary, consistent with expectation, the two sets of tests both show that the 2009 Regulation does not change the growth rates of cash compensation or CWP ratios. Hence, H2 is supported. 4.4 The impact of 2009 Regulation on pay-performance relation Table 7 reports the effect of 2009 Regulation on the pay-performance relation in Chinese SOEs using the SOE full sample. Panels A and B show the relation between pay and subsequent one and two year performance respectively. Table 7 shows the F-statistics are highly significant in all eight models. The adjusted R- Square values vary across the models with stock performance models tend to have much higher explanatory power (from to 0.374) than operating performance models (from Page 19 of 35

20 0.064 to 0.080). Consistent with expectation, none of coefficient in the eight models is significant. This shows the pay and the operating (stock) performance link does not increase after the 2009 Regulation. Therefore, H3 is supported. For the subsequent one-year operating performance, the coefficient on total cash compensation is positive and significant (p<0.01). This means consistent with findings of Firth et al (2010), there is a positive link between total cash compensation and operating performance in China. However, the coefficient on the CWP is not significant though positive. Regarding the subsequent one-year stock performance, neither the coefficient on the Ln(Comp) nor CWP is significant. Meanwhile, the coefficients on Post09dum are negative and highly significant in both Models 3 and 4, which means the post-2009 stock performance in Chinese SOEs is worse than pre This is consistent with the descriptive statistics. Panel B reports the pay and subsequent two-year performance link. Only the coefficient on total cash compensation is highly significant in Model 5, which suggest a positive payoperating performance link. Post09dum are again significant in Models 7 and 8, which indicate the stock returns dropped significantly after Table 8 examines the pay-performance relation difference between SOEs and matched SOEs for the post-2009 period. The coefficients on the interactions are insignificant in all eight models. This indicates that the pay-performance link in the SOEs is not different to the matched non-soes after The 2009 Regulation seems to not to have an effect on the pay-performance relation. H3 is, therefore, accepted. 5. Robustness checks A number of sensitivity tests were conducted to ensure the robustness of the main results. First of all, for the second set of regressions, the pre-2009 non-soe data are also included in the regressions to ensure the post-2009 differences between SOEs and non-soes are due to the 2009 Regulation, not because of pre-2009 differences. An interaction term of Post09dum *SOE is created to re-run regressions testing H1 and H2. Results are reported in Table 9. As shown in Table 9, the only significant interaction term is in Model 2. This means the CWP ratios in the SOEs did decrease after 2009 relative to the non-soes, which confirms the findings from the main analysis in Table 5. Page 20 of 35