The Effect of Compensation Disclosure on Compensation Benchmarking: Evidence from China

The Effect of Compensation Disclosure on Compensation Benchmarking: Evidence from China Wei Jiang Department of Accounting, School of Management Center for Management Accounting Research Jinan University Guangzhou 510275, China Email: tweijiang@jnu.edu.cn Xinxin Liao School of Business Sun Yat-Sen University Guangzhou 510275, China Email: lxinx@mail2.sysu.edu.cn Bingxuan Lin College of Business Administration University of Rhode Island Kingston, RI 02881 Email: blin@uri.edu 1-401-874-4895 Yunguo Liu* School of Business Sun Yat-Sen University Guangzhou 510275, China Email: mnsygliu@mail.sysu.edu.cn Wei Jiang acknowledges the financial support of the National Natural Science Foundation of China (71272212), the Humanities and Social Science Foundation of the China Ministry of Education (11YJC630076), Institute of Enterprise Development at Jinan University (2014ZD001) and the Fundamental Research Funds for the Central Universities at Jinan University (12JNYH003). * Contact Author 1

The Effect of Compensation Disclosure on Compensation Benchmarking: Evidence from China Abstract Improved compensation disclosure might minimize unscrupulous compensation behavior by entrenched executives, and it could allow better benchmarking against peer groups. Meanwhile, with improved disclosure, executives can collectively defend their high salaries by engaging in opportunistic peer selection behavior. Better disclosure also forces companies to pay their executives market salaries. Using a sample of Chinese companies, we find that industry benchmarking is prevalent in China. Moreover, since the amended regulation for executive compensation disclosure in 2005, executives whose compensations are above the industry average have experienced much smaller pay raises, and executives whose compensations are below the industry average have had much higher pay raises. The results of our study are robust after controlling for various firm and industry characteristics. The results also show that companies controlled by different entities (i.e., central government, local government or non-government) behave very differently in response to enhanced compensation disclosure. These findings highlight the importance for policymakers of understanding how different firms react to improved disclosure, and how various firms face different incentives concerning disclosure. 2

The demand for greater corporate compensation disclosure has increased considerably in recent years, especially since the 2008 financial crisis. However, studies on executive compensation disclosure have yielded many different results. On the one hand, improved compensation disclosure is associated with better governance, higher pay-for-performance sensitivity and improved monitoring by stakeholders (Lo, 2003; Perry and Zenner, 2001; Vafeas and Afxentiou, 1998). On the other hand, compensation disclosure regulations appear to have very little effect in mitigating the problems involved in corporate compensation practices (Faulkender and Yang, 2013). There might also be unintended consequences of better disclosure, as disclosure may aggravate agency problems and affect related costs (Hermalin and Weisbach, 2012). Compensation disclosure in emerging markets can be especially difficult to analyze due to weak corporate governance, lax enforcement of security laws and a poor disclosure environment. Leuz and Wysocki (2008) point out that there have been major changes in disclosure regulations in many emerging markets. However, these regulations may have different effects in various markets due to the diversity of institutional and economic factors. Executive compensation disclosure in China is fairly limited. In 1998, the China Security Regulatory Commission (CSRC) required listed firms in China to disclose information about top executive compensation. However, this regulation only required companies to report the total compensation for the three highest-paid managers. As corporate governance in China slowly improved and demand for corporate disclosure became greater, the CSRC issued new rules in 2005. Under the 3

new disclosure regulation, all listed firms are required to report the compensation, including the salaries, bonuses, stipends and other benefits for the top three executives separately. This disclosure regulation of 2005 is in many ways comparable to the 1992 United States Securities and Exchange Commission (SEC) revisions to its rules governing disclosure of executive compensation. Although the CSRC 2005 disclosure regulation does not require firms to reveal their compensation peers (as the 2006 U.S. SEC rule does), it does give firms access to peer compensation information at the individual level, which allows for better benchmarking. The extant literature suggests that firms are likely to benchmark against their peers, and the use of benchmarking has a significant effect on CEO compensation (Bizjak et al., 2008; Bizjak et al., 2011; Albuquerque et al., 2013). In this study, we examine how executive compensation benchmarking behavior has changed in response to the new 2005 regulation. We find that under the amended regulation, executives whose compensations are above the industry average have experienced much smaller pay raises, and executives whose compensations are below the industry average have seen much higher pay raises. The results of our study are robust after controlling for various firm and industry characteristics. Furthermore, we find that companies controlled by different entities (i.e., central government, local government or non-government) behave very differently in response to enhanced compensation disclosure. Specifically, executives whose compensations are above the industry average generally receive lower pay increases in the subsequent year. After 2005, this effect has been stronger for firms controlled by the central government, but 4

weaker for firms controlled by local governments or non-government entities. Meanwhile, executives whose compensations are below the industry average have received higher pay increases in the subsequent year. This effect, however, has been weaker for firms controlled by the central government and stronger for firms controlled by local governments. Also, the 2005 regulation does not seem to have any effect on firms controlled by private entities. Our findings suggest that due to agency conflicts and labor market competition, firms controlled by different entities tend to react differently to the disclosure policy change. Our study contributes to the literature in the following ways. First, we provide a direct analysis of the 2005 CSRC compensation disclosure, showing that increased disclosure leads to both intended and unintended economic consequences that vary among different types of firms. Although China has become the world s largest economy, its governance and disclosure environment remain weak (Fan et al., 2007; Jiang et al., 2010). It is therefore important to understand how the requirement for compensation disclosure might have varying effects on corporate behavior. Second, we extend the study of Ezzamel and Watson (1998), which suggests that managers who are paid more than their peers and managers who are paid less than their peers face different pressures and incentives for inflating their pay, and they often do so by using peer groups as benchmarks. Finally, we contribute to the growing literature that examines CEO compensation in China. Previous studies on the Chinese market have examined many aspects of executive compensation, including pay-for-performance sensitivity (Firth et al., 2006; Gu et al., 2010), corporate governance (Conyon and He, 5

2011), managerial power and entrenchment (Chen et al., 2011; Lin and Lu, 2009) or executive compensation in family firms (Cheng et al., 2014). However, very few studies have explored benchmarking behavior in China. Our study fills this gap in the research and shows how benchmarking behavior can change in response to the new disclosure requirement. The remainder of this study is organized as follows. Section 2 discusses the related literature and introduces our hypotheses. Section 3 presents our sample description, the variable definitions and the empirical methodology. Section 4 reports our empirical analysis. Section 5 gives the results of robustness tests, and Section 6 offers conclusions from the study. 2. Literature Review and Hypotheses Development Corporate disclosure regulation can result in both firm-specific benefits and costs. On the one hand, greater disclosure is often associated with improved market liquidity (Verrecchia, 2001), reduced agency cost (Shleifer and Wolfenzon, 2002) and lower cost of capital (Lambert et al., 2007). On the other hand, improved disclosure exposes firms to indirect costs, as firm-specific information can disclose the firm s disadvantages to competitors or regulators (Verrecchia, 1983; Feltham et al., 1992). Ernstberger and Gruning (2013) show that a country s regulatory environment interacts with firm governance arrangements to affect the quality of disclosure. It is therefore helpful to examine the pros and cons of disclosure in different regulatory settings. Leuz and Wysocki (2008) review the literature on disclosure regulation and 6

suggest that the extant literature focuses heavily on regulatory changes in the U.S. market, and that the major regulatory or enforcement changes in other countries are largely ignored. Examining the effect of regulatory disclosure in China can therefore provide us with new insights with respect to how regulatory disclosure might result in different kinds of corporate behavior in regulatory environments outside the U.S. We specifically focus on the effect of compensation disclosure on compensation peer benchmarking in China 1. In the U.S., market compensation disclosure has become a focal point of public interest since the early 1990s. The SEC disclosure rules of 1992 and 2006 are major regulations that specify the information to be released in company compensation disclosures. Many studies have shown that better compensation disclosure results in improved corporate governance and better pay-for-performance sensitivity (Franco et al., 2013; Vafeas and Afxentiou, 1998; Ke et al., 1999; Lo, 2003). The SEC rule of 2006 requires firms to disclose the paysetting process by revealing which compensation peer groups are considered. This disclosure allows researchers to look inside the black box and explore how firms pick their peer groups. Faulkender and Yang (2010) analyze firms in the S&P 500 and the S&P 400 Midcap firms. These researchers find that firms tend to select highly paid peers to justify their CEO compensations. In a subsequent study, Faulkender and Yang (2013) conclude that strategic peer benchmarking has remained prevalent since the SEC 2006 disclosure requirement. The findings of these researchers suggest that 1 For a comprehensive review of compensation policy in China, please refer to Conyon and He (2011) and Beaulier et al. (2012). 7

disclosure regulation does very little to mitigate the agency problems involved in compensation practice. However, Cadman and Carter (2013) use a broader sample and find that opportunism is not the main motive behind such peer group selection. Compensation disclosure in China started in 1997, when the CSRC required all listed firms to disclose compensation information for their executives. However, the 1997 regulation was so vague that companies could often bury the compensation disclosure amidst lengthy corporate annual reports. The CSRC modified its rule in 2001 and required a separate section in the annual report dedicated to compensation disclosure. Furthermore, the CSRC 2001 rule required firms to disclose more specific compensation information, such as the process of setting the compensation, the total compensation, the sub-totals of compensation for the three most highly paid managers, the allowance for independent directors and the intervals of compensation (Beaulieu et al., 2012). In 2005, the CSRC issued another update and required that listed companies report individual executive compensations, instead of the aggregate compensations of the top three executives. Following the 2005 update, corporate reports disclosed executive compensation at the individual level for the first time. This change allowed companies to figure out the exact compensation earned by other executives in firms of the same industry or of similar size. Bizjak et al. (2008) show that the category of firms in the same industry is one of the most popular benchmarks used by companies, and that managers usually target their pay at or above the median (mean) level of their industry peers. We hypothesize that Chinese companies normally use the industry average compensation as their 8

benchmark to set executive compensation. In many cases, we have observed companies such as Shenzhen Wanke (ticker 000002) explicitly stating that they set their executive compensation based on the compensation level in the same industry. Given the more detailed disclosure requirements since 2005, firms have had easier access to executive pay information about their peers. Therefore, many firms find it convenient to benchmark against their peers in the same industry. Hence, we propose the following hypotheses: H1a: Chinese companies use industry benchmarking to determine executive compensation. H1b: This benchmarking behavior has grown more prevalent since 2005. Compensation benchmarking might result in different outcomes for executives who are paid above the benchmark and for those who are paid below it. Using social comparison and equity theories, Ezzamel and Watson (1998, 2002) find that external labor markets and internal pay comparisons are critical factors in determining executive pay. Furthermore, these researchers find evidence that shows asymmetric adjustment to prior-period pay anomalies. Specifically, the pay of relatively underpaid executives displays much higher sensitivity to comparison with external market pay levels. With greater compensation disclosure, we should expect that the underpaid executives 2 have a much stronger case for requesting higher pay. At the same time, we should expect that the over-paid executives are under greater pressure to 2 In this study, the terms under-/over-paid refer to compensation levels below/above industry average. 9

curb excessive compensation. We therefore suggest the following additional hypotheses: H2a: Executives whose compensations are above the industry average will receive lower pay increases in the subsequent year, and executives whose compensations are below the industry average will receive higher pay increases in the subsequent year. H2b: The asymmetric adjustments in salary between the over-paid and underpaid executives are more striking since 2005. Adjustments in salary are also closely linked to various other factors such as management incentives and labor market competition. One of the unusual characteristics of Chinese companies is their diversity in forms of ownership control. Firms can either be controlled by the government as state-owned enterprises (SOEs) or owned by private entities. Among SOEs, there are also major differences between SOEs affiliated with the central government (SOECGs) and SOEs affiliated with local governments (SOELGs). Chen et al. (2009) show that SOECGs are subject to strict supervision. Executives in SOECGs are appointed by the central government, and many of them eventually become vice ministers of state. The levels of compensation in these firms are thus less important to the executives than their political careers. In extreme cases, we can expect SOECG executives to sacrifice their compensation levels for the sake of career advancement. Hence, we would expect the pay of SOECG executives to 10

adjust more slowly, even if they have been relatively under-paid in the past. When compensation disclosure becomes more transparent after 2005, we expect managers in SOECGs to have fewer incentives to increase their monetary compensation, since they derive greater benefit from political advancement rather than direct compensations. Meanwhile, the central government has also issued several regulations to limit the compensation for SOECG executives, and this in turn would create greater pressure for over-paid executives to receive lower pay increases (even pay cut) if compensation information becomes public. Therefore, if there is greater transparency in compensation disclosure after 2005, SOECG executives who were over-paid compared to the industry average would expect to receive much lower pay increases. SOELGs, however, are subject to weaker supervision and management (Chen et al., 2009). Jiang et al. (2010) find that tunneling behavior is more severe for SOELGs than for SOECGs, which suggests a more severe agency problem for SOELGs. We therefore hypothesize that executives of SOELGs are more likely to engage in opportunistic benchmarking behavior. If peer compensation information becomes available, SOELG executives can justify a higher pay raise using selective benchmarks. We expect under-paid SOELG executives to increase their compensation faster after 2005. For over-paid executives, we also expect their salaries to be less responsive to the industry benchmarks, due to the heightened agency conflicts within SOELGs. They will have greater incentives to engage in selective benchmarking behavior in order to maintain their higher pay level. 11

For private non-soe firms, the compensation for executives is more directly driven by the competitive labor market, and we expect under-paid executives to have greater salary increases after 2005 when peer compensation information becomes more available. Overall, we propose the following hypotheses: H3a: Over-paid (under-paid) executives in SOECGs will receive less (more) pay increase in the following year; however, this pattern is more (less) pronounced since 2005. H3b: Over-paid (under-paid) executives in SOELGs and non-soe firms will receive less (more) pay increases in the following year, and this pattern is less (more) pronounced since 2005. 3. Data, Variable Definition and Empirical Methodology 3.1 Data and Definitions of Variables Our sample includes all firms listed in the China Stock Market and Accounting Research (CSMAR) database during the period between 2003 and 2007. We use data from the two years before and after 2005 to construct the sample for comparison. We exclude firms in the financial sector, because their financial data are not directly comparable to those of other firms. We also remove observations with missing financial information, ownership data or compensation information, and firms missing two consecutive years of information on managerial compensation, sales growth, debt, ROA or firm size. To mitigate the influence of possible spurious outliers, 12

we also winsorize all variables at the 1% and 99% level. We obtain a final sample of 2878 firm-year observations. One of the key variables of interest is the change in compensation from year t- 1 to year t. As companies only disclosed the sum of their top three executives compensations prior to 2005, we can only examine the compensation benchmarking behavior for the top three executives as a whole. We define Compen as the total compensation of the three highest-paid executives in the firm. Compen is computed as Compen in the current year (t) minus Compen in the previous year (t-1), scaled by Compen in year t-1. To measure the effect of peer benchmarking, we first identify peer firms as those operating in the same industry (as classified by the CSRC). We then define the mean and the median of the total compensation reported by these firms as PeerMean and PeerMed. The measure of peer benchmarking (BMark) is the difference between company compensation and peer compensation, which are defined as BmarkMean = (PeerMean-Compen)/PeerMean, or BmarkMed = (PeerMed- Compen)/PeerMed. Hence, if executives are paid above the industry average, BmarkMean and BmarkMed should be negative. To facilitate the interpretation of our results, we use the absolute value of these variables when conducting the empirical tests. Following previous studies, we control for corporate governance and company financial characteristics. As in the studies by Albuquerque et al. (2013), Bizjak et al. (2008; 2011) and Cadma and Carter (2014), we control for the ultimate owner of the firm. Ownership equals 1 if a firm is controlled by the state, and 0 otherwise. We also 13

control for ownership by the largest shareholders. Topshare is measured as the number of shares owned by the largest shareholder divided by the total shares outstanding. Duality equals 1 if the CEO is also the chairman of the board, and 0 otherwise. Board independence (BIndepen) is proxied by the number of independent directors in relation to the total number of directors. Board size (Bsize) is the natural log of the number of directors on the board. Growth represents change in sales growth, with sales growth measured by (Sales t -Sales t-1 )/Sales t-1. Lev represents change in leverage, and is measured by (Lev t -Lev t-1 )/Lev t-1. Lev is defined as total debts over total assets. ROA represents change in return on total assets, and is measured by NI t /Total Asset t NI t-1 /Total Asset t-1. Size is the change in firm size, and firm size is measured as the natural log of total assets. To control for the effect of the 2005 split-share structure reform (Liao et al., 2014), we also control for the number of non-tradable shares (Ntradeshares) issued by the firm. Finally, we define a dummy variable Period05 that equals 1 if the observation occurs after 2005, and 0 otherwise. Insert Table 1 Here Table 2 shows the summary of statistics for our sample. The mean (median) value for the dependent variable, Compen, is 0.0175 (0.0114). BmarkMean has a mean of 0.0271 and a median of 0.0245. BmarkMed has very similar results, suggesting that very little difference exists between the two industry peer benchmarks. We also find that 69.85% of our observations are SOEs. We therefore further classify our sample into SOEs controlled by the central government and SOEs controlled by 14

local governments. We classify 11.02% of our observations as SOECGs and 58.83% as SOELGs. Insert Table 2 Here 3.2 Empirical Models Following Bizjak, Lemmom and Naveen (2008), we use the following two models to test the effects of industry compensation benchmarking: Compen = β 0 + β 1 BMark + β 2 Period05 + β 3 Ownership + β 4 Topshare + β 5 Duality + β 6 BIndepen + β 7 Bsize + β 8 Growth + β 9 Lev + β 10 ROA + β 11 Size + β 12 Ntradeshares + Σ Year + Σ Industry (1) Compen = β 0 + β 1 Bmark + β 2 Bmark Period05 + β 3 Period05 + β 4 Ownership + β 5 Topshare + β 6 Director + β 7 Indepen + β 8 Bsize + β 9 Growth + β 10 Lev + β 11 ROA + β 12 Size +β 13 Ntradeshares + Σ Year + Σ Industry (2) In both models we control for the year and industry fixed effects. The industry benchmarking measure Bmark is either the mean compensation benchmark (BmarkMean) or the median compensation benchmark (BmarkMed). 4. Empirical Results 15

The results for model 1 using the full sample are shown in Table 3. The coefficient for BmarkMean (BmarkMed) is 0.1747 (0.1871) and is significant at the 1% level. This result suggests that industry benchmarking is evident in the Chinese market. We then divide the full sample into two groups: those in which executives are paid above the mean level of their industry peers, and those in which executives are paid below that level. In general, increases in size and firm performance ( ROA) are associated with positive pay increases. It is worth noting that BmarkMean is negative if the executives are paid more than their peers. To make interpretation easier, we use the absolute value of BmarkMean in these tests. We can then interpret the negative and significant coefficient for BmarkMean as evidence suggesting that executives who are paid above the industry average tend to subsequently receive lower pay raises. The coefficients for BmarkMean and BmarkMed are -0.4711 and -0.4823, respectively. For the group of executives paid below the industry average, the coefficients for BmarkMean and BmarkMed are both positive and they are significant. These results suggest that the more an executive is under-paid, the greater a pay increase she will receive in the subsequent year. These results are consistent with hypotheses 1a and 2a. Insert Table 3 Here We next run model 2 and examine the effects of the 2005 compensation disclosure requirement. The results are shown in Table 4. Interestingly, the 16

coefficients for BmarkMean 3 are positive and significant for the full sample and for two sub-samples. However, the coefficient for the interaction term between BmarkMean and the Period05 dummy variable is insignificant, which suggests no significant effect since the 2005 compensation disclosure requirement. Further analysis using subsamples reveals that the coefficients for the interaction term are both significant. If executives are paid above the industry average, they will thus receive lower pay raises in the subsequent year (negative coefficient for BmarkMean). However, this effect is weaker since 2005 (positive coefficient for the interaction term). These findings suggest that in response to requirements for more transparent disclosure, executives might engage in opportunistic benchmarking, and therefore find it easy to maintain their high compensation levels. This result is also consistent with a greater completion in the labor market after 2005. In order to retain talented executives, firms have to offer a more competitive compensation package (Bryson et al. 2014). Meanwhile, those executives who are paid below the industry average will receive higher pay raises in the subsequent year (the coefficient for BmarkMean is 0.1753). After 2005, we observe a faster adjustment rate (the coefficient for the interaction term is -0.0014). If an executive perceives her pay as below the market average, then compensation disclosure by her peer companies allows her to justify a greater increase in compensation. It is therefore important to realize that compensation disclosure might have different effects on the benchmarking behavior 3 In the following analysis, the results using BmarkMean and BmarkMed are all qualitatively similar. Therefore, we only report results using BmarkMean in the subsequent portions of our study. 17

of over-paid and under-paid executives. Overall, our results as shown in Table 4 support hypotheses 2b. Insert Table 4 Here Our final test examines the differences in benchmarking behavior among SOECG, SOELG and Non-SOE firms. As the over-paid and under-paid executives display different benchmarking characteristics, we first divide our sample into these two groups. Then we run the tests for the SOECG, SOELG and Non-SOE samples separately. In Table 5, we see that if executives are paid above the industry average, they will receive lower pay increases subsequently (the coefficients for BmarkMean are negative for all three sub-samples). Interestingly, the interaction term between BmarkMean and Period05 is negative for the SOECG sample. This result shows the executives of SOECGs have received lower increases in compensation since the improved disclosure requirement of 2005. This finding is consistent with the common observation that SOECGs are highly monitored, and their executives are not mainly incentivized by levels of compensation. However, the opposite situation applies for SOELGs and Non-SOEs. It seems that executives in these types of firms are more likely to engage in opportunistic benchmarking to secure their higher pay levels. For the under-paid executives, we see in Column B of Table 5 that the coefficients for BmarkMean are all positive, which suggests an upward pay adjustment in the subsequent year. Although the executives of SOECGs received lower pay raises after 2005, the executives of SOELGs received greater increases in pay. For Non-SOEs, the coefficient for the interaction term is insignificant. The striking differences among 18

the SOECG, SOELG and Non-SOE firms suggest that different agency issues and incentives directly affect the outcomes of the disclosure policy. Specifically, transparency in compensation deters executives in SOECGs from acquiring excessive compensation, but executives in SOELGs and Non-SOEs may use the disclosed information to engage in opportunistic peer benchmarking. Insert Table 5 Here 5. Robustness Check In the previously reported tests, we use all firms in the same industry as benchmark firms. Some observers, however, might argue that firms tend to benchmark against peers with the same ownership types. To ensure that our results are robust, we re-run all of the tests using only firms with the same ownership types within the same industry as benchmark firms. Although this method reduces the number of peer firms in the same industry, our results as shown in Tables 6 through 8 are very similar to those reported in Tables 3 through 5. Instead of using all firms in the same industry, we follow the approach taken by Brookman and Thistle (2013) and define peer companies as firms of comparable size (0.5-2 times the firm size) in the same industry. The results, shown in Tables 9, 10 and 11 are also very similar to those reported earlier. 6. Conclusions 19

Improved compensation disclosure allows firms to better benchmark executive compensation against their peers. However, such benchmarking can also result in opportunistic behavior in which managers strategically choose their peers to inflate their overall compensation. Studies on the 2006 SEC regulation have found mixed evidence with respect to the effect of greater disclosure on compensation contracting. As the pros and cons of disclosure are closely related to institutional and market conditions, we examine the effects of the compensation disclosure rule in China as issued by the CSRC in 2005. We find evidence suggesting industry peer benchmarking in China. Over-paid executives tend to receive lower pay increases in the subsequent year. However, it seems that these over-paid executives have been able to use peer disclosure to justify their compensation and to reduce the effect of downward pay adjustment. We find that under-paid executives tend to receive significant pay increases in the subsequent period, and this pattern has grown stronger since 2005 when detailed information on executive compensation became available. Overall, it seems that improved compensation disclosure has had an overall positive effect on the levels of executive compensation. More importantly, we show that firms with different ownership types behave differently toward compensation disclosure. We provide additional evidence showing that executives in SOECGs are under stricter monitoring, and both the under-paid and over-paid executives have been more likely to reduce their pay increases since 2005. However, executives of SOELG and non-soe firms are more likely to use improved compensation disclosure to secure higher pay. 20

Our study explores the effect of compensation disclosure in an emerging market, and the results show how firms with different forms of ownership benchmark executive compensation differently for their under- or over-paid executives. In future studies, it would be interesting to examine whether China continues to improve its compensation disclosure and how compensation contracting, specifically benchmarking behavior, might further evolve as more and more Chinese companies are listed on foreign exchanges. 21

References: Albuquerque, A., Franco, G. and Verdi, R., 2013, Peer Choice in CEO Compensation. Journal of Financial Economics, 108, 160-181. Beaulieu P, Ding S, Qu B. 2012, Compensation disclosure in China., Asia-Pacific Journal of Accounting & Economics, 19, 253-277. Bizjak, J., Lemmom, M. and Naveen, L., 2008, Does the Use of Peer Groups Contribute to Higher Pay and Less Efficient Compensation? Journal of Financial Economics, 90, 152-168. Bizjak, J., Lemmom, M. and Nguyen,T., 2011, Are All CEOs above Average? An Empirical Analysis of Compensation Peer Groups and Pay Design. Journal of Financial Economics, 100, 538-555. Brookman J T, Thistle P D. 2013, Managerial compensation: Luck, skill or labor markets? Journal of Corporate Finance, 21, 252-268. Bryson A, J Forth and M Zhou (2014), "Same or Different? The CEO Labour Market in China's Public Listed Companies", The Economic Journal, 124, F90-F108 Cadman B, Carter M E. 2013, Compensation peer groups and their relation with CEO pay. Journal of Management Accounting Research, 26, 57-82. Chen, G, Firth, M, Xu, L,2009. Does the type of ownership control matter? Evidence from China s listed companies. Journal of Banking and Finance, 33, 171-181. Chen, J., Ezzamel M. and Cai, Z., 2011, Managerial power theory, tournament theory, and executive pay in China., Journal of Corporate Finance, 17, 1176 1199. Cheng, M.Y, Lin, B.X., and Wei. M.H., 2014, Executive compensation in family firms: The effect of multiple family members., Journal of Corporate Finance, forthcoming. Conyon, M. and He, L., 2011. Executive compensation and corporate governance in China Journal of Corporate Finance, 17, 1158 1175. Ernstberger, J. and Grüning, M., 2013. How do Firm- and Country-Level Governance Mechanisms Affect Firms Disclosure? Journal of Accounting and Public Policy, 32, 50-67. Ezzamel M., and Watson, R., 1998, Market Comparison Earnings and the Bidding-Up of Executive Cash Compensation: Evidence from the United Kingdom. Academy of Management Journal, 41, 221-231. Ezzamel M, Watson R. 2002, Pay comparability across and within UK boards: An empirical analysis of the cash pay awards to CEOs and other board members. Journal of Management Studies, 39, 207-232. Fan, J., Wong, T. and Zhang, T., 2007, Politically connected CEOs, corporate governance, and Post-IPO performance of China s. Journal of Financial Economics, 84, 330-357. Faulkender, M., and Yang, J., 2010, Inside the Black Box: the Role and Composition of Compensation Peer Groups. Journal of Financial Economics, 96, 257-270. Faulkender, M., and Yang, J., 2013. Is Disclosure an Effective Cleansing Mechanism? The Dynamics of Compensation Peer Benchmarking. Review of Financial Studies, 20, 806-839. Feltham, G.A. and Xie J.Z., 1992, Voluntary financial disclosure in an entry game with continua of types. Contemporary Accounting Research, 9, 46-80. Firth, M., Fung, P., & Rui, O. M. 2006. Corporate performance and CEO compensation in China. Journal of Corporate Finance, 12, 693 714. 22

De Franco G, Hope O K, Larocque S. 2013, The effect of disclosure on the pay-performance relation. Journal of Accounting and Public Policy, 32, 319-341. Gu, Z., Wang, K., Xiao, X., 2010. Government Control and Executive Compensation: Evidence from China. Working Paper. Hermalin B E, Weisbach M S. 2012, Information disclosure and corporate governance. Journal of Finance, 67, 195-233. Jiang, G., Lee, C., Yue, H., 2010. Tunneling through intercorporate loans: The China experience. Journal of Financial Economics, 98, 1 20. Ke, B., Petroni, K. & Safieddine, A, 1999. Ownership concentration and sensitivity of executive pay to accounting performance measures: Evidence from publicly and privately-held insurance companies. Journal of Accounting and Economics, 28, 185 209. Lambert R, Leuz C, Verrecchia R E. 2007, Accounting information, disclosure, and the cost of capital. Journal of Accounting Research, 45, 385-420. Leuz, C., and P. Wysocki. 2008. Economic consequences of financial reporting and disclosure regulation: A review and suggestions for future research. Working Paper. Liao L, Liu B, Wang H. 2014, China's Secondary Privatization: Perspectives from the Split- Share Structure Reform. Journal of Financial Economics, 113, 500-518. Lin B X, Lu R. 2009, Managerial power, compensation gap and firm performance Evidence from Chinese public listed companies. Global Finance Journal, 20, 153-164. Lo K. 2003, Economic consequences of regulated changes in disclosure: the case of executive compensation. Journal of Accounting and Economics, 35, 285-314. Perry T, Zenner M. 2001, Pay for performance? Government regulation and the structure of compensation contracts. Journal of Financial Economics, 62, 453-488. Shleifer A, Wolfenzon D. 2002, Investor protection and equity markets. Journal of Financial Economics, 66, 3-27. Vafeas N, Afxentiou Z. 1998, The association between the SEC's 1992 compensation disclosure rule and executive compensation policy changes. Journal of Accounting and Public Policy, 17, 27-54. Verrecchia R E. 1983, Discretionary disclosure. Journal of Accounting and Economics, 5: 179-194. Verrecchia R E. 2001, Essays on disclosure. Journal of Accounting and Economics, 32, 97-180. 23

Table 1: Variable Definition Variables Compen Definition computed as Compen in the current year (t) minus Compen the previous year (t-1), and scaled by Compen in year t-1, where Compen is the total compensation of the three highestpaid executives in the firm The difference between managerial pay and industry peers BmarkMean BmarkMed (PeerMean-Compen)/PeerMean, where PeerMean is the mean of the total compensation reported by firms operating in the same industry (PeerMed-Compen)/PeerMed, where PeerMed is the median of the total compensation reported by firms operating in the same industry Period05 Ownership SOECG SOELG Topshare Duality BIndepen Bsize Growth Lev ROA Size Ntradeshares dummy variable, taking the value 1 if the year is 2006 or 2007, and 0 otherwise dummy variable, taking the value 1 if the ultimate controller is the state, and 0 otherwise dummy variable, taking the value 1 if the ultimate controller is the central-government, and 0 otherwise dummy variable, taking the value 1 if the ultimate controller is the local-government, and 0 otherwise proportion of shareholdings of the largest shareholder dummy variable, taking the value 1 if CEO is the Chairman, and 0 otherwise ratio of number of independent directors to number of total directors natural log size of the board change of sales growth during period t and period t-1, where sales growth is change of sales during period t and period t-1, and scaled by sales in year t-1 change of leverage during period t and period t-1, where leverage is the ratio of total debts to total assets change of ROA during period t and period t-1, where ROA is the ratio of net incomes to total assets change of firm size during period t and period t-1, where firm size is the natural log of total assets dummy variable, taking the value 1 if non-tradable shares of the firm are above the mean of all firms in the same industry, and 0 otherwise 24

Table 2: Descriptive Statistics Please refer to table 1 for variable definitions. The total number of observation for the sample is 3830. We report the minimum (Min), maximum (Max), mean, median and standard deviation (STD) for the complete sample here. Variable Min Max Mean Median STD Compen (ten thousands RMB) 18 1400 73 54 75 Compen -0.0795 0.1431 0.0175 0.0114 0.0344 BmarkMean -0.1729 0.2047 0.0271 0.0245 0.0655 BmarkMed -0.1682 0.2097 0.0330 0.0298 0.0654 Ownership 0.0000 1.0000 0.6985 1.0000 0.4590 SOECG 0.0000 1.0000 0.1102 0.0000 0.3132 SOELG 0.0000 1.0000 0.5883 1.0000 0.4922 Topshare 0.0899 0.7678 0.3928 0.3721 0.1631 Duality 0.0000 1.0000 0.1151 0.0000 0.3192 BIndepen 0.0000 0.5455 0.3445 0.3333 0.0521 Bsize 5.0000 15.0000 9.5961 9.0000 2.0385 Growth -2.9544 2.7859 0.0263 0.0191 0.5858 Lev -0.2219 0.2910 0.0171 0.0111 0.0835 ROA -0.2493 0.1760-0.0009 0.0008 0.0532 Size -0.4192 0.9642 0.1230 0.0962 0.2121 Ntradeshares 0.0000 1.0000 0.4962 0.0000 0.5000 25

Table 3: Test on Compensation Industry Benchmarking Please refer to table 1 for variable definition. Benchmark peer firms are firms within the same industry. We run the following model here Compen = β 0 + β 1 BMark + β 2 Period05 + β 3 Ownership + β 4 Topshare + β 5 Duality + β 6 BIndepen + β 7 Bsize + β 8 Growth + β 9 Lev + β 10 ROA +β 11 Size +β 12 Ntradeshares +Σ Year +Σ Industry. P-values are shown in parenthesis. *** significant at the1% level; ** significant at the 5% level, * significant at the 10% level. Full sample Compen > PeerMean Compen > PeerMed Compen < PeerMean Compen < PeerMed (1) (2) (3) (4) (5) (6) _cons 0.0026 0.0009 0.0196** 0.0187** -0.0127-0.0117 (0.716) (0.897) (0.027) (0.032) (0.121) (0.158) BmarkMean 0.1747*** -0.4711*** 0.3573*** (0.000) (0.000) (0.000) BmarkMed 0.1871*** -0.4823*** 0.3404*** (0.000) (0.000) (0.000) Period05 0.0011 0.0020 0.0047* 0.0063** 0.0014 0.0020 (0.556) (0.278) (0.057) (0.013) (0.512) (0.356) Ownership 0.0047*** 0.0049*** 0.0012 0.0010 0.0021 0.0016 (0.005) (0.003) (0.547) (0.607) (0.274) (0.402) Topshare 0.0002 0.0002-0.0016-0.0010-0.0016-0.0047 (0.970) (0.964) (0.762) (0.861) (0.768) (0.369) Duality -0.0016-0.0016 0.0054** 0.0053** -0.0041* -0.0042* (0.424) (0.406) (0.022) (0.030) (0.058) (0.053) BIndepen 0.0045 0.0058 0.0038 0.0031 0.0006 0.0001 (0.723) (0.645) (0.809) (0.851) (0.965) (0.992) Bsize 0.0000 0.0001 0.0010*** 0.0011*** 0.0004 0.0003 (0.879) (0.743) (0.006) (0.003) (0.359) (0.389) Growth -0.0010-0.0011-0.0003-0.0001-0.0025** -0.0026** (0.399) (0.350) (0.817) (0.966) (0.043) (0.041) Lev -0.0173* -0.0179** -0.0449*** -0.0428*** -0.0137-0.0163* (0.051) (0.041) (0.000) (0.000) (0.175) (0.097) ROA 0.0437*** 0.0430*** -0.0076 0.0026 0.0343** 0.0388*** (0.001) (0.001) (0.669) (0.892) (0.020) (0.007) Size 0.0158*** 0.0166*** 0.0239*** 0.0194*** 0.0156*** 0.0176*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Ntradeshares -0.0014-0.0015-0.0030* -0.0032* 0.0010 0.0015 (0.380) (0.338) (0.096) (0.082) (0.579) (0.401) Year Y Y Y Y Y Y Industry Y Y Y Y Y Y N 2878 2878 1253 1147 1625 1731 adj. R2 0.075 0.086 0.445 0.453 0.293 0.272 26

Table 4: Effects of the 2005 Compensation Disclosure on Compensation Benchmarking Please refer to table 1 for variable definition. Benchmark peer firms are firms within the same industry. We run the following test: Compen = β 0 + β 1 BMark + β 2 BMark Period05 + β 3 Period05+ β 4 Ownership+ β 5 Topshare + β 6 Director + β 7 Indepen + β 8 Bsize + β 9 Growth+ β 10 Lev + β 11 ROA + β 12 Size +β 13 Ntradeshares + Σ Year + Σ Industry. P-values are shown in parenthesis. *** significant at the1% level; ** significant at the 5% level, * significant at the 10% level. Full Sample Compen > PeerMean Compen <PeerMean (1) (2) (3) _cons 0.0026 0.0198* -0.0068 (0.722) (0.093) (0.426) BmarkMean 0.1753*** -0.1388*** 0.2980*** (0.000) (0.007) (0.000) BmarkMean Period05-0.0014 0.1224** 0.0957*** (0.962) (0.045) (0.005) Period05 0.0012 0.0006-0.0069** (0.641) (0.882) (0.035) Ownership 0.0047*** 0.0040 0.0020 (0.005) (0.123) (0.315) Topshare 0.0002 0.0014-0.0007 (0.970) (0.843) (0.902) Duality -0.0015 0.0044-0.0049** (0.426) (0.163) (0.030) BIndepen 0.0045-0.0010-0.0009 (0.723) (0.962) (0.950) Bsize 0.0000-0.0003 0.0003 (0.878) (0.524) (0.453) Growth -0.0010 0.0026-0.0025* (0.399) (0.172) (0.054) Lev -0.0173* -0.0255* -0.0162 (0.051) (0.065) (0.123) ROA 0.0437*** 0.0215 0.0325** (0.001) (0.361) (0.034) Size 0.0158*** 0.0096* 0.0160*** (0.000) (0.073) (0.000) Ntradeshares -0.0014-0.0014 0.0009 (0.380) (0.561) (0.617) Year Y Y Y Industry Y Y Y N 2878 1253 1625 adj. R 2 0.075 0.031 0.234 27

Table 5: Effects of the 2005 Compensation Disclosure on Compensation Benchmarking for Different Types of Firms Please refer to table 1 for variable definition. Benchmark peer firms are firms within the same industry. We run the following test: Compen = β 0 + β 1 BMark + β 2 BMark Period05 + β 3 Period05+ β 4 Ownership+ β 5 Topshare + β 6 Director + β 7 Indepen + β 8 Bsize + β 9 Growth+ β 10 Lev + β 11 ROA+ β 12 Size +β 13 Ntradeshares + Σ Year + Σ Industry. P-values are shown in parenthesis. *** significant at the1% level; ** significant at the 5% level, * significant at the 10% level. Compen > PeerMean Compen < PeerMean SOECG SOELG Non-SOEs SOECG SOELG Non-SOEs (1) (2) (3) (4) (5) (6) _cons 0.0375 0.0288** 0.0121-0.0312-0.0081 0.0004 (0.165) (0.013) (0.525) (0.269) (0.430) (0.980) BmarkMean -0.2794*** -0.5873*** -0.6196*** 0.4735*** 0.2741*** 0.3876*** (0.006) (0.000) (0.000) (0.000) (0.000) (0.000) BmarkMean Period05-0.1891* 0.1834*** 0.1522** -0.2005* 0.1895*** 0.0624 (0.079) (0.000) (0.011) (0.068) (0.000) (0.246) Period05-0.0106 0.0060* -0.0037 0.0243** -0.0118*** -0.0002 (0.423) (0.052) (0.476) (0.040) (0.003) (0.964) Topshare 0.0029 0.0005 0.0019 0.0020-0.0008-0.0057 (0.845) (0.940) (0.855) (0.916) (0.892) (0.602) Duality 0.0227* 0.0041 0.0030-0.0018-0.0047 0.0001 (0.092) (0.219) (0.414) (0.822) (0.106) (0.985) BIndepen -0.0188-0.0123 0.0212-0.0044 0.0085-0.0308 (0.674) (0.541) (0.535) (0.946) (0.617) (0.265) Bsize 0.0003 0.0010** 0.0016* 0.0009 0.0001 0.0003 (0.715) (0.040) (0.076) (0.480) (0.762) (0.694) Growth 0.0034-0.0034 0.0007-0.0040-0.0013-0.0030 (0.280) (0.120) (0.758) (0.375) (0.492) (0.107) Lev 0.0096-0.0550*** -0.0601*** -0.0318-0.0241* 0.0004 (0.670) (0.001) (0.001) (0.418) (0.087) (0.983) ROA 0.0808** -0.0129-0.0609** -0.0008 0.0404* 0.0289 (0.045) (0.635) (0.041) (0.988) (0.053) (0.218) Size 0.0065 0.0310*** 0.0254*** -0.0003 0.0195*** 0.0169** (0.440) (0.000) (0.000) (0.988) (0.000) (0.012) Ntradeshares 0.0057-0.0050** 0.0020-0.0078 0.0023 0.0013 (0.226) (0.046) (0.603) (0.255) (0.372) (0.693) Year Y Y Y Y Y Y Industry Y Y Y Y Y Y N 188 727 338 145 976 504 adj. R 2 0.455 0.448 0.517 0.188 0.286 0.344 28

Table 6 Test on Compensation Industry Benchmarking Please refer to table 1 for variable definition. Benchmark peer firms are firms with the same ownership property within the same industry. We run the following model here Compen = β 0 + β 1 BMark + β 2 Period05 + β 3 Ownership + β 4 Topshare + β 5 Duality + β 6 BIndepen + β 7 Bsize + β 8 Growth + β 9 Lev + β 10 ROA +β 11 Size +β 12 Ntradeshares +Σ Year +Σ Industry. P-values are shown in parenthesis. *** significant at the1% level; ** significant at the 5% level, * significant at the 10% level. Full Sample Compen > PeerMean Compen < PeerMean (1) (2) (3) _cons -0.0004 0.0192* -0.0165* (0.963) (0.071) (0.094) BmarkMean 0.1959*** -0.4954*** 0.3800*** (0.000) (0.000) (0.000) Period05 0.0044** 0.0035 0.0052** (0.024) (0.146) (0.016) Ownership 0.0057*** 0.0058** 0.0039* (0.005) (0.024) (0.078) Topshare -0.0038-0.0028-0.0073 (0.450) (0.648) (0.199) Duality -0.0007 0.0067** -0.0031 (0.754) (0.013) (0.230) BIndepen -0.0014 0.0002-0.0166 (0.929) (0.992) (0.342) Bsize 0.0001 0.0011** 0.0003 (0.771) (0.012) (0.562) Growth -0.0017 0.0006-0.0022 (0.224) (0.730) (0.132) Lev -0.0250** -0.0452*** -0.0204* (0.011) (0.000) (0.052) ROA 0.0338** -0.0214 0.0426*** (0.030) (0.328) (0.008) Size 0.0146*** 0.0192*** 0.0144*** (0.000) (0.000) (0.000) Ntradeshares -0.0008-0.0053** 0.0004 (0.684) (0.023) (0.833) Year Y Y Y Industry Y Y Y N 2071 825 1246 adj. R 2 0.080 0.464 0.267 29

Table 7: Effect of the 2005 Compensation Disclosure on Compensation Benchmarking Please refer to table 1 for variable definition. Benchmark peer firms are firms with the same ownership property within the same industry. We run the following test: Compen = β 0 + β 1 BMark + β 2 BMark Period05 + β 3 Period05+ β 4 Ownership+ β 5 Topshare + β 6 Director + β 7 Indepen + β 8 Bsize + β 9 Growth+ β 10 Lev + β 11 ROA+ β 12 Size +β 13 Ntradeshares + Σ Year + Σ Industry. P-values are shown in parenthesis. *** significant at the1% level; ** significant at the 5% level, * significant at the 10% level. Full Sample Compen > PeerMean Compen < PeerMean (1) (2) (3) _cons 0.0014 0.0220** -0.0123 (0.878) (0.038) (0.216) BmarkMean 0.1648*** -0.6314*** 0.3142*** (0.000) (0.000) (0.000) BmarkMean Period05 0.0476 0.1763*** 0.1067*** (0.190) (0.000) (0.006) Period05 0.0015 0.0011-0.0022 (0.601) (0.658) (0.517) Ownership 0.0057*** 0.0056** 0.0040* (0.005) (0.027) (0.069) Topshare -0.0037-0.0047-0.0069 (0.464) (0.441) (0.229) Duality -0.0008 0.0066** -0.0033 (0.710) (0.014) (0.198) BIndepen -0.0014-0.0019-0.0168 (0.931) (0.925) (0.336) Bsize 0.0001 0.0010** 0.0003 (0.766) (0.019) (0.498) Growth -0.0016 0.0004-0.0021 (0.232) (0.821) (0.146) Lev -0.0248** -0.0425*** -0.0203* (0.012) (0.001) (0.052) ROA 0.0335** -0.0261 0.0429*** (0.031) (0.230) (0.008) Size 0.0146*** 0.0196*** 0.0145*** (0.000) (0.000) (0.000) Ntradeshares -0.0008-0.0049** 0.0001 (0.680) (0.032) (0.955) Year Y Y Y Industry Y Y Y N 2071 825 1246 adj. R 2 0.080 0.473 0.271 30

Table 8: Effects of the 2005 Compensation Disclosure on Compensation Benchmarking for Different Types of Firms Please refer to table 1 for variable definition. Benchmark peer firms are firms with the same ownership property within the same industry. We run the following test: Compen = β 0 + β 1 BMark + β 2 BMark Period05 + β 3 Period05+ β 4 Ownership+ β 5 Topshare + β 6 Director + β 7 Indepen + β 8 Bsize + β 9 Growth+ β 10 Lev + β 11 ROA + β 12 Size +β 13 Ntradeshares + Σ Year + Σ Industry. P-values are shown in parenthesis. *** significant at the1% level; ** significant at the 5% level, * significant at the 10% level. Compen > PeerMean Compen < PeerMean SOECG SOELG Non-SOEs SOECG SOELG Non-SOEs (1) (2) (3) (4) (5) (6) _cons 0.0308 0.0360*** 0.0125-0.0325-0.0061-0.0132 (0.558) (0.006) (0.600) (0.378) (0.645) (0.503) BmarkMean -0.3448-0.6016*** -0.7701*** 0.7803*** 0.2922*** 0.3649*** (0.581) (0.000) (0.000) (0.003) (0.000) (0.000) BmarkMean Period05-0.2070 0.1552*** 0.2629*** -0.4542* 0.1627*** 0.0609 (0.749) (0.006) (0.003) (0.095) (0.001) (0.415) Period05-0.0018 0.0048-0.0075 0.0259-0.0048-0.0018 (0.965) (0.101) (0.180) (0.201) (0.283) (0.782) Topshare -0.0072-0.0003 0.0026-0.0095-0.0036-0.0072 (0.767) (0.969) (0.849) (0.615) (0.631) (0.548) Duality -0.0008 0.0042 0.0032-0.0207** -0.0031 0.0015 (0.960) (0.240) (0.513) (0.046) (0.384) (0.725) BIndepen 0.1236-0.0118 0.0040 0.0330-0.0042-0.0546* (0.209) (0.634) (0.923) (0.574) (0.858) (0.098) Bsize -0.0030 0.0010* 0.0012 0.0001-0.0003 0.0011 (0.101) (0.051) (0.261) (0.929) (0.630) (0.253) Growth 0.0036-0.0038 0.0018 0.0014-0.0023-0.0027 (0.447) (0.122) (0.551) (0.704) (0.326) (0.249) Lev -0.0777-0.0463*** -0.0465* -0.0401-0.0277* -0.0191 (0.108) (0.009) (0.051) (0.137) (0.084) (0.265) ROA 0.0753-0.0174-0.0578-0.0173 0.0451* 0.0403 (0.408) (0.578) (0.105) (0.734) (0.067) (0.114) Size -0.0152 0.0282*** 0.0215** 0.0102 0.0157** 0.0127* (0.359) (0.000) (0.015) (0.321) (0.011) (0.078) Ntradeshares -0.0138-0.0086*** 0.0058-0.0005-0.0009 0.0004 (0.104) (0.003) (0.257) (0.931) (0.764) (0.923) Year Y Y Y Y Y Y Industry Y Y Y Y Y Y N 77 532 216 146 692 408 adj. R 2 0.505 0.468 0.530 0.268 0.284 0.247 31