Institutional Background

The ASMM is a penalty for material misstatements made by companies, with executives and others responsible held accountable and the outcome made public. The Chinese capital market has experienced countless misstatement scandals, drawing constant attention from regulators. As shown in Figure 1, penalties imposed by the CSRC on listed companies for information disclosure violations have increased considerably in the past two decades and accounted for a significant proportion of corporate violations.

To enhance executives’ information disclosure responsibilities and quality in annual reports, the ASMM was first introduced in late 2009, with the CSRC's Announcement on Preparation of the Annual Reports and Related Work of Listed Firms in 2009 (Announcement No. 34)5 and has been established by listed firms since 2010. With an ASMM, companies are required to truthfully provide an itemized disclosure of the causes and effects of misreporting in their annual reports, the accountability measures taken by the board against those responsible, and the outcomes of handling. Compared with previous regulatory reforms, which mainly were aimed at the board of directors (Kumar and Sivaramakrishnan, 2008; Liao et al., 2022) and focused on financial fraud (Yang et al., 2017), this approach not only increases executive responsibility for information disclosure but also represents a new chapter in the governance of companies’ disclosure of irregularities in the Chinese capital markets.

The ASMM has three main features. First, the scope of accountability for misstatements is broad. The accountability scope includes material misstatements, including corrections of material accounting errors, additions of material omissions, and revisions of earnings forecasts during the annual, semi-annual, and quarterly reports. Second, the target of accountability is comprehensive. Executives such as the chairman, CEO, CFO, and the head of the accounting department are responsible for the truthfulness, accuracy, completeness, timeliness, and fairness of the information disclosed and the financial reporting.6 Third, the accountability measures are severe. These accountability measures include, but are not limited to, financial penalties, notification of criticism, demotion, and dismissal. Depending on the circumstances of the misstatement, accountability measures can be combined, and misstatements can even be referred to the judicial authorities if they are serious and suspected of being criminal. As a result, those responsible for disclosure violations have greater accountability, and the accountability results disclosed in annual reports are more widely disseminated under the ASMM than in the past.

The ASMM also reforms corporate governance by combining public and private enforcement. On the one hand, the ASMM proposed by the CSRC is mandatory to a certain extent. Ideally, all companies should actively follow the call to establish an ASMM. However, due to China's weak law enforcement environment, listed companies did not implement the policy simultaneously when it was introduced. Instead, the introduction was staggered and some companies have yet to establish it, despite the CSRC stressing in several announcements that listed companies should establish it. Moreover, due to resource constraints, CSRC supervision of companies is selective and lagging (Xu and Xu, 2020). The supervision of companies’ establishment of ASMM may need to be delegated to local regulatory bureaus. Some local regulators of the CSRC (e.g., the Shenzhen Bureau) have issued announcements clearly indicating that they will strictly review and punish companies that fail to establish an ASMM,7 but more regional regulators do not require this. As a result, about 30.4% of companies still have not established an ASMM (as shown in Table 1). Our study uses companies with an ASMM in 2010–2021 as the treatment group and companies without an ASMM as the control group.

Hypothesis Development

A weak financial reporting environment encourages executive misstatement (Jongjaroenkamol and Laux, 2017) and suboptimal decisions not consistent with firm value (Cheng et al., 2013), causing accounting earnings to deviate in a manner that truly reflects executive effort. We argue that the ASMM limits executives’ opportunistic behaviour and promotes the board's use of earnings-based compensation contracts. Compensation contracts based on accounting earnings are widely used in the Chinese capital market. Nevertheless, a lack of effective monitoring in the Chinese market can motivate executives to act opportunistically, leading to repeated misstatements and strong incentives to conceal information. These motivations, which are complex, may include meeting earnings expectations (Barua et al., 2010), maximizing short-term bonuses (Guidry et al., 1999), appropriation (Nagar and Schoenfeld, 2021), empire building (Hope and Thomas, 2008), insider trading (Frankel and Xu, 2004), tax avoidance (Kim et al., 2011), and external funding (Dordzhieva et al., 2022). Consequently, accounting information can distort agency conflicts between shareholders and executives. Information disclosure plays a significant role in the board's supervision of executives (Franco et al., 2013), but the board's access to internal information is limited. This impedes its oversight of executive behaviour and leads to the failure of compensation contracts that are based on accounting earnings.

The ASMM, with its broader scope of accountability, more comprehensive accountability targets, and more severe accountability measures, closely links executives’ personal wealth and professional reputation with future development in terms of the quality of the information they provide. This can increase the costs of misstatement, encouraging rational executives not only to reduce any incentives to misstate but also to improve the accounting information system, thereby enhancing the quality of accounting information. When the accuracy of reported earnings improves, its weight in compensation contracts can increase (Baiman and Verrecchia, 1995).

Moreover, when the ASMM strengthens accountability penalties for executive misstatements and limits executives’ use of private information to obtain excessive returns, it encourages executives to abandon their hold on internal information, reducing information asymmetry between shareholders and management and strengthening the board's regulatory function (Franco et al., 2013). Accordingly, if the ASMM's governance of information disclosure is conducive to limiting the distortion of accounting earnings, this will improve accounting earnings’ accurate reflection of executive efforts and strengthen the board's oversight of executive conduct. Consequently, it will facilitate the use of accounting earnings in compensation contracts, as reflected in the increased sensitivity of executive compensation to accounting earnings. Therefore, Hypothesis 1a is formulated as follows.

However, the restriction of ASMM on corporate misreporting may also reduce executive pay-performance sensitivity, as empirical evidence shows that executives manipulate earnings for high-performance compensation (Guidry et al., 1999; Cornett et al., 2008). Not only that, the ASMM may intensify risk-aversion behaviour among executives, prompting the board to reduce the weight of accounting earnings in executive compensation contracts.

Excessively high-powered incentive compensation will cause aggressive misreporting (Marinovic and Povel, 2017; Jongjaroenkamol and Laux, 2017). Several studies provide evidence that bonus payments provide executive incentives to meet earnings benchmarks (Matsunaga and Park, 2001; Dechow et al., 2010). To achieve the potential earnings benchmark for bonus payments, executives are motivated to manage earnings (Guidry et al., 1999) and reduce the board's ability to detect and dismiss poor-performing executives (Jongjaroenkamol and Laux, 2017). Not only that, boards were also found to conspire with executives to report more earnings. Dechow et al. (2010) show that managers’ report securitization gains instead of accounting earnings when earnings are low or negative, in order to report larger gains, and that more independent or informed boards do not intervene in the CEO's high compensation sensitivity to reported securitization gains. Moreover, Chen and Lu (2011) find in the context of China that companies manage earnings to enhance compensation, and earnings-based incentive compensation leads to more income-increasing earnings management. We argue that if executives’ ability to manage reported earnings is restricted, high compensation sensitivity to reported earnings, usually overstated for more bonuses, will reduce.

More specifically, the ASMM may restrict executive manipulation of discretionary accruals to overstate reported earnings when executives are incentivized by bonus-based compensation. This will reduce their compensation sensitivity to reported earnings. Accrual-based earnings management, which is the most common tool to manage earnings, often contradicts the Accounting Standards for Business Enterprises (ASBE) (Ho et al., 2015) and is susceptible to producing misstatements (Dechow et al., 1996), and thus is more likely to be penalized by the ASMM. While boards may not sufficiently implement the ASMM because they are aware of and tolerating executives’ earnings management for more reported earnings (Dechow et al., 2010), the ASMM may also play a governance role as external stakeholders will monitor the execution of the system. The CSRC requires companies to disclose their establishment of ASMM publicly, and some local supervision bureaus (e.g., the Shenzhen Bureau) of the CSRS also include the implementation of ASMM in their routine inspections of listed companies. The ASMM implemented by companies suffers from public supervision. Although securities regulation is not efficiently enforced in China (Song and Ji, 2012), stakeholders and potential investors of companies also monitor the execution of the ASMM and will consider any misstatement they uncover when making investment decisions. Hence, companies establishing an ASMM do so with increased scrutiny, which means executives have to give up their manipulation of discretionary accruals. This will reduce executives’ compensation sensitivity to reported earnings if they previously were managing earnings for high bonus-based compensation.

In addition, executives sacrifice relatively valuable long-term investments to meet short-term performance goals (Graham et al., 2005; Chen, Cheng, Lo, and Wang, 2015; Kraft et al., 2018). Even in the absence of earnings manipulation, widespread risk aversion and moral hazard issues among executives can still lead to a biased reflection of accounting earnings in a manner that obscures executive efforts. Given a strong incentive to circumvent oversight for self-interested purposes, executives whose opportunistic behaviours are limited by the ASMM may engage in alternative undesirable behaviours. Specifically, executives are found to shift to a more myopic investment strategy to preserve earning-based compensation when accrual-based earnings management is dissuaded (Biddle et al., 2024). The ASMM restricts companies from hiding negative information and increases executive pressure to maximize short-term performance, which may lead to executives’ increased risk aversion and moral hazard. As a result, boards reduce the weighting of accounting earnings in compensation contracts (Peek and Moers, 2001). This is because positive net present value investments that enhance firm value are often high-risk projects that involve expensing upfront investments and long payback periods, resulting in poor short-term accounting earnings. High risk also implies uncertainty about high returns and the risk that firms will invest unsuccessfully and experience financial distress (Holmstrom, 1989; Tian and Wang, 2014), along with the possibility that executives will suffer reputational damage and mandatory departures because of their failed decisions (Lehn and Zhao, 2006). Furthermore, the ASMM may increase the likelihood that external information users will view stringent governance of misstatements as a signal of high quality. If establishing the ASMM attracts attention from investors and increases the consideration of accounting earnings in investment decisions, it may also increase the pressure on executives for short-term performance and prompt short-termism.

To avoid an earnings decrease and improve short-term performance, executives may reduce their exposure to high-risk investments and choose suboptimal investments (Chen, Cheng, Lo, and Wang, 2015). To reduce executives’ risk aversion and myopic behaviour, the board will reduce the weight of accounting earnings in compensation contracts, manifested in executive compensation's reduced performance sensitivity. Based on the above analysis, we propose Hypothesis H1b.

Sample

China's accounting standards, implemented in 2007, have converged with the International Financial Reporting Standards (Zhang, 2022). Considering the impact of the differences in calculation standards for financial indicators, our sample begins in 2007. Our sample selection is described in detail in Table A2 of the Appendix. First, to explore the impact of ASMM establishment on executive compensation contracts, we select a sample from 2007 to 2021 (44,303). Second, we eliminate companies listed on the B-share market (1,549), companies from financial industries (943), and observations with missing financial data (5,826). As in Nienhaus (2022), firms without at least one observation before and after the year of ASMM establishment are eliminated to improve the validity of the DiD model (9,954). We eventually obtain 26,031 valid firm-year observations. Among them, 18,120 observations, representing 1,359 listed companies, are in the treatment group. The observations for the control group are 7,911, representing 1,604 listed companies. The data relating to ASMM establishment are manually extracted from the Cninfo website,8 and other data are from the China Stock Market & Accounting Research (CSMAR) Database. In addition, to exclude the effect of outliers, all continuous variables are winsorized at the 1% and 99% percentiles, and all regressions are clustered at the firm level for standard errors.

Model Specification

The dependent variables, TPAY1 and TPAY2, refer to the logarithm of executives’ annual compensation and the logarithm of the annual compensation of the top three executives, respectively.10 Referring to the method of Chan et al. (2012), TREAT equals one if the company is an ASMM adopter (treatment group) and equals zero if it is instead a non-adopter (control group). POST is an indicator variable that identifies the pre- and post-adoption periods, which equals one for a firm year in which the ASMM adopters have ASMM provisions in place, and zero otherwise. To identify the sensitivity of executive compensation to accounting earnings, we construct the variable return on assets (ROA), defined as earnings before interest, taxes, depreciation, and amortization (EBITDA) divided by total assets. The coefficient α₁ of TREAT×POST×ROA is the critical indicator of interest in this paper, measuring how and to what extent ASMM affects the sensitivity of executive compensation to accounting earnings.

Further, referring to Ke et al. (1999) and Rhodes (2016), we control for a range of corporate finance and governance characteristics, including firm size (SIZE), leverage (LEV), sales growth (GROWTH), market-to-book ratio (MTB), volatility of return on assets (varROA), R&D investment (RND), nature of property rights (SOE), years of CEO tenure (CEOTENURE), duality of chairman and CEO (DUAL), board size (BOARD), the proportion of independent directors (INDE), institutional shareholdings (INSSHR), and whether the auditors are from Big 4 accounting firms (BIG4). Following Banker et al. (2009) and Rhodes (2016), we control for the interaction term between the above control variables and ROA to capture cross-sectional differences in executive pay-performance sensitivity. The variables are defined in Table A1 of the Appendix. There may be unique characteristics of firms and years that are not time-varying and unobservable, leading to endogeneity issues in the regressions. For this reason, we control for firm and year-fixed effects.

Table 1 shows the distribution of the sample by year. Since 2010, companies’ establishment of the ASMM has been staggered, facilitating our construction of a DiD model and alleviating the endogeneity problem. Although a considerable proportion of enterprises has not established the ASMM, the annual distribution shows that the sample of the treatment group in each year accounts for 69.60%, indicating that most firms have complied with the requirements of the CSRC and declared their intention to establish the ASMM.

Descriptive Statistics of Main Variables

Table 2 presents the descriptive statistics of the variables in Model (1). Panel A shows the descriptive statistics of the full sample. The mean of the logarithm of the compensation and the logarithm of the top three executive compensation are 16.86 and 16.30, respectively, with standard deviations of 2.603 and 2.877, showing a large gap in executive compensation between different firms. The ASMM has been established by 69.6% of the firms, which are classified as the treatment group, whereas 55% of the observations are from the year the ASMM was established or in subsequent years. In addition, the descriptive statistics of other variables are consistent with recent studies focused on the Chinese capital market (Yao et al., 2020; Chen et al., 2020).

Panel B shows the descriptive statistics between the treatment and control groups. The mean value of the logarithm of executive compensation (TPAY1) for the treatment group is 16.30, compared to 18.13 for the control group. This indicates that the mean values of the logarithm of executive compensation are lower for the treatment group than for the control group. When one observes the medians and the compensation of the top three executives, the conclusion is also consistent. In terms of the mean and median values of the control variables, all of the variables differ significantly between the treatment and control groups. Accordingly, it is necessary to put these control variables into the model. We also mitigate this problem later using propensity score matching, the details of which are described in a later section.

The Baseline Results

The results for H1a and H1b are presented in Table 3, Panel A. Columns (1) and (2) show the relationship between accounting earnings and executive compensation. When using TPAY1 and TPAY2 as the dependent variables and ROA as the independent variable, the coefficients of ROA are 1.183 and 1.207 at the 1% significance level. This indicates that accounting earnings are a commonly used performance indicator for listed companies in China. Columns (3) and (4) show the impact of the ASMM on executive total compensation. The coefficients of TREAT×POST show that executive compensation significantly decreases after firms establish an ASMM.

Columns (5)–(8) of Table 3 Panel A show the results of the tests of H1a and H1b. In columns (5) and (6), excluding the effect of the interaction term between ROA and the control variables, the coefficients of ROA are 1.733 and 1.785 and significant at the 1% level. This indicates that the accounting earnings of firms without the ASMM are significantly and positively related to executive compensation. The coefficients of TREAT×POST×ROA are –0.936 and –0.984, respectively, and significant at the 1% level. Further, the results of Model (1) are presented in columns (7) and (8). Regardless of whether TPAY1 or TPAY2 is the dependent variable, the coefficient of ROA is positive but insignificant,11 and the coefficients of TREAT×POST×ROA are –0.831 and –0.849, respectively, at the 1% level. Consequently, the sensitivity of executive compensation to accounting earnings significantly decreases after establishing the ASMM, and thus H1b is supported.

Omitted Variable Issue

Referring to Donohoe et al. (2022), we use Oster's (2019) approach to explore the impact of omitted factors in Model (1). The examination is divided into two parts: whether there are any omitted variables equivalent in importance to the observed variables in Model (1) that have an impact on the regression results, and how many times the impact of the unobserved factors needs to be greater than that of the observed variable to overturn the findings in the baseline results. First, it assumes that the R² of the main test becomes 1.3 times its original value after introducing an omitted factor of equal importance (δ = 1) to the observed factor (TREAT×POST×ROA), whereby the interval of the true α₁ of the main test is restored. Suppose the interval of the restored true α₁ is within the 99.5% confidence interval of the coefficient of TREAT×POST×ROA in Model (1) and does not contain the value of zero. In that case, the coefficient of TREAT×POST×ROA in Model (1) is considered stable. It is unlikely that there are unobserved factors as important as the observed variables.

Second, assuming that the coefficient of TREAT×POST×ROA in Model (1) becomes zero or significantly positive after the inclusion of the omitted factor, the importance (δ) of the omitted factor relative to the observed factor (TREAT×POST×ROA) is calculated. Oster (2019) argues that when the absolute value of δ exceeds one, it is unlikely that there are unobserved factors that significantly affect the main test. The omitted variable tests for Model (1) are presented in Panel B. In columns (1) and (3), regardless of whether TPAY1 or TPAY2 are taken as the dependent variable, the interval of true α₁ does not contain zero and is within the 99.5% confidence interval of the coefficient of TREAT×POST×ROA in Model (1). The absolute value of δ is 221.25 and 24.61, respectively, much greater than 1. Given the evidence above, we believe that omitted variables are unlikely to be a concern in our model.

Parallel Trend Assumption Test

The DiD model can only be valid when the assumption of parallel trends is satisfied. If the trends of the dependent variables were systematically different between the treatment and control groups, our conclusions would be challenged by incorrect research design. Therefore, we construct the dummy variables PRE_t (t = 1, 2, 3), CURRENT, and POST_t (t = 1, 2, 3, 4) to identify the changes in executive compensation in each year before and after the point of establishment of the system to test the parallel trend assumption.

The variable PRE_t (t = 1, 2, 3) is equal to one if the firm-year is in the third year and before, the second year, and the first year before the treatment group established the ASMM, and zero otherwise. In contrast, CURRENT and POST_t (t = 1, 2, 3, 4) are equal to one when the firm-year is in the current year, the first year, the second year, the third year, the fourth, and the subsequent years when the treatment group establishes the ASMM, and zero otherwise. Using PRE_3 as the benchmark and replacing POST in Model (1) with PRE_t (t = 1, 2), CURRENT, and POST_t (t = 1, 2, 3, 4), we construct a Model (2). The control variables are the same as in Model (1).

The regression results are presented in Table 3, Panel C, and Figure 2, respectively. The coefficients of TREAT×PRE_t × ROA are not significant, indicating no significant difference in the pay-performance sensitivity of executives between the treatment and the control groups before establishing the ASMM. Although the coefficients of TREAT×CURRENT×ROA, TREAT×POST_1 × ROA, and TREAT×POST_2 × ROA are not significant, the coefficients of TREAT×POST_3 × ROA and TREAT×POST_4 × ROA are significantly negative. This indicates that executive pay-performance sensitivity declines significantly after three years since the establishment of the ASMM.

Propensity Score Matching

Significant differences in firm characteristics may exist between the treatment and the control groups, which could affect the validity of the regression results. Using propensity score matching to screen the sample, this problem can be mitigated. Referring to the method of Kubick et al. (2020), we use the treatment group in the year prior to the establishment of the ASMM (1,307 observations) and all the control group firms (7,910 observations) as matched subjects. We use firm size (SIZE), leverage (LEV), return on assets (ROA), sales growth (GROWTH), market-to-book ratio (MTB), whether the firm is loss-making (LOSS), fixed asset ratio (TANGIBLES), R&D investment (RND), agency cost (MFEE), firm age (AGE), nature of property rights (SOE), board size (BOARD), the proportion of independent directors (INDE), institutional shareholdings (INSSHR), CEO tenure (CEOTENURE), and Big-4 accounting firms (BIG4) as matching indicators (Bao et al. 2018), with the threshold of 0.005 for a 1 to 1 match without putback in the same year. As a result, 431 firms in the control group and 431 firms in the treatment group are successfully matched, representing 3,803 and 5,536 observations, respectively. Based on the matched observations, we retest Model (1) with the matching sample. The results are presented in Table 5.

Panel A reports the results of the test. Columns (1)–(3) show the differences in the matching indicators between the treatment and the control groups before matching, with a sample of the treatment group in the year prior to the establishing year and the entire control group. Columns (4)–(6) show the differences in firm characteristics between the groups after matching. According to columns (3) and (6), most of the firm characteristic indicators are significantly different before matching, whereas the difference disappears after matching, which indicates that our matching is effective. Panel B shows the regression result of Model (1) using the matching sample. The coefficients of TREAT×POST×ROA are all significantly negative, which means that the results of the baseline test hold.

Placebo Test

Although the ASMM established by Chinese listed companies demonstrates staggered establishment characteristics, which greatly alleviates our concerns about omitted factors contributing to the findings, we use a placebo test to investigate this issue further. Placebo tests are conducted by randomly generating treatment groups.

Referring to Zhang et al. (2022), we use the treatment group sample for the year prior to the establishment of the system and all the control group samples to construct the placebo test of a randomly selected treatment group. We randomly select 1,500 firms as the pseudo-treatment group and the remaining firms as the pseudo-control group. Based on the pseudo-treatment and pseudo-control groups, we generate the pseudo-dummy variable TREAT×POST and the key variable TREAT×POST×ROA. After repeating the random sampling 500 times, we plot the distribution of the estimated coefficients and p-values of pseudo-TREAT×POST×ROA from the 500 regressions, and the results are presented in Figure  3. The coefficients of pseudo-TREAT×POST×ROA are mainly distributed around zero, and the p-values are also mostly above the 10% significance level, implying that our conclusions are relatively unlikely to be driven by other policies or unobserved factors.

Heckman Test

We use the two-stage Heckman treatment effects model to mitigate the self-selection problem in establishing the ASMM. We construct the selection model for the ASMM in the first stage, as shown in Model (7), using the establishment ratio of the ASMM in each industry among years (IV) as the instrumental variable, and the Inverse Mills Ratio (IMR) is calculated accordingly. The control variables are the same as in Model (1), with industry and year-fixed effects. After adding the IMR to Model (1), we construct Model (8), and the results are presented in Table 6. Column (1) shows that the coefficient of IV is significantly positive, indicating that IV is positively correlated with TREAT. After adding the IMR into Model (1), the results in columns (2) and (3) show that the results of the main test hold.

Concerns about the Nature of Property Rights

After we add SOE and its interaction terms with TREAT×POST×ROA, TREAT×POST and ROA to Model (1), Model (9) is constructed to identify the moderating effect of the nature of property rights on the relationship between ASMM and executive pay-performance sensitivity. The results are presented in columns (1) and (2) of Table 7. The coefficients of SOE×TREAT×POST×ROA are not significant, suggesting that the reduction effect of the ASMM on the performance sensitivity of executive compensation is common in state-owned and non-state-owned enterprises.

Concerns about Other Confounding Events

Given the extensive time span covered by our sample, it is possible that significant events occurring during the period could have influenced the findings presented. Although we have employed placebo tests to address this concern, we acknowledge the need to explore notable events that may potentially impact the conclusions.

Firstly, the General Office of the State Council of China issued a directive in 2016 mandating state-owned enterprises to establish a system for holding accountable those responsible for violations and misconduct in business operations and investments (the SOE clawback system). This system entails imposing various measures on culpable individuals, such as organizational actions, salary deductions, entry restrictions, disciplinary actions, or referring them to judicial authorities. Individuals in charge of major decisions that contravene regulations during their tenure are subject to lifetime accountability. This policy enhances the responsibility of executives in state-owned enterprises regarding their daily business operations and investments. More importantly, the SOE clawback system requires executives to be held accountable for misreports, as ASMM specifically focuses on disclosure responsibilities. The potential concern is that the reduced executive compensation sensitivity to performance is driven by the enhanced information responsibilities of the SOE clawback system rather than the ASMM. As we find no differences in the impact of the ASMM on executives’ compensation sensitivity to accounting earnings between SOEs and non-SOEs, we exclude this concern. We are still concerned about whether the SOE clawback system will affect our main findings.

Finally, it has been found that firms engage in earnings management prior to undertaking seasoned equity offerings (Rangan, 1998; Cohen and Zarowin, 2010). Notably, the CSRC has established earnings thresholds for listed companies conducting equity issuances in the Measures for the Administration of Securities Issuance by Listed Companies. According to this regulation, the average return on weighted average net assets for the most recent three accounting years should not be lower than 6%. This requirement may create a stronger incentive for firms to manipulate earnings in order to meet the threshold. The earnings management related to seasoned equity offerings may potentially affect our results. From the distribution of return on equity (ROE) for the treatment group and control group, as depicted in Figure 4, it can be observed that the peak of the ROE distribution for both groups occurs around 6%. This suggests that firms may engage in earnings management for seasoned equity offerings, but there is no noticeable difference in this behaviour between the treatment and control groups.

Further, we eliminate the observations that may be affected by the earnings management incentive for seasoned equity offerings. Companies may manage earnings for three years prior to the seasoned equity offering issuance, as the earnings threshold is calculated based on three years. Moreover, accrual, as the most common method of earnings management based on assumptions and estimates, needs to be corrected in future accruals and earnings (Dechow and Dichev, 2002). Therefore, we eliminate companies’ observations three years before or after each seasoned equity offering issuance and retest Model (1). The results are displayed in columns (7) and (8) of Table 7. The findings indicate that the mitigating effect of ASMM on the sensitivity of executive compensation to performance exists.

Re-measuring Key Variables

We re-measure the independent and dependent variables and then re-examine them. First, we use the balance of net profit after deducting non-recurring gains and losses divided by net assets (ROE) (Banker et al., 2013), economic value added (EVA) (Biddle et al., 1997), and industry-adjusted ROA (ADJ_ROA) (Cadman et al., 2010) to measure firm performance and replace ROA with ROE, EVA, and ADJ_ROA, respectively, then modify and retest Model (1). The regression results in columns (1)–(6) of Table 8 show that the main results hold. It also suggests that our main findings hold.

In addition, we explore how the ASMM affects the sensitivity of executive cash compensation to accounting earnings. We find a significant reduction in the sensitivity of executive cash compensation to accounting earnings after adopting an ASMM when we use the logarithm of total salary and bonus (CPAY1) and the logarithm of total salary and bonus of the top three executives to measure cash compensation (CPAY2) (Rhodes, 2016). As shown in columns (7)–(8) of Table 8, our main findings remain.

Replacing the Regression Model

Agency Cost

Audit Quality

We conduct Model (15) to test whether Big 4 auditors affect ASMM's reduction effect on executives’ pay-performance sensitivity, and the results are presented in columns (3) and (4) of Table 9. The ASMM's reduced effect on the performance sensitivity of executive compensation is significantly higher in companies audited by non-Big 4 firms, which is consistent with our inference.

Analyst Coverage

We construct Model (16) to identify how analyst coverage influences the reduction effect of ASMM on executive pay-performance sensitivity. The indicator ANA refers to the logarithm of one plus the number of analysts following the firm in the current year. From columns (5) and (6) in Table 9, it is evident that the coefficients for ANA×TREAT×POST×ROA are both significantly positive. This indicates that the reduction effect of ASMM on executive pay-performance sensitivity is stronger in companies with less analyst coverage.

The ASMM and the Shifting of Performance Indicators and Compensation Manners in Executive Compensation Contracts

Accounting and market performance are two measures of both firm performance and management effort (Baiman and Verrecchia, 1995; Davila and Penalva, 2006). Agency models based on multiple performance indicators suggest that optimal compensation contracts should rely on performance indicators that are precise and sensitive to the agent's efforts (Engel et al., 2003). We argue that the ASMM may restrict executives’ manipulation of earnings but prompt them to risk aversion, distorting accounting-based performance indicators in reflecting their effort. Thus, boards reduce the weight of accounting earnings in compensation contracts.

To test the speculation set forth above, we construct the variable stock returns (RET) and add it to Model (1), along with the interaction term of RET and TREAT×POST, TREAT×POST×RET, to investigate the impact of the ASMM on the sensitivity of executive compensation to stock returns. Column (1) of Table 10 reports the results, which show that the coefficient of TREAT×POST×ROA remains significantly negative and the coefficient of TREAT×POST×RET is insignificant. This indicates that the ASMM reduces the sensitivity of executive compensation to accounting earnings but does not increase executive compensation's sensitivity to stock returns.

We further split executives’ total compensation (TPAY) into cash compensation (CPAY) and equity compensation (SPAY). After we modify the dependent variable of Model (17) with CPAY and SPAY, respectively, the results are listed in columns (2) and (3) of Table 10. We find that the ASMM reduces the sensitivity of executive cash compensation to accounting earnings and increases the sensitivity of executive cash compensation to stock returns. Although we do not find the reduced sensitivity of executive equity compensation to accounting earnings after establishing the ASMM, the sensitivity of executive equity compensation to stock returns increased. This is consistent with the view that it is necessary to use multiple performance indicators to assess executive effort and to dynamically adjust incentive-compatible compensation contracts (Harris and Holmstrom, 1982; Tirole and Laffont, 1988; He, 2011).

The ASMM and the Shifting of Earnings Management Methods

We constructed Model (18) to test the relationship between the ASMM and real earnings management,16 and the result is shown in column (4) of Table 10. It is found that real earnings management does not reflect an improvement after establishing an ASMM, which shows that our main findings may not be driven by increased real earnings management.

Moreover, even if real earnings management increased, it is still not the reason executives’ pay-performance sensitivity to accounting earnings decreased. The reason is that the potentially competing explanation argues that executives will transform the earnings management behaviour from accrual-based to real-based, which distorts the accurate reflection of accounting earnings to executives’ efforts. The boards reduce the weight of accounting earnings in bonus contracts. However, it is difficult to explain why the compensation committee on the board, which cannot identify or restrict executives’ accrual earnings management, can effectively recognize and forbid the more hidden real earnings management. For the reasons above, the alternative explanation above is ruled out.