State Ownership and Earnings Management around Initial Public. Offerings: Evidence from China

Transcription

1 State Ownership and Earnings Management around Initial Public Offerings: Evidence from China C.S. Agnes Cheng The Hong Kong Polytechnic University Jing Wang Queen's University Steven X. Wei The Hong Kong Polytechnic University This Version: May 2014 We are grateful for comments from Donal Byard, Jun Cai, Peter Chen, Yongtae Kim, James Ohlson, Nancy Su, Wilson Tong, Steven Wang, the workshop participants at The Hong Kong Polytechnic University, Nagoya University, the 2012 American Accounting Association Annual Meeting, the 2013 Financial Accounting and Reporting Section Midyear Meeting, and two anonymous referees of the 2014 Journal of International Accounting Research conference. We acknowledge the financial support from The Hong Kong Polytechnic University Research Grant G-U179. All remaining errors are ours. 1

2 State Ownership and Earnings Management around Initial Public Offerings: Evidence from China Abstract This study investigates earnings management by firms engaging in initial public offerings (IPOs) in domestic Chinese equity markets. Using a sample of 437 IPO firms, we find that Chinese firms tend to inflate earnings around their IPOs. We also show that state-owned enterprises (SOEs) manage earnings to a lesser degree than non-state-owned enterprises (NSOEs) around IPOs. Furthermore, using the path analysis, we find that two incentive factors, CEO shareholding and accessibility to bank loans, explain 48% of the correlation between state ownership and earnings management for IPOs. In particular, the accessibility to bank loans is a more important incentive factor that leads to less earnings management for SOEs than NSOEs. 2

3 1. Introduction Do firms manage earnings around the time of initial public offerings (IPOs)? In early studies for the US market, Teoh, Welch and Wong (1998) and Teoh, Wong and Rao (1998a) find that IPO firms engage in earnings management around the issue year. This view has been recently revised by Ball and Shivakumar (2008) and Armstrong, Foster and Taylor (2009), who claim that issuing firms do not inflate earnings around IPOs in the UK and US markets. Ball and Shivakumar (2008) argue that this is due to the strong monitoring offered by internal and external parties (e.g. auditors, boards, analysts, rating agencies and the press) who scrutinize and penalize earnings manipulators around IPOs in the markets. 1 However, the strong monitoring cannot be universally assumed, in particular, in emerging markets. This paper examines the same issue in China s market, the largest stock market among all emerging economies and the second largest equity market in the world. 2 The strong monitoring argument for the UK or the US shall not hold in China s market. The literature provides strong empirical evidence that monitoring is weak in China. For example, Allen, Qian and Qian (2005) find that China s law and institutions are underdeveloped; in particular, the standard corporate governance mechanisms are weak, law enforcement is ineffective due to the lack of an independent and efficient judicial system, and the lacking of independent and professional auditors. As pointed out by Piotroski and Wong (2011), China s financial markets continue to be plagued by weak information systems, resulting in low quality financial reporting, which substantially increase analysts earnings forecasting errors. In addition, the media and internet are highly regulated in China. The strong monitoring, as argued by Ball and Shivakumar (2008) in the UK market, does not 1 Ball and Shivakumar (2008) re-examine the findings of Teoh et al. (1998) using the Jones model (Jones, 1991) and a piecewise linear variant suggested by Ball and Shivakumar (2006). They find that the sample firms in Teoh et al. (1998), on average, do not inflate earnings opportunistically for the years prior to their IPOs. They attribute the findings of Teoh et al. (1998) managers opportunistically inflate earnings to influence IPO pricing to unreliable estimates of discretionary accruals. Teoh et al. s (1998) estimates of discretionary accruals contain substantial endogenous effects of the IPOs. 2 By the end of 2009, the total market capitalization of Shenzhen and Shanghai Stock Exchanges has exceeded 3573 billion US dollars, surpassing that of Tokyo Stock Exchange (3306 billion US dollars). China s stock market became the second largest after the US s. The data are taken from 3

4 apply to China s market. Therefore, we predict that Chinese issuing firms engage in earnings management around IPOs. Using a sample of 437 Chinese IPO firms listed in the domestic markets over the period from 2003 to 2009, 3 we find that Chinese IPO firms do manage earnings opportunistically. Following Armstrong et al. (2009), we estimate discretionary accruals of the IPO firms using the modified cross-sectional Jones model (Dechow, Sloan and Sweeney, 1995). Both the mean and median of discretionary accruals of the IPO firms are significantly positive surrounding the issuing years. Our results are consistent with the suggestion that weak monitoring and scrutiny by Chinese equity markets result in earnings management of IPO firms. Next, we examine whether the degrees of earnings management for IPO firms differ between state-own enterprises (SOEs) and non-state-own enterprises (NSOEs). 4 Chinese firms are generally classified into two types: SOEs and NSOEs. Prior to year 2000, since China s stock market was re-established in early 1990s for solving the financing problem of SOEs, the IPOs were overwhelmingly dominated by SOEs. With the rapid economic growth in China, more and more NSOEs become listed. The majority of our sample from 2003 to 2009 is NSOE (260 IPOs for NSOEs and 177 IPOs for SOEs). Prior literature suggests that the incentives for, and monitoring of, earnings management are different between the two types of firms (Aharony, Lee and Wong, 2000; Liu and Lu, 2007; Yang, Chi and Young, 2011; Fan, Wong and Zhang, 2007; Chen, Chen, Lobo and Wang, 2011). The increasing trend of NSOE IPOs in China suggests the importance to examine if NSOEs differ from SOEs in manage earnings around IPOs. The degrees of incentives for earnings management vary between SOEs and NSOEs around IPOs for three reasons. First, for SOEs, top managers who own few shares and no 3 We start our sample period from 2003 because most of the IPO firms have insufficient data to estimate discretionary accruals in the issue year before SOE firms are defined as those firms whose ultimate owners are state asset management bureaus, the central government or a local government. NSOEs are defined as those firms whose ultimate owners are individuals or private investors. 4

5 stock options will not directly or monetarily gain from high IPO prices. This means that they have no direct incentives to manage earnings. However, they have indirect incentives to do so since the listing status of SOEs may bring them higher prestige and other non-pecuniary benefits, such as political promotion. However, for NSOEs, top managers are usually the founders of IPO firms and have direct and substantial monetary gains from high IPO prices. Thus, they have direct incentives for managing earnings opportunistically. Second, SOEs have favorable access to the equity market (see Section 2), which enhances their chances of going public, so they have weaker incentives for managing earnings around IPOs than NSOEs do. Third, since the state-owned banks dominate the banking industry and they prefer to grant loans to SOEs than NSOEs (Brandt and Li, 2003), SOEs have favorable access to bank loans and are less likely to have financial constraints. Moreover, non-state-owned banks also put their lending priority to SOEs. As a result, SOEs have weaker incentives for managing earnings around IPOs than NSOEs do. Ceteris paribus, the weaker incentives of SOEs should lead to less earnings management in comparsion with NSOEs. On the other hand, the degrees of monitoring of earnings management around IPOs may also vary between SOEs and NSOEs. The existing empirical evidence shows that the monitoring is weaker for SOEs: governance characteristics are weaker and less professional (Fan et al., 2007); SOEs prefer to hire small local auditors (Wang, Wong and Xia, 2008); state-owned banks grant loans imprudently to SOEs (Chen et al., 2010); the role of underwriters in restricting pre-ipo earnings management is limited (Chen, Shi and Xu, 2013). The weaker monitoring of SOEs predicts more earnings management for SOEs than NSOEs. Since SOEs have less incentive to manage earnings but weaker monitoring to reduce earnings management than NSOEs, it is an empirical question whether the two types of firms, SOEs and NSOEs, manage earnings differently. Our empirical results demonstrate that the degrees of earnings management for IPO firms differ between SOEs and NSOEs. Specifically, SOEs manage earnings to a lesser degree than NSOEs around IPOs, even after controlling for firm characteristics. This seems to 5

6 suggest that the weaker incentives for SOEs play a stronger role in earnings management around IPOs than the weaker monitoring. Next, we further investigate how incentives CEO shareholding and accessibility to bank loans contribute to our findings. We use path analysis to evaluate the relative importance (mediating effects) of the two variables, CEO shareholding and accessibility to bank loans, to the relationship between SOE and DA (discretionary accruals). 5 Our result shows that the two mediating variables explain 48.1% of the correlation between state ownership and earnings management in the IPO years. The accessibility to bank loans is the more important mediator, accounting for 26.5% of the correlation. The rest, 51.9% of the correlation, should be due to the direct link between state ownership and earnings management. This study makes a significant contribution to the literature at least in three ways. First, unlike the recent findings for the UK or US IPO firms, we find that Chinese IPO firms do manage earnings. Ball and Shivakumar (2008) argue that strong monitoring prohibits IPO firms from managing earnings in the UK. This implies that IPO firms under weak monitoring are likely to inflate earnings upwards. Our finding that the weak monitoring in China (Piotroski and Wong, 2011) cannot prevent earnings management for IPO firms confirms the strong monitoring argument. As a policy implication, our finding calls for strengthened regulation to reduce or prohibit earnings management for IPO firms. Second, this paper complements the early study by Aharony et al. (2000) in a few aspects: (1) while they consider the Chinese IPO firms listed in foreign markets, we focus on those listed in the domestic markets. Chinese domestic equity markets have played an increasingly important role in China as well as in the global economy; (2) we examine earnings management around IPOs for both SOEs and NSOEs, but they only study SOEs 5 Path analysis is a statistical tool commonly used to describe the casual dependencies, and examine the relative strength of direct and indirect relationships among variables (e.g. between the dependent and independent variables). In specific, path analysis allows researchers to decompose the various factors that affect an outcome into direct and indirect effects. It has been widely used in auditing and managerial accounting research. In the literature of capital market research, Bhattacharya, Ecker, Olsson and Schipper (2012) use path analysis to investigate the direct and indirect links between earnings quality and the cost of equity. In this paper, we use path analysis as complementary to our regression analysis since there are more than one incentive factors that lead SOEs to engaging in earnings management around IPOs, and we want to know the relative importance of them. 6

7 since the IPO firms in their sample period were overwhelmingly dominated by SOEs; (3) in terms of methodology, we adopt several accruals expectation models (e.g. the cross-sectional Modified Jones model) to estimate the discretionary accruals in order to provide a more powerful test of earnings management (Dechow et al., 1995). Last, we introduce the path analysis to anatomize the mediating roles played by the two incentive factors in the relationship between state ownership and earnings management around IPOs. To our knowledge, this is the first paper to identify accessibility to bank loans as an important mediator leading to a lesser degree of earnings management for SOEs than for NSOEs around IPOs. The rest of this paper proceeds as follows. Section 2 describes the institutional background in China. Section 3 develops three hypotheses on earnings management around IPOs. Section 4 describes data and sample, and discusses our empirical research design. We report and analyze our empirical results in Section 5. After additional and robust analyses in Section 6, Section 7 concludes the paper. 2. Institutional background In the early 1990s, the Chinese government reestablished the stock market to facilitate the financial system and solve the financing problems for SOEs. Over the last two decades, the number of listed firms (including A-shares and B-shares) has reached 1754, and the total market value of publicly traded shares has exceeded 3573 billion US dollars by the end of The Chinese stock market has become an increasingly important financing source for Chinese firms. In the early stage of Chinese stock market, some policies associated with China s stock issuance procedure are expedient. For instance, prior to 2001, without approval by a local government or the central government, a sponsor was not permitted to apply for an 6 Data are collected from 7

8 IPO. 7 To improve the successful rate of IPOs, the China Securities Regulatory Commission (CSRC) regulated the IPO pricing methods. The pricing methods over this period were dominated by fixed price and relative fixed price-earnings (P/E) ratio. 8 Since 1996, the CSRC has implemented a series of reforms to establish a market-oriented IPO procedure. In 2001, a firm did not need to obtain approval from a local government or the central authorities before it applied for an IPO. Since 2005, the accumulated bidding inquiry pricing method has been used to replace the controllable P/E ratio pricing method. 9 Yet, government intervention has still been an important feature of the Chinese stock market. Characteristic of the Chinese capital market is that historically more than half of the listed firms are controlled by the government. Out of the 1,754 firms listed in Shanghai and Shenzhen Stock Exchanges, 964 were SOEs by the end of On the one hand, the Chinese government, as the largest shareholder of SOEs, has adopted a number of policies to aid these SOEs when they are in trouble. On the other hand, chief executive officers (CEOs) of SOEs are generally appointed by the government, and most of them are current or former government officials. These CEOs are very likely to take advantage of their own political connections to benefit the SOEs. Therefore, compared with NSOEs, SOEs peculiarly have favorable access to the stock market (Aharony et al., 2000), and enjoy government subsidies and lighter taxation. In addition, SOEs have favorable access to bank loans in contrast to NSOEs. This is because four state-owned banks largely dominate China s banking sector, and state-owned banks 10 typically prefer to grant loans to SOEs rather than NSOEs, sometimes for non-profit reasons, such as political, ideological or personal goals (Brandt and Li, 2003). 7 This is called an examination and approval system. 8 Under the fixed price method, the IPO issue price is calculated mainly on the basis of earnings forecast estimated according to prior earnings. Under the relative fixed P/E ratio method, the issue price is calculated as the average of realized profits per share after taxes for the latest three years multiplied by a P/E ratio between 13 and Under the controllable P/E ratio approach, the IPO issue price is determined by the accumulated bidding inquiry pricing method, but the issue price should be less than 20 times the P/E ratio. To test whether the change in pricing method affects earnings management patterns for IPOs, we rerun our main test using two subsamples: IPOs from 2003 to 2004 and IPOs after We find that the empirical results for each subsample are consistent with our main results. 10 For example, the Bank of China, Industrial and Commercial Bank of China, Construction Bank of China, and the Agricultural Bank of China. 8

9 As mentioned before, the objectives of CEOs (or top management teams) can differ between SOEs and NSOEs. The CEOs of SOEs usually hold government positions and some of them would receive political promotions in the future. Consequently, these CEOs act more as politicians than as businessmen, and their performances are evaluated on the basis of not only their financial performance but also various political and social achievements, such as improvement in the employment rate and contribution to the national tax revenue (Fan et al., 2007). Therefore, other than maximizing their firms values, the CEOs of SOEs have various other objectives. In contrast to SOEs, financial performance is the most important criterion for evaluating CEOs performances in NSOEs. Therefore, SOEs should have weaker incentives to manage earnings than NSOEs. 3. Hypothesis development 3.1 Do Chinese IPO firms manage earnings? In the UK and the US, top managers in a firm, who typically hold a certain fraction of the firm s equity, have incentives to manage earnings upwards to influence IPO pricing. However, the recent studies by Ball and Shivakumar (2008) and Armstrong et al. (2009) report no evidence of earnings management around IPOs in both the US and the UK. Ball and Shivakuma (2008) interpret their findings as follows: We attribute this to the higher quality reporting demanded of pubic firms by financial statement users and consequentially higher monitoring by auditors, boards, analysts, rating agencies, press, and litigants, and to greater regulatory scrutiny. (p. 324) For US and UK firms, even though managers have the incentives to manage earnings upwards around IPOs, they will not do so due to strong monitoring and great regulatory scrutiny. Similar to the listed firms in the US and the UK, Chinese firms, both SOEs and NSOEs, have incentives for managing earnings upwards around IPOs. For an NSOE firm, its top managers wealth is usually concentrated on the firm s equity prior to its IPO and would 9

10 be evaluated on its stock price after its IPO. Thus, the managers have strong incentives to manage earnings upwards to boost the issuing price. For a SOE firm, its top managers, who have few shares and no stock options of the firm, may not have direct incentives to manage earnings. Yet, the listing status of SOEs may bring the managers higher prestige and other non-pecuniary benefits, such as political promotion (Aharony et al., 2000). Consequently, top managers of SOEs may also have incentives to manage earnings for IPOs. Unlike the stock markets in the US and the UK, the strong monitoring argument does not apply to China s stock market. As documented in Allen et al. (2005) and Piotroski and Wong (2011), China s stock market exhibits typical features of an emerging equity market: ineffective internal and external monitoring, weak legal systems, a low level of investor protection, and an underdeveloped institutional environment. Given the weak monitoring and regulatory scrutiny in Chinese equity markets, we form our first hypothesis as follows: H1: IPO firms engage in accrual-based earnings management (in China). 3.2 Earnings management for IPO firms: SOEs versus NSOEs Compared with NSOEs, internal and external monitoring is weak for SOEs. Fan et al. (2007) argue and document that firms governance characteristics are weak and less professional when their CEOs have political connections. We can infer that corporate governance is weaker in SOEs than NSOEs since the former s CEOs are connected with the government. Wang et al. (2008) find that local and central SOEs tend to hire small local auditors who, on average, provide lower-quality auditing services. Chen et al. (2010) argue that state-owned banks often make favorable decisions of granting loans to SOEs based on political considerations, rather than on profitability and the credibility of the accounting information. In addition, Chen et al. (2013) find that the role of underwriters in restricting pre-ipo earnings is limited for SOEs. To sum up, monitoring for SOEs is weaker than that for 10

11 NSOEs. Consequently, CEOs of SOEs should have more chances to manage earnings than NSOEs around IPOs. On the other hand, SOEs incentives to inflate earnings are weaker than NSOEs for three reasons. First, as argued by Chen et al. (2011), the compensation contracts for CEOs in SOEs are usually measured by various social and political objectives besides firm value maximization, and these contracts place relatively less weights on firms accounting performances when compared with compensation contracts for managers of NSOEs. In addition, there are little pecuniary benefits from IPOs for CEOs of SOEs, whose wealth is less likely concentrated on the firm s equity. As a result, CEOs of SOEs have weaker incentives to manage reported earnings around IPOs than those of NSOEs. Second, as pointed out by Aharony et al. (2000), a firm s chance of being selected to go public is also related to noneconomic factors, such as political connections and government policies. It is known that SOEs are more closely related to the government than NSOEs so that SOEs have more favorable access to the equity market. Consequently, SOEs have weaker incentives to manage reported earnings than NSOEs. Last, the state-owned banks, dominating the banking industry, prefer to grant loans to SOEs rather than NSOEs (Brandt and Li, 2003). Therefore, SOEs, which have favorable access to bank loans (so as to be less likely to have financial constraints), have weaker incentives to manage earnings around IPOs. Based on the above analyses, both monitoring and incentives are weaker for SOEs than for NSOEs. Weaker monitoring should result in more earnings management for SOEs than for NSOEs while weaker incentives should result in less earnings management for SOEs than for NSOEs. We do not have a prediction on the overall effects of weaker monitoring and weaker incentives on earnings management for SOEs versus NSOEs. We form our second hypothesis (in its null form) as follows: and NSOEs. H2: The degrees of earnings management for IPO firms are the same between SOEs 11

12 4. Variable measurement and research design 4.1 Data and sample selection Our original data consist of all Chinese IPOs over the period, and are obtained from the China Stock Market and Accounting Research (CSMAR) Database. All the IPOs in our sample are required to issue domestic A-shares in the Shanghai Stock Exchange or Shenzhen Stock Exchange, 11 and are also required to have annual financial statement data for calculating the discretionary accruals (earnings management proxy) for the IPO year and one year preceding the IPO. Moreover, we restrict our sample to all nonfinancial firms based on the Industry Classifying Index Code of Listed Companies released by the CSRC. We collect our financial data from the CSMAR database. We require at least 10 firms in each CSRC industry per year, and the firm-year observations in each industry should have sufficient data to calculate the discretionary accruals. We measure total accruals (TA) using the statements of cash flow data as the earnings before extraordinary items (operating income) less operating cash flows (CFO). Our study requires firms ownership information, which is hand-collected from their prospectuses. We identify a firm s ownership type, SOE or NSOE, according to the identity of the ultimate owner disclosed in the firm s prospectus. We define SOEs as those firms whose ultimate owners are state asset management bureaus, the central government or local governments; whereas, NSOEs are defined as those firms whose ultimate owners are individuals or private investors. Following Chen et al. (2011), township-village enterprises are excluded from our sample. 12 Panel A of Table 1 summarizes our sample selection procedures. Following the sample selection criteria, we obtain 437 IPOs in our final sample over the period from 2003 to 11 Firms listed in the Growth Enterprise Market (GEM) board do not fulfill the requirements of profitability or track record as required by the main board of the exchanges. The GEM board is established to provide financing sources for small- and medium-sized enterprises and high-growth companies. We exclude IPOs in the GEM board from our sample since their financial requirements for going public are quite different from those in the main board. 12 A township-village enterprise refers to a business unit that belongs to all residents of a rural community. It is a community enterprise controlled by a community government. Che and Qian (1998) point out that these township-village enterprises are neither SOEs nor NSOEs. 12

13 2009. Panel B of Table 1 reports the distribution of our sample by year. There are only 13 IPOs in 2005 since IPOs were suspended for more than one year starting from May of 2005 due to the Non-tradable Share Reform. 13 Consistent with the belief that firms are more likely to issue equity when their market values are high, there are 106 IPOs in the year of 2007 when the Chinese capital market is booming. Since 2006, the number of NSOEs going public exceeds that of SOEs. Panel C of Table 1 reports the distribution of our sample by industry, showing that IPOs distribute nearly in all industries. However, they cluster in a few industries, such as machinery manufacturing, chemical products manufacturing, metal and non-metal manufacturing, and information technology. Together, IPOs in these four industries comprise more than 45% of our sample. In addition, IPO firms in some industries are dominated by SOEs, such as mining, utilities, transportation and warehousing, and construction. 4.2 Measuring discretionary accruals We use a modified cross-sectional Jones model to estimate the discretionary accruals. 14 For each year, we estimate the model for each industry using non-ipo firms (excluding all observations within two years of an IPO), allowing the coefficients to vary across industries and over time (DeFond and Jiambalvo, 1994). Our primary model for estimating discretionary accruals is based on the following cross-sectional model for each CSRC industry: ACC 1 Sales PPE it, (1) ATA ATA ATA ATA it it it it it it it 13 The Non-tradable Share Reform refers to a process of making state shares and other non-tradable shares tradable on the equity markets. Before the reform, Chinese listed firms generally have nontradable shares and tradable shares. The two types of shares have different prices and benefits. China's stock markets had been in distress for five years before the reform, but they became extremely bullish starting from late As a sensitivity test, we use the performance-matched model suggested by Kothari, Leone and Wasley (2005) to calculate the discretionary accruals. The main results are unchanged when we use performance-matched abnormal accruals as a measure of earnings management. 13

14 where for fiscal year t and firm i, ACC it = OPE it CFO it, where OPE it is earnings before extraordinary items (operating income) and CFO it is operating cash flows taken from the statement of cash flows; ATA it is average total assets ((Assets it-1 + Assets it ) / 2); Sales it is the change in total sales; PPE it is the value of net fixed assets. The coefficient estimates of Equation (1) for each industry and each year are used to calculate the firm-specific non-discretionary accruals (NDA) and discretionary accruals (DA) for the IPO firms in our sample: 15 ( Sales AR ) PPE NDAit ATA ATA ATA 1 it it it ACCit DAit NDA ATA it it it it it (2) where AR it is the change in accounts receivable. 4.3 Regression model Hypothesis H2 states that the levels of earnings management for IPO firms are the same for SOEs and NSOEs. For SOEs, both monitoring and incentives of earnings management are weaker than those for NSOEs; however, they have opposite effects. Weaker monitoring means that SOEs should engage in more earnings management than NSOEs while weaker incentives induce SOEs to engage in less earnings management than NSOEs around IPOs. We cannot predict which of the two effects weaker incentives or weaker monitoring plays a more dominating role in influencing earnings management for SOEs. Accordingly, we 15 We fit Equation (1) to non-ipo firms, and then apply the fitted model to IPO firms to separate the total accruals into discretionary and non-discretionary components. This method might be inappropriate when non-discretionary accruals of IPO and non-ipo firms are differently determined (for example, with different model parameters). To alleviate this problem, we re-estimate Equation (1) using data including IPO firms and non-ipo firms, but allow the coefficients to vary between IPO and non-ipo firms. Then, we apply the estimated model to obtain the discretionary and nondiscretionary components. Using this method to estimate the discretionary accruals does not change our main cinclusion. 14

15 rely on empirical test. We employ a regression model, modified from those of Teoh et al. (1998b) and Fan (2007): 16 DA SOE MARKETI RETENTION AGE it 0 1 it 2 it 3 it 4 it AUDITOR REPUTATION SIZE LEV CFOA 5 it 6 it 7 it 8 it 9 it , (3) GROWTH YEAR INDUSTRY it t t j j it t 1 j 1 where the dependent variable, DA, is discretionary accruals. SOE is a dummy variable that equals one if a firm s ultimate owner is the state asset management bureau, the central government or a local government; and zero if a firm s ultimate owner is an individual or a private investor. MARKETI is the Marketization Index for each province or provincial region constructed by Fan et al. (2009). RETENTION is ownership retention, measured as 1 (number of public shares offered/number of total shares offered). AGE is equal to ln(1 + firm age), where firm age is the difference between the founding year and IPO issue year. AUDITOR is a dummy variable that equals one if the firm s auditor belongs to the Big Eight auditing firms, and zero otherwise. 17 REPUTATION is a dummy variable that gets the value of one if the market share of an IPO firm s leading underwriter is within the top 20, and zero otherwise. 18 This variable is a proxy for the leading underwriter s reputation. SIZE is the natural logarithm of total assets measured at the beginning of the year. LEV is leverage ratio, measured as total liabilities divided by total assets. CFOA is the cash flow on assets defined as operating cash flow divided by average total assets. GROWTH is the growth in sales, measured as the change in sales divided by total sales at the beginning of the year. In Equation (3), the coefficient 1 indicates the difference in the levels of earnings management between SOEs and NSOEs. If SOEs engage in more (less) earnings management than NSOEs in the issue year, we expect 1 to be positive (negative). 16 We also form a Tobit model to replace model (3). If the discretionary accruals, DA, are less than zero, we set DA as zero. Our empirical results are stronger than those obtained from model (3), but our main claim remains unchanged. 17 The Big Eight in China include PwC, E&Y, Deloitte, KPMG, RSM China, Shu Lun Pan, Zhejiang Pan-China, and Shine Wing. 18 Because there is, to date, no IPO underwriter reputation ranking reported by any prestigious research institutes in China, we use each underwriter s market share in the stock underwriting market as a proxy for the underwriter s reputation. 15

16 5. Empirical results 5.1 Earnings management around IPOs Panel A of Table 2 presents the mean and median discretionary accruals over the years from -2 to +2 relative to the IPO year (year 0). We find that the mean discretionary accruals are significantly positive in the IPO year and the two years preceding the IPO. 19 The median discretionary accruals, which are less likely to be influenced by extreme values, are also significantly positive. Therefore, our results are consistent with hypothesis H1, indicating that China s IPO firms engage in accrual-based earnings management. Panel B of Table 2 reports the mean and median discretionary accruals separately for SOEs and NSOEs around IPOs. We find that both the mean and median of discretionary accruals are significantly positive for both SOEs and NSOEs around the issue year. This result indicates that both SOEs and NSOEs manage earnings by taking positive discretionary accruals. It is interesting that the mean and median discretionary accruals for NSOEs are much greater than those for SOEs over the years from -2 to +2. This finding indicates that NSOEs tend to engage in more accrual-based earnings management than SOEs around IPOs. This is just the prima facie evidence for analyzing H2. We will examine the hypothesis in depth by using regression analyses. To examine the earnings performances for IPO firms, Table 3 reports the mean and median of both return on assets (ROA) and difference in ROAs (DROAs) around the issue year. 20 Panel A presents the mean and median of ROA and DROA for the whole sample from year -2 to year +2 relative to the IPO year. We find that both the mean and median ROAs for the post-ipo years are much lower than those for the IPO year and two years preceding it. Also, we find that the DROAs for years -2 and -1 are positive, but changes are negative for the two years after the IPO year. This post-issue poor earnings performance is consistent with 19 It is surprising to note that the abnormal accruals are still significantly positive for the years following the issue year. This result may be driven by the fact that the lockup period for controlling shareholders is 36 months after IPOs. Therefore, IPO firms still have incentives to inflate earnings to prevent their stock prices from dropping sharply within three years after their IPOs. 20 DROA is calculated as the annual ROA minus the ROA in the IPO year. 16

17 the findings in prior research (Aharony et al., 2000), and can be partially explained by the reversal of accruals. Panel B reports the mean and median of both ROA and DROA for SOEs and NSOEs, respectively. It is noted that the post-issue earnings performances for IPO firms, both SOEs and NSOEs, are poor in our sample. To sum up, the empirical results indicate that China s IPO firms, both SOEs and NSOEs, engage in accrual-based earnings management. This leads to poor post-issue earnings performances. Thus, our findings are consistent with H State ownership and earnings management around IPOs Panel A of Table 4 presents the descriptive statistics of firm characteristics for the full sample in the IPO year. 21 The mean (median) discretionary accruals (DA) are (0.069). SOEs make up 39.7% of the firms in our sample. On average, 16.1% of firms shares are nontradable in the equity market. Since the year of establishment, firms in our sample, on average, choose to go public in five years. The Big 8 auditors audit 31.2% of all the firms in our sample. The top 20 investment banks, as the leading underwriters in China, underwrite 56.0% of all the IPO firms. On average, 44.4% of CEOs hold their own firms stocks; 11.3% of the liabilities are long-term. Panel B of Table 4 reports the descriptive statistics for SOEs and NSOEs separately. The mean (median) of discretionary accruals for SOEs is (0.046), lower than that for NSOEs (0.081). Compared with SOEs, NSOEs tend to locate in regions with a higher marketization index. The sizes of SOEs are, on average, much larger than those of NSOEs in our sample. We also note that the debt-to-assets ratio is higher for SOEs (0.380) than for NSOEs (0.324). In addition, the ratio of long-term debt to total liabilities for SOEs (0.177) is higher than that for NSOEs (0.071). About 59.2% of CEOs in NSOEs hold their own firms shares, by contrast, only 22.0% of CEOs in SOEs have stock compensation. 21 We require that the offering characteristic variables and other control variables are available for IPO firms. Thus, our sample used in the regression analyses is 423 IPO firms. 17

18 In Panel C of Table 4, we report the Pearson correlation matrix for variables used in the regression analysis and the path analysis. We note that SOE is negatively correlated with the discretionary accruals (DA). We observe that discretionary accruals (DA) are not significantly correlated with the marketization index (CREDTMI), ownership retention (RETENTION), firm age (AGE), auditor choice (AUDITOR) and the leading underwriter s reputation (REPUTATION). Unsurprisingly, discretionary accruals (DA) are significantly and positively associated with the CEO shareholding dummy (CEOSH). The ratio of longterm debt to total liabilities (LTD) is significantly and negatively related with discretionary accruals (DA). This finding seems to suggest that IPO firms, having favorable access to bank loans (i.e., with higher LTD), are less likely to manage earnings opportunistically. In addition, the correlation coefficients among the independent variables (excluding DA) are less than 0.5. This means that our regression analyses do not suffer seriously from the multilinearity problem. Table 5 reports the OLS regression results of Equation (3). In Panel A, the dependent variable is positive discretionary accruals (DA>0), which reflects the level of incomeincreasing earnings management. The coefficient 1 is significantly negative ( 1 = , t=2.78), suggesting that the discretionary accruals for SOEs are significantly smaller than those for NSOEs. Our second hypothesis does not predict that the levels of earnings management for IPO firms are higher or lower for SOEs than NSOEs. The regression results show that the level of income-increasing earnings management is lower for SOEs than that for NSOEs. In addition, we find that the level of income-increasing earnings management decrease with SIZE, LEV and CFOA. The coefficients for AUDITOR and REPUTATION are not significantly negative, suggesting that the auditors and underwriters do not play a significant role in restricting IPO firms earnings management. This finding is consistent with the argument that the institutional environment is underdeveloped in China. In Panel B of Table 5, the dependent variable is negative discretionary accruals (DA<0), which reflects the level of income-decreasing earnings management. We note that 18

19 the coefficient for SOE ( 1 ), which reflects the difference in levels of earnings management between SOEs and NSOEs, is insignificant ( 1 =0.004, t=0.30), consistent with the conjecture that IPO firms do not have incentives to manage earnings downward. In short, the results in Table 5 reject Hypothesis H2, indicating that the level of income-increasing earnings management is lower for SOEs than that for NSOEs Path analysis In this section, we use path analysis to decompose the correlation between state ownership and IPO firms earnings management into direct and indirect links. This method can be used not only to examine the existence of direct and indirect paths between state ownership and earnings management, but also to provide evidence on the relative importance of each link. We posit a direct link (path) between state ownership and earnings management. As argued by Aharony et al. (2000), a firm's chance of being selected to go public is related to some non-economic factors, such as political connections and government policies. It is known that SOEs are more closely related to the government than NSOEs so that SOEs have more favorable access to the equity market. As a result, SOEs have weaker incentives to manage earnings to enhance the chance for public listing. We posit two indirect links (paths) between state ownership and earnings management that are mediated by CEO shareholding (CEOSH) and accessibility to bank loans (LTD) respectively. First, Armstrong et al. (2009) claim that managers, who usually hold their own firms shares, have incentives to manage earnings around IPOs. Once companies go public, managers wealth will be evaluated on the basis of the firms stock prices. In China, managers of SOEs hold few shares and no stock options, while managers of 22 In Table 5, we only examine whether our hypothesis H2 holds in the IPO year. Chinese IPO firms may have strong incentives to manage earnings in years immediately prior to the issue year. To examine this issue, we rerun our main test using data for two years preceding the IPO year. Our (untabulated) results show that SOEs also manage earnings to a lesser degree than NSOEs before the year of IPO. 19

20 NSOEs usually hold a considerable number of shares of their own firms. Therefore, state ownership can affect IPO firms earnings management indirectly through the link of CEO shareholding. Second, as we argued, SOEs have more favorable access to bank loans from state-owned banks. They are less likely to be financially constrained. Thus, SOEs have weaker incentives to manage earnings around IPOs. Therefore, state ownership can impact IPO firms earnings management indirectly through its link to firms accessibility to bank loans. Table 6 reports the results for the path analysis. The first two columns show results when only one mediated (by CEO shareholding (CEOSH)) link between state ownership and earnings management is considered. Starting with the Pearson correlation between state ownership (SOE) and earnings management (DA), r[soe, DA] equals , significant at the level. p[soe, DA] is the direct path coefficient, and the direct path captures 72.0% of the correlation between state ownership and IPO firms earnings management. p[soe, CEOSH] is the path coefficient between state ownership and CEO shareholding, and p[ceosh, DA] between CEO shareholding and earnings management. The indirect path mediated CEO shareholding (CEOSH) is calculated as the product of p[soe, CEOSH] and p[ceosh, DA]. The indirect path coefficient is , capturing 28% of the total correlation. The last two columns present the path analysis when we consider the two indirect links between state ownership and earnings management. Again, we begin our analysis with the Pearson correlation coefficient, r[soe, DA], which is (t=3.18). 51.9% of the correlation between SOE and DA is attributed to a direct path p[soe, DA], 21.6% to an indirect path mediated by CEO shareholding and 26.5% to an indirect path mediated by the accessibility to bank loans (LTD). This result suggests that the two mediators (CEOSH and LTD) account for 48.1% of the total correlation between SOE and DA, and that the accessibility to bank loans (LTD) is a more important mediator which leads to less earnings management for SOEs than NSOEs around IPOs. 20

21 6. Additional and sensitivity tests 6.1 Central SOEs and local SOEs In this section, we classify SOEs into two groups: central SOEs and local SOEs. 23 This is because some differences between the two groups may affect earnings management. For example, Wang et al. (2008) find that local SOEs have a stronger propensity to hire small local auditors than central SOEs. This is probably because the local governments have more direct political power over small local auditors than the central government. Jian and Wong (2010) document that central SOEs are more likely to have higher abnormal related sales than local SOEs and non-state-owned enterprises. Considering the possible difference in the levels of earnings management between central SOEs and local SOEs, we estimate Equation (3) where the variable SOE is replaced by CSOE, which is a dummy variable that equals one if a firm is identified to be a central SOE, and zero otherwise. Table 7 reports the regression results. If the dependent variable is positive abnormal accruals (DA>0), the coefficient on CSOE is not significantly different from zero ( 1 = (t = -0.72)), indicating that the levels of income-increasing earnings management for central SOEs are not different from those for local SOEs. Also, the levels of income-decreasing earnings management do not differ between central SOEs and local SOEs because the coefficient for CSOE is not different from zero ( 1 = (t = -0.53)) when the dependent variable is negative abnormal accruals. In sum, the results in Table 7 indicate that there is no difference in the levels of earnings management between central SOEs and local SOEs in the IPO year. 23 Central SOEs are defined as those SOEs owned by departments of the central government, such as the State-owned Assets Supervision and Administration Commission (SASAC) or Ministry of Finance; local SOEs are owned by agencies of the local government, such as the provincial state asset management bureaus or finance bureau. 21

22 6.2 Real earnings management One may argue that SOEs engage in less accrual-based earnings management than NSOEs since the former prefer to prop up earnings using real activities, such as related-party sales (Jian and Wong, 2010; Aharony, Wang and Yuan, 2010). To address this concern, we examine real earnings management for Chinese IPO firms. Following Roychowdhury (2006), we examine the following manipulation of real activities: acceleration of the timing of sales through increased price discounts; more lenient credit terms or related-party sales; reducing the cost of goods sold by overproduction; and cutting discretionary expenditures, including advertising expenses, research and development (R&D) and selling, general and administrative (SG&A) expenses. Given the sales levels, firms that manage earnings upwards are likely to have one or all of the following: abnormal low cash flows from operations, and/or abnormal low discretionary expenses, and/or abnormal high production costs. Table 8 presents the mean of abnormal cash flows from operations (Panel A), the mean of abnormal production costs (Panel B), and the mean of abnormal discretionary expenses (Panel C) for SOEs and NSOEs around IPOs, respectively. We find that the mean of abnormal cash flows from operations and the mean of abnormal discretionary expenses, for both SOEs and NSOEs, are not significantly negative, and the mean of abnormal production costs, for both SOEs and NSOEs, is not significantly positive. This finding indicates that both SOEs and NSOEs do not manage earnings upwards via real activities. 6.3 Market reactions to SEO Announcements Earnings management around IPOs can mislead investors for a short time, but the market may learn of the manipulation over time. If this is the case, the market may punish the manipulators in subsequent Seasoned Equity Offerings (SEOs). To test whether this unraveling effect exists, we regress of a firm's three-day abnormal returns around SEOs on its discretionary accruals in the IPO year and other control variables. Table 9 presents the 22

23 regression results. We find that the coefficients of DA are not significantly different from zero for both the SOE group ( 1 =0.031 (t=0.44)) and NSOE group ( 1 = (t=-0.26)), indicating that the market seems not to punish the manipulators in IPOs in China. 6.4 Ball and Shivakumar s (2008) model We also estimate the abnormal accruals for our sample using Ball and Shivakumar s (2008) model. Panel A of Table 10 reports the regression results and Panel B reports the abnormal accruals estimated by Ball and Shivakumar s (2008) model. We find that the mean and median of abnormal accruals are positive and significant. This finding is consistent with our main results and indicates that China s IPO firms in our sample, on average, engage in earnings management through taking positive abnormal accruals. In addition, the regression results in Panel A show that the coefficients on DCFO*CFO, which reflects conditional conservatism, are (t= -8.20) for IPO year 0, (t= -5.03) for year -1, and (t= -7.81) for year -2. This result suggests that Chinese IPO firms report earnings aggressively in the issue year and years preceding the IPO. 6.5 Performance-matched abnormal accruals We also repeat our analyses using the performance-matched abnormal accruals suggested by Kothari et al. (2005). Following Kothari et al. (2005), we match each observation in our IPO sample with another non-ipo firm with the closest ROA (earnings before extraordinary items scaled by average total assets) in the same industry and the same year. Then, we estimate the Modified Jones model for both IPO firms and their matched non- IPO firms. The performance-matched abnormal accruals for the IPO firm i are measured as the abnormal accruals for firm i minus the matched firm s abnormal accruals. We repeat our analyses using performance-matched abnormal accruals as a measure of earnings management. The results (untabulated) are consistent with our main findings. 23

24 6.6 Other tests In 2005, the IPO pricing method was changed from the controllable P/E ratio approach to accumulated bidding inquiry approach. To test whether the change in pricing method affects earnings management patterns for IPOs, we rerun our main test using two subsamples: IPOs from 2003 to 2004 and IPOs after The untabulated results show that IPO firms for each subsample engage in earnings management. In addition, IPOs were suspended for more than one year starting from May of 2005 due to the Non-tradable Share Reform. Before the Non-tradable Share Reform, a firm's value consists of the value of tradable shares and the value of non-tradable shares. Only the value of tradable shares is determined by stock prices. While after the reform, IPO firms can issue tradable shares only. As a result, a firm's value is more closely related to its stock price. Therefore, IPO firms may have stronger incentives to manage earnings upwards to boost their stock prices. To test whether the Non-tradable Share Reform affects IPO firms' level of earnings management, we split our sample into two groups: IPOs from 2003 to 2004 and IPOs after We rerun our main test for each group and find that the level of earnings management does not differ between the two groups. The results are not tabulated. 7. Conclusion This study investigates whether issuing firms inflate earnings around IPOs in China, and if so, which type of IPO firms, SOEs or NSOEs, engages in more earnings management. We find that Chinese IPO firms, on average, have high positive discretionary accruals in the issuing years and the two years before the listing years. This finding is consistent with the strong monitoring argument made by Ball and Shivakumar (2008). In contrast to the strong monitoring in the US or UK equity markets, the monitoring and regulatory scrutiny are weak in China (Piotroski and Wong, 2011). The weak monitoring cannot prohibit IPO firms from managing earnings. 24