Earnings Management and the Operating Performance around the Private Placement of Equities (PPEs)

Transcription

1 Earnings Management and the Operating Performance around the Private Placement of Equities (PPEs) Binqing Xiao School of Management and Engineering, Nanjing University, , China Panagiotis Andrikopoulos School of Economics, Finance and Accounting, Coventry University, CV1 1FB, UK Xiaowei Liu School of Economics, Finance and Accounting, Coventry University, CV1 1FB, UK Jun Wang * School of Economics, Finance and Accounting, Coventry University, CV1 1FB, UK Abstract This paper investigates the post-issue operating performance of private placement of equities (PPEs) and its relationship with pre-issue earnings management. We find that 1) there is persistent post-ppe long-term operating performance decline for issuing firms, 2) discretional accruals are significantly positive in issuing year and after for issuing firms, 3) post-ppe long-term operating performance are driven, at least in part, by pre-ppe upward earnings management. Overall, these results suggest that firms manipulate earnings prior to PPEs. We also investigate the alternative explanations including overinvestment effect, agency problem, and leverage effect, all of which are ruled out in this study. The main results are robust to a variety of operating performance measures, earnings management measures, matching methods, and the effect of split share structure reform in China. JEL classification: G32 Keywords: Earnings management; private placement of equities (PPEs); discretional accruals, Chinese listed companies. * Corresponding author. Tel: ; ab9571@coventry.ac.uk addresses: bengking@nju.edu.cn (Binqing Xiao), ab7724@coventry.ac.uk (Panagiotis Andrikopoulos), liux55@uni.coventry.ac.uk (Xiaowei Liu) 1

2 I. Introduction Previous studies have documented the operating performance following seasoned equity offerings (SEOs) and initial public offerings (IPOs) (e.g.,jain and Kini, 1994; Loughran and Ritter, 1997; Teoh et al., 1998a; Teoh et al., 1998b; Rangan, 1998; Shivakumar, 2000; Kao et al., 2009; Cohen and Zarowin, 2010; Fu, 2010; Bae et al., 2013). However, little attention has been paid to the operating performance of the private placement of equits (PPEs). In this study, we investigate the operating performance around the private placement of equities, particularly, we explore whether the operating performance deteriorates after the issuing of PPEs. Furthermore, we investigate the determinants of post-issue operating performance of PPE, partuclarly focusing on pre-issue earnings managagement. Previous stuides have documented earnings management behaviour around seasoned equity offerings (SEOs) and earnings management behaviour around initial public offerings (IPOs) (e.g.,teoh et al., 1998a; Teoh et al., 1998c; Aharony et al., 2010). They argue that in order to increase the offering proceeds, issuing firms manage earnings upward prior to SEOs and IPOs. Given that high pre-issue earnings are not sustainable, the operating performance declines in the post-offering period, which is termed as earnings management hypothesis in this study. Earnings management hypothesis predicts worse post-issue operating performance for issuers with unusually large income-increasing accounting adjustments prior to the SEO and IPO. Similar to SEO and IPO, we conjecture that firms manipulate earnings upward prior to new shares issuing in PPEs to increase the offering proceeds. Due to the manipulated earnings are not sustainable, the operating performance deteriorates after issuing year. The aim of this study is to test whether earnings management hypothesis applies to PPE. In other words, we attempt to explore whether firms conduct earnings management around the period of PPE. To achieve this aim, the objective of this study is to address the following specific research questions. Firstly, what is operating performance around PPEs, particularly post-issue operating performance of PPEs? Secondly, what is earnings management behaviour around PPEs, particularly prior to PPEs. Thirdly, does pre-issue earnings management explain post-issue long-term operating performance of PPEs? To provide compelling empirical evidence that firms manipulate earnings around the private placement of equity, three hypotheses must be tested. Firstly, earnings increase prior to PPEs and reach its peak 2

3 in issuing year, and then decline afterwards. Secondly, upward earnings management exists prior to PPEs. Thirdly, pre-issue earnings management proxies can explain post-issue operating earnings change. We find evidence to show empirically that these hypotheses hold true in this study. To address these research questions, we focus on publicly traded companies in China for the following reasons. Firstly, private placement of equity is pervasive in China. More than 1,600 private placement of equities occur between year 2006 and 2014 in China, which is comparable to previous studies (e.g.,rangan, 1998). Table 1 and Figure 1 shows that there was only 46 PPEs in year The total number of offering increases gradually since then and in year 2016, there are 265 PPEs in China. Secondly, earnings management behaviours are not uncommon in China. Yu et al. (2006) and Haw et al. (2005) document earnings management in Chinese listed companies during the period of rights offerings. Aharony et al. (2010) investigate earnings management during the IPO process in China and find evidence that Chinese listed companies conduct earnings management during the process of IPO issues. Chen et al. (2008) examine how local government in China facilitate publicly traded companies in earnings management to circumvent relevant regulations, such as the minimum threshold for rights offerings and de-listings. Overall, China provides an ideal setting to investigate earnings management hypothesis in PPEs. We find the following results. Firstly, we find that the pre-issue operating performance increase up to the issuing year of PPEs, reach its peak in issuing year, and then the post-issue operating performance of issuing firms deteriorates afterwards. Particularly, the earnings of issuing firms are significant higher in issuing year relative to that in three years before and after offering. The mean (and median) ROA of issuing firms is 4.59% (3.98%) in issuing year (i.e., year T), while they are only 1.67% (2.82%) in three years before offering (i.e., year T-3) and 3.02% (2.73%) in three years after. Moreover, the earnings of issuing firms are significant higher in issuing year relative to that of control firms constructed based on issuing year, industry, firm size and operating performance. The earnings of controlling firms in issuing year is significantly lower relative to that in issuing firms in issuing year. The mean and median ROA of control firms are 3.30% and 3.17%, which are significantly lower than that for issuing firms in issuing year. The main results do not change when using alternative operating earnings measures, such as ROE, profit margin, EBITDA relative to total assets or annual sales etc. Overall, these results indicate that operating performance of issuing firms increase significantly prior to PPEs, and reach its peak in in the issuing year and then declines subsequently to the offering. 3

4 Secondly, the issuers manipulate their earnings upwards prior to PPEs and in issuing year. The discretionary accruals are positive and statistically significant from zero in the issuing year. The mean (and median) discretionary accruals are 2.65% (1.85%) in issuing year when using Jones model. Both mean and median are significantly different from zero at 1% significance level using T-test and Wilcoxon rank test respectively. These results suggest that issuing firms manipulate their earnings upwards in issuing year. The level of earnings management is slightly less than counterparts in existing (e.g.,chou et al., 2010). The results about discretionary accruals based on modified Jones model is similar to the results based on Jones model. Finally, we test the relationship between earnings management and operating performance of PPEs firms, particularly the relationship between pre-issue earnings management and (long-term) post-issue operating performance of issuing firms. We test whether pre-issue earnings management can explain the post-issue operating performance deterioration 2. We find a significantly negative association between pre-issue abnormal accruals and post-issue ROA and ROE change, suggesting that the deterioration in operating performance is due, at least in art, to pre-issue earnings management (rather than genuine growth in profitability). Our results indicate that earnings management explain the underperformance of post-issue firms as earnings management is negatively and significantly related to the post-issue operating performance. There are a couple of alternative explanations for the decline in the post-issue operating performance of PPEs. A first alternative explanation is related to overinvestment (Fu, 2010). Fu (2010) argues that poor post-issuing operating performance is due to the overinvestment of the issuing firms as overinvestment significantly reduces the asset productivity of issuing firms, which leads to poor operating performance. To test this argument, we add additional control variable controlling the effect of overinvestment on post-issue operating performance and we find no evidence on overinvestment argument. A second alternative explanation comes from agency problem. The poor post-issuing operating performance might be related to the potential for increased agency conflict between controlling shareholders and minority shareholders. The agency problem could deteriorate due to entrenchment effect and alignment effect. To test this argument, we add the change of shareholding owned by 2 We model changes in the firms operating performance as a dependent variable and discretional accruals as key independent variable controlling for various microeconomic characteristics such as current financial ratio, leverage, sales growth, capital expenditure and book-to-market. 4

5 controlling shareholders as the additional control variable and re-run main regression. The coefficients on this control variable are not significant, which indicates that agency problem story does not satisfied. The coefficients on earnings management variables are still negative and significant indicating that the earnings management hypothesis still holds. A third explanation for the operating performance decline is leverage effect. The leverage effect comes from the benefit of tax shield of debt (Modigliani and Miller, 1963). It argues that when firms issue new shares, the level of leverage will be reduced. Thus, the benefit of debt tax shield will be reduced, which leads to lower net earnings after new shares issued. We test this argument by adding the change of leverage of issuing firms between before and after the new shares issuing as additional control variable and re-run main regression. The coefficients on this control variable are not significant, which indicates that the story of leverage effect does not exist. The coefficients on pre-issue earnings management variables are still negative and significant indicating that the earnings management hypothesis is still satisfied. This study adds to the literature about the operating performance around new shares issuing. Previous studies document that listed companies have poor operating performance after stock issues (e.g.,ritter, 1991; McLaughlin et al., 1996; Rangan, 1998; Teoh et al., 1998b; Allen and Soucik, 2008; Loughran and Ritter, 1997) 3. Hansen and Crutchley (1990) and Loughran and Ritter (1997) document that seasoned equity offerings are followed by significant earnings declines. Jain and Kini (1994) investigate the postissue operating performance of IPO firms. They find that there is a significant decline in operating performance subsequent to firm s IPO. Spiess and Affleck-Graves (1995) document that operating performance declines significantly after additional share issuance. McLaughlin et al. (1996) report a significant decline in the operating performance of companies following a seasoned equity offerings (SEOs). The post-issue underperformance is also corroborated by Loughran and Ritter (1997) who show that U.S. listed companies have poor operating performance (and stock returns) during the five years after SEOs. Rangan (1998) and Teoh et al. (1998b) report a deterioration in both the operating and market performances after shares issuing. Allen and Soucik (2008) find that issuing companies significantly underperform their non-issue counterparts. (Loughran and Ritter, 1997)report that the operating performance of issuing firms is better than that of a control group just prior to the issue, but 3 Previous literature also documents the poor post-issue stock performance after stock issues (e.g., Ritter, 1991; ). 5

6 deteriorates afterwards. This study also adds to the literature about the earnings management, particularly around the new share issuing. Firms manage their reported earnings prior to corporate events such as management buyouts (Perry and Williams, 1994), initial public offerings (IPOs) (Teoh et al., 1998a), seasoned equity offerings (SEOs) (Teoh et al., 1998b; Shivakumar, 2000), and stock-for-stock mergers (Erickson and Wang, 1999; Louis, 2004), openmarket repurchases (Gong et al., 2008). This study investigates the earnings management behaviour around PPEs, which are ignored in previous literature. In addition, this study extends prior research on the relationship between earnings management and share issuing to developing countries, which is China in this study. Most existing studies investigating the relationship between earnings management and new share issuing (e.g., SEOs or IPO) focus on U.S, where the law system, corporate governance and regulation are completely different from that in China. Given that China is the second largest economy around the world now, it is very worthwhile to test whether previous main results still hold in developing countries. The remainder of this paper is organised as follows. Section II briefly discusses the institutional background of private placement of equity in the Chinese stock market. Section III describes the data. The analyses about the operating performance and discretional accruals around PPEs are presented in Section IV and V respectively. Section VI presents empirical findings about the relationship between pre-issue earnings management and post-issue operating performance. Alternative explanations about the poor post-issuing operating performance are discussed in Section VII. Robustness check and conclusion are showed in Section VIII and IX. 6

7 II. Institutional Background Compared with its western counterparts, PPE in China has shorter history. On 8 May 2006, the China Securities Regulatory Commission (CSRC) issued The Administration of the Issuance of Securities by Listed Companies. Since then, PPEs have become the main methods of equity refinancing for listed companies in China 4. The figure 1 shows the development of PPEs in the Chinese stock market during the period between 2006 and From this figure, we can find that the popularity of PPEs in China increases sharply over the past ten years since it was introduced, particularly in recent years. In Panel B and Panel C in Table I, it shows that the number of PPEs in China was 49, 129, and 95 in years 2006, 2007, and 2008 respectively. These figures increase gradually to 208, 288, 265 in year 2014, 2015, 2016 respectively. In addition, less than 5 percent firms issued new shares through PPEs in the beginning year of Since PPEs were introduced in China, now more than 20 percent firms issued new shares through PPEs in Chinese stock markets in latest year. Furthermore, PPEs have become the only refinancing method used for listed companies in China. [Insert Figure 1 here] PPEs in China are highly regulated and have some unique features. Firstly, companies can only sell new shares to pre-specified investors for private placement of equity, while for public equity offering, companies can sell new shares to the public and for rights offering, only existing shareholders of the issuing companies can purchase new shares. Secondly, the total number of investors allowed to purchase new shares through PPEs must be no more than ten investors each time. Thirdly, there is no financial requirement for companies applying for share issuance through PPE. Companies can issue shares through PPE even if the issuing firms suffer from a profit loss, while firms have to meet certain financial requirements if they wish to issue share through public equity offering or right offering in China 5. Fourthly, there is no constraint to re-sell new shares for public equity offering and rights offering. However, there is a lockup period for PPEs. New shares are not allowed to re-sell within the first 36 4 China's Securities Regulatory Commission (CSRC) establish Chinese stock market in the early 1990s, and regulate share issuance with merit-based system. The purpose of the establishment of the Chinese stock market is to help state-owned enterprises raise capital and improve operating performance GREEN, S CHINA'S STOCK MARKET EIGHT MYTHS AND SOME REASONS TO BE OPTIMISTIC.. In the early 1990s, Chinese listed companies were allowed to issue additional shares to their existing shareholders with pre-emptive rights in the 3 years after their IPOs as rights offering CHEN, K. C. & YUAN, H Earnings management and capital resource allocation: Evidence from China's accounting-based regulation of rights issues. The Accounting Review, 79, In order to create more financing opportunities and ability, the large-scale seasonal equity offerings (SEOs) were permitted by China's Securities Regulatory Commission (CSRC) in 2000, which including public equity offering and rights offering CHEN, K. C. & WANG, J Accounting-based regulation in emerging markets: The case of China's seasoned-equity offerings. The International Journal of Accounting, 42, Rights offering was the main way of refinancing for Chinese listed companies in early 1990s. 5 The financial requirement for rights offering is as follows. Before 1999, the average return on equity (ROE) must be higher than 10% over the past three years (Corporate Law, 1994). During the period of 1999 to 2001, the average ROE must exceed 10% in the past three years, but not lower than 6% in any of these years (1999). After 2001, the average ROE must exceed 6% in the past three years (2001). The financial requirement for public equity offering is the same as the financial requirement for initial public offering. 7

8 months for controlling shareholder purchasers and not allowed to re-sell within the first 12 months for other purchasers. Fifthly, the issuing price through PPEs must be not less than 90% of the current stock price of issuing firms. Lastly, each investor is not allowed to purchase new shares more than 5% of total number of outstanding shares. The issuing procedure of PPEs in China includes several steps. Firstly, the proposal of PPEs has to be approved by a board of directors through a board meeting first. Secondly, the proposal of PPEs has to be approved by majority shareholders through annual general meeting. After that, the proposal of PPEs has to be approved by CSRC, the regulator in China. Once the proposal of PPE is approved by CSRC, the issuing firms can issue new shares through stock exchanges. 8

9 III. Data This study focuses on all Chinese listed companies listed in main board of Shanghai Stock Exchange and Shenzhen Stock Exchanges during the period of 2006 to The main data stems from the China Stock Market and Accounting Research database (CSMAR). Our initial sample consists of 1,627 private placement companies. In line with prior literature (e.g.,loughran and Ritter, 1997; Shivakumar, 2000; Koeplin et al., 2000; Cronqvist and Nilsson, 2005), all financial companies are excluded as their business operations and their financial statements are different to all other companies. In addition, in order to rule out possible influence of other equity offering methods, we also exclude all companies that have rights offers, public offers and initial public offering (IPOs) at the same year with the private equity placements. Given that this study investigates operating performance around PPEs including both the pre- and the post-issue years, all of these companies have complete financial statement data from three years before private placement (T-3) to three years after issue (T+3). Due to the fact that some companies miss accounting data during the sample period, the final issuing companies is 804. Table 1, Panel A reports the distribution for all private equity placements companies by year and industry. All 1,629 private placements are associated with nineteen sectors including Agriculture, Mining and Manufacturing etc. This provides a good representation of the whole market. In addition, more than half (56.61%) of issuing companies are from the manufacturing sector, which is hardly a surprise as most Chinese listed companies are in the manufacturing sector. [Insert Table 1 here] Table 1, Panel B presents the descriptive statistics of the number of issues and its percentage during the period There are only 46 private placements in the year of 2006, however, the number of private placements increases gradually from 95 in year 2008 to 265 in The largest number of issuing companies is 288 in year In addition, Panel B shows that only 3.55% firms issued new shares through PPEs in year Now this percentage increased to more than 20% after Also, the percentage of new share issuings through PPEs relative to total number of issuings through PPEs increase gradually. It was 2.83% in year Now it increases to 17.70% in year Table 1, panel C presents the descriptive statistics of PPE issuing size. The issuing size of PPEs is only 12,826 million dollars in year This figure increased by ten time in ten years and it reached 138,143 million dollars in year In addition, about 87.74% refinancing in listed companies in China were 9

10 done through PPEs. Now 100% refinancing in listed companies in China were done through PPEs. This is consistent with the fact that private equity issue becomes the most popular method for listed companies in China to raise capital through stock exchanges since It is attributed that there is no financial requirement for companies issuing new shares through private placement, even the firms with negative profits can issue new shares through private placement of equity. 10

11 IV. Operating performance around PPE To test earnings management hypothesis, the first step is to examine the operating performance around PPEs, particularly whether the post-issue operating performance deteriorate. To gauge the operating performance of our sample firms, we computer the ratio of net income to total asset (ROA) and the ratio of net income to equity of the firm (ROE) for the years surrounding the PPEs 6. To investigate the operating performance change, we calculate both ROA and ROE change three years before and three years after the firm issue new shares through PPEs. Because of the skewness of accounting ratios, we report both mean values as well as median values on operating performance 7. Table II shows both the mean (and median) of ROA and ROE for issuing firms from year T-3 to year T+3, where year T represents the issuing year of private placement of equity, and T-3 and T+3 represent three years before and three years after issuing. It shows that both ROA and ROE increase from year T-3 up to year T and reach its maximum in the issuing year T. It then declines until the year T+3 afterwards. In particular, the mean (median) ROA is 1.67% (2.82%) in year T-3 and increases to 4.59% (3.98%) in year T, which is comparable to the findings documented by previous studies 8, and then decreases to 3.02% (2.73%) in year T+3. The mean (median) of ROA reaches its peak in year T, which is issuing year. Similar results can be found in Table II when we use ROE as measurement of operating performance. The time-series patterns of operating performance are very similar to the findings of Loughran and Ritter (1997) and Fu (2010). In addition, to eliminate the influence of industry, size and firm value, we construct control groups for issuing firms based on issuing year, industry, firm size and operating performance measures. We choose firms from the same issuing year, same industry, similar size and operating performance measures as control firms of the issuing firms. Based on control firms, we calculate the operating performance of control firms and compare these with that of issuing firms. Table II shows both the mean (and median) ROA and ROE from year T-3 to year T+3 for control firms. The values both for mean (and median) ROA and ROE are relative stable for control firms before to and after PPE issuing. The mean (and median) ROA for control firms is 3.98% (3.54%) in year T-3, and 3.30% (3.17%) in year T, and 2.56% 6 We use various measures as operating performance in robustness check. 7 DeAngelo (1998), Kaplan (1989), Healey and Palepu (1990), Degeorge and Zeckhauser (1993), Jain and Kini (1994), Mikkelson, Partch, and Shah (1997), and McLaughlin, Safieddine, and Vasudevan (1996), Loughran and Ritter (1997), Fu (2010), among others, all report median values. 8 Fu (2010) reports that the ROA in issuing year is 5.28%. 11

12 (2.47%) in year T+3. The similar results can be found in ROE for control firms. Figure 2 shows the trend of operating performance from year T-3 to year T+3. The results confirm the conjecture that the operating performance of issuing firms increases prior to PPEs and then decreases afterwards, and reaches its peak in issuing year of new shares. [Insert Figure 2 here] Furthermore, to compare the difference in mean (median) in ROA between issuing firms and control firms, we carry out T (Mann-Whitney) test in ROA between issuing firms and control firms. It shows that the p values are significantly less than 1% for year T and year T+1 indicating that the operating performance of issuing firms in year T and year T+1 are significantly higher than its counterparts. [Insert Table II here] To further check whether the post-offering operating performance of issuing firms deteriorate or not, we calculate the operating performance change in return on assets (ROA) and return on equity (ROE) for issuing firms between post-issue year and issuing year. Table II reports the results. It shows both return on assets and return on equity decline significantly after share issuing. In particularly, the differences between return on assets in year T+2 and T+3 and that in year T ( ROA2, ROA3) are significantly negative. The difference between ROA in year T+2 and year T is -1.46% (the p-value is ), while the difference between ROA in year T+3 and T is -1.62% (the p-value is ), which indicates that the operating performance decline significantly two (three) years after new shares issuing 9. The results still hold when operating performances are proxied by return on equity. [Insert Table III here] 9 The magnitude of the size is relatively smaller than that documented by Fu (2010). Fu (2010) documents the median difference in raw operating performance between the post- and pre-issue years. The pre-issue years are defined to be the three years before the offering, and the post-issue years are the five years after the offering. The median change in ROA is -1.88%. However, the magnitude is relatively larger than that documented by Rangan (1998). Rangan (1998) find that the median change in ROA is -0.97% in year 1 (significant at 1% significance level) and -0.67% in year 2 (significant at 1% significance level), and -0.23% in year 3 (significant at 10% significance level). Given that the quarterly data was collected, Rangan (1998) defines ROA in year 0 as the sum of income before extraordinary items over quarter -1 to quarter 2, scaled by assets at beginning of four-quarter period. Performance over four quarter intervals prior to and subsequent to year 0 are similarly aggregated to construct ROA for years -2 to 3. ROA are computed as first differences in annual ROA. 12

13 Our results about the post-issuing operating performance decline are similar to findings in previous literature. (Jain and Kini, 1994) document a significant decline in operating performance subsequent to the initial public offering (IPO). They find that the median changes (industry-adjusted change) in return on assets are -3.58% (-2.91%), -7.60% (-6.24%), % (-8.12%), and -9.09% (-6.81%) (all significantly different from zero at the 0.01 level) for years T, T+1, T+2, and T+3 relative to year T-1. We find a similar trend in this study 10. Rangan (1998) documents post-issuing operating performance of seasoned equity offering firms. He calculates median changes in return on assets ( ROA) for years 1 to 3 relative to year 0, and finds that issuing firms experience earnings increases up to year 0 and earnings declines in subsequent years. Particularly, median ROA is negative and statistically significant in each of years 1-3. These results are similar to that documented by Hansen and Crutchley (1990), and Loughran and Ritter (1997). In addition, Hertzel et al. (2002) also documents poor post-issue operating performance of PPE firms. They find that the post-issuing operating performance for issuing firms are significantly lower than that of industry counterparts. However, the difference in operating performance becomes smaller indicating that the post-issuing operating performance of issuing firms is improving. 10 We find a similar trend when we choose year T as the bench year to construct control firms. 13

14 V. Discretional Accruals around Private Placement of Equities A. Jones model and modified Jones model In this study, we calculate the accrual-based earnings management using the Jones model (1991) and the modified Jones model (Dechow et al., 1995). Jones model and modified Jones model are widely used in previous literature measuring the degree of earnings management (e.g.,rangan, 1998; Teoh et al., 1998b; DuCharme et al., 2004; Nagata and Hachiya, 2007; Kim and Park, 2005). Jones model suggests that discretionary accruals can be used as a proxy for company s earnings management, and the discretionary accruals are obtained by subtracting the non-discretionary accruals from the total accruals. The Jones model captures non-discretionary accruals using the following regression equation:,, = β, + β,, + β,, + ε, (1) where TA, = NI, CFO, ; TA represents the total accruals in year t for firm i; NI, represents the net income in year t for firm i; CFO, represents the operating cash flows in year t for firm i; REV represents changes in net revenue from year t-1 to year t for firm i; PPE, represents the value for property, plants, and equipment at the end of year t; A, represents the total assets in year t - 1 for firm i, ε, represents the error term in year t for firm i. Dechow et al. (1995) argue that the traditional Jones model tends to underestimate the earnings management when companies control the operating income with account receivables. To address this problem, Dechow et al. (1995) modify the Jones model retaining the main assumptions of the original model but subtracting the changes in account receivables from the changes in the net revenues. It is generally acknowledged that this modified Jones model has stronger explanatory power in capturing the level of earnings management compared to its original version (Bartov et al., 2000; Kothari et al., 2005). In this modified version, non-discretionary accruals are estimated using the following specification:, = β + β,,, + β, + ε,, (2),, where REV, represents the net account receivables in year t defined as the account receivables in year t less the account receivables in year t-1. All other variables are the same as in the original model. 14

15 B. Discretional accruals around PPE We calculate the discretional accruals based on Jones model and modified Jones model discussed above. Particularly, we calculate the earnings management both before shares issuing and after shares issuing. The main results are reported in Table III. Panel A shows that the mean discretional accruals are significantly positive no matter which model we use. The mean (median) discretional accruals is 2.65% (1.85%) when we use Jones model and is 2.96% (1.84%) when we use modified Jones model. This indicates that issuing firms manipulate earnings upwards in the issuing year. In addition, we calculate the discretionary accruals around the period of issuing. Particularly, we calculate the discretionary accruals three years before and three years after issuing. The results are reported in Panel B. It shows that the mean (median) discretionary accruals increase from year T-3 up to year T+1 and reach its maximum in the year T+1. It then declines afterwards. In particular, the mean (median) discretionary accruals is ( ) in year T-3 and increases to (0.0185) in year T. It reaches its peak at (0.0190) in year T+1 and then decreases to (0.0110) in year T+3. Similar results can be found in panel C in table III except that discretional accruals reach its peak in issuing year instead of year T+1 when we use modified Jones model to calculate the discretionary accruals as measurement of earnings management 11. Furthermore, we carry out T test (Wilcoxon test) to examine whether the mean (median) of the discretionary accruals are significant or not around the period of PPEs. It shows that the discretionary accruals are significant different from zero at 1% significance level in T-1 year and afterwards when we use Jones model to calculate the discretionary accruals. We find a similar result when we use modified Jones model. These results indicate that firms manipulate earnings upwards one year before issuing. [Insert Table III here] [Insert Figure 3 here] Overall, these results indicate that companies conduct earnings management before and in the year of private placement of equity and the degree of earnings management reduces after private placement 11 These figures are slightly higher than the counterparts reported by Rangan (1998) when he studies earnings management behaviour around seasoned equity offering for American firms. The median discretionary accruals Rangan (1998) reported in issuing year is 0.83% and is 1.15% in year T+1. However, both Rangan and us find a similar result that issuing firms manipulate earnings significantly upwards in issuing year and after. 15

16 of equity. Our results are different from previous studies, such as Rangan (1998), Teoh et al. (1998b) and Cohen and Zarowin (2010), as they find that companies manipulate the earnings in the year of the offering instead of one year before in this study. 16

17 VI. Earnings management and PPE While it is difficult to pinpoint the exact reason for the inferior operating performance of PPE firms, several possibilities come to mind. We test earnings management hypothesis in this section and discuss alternative explanations in next section. A. Methodology If issuing firms use discretionary accruals to shift current earnings from the future, we expect a negative relationship between discretionary accruals around the offering and subsequent earnings changes (e.g.,rangan, 1998; Teoh et al., 1998b; Shivakumar, 2000; Chou et al., 2010). To test this prediction, we run multivariate regression. Following Lie (2005) and Chen et al. (2010), we calculate post-issue operating performance as the difference between return on assets (ROA) in year T+1 and that in year T. The key independent variable is discretionary accruals and we calculate the discretionary accruals in issuing year representing the level of earnings management. Moreover, we add a set of control variables based on the prior studies. The regression model is set as the following specification: ROA, = β + β DA, + β CR, + β SG, + β SIZE, + β BM, + ε, Where, the dependent variable ( ROA i,t ) represents the change of return on asset ratio (return on equity) at year t (Lie, 2005; Chen et al., 2010); DA i,t, is defined as the abnormal accruals during the equity issuance year and is used as a proxy for the extent of earnings management. Other control variables include the current ratio (CR i,t ) which is defined as the ratio of current assets over current liabilities; sales growth (SG i,t ) which measures the percentage increase in corporate sales from year T-1 to year T; SIZE i,t represents firm size and defined as the log of total assets of issue firms and lastly the book-to-market ratio (BM i,t ) 12. Finally, statistical inferences are calculated using the Newey and West (1987) procedure in order to account for possible heteroskedasticity and serial correlation effects in the time series. B. Regression Results To evaluate the ability of earnings management to explain the post-issue firm performance decline after 12 Seasoned equity issues coincide with high valuations in Taggart (1977), Marsh (1982), Asquith and Mullins (1986), Korajczyk, Lucas, and McDonald (1991), Jung, Kim, and Stulz (1996), and Hovakimian, Opler, and Titman (2001). Initial public equity issues coincide with high valuations in Loughran, Ritter, and Rydqvist (1994) and Pagano, Panetta, and Zingales (1998). Repurchases coincide with low valuations in Ikenberry, Lakonishok, and Vermaelen (1995). 17

18 the private placement of equity. This study runs multi-variable regression with dependent variable ΔROA, which is defined as the difference between return on assets (return on equity) in year T+1 and that in year T, key independent variables are discretional accruals (DA1 representing discretionary accruals based on Jones model and DA2 representing discretionary accruals based on modified Jones model) in issuing year. On the other hand, we choose some control variables depend on the previous studies. Rangan (1998) argue that post-issue firm performance declines could be related to growth in sales and growth in assets. Particularly, he argues that issuing firms that invest in projects that generate profits only from year T+2 onward, the increase in assets in year T will induce mechanical declines in earnings in year T+1. In addition, issuing firms experiencing rapid sales growth are likely to attract new entrants into their sectors, which cause the increase in competition and in turn causing issuing firms to experience earnings declines in year T+1. Moreover, Loughran and Ritter (1997) find that issuing firms growing rapidly in sales and capital expenditure in the year of shares offering experience large postissue firm performance declines than slowly growing issuing firms. Therefore, this study includes sales growth (SALE_GROW) as control variables. Variable SALE_GROW is calculated as percentage growth rate in sales from year T-1 to year T. [Insert Table IV here] Table IV shows that the coefficients of DA1 and DA2 are negative and statistically significant at 0.01 level in model specification from (1) to (2). To assess economic significance, this study calculates the one-standard-deviation influence of discretional accruals by multiplying the coefficients of discretional accruals (i.e., DA1 and DA2) by its sample standard deviation. The coefficient estimate implies that a one-standard-deviation increase in DA1 (DA2), which is (0.1383) in this study, is associated with 0.4% (0.029*0.1377) decline in return of assets in year T+1. Thus, discretional accruals are associated with both economically and statistically significant earnings declines in year T+1. In addition, the coefficients on SALE_GROW and BTM are all statistically significant in model specification from (1) to (4). Moreover, a one-standard-deviation increase in sales growth in year T is associated with 0.63% (=0.001*6.2928) decline in return of assets in year T+1; a one-standard-deviation increase in book-tomarket ration in year T implies 0.42% (=0.005*0.8457) increase in return of assets in year T+1. These results suggest that both discretionary accruals and firm growth in year T are associated with earnings declines in the following year. To check the robustness of the results, variable ΔROE is used as alternative measurement of firm 18

19 performance and regression are re-run by using ΔROE as dependent variable. ΔROE is defined as the difference between return on equity in year T+1 and year T. The main results do not change qualitatively. The results in model specification from (3) to (4) show that the coefficients on discretionary accruals (DA1 and DA2) are all negative and statistically significant at 0.01 level. In addition, one standarddeviation increase in discretionary accruals implies 0.84% (0.095*0.0893) declines in return of equity in year T+1. The signs and magnitude of main control variables do not change in model specification from (3) to (4). Rangan (1998) documents that discretionary accruals in the year around the offering are reversed in the following year, thereby explaining a portion of the earnings declines in that year. Further, Rangan (1998) find that the reversal of discretionary accruals is only concentrated in the year following the offering year and earnings changes in the second and third year after the offering year are not related to offeringyear discretionary accruals. We find that the relations between pre-issue earnings management and postissue operating performance are only significant for earnings changes between T+1 and T rather than between T+2 or T+3 and T. Overall, these results imply that discretionary accruals in year T has explanatory power to the decline of post-issue earnings in year T+1. This conclusion is consistent with the results of previous studies investing the relationship between earnings management and seasoned equity offering (Loughran and Ritter, 1997; Rangan, 1998; Teoh et al., 1998b), and the relationship between earnings management and initial public offering (Teoh et al., 1998b). 19

20 VII. Alternative Explanations A. Overinvestment and Post-issue Operating Performance (Fu,2010) Fu (2010) presents an overinvestment explanation for poor firm performance after SEOs. The overinvestment explanation argues that poor post-issuing operating performance is due to the overinvestment of the issuing firms as overinvestment significantly reduces the asset productivity of SEO firms, which leads to poor operating performance. Based on the data, Fu finds that subsequent to the offering, SEO firms invest more heavily than non-issuing control firms. In addition, there is a negative correlation between post-offering investment and operating performance, controlling for investment opportunities and pre-issue performance. Furthermore, he finds evidence that overinvestment significantly reduces the asset productivity of SEO firms. In this study, we test the overinvestment explanation about the issuing shares in PPEs. Similar to the above argument, we argue that the poor post-issuing operating performance of PPEs might due to the overinvestment of issuing firms after PPEs. To test this explanation, we add post-issue investment and an interactive term between post-issue investment and investment opportunities as additional control variables and re-run main regression. Table V shows that the coefficients on I/A and I/A_DUM are not significant at 1% significance level. These results indicate that overinvestment argument does not hold in this study. In addition, the coefficients on discretionary accruals DA1 and DA2 are all negative and statistically significant at 1% significance level, which indicating that earnings management hypothesis still holds after we control overinvestment factor. B. Ownership Structure and Post-issue Operating Performance (Agency problem) Another explanation to post-issuing operating performance decline is related to the potential for increased agency conflict between controlling shareholders and minority shareholders. The agency problem could deteriorate when shareholdings held by controlling shareholders increase due to new shares issued through PPEs. One the one hand, the percentage of shares owned by controlling shareholders could increase if controlling shareholders are involved with new shares purchasing. The increase of shareholdings owned by controlling shareholders could induce the entrenchment effect, which leads to the decrease of firm performance (Claessens et al., 2002). On the other hand, the percentage of shares owned by controlling shareholders could decrease if new shares are sold only to 20

21 other investors. The decrease of shareholdings owned by controlling shareholders could reduce the alignment effect, which also leads to the decline of firm performance (Fan and Wong, 2002). To test the above arguments, we create a dummy representing whether the percentage of shares owned by controlling shareholders increase or decrease after PPEs. In addition, we use difference in share percentage between pre- and post-issue PPEs to represent the change of controlling shareholders. We add both variables as additional control variables separately and re-run main regression. Table VI shows that the coefficients on DIFF_V1 and DUM_DIFF_V1 are not significant at 1% significance level. These results indicate that agency problem argument does not hold in our study. In addition, the coefficients on discretionary accruals DA1 and DA2 are all negative and statistically significant at 1% significance level, which indicating that earnings management hypothesis still holds after we control ownership of controlling shareholders. C. Leverage and Post-issue Operating Performance We know that one advantage of issuing debt is the interest on debt is deductible on the corporation s income return (Miller, 1977), in other words, issuing more debt can reduce the tax paid to tax authorities, which is called as debt tax shield (is the amount of tax saved due to incurring of interest on debt). When firms issue new shares, the level of leverage will be reduced. Thus, the benefit of debt tax shield will be reduced, which leads to the lower earnings. This indicates that after issuing new shares, firm performance will reduce because the debt tax shield enjoyed by the firms will be reduced. To test this argument, we create a dummy representing whether the leverage of the firm increase or decrease after PPEs. In addition, we use difference in leverage between pre- and post-issue PPEs to represent the change of the level of leverage. We add both variables as additional control variables separately and rerun main regression. Table VII shows that the coefficients on DIFF_LEV are not significant and the coefficients on DUM_DIFF_LEV are significant at 10% significance level. These results indicate that agency problem argument does not hold in our study. In addition, the coefficients on discretionary accruals DA1 and DA2 are all negative and statistically significant at 1% significance level, which indicating that earnings management hypothesis still valid after we control leverage effect. 21

22 VIII. Robustness Check In this section, we address several issues that might have impact on the robustness of our results. Specifically, we consider (1) matching methods; (2) alternative operating performance measures; (3) operating performance relative to control firms; (4) the effect of split share structure reform. A. Matching methods We use propensity score matching methods to construct control firms. We first run probit (and logit) model to calculate the probability of issuing new shares through PPE. The probit (and logit model) as specified below: Dum = β + β Issue + β Industry + β SIZE, + β BTM, + β LEV, + β QR1 + β RED, + β SA, + β TR, + β BCHG1, + β LOG(AGE) + ε, (9) Where Dum is 1 if a firm have private placement issue in a certain year and 0 otherwise. The independent variable including Issue year; Industry; SIZE represents firm size and defined as the log of total assets of issue firms; BTM i,t represents book to market ratio and is defined as the ratio of book value of the firm over the market value of the firm; LEV i,t represents leverage and is defined as the ratio of total debt over total assets; QR1 represents stock return in year T-1; RED i,t represents research development expenses over total assets; SA i,t represents sales revenue over total assets; TR i,t represents income taxes divided by pre-tax income; BCHG1 i,t represents the net income change, is defined as the difference of net income between year T and year T-1. LOG(AGE) defined as the log of firm age, that number of years since a firm was incorporated. Table VIII shows the results on probit (and logit) model. Most variables are significant at 5% or 1% significance level indicating these factors are potentially affect whether firms choose to issue shares through PPEs or not. Based on the probabilities we obtained, we use these probabilities as the score of firms and find the closest scores of non-issuing firms to that of issuing firms. We use that non-issuing firms as control group based on propensity score matching method. Untabulated results show the post-issue operating performance change based on new control firms. Similar to the previous results, the operating performance declines after issuing new shares through PPEs. In addition, the relationship between pre-issue earnings management and post-issue operating performance change still negative and statistically significant. [Insert Table VIII here] 22

23 B. Alternative operating performance measures We consider return on equity (ROE), return on sales (ROS), and return on total assets considering taxes (ROTA) as alternative operating performance measures. ROE is defined as the ratio of net income to the value of equity. ROS is defined as the ratio of net income to net sales. ROTA is defined as the ratio of net income and total taxes over the total assets. We recalculate operating performance using these measures and the results about the deteriorates of operating performance following PPEs do not change. In addition, the main regression results about the negative association between earnings management and post-issue operating firm performance do not change. [Insert Table XIIII here] C. Operating performance relative to control firms We further use the relative operating performance checking the post-issue operating performance of issuing firms of PPEs. We calculate the relative operating performance by subtracting the control firms operating performance as the new dependent variable. We then re-run the main regression and find that the main results do not change. Particularly, the coefficients on DA1 and DA2 are both negative and significant at 1% significance level. Thus, these results confirm the earnings management hypothesis. [Insert Table X here] D. The effect of split share structure reform To examine the potential effect of split share structure reform launched in 2005, we split the sample into two categories, one is before and the other is after split share structure reform. The results are showed in Table XI. In both periods, there is still evidence about the earnings management hypothesis. However, the strength of the evidence is weaker in the latter period. It appears that the split share structure reform made some progress toward reducing the influence of earnings management around the PPEs. [Insert Table XI here] 23

24 IX. Discussion and Conclusion Previous studies document the operating performance around seasoned equity offering and initial public offerings. However, little literature investigates the operating performance around private placement of equities. In addition, most existing studies investigate the relationship between earnings management and seasoned equity offering or the relationship between earnings management and initial public offering. However, few investigate the relationship between earnings management and private placement of equity. This prompts this study to fill in this gap in literature. To achieve the above objective, this study focuses on Chinese listed companies as China provides an ideal setting to test the relationship between earnings management and private placement. This study finds that (1) there is a persistent long-term post-issue operating performance decline. This finding is consistent with previous studies. (2) there is a significant earnings management upwards prior to PPEs, especially in the year of private placement. This result is similar to the research about rights offerings, public equity offerings and initial public offerings in the listed company; (3) there is a negative and statistically significant association between pre-issue discretionary accruals and post-issue operating performance, which indicates that companies inflate their earnings through discretionary accruals before and in the year of private placement of equity, which causes the decline of earnings after private placement of equity. Overall, these results suggest that companies tend to manipulate earnings through accruals-based earnings management before and in the year of private placement of equity. The contribution of this study is as follows. Firstly, this research add literature in operating performance and earnings management around corporate events. Secondly, this study extends the existing study on earnings management to the international literature. Finally, the findings of this study have policy implication in China as it urges the regulator in China to strictly supervise, monitor the earnings management behaviour in China in order to protect shareholders' interests, particularly small investors interests. 24

25 Appendix A Table I. Private placement of equities in Chinese stock market from 2006 to 2016 The sample consists of 1,627 private placement of equities over the period of from CSMAR database. The distribution of the sample is reported in Panel A by industry code, and in Panel B by PPEs calendar year, and in Panel C by PPEs characteristics. Panel A : Industry Distribution of Private Placement Industry No. of Offerings 100% Cumulative percentage Agriculture % 1.91% Mining % 5.35% Manufacture % 61.95% Production % 67.49% Construction % 70.07% Wholesale % 77.44% Transport % 81.07% Catering/Lodging % 81.62% Telecom % 84.51% Financial % 86.79% Real estate % 93.67% Tenancy % 94.84% Research % 94.96% Environment/Public facilities % 95.57% Services % 95.64% Education % 95.76% Health % 95.76% Media % 97.23% Conglomerates % % Total 1, % % Panel B: Descriptive Statistics on PPEs over the period of Year No. of Percentage relative to total Percentage relative to total Cumulative percentage Offerings number of firms number of offerings % 2.83% 2.83% % 7.93% 10.76% % 5.84% 16.59% % 6.27% 22.86% % 6.76% 29.63% % 7.38% 37.00% % 6.39% 43.39% % 9.83% 53.23% % 12.78% 66.01% % 17.70% 83.71% % 16.29% % Total 1, % % % Panel C: Descriptive Statistics on the Characteristics of PPEs over the period of Year PPE Issue Size Percentage of PPEs in Percentage relative to total Cumulative percentage (million $) refinancing refinancing , % 1.95% 1.95% , % 5.71% 7.65% , % 4.82% 12.48% , % 6.39% 18.86% , % 7.74% 26.60% , % 9.71% 36.31% , % 6.20% 42.51% , % 7.87% 50.38% , % 11.54% 61.92% , % 17.11% 79.04% , % 20.96% % Total 658, % % % 25

26 Table II. Operating performance around the private placement of equities This table reports the operating performance around the private placement of equity both for issuing firms (i.e., PPE firms) and control firms. Return on assets (ROA) and return on equity (ROE) are used as the measurement of operating performance. T represents issuing year. ROA represents return on assets and is defined as the ratio of net income divided by fiscal year-end total assets. ROE represents return on equity and is defined as the ratio of net income divided by fiscal/calendar year-end market value of equity. The control firms are matched on the basis of year, industry, size and operating performance the same issuing year of private placement of equity. Both mean and median of yearly operating performance are shown. T-tests are used to test mean difference between issuing firms and control firms. Mann-Whitney tests are used to test median difference between issuing firms and control firms. ***, **, * represents significance level at 1%, 5%, and 10% significance level respectively. Obs. T-3 T-2 T-1 T T+1 T+2 T+3 ROA1 ROA2 ROA3 ROA PPE firms Mean Median % negative % 10.07% 9.33% 4.10% 7.09% 9.95% 8.58% 57.71% 62.94% 65.05% Control firms Mean Median % negative % 8.71% 10.07% 10.57% 11.69% 13.18% 13.68% 53.86% 56.47% 58.58% T-test (p-value) *** *** *** *** Mann-Whitney test (p-value) *** *** *** Obs. T-3 T-2 T-1 T T+1 T+2 T+3 ROE1 ROE2 ROE3 ROE PPE firms Mean Median % negative % 11.44% 9.58% 4.10% 6.97% 9.95% 8.46% 55.22% 61.32% 60.45% Control firms Mean Median % negative % 8.33% 9.70% 10.70% 10.95% 12.56% 13.06% 54.23% 56.22% 59.20% T-test (p-value) *** *** *** *** Mann-Whitney test (p-value) *** ***

27 Table III. Distribution of accruals-based earnings management around PPE This table shows descriptive statistics on discretional accruals around the private placement of equity. T-tests in Panel A are used to test whether the mean of discretional accruals in issuing year is statistically different from zero. Mann-Whitney tests in Panel B are sued to test whether the median of discretional accruals in issuing year is statistically different from zero. T-tests in Panel B are used to test whether the mean of discretional accruals is statistically different from zero in years around the private placement of equity. Mann- Whitney tests in Panel B are used to test whether the median of discretional accruals is statistically different from zero in years around the private placement of equity. DA1 represents discretional accruals based on Jones Model, while DA2 represents discretional accruals based on Modified Jones Model. ***, **, * represents significance level at 1%, 5%, and 10% significance level respectively. Panel A: Descriptive statistics for discretional accruals in issuing year Variable Obs. Mean Median Std. Dev. Min Max T-test Mann- Whitney tests DA1(Jones Model) *** *** DA2(Modified Jones Model) *** *** Panel B: Discretional accruals change around private placement offering DA1: Jones model Variable Obs. T-3 T-2 T-1 T T+1 T+2 T+3 Mean T-test *** *** *** *** Median Wilcoxon test *** *** *** *** DA2: Modify Jones model Variable Obs. T-3 T-2 T-1 T T+1 T+2 T+3 Mean T-test *** *** *** *** Median Wilcoxon test *** *** *** *** 13 The test for differences between the two groups is performed using the Wilcoxon two-sample signed rank test, which assumes that the observations are independent. 27

28 Table IV. Regression on the relationship between accrual-based earnings management and postissue operating performance change (Basic Model) This table reports the results from cross-sectional regression of ΔROA (and ΔROE) on discretionary accruals and other control variables. The sample period of from year 2006 to year The statistical inference is based on standard errors estimated following the approach proposed in Newey and West (1987), which is robust to both heteroskedasticity and serial correlation. The dependent variable ΔROA (ΔROE) in model specifications, which represents the post-issue operating performance change, is defined as the difference between return on assets (return on equity) in year T+1 and that in year T. The key independent variable DA1 (DA2) represents the discretionary accruals calculated by using Jones model (modified Jones model). CUR_RATIO represents financial ratio that measures whether or not a firm has enough resources to pay its debts over the next 12 months. It is defined as the ratio of current assets over current liability. SALE_GROW represents the sale growth rate and is defined as the percentage growth rate in sales from year T-1 to year T. SIZE represents firm size and defined as the log of total assets of issue firms. BTM represents book to market ratio and is defined as the ratio of book value of the firm over the market value of the firm. Robust t-statistics are reported in parentheses, ***, **, and * denotes the significance level at 1%, 5% and 10% respectively. VARIABLES ΔROA ΔROA ΔROE ΔROE DA *** *** (-2.794) (-3.464) DA *** *** (-2.748) (-3.430) CUR_RATIO ** 0.005** (-0.192) (-0.238) (2.172) (2.120) SALE_GROW *** *** *** *** (-3.178) (-3.144) (-3.350) (-3.311) SIZE (-0.921) (-0.900) (-0.724) (-0.704) BTM 0.005*** 0.005*** (2.890) (2.852) (1.542) (1.507) Constant (0.619) (0.606) (0.419) (0.406) Observations R-squared Adj. R-squared

29 Table V. Regression on the relationship between accrual-based earnings management and postissue operating performance change (Overinvestment) This table reports the results from cross-sectional regression of ΔROA (and ΔROE) on discretionary accruals and other control variables. The sample period of from year 2006 to year The statistical inference is based on standard errors estimated following the approach proposed in Newey and West (1987), which is robust to both heteroskedasticity and serial correlation. The dependent variable ΔROA (ΔROE) in model specifications, which represents the post-issue operating performance change, is defined as the difference between return on assets (return on equity) in year T+1 and that in year T. The key independent variable DA1 (DA2) represents the discretionary accruals calculated by using Jones model (modified Jones model). CUR_RATIO represents financial ratio that measures whether or not a firm has enough resources to pay its debts over the next 12 months. It is defined as the ratio of current assets over current liability. LEVERAGE represents leverage and is defined as the ratio of total debt over total assets. SALE_GROW represents the sale growth rate and is defined as the percentage growth rate in sales from year T-1 to year T. I/A represents the investment and is defined as ratio of capital expenditure over total assets in year T. SIZE represents firm size and defined as the log of total assets of issue firms. DUM is a dummy variable and represent investment opportunities and it is calculated based on market to book ratio. It is defined as one if a firm s market to book ratio is higher than the median, zero otherwise. Robust t-statistics are reported in parentheses, ***, **, and * denotes the significance level at 1%, 5% and 10% respectively. VARIABLES ΔROA ΔROA ΔROE ΔROE DA *** *** (-2.812) (-3.435) DA *** *** (-2.759) (-3.397) CUR_RATIO ** 0.005** (-0.110) (-0.155) (2.167) (2.115) SALE_GROW *** *** *** *** (-3.096) (-3.064) (-3.291) (-3.253) SIZE (-0.922) (-0.898) (-0.764) (-0.740) BTM 0.004*** 0.004** (2.595) (2.567) (1.331) (1.305) I/A (0.047) (0.011) (0.607) (0.550) I/A*DUM (-0.613) (-0.577) (-0.751) (-0.700) Constant (0.642) (0.626) (0.462) (0.446) Observations R-squared Adj. R-squared

30 Table VI. Regression on the relationship between accrual-based earnings management and postissue operating performance change (Agency Problem) This table reports the results from cross-sectional regression of ΔROA (and ΔROE) on discretionary accruals and other control variables. The sample period of from year 2006 to year The statistical inference is based on standard errors estimated following the approach proposed in Newey and West (1987), which is robust to both heteroskedasticity and serial correlation. The dependent variable ΔROA (ΔROE) in model specifications, which represents the post-issue operating performance change, is defined as the difference between return on assets (return on equity) in year T+1 and that in year T. The key independent variable DA1 (DA2) represents the discretionary accruals calculated by using Jones model (modified Jones model). CUR_RATIO represents financial ratio that measures whether or not a firm has enough resources to pay its debts over the next 12 months. It is defined as the ratio of current assets over current liability. SALE_GROW represents the sale growth rate and is defined as the percentage growth rate in sales from year T-1 to year T. I/A represents the investment and is defined as ratio of capital expenditure over total assets in year T. SIZE represents firm size and defined as the log of total assets of issue firms. DUM is a dummy variable and represent investment opportunities and it is calculated based on market to book ratio. It is defined as one if a firm s market to book ratio is higher than the median, zero otherwise. DIFF_V1 represents the change of the shareholdings owned by controlling shareholders between pre- and post-issuing. DUM_DIFF_V1 is a dummy variable and represents whether the change of the shareholdings owned by controlling shareholders between pre- and post-issuing is larger than zero or not. It takes one if it is larger than zero and zero otherwise. Robust t-statistics are reported in parentheses, ***, **, and * denotes the significance level at 1%, 5% and 10% respectively. VARIABLES ΔROA ΔROA ΔROA ΔROA DA *** *** (-2.789) (-3.402) DA *** *** (-2.727) (-3.356) CUR_RATIO * 0.005* (-0.278) (-0.320) (1.884) (1.835) SALE_GROW *** *** *** *** (-2.632) (-2.609) (-2.925) (-2.897) SIZE (-1.020) (-0.994) (-0.866) (-0.839) BTM 0.005*** 0.005*** (2.719) (2.689) (1.450) (1.424) I/A (0.057) (0.023) (0.635) (0.583) I/A*DUM (-0.602) (-0.566) (-0.746) (-0.697) DIFF_V (0.182) (0.182) (0.411) (0.413) DUM_DIFF_V (-1.076) (-1.059) (-1.360) (-1.341) Constant (0.796) (0.777) (0.627) (0.607) Observations R-squared Adj. R-squared

31 Table VII. Regression on the relationship between accrual-based earnings management and postissue operating performance change (Leverage Effect) This table reports the results from cross-sectional regression of ΔROA (and ΔROE) on discretionary accruals and other control variables. The sample period of from year 2006 to year The statistical inference is based on standard errors estimated following the approach proposed in Newey and West (1987), which is robust to both heteroskedasticity and serial correlation. The dependent variable ΔROA (ΔROE) in model specifications, which represents the post-issue operating performance change, is defined as the difference between return on assets (return on equity) in year T+1 and that in year T. The key independent variable DA1 (DA2) represents the discretionary accruals calculated by using Jones model (modified Jones model). CUR_RATIO represents financial ratio that measures whether or not a firm has enough resources to pay its debts over the next 12 months. It is defined as the ratio of current assets over current liability. SALE_GROW represents the sale growth rate and is defined as the percentage growth rate in sales from year T-1 to year T. I/A represents the investment and is defined as ratio of capital expenditure over total assets in year T. SIZE represents firm size and defined as the log of total assets of issue firms. DUM is a dummy variable and represent investment opportunities and it is calculated based on market to book ratio. It is defined as one if a firm s market to book ratio is higher than the median, zero otherwise. DIFF_V1 represents the change of the shareholdings owned by controlling shareholders between pre- and post-issuing. DUM_DIFF_V1 is a dummy variable and represents whether the change of the shareholdings owned by controlling shareholders between pre- and post-issuing is larger than zero or not. It takes one if it is larger than zero and zero otherwise. DIFF_LEV represents the change of leverage between pre- and post-issuing. Leverage is defined as the ratio of total debt over total assets. DUM_DIFF_V1 is a dummy variable and represents whether the change of the leverage between pre- and post-issuing is larger than zero or not. It takes one if it is larger than zero and zero otherwise. Robust t-statistics are reported in parentheses, ***, **, and * denotes the significance level at 1%, 5% and 10% respectively. VARIABLES ΔROA ΔROA ΔROA ΔROA ΔROA ΔROA ΔROA ΔROA DA *** *** *** *** (-2.774) (-2.785) (-2.794) (-2.801) DA *** *** *** *** (-2.718) (-2.707) (-2.735) (-2.720) CUR_RATIO (-0.213) (-0.147) (-0.289) (-0.216) (-0.256) (-0.194) (-0.332) (-0.263) SALE_GROW ** ** *** *** ** ** ** *** (-2.529) (-2.543) (-2.585) (-2.607) (-2.505) (-2.521) (-2.558) (-2.583) SIZE (-1.024) (-0.655) (-1.090) (-0.722) (-1.006) (-0.638) (-1.070) (-0.704) BTM 0.004*** 0.004** 0.005*** 0.005*** 0.004** 0.004** 0.005*** 0.005*** (2.603) (2.577) (2.701) (2.658) (2.574) (2.550) (2.671) (2.630) I/A (-0.032) (-0.029) (0.022) (0.016) (-0.068) (-0.059) (-0.014) (-0.015) I/A*DUM (-0.557) (-0.568) (-0.571) (-0.581) (-0.520) (-0.534) (-0.534) (-0.546) DUM_DIFF_V (-1.426) (-1.150) (-1.399) (-1.131) DIFF_V (-0.929) (-0.700) (-0.905) (-0.686) DUM_DIFF_LEV * * * * (-1.808) (-1.699) (-1.759) (-1.651) DIFF_LEV (0.225) (0.253) (0.247) (0.274) Constant (0.760) (0.432) (0.868) (0.531) (0.750) (0.421) (0.855) (0.518) Observations R-squared Adj. R-squared

32 Table VIII. Probit/ logit model on the determinants of Private Placement of Equity This table reports the results from logit model of private placement decision. The data set is 1008 private equity placement and 10,058 non-issue firms in the period of year 2006 to The dependent variable is 1 if a firm have private placement issue in a certain year and 0 otherwise. The independent variable including ISSUE YEAR; INDUSTRY; LEVERAGE represents leverage and is defined as the ratio of total debt over total assets; QR1 represents stock return in year T-1; RESEARCH EXPENSES represents research development expenses over total assets; BTM represents book to market ratio and is defined as the ratio of book value of the firm over the market value of the firm; SALE/ASSETS represents sales revenue over total assets; TAX RATE represents income taxes divided by pretax income; BCHG1 represents the net income change, is defined as the difference of net income between year T and year T-1. SIZE represents firm size and defined as the log of total assets of issue firms. LOG(AGE) defined as the log of firm age, that number of years since the firm was incorporated. Robust t-statistics are reported in parentheses, ***, **, and * denotes the significance level at 1%, 5% and 10% respectively. PPEs vs. Non-PPEs (PPEs =1; Non-PPEs=0) PPEs vs. Non-PPEs (PPEs =1; Non-PPEs=0) Probit model Logit model VARIABLES Coefficient P-value Coefficient P-value SIZE 0.297*** *** BTM *** *** LEVERAGE *** *** QR *** *** RESEARCH EXPENSES *** *** SALE/ASSETS * * TAX RATE BCHG *** *** LOG(AGE) Constant *** *** Year fixed effects Yes Yes Industry fixed effects Yes Yes Observations R-squared Prob > chi

33 Table IX. Regression of earnings management on post-issuing operating performance change (return on sales and return on total assets) This table reports the results from cross-sectional regression of ΔROS and ΔROTA on discretionary accruals and other control variables. The sample period of from year 2006 to year The statistical inference is based on standard errors estimated following the approach proposed in Newey and West (1987), which is robust to both heteroskedasticity and serial correlation. The dependent variable ΔROS (ΔROTA) in model specifications, which represents the post-issue operating performance change, is defined as the difference between return on sales (return on total assets) in year T+1 and that in year T. The key independent variable DA1 (DA2) represents the discretionary accruals calculated by using Jones model (modified Jones model). CUR_RATIO represents financial ratio that measures whether or not a firm has enough resources to pay its debts over the next 12 months. It is defined as the ratio of current assets over current liability. LEVERAGE represents leverage and is defined as the ratio of total debt over total assets. SALE_GROW represents the sale growth rate and is defined as the percentage growth rate in sales from year T-1 to year T. CAPEX represents the capital expenditure rate and is defined as the percentage growth rate in assets from year T-1 to year T. SIZE represents firm size and defined as the log of total assets of issue firms. BTM represents book to market ratio and is defined as the ratio of book value of the firm over the market value of the firm. Robust t-statistics are reported in parentheses, ***, **, and * denotes the significance level at 1%, 5% and 10% respectively. VARIABLES ΔROS ΔROS ΔROTA ΔROTA DA ** ** (-2.035) (-2.013) DA ** * (-2.121) (-1.924) CUR_RATIO (-0.466) (-0.481) (-0.692) (-0.733) SALE_GROW * * *** *** (-1.738) (-1.713) (-3.405) (-3.383) SIZE (-0.598) (-0.588) (-1.097) (-1.080) BTM *** 0.007*** (1.385) (1.350) (3.409) (3.382) Constant (0.464) (0.460) (0.798) (0.785) Observations R-squared Adj. R-squared

34 Table X. Regression of pre-issue earnings management on post-issue operating performance change (relative to control firms) This table shows results of regression of post-issue firm performance change (both ROA and ROE) relative to matched firms on discretionary accruals and other control variables. The sample period of from year 2006 to year The statistical inference is based on standard errors estimated following the approach proposed in Newey and West (1987), which is robust to both heteroskedasticity and serial correlation. The dependent variable ΔROA_M (ΔROE_M) in model specifications, which represents the post-issue operating performance change relative to matched firms, is defined as the difference between return on assets (return on equity) in year T+1 and that in year T relative to matched firms. The key independent variable DA1 (DA2) represents the discretionary accruals calculated by using Jones model (modified Jones model). CUR_RATIO represents financial ratio that measures whether or not a firm has enough resources to pay its debts over the next 12 months. It is defined as the ratio of current assets over current liability. LEVERAGE represents leverage and is defined as the ratio of total debt over total assets. SALE_GROW represents the sale growth rate and is defined as the percentage growth rate in sales from year T-1 to year T. CAPEX represents the capital expenditure rate and is defined as the percentage growth rate in assets from year T-1 to year T. SIZE represents firm size and defined as the log of total assets of issue firms. BTM represents book to market ratio and is defined as the ratio of book value of the firm over the market value of the firm. Robust t-statistics are reported in parentheses, ***, **, and * denotes the significance level at 1%, 5% and 10% respectively. VARIABLES ΔROA_M ΔROA_M ΔROE_M ΔROE_M DA *** (-1.379) (-2.588) DA ** (-1.246) (-2.494) CUR_RATIO (-1.023) (-1.064) (0.413) (0.371) SALE_GROW ** ** (-0.805) (-0.785) (-2.378) (-2.345) SIZE (0.274) (0.287) (0.358) (0.374) BM (-0.566) (-0.581) (-0.247) (-0.278) Constant (-0.179) (-0.189) (-0.365) (-0.374) Observations R-squared Adj. R-squared

35 Table XI. Regression results on the relationship between pre-issue earnings management and post-issue operating performance before and after split share structure reform. This table reports the results from cross-sectional regression of ΔROA (and ΔROE) on discretionary accruals and other control variables. The sample period of from year 2006 to year 2010 (during the split share structure reform); and year 2011 to year 2016 (completed the split share structure reform). The statistical inference is based on standard errors estimated following the approach proposed in Newey and West (1987), which is robust to both heteroskedasticity and serial correlation. The dependent variable ΔROA (ΔROE), which represents the post-issue operating performance change, is defined as the difference between return on assets (return on equity) in year T+1 and that in year T. The key independent variable DA1 (DA2) represents the discretionary accruals calculated by using Jones model (modified Jones model). CUR_RATIO represents financial ratio that measures whether or not a firm has enough resources to pay its debts over the next 12 months. It is defined as the ratio of current assets over current liability. LEVERAGE represents leverage and is defined as the ratio of total debt over total assets. SALE_GROW represents the sale growth rate and is defined as the percentage growth rate in sales from year T-1 to year T. CAPEX represents the capital expenditure rate and is defined as the percentage growth rate in assets from year T-1 to year T. SIZE represents firm size and defined as the log of total assets of issue firms. BTM represents book to market ratio and is defined as the ratio of book value of the firm over the market value of the firm. Robust t-statistics are reported in parentheses, ***, **, and * denotes the significance level at 1%, 5% and 10% respectively. VARIABLES ΔROA ΔROA ΔROE ΔROE ΔROA ΔROA ΔROE ΔROE Period DA *** *** * (-2.908) (-2.934) (-1.061) (-1.857) DA *** *** * (-2.911) (-2.957) (-1.003) (-1.802) CUR_RATIO ** 0.010** (1.412) (1.396) (2.511) (2.497) (-1.175) (-1.225) (0.774) (0.707) SALE_GROW ** ** ** ** ** ** ** ** (-2.123) (-2.101) (-2.383) (-2.368) (-2.252) (-2.234) (-2.365) (-2.321) SIZE * * (0.064) (0.083) (-0.207) (-0.191) (-1.803) (-1.793) (-1.088) (-1.077) BM 0.007*** 0.007*** 0.018** 0.018** 0.005** 0.005** (2.767) (2.742) (2.435) (2.413) (2.236) (2.216) (0.550) (0.533) Constant (-0.405) (-0.418) (-0.122) (-0.132) (1.611) (1.606) (0.905) (0.900) Observations R-squared Adj. R-squared

36 Appendix B Figure 1. The development of PPEs in Chinses stock market 36

37 Figure 2. ROA before and after private placement of equity 37

38 Figure 3. Mean (Median) discretional accruals around private placement of equity 38