Why Don t Share Issue Privatizations Improve Profitability in China?

Transcription

1 Why Don t Share Issue Privatizations Improve Profitability in China? Bo Li School of Business, Shantou University (boli@stu.edu.cn) William L. Megginson Professor and Price Chair in Finance, University of Oklahoma Saudi Aramco Chair in Finance, King Fahd University of Petroleum and Minerals (wmegginson@ou.edu) Zhe Shen School of Management, Xiamen University (z.shen@xmu.edu.cn) Qian Sun Department of Finance, School of Management, Fudan University (sunqian@fudan.edu.cn) July 29, 2017 Abstract We document significant post-share issue privatization (SIP) increases in profitability of divested Chinese state-owned enterprises (SOEs), once the negative IPO listing effect is accounted for, contradicting previous studies showing profitability declines after privatization. We find robust evidence of significant post-listing declines in profitability for both SOEs and privately-owned firms (POs), and evidence of a positive relationship between the degree of privatization and the profitability improvement for SIP firms. Firms privatized without public listing out-perform matched SOEs that remain state controlled. Further analysis suggests declining profitability is more pronounced for those SIP firms more engaged in earnings management and agency problems. JEL Classification: G32, G38, G15 Keywords: Privatization, International financial markets, Government policy and regulation Please address correspondence to: William L. Megginson Price College of Business 307 West Brooks, 205A Adams Hall The University of Oklahoma Noman, OK Tel: (405) ; Fax: (405) wmegginson@ou.edu 1

2 Why Don t Share Issue Privatizations Improve Profitability in China? Abstract We document significant post-share issue privatization (SIP) increases in profitability of divested Chinese state-owned enterprises (SOEs), once the negative IPO listing effect is accounted for, contradicting previous studies showing profitability declines after privatization. We find robust evidence of significant post-listing declines in profitability for both SOEs and privately-owned firms (POs), and evidence of a positive relationship between the degree of privatization and the profitability improvement for SIP firms. Firms privatized without public listing out-perform matched SOEs that remain state controlled. Further analysis suggests declining profitability is more pronounced for those SIP firms more engaged in earnings management and agency problems. JEL Classification: G32, G38, G15 Keywords: Privatization, International financial markets, Government policy and regulation The paper was previously circulated under the title Do Share Issue Privatizations Really Improve Firm Performance in China?. We wish to thank Wolfgang Bessler, Jerry Cao, Andrea Carosi, William Mingyan Cheung, Jerry Coakley, Joseph P.H. Fan, Veljko Fotak, Michael Firth, Edith Ginglinger, C.Y. Hwang, Patrick Jaslowitzer, Ping Jiang, Woochan Kim, Tiecheng Leng, Li Liao, Bibo Liu, Meijun Qian, Marc Steffen Rapp, Christo Shemtov, Wilson Tong, Oscar Varela, Hao Wang, Steven Wang, Donghui Wu, Yexiao Xu, Chengxi Yin, Bohui Zhang, Chu Zhang and seminar participants at Tsinghua University, Fudan University, Université Paris Dauphine, Shanghai University of Finance and Economics, Renmin University of China, Zhejiang University, University of Oklahoma, Xiamen University, King Fahd University of Petroleum and Minerals, Peking University, the 2015 Indonesia Financial Management Association Conference [keynote speech], the 2015 Auckland Finance Meeting, the 2016 FMA European Conference (Helsinki), the 2016 INFINITI Conference (Dublin), the 2016 EFMA Conference (Basel), the 2016 FMA Asia/Pacific (Sydney), the 2016 World Finance Conference New York, the 2017 Asian Finance Association Conference (Seoul), and the 2017 China Financial Research Conference (Beijing) for their helpful comments. 2

3 Why Don t Share Issue Privatizations Improve Profitability in China? The privatization of state-owned enterprises (SOEs) by governments has become a global phenomenon since its introduction by Margaret Thatcher s British government during the 1980s. The cumulative proceeds raised through privatization sales by governments selling their SOE shareholdings to private investors or by SOEs themselves selling new primary share issues reached $3.50 trillion (Megginson, 2017) in As the largest economy in transition, China has the largest privatization program (Gao and Megginson, 2016). Share issue privatizations () have been an important method used by Chinese government to (partially) privatize its SOEs. In fact, the Chinese stock market was initially set up for listing firms privatized through public share offering and, until 1998, almost all listed firms were. Today, about half of the 3000 firms listed on the Shanghai Stock Exchange (SSE) and the Shenzhen Stock Exchange (SZSE) are SIP firms. Besides being large in scale, China s also differ in that the vast majority has been primary, capital-raising share offerings by SOEs themselves, rather than secondary share offerings where sale proceeds flowed to the divesting government. This was supposed to help Chinese SIP firms raise capital to smooth their transitions to private ownership, but the existing empirical research examining Chinese finds either no profitability improvements or even outright declines in profitability for privatized companies. 1 This is puzzling as it is inconsistent with SIP studies elsewhere, which typically document evidence of significant profitability improvements after privatization. 2 Some authors suggest that the fact that the Chinese government still remains the controlling shareholder for most of its SIP firms could be the reason behind the profitability decline. Sun and Tong (2003) and others find that post-ipo profitability is negatively related to the government s retained ownership in the SIP firms. However, if partial privatization is better than no privatization, we should still see some improvement in profitability. In fact, Gupta (2005) documents evidence of profitability improvements for a sample of Indian SIP firms with an average of less than 10% of shares privatized. If the were representative of the whole privatization program in China, then it would also be difficult to understand why the Chinese government keeps privatizing its SOEs while 1 Aharony, Lee and Wong (2000) find that ROA for the median firm tends to peak in the IPO year and decline thereafter for a sample of 83 Chinese SOEs which issue shares to foreign investors in the period. Sun and Tong (2003) find that several profitability measures do not improve after 634 SOEs list their shares on domestic stock exchanges during the period. Subsequent studies that confirm these findings include Wei, Varela, D Souza and Hassan (2003), Wang, Xu and Zhu (2004), Wang (2005), Jia, Sun and Tong (2005), Chen, Firth and Rui (2006), Fan, Wong and Zhang (2007), Jiang, Yue and Zhao (2009). 2 See the studies summarized in Megginson and Netter (2001), Djankov and Murrell (2002), and Estrin, Hanousek, Kòcenda and Svejnar (2009). 3

4 the privatization does not improve profitability. 3 Dong, Putterman and Unel (2006) examine the impact of privatization on performance improvement using 165 unlisted firms located in Nanjing and its outskirts and find evidence of improvement in profitability for these firms. Gan, Guo and Xu (2017) use survey data to show that privatization via management buyouts (MBOs) improves firm profitability in China. These findings suggest that privatization without listing may improve profitability. Hence, we argue in this study that the underperformance of SIP firms in profitability is due to a negative listing effect. Indeed, there is a growing literature documenting a negative listing effect. For example, Jain and Kini (1994) find that firms show disappointing operating performance, as measured by EBIT/total assets and operating cash flows/total assets, after their IPOs. Mikkelson, Partch and Shah (1997) find that worsening agency problems can account for declining operating performance after IPOs. Teoh, Welch and Wong (1998) find that earnings management is responsible for the post-ipo decline in operating performance and, more generally, the long-run underperformance of US-listed IPOs. Pastor, Taylor and Veronesi (2009) find that managers can time their company s IPO when ROE reaches its peak. It is very possible that agency problems and earnings management associated with listing are more serious in China than elsewhere, and lead to a more negative listing effect as the overall legal and financial environment and the market infrastructure is poor (Allen, Qian, and Qian, 2005). On the other hand, the fact that the government remains the controlling shareholder and exercises that control without challenge after the listing is another distinctive feature of China s SIP program. In fact, the government is not expected to transfer its control of these firms in the foreseeable future (Gan, Guo and Xu, 2017). Such determinedly partial privatization may create more serious agency problems for listed firms and dampen the positive privatization effect to some extent. A more serious negative listing effect combined with a less positive partial privatization effect is likely to produce an overall negative SIP effect on post-ipo profitability in China. Using the National Bureau of Statistics (NBS thereafter) database which contains financial information for all manufacturing firms with annual sales of more than RMB 5 million in China from 1998 to 2013 and the China Stock Market and Accounting Research (CSMAR) database, we identify 498 SIP firms that go through privatization via issuing shares to the public over the period, 4,895 privatized SOEs (PSOEs thereafter) that go through privatization using means other than issuing shares to the public, 442,725 unlisted SOEs that do not go public or experience any form of partial privatization over the sample period, 1,079 privately-owned firms (POs thereafter) that go public over the sample period, and 3 Chinese leaders decided in the Third Plenum of the 18th Communist Party Congress to accelerate the privatization of SOEs. 4

5 2,726, 702 unlisted privately-owned firms (UPOs thereafter) that remain unlisted throughout the sample period. Based on these five different firm groups, we construct a number of matched samples to disentangle the privatization effect from the listing effect and examine whether the positive privatization effect is dominated by the negative listing effect. We start our empirical analysis by first examining whether the overall SIP effect on profitability is negative or insignificant, as documented previously in the literature. Following previous studies such as Megginson, Nash and Randerborgh (1994), we compare profitability three years before and after the SIP event for 498 firms which complete during the period Consistent with previous studies, we initially find that profitability declines significantly relative to the pre- SIP level. The median decline in ROS (EBIT/Sales) is 2.87 (3.84) percentage points, from a median 11.39% (14.06%) before sale to 7.44% (9.02%) post-ipo. Even after controlling for the market-wide impact which has nothing to do with the SIP event using a matched sample of SIP firms and their otherwise comparable unlisted SOEs and a difference-in-differences (DID) approach used by Jiang, Yue and Zhao (2009), we still find no evidence of profitability for SIP firms. The median decline in ROS (EBIT/Sales) of SIP firms is 0.14 (0.77) percentage points more for than for unlisted SOEs post-ipo, but statistically insignificant. These findings are consistent with previous studies in the empirical literature. Having confirmed that the overall SIP effect on profitability is non-positive, we move on to examine whether there is a positive privatization effect on profitability. For 4,701 out of 4895 PSOEs that went through privatization (or partial privatization) without listing during the period , we find their unlisted comparables. Our DID analysis with this large matched sample provides strong evidence of a significantly positive privatization effect on profitability. The median improvement in ROS (EBIT/Sales) of PSOEs relative to SOEs is 1.15 (1.35) percentage points. This clearly indicates that privatization without listing improves profitability, which is consistent with Dong, Putterman and Unel (2006) and Gan, Guo, and Xu (2017) albeit our sample is much larger than theirs. Further analysis reveals that profitability improvement in the post-privatization period is associated with the decline in the percentage of state ownership, that indicating that the positive privatization effect tends to increase with the extent of partial privatization. We then examine whether there is a negative listing effect on profitability. We measure the listing effect as the profitability difference before and after the SIP year for SIP firms benchmarked against the profitability change over the same period for comparable PSOEs. It is reasonable to argue that the positive privatization effect may be specific to PSOEs but not applicable to SIP firms as SIP firms are much larger, more profitable before listing/partial privatization, and with more government ownership after the listing/partial privatization. 4 We alleviate this concern by choosing comparable PSOEs based not only on 4 See Tables 1 and 2 for the summary statistics. 5

6 industry, sales revenues, pre-listing ROS, but also on change in government ownership. Using the similar DID analysis, we find robust evidence of profitability declines in the sense that 142 SIP firms underperform their comparable PSOEs by 3.78 (4.50) percentage points in ROS (EBIT/Sales). This finding shows that the privatization effect is positive while the listing effect is negative as listing is the main, if not the only, difference between these matched SIP firms and PSOEs. We also show that the negative listing effect is common to POs. The median decline for relative change in ROS (EBIT/Sales) is 3.27 (4.41) percentage points for a matched sample of 863 POs and their comparable UPOs. These findings strongly confirm the existence of a negative listing effect in China. Further regression analysis suggests that the listing effect is negatively related to discretionary accruals in the pre-ipo period and related-party transaction (RPT) volume in the post-ipo period, consistent with previous studies such as Mikkelson, Partch, and Shah (1997) and Teoh, Welch and Wong (1998), and Cheung, Jing, Lu, Rau, and Stouraitis (2009) indicating that earnings management (EM) for IPOs and the worsening agency problem associated with listing contribute to the negative listing effect. Finally, we examine whether privatization improves profitability for SIP firms after controlling for the negative listing effect. We match a group of 237 SIP firms to their comparable listed PO firms so that the major difference between the two groups is privatization. We further control for EM and RPT to account for possible differences in earnings management and related party transactions between SOEs and POs, and the degree of privatization around the IPO period in the DID analysis. We find clear evidence that SIP firms outperform their private counterparts in ROS (EBIT/Sales), by 0.89 (1.64) percentage points. Further analysis suggests that EM and RPT are more pronounced in Chinese provinces where the legal and market infrastructure is poorer, which is in line with our argument that these two leading explanations for the negative listing effect can be more serious in China where the legal system and market infrastructure are relatively less developed than they are in other countries. Our main findings are robust to alternative regression specifications and different matching thresholds. Our study contributes to the existing literature in several important aspects. First, we provide a novel explanation to the puzzling observation that do not improve profitability in China. We are the first to disentangle the listing effect from the privatization effect and empirically show that while the privatization effect is positive, it is dominated by the negative listing effect which leads to an overall negative SIP effect. Second, people may wonder why the Chinese government keeps privatizing its SOEs while do not seem to improve the profitability of privatized companies. Given the estimation by Gan, Guo and Xu (2017) that only account for 10 percent in terms of the total privatization assets in China, our large sample results for PSOEs suggest that the overall privatization program in China is successful, despite not improving profitability. Third, we provide strong evidence that partial privatization can improve profitability in China, and the larger the degree of partial privatization, the better the profitability 6

7 improvement. Our study is related to but distinguished from several strands of literature. First, previous studies on the performance of China s privatization program mainly focus on. An early exception is Dong, Putterman and Unel (2006), who find privatization without listing can improve profitability. However, their sample is small with only 165 firms concentrated in the Nanjing area. A more recent exception is Gan, Guo and Xu (2017), who use survey data and find that privatization via MBOs can improve profitability. We use a much larger sample of 4,701 firms with a matched control group across the country to show that privatization without listing improves profitability in China. Second, while many studies (see footnote 1) document the puzzling phenomenon that do not improve profitability in China, few offer any solid explanation why this is the case, besides suggesting that the nature of partial privatization in which government still retains the control power could be the reason. We propose a new explanation and provide consistent evidence that although partial privatization may dampen the positive privatization effect, the negative listing effect associated with is the main reason for the decline in profitability of SIP firms. Third, previous research documents that Chinese firms aggressively engage in earnings management for IPOs, SEOs, and to meet various regulatory requirements (Aharony, Lee and Wong, 2000; Yu, Du and Sun, 2006: Kao, Wu and Yang, 2009; Wang and Yung, 2011; Shen, Coakley and Instefjord, 2014). In addition, listed Chinese firms heavily engage in related party transactions with their parents and sister companies at the expense of individual and non-controlling shareholders (Jiang, Lee and Yue, 2010; Peng, Wei and Yang, 2011; Jiang, Rao and Yue, 2015). We explicitly show that earnings management and related party transactions are partially responsible for the negative listing effect of SIP firms in China. Finally, La Porta, Lopez-de-Silanes, Shleifer and Vishny (1998, hereafter LLSV) point out that the legal and market environment affects opportunistic behavior of managers and investors. We use the marketization index compiled by Fan, Wang and Zhu (2011) to show that the earnings management and related party transactions are more serious in provinces with poorer legal and market infrastructure. This helps to explain why in other countries are more successful than in China. Since we do not have the necessary data, it is impossible for us to do a cross country comparison of the listing effect. However, according to LLSV (1998), China s legal and market environment is ranked quite low in the world. Allen, Qian and Qian (2005) also point out that China s formal sector performs poorly due to the weak market and legal environment and the private sector is the main driving force for China s fast-growing economy. Many other studies, including Fernald and Rogers (2002) and Mei, Scheinkman and Xiong (2009), show that the Chinese stock market is highly speculative and subject to heavy interventions by the government. If we agree that China s legal and market infrastructure is poor relative to many other countries, then we can reasonably infer that earnings management and other agency problems associated with should be more 7

8 serious, which leads to a more severe negative listing effect. Our study not only has theoretical implications for the privatization literature, but also has very practical implications for China, as the truly massive (over $100 billion per year, as documented by Megginson, 2017) privatization program is still on-going. To make more successful, the Chinese government should tackle the serious negative listing effect and consider further relinquishing control of its listed, partially privatized companies. The rest of this article is organized as follows. Section 1 describes data, sample and variables used in our analysis. Section 2 examines whether there is an overall negative SIP effect on profitability as documented in the literature during our sample period. Section 3 examines whether privatization without listing can improve profitability and explores whether the profitability improvement is related to the degree of privatization. Section 4 further examines whether there is a negative listing effect and relates the negative listing effect to EM and RPT. Section 5 disentangles the privatization effect from the listing effect to show that the negative listing effect is the main reason for lack of improvement in the profitability of SIP firms. Section 6 examines whether our main findings are robust to alternative matching algorithms. Section 7 concludes. 1. Data, Sample and Variables 1.1 Data We obtain the IPO and associated pre- and post-listing financial data from the China Stock Market and Accounting Research (CSMAR) database. We further obtain the relevant financial data for non-listed companies from the National Bureau of Statistics (NBS hereafter) database. This database is constructed and maintained by the National Bureau of Statistics of China, providing similar financial information on all manufacturing firms with annual sales of more than RMB 5 million, be they SOEs or privately-owned. These firms operate in 37 two-digit manufacturing industries and they are geographically located in 31 province or province-equivalent autonomous regions. The NBS database has been used to complement the CSMAR in many studies to address research questions that could not have been properly explored in the past, including Cai and Liu (2009), Li, Yue and Zhao (2009), Hsieh and Klenow (2009), Jiang, Yue and Zhao (2009), Liu and Siu (2011), Ding, Guariglia and Knight (2013), Fan, Huang and Zhu (2013), Piotroski and Zhang (2014), Feenstra, Li and Yu (2014), Lin, Sun and Wu (2015), and Gan, Guo and Xu (2017). *** Insert Table 1 around here *** Since the data availability in NBS is from 1998 to 2013, our sample period for SIP firms starts from 1999 and ends in 2012 so that we have at least one year before or after privatization/listing data for these firms to compare profitability changes. Table 1 presents descriptive statistics for the data. Panel A of Table 1 presents the year distribution of various groups of firms included in our analysis. Column 1 shows the total number of IPOs listed on the SSE and the SZSE across the years. There are 1,729 IPOs in total from 8

9 1999 to 2012 but these IPOs are not evenly distributed across years has only 15 IPOs. Part of the reason behind this variation is the split share structure reform which started in early 2005 and ended in mid (Firth, Lin and Zou 2010; Li, Wang, Cheung and Jiang 2011; Liao, Liu and Wang 2014). The reform aimed to make non-tradable state and legal-person shares tradable. The CSRC (China Securities Regulatory Commission) suspended IPOs for that period to reduce the supply of shares coming to market. From these IPO firms, we identify SIP and privately owned (PO) firms that went public. We define a firm as an SIP if the status of its ultimate owner in the pre-ipo year is the state and the percentage of state ownership is greater than 50%. We define a company as a PO if the percentage of state ownership is less than 10%. We drop firms with state ownership in the grey area between 10-50% because we cannot classify them accurately into either type. 5 The number of SIP firms and private firms going public in each year are shown in Columns 2 and 3. In general, we can see that the number of tends to decrease while the number of listed POs tends to increase during the sample period, which is consistent with the trend that the state sector has been shrinking while the private sector growing. Column 4 presents the total number of firms included in the NBS database. It varies greatly from a low of 165,118 firms in 1998 to a high of 412,212 in 2008, and 344,875 in For firms that remain unlisted throughout the sample period, we define three different groups of firms. PSOEs (privatized SOEs) refer to those firms that went through privatization via a method other than public share offering (management buyout, direct sale of stock ownership, sale of assets) during the sample period. We require that a PSOE must have 50% or more state ownership before privatization. The year of privatization is recorded if the percentage of state ownership goes down by more than 15% in that year. The 15% threshold is chosen because we want to match PSOEs with in the study as many SIP firms sell less than 20% of shares to the public upon their IPOs. Unlisted SOEs (SOEs) and unlisted privately-owned firms (UPOs) refer to firms with state ownership greater than 90% and less than 10% throughout our sample period, respectively. We choose 90% because this threshold can ensure that SOEs do not experience significant privatization in previous years, and also because the median state ownership before is more than 90%. The distributions of PSOEs, SOEs, and UPOs over the years are shown in Columns 5, 6, and 7 respectively. There are 4,895 PSOEs from 1999 to The number of SOEs and UPOs are 17,304 and 37,918 in 1998, respectively. There are firms newly added to these two groups over years. Altogether, there are 36,674 SOEs and 338,291 UPOs in the final sample year. There is no overlap of the firms between any two groups. 5 Using greater than 60% government ownership before listing as the screen for SIP before listing as the screen for private firm reduces the sample size but does not change the principal results obtained in the paper. 6 Gan, Guo, and Xu (2017) provide a detailed discussion on the NBS data quality, especially the change of the number of firms included in the database over the years in their appendix. 9

10 Panel B of Table 1 presents the summary statistics of firm size and profitability for all five groups (SIP, PO, PSOE, SOE, and UPO) of firms. For listed firms including both and listed POs (PO thereafter), the summary statistics are based on seven years of data from three years before the IPO to three years after. Similarly, for PSOEs, the statistics are based on the data from three years before the privatization to three years after. 7 For unlisted SOEs and UPOs, the statistics are based on firms with at least three consecutive years of data during the whole sample period. It is obvious that listed firms are much larger in total assets and total sales (both in 2010 RMB) than are unlisted companies. Within the listed firms, are larger than POs. Within unlisted firms, unlisted SOEs are larger than UPOs, while PSOEs are comparable to UPOs. The primary focus of this study is on the profitability improvement for SIP firms and PSOEs. We measure profitability as return on assets (ROA), return on equity (ROE), return on sales (ROS) and EBIT/Sales. The first three are defined as net income over total assets, total equity, and sales revenue, respectively, while the last is defined as EBIT standardized by sales revenues. Under all these measures, profitability is higher for listed firms than those unlisted on average. However, within listed firms, profitability is lower for than for POs in general. Among unlisted firms, UPOs have the highest profitability, while PSOEs have higher profitability than SOEs. These findings suggest that private firms are more profitable than SOEs and partially privatized SOEs. Panel C of Table 1 presents the distribution of firms across state ownership deciles before and after listing or privatization for and PSOEs, respectively. The median (mean) state ownership for before listing is 87% (94%) and 59% (61%) after listing, with the median (mean) decrease being 21% (33%). Note that once a firm with more than 50% state ownership is listed over the sample period, it is considered a SIP. Hence, there is still one case in which the firm with its post-listing state ownership of more than 90% and there are also 35 cases in which the decline in state ownership during the SIP is less than 10%. For PSOEs, the median (mean) state ownership changes from 100% (91%) before the privatization to 0% (23%) after, and the majority of these firms have more than 90% government ownership before privatization but less than 10% after the privatization. Yet, there are still more than 650 firms with more than 50% state ownership after privatization. These firms can, however, match to SIP firms in our empirical analysis as SOEs, since government usually retains more than 50% ownership in SIP firms after their listing. 2. The SIP Effect Although studies on China s privation unanimously document that do not improve firm 7 For firms privatized in 1999 or 2000, only 1 or 2 years of pre-privatization data are available. Similarly, for firms privatized in 2012 and 2011, only 1 or 2 years of post-privatization data are available. 10

11 profitability, we still want to confirm that this is indeed the case during our sample period. Following the before-and-after methodology developed by Megginson, Nash and Randerborgh (1994) and used by many others, we define the year of going public as the SIP year (t=0) and compare performance measures three years before and after the SIP year for 498 SIP firms. Our major focus is on profitability measures such as ROS and EBIT/Sales, but we also report ROA and ROE for reference. We define the difference in profitability, or profitability improvement, as the average profitability over the three-year post-sip period minus the average profitability over the three-year pre-sip period. To be consistent with previous studies, we also compare the 3-year average real total assets, real sales, real net profits, capital expenditure over total assets, leverage, turnover and the state ownership before and after the SIP. The real total assets, real sales and real net profits are the annual total assets, sales and net profits adjusted by the Chinese producer price index with 2010 as the base. Leverage is total liabilities over total assets, and turnover is total sales over total assets. The univariate test results for the before-and-after comparison of these variables are reported in Panel A of Table 2. The pre-sip median and mean of each variable are shown in Column 1, while the post- SIP median and mean are in Column 2. The median difference (post pre) for each variable and its associated Wilcoxon Z statistics are shown in Column 3. Following the literature, we draw inferences from the median rather than from the mean because the former is less affected by the presence of outliers. The results clearly indicate that increase total assets, total sales, total net profit, and capital expenditure, but decrease profitability, leverage, turnover, and state ownership; all these changes are highly significant statistically. In particular, the median ROS drops from percent before the SIP to 7.44 percent after the SIP. The median EBIT/Sales also declined substantially, from percent in the pre-sip period to 9.02 percent in the post-sip period. We do not draw inference from the significant decline in ROA and ROE as in China are mostly primary issues which mechanically increase total equity and total assets and thus decrease ROA and ROE. These findings are very much consistent with the existing literature on the performance of China s SIP firms (see footnote 1 for a summary of relevant papers). *** Insert Table 2 around here *** Evidence of declined profitability relative to the pre-sip level for does not necessarily mean a negative SIP effect, since profitability changes before and after the SIP event are also driven by marketwide factors which have nothing to do with. It is possible that SOEs which are not privatized may have even worse profitability than. So even if have a profitability decline, compared to their own past, they may still outperform SOEs without privatization in a relative sense. To explore this possibility, we further refer to the most comparable unlisted SOEs which do not experience any SIP or partial privatization over the same period and use their profitability changes as the benchmark to assess the profitability improvement for SIP firms. This is effectively a difference-in-differences analysis, which can 11

12 address a number of econometric and empirical challenges described in existing studies, explicitly or implicitly. For example, using a matched sample, the results are less likely driven by selection bias since we construct a random sample of and unlisted SOEs in which 50% of SOEs went through issuing shares while the other 50% did not. The results are also less likely driven by the tendency of Chinese governments to privatize worse performing or better performing SOEs, because our control group is chosen on the basis of sales revenues and profitability in the pre-sip period. Finally, the results should be less vulnerable to the omitted-variable problem since the DID approach also attempts to minimize potential effects on profitability changes for SIP firms due to unobservable changes in the macroeconomic, marketwide, or industry-wide conditions over the 7-year event window. These effects, if any, should be removed, to a great extent, from the DID analysis when we include a temporally matched control sample and use their performance changes for adjustment. For this purpose, we match each SIP with an unlisted SOE which operates in the same industry and has similar sales revenue and ROS in the three-year period before the SIP year but does not experience any form of privatization over the sample period. This is a common matching strategy used by, among others, Mortal and Reisel (2013), Gao, Harford and Li (2013), and Asker, Farre-Mensa and Ljungqvist (2015). Our baseline matching algorithm requires that the characteristics of a comparable firm should not be 30% greater or smaller than the corresponding characteristics of the SIP firm in relative terms. Using the nearest neighborhood principle, we identify a control sample of 297 unlisted SOEs over the period. Panel B reports results based on 297 SIP firms and their comparable unlisted SOEs. Note the mean and median sales and ROS are very similar between SIP firms and unlisted SOEs in the pre-sip period indicating that the match is reasonably good. The pre-sip difference between these two groups (SIP SOE) shows that SIP firms are more profitable in terms of ROA, ROE, and EBIT as the median difference under these measures are statistically significant at the 1 percent level, but the median difference in ROS is statistically insignificant. However, the differences in these profitability measures in the post-sip years are largely reversed. ROA and ROE of SIP firms are significantly lower than those of SOEs. The EBIT/Sales of SIP firms is also lower than that of SOEs, although it is statistically insignificant. The difference in ROS is almost zero and statistically insignificant. Using the (post pre) difference of SIP firms minus the (post pre) difference of SOEs, we obtain the difference-in-differences results. Similar to those shown in Panel A, we find that significantly increase real sales, profits, and capital expenditure but significantly decrease ROA, ROE, leverage and turnover. ROS and EBIT/Sales show a negative change but are statistically insignificant. Therefore, we conclude that share issue privatization does not improve profitability for SIP firms over the period. 3. The Privatization Effect 12

13 After confirming that the SIP does not improve profitability, we now examine whether privatization without listing improves profitability. Following the similar matching criteria specified in the previous section, we match PSOEs with similar unlisted SOEs that do not go through privatization in terms of industry, sales revenues and ROS in the pre-privatization period. Out of 4,895 PSOEs and tens of thousands SOE firms, we obtain 4,701 pairs, giving 4,701 PSOEs in the treatment group and 4,701 comparable unlisted SOEs in the control group. We then perform a DID analysis on the matched sample. *** Insert Table 3 around here *** Table 3 presents the univariate test results. For simplicity, we only report the results for real sales, real net profit, the other profitability measures and the state ownership. For the other variables, the results are available upon request. First, we find that profitability of PSOEs tends to improve relative to their preprivatization levels, since the median increase in ROA, ROE, ROS and EBIT/Sales are 0.45, 0.87, 0.32 and 0.39 percentage points, respectively. Second, we find that sales revenues, ROS and state ownership in the pre-privatization are reasonably similar between PSOEs and SOEs due to the matching. The median difference in ROS and state ownership is not significant. While the median difference for real sales is statistically significant, the median difference relative to the median real sales for PSOEs is very small (0.14% = 0.03/20.94). Third, ever after making adjustments for the impacts unrelated to partial privatization, the privatization effect on profitability is still positive and significant under all profitability measures. The difference-in-differences results, (post-psoe pre-psoe) (post-soe pre-soe), show that the median improvements for ROA, ROE, ROS, and EBIT/Sales are 0.91, 2.37, 1.15 and 1.35 percentage points, respectively. Therefore, no matter comparing performance relative to their own past or relative to their matched peers, privatization without listing improves profitability. To alleviate the concern that our matching algorithm is rough, as it is only based on the industry and two pre-privatization variables with up to 30% deviation between a PSOE and its matched SOE. We further perform a regression analysis which can better control for more pre-privatization firm characteristics. In addition, we can also examine whether profitability improvement is negatively related to the state ownership change or to the remaining state ownership after privatization. This is an interesting issue as some authors argue that partial privatization may be better than full privatization in China. For example, Sun, Tong and Tong (2002) find that partial government ownership has a positive impact on SOE performance for a sample of firms listed on the SSE and the SZSE during the period Tian and Estrin (2008) document a non-linear relationship between remaining state ownership and the performance for SIP firms after their listing. The baseline regression we use is specified as follows: ROS (or EBIT Sales ) = DPSOE State Ownership i i 0 1 i 2 Post _ State Ownership Pre _ Ln Total Assets Pre _ ROA 3 i 4 i 5 i i 13

14 Leverage Pre _ CAPINV Pre _ Ln 1 Age (1) 6 i 7 8 where the dependent variable is the profitability improvement (post pre) for firm i measured as ΔROS or ΔEBIT/Sales; Pre_Ln(Total Assets), Pre_ROA, Leverage and Pre_CAPINV are the natural logarithm of the average of total assets, ROA, leverage and capital expenditure in the pre-privatization period for firm i; Pre_Ln(1+Age i) is the natural logarithm of firm i s age measured as the number of years since the registration of the firm; Privatization year dummies are included to control for the year-specific effect; and DPSOE i captures the general privatization effect on profitability, which is set to 1 if the firm is a PSOE and zero otherwise. We further include the change in state ownership Δ(State Ownership i) from the preprivatization period to the post-privatization period, (post pre), which is negative for all PSOEs and the average remaining state ownership, Post_State Ownership i, for PSOE i after the privatization, which is nonnegative for all PSOEs. Both Δ(State Ownership i) and Post_State Ownership i are used to examine whether the profitability improvement is affected by the degree of privatization. Since Δ(State Ownership i) and Post_State Ownership i are correlated to a certain extent, we do not include both of them in the same regression. *** Insert Table 4 around here *** Table 4 presents regression results for the matched sample which contains 4,701 PSOEs and their 4701 comparable unlisted SOEs. The first 3 columns show the results using ΔROS i as the dependent variable. The Column 1 regression is actually a DID analysis with more controls on firm specific characteristics. The PSOE dummy (DPSOE) captures the relative profitability improvement for PSOE firms in general. The coefficient for DPSOE is , which is significant at the 1 percent. This confirms the univariate DID results for ROS shown in Table 3. In fact, the relative ROS improvement in the regression is 3.78% which is much larger than 1.15% shown Table 3. This is understandable as our univariate analysis is based on median values while regression analysis based on mean values and also with more controls. We further add in Δ(State Ownership i) in the Column 2 regression. Interestingly, both DPSOE and Δ(State Ownership i) enter the regression significantly. The coefficient for DPSOE is positive and the coefficient is negative for Δ(State Ownership i). This indicates while privatization improves the profitability in general, the profitability improvement is negatively related to the change (post pre) in state ownership, which suggests the degree of privatization matters, the larger the reduction in state ownership, the better the profitability improvement. When we include Post_State Ownership i in the regression with DPSOE as shown in Column 3, it also enters significantly negatively, suggesting the profitability improvement is negatively associated with the remaining state ownership in a PSOE firm. Regressions reported in Columns (4) through (6) where the dependent variable is ΔEBIT/Sales provide similar results. The coefficients on DPSOE are all positive and significant, greater than the 1.35% median profitability improvement indicated in Table 3. Both the coefficient on Δ(State Ownership i) in Column (5) and the coefficient on Post_State 14 i

15 Ownership in Column (6) are negative and statistically significant. Overall, we find strong evidence in support of a significantly positive privatization effect associated with SOEs going through privatization without listing. In addition, the higher the degree of privatization, the larger is the profitability improvement for PSOEs. 4. The Listing Effect With a non-positive SIP effect documented in Section 2 and a positive privatization effect documented in Section 3, one may infer there is a negative listing effect. However, such an inference may be premature. As shown in Tables 2 and 3, there are some obvious differences between listed firms and unlisted firms. Listed firms, especially are much larger and more profitable on average than the 4,895 PSOEs. The pre-listing median (mean) real sales and ROS for the 498 SIP firms are RMB (4,935.78) million and 11.39% (14.15%), respectively, while the corresponding median (mean) values for 4,895 PSOEs are only RMB21.45 (97.76) million and 1.75% (0.63%). In addition, the government is still the controlling shareholder for the typical SIP firms while most PSOEs have zero state ownership after privatization. The median (mean) state ownership after privatization for the 498 SIP firms is 61.08% (59.60%) while for 4,895 PSOEs it is 0.00% (22.96%). To see whether the listing effect is really negative, we do a further matching between SIP firms and PSOEs to make them comparable in terms of industry, sales revenues, profitability, as well as state ownership. Using the same matching algorithm, we obtain 142 SIP firms and 142 similar PSOEs. *** Insert Table 5 around here *** Panel A of Table 5 reports the comparison results between 142 SIP firms with 142 matched PSOEs. Notice that the median (mean) pre-listing/privatization real sales, ROS, and state ownership for SIP firms are RMB (316.33) million, 10% (11%), and 95% (87%), respectively, and the median (mean) preprivatiztion real sales, ROS, and state ownership for PSOEs are RMB (542.75) million, 9.78% (11.08%), and 94.82% (87.07%). The two groups are now much more comparable. In addition, the postprivatization state ownership between the two groups is also more comparable. The median (mean) state ownership is 65.90% (60.75%) for the 142 SIP firms and 59.82% (55.72%) for the 142 matched PSOEs. The comparison of the profitability improvement between SIP firms and PSOEs clearly indicates SIP firms underperform. The median relative underperformance (DID) of versus PSOEs in terms of ROS, EBIT/Sales, ROA and ROE are -3.78%, -4.50%, -3.87% and %, respectively, all statistically significant at the 1% level. We attribute such underperformance of SIP firms to the event of public listing, as it is the main, if not the only, difference between the 142 SIP and 142 comparable PSOEs. We admit that the sample size is a bit small in the above comparison. To remedy this concern, we repeat the comparison with an alternative sample. The listing effect, if any, should affect not only but 15

16 also privately-owned firms that become listed through an IPO (become listed POs). We manage to match 863 listed POs with the same number of unlisted private firms (UPOs) which are similar in industry, sales revenues and ROS. The main difference between these two groups is again listing, as there is no privatization involved. Using this matched sample, we perform the same test to examine if the listing effect is indeed negative. Panel B reports the comparison statistics for this matched sample. We find robust evidence that listed POs underperform their matched UPOs in all four profitability measures with statistical significance at the 1 percent level, which also shows the existence of a negative listing effect. Although the 142 SIP sample is not matched to the 863 PO sample, the magnitude of the listing effect for SIP firms and for listed POs are not very different. The median underperformance of listed POs in ROS and EBIT/Sales, from the pre-listing period to the post-listing period is -3.27% and -4.41%, respectively. We further perform regression analyses to control for more pre-listing variables. The baseline regression is specified as follows: ROS (or EBIT Sales ) = List Pre _ Ln Total Assets Pre _ ROA i i 0 1 i 2 i 3 i Leverage Pre _ CAPINV Pre _ Ln 1 Age (2) 4 i 5 8 where List i is a dummy variable which takes the value of 1 for listed firms and zero, otherwise. The other variables are defined the same as those in Equation (1). *** Insert Table 6 around here *** Table 6 reports the regression results. Panel A presents the results for 142 SIP firms and their matched peers. Column 1 shows the estimation of Equation (2) with ΔROS i as the dependent variable. The estimated coefficient for List is , which is significant at the 1 percent level, indicating an average negative listing effect of about -4.61% for SIP firms, which is more negative than the univariate estimate of -3.78% shown in the Panel A of Table 5. However, it is not directly comparable as the former captures the mean difference after control for various variables in the regression while the latter is the median difference in a univariate test. We add Δ(State Ownership i) and Post_State Ownership i, the two alternative proxies for the degree of privatization, in Columns 2 and 3 regressions, respectively. The estimated coefficient for List is almost the same as that in Column 1, showing a negative listing effect of around 4.70%. However, neither Δ(State Ownership i) nor Post_State Ownership i enters the regression significantly, indicating the degree of privatization has no impact on profitability which is inconsistent with our earlier finding in Table 4. It seems to suggest that while the privatization matters for the SIP firms and their comparable PSOEs, the degree of privatization does not matter for this group. A unique feature for this sample is that the degree of privatization is small and most of these firms still have more than 50% of state ownership after the privatization. In fact, some studies (see Sun and Tong (2003) and Tian and Estrin (2008) as examples) find the relationship between state ownership and profitability for SIP firms after listing is i 16

17 either insignificant or non-linear. We repeat these regressions using ΔEBIT/Sales i in place of ΔROS i and results are reported in Columns 4, 5 and 6 of Table 6. They are largely the same as those reported in Columns 1, 2 and 3. Panel B presents the regression results for 863 listed POs and their matched peers. Since these are private firms, we do not include any privatization proxy in the regressions. The results clearly show a negative and significant listing effect. However, the magnitude of the listing effect is about 2%, which seems smaller than those reported in Panel B of Table 5. Overall, our analysis in this section provides strong evidences for the existence of a negative listing effect in China. This is consistent with Degeorge and Zeckhauser (1993), Jain and Kini (1994), Holthausen and Larcker (1996), and Pagano, Panetta and Zingales (1998), who report considerable evidence that there is generally a significant decline in operating performance post IPOs. Plausible explanations for declining performance in the post-ipo period include worsening agency problems (Mikkelson, Partch, and Shah, 1997), earnings management (Teoh, Welch and Wong, 1998), and rational timing of IPO decisions by an entrepreneur who can learn about average IPO profitability (Pastor, Taylor and Veronesi, 2009). We further examine if the listing effect is related to earnings management and worsening agency problems. We do not consider rational timing as a determinant for the listing effect because China has an IPO approval regime where the firm managers do not have the capability to decide when to list. Following the existing literature (Teoh, Welch and Wong, 1998), we use discretionary accruals as our proxy for the level of earnings management. We obtain relevant financial data from CSMAR and use the modified Jones model (Dechow, Sloan and Sweeney, 1995) to estimate the discretionary accruals (DAs) in the pre-ipo year for 142 SIP firms and 863 listed POs. It is well documented that EM is widely used in China for the listing purpose, and it negatively affects the firm performance after listing. Hence, we expect a negative relationship between DAs and the listing effect, i.e., the more serious the EM is, the more negative the listing effect. We provide a description on how we use the modified Jones model to derive DAs in the Appendix. Several studies show that wealth can be transferred between listed firms and their large shareholders through related-party transactions such as inter-corporate loans (Jiang Lee and Yue (2010) and Aharony, Wang and Yuan (2010)), propping up and tunneling (Cheung, Jing, Lu, Rau and Stouraitis (2009) and Peng, Wei and Yang (2011)), and non-operational fund occupancy (Jiang, Rao and Yue, 2015) in China. Following these studies, we use related-party transaction to proxy for agency problems. We expect declined profitability is associated with related party transactions, an important way that controlling shareholders tunnel the assets or profits from the listed firms, be it the government or a family. They can inject good quality assets or money to prop up the firm for listing (Peng, Wei and Yang, 2011). However, after these firms go public, they tend to get back financial resources from RPT (Jiang, Lee and Yue, 2010; Peng, Wei and Yang, 2011; Jiang, Rao and Yue, 2015). Hence, we expect that a larger post-listing RPT value would 17