Journal of Banking & Finance 37 (2013) 1460–1474

Contents lists available at SciVerse ScienceDirect

Journal of Banking & Finance journal homepage: www.elsevier.com/locate/jbf

Overseas listing as a policy tool: Evidence from China’s H-shares
Qian Sun a,⇑, Wilson H.S. Tong b, Yujun Wu c a Department of Finance, School of Management, Fudan University, Shanghai 200433, China School of Accounting and Finance, Faculty of Business and Information Systems, Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong Special Administrative Region c Wealth Management Institute of Lujiazui, Shanghai 200122, China b a r t i c l e

i n f o

a b s t r a c t
We investigate why the Chinese government chooses to perform share issue privatization (SIP) of its state-owned enterprises (SOEs) in Hong Kong, despite the benefit of facilitating the domestic stock market development if performing SIP in China (Subrahmanyam and Titman, 1999) and the higher cost to list in Hong Kong. We address this issue by arguing that the positive effect of SIPs on the development of the domestic market may have limitations, especially when the domestic market is not well developed and cannot absorb rapid and large-scale SIP activities. To maintain domestic market order, it may be optimal to carry out SIP in overseas markets. Furthermore, by listing shares in developed overseas markets, SOEs from the less developed countries could leverage on the overseas markets’ better accounting, governance, and legal standards. By examining a sample of 92 Chinese firms listed in Hong Kong and the relevant control samples of purely domestically listed Chinese firms during the period of 1993–2006, we find supporting evidence for both arguments. Ó 2012 Elsevier B.V. All rights reserved.

Article history: Available online 7 June 2012 JEL classification: G34 G39 Keywords: Privatization Overseas listing Market order Corporate governance SOE

1. Introduction This study investigates why the Chinese government chooses to perform share issue privatization (SIP) of many of its state-owned enterprises (SOEs) in an overseas market, particularly in the Stock Exchange of Hong Kong (SEHK),1 instead of in the domestic market. Such an investigation is important, because SIP in Hong Kong started in 1993 and has become increasingly prominent, even though it appears to be inconsistent with existing finance theories, at least on the surface. First, in the cross-listing literature, one popular explanation for firms listing abroad is the lowering of the cost of capital by breaking down the barriers of investment (Stapleton and Subrahmanyam, 1977; Errunza and Losq, 1985; Alexander et al., 1987). Yet a well-documented but puzzling fact is that the shares of the Chinese firms cross-listed in Hong Kong (the ‘‘H-shares’’) typically trade at a price discount relative to their ‘‘A-share’’ counterparts in the Chinese domestic markets (Sun and Tong, 2000; Wang and Jiang, 2004).2 If anything, Chinese firms are subject to a higher cost of
⇑ Corresponding author. Tel.: +86 21 65640983; fax: +86 21 65648384.
E-mail addresses: sunqian@fudan.edu.cn (Q. Sun), afwtong@inet.polyu.edu.hk (W.H.S. Tong), wuyujun@wmichina.org (Y. Wu). 1 Although Hong Kong was handed over to China in 1997, under the ‘‘one-countrytwo-systems’’ policy, the Hong Kong stock market is a ‘‘foreign’’ market to Chinese companies for all practical purposes. 2 Admittedly, the price discount has been narrowed in recent years and a few stocks have even shown some price premiums. But discount is still a general phenomenon. See more discussion below. 0378-4266/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jbankfin.2012.05.013

capital when conducting their IPOs in Hong Kong, because their issuing P/E multiples have been lower than in the domestic market. Yet we continue to see Chinese SOEs doing IPOs in Hong Kong.3,4 Second, in the privatization literature, Megginson and Netter (2001), Megginson et al. (2004) found that SIPs tend to occur in countries with less developed capital markets, probably due to the need and desire of the governments to use SIPs to develop the national market’s liquidity and absorptive capacity. This echoes Subrahmanyam and Titman’s (1999) argument that SIPs facilitate national stock market development. This is also consistent with the fact that firms listing their shares abroad typically list in their home markets first through an IPO, before cross-listing their shares abroad. This is even truer for privatized firms. Although privatized firms tend to list overseas (Pagano et al., 2002), Bortolotti et al. (2002) found that out of a sample of 392 public offerings in the Organisation for Economic Co-operation and Development (OECD) and non-OECD countries, only 11 privatized firms had their primary listing abroad. On the contrary, all Chinese SOEs that listed in Hong Kong as H-shares had their initial offerings in Hong Kong during our sample period of 1993–2006, and

3 In fact, the large-scale overseas listings of Chinese SOEs came in the last several years when the Chinese government privatized its giant state banks, such as the Bank of China (BOC), the Industrial and Commercial Bank of China (ICBC) in 2006, and most recently, the Agricultural Bank of China (ABC) in 2010. However, the IPOs for these banks were, for the first time, simultaneously issued in both the Hong Kong and Shanghai stock markets. 4 Since China has capital control, no arbitrage is allowed between A- and H-share markets and between A- and B-share markets. Hence, A-, H-, and B-shares issued by the same company can be traded at different prices.

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

1461

only half of them subsequently cross-listed back in the Chinese home market as A-shares. To explain this apparently odd phenomenon, we propose two hypotheses. First, we suggest a ‘‘market order’’ hypothesis. Although there are positive impacts of domestic SIPs on market development, such benefits likely have limitations. If the SIP progresses more rapidly than the development of the domestic market, and if the SIP scale is larger than what the domestic market can absorb, SIPs may hinder rather than facilitate market development. SIPs in overseas markets can be an alternative choice for governments. For instance, in Bortolotti et al.’s (2002) cross-country study on why governments sell privatized companies abroad, it was found that the low liquidity of emerging markets induces governments to cross-list in order to ‘‘import’’ liquidity. Furthermore, the transition of the Chinese economy from central planning to market mechanisms has caused most SOEs to fall heavily into debt and short of equity capital. Many had an urgent need to list and raise equity capital in the stock market. Yet the Chinese stock market was still in an early stage of development with limited depth, and speculative bubbles abounded (Mei et al., 2009). To maintain market order and to prevent the market from crashing,5 the Chinese government imposed limitations, including a quota system in the early years, to regulate IPOs in the market.6 Since the demand to list was much larger than the quota allowed, many firms had to wait in the queue for years. As a way to relieve listing pressure in the domestic market, the government has chosen to list many of its large SOEs overseas, even at a large price discount. Second, we suggest a ‘‘modern enterprise’’ or ‘‘governance’’ hypothesis. To convert a central planned economy to a marketoriented one is a difficult task. Due to the weak legal and market infrastructure (Berkman et al., 2010), it would take a long time to establish a modern corporate system with international standards. By selectively listing SOEs overseas and exposing them to an international business environment, especially to a more stringent legal and monitoring environment, the Chinese government forces these overseas-listed companies to conform to the international norm. These companies can then set examples for other Chinese companies to follow. In fact, the chief of the State Asset Commission, Li Rongrong, has repeatedly stated that it is China’s long-term strategy to list its large SOEs overseas continuously, because this will help establish a modern enterprise system in China. This relates to, but is not the same as, the bonding argument in the cross-listing literature. Coffee (1999, 2002) and Stulz (1999) argue that good firms from a country with poor legal, accounting, and governance standards can distinguish themselves from other firms in the home country by cross-listing their shares in a foreign market with higher standards. Such cross-listings effectively bond these firms to higher standards and, hence, help to improve their credibility and prestige among investors. This bonding hypothesis has a subtle difference with our governance hypothesis, which is built on a government policy objective imposing on individual SOEs, aiming to improve their efficiency and quality through leveraging on the high quality of the cross-listed foreign market. The bonding hypothesis, however, postulates that firms self-selectively cross-list to a high quality foreign market to leverage on the better governance standard there. As a point of illustration, the bonding literature argues that firms listing on lower-quality exchanges do not raise as much capital in follow-on offerings as on higher-quality foreign exchanges. For example, Reese and Weisbach (2002) find that firms listing on high-quality exchanges are more likely to undertake
This is particularly conceivable for a central-planned economy like China’s, in which the government prefers everything to be under control. Market volatility and ‘‘instability’’ need to be avoided. 6 See Section 2 for details.
5

subsequent equity offerings. They attribute this ex post financing pattern to the markets assigning greater credibility to firms that are subject to the more stringent standards of high-quality exchanges. However, the fact is that H-share firms rarely have seasoned equity offerings in Hong Kong. They typically do seasoned issues back in China, which again is subject to government control. Hence, the ‘‘standard’’ bonding argument does not seem to provide a good explanation of the phenomenon. We argue that the Chinese government was keen on reforming its inefficient SOEs7 and, hence, has the incentive to make use of the bonding mechanism to advance their goal of SOE reform in establishing a modern corporate system in China.8 By comparing a sample of 92 Chinese firms listed in Hong Kong with a control sample of purely domestically listed Chinese firms during the period of 1993–2006, we show evidence supportive to our two hypotheses. We find that firms with large IPO issues are more likely to list in Hong Kong. On the other hand, even the Chinese domestic IPO issues, which are much smaller than the H-share firms in terms of issuing proceeds, exert downward pressure on the price and trading volume in the A-share market. By diverting large IPOs overseas, this downward pressure can be reduced. This is consistent with our ‘‘market order’’ argument. We also find evidence consistent with the ‘‘governance’’ argument. There is a valuation premium in the A-share market for firms issuing both H- and A-shares relative to the control samples of pure A-share firms. Since H-shares are traded at a price discount relative to the A-shares of the same company, dual listings should have a value transfer from A-shareholders to H-shareholders. Yet we observe a premium. We argue that such a premium is due to H-share firms having better corporate governance than pure Ashare firms, and we find that this premium decreased after 2002, the year designated by the China Securities Regulatory Commission (CSRC) as the year of corporate governance in China. CSRC announced and enforced a series of tough regulations and measures in 2001 and 2002 which dramatically raised corporate governance standards for listed firms in China. This clearly indicates that the premium relates to governance quality. We further show some evidence that this premium is likely to be the result rather than the cause of overseas listing. Through reconciling the apparent inconsistencies of the foreign listing phenomenon of Chinese firms with existing finance theories, we make important contributions to the overseas listing literature. First, by suggesting that the Chinese government uses overseas listing as a policy tool to establish a modern corporate system and to maintain market order in their relatively immature domestic stock market, our study enriches the literature on privatization, especially on foreign SIPs, studies of which are very limited. Second, we provide direct evidence that a lower cost of capital needs not be an important motive for cross-listings. Although Karolyi (2004) has already pointed out facts inconsistent with the argument of the lower cost of capital, these inconsistent facts are only indirect evidence, but our results are direct evidence. Third, although studies on cross-listings are voluminous, those on foreign primary listings are limited. The study by Blass and Yafeh (2001) on Israeli IPOs in the United States (US) seems to be the only work specifically in this area. However, their sample is very specialized and the phenomenon is likely to be temporary. According to Yehezkel (2006), Israeli companies went public in the US mostly in the 1990s, and most of them were young and from the high-tech industry. After the high-tech bubble burst in 2000, and

See Sun and Tong (2003) for a brief history of the Chinese SOE reform process. Hung et al. (2008) have an interesting paper that examines a similar issue. They argue that the overseas listing of Chinese SOEs is mainly motivated by political needs. Please refer to their footnote number 5, which highlights the differences between their approach and ours.
8

7

1462

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

the enactment of the Sarbanes–Oxley Act in 2002, the situation changed quickly. The phenomenon in this study is a persistent one, hence, our study fills the void in a more significant way. This study is also important because of the sheer size of the phenomenon. Listings of Chinese SOEs in Hong Kong have become even hotter and the size becomes ever bigger. In fact, Hong Kong raised a total amount of US$42.4 billion in 2006, which makes it comparable for the first time to the top capital-raising markets, such as London (US$48.3 billion) and the US (US$45.8 billion, combining NYSE and NASDAQ). Among the biggest IPOs in 2006 and 2007, four were H-shares.9 Hong Kong has also become the seventh largest stock market in the world with a market capitalization of US$1.55 trillion. The significance of these IPO events was written about in a Wall Street Journal article by New York Mayor Mr. Bloomberg and New York Senator Mr. Schumer on November 1, 2006, in which they raised the concern that the NYSE may lose its leading position in the global financial market. Finally, our study has practical relevance. As China’s economic importance and worldwide influence increases day by day,10 many developing countries and emergent markets are keen to know about and learn from China’s experience. Doing SIP through listing overseas is definitely an experience worth investigating. Our study helps understand this strategy better. In the following sections, we provide some background information regarding China’s overseas listings. We develop and test the market order and governance hypotheses in Sections 3 and 4. Section 5 concludes the paper. 2. Background information11 In this section, we provide brief introductions to H-share market development, China’s IPO quota system, China’s share–split system and market depth, and Chinese efforts to establish a modern enterprise system. 2.1. The development of the H-share market Mainland Chinese firms can list shares in overseas markets, such as Hong Kong (H-shares), New York (N-shares), Singapore (S-shares), London, and so forth. However, H-share firms dominate, both in terms of number and issuing size. In fact, most N-shares are traded in the form of American depositary receipts (ADRs), with the underlying shares listed in Hong Kong.12 Overseas listings were not originally planned by the government. Inspired by China’s growth potential after Deng Xiaoping’s grand tour of South China in early 1992, international investors wanted to hold Chinese equity. Investment banks foraged around China for restructured SOEs that wanted to raise capital overseas. In response, the State Council quickly issued its first regulation on listing overseas, ‘‘A Special Regulation on Raising Capital and Listing Overseas by a Joint-Stock Company,’’ on April 19, 1993, only three months before the first H-share listing by Tsingtao Brewery. Initially, the Chinese government wanted New York to be the overseas listing hub for its SOEs, but it ended up being Hong Kong. The
9 The Industrial and Commercial Banking Corporation’s IPO in October 2006 was record-breaking. The debut valued at about US$139 billion, ranking it fifth among global banks. It is the world’s largest IPO so far, raising an amount up to US$21.9 billion. The second largest IPO belongs to another Chinese state bank, the Bank of China, which debuted in May 2006. 10 China’s GDP surpassed Japan in 2010 and became the second largest economy in the world. HSBC’s global research department has recently released a report predicting that China will have the world’s top economy in 2050 (cnn.com, January 12, 2012). 11 Some of the information in Sections 2.1 and 2.2 is from Greene (2002, Chapter 3). 12 For details of the distribution of China’s overseas-listed firms, see Table 3.3 under ‘‘Statistical Information’’ in the CSRC’s website: http://www.csrc.com.cn/.

policy on listing overseas was also not well coordinated, because China’s ‘‘B-share’’ market had just opened to foreign investors in early 1992. The strict segmentation of B-shares as only for foreign investors and A-shares as only for Chinese local investors led to large price disparities. The price discount on B-shares relative to A-shares quickly increased after the issuing of H-shares, because H-shares provide a better alternative for foreign investors (Sun and Tong, 2000). Yet, the H-share prices themselves are also traded at substantial discounts relative to their A-share counterparts, although the discount tends to be narrow, and several of them have been evenly traded in recent years. As of July, 2010, 120 Hshare firms were listed on the main board of the Hong Kong Exchange and 37 were listed on the GEM board.13

2.2. The IPO quota system China’s IPO quota system was first adopted in 1993 to maintain market order and to prevent cash-starved, poor-quality SOEs from flooding the market with shares. The State Planning Commission determined the quantity of equity to be issued each year and CSRC would then divide this quota up among the provinces and ministries. A company seeking to list would have to be selected by a provincial government or ministry with a quota before asking CSRC for approval. Local authorities often reduced the issuing proceeds for each firm, in order to let more firms list within the quota limit. In 1996, the quota was changed from restricting the quantity of equity to be issued to restricting the number of firms to be listed. Such a quota system prevented many large firms from getting listed, and CSRC quickly found that overseas listings could be a partial solution. Eventually, firms with a net worth above RMB400 million, the previous year’s net profit greater than RMB60 million, and issuing proceeds of more than US$50 (about RMB400) million were allowed to apply for an overseas listing.14 By doing so, the Chinese government effectively diverted some big issues to the bigger overseas markets, although the CSRC has never admitted that it uses overseas markets to relieve domestic issuing pressure. In February 2001, CSRC replaced the IPO quota system with new rules, which reduced CSRC’s power to approve IPOs and increased the responsibility of the lead underwriters. First, quota allocations for listing were abolished. Second, CSRC approval was replaced with a CSRC confirmation requirement for IPOs. Third, CSRC established a review committee, consisting of both its own professionals and external specialists. Fourth, issuers and underwriters could now negotiate IPO prices on their own, although these would be subject to the CSRC’s approval. The rules were further revised in February 2004, which put more responsibility onto the securities firms. However, IPO firms still need to get a recommendation from qualified recommenders from one of the 29 major securities houses. A recommender can recommend only one firm at a time, and the number of qualified recommenders in a particular securities house was usually fewer than four in 2006, after the end of our sample period. After a firm is recommended, it has to go through a lengthy restructuring process for no less than a year before it can launch an IPO. As a result, a securities house can, at most, recommend four firms for IPOs during a particular year. Effectively, the explicit and strict quota system set by CSRC has been replaced by the new quota system of the securities houses. Hence, there are still limitations on how many IPOs can be issued under the new system.
13 The Hong Kong Growth Enterprise Market (GEM) was established in November 1999 to cater to small and high-tech companies. The Chinese firms listed on GEM are mostly private companies and are not the focus of our study. Up until June 2004 when the Small and Medium Enterprise Sector was opened in the Shenzhen Stock Exchange (SZSE), small private firms in China had no access to the domestic market. 14 These requirements do not apply to firms listed on the Hong Kong GEM board.

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

1463

2.3. The share–split system and market depth China had a unique share–split system during our sample period: the shares of listed firms were segmented into two categories, negotiable and nonnegotiable. The former was tradable on the stock market and mainly held by individual investors, while the latter was non-tradable and mainly held by the government (state shares) and other legal entities (legal person shares). The government and other legal entities that helped to establish or restructure a company in the process of SIP converted their contributions of existing assets (both tangible and intangible) into state shares and legal person shares, respectively. On average, the government owned about one-third of the total shares. Legal persons held another one-third and the rest was held by public (negotiable) shareholders. The situation changed when the so-called ‘‘share–split structure reform’’ was launched in April 2005 and completed at the end of 2006. The reform allows the nonnegotiable state and legal person shares to be gradually negotiable after two years. Under this reform plan, the nonnegotiable shareholders gave out about 30% of the shares to the (tradable) A-share shareholders as a means of compensation, in anticipation that once the non-tradable shares became tradable, the stock market would plummet. In fact, CSRC froze all of the IPOs and dramatically reduced SEO during most of 2005–2006. Hence, maintaining market order was a key concern of the government throughout our sample period. Even now, a certain amount of shares are still not allowed to trade because CSRC has stipulated that only a certain percentage of non-tradable shares can enter the market every year.

2.5. Comparative statistics for firms listed in China and Hong Kong We present comparative market statistics for firms listed in China and Hong Kong in Table 1. As shown in Panel A, the mainland Chinese stock market has been growing much faster than the Hong Kong market. The number of firms listed in China’s A-share market increased by almost nine times from 177 in 1993 to 1527 in 2007, while for the same period firms listed on the main board of the Hong Kong Exchange only increased by a little more than double, from 477 to 1048. Of these 1048 firms, 105 were H-share firms, and 51 of those were also cross-listed on either the Shanghai Stock Exchange (SHSE) or the Shenzhen Stock Exchange (SZSE). The number of firms listed in China’s B-share market also increased from 41 in 1993 to 109 at the end of 2007. 86 out of 109 B-share companies had dual-listings in the A-share market. As of 2007, there were 1390 pure Ashare firms (i.e., A-share firms without dual-listings in either the B- or H-share markets). IPO proceeds raised from the H-share market were also larger than those raised from the B- and A-share markets, especially when taking into consideration the number of H-share firms relative to the A-share firms (Panel B). In 1994, and from 2003 to 2006, H-share IPO proceeds were even larger than A-share IPO proceeds. Apparently, a very significant portion of the amounts issued for IPOs was diverted to the Hong Kong stock market during our sample period. On the other hand, the market P/E ratio in Hong Kong was much lower than that in the A-share market during most of the years in our sample period. However, most of the time it was higher than the P/E ratio in the B-share market, at least until 2000. After 2000, the higher P/E ratio for B-shares was most likely due to the partial liberalization of the B-share market in early 2001, when the B-share market was opened to local Chinese investors having US Dollar or Hong Kong Dollar bank accounts. Finally, Panel C shows that the market capitalization of the Hong Kong stocks was much larger than the combination of both of the A- and B-share markets. Yet, Hong Kong’s market trading value has been generally lower than, or similar to, the value of the Ashare market since 1996. Consequently, Hong Kong’s market turnover rate was much lower than either the A- or B-share markets.

2.4. Establishing a modern corporate system Since most of the listed companies in China were converted from SOEs, it is understandable that corporate governance was not well developed. There have been many stories about insider trading, related-party transactions, ineffective or even collusive boards of directors, the expropriation of minority shareholders, and accounting and disclosure irregularities (e.g., see Berkman et al., 2009; Liu and Lu, 2007; and Chen and Yuan, 2004). According to MacNeil (2002), the investor protection index score for China is ‘‘2,’’ compared to a world average of ‘‘3’’ and a maximum of ‘‘6.’’ Hence, establishing a modern enterprise system with good corporate governance is one of China’s objectives for SOE reform. As more and more Chinese firms listed overseas, the Chinese government started to emphasize the strategic role played by overseas listings in establishing a modern corporate system. In fact, the Chinese government has stated all along that one purpose of listing firms overseas is to bring the management and performance of Chinese firms up to international norms. By listing overseas, Hshare firms are forced to undergo a thorough restructuring, to be audited according to international standards, and to be disciplined and monitored by a more demanding investment community. On the other hand, CSRC has also made great efforts in improving corporate governance for domestically listed firms. In 2001, CSRC cracked down on the irregularities and fraudulent activities of listed firms and severely punished 30 firms. In August 2001, CSRC developed a guideline for establishing an independent director system. On January 7, 2002, CSRC formally issued the Corporate Governance Code for firms listed in China. The code is quite close to the international norm. 2002 was then designated as the year of corporate governance. In April, CSRC announced that it would send inspection teams out to all of the listed companies to see if the standards set by the Corporate Governance Code were met by the end of 2002. As a result, there were obvious improvements of corporate governance among listed firms in China after 2002.

3. ‘‘Market order’’ hypothesis Subrahmanyam and Titman (1999) argue that SIPs facilitate the development of stock markets. Megginson et al. (2004) find evidence to support the view that SIPs tend to occur in countries with less developed capital markets. However, if the domestic market is relatively undeveloped, as in the case of China, in which both the SHSE and SZSE were established only in the early 90s, it may not be able to absorb large and continuous IPO pressure, leading to price and trading suppression of existing stocks. A vivid example of this is the reform of the share–split system mentioned earlier. When the CSRC planned to sell state shares and make them tradable, it met with fierce resistance from various interest groups, which caused panic and a strong negative reaction from the market. This had to be cancelled after two failed attempts in 1999 and 2001. The final success of the reform implementation that started in 2005 had to do with nontradable shareholders giving certain compensation to tradable shareholders. One suggested explanation for the need of such compensation by Firth et al. (2010) and Li et al. (2011) rests precisely with the view that the tradable shareholders may suffer from the adverse price impact associated with a large increase in the supply of tradable shares. In view of this, we propose that the ‘‘market order hypothesis’’ is the motive behind foreign listings: diverting large IPOs to overseas markets releases issuing pressure in the domestic market.

1464

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

Table 1 Comparison of the Mainland China and Hong Kong Stock Exchanges. This table provides summary statistics of some variables of interest for Chinese domestic A-share and B-share firms, as well as the Hong Kong stock markets. Only the main-board listed firms are counted in the Hong Kong stock market (SEHK). The GEM listed firms are excluded. The variables examined here include the number of listed companies; proceeds raised through IPOs; market P/E ratios; total market capitalization; trading values, and market turnover rate (defined as total trading value / total market capitalization) for the sample period 1993–2007. The IPO funds raised, the market value and trading value of H-shares and the Hong Kong stock market are converted into RMB using the year-end official exchange rate between the Hong Kong Dollar and the Chinese RMB. Only the main-board listed firms are included in the statistics. The total market value in the A-share market and the B-share market is the value of the tradable shares. Year Number of Listed Companies A-share market B-share market H-share 6 14 16 22 37 39 41 44 47 53 62 71 80 99 105 Pure-A share 139 227 242 431 627 728 822 955 1024 1086 1147 1237 1240 1288 1390 A–B share 35 54 58 69 76 80 82 86 88 87 87 86 86 86 86 Pure-B share 6 4 12 16 25 26 26 28 24 24 24 24 23 23 23 A–H share 3 6 11 14 17 18 19 19 24 27 29 30 31 37 51 Pure-H share 3 8 5 8 20 21 22 25 23 25 33 41 49 62 54 Hong Kong 477 529 542 583 658 680 701 736 756 812 852 892 934 975 1048

Panel A: Number of listed firms 1993 177 41 1994 287 58 1995 311 70 1996 514 85 1997 720 101 1998 826 106 1999 923 108 2000 1060 114 2001 1136 112 2002 1200 111 2003 1263 111 2004 1353 110 2005 1357 109 2006 1411 109 2007 1527 109

Year

Fund raised (IPO) (billion) A-share B-share Hong Kong P/E ratios 66.972 57.837 42.322 107.530 265.143 40.887 157.985 479.249 62.159 107.539 222.187 293.757 314.059 529.657 556.680 H-share 6.058 11.018 2.171 7.347 34.311 2.215 4.548 54.958 5.910 17.899 49.144 42.511 144.462 297.688 72.904

As a percentage A-share IPO proceeds B-share 22.83% 71.39% 129.51% 18.55% 12.02% 4.23% 0.32% 1.70% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Hong Kong 268.73% 1064.83% 1931.15% 420.61% 401.25% 99.95% 310.04% 585.13% 110.37% 201.45% 489.93% 792.47% 5449.51% 317.59% 121.18% H-share 24.31% 202.86% 99.08% 28.74% 51.92% 5.41% 8.92% 67.10% 10.49% 33.53% 108.36% 114.68% 2506.69% 178.50% 15.87%

Market P/E ratios A-share 44.210 10.670 9.800 38.880 42.660 32.310 37.850 58.945 39.175 36.355 37.035 24.965 16.670 33.495 65.675 B-share 20.110 7.020 6.010 14.070 10.670 5.710 10.215 19.145 34.345 24.060 25.620 16.525 10.755 22.490 43.005 Hong Kong 21.630 10.710 11.440 16.690 12.100 10.660 26.730 12.800 12.180 14.890 18.960 18.730 15.570 17.370 22.470 H-share 5.863 12.216 16.147 17.378 15.319 12.699 14.655 8.120 8.039 10.122 14.896 20.739 17.700 19.278 24.384

Panel B: Issuing proceeds and market 1993 24.922 5.689 1994 5.432 3.878 1995 2.192 2.838 1996 25.565 4.742 1997 66.079 7.941 1998 40.909 1.729 1999 50.957 0.164 2000 81.904 1.391 2001 56.318 0.000 2002 53.383 0.000 2003 45.351 0.000 2004 37.069 0.000 2005 5.763 0.000 2006 166.772 0.000 2007 459.398 0.000 Year

Total market value (billion) A-share B-share Hong Kong H-share 13.56 22.28 17.77 33.90 52.07 35.84 44.68 90.42 105.89 137.10 428.31 483.52 1348.43 3452.65 4930.40

Trading value (billion) A-share 352.26 800.59 395.82 2105.33 3029.52 2340.95 3104.37 6029.82 3327.98 2714.82 3124.83 4156.70 3108.74 8920.31 45474.25 B-share 47.14 95.11 52.37 96.38 197.22 72.23 128.57 54.83 515.29 84.98 84.48 75.64 55.71 124.92 574.40 Hong Kong 911.56 1268.49 892.61 1518.31 4057.79 1818.06 2043.54 3236.44 2068.78 1696.22 2704.80 4194.46 4736.72 8552.75 20968.61 H-share 24.58 38.15 18.67 26.76 318.90 78.59 109.63 174.49 260.13 148.20 532.84 992.07 999.51 2588.38 7555.18

Turnover rate A-share 5.14 9.96 4.97 8.35 6.21 4.22 3.89 3.88 2.50 2.31 2.53 3.76 3.10 3.74 5.01 B-share 3.91 6.18 3.63 2.77 5.67 3.71 4.65 0.97 4.55 1.11 0.96 1.09 0.92 0.98 2.29 Hong Kong 0.41 0.55 0.35 0.41 1.18 0.64 0.41 0.64 0.50 0.45 0.46 0.60 0.55 0.63 1.05 H-share 1.81 1.71 1.05 0.79 6.12 2.19 2.45 1.93 2.46 1.08 1.24 2.05 0.74 0.75 1.53

Panel C: Market capitalization and turnover 1993 68.53 12.06 2213.95 1994 80.39 15.38 2325.48 1995 79.68 14.43 2535.22 1996 252.20 34.78 3737.04 1997 488.18 34.76 3429.86 1998 554.78 19.49 2844.71 1999 797.74 27.66 5042.38 2000 1555.06 56.68 5092.33 2001 1332.81 113.36 4121.83 2002 1177.06 76.62 3775.31 2003 1236.36 87.74 5820.07 2004 1106.64 69.09 7042.40 2005 1003.28 60.26 8543.75 2006 2384.67 127.13 13598.81 2007 9081.83 251.06 20023.05

3.1. Methodology and data Our key direct measure of the issuing pressure is the IPO issuing amount. The larger the IPO amount is, the bigger the pressure the capital market will face. Listing abroad could relieve the domestic market of such issuing pressure. Hence, we postulate that the IPO proceeds should be positively related to the overseas listing decision. We construct two proxies. ‘‘MP1’’ is the ratio of a firm’s IPO

issuing proceeds over the average A-share market capitalization in the previous three months. The higher the IPO proceeds relative to the whole market are, the bigger the potential pressure on the market there will be. ‘‘MP2’’ is the ratio of IPO issuing proceeds over the total A-share issuing amount for that year. If the total A-share issuing amount during a particular year can proxy for the issuing quota for the year, then MP2 is a measure of the Hshare issuing amount relative to the yearly quota. All H-share

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

1465

and A-share IPO firms are included in computing MP1 and MP2. Hshare IPO proceeds are converted into RMB in the computation. We then run the following probit model built on the one used in Pagano et al. (1998), hereafter called PPZ:

ProbðHi ¼ 1Þ ¼ a0 þ a1 LTAi þ a2 LEVi þ a3 CAPEXi þ a4 GSalesi þ a5 ROAi þ a6 MP1 ðor MP2i Þ þ Rbj INDj þ a7 RELPEi ðor Rct YRt Þ þ ui ; ð1Þ
The choice variable Hi takes the value of 1 if a firm issues Hshares and zero if a firm issues A-shares. Note that our decision is between listing in the Hong Kong market and listing in China’s market, whereas PPZ’s decision is between listing and not listing. As such, our model carries somewhat different meanings, and has effectively controlled for pure listing motives captured in their model. For instance, PPZ’s firm size proxy, LTA, captures the adverse selection cost of listing in their IPO decision model, whereas in our model, it captures the possible difference in the adverse selection cost between listing in Hong Kong and listing in China, and the economies of scale. Similarly, they use capital expenditure (Capex), leverage (LEV), and growth (GSales) to capture the funding needs of a private firm to go public. In our model, such need is assumed and, hence, these variables capture the need to be fulfilled through listing in the Hong Kong market only, and not in China’s market. That is, given the waiting time needed under the IPO quota system in China, firms with more urgent funding needs may have stronger incentives to choose Hong Kong listing in order to skip the queuing cost. ROA (Return-on-Asset) also appears in the PPZ model as a control variable for profitability. However, the percentage profitability requirement set by CSRC for overseas listing is the same as that for domestic listing, that is, ROE equals 10% (6% after 2001). So the profitability requirement per se does not guarantee that H-share firms are more profitable upon listing than their domestic counterparts. ‘‘INDj’’ is the dummy variable for industry j to see if some key industries supported by the government are more likely to list overseas. Bortolotti et al. (2002) find that telecommunications companies tend to seek foreign listings, whereas energy companies are seldom floated abroad. In our case, in contrast, the Chinese government has an industry policy to support energy, basic materials, and transportation companies, and more recently technology firms, to raise funds abroad. Finally, as mentioned before, foreign listings face the cost of low issuing P/E multiples in Hong Kong as compared to China. If the difference in issuing P/E multiples between Hong Kong and China becomes smaller, more mainland Chinese firms will be attracted to float their issues in Hong Kong. To control this possible effect, we use RELPEi, which is the three-month average P/E ratio in the Hong Kong market over that in the A-share market. It also controls for possible time-specific effects due to windows of opportunity and business cycles, as suggested in Ritter (1991).15 ‘‘YRt’’ is the IPO year dummy, an alternative way to control for time-specific effects. Our sample starts with 99 H-share firms and 1411 A-share firms listed at the end of 2006. We cut the sample period off at the end of 2006 because many firms had simultaneous issues in the A and H markets thereafter. Also, non-tradable shares would become tradable starting in 2007, due to the split-share reform mentioned in Section 2.3. These factors may have confusing effects if we include H-shares listed after 2006. We also exclude A-share firms that cross-list in either the B- or H-share market, since we examine the choice between listing in the
Pagano et al. (1998) use the median market-to-book ratio of the public companies in the same industry of the IPO firm to capture this effect in their probit model.
15

domestic or Hong Kong stock markets.16 We further drop 30 pure A-share firms that went public before 1993, because overseas listings were not possible then. We exclude 7 financial firms from the H-share firms and 13 from the pure A-share firms because their capital structure was not comparable to other firms. Finally, we exclude firms with missing data, which were mostly firms with less than two years of pre-listing leverage data. This results in 92 H-share firms and 1086 pure A-share firms. Since not all of the 1086 A-share firms are comparable to the Hshare firms, we construct two control samples out of these A-share firms to enhance the comparability. The first control group, the eligible sample, consists of 126 pure A-share firms eligible to list in Hong Kong, that is, they have a net worth of above RMB400 million, their previous year’s net profit was greater than RMB60 million, and they have issuing proceeds of more than US$50 (about RMB400) million. The second control group, the matched sample, consists of the pure A-share firms matched to the H-share firms by listing year, industry, and the closest size. We can find only 86 A-share firms that match with corresponding H-share firms. We obtain accounting and market data mainly from the WIND and RESSET Databases, supplemented by Datastream and Bloomberg. We measure LTA as the log of inflation-adjusted prelisting total assets; Capex as a firm’s investment on property, plant, and equipment divided by its total assets; LEV as the debt-equity ratio: and GSales as the percentage change of inflation-adjusted sales as a proxy for firm growth rate. We compute LTA, LEV, GSales, and ROA using three-year prelisting data and estimate the models by using the Huber–White-Sandwich robust estimator. We examine Pearson correlations and find that all of the correlations between independent variables are lower than absolute 0.4, except the one between TA and MP2 for the eligible sample, which is 0.53. We further calculate the variance inflation factors (VIF) for each independent variable and find that all VIFs are significantly lower than the typical threshold of 10. 3.2. Empirical results Panel A of Table 2 provides summary statistics and univariate test results of all independent variables in our probit model (1) for the 92 H-share firms and 126 eligible pure A-share firms. Largely consistent with our expectations, the mean and median total assets, MP1, MP2, LEV, and GSales of the 92 H-share firms are much larger than those of the 126 pure A-share firms, and the Wilcoxon test for the median difference is significant at the 1% level for all of these variables. On the other hand, the mean and median issuing PE ratios of the H-share firms are significantly smaller than those of the pure A-share firms. However, the H-share firms have lower mean and median ROA than the pure A-share firms, and the median difference is statistically significant at the 5% level. Similarly, the H-share firms have lower mean and median Capex than the pure A-shares, albeit the difference is not statistically significant. These could be due to the fact that the H-share firms are still much larger in terms of total assets. Notice that in Panel B, these variables are more comparable, in general, under the matched sample of 86 pure A-share and H-share firms, and the mean and median Capex for the H-shares are a bit higher than that of the pure A-shares, although the H-share firm size is still larger than the pure A-share firm size. On the whole, Table 2 indicates that the H-share firms are much larger in total assets and issuing amount, and higher in growth rate and leverage, but lower in issuing P/E and more similar in Capex ratios than their pure A-share counterparts.
16 All of the cross-listed H-share firms had their IPOs in the SEHK during our sample period, and more than half of the cross-listed B-share firms had their IPOs in the Bshare market.

1466

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

Table 2 Statistics for variables in probit regressions. The table presents statistics for the independent variables used in the probit regressions. The whole sample consists of 92 H-share firms and pure A-share firms listed in the period of 1993–2006 with available data. The eligible sample consists of 92 H-share firms and 126 pure A-share firms that met the CSRC requirements to list in Hong Kong. The matched sample consists of 86 H-share firms and 86 pure A-share firms matched by listing year, industry, and closest size. Panel A shows the descriptive and univariate test statistics for the eligible sample. Total assets (TA), sales growth rate (GSales), debt/equity (LEV), capital expenditures normalized by total assets (Capex/ TA), and return-on-assets (ROA) are the three-year average of the prelisting data. MP1 is the IPO issuing proceeds of a firm divided by the average market capitalization of the A-share market three months before the IPO. MP2 is the IPO issuing proceeds of a firm divided by the total A-share issuing proceeds during the same calendar year. PE refers to the issuing P/E ratio. Wilcoxon Z-statistics are used to test the median difference between the H- share group and the A-share group for each variable. Panel B shows the same statistics for the matched sample. For H-share firms, the currency is converted into RMB using the year-end official exchange rate between the Hong Kong Dollar and the Chinese RMB. Variables Group Min. Mean Median 1425.07 2952.98 0.153 0.128 0.117 0.312 0.015 0.032 0.177 0.305 0.599 0.662 0.086 0.064 16.715 11.060 1244.02 2672.37 0.133 0.141 0.111 0.316 0.013 0.031 0.187 0.306 0.618 0.651 0.069 0.072 16.360 11.462 Max. 86901.94 197253.45 0.769 0.749 1.049 5.000 0.215 0.950 6.000 2.139 0.910 0.910 0.600 0.594 52.370 43.900 25140.74 25140.74 0.835 0.749 1.225 1.670 0.215 0.250 3.000 2.139 0.858 0.880 0.441 0.460 50.858 43.900 Std dev. 8610.51 36432.87 0.171 0.148 0.152 1.038 0.025 0.199 0.912 0.377 0.146 0.163 0.084 0.105 6.318 6.426 4850.99 7822.28 0.168 0.152 0.223 0.540 0.030 0.080 0.496 0.390 0.150 0.163 0.060 0.096 6.421 6.573 Median diff. 1527.91 À0.025 0.195 0.017 0.127 0.063 À0.021 À5.655 Wilcoxon Z test 3.13*** À0.995 5.90*** 4.98*** 3.34*** 3.38*** À2.43** À8.18***

Panel A: Summary statistics of independent Total assets (million) A H Capex/TA A H MP1 (Â100) A H MP2 A H GSales A H LEV A H ROA A H PE A H Panel B: Summary statistics of independent Total assets (million) A H Capex/TA A H MP1 (Â100) A H MP2 A H GSales A H LEV A H ROA A H PE A H
* ** ***

variables (eligible sample) 221.13 3236.60 162.75 13804.76 0.002 0.223 0.002 0.183 0.025 0.160 0.009 0.711 0.003 0.022 0.002 0.113 À0.090 0.371 À0.090 0.391 0.192 0.565 0.190 0.633 0.014 0.104 0.010 0.099 7.660 18.329 3.315 12.191 variables (matched sample) 72.17 2735.91 146.13 6062.97 0.002 0.209 0.002 0.245 0.013 0.192 0.009 0.535 0.001 0.022 0.001 0.068 À0.170 0.324 À0.120 0.401 0.195 0.583 0.195 0.627 0.023 0.083 0.023 0.101 4.232 17.011 3.315 12.466

1428.35 0.008 0.206 0.018 0.120 0.034 0.003 À4.898

3.47*** 0.743 4.93*** 4.74*** 2.66** 1.82* 0.36 À5.81***

Statistical significance at the 10% level. Statistical significance at the 5% level. Statistical significance at the 1% level.

We hence run the probit regressions based on the two control samples. The results for the eligible sample are presented in Panel A of Table 3. Model 1 uses MP1 and Model 2 uses MP2 as the issuing pressure proxy. Both models use RELPE to control for time-specific effects. Models 3 and 4 are similar to Models 1 and 2, respectively, except that the time-specific effect is controlled by year dummies. In all of the models except for Model 4, MP1 and MP2 are positively related to the H-share listing, with MP1 being statistically significant at the 1% level and MP2 marginally significant at the 10% level. The size variable, LTA, is never significant in the regression models, which is inconsistent with the existing literature. PPZ argue that firm size captures the adverse selection cost of listing and find that it is positively related to a firm’s IPO decision. Saudagaran (1998), on the other hand, argues that size should have a positive impact on the overseas listing, as there is an effect of the economies of scale, due to the higher listing cost. He also finds that size is the most important determinant for overseas listing decisions. The fact that the H-share listing is more related to market pressure rather than to firm size for our eligible sample lends support to our market order hypothesis that CSRC uses H-share listings as a means to divert large IPO issues overseas, thus relieving domestic issuing pressure. For other control variables, we find that

the estimates for LEV and ROA are positive and significant, which are consistent with the argument that firms with an urgent need for cash and that are more profitable are more likely to list overseas. Note in Table 2 that the ROA for the pure A-shares is higher than that for the H-shares. However, after controlling for other variables, ROA is positively associated with the likelihood to list overseas. The estimates for GSales and Capex are insignificant. The industry dummy for transportation is positive and significant, for energy insignificant, and for basic materials negative and marginally significant, indicating that firms in the transportation industry are more likely to list overseas while others are not. Panel B presents the same probit regressions for the matched sample. Now, the market pressure proxies of both MP1 and MP2 are all statistically significant at the 1% level. The estimated coefficients for all of the control variables are qualitatively the same, except that the industry dummies are insignificant in all four models. This is understandable, as this is for the matched sample, so the industry effect is controlled to a great extent. However, LTA becomes positive and significant at the 5% level in Model 1.17
17 To address the possible multicollinearity between TA and MP1 (MP2), we further make TA orthogonal to MP1 (MP2) and use the residual in place of TA and repeat all the probit regressions. The results are largely the same and are available upon request.

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

1467

Table 3 Probit regression results. This table reports the results of the following probit model: ProbðHi ¼ 1Þ ¼ a0 þ a1 LTAi þ a2 ROAi þ a3 LEVi þ a4 GSalesi þ a5 CAPEXi þ a6 MP1i ðor MP2i Þþ Rbj INDj þ a7 RELPEi ðor Rct YRt Þ þ ei , Hi is a binary variable that takes the value of 1 if a firm issues H-shares and zero if a firm issues A-shares. LTA is the log of total assets. MP1 is defined as the IPO issuing amount divided by the average market capitalization of the A-share market for the previous three months. MP2 is defined as the IPO issuing amount/ total A-share issuing amount for the year. Both measures are used to proxy for the IPO issuing pressure on the domestic secondary market. GSales is the sales growth rate. LEV is the debt/equity ratio. ROA is return-on-assets. Capex is the capital expenditure divided by total assets. Materials, energy, and transport are dummies of the respective industries. RELPE takes the value of the yearly average Hong Kong market P/E over the average of the A-share market P/E in the previous three months. YR is the yearly dummy. The figures inside the parentheses are heteroscedasticity-consistent t-values. Model Constant LTA MP1 (Â100) 0.18 (3.96)*** 0.47 (1.66)* 0.21 (3.93)*** 0.43 (1. 60) 0.33 (3.59)*** 2.85 (3.25)*** 0.47 (3.83)*** 2.94 (3.04)*** MP2 GSales LEV ROA Capex Material Energy Transport RELPE YR Pseudo R2 0.265 0.219 Yes Yes 0.270 0.223 Obs

Panel A: Eligible sample 1 À0.72 0.03 (À1.62) (1.12) 2 À0.90 0.05 (À1.74)* (1.38) 3 À0.24 0.02 (À1.19) (0.49) 4 0.16 0.05 (0.95) 1.59 Panel B: Matched sample 1 À1.15 0.07 (À2.34)** (2.32)** 2 À0.44 0.04 (À0.70) (0.93) 3 À0.38 0.03 (À1.77)* (0.88) 4 À0.46 0.03 (À0.26) (0.70)
* ** ***

0.00 À0.02 0.00 À0.09 0.00 À0.08 0.00 À0.03 0.10 (1.25) 0.12 (1.46) 0.10 (1.24) 0.12 (1.46)

0.59 (2.85)*** 0.62 (2.90)*** 0.61 (2.94)*** 0.60 (2.80)*** 0.42 (1.79)* 0.38 (1.65)* 0.44 (1.89)* 0.38 (1.63)

0.86 (2.28)** 0.94 (2.42)** 0.85 (2.28)** 0.92 (2.36)** 1.54 (3.18)*** 1.48 (2.99)*** 1.49 (3.08)*** 1.47 (2.97)***

À0.14 À0.87 À0.11 À0.67 À0.16 À1.00 À0.10 À0.60 À0.15 (À1.19) À0.25 (À1.64) À0.20 (À1.53) À0.25 (À1.64)

À0.14 (À1.89)* À0.12 (À1.61) À0.14 (À1.93)* À0.12 À1.61 À0.06 (À0.66) À0.04 (À0.44) À0.05 (À0.61) À0.04 (À0.42)

À0.08 À0.71 À0.04 À0.32 À0.10 À0.87 À0.03 À0.25 0.15 (0.99) 0.19 (1.23) 0.14 (0.92) 0.19 (1.23)

0.22 (2.20)** 0.26 (2.52)** 0.21 (2.11)** 0.26 (2.52)** 0.15 (1.33) 0.16 (1.42) 0.14 (1.22) 0.16 (1.42)

0.43 (3.16)*** 0.40 (2.58)***

218 218 218 218

0.10 (0.62) 0.18 (0.92) Yes Yes

0.247 0.240 0.259 0.240

172 172 172 172

Statistical significance at the 10% level. Statistical significance at the 5% level. Statistical significance at the 1% level.

3.3. IPO impact on market prices and liquidity Our market pressure hypothesis builds on the argument that the SHSE and SZSE have low market depth, so that large IPO issues might depress the price and trading volume of the stocks trading in these exchanges. In this section, we provide evidence by comparing the response of the secondary market to the issuing of IPOs in the China and Hong Kong stock exchanges. First, we identify all of the trading days with IPOs in both the SHSE and SZSE, and compute the IPO proceeds for each of these days. If there are multiple IPOs on the same day, their proceeds are summed up. Next, we compute two measures of secondary market activity. VWRET is the market capitalization weighted market return on the IPO day and TURNOVER is the IPO day market turnover, defined as the total market trading value divided by total market capitalization.18 We then run the following regression model separately on each of these two variables:

VWRET=TURNOVER ¼ ai þ b1 LnðIPO Issuing ProceedsÞ or MP ð2Þ þ Control Variables þ ei
The independent variable is the natural log of the inflationadjusted IPO issuing proceeds on the same IPO day. We also use MP1, which standardizes the issuing proceeds as our alternative measure for issuing pressure. If the depth of the Chinese stock market is really low, then even domestic IPOs, which have a much smaller average issuing size than H-shares, may have a negative impact on secondary market return and/or turnover. We repeat the above exercise on all IPOs in Hong Kong to examine the response of the Hong Kong market to IPOs. As such, we can contrast the market responses of IPOs in China and Hong Kong. There are a total of 1274 observations for A-share firms and 620 for Hong Kong firms.19 The regression results are presented in Table 4.
18 All IPO firms are excluded in the VWRET and TURNOVER calculations and, hence, the two variables are computed based on all of the firms already listed in both the SHSE and SZSE. 19 About 620 observations contain all IPOs during the period of 1993–2006 in Hong Kong, and some of the firms were delisted during this period.

For the A-share market, the amount of the IPO proceeds tends to exert a downward pressure on the return in the A-share market on the IPO date, although the t-value is too low to claim statistical significance. However, the IPO proceeds are negatively correlated with the market turnover, which is the total market trading value scaled by the total market capitalization (excluding IPO volume). The coefficient is À0.0052, with a highly significant t-value of À3.98. This indicates that when A-share firms raise new capital in the domestic market, there is a negative impact on the trading of existing shares in the market. As H-share firms typically raise much more IPO proceeds, their negative impact on the domestic market would be conceivably larger if they were to float their shares in China instead of in Hong Kong. The results are qualitatively similar when we use MP1 as the independent variable. The coefficient for VWRET is not statistically significant, while for TURNOVER, the coefficient is À0.0733 with a t-value of À1.90, which is significant at the 10% level. On the other hand, H-share IPOs have no significant impact on both the return and turnover in the Hong Kong market, indicating that the Hong Kong market is much deeper than the Chinese Ashare market. We have also used equal-weighted market return in place of value-weighted market return, and three-day window instead of one-day window in Eq. (2).20 The results are qualitatively the same and not reported here to save space. We conclude that the results in Tables 2–4 are supportive of our market order hypothesis.

4. The governance hypothesis Other than relieving the pressure on domestic issues, the Chinese government also uses overseas listings as a means to force SOEs to conform to ‘‘international standards.’’ If listing H-share firms is driven by the motivation to establish a modern enterprise system in China, then H-share firms should exhibit better corporate
20 We do not use longer windows due to the overlapping of IPOs in the China market.

1468

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

Table 4 Impact of IPOs on Mainland China and Hong Kong Stock Markets. This table reports the estimated coefficients from the following cross-sectional regressions of market return or turnover on IPO issuing amounts. VWRETi ðor TURNOVERi Þ ¼ a0 þ a1 LnProceedsi ðor MP1i Þ þ a2 Lag VWRETi ðor Lag TURNOVERi Þ þ a3 Year dummies þ ei , VWRET is the valueweighted market return and TURNOVER/TOVR is the total market trading value scaled by total market capitalization on the IPO day. IPO firms are excluded in computing VWRET and TURNOVER. LnProceeds is the logarithm of the inflation-adjusted IPO issuing amount on the IPO day and MP1 is the IPO issuing amount divided by the average market capitalization of the market for the previous three months in the relevant market. The figures inside the parentheses are heteroscedasticity-consistent t-values. Variables A-share market VWRET Intercept LnProceeds MP1 (Â100) Lag VWRET Lag Turnover Year dummy No. obs Adjusted R-square
* ** ***

Hong Kong Market TOVR 0.0570*** (7.40) À0.0052*** (À3.98) VWRET À0.0003 (À0.17) TOVR 0.0266*** (À11.85) VWRET 0.0050 (1.12) À0.0005 (À0.69) TOVR 0.0059*** (5.86) À0.0001 (À0.67) VWRET 0.0026 (1.53) TOVR 0.0054*** (9.22)

0.0074 (1.13) À0.0013 (À1.08)

0.0470 (1.36) 0.1780*** (3.75) 1.9297*** (30.78) Included 1274 0.0437 Included 1274 0.6275 Included 1274 0.0443 0.1709*** (3.60)

À0.0733* (À1.90) 0.0873** (2.09) 1.9584*** (31.36) Included 1274 0.6239 Included 620 0.0064 0.6833*** (23.14) Included 620 0.552

À0.0184 (À1.3) 0.0850** (2.04)

À0.0016 (À0.55)

0.6814*** (23.05) Included 620 0.0084 Included 620 0.5519

Statistical significance at the 10% level. Statistical significance at the 5% level. Statistical significance at the 1% level.

governance than their domestic counterparts after the listing, and this is what we examine in this section. 4.1. Governance and cross-listing premium In testing our governance hypothesis, we face the problem that there is no single, convincing, catch-all proxy for corporate governance. Conventional proxies, such as ownership structure, CEO duality and compensation, board size and independence, earnings quality, related party transactions, and transparency and disclosure, are noisy, correlated with each other, and subject to many different interpretations. In fact, we have no reliable data on the governance structure of SOEs prior to their IPOs, since all Chinese firms undergo restructuring before they are listed in domestic or foreign markets. Therefore, we focus on comparing the ‘‘cross-listing premium’’ of A–H firms and the pure A-share firms in the A-share market, particularly before and after 2002, the year of corporate governance in China. Stulz (1999) and Coffee (1999, 2002) argue that foreign firms listed in the US market are subject to more stringent listing requirements, enabling investors to use increased disclosure requirements to monitor firms more closely and to take advantage of better protection from stronger legal systems. As such, these firms might earn a cross-listing premium in the form of a higher firm valuation relative to their domestic peers. Nenova (2003), Doidge et al. (2004), and Dyck and Zingales (2004) find supporting evidence when using the market-to-book ratio (MBR) as the proxy for firm valuation. We know that a firm’s H-share is usually traded at a discount relative to its dual-listed A-share. If we still observe a cross-listing premium, and this premium changes as the relative corporate governance quality changes, then it offers strong support to our governance argument. To look for evidence, we examine 37 H-share firms that also issued A-shares in the domestic market as of 2006. We expect the MBR of this dual-listed A-share to be higher than the pure A-share firms.21 We should also observe a significant
21 Sun et al. (2009) have documented such premiums for dual-listed A–H shares over the pure A-share firms. Tong and Yu (2012), on the other hand, examine the price discount of B-shares relative to their corresponding A-shares, and show that governance quality has strong explanatory power of such a price discount phenomenon, which supports our argument that governance quality affects firm valuation in another dimension.

decrease in this premium after 2002, when corporate governance dramatically improved in China. By examining this premium and its change after China’s corporate governance year, we can avoid the difficulties of searching for appropriate governance proxies and having no reliable data on the governance structure of Chinese firms before listing. 4.2. Methodology As mentioned before, the A-share, B-share, and H-share markets are segmented, and the B-share and H-share prices are traded at a large discount relative to their A-share counterparts. To measure the possible cross-listing premium in the A-share market, we use the A-share price of the cross-listing firms to compare against the pure A-share companies so that we can determine how A-share market investors evaluate these two groups of firms differently. No shares dual-listed in the B-share and Hong Kong markets. Since Chinese firms have both tradable and nontradable shares during our sample period, we compute the yearly MBR of equities using only the tradable shares for each and every A–H firm and the relevant pure A-share firms in our sample. We then run the following pooled regression model:

MBRit ¼ a0 þ a1 LTAit þ a2 ROAit þ a3 LEVit þ a4 GSalesit þ a5 DIVit þ a6 PostGOVt þ a7 Hi þ a8 PostGOVÃ Hi t þ a9 Trdi t þ a10 IndðMBRit Þ þ Rbj YRt þ eit : ð3Þ

where Size (LTA), Profitability (ROA), Growth (GSales), Leverage (LEV), and Dividend Payout Ratio (DIV) are conventional control variables used in MBR analysis.22 The proportion of shares that is tradable (Trd) is used to control the unique share-segmentation phenomenon. Our key testing variables are H-share Dummy (H), which takes the value 1 if the firm is dual-listed in the Hong Kong market and zero if it is a pure A-share firm; Post Corporate Governance Year Dummy (PostGOV), which takes the value 1 if it is from 2002 onwards and zero otherwise; and the interactive term PostGOV Ã H. If there is an overseas listing premium, H should be significantly positive as MBR should be higher for AH firms. If corporate governance improved, in general, for all A-share firms after 2002, then PostGOV should be significantly positive as it will raise MBR for all A-share
22

See Doidge et al. (2004), for example.

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

1469

Table 5 Summary statistics of the variables used in the governance regressions. The table presents statistics for the variables used in the regression, examining the effect of the corporate governance change that occurred in 2002. Panel A presents the statistics of the yearly observations of 37 firms issuing both H- and A-shares (A–H shares) and the yearly observations of 126 pure A-share firms that met the SEHK listing requirement. The variables are market-to-book ratio (MBR) of tradable shares in the A-share market, log of total assets (LTA), return-on-assets (ROA), debt/equity ratio (LEV), sales growth rate (GSales), dividend payout ratio (DIV), and tradable share ratio (Trd). Wilcoxon Z-statistics are used to test the median difference between the A–H share group and the A-share group for each variable. Panel B further presents the statistics of the yearly observations for the 37 A– H firms and the 37 pure A-share firms matched to the A–H firms by industry, issuing year, and size. The pure A-share firms are matched to A–H firms by industry, listing year, and closest size. A/H N Mean Min. À4.625 À4.625 17.922 19.480 À2.227 À0.705 0.075 0.075 À0.798 À0.560 0.000 0.000 0.098 0.226 0.486 À4.625 19.484 19.480 À0.236 À0.705 0.075 0.075 À0.694 À0.560 0.000 0.000 0.098 0.226 Median 2.225 2.526 21.499 22.131 0.042 0.032 0.482 0.474 0.155 0.165 0.385 0.304 0.300 0.458 2.361 2.526 21.812 22.131 0.051 0.032 0.506 0.474 0.225 0.165 0.299 0.304 0.356 0.458 Max. 22.499 16.496 23.338 23.338 0.666 0.235 3.599 1.159 2.111 2.111 1.538 1.538 0.858 0.798 20.599 16.496 23.338 23.338 0.284 0.235 0.978 1.159 2.111 2.111 1.538 1.538 0.836 0.798 Std. 1.594 2.279 0.809 1.084 0.110 0.077 0.254 0.214 0.276 0.284 0.328 0.297 0.136 0.112 2.039 2.279 0.990 1.084 0.056 0.077 0.185 0.214 0.405 0.284 0.302 0.297 0.140 0.112 Median diff. 0.301 0.632 À0.011 À0.008 0.010 À0.080 0.158 Wilcoxon Z test 4.907*** 5.731*** À3.892*** À0.379 0.272 À2.789*** 14.441***

Panel A: A–H shares versus qualified pure A-shares MBR A 852 2.521 AH 301 3.155 LTA A 857 21.598 AH 301 21.984 ROA A 857 0.041 AH 301 0.028 LEV A 857 0.485 AH 301 0.488 GSales A 856 0.185 AH 300 0.195 DIV A 790 0.399 AH 266 0.336 Trd A 850 0.332 AH 288 0.457 Panel B: Matched sample MBR A AH LTA A AH ROA A AH LEV A AH GSales A AH DIV A AH Trd A AH
⁄

301 301 301 301 301 301 301 301 299 300 281 266 287 288

2.888 3.155 21.757 21.984 0.050 0.028 0.504 0.488 0.288 0.195 0.333 0.336 0.372 0.457

0.166 0.319 À0.020 À0.032 À0.060 0.005 0.102

2.208** 2.695*** À5.069*** À1.445 À3.027*** 0.175 7.983***

Statistical significance at the 10% level. Statistical significance at the 5% level. *** Statistical significance at the 1% level.
**

firms. If the overseas listing premium is due to better corporate governance, then PostGOV Ã H should be significantly negative as the difference in corporate governance standards has been narrowed. Industry average MBR (Ind(MBR)) is used to control for the possible industry effect, while YR dummies are used to control the possible year-specific effect. As before, we also construct two samples. The eligible sample consists of 37 H-share firms also cross-listed in the A-share market during the period 1993–2006, and 126 pure A-share firms eligible to list overseas during the same period. The matched sample consists of the same 37 A–H firms and 37 pure A-share firms matched to the A–H firms by listing year, industry, and closest firm size. Note that the independent variables used in the probit analysis in the previous section are constructed by taking the average of up to three years pre-IPO data, while the variables used here are the annual data after listing. In addition, the data used in the probit analysis are cross-sectional, while the data used here are pooled across firms and over years. Table 5 presents the summary statistics for the variables used in the regression equation (3). Panel A shows 301 firm-year observations for A–H firms and 857 firm-year observations for eligible pure A-share firms. The mean and median MBR, LTA, and Trd (the ratio of tradable shares) of the A–H firms are much higher than that of the pure A-share firms, and the median difference between them are all significant at the 1% level. On the other hand, the mean and median ROA and DIV (dividend payout ratio) of the A–H firms are lower than that of the pure A-share

firms, and the median difference is also significant at the 1% level. The difference in LEV and GSales between the A–H and pure A-share firms are not statistically different. All variables have quite wide variations, as evidenced by comparing the standard deviation to the mean. We also notice that the minimum MBR for both of the A–H and pure A-share firms is negative. We set all negative MBR to zero in the regressions.23 Panel B shows 301 firm-year observations for the matched 37 A–H and pure A-share firms. The comparisons of the variables between the two groups are similar to those reported in Panel A. The only difference is that the median difference between GSales becomes negative and significant, while the median difference between DIV becomes insignificant. The Pearson correlation coefficients between all of these variables are lower than 0.3 in absolute value. The VIF for all of these variables is also significantly lower than 10. Hence, multicollinearity should not be a serious problem.

4.3. Results The regression results of Eq. (3) are reported in Table 6. In Model 1 of the eligible sample, we see that the coefficient for H is 0.875 and is highly significant with a t-value of 9.46, indicating
23 Altogether there are 7 negative MBR observations from one A-H firm and two pure A-share firms.

1470

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

Table 6 Impact of corporate governance improvement on H-share premium (37 A–H shares listed during 1993–2006, versus the control sample). This table reports the results of the following regression: MBRit ¼ a0 þ a1 LTAit þ a2 ROAit þ a3 LEVit þ a4 GSalesit þ a5 DIVit þ a6 PostGOVt þ a7 Hi þ a8 PostGOVÃ Hi þ a9 Trdit þ a10 IndðMBRit Þ þ Rbj YRt þ eit . MBR is the t market-to-book ratio of tradable shares in the A-share market. LTA is the log of total assets; ROA is return-on-assets; LEV is the debt/equity ratio; GSales is the sales growth rate; DIV is the dividend payout ratio; PostGOV is an indicator variable, which is set to 1 after 2002 and zero otherwise; H is also an indicator for A–H firms; PostGOV Ã H is an interactive term; Trd is tradable share ratio; Ind(MBR) is the industry median of MBR each year; and YR is the year dummy. The figures inside the parentheses are heteroscedasticity-consistent t-values. Variables MBR (eligible sample) Model 1 Intercept LTA ROA LEV GSales DIV PostGOV H PostGOV Ã H Trd Ind(MBR) YR Obs. Adjusted R-square
* ** ***

MBR (matched sample) Model 3 8.442 (8.29)*** À0.468 (À9.45)*** 12.257 (12.17)*** 2.392 (9.6)*** À0.172 (À1.22) À0.042 (À0.36) 0.568 (5.54)*** 0.987 (10.57)*** Model 4 8.323 (8.17)*** À0.466 (À9.43)*** 12.503 12.34)*** 2.424 (9.73)*** À0.192 (À1.37) À0.042 (À0.37) 0.662 (5.89)*** 1.201 (8.54)*** À0.343 (À2.03)** À1.903 (À6.40)*** 0.886 (23.40)*** NO 1158 0.4945 Model 1 9.309 (6.28)*** À0.481 (À7.05)*** 13.088 (7.12)*** 2.509 (6.05)*** À0.005 (À0.02) À0.336 (À1.46) Model 2 13.53 (7.67)*** À0.628 (À8.68)*** 11.442 (6.13)*** 2.061 (4.86)*** 0.003 (0.01) À0.39 (À1.70)* Model 3 10.908 (7.36)*** À0.575 (À8.31)*** 11.305 (6.18)*** 2.089 (5.05)*** À0.02 (À0.10) À0.39 (À1.73)* 0.866 (5.07)*** 0.81 (5.89)*** Model 4 10.85 (7.35)*** À0.581 (À8.43)*** 11.862 (6.45)*** 2.169 (5.24)*** À0.048 (À0.26) À0.434 (À1.93)* 1.153 (5.45)*** 1.168 (5.62)*** À0.585 (À2.29)** À2.375 (À4.49)*** 1.002 (16.17)*** NO 602 0.4486

Model 2 10.356 (9.02)*** À0.502 (À9.92)*** 12.258 (12.07)*** 2.4 (9.48)*** À0.142 (À1.01) À0.061 (À0.53)

7.573 (7.42)*** À0.406 (À8.29)*** 13.03 (12.88)*** 2.572 (10.26)*** À0.15 (À1.05) À0.074 (À0.63)

0.875 (9.46)***

1.056 (10.86)***

0.752 (5.35)***

0.834 (5.98)***

À1.348 (À4.71)** 0.753 (24.96)*** NO 1158 0.4783

À2.143 (À6.72)** 0.795 (10.04)*** Include 1158 0.495

À1.866 (À6.28)*** 0.883 (23.30)*** NO 1158 0.493

À1.548 (À2.97)*** 0.824 (15.60)*** NO 602 0.4172

À2.577 (À4.55)*** 0.933 (7.81)*** Include 602 0.4437

À2.289 (À4.32)*** 1.000 (16.07)*** NO 602 0.444

Statistical significance at the 10% level. Statistical significance at the 5% level. Statistical significance at the 1% level.

that there is an H-share premium. Adding year dummies in Model 2 makes the H-share premium even higher (with a coefficient of 1.056) and more significant (with a t-value of 10.86), indicating that the H-share premium is not affected by the year-specific effect. We take off the year-specific dummies but add the PostGOV dummy, which takes the value 1 from 2003 to 2006 in Model 3.24 We find that both the H-share premium and PostGOV dummy are positive. Specifically, the H dummy coefficient is 0.987 and the PostGOV dummy coefficient is 0.568, and both are highly significant. This suggests that the significant improvement in corporate governance in 2002 raised the MBR for all firms listed in the Ashare market. We further add the interactive dummy PostGOV Ã H in Model 4. We find that while the coefficients for H and PostGOV are still positive and highly significant, the coefficient for the interactive dummy is À0.343 with a t-value of À2.03, which is significant at the 5% level, indicating that the general improvement in the corporate governance of the A-share firms after 2002 has reduced the valuation premium of A–H shares over their pure Ashare counterparts. These results lend support to our hypothesis that listing in Hong Kong creates a premium, and that this premium is related to the better corporate governance of these H-share firms. The estimates for the control variables are very consistent across the four models, in terms of sign and statistical significance. Specifically, Size has a negative and highly significant impact on MBR, indicating that large firms may have a more serious agency problem. ROA has a positive and highly significant
Including both year dummies and PostGOV in the regression led to perfect multicollinearity.
24

impact on MBR, as high profit may foster the stock price. LEV has a positive and highly significant impact on MBR. This probably indicates that high leverage may reduce the agency problem due to stringent monitoring and less free cash flow. GSales and Div have negative coefficients but are statistically insignificant, indicating that the current sales growth and dividend payout ratio have no impact on MBR. The tradable share ratio (Trd) has a negative and significant impact on MBR. This is due to the fact that higher Trd means more supply of the shares. As expected, the industry average MBR is significantly positive on MBR for individual firms. The results for the matched samples are qualitatively the same, except that DIV is negative and marginally significant for Models 2–4. We conducted several robustness checks. First, it is possible that the above results are influenced by firms listed after 2002. We repeat our regressions by excluding firms listed after 2002 (inclusive), leaving 24 A–H firms and 86 pure A firms in the eligible sample, and 24 pairs in the matched sample. The results on this pre-2002 listing sample are qualitatively the same. As mentioned before, Chinese firms have tradable and non-tradable shares, and our tests are on tradable shares only. We aim to examine if including non-tradable shares affects our results or not, and hence re-compute MBR by including non-tradable shares as follows:

MBR ¼fA-share price xðtradable A þ H sharesÞ þ Net Asset Value x nontradable shares þ book v alue of liabilitiesg=fbook v alue of total assetsg ð4Þ

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

1471

Table 7 Summary statistics of the variables used in listing effect regressions. The table presents statistics for the variables used in the regression examining the effect of the B-share listing on the MBR. Panel A presents the statistics of yearly observations of 28 firms first issuing A-shares and then B-shares, and the yearly observations of 1086 pure A-share firms. The variables are market-to-book ratio (MBR) of tradable shares in the A-share market, log of total assets (LTA), return-on-assets (ROA), debt/equity ratio (LEV), sales growth rate (GSales), dividend payout ratio (DIV), and tradable share ratio (Trd).Wilcoxon Z-statistics are used to test the median difference between the A–B share group and the A-share group for each variable. Panel B further presents the statistics of yearly observations for the 28 A–B firms and the 28 pure A-share firms matched to the A–B firms by industry, issuing year, and closest size. A/B N Mean 3.728 4.725 20.174 20.746 0.029 0.031 0.475 0.480 0.199 0.161 0.249 0.216 0.387 0.454 3.291 4.725 20.634 20.746 0.040 0.031 0.506 0.480 0.255 0.161 0.274 0.216 0.354 0.454 Min. À2.754 0.919 18.126 19.003 À0.355 À0.245 0.073 0.048 À0.667 À0.529 0.000 0.000 0.136 0.096 À4.868 0.919 18.963 19.003 À0.309 À0.245 0.078 0.048 À0.537 À0.529 0.000 0.000 0.072 0.096 Median 3.053 3.379 20.116 20.726 0.038 0.032 0.463 0.490 0.149 0.137 0.000 0.000 0.370 0.447 2.778 3.379 20.651 20.726 0.044 0.032 0.519 0.490 0.168 0.137 0.156 0.000 0.346 0.447 Max. 20.039 28.986 22.961 22.964 0.184 0.203 1.596 0.892 2.035 1.778 1.481 1.046 0.842 0.854 15.818 28.986 22.853 22.964 0.185 0.203 1.009 0.892 2.641 1.778 1.327 1.046 0.862 0.854 Std. 2.965 4.332 0.913 0.813 0.074 0.075 0.223 0.169 0.367 0.328 0.335 0.293 0.126 0.157 2.398 4.332 0.748 0.813 0.064 0.075 0.184 0.169 0.445 0.328 0.325 0.293 0.145 0.157 Median diff. 0.326 0.610 À0.006 0.027 À0.012 0.000 0.077 Wilcoxon Z test 4.622*** 12.446*** À0.872 2.163** À1.987** À1.251 8.686***

Panel A: A–B shares versus all pure A-shares MBR A 10,808 AB 392 LTA A 10,808 AB 392 ROA A 10,808 AB 392 LEV A 10,808 AB 392 Gsales A 10,808 AB 392 DIV A 10,808 AB 392 Trd A 10,808 AB 392 Panel B: Matched sample MBR A AB LTA A AB ROA A AB LEV A AB Gsales A AB DIV A AB Trd A AB
* ** ***

392 392 392 392 392 392 392 392 392 392 392 392 392 392

0.601 0.075 À0.012 À0.029 À0.031 À0.156 0.101

5.225*** 1.907* À2.757*** À1.985** À2.663*** À2.651*** 8.486***

Statistical significance at the 10% level. Statistical significance at the 5% level. Statistical significance at the 1% level.

In a limited number of transfer transactions of non-tradable shares especially approved by the CSRC, the transfer price is from 70% to 130% of the net asset value (see Huang and Fung, 2004). Hence, it is reasonable to assume that they are traded at the net asset value. Using this MBR measure, we repeat Eq. (3) regressions and obtain similar results. We further use PE ratio as a measure for valuation (which reduces the number of observations by 20%, due to negative values and outliers), and we obtain, again, qualitatively the same results.25 All in all, our evidence is consistent with our second hypothesis that the government uses overseas listing as a means to help establish a modern enterprise system in China.26

4.4. Endogeneity problem Our governance hypothesis may have an endogeneity problem, as firms chosen to list in Hong Kong may have better corporate governance to begin with. However, we argue that the problem is less serious than it appears. First, most of the prelisting firms in China did not have corporate governance that was close to the international standards. Otherwise, the Chinese government would
25 The results of these robustness checks are not reported to save space, but are available upon request. 26 In an earlier version of this paper, we also compared company performance in terms of ROS, ROA, ROE, and earnings quality, using the measure proposed by Leuz et al. (2003) and find that A–H firms do have better performance and earnings quality than pure A-share firms.

not need to spend so much effort to reform ailing SOEs. Because of the more thorough pre-listing restructuring to meet the more demanding Hong Kong listing requirements, the corporate governance of the H-share firms is improved. Also, the H-share firms must stick to the higher corporate governance standards after listing, because these firms are exposed to a more stringent monitoring and law enforcement environment. Hence, it is reasonable to believe that the valuation premiums found in Table 6 are, at least partly, due to the cross-listing effect. Of course, one can never preclude the possibility that some firms with a better governance structure choose to list or cross-list in foreign markets. But even that, we want to emphasize, need not invalidate our proposition that the Chinese government wants its SOEs to practice good governance through listing abroad. If overseas listings can help distinguish these firms from other firms with poorer governance structures and get higher valuations, then it can also help to establish modern enterprises in China by showing that good corporate governance is valuable. That said, we still try to tackle this endogeneity problem empirically. An ideal setting to address the issue is to identify a group of firms that were first listed in the A-share market and then crosslisted in Hong Kong. If the premium is really driven by bonding after the listing, we should observe that these firms earn premiums only after the cross-listings and not before. But as said earlier, all of the H-share firms during our sample period had their IPOs done first in Hong Kong and, hence, we have to take a less direct route and choose a sample of firms listed in the A-share market first and cross-listed in the B-share market later to do the test. Arguably,

1472

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

Table 8 Listing effect on cross-listing premium. This table reports the pooled regressions of the market valuation measures of firm characteristics for the 28 A–B firms and the two control samples of the pure A-share firms during the period of 1993–2006.MBRit ¼ a0 þ a1 LTAit þ a2 ROAit þ a3 LEVit þ a4 GSalesit þ a5 DIVit þ a6 Pre-Xlistt þ a7 Post-Xlistt þ a8 Trdit þ a9 IndðMBRit Þ þ Rbj YRt þ eit . The eligible sample here contains 1086 pure A-share firms, while the matched sample has 28 pure A-share firms matched to the A–B firms by industry, listing year, and the closest size. MBR is the market-to-book ratio of tradable shares in the A-share market. LTA is the log of total assets; ROA is return-on-assets; LEV is the debt/equity ratio; GSales is the sales growth rate; and DIV is the dividend payout ratio. Pre-XList and Post-XList are, respectively, dummy variables that take the value of one for the period before and after an A-share firm cross-listed in the B-share market, and zero otherwise. Trd is tradable share ratio; Ind(MBR) is the industry median of MBR each year; and YR is the year dummy. The figures inside the parentheses are heteroscedasticity-consistent t-values. Variables MBR (eligible sample) 1993–2002 Intercept LTA ROA LEV GSales DIV Pre-XList Post-XList Trd Ind(MBR) YR Obs. Adjusted R-square
* ** ***

MBR (matched sample) 1993–2006 22.77 (31.20)*** À1.113 (À37.40)*** À0.242 (À0.55) 1.719 (12.54)*** 0.420 (5.84)*** À0.581 (À7.35)*** 0.143 (0.23) 1.983 (13.86)*** À3.403 (À16.72)*** 0.865 (17.13)*** Include 11,200 0.3568 1993–2002 35.74 (8.13)*** À1.893 (À9.47)*** 6.254 (2.34)** 8.802 (8.80)*** À0.552 (À1.38) À0.262 (À0.57) À0.263 (À0.35) 2.473 (8.26)*** À7.330 (À7.76)*** 1.013 (4.30)*** Include 560 0.4519 1993–2006 37.953 (10.22)*** À1.961 (À12.17)*** 3.112 (1.35) 8.834 (10.99)*** À0.551 (À1.68)* 0.152 (0.359) À0.042 (À0.05) 2.612 (10.32)*** À7.182 (À8.52)*** 0.886 (3.53)*** Include 784 0.4594

36.21 (36.95)*** À1.792 (À40.30)*** À0.673 (À1.02) 4.480 (20.93)*** 0.387 (3.97)*** À0.514 (À4.82)*** 0.223 (0.35) 2.172 (11.82)*** À4.801 (À15.71)*** 0.758 (14.19)*** Include 6247 0.4034

Statistical significance at the 10% level. Statistical significance at the 5% level. Statistical significance at the 1% level.

the B-share market also has more stringent listing requirements than the A-share market, though not as stringent as the Hong Kong market.27 Without reporting the details here, a test on the full sample of 86 A–B pairs does show cross-listing premiums. If our governance hypothesis is correct, and if the B-share market has better governance quality than the A-share market, we should observe a similar valuation premium for B-shares and, more importantly, such premium should occur only after the company has cross-listed to the B-share market. Among the 86 A–B dual-listed firms, there are 42 firms that listed first in the B-share market; the remaining 44 firms listed first in the A-share market. Further dropping of the firms with A and B shares listed in the same year, we have 28 firms that listed first in the A-share market and subsequently cross-listed in the B-share market. Using these 28 firms, we run the following regression over the sample period of 1993–2006:

MBRit ¼ a0 þ a1 LTAit þ a2 ROAit þ a3 LEVit þ a4 GSalesit þ a5 DIVit þ a6 Pre-Xlistt þ a7 Post-Xlistt þ a8 Trdit þ a9 IndðMBRit Þ þ Rbj YRt þ ei ð5Þ

27 There are special requirements on the B-share companies regarding information disclosure. First, the companies have to prepare disclosures in both Chinese and English, and must publish them in local and overseas newspapers (Articles 35 and 37, Operating Rules on Issuing B-shares by Joint Stock Limited Companies). The interim and annual reports are prepared according to the International Accounting Standard, not the China Accounting Standard. Any material differences between the two standards must be explained in the financial reports (Article 36, Operating Rules on Issuing B-shares by Joint Stock Limited Companies). The companies can hire stateapproved foreign audit firms to audit the company accounts (Article 42). So, in effect, all B-share companies receive audit services from one of the Big Five (now Big Four) international audit firms. These special provisions might pressure the B-share companies to observe and to conform to international governance standards more closely. Also, foreign investors can be represented on the board of directors for the Bshare firms, but usually not for the pure A-share firms. Hence, it is reasonable to assume that the B-share market would have a better corporate governance environment.

We use the regression to contrast the 28 firms against all pure A-share firms, as all A-share firms are eligible to list on the B-share market as long as they can generate enough foreign exchange to pay cash dividends. In addition, we also contrast the 28 firms with a matched sample of 28 pure A-share firms (matched by industry, listing year, and closest size). All of the variables are defined as those in Eq. (3), except that we have two new cross-listing dummy variables. ‘‘Pre-XList’’ and ‘‘Post-XList’’ take the value of one, respectively, for the period before and after firm i cross-listed in the B-share market, and are zero otherwise. If higher valuation indeed comes from cross-listing, the coefficient of Post-XList, a2, will be significantly positive, while that of Pre-XList, a1, should be insignificantly different from zero. Table 7 provides the summary statistics for all but the dummy variables used in Eq. (5). Panels A and B contrast the 28 A–B share firms with the 1086 pure A-share firms (eligible sample) and the 28 matched pure A-share firms, respectively. In general, we find that MBR, size, leverage, and tradable share proportion for the A– B firms are significantly larger/higher than those for the pure A firms, although the size in Panel B is more comparable between the two groups. The A–B firms have lower ROA and dividend payout than pure A firms, but the differences are statistically insignificant in Panel A and significant in Panel B. The only thing inconsistent across the two panels is leverage. It is statistically significantly higher for A–B share firms than for pure A-share firms in Panel A, but it is statistically significantly lower in Panel B. On the whole, the summary statistics reported in Table 7 are similar to

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474

1473

those reported in Table 5. A further examination of the correlation matrix and VIF for all of these variables shows that multicollinearity is not a serious problem.28 Table 8 presents the regression results of Eq. (5) for both of the eligible and matched samples. For each sample, we further run regressions using the data for the full sample period (1993– 2006), as well as for the period from 1993 to 2002. This is because no firm listed in the B-share market after 2000. Although we put year dummies in the regressions, the estimated results can still be affected by noise if the post-cross-listing period is too long. We find evidence in Table 8 to support our claim that endogeneity is not a serious problem affecting our results. The Pre-XList dummy coefficients have small t-values of no statistical significance in all four regressions, indicating that before cross-listing in the B-share market, there is no value premium between A–B share firms and the pure A-share firms. However, the Post-XList dummy coefficients are positive and highly significant at the 1% level in all regressions, indicating that the A–B share firms have a more positive MBR than their pure A-share peers in the post-listing period and their own pre-listing past. These results are clear evidence that firms enjoy a higher valuation subsequent to, but not before, their cross-listings, thereby indicating that the premium is a cross-listing phenomenon.29 The estimated coefficients for the control variables are largely consistent across the sample periods and control samples. They are also qualitatively similar to those reported in Table 6. Since our focus is on the cross-listing premium, we do not discuss them in detail here to save space. We also performed robustness checks. First, we run regressions by omitting observations after 2001, as the B-share market was partially open to domestic investors. Second, we use the PE ratio in place of MBR as the dependent variable. The results are qualitatively the same. Yet, we admit that whether A–B firms can be good proxies for A–H firms is still debatable. 5. Conclusion We observe that the Chinese government continues to sell its SOEs in the Hong Kong market, even though the phenomena look inconsistent with traditional finance theory. We reconcile this inconsistency by providing new perspectives. We suggest a market order hypothesis and a governance hypothesis, which are supported by evidence from testing a sample of 92 Chinese firms listed in Hong Kong and several control samples of purely domestically listed Chinese firms during the period of 1993–2006. Consistent with the market order argument that the relatively young Chinese stock market has difficulties meeting the equitycapital-raising demand of many SOEs that are heavily in debt, so they have to tap the Hong Kong capital market, we find that the Chinese domestic market as a whole responds negatively to domestic IPO issues, and that H-share firms are large in terms of total assets, sales, and issuing proceeds. Listing these firms overseas can help divert the supply and help to prevent a possible crash in prices in the domestic market. We also find that H-share firms have higher pre-listing leverage than their domestic counterparts. This indicates that H-share firms have a more urgent need for equity capital before their listing and, thus, would be willing to list overseas despite issuing at a large discount. Consistent with the governance argument that firms that crosslist in the markets and have high accounting, legal, and governance standards effectively bond themselves to the higher standards to help improve their credibility and prestige among international
The results are not reported to save space. Due to the lack of good instrument variables, we do not use two-stage least squares (2SLS) and Heckman’s (1979) two-step procedure to correct for possible selfselection bias.
29 28

as well as domestic investors, we find that the bonding premium for firms issuing both H- and A-shares relative to the control samples of the pure A-share firms decreased after 2002. CSRC announced and enforced a series of tough regulations and measures in 2001 and 2002, which dramatically raised the corporate governance standards for firms listed in China. When the general governance standard improved after 2002, the premium enjoyed by the cross-listing firms through bonding with the Hong Kong market was reduced. Since the second half of 2005, the China A-share market has gone through a drastic change. The size of the SHSE and SZSE markets combined has jumped from US$0.39 trillion to US$2 trillion. The total number of investor accounts in these two markets has also increased from around 73 million to nearly 100 million. As such, our market pressure hypothesis will predict a slowdown of Chinese SOEs’ foreign IPOs. In fact, a trend of ‘‘reverting back’’ to the A-share market started a year ago, in which pure H-share companies and red chip companies planned to issue back A-shares in the Mainland. On the other hand, as long as the Hong Kong market maintains its ‘‘governance premium’’ over China’s domestic market, our governance hypothesis will predict a continuation of Chinese SOEs listing in Hong Kong. A recent discussion among the China and Hong Kong government officials in charge of economic and financial affairs concerns allowing small and medium mainland Chinese enterprises to list in Hong Kong. It will be interesting to see how the various forces studied in this paper will shape the future development of this cross-listing phenomenon between China and Hong Kong. Acknowledgements For helpful comments and discussion, we would like to thank Warren Bailey, William Megginson, Jay Ritter, Andrew Karolyi and the participants of our presentations at the 2006 FMA annual meeting in Salt Lake City, the First Korea International Finance Conference in Seoul, the 2008 Asian Finance Association annual meeting in Yokohama, and the 2011 FMA Asian meeting in Queenstown. Sun and Wu gratefully acknowledge the financial support from the National Natural Science Foundation of China (Grant No. 70872095) and from the Center for Financial Studies at Fudan University. Tong would like to acknowledge the Central Research Grant (PolyU 5436/07H). References
Alexander, G.J., Eun, C.S., Janakiramanan, S., 1987. Asset pricing and dual listing on foreign capital markets: a note. Journal of Finance 42, 151–158. Berkman, H., Cole, R., Fu, J., 2009. Expropriation through loan guarantees to related parties: evidence from China. Journal of Banking and Finance 33, 141–156. Berkman, H., Cole, R., Fu, J., 2010. Political connections and minority-shareholder protection: evidence from securities-market regulation in China. Journal of Financial and Quantitative Analysis 45 (6), 1391–1417. Blass, A., Yafeh, Y., 2001. Vagabond shoes longing to stray: why foreign firms listed in the United States. Journal of Banking and Finance 25, 555–572. Bortolotti, B., Fantini, M., Scarpa, C., 2002. Why do governments privatize abroad? International Review of Finance 3 (2), 131–161. Chen, K., Yuan, H., 2004. Earnings management and capital resource allocation: evidence from China’s accounting-based regulation of rights issues. Accounting Review 74 (3), 645–665. Coffee, J., 1999. The future as history: the prospects for global convergence in corporate governance and its implications. Northwestern University Law Review 93, 641–708. Coffee, J., 2002. Racing towards the top? The impact of cross-listings and stock market competition on international corporate governance. Columbia Law Review 102, 1757–1831. Doidge, C., Karolyi, A., Stulze, R., 2004. Why are foreign firms listed in the US worth more? Journal of Financial Economics 71 (2), 205–238. Dyck, A., Zingales, L., 2004. Private benefits of control: an international comparison. Journal of Finance 59 (2), 537–600. Errunza, V., Losq, E., 1985. International asset pricing under mild segmentation: theory and test. Journal of Finance 40, 105–124.

1474

Q. Sun et al. / Journal of Banking & Finance 37 (2013) 1460–1474 Pagano, M., Panetta, F., Zingales, L., 1998. Why do companies go public? An empirical analysis. Journal of Finance 53 (1), 27–64. Pagano, M., Röell, A.A., Zechner, J., 2002. The geography of equity listing: why do companies list abroad? Journal of Finance 52 (6), 2651–2694. Reese, W.A., Weisbach, M.S., 2002. Protection of minority shareholder interests, cross-listings in the United States, and subsequent equity offerings. Journal of Financial Economics 66 (1), 65–104. Ritter, J., 1991. The long-run performance of initial public offerings. The Journal of Finance 46 (1), 3–27. Saudagaran, S., 1998. An empirical study of selected factor influencing the decision to list on foreign stock exchanges. Journal of International Business Studies 19 (1), 101–127. Stapleton, R., Subrahmanyam, M., 1977. Market imperfections, capital market equilibrium and corporation finance. Journal of Finance 32, 307–319. Stulz, R., 1999. Globalization, corporate finance, and the cost of capital. Journal of Applied Corporate Finance 12, 8–25. Subrahmanyam, A., Titman, S., 1999. The going public decision and the development of financial markets. Journal of Finance 54, 1045–1082. Sun, Q., Tong, W., 2000. The effect of market segmentation on stock prices: the China syndrome. Journal of Banking and Finance 24 (12), 1875–1902. Sun, Q., Tong, W., 2003. China share issue privatization: the extent of its success. Journal of Financial Economics 70 (2), 183–222. Sun, Q., Tong, W., Yan, Y., 2009. Market liberalization within a country. Journal of Empirical Finance 16 (1), 18–41. Tong, W., Yu, W., 2012. A corporate governance explanation of the A–B share discount in China. Journal of International Money and Finance 31 (2), 125–147. Wang, S., Jiang, L., 2004. Location of trade, ownership restriction and market lliquidity: examining Chinese A- and H-shares. Journal of Banking and Finance 28 (6), 1273–1297. Yehezkel, A., 2006. Foreign corporations listing in the United States – Does law matter? Testing the Israeli phenomenon. NYU Journal of Law & Business 2, 351.

Firth, M., Lin, C., Zou, H., 2010. Friend or foe? The role of state and mutual fund ownership in the split share structure reform in China. Journal of Financial and Quantitative Analysis 45, 685–706. Greene, W.H., 2002. Econometric Analysis, fifth ed. Prentice Hall International, Inc. Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47 (1), 153–161. Huang, A.G., Fung, H.G., 2004. Stock ownership segmentation, floatability, and constraints on investment banking in China. China and World Economy 12, 55– 78. Hung, M., Wong, T.J., Zhang, T., 2008. Political Relations and Overseas Stock Exchange Listing: Evidence from Chinese State-owned Enterprises, Working paper. Leuz, C., Nanda, D., Wysocki, P., 2003. Earnings management and investor protection: an international comparison. Journal of Financial Economics 69, 505–527. Li, K., Wang, T., Cheung, Y., Jiang, P., 2011. Privatisation with negotiation: evidence from the split share structure reform in China. Review of Financial Studies 24, 2499–2525. Liu, Q., Lu, Z., 2007. Corporate governance and earnings management in the Chinese listed companies: a tunneling perspective. Journal of Corporate Finance 13, 881–906. MacNeil, I., 2002. Adaptation and convergence in corporate governance: the case of Chinese listed companies. Journal of Corporate Law Studies 2, 289–344. Megginson, W.L., Netter, J.M., 2001. From state to market: a survey of empirical studies on privatization. Journal of Economic Literature 39, 321–389. Megginson, W., Nash, R., Randenborgh, M., Poulsen, A., 2004. The choice of private versus public capital markets: evidence from privatizations. Journal of Finance 59 (6), 2835–2870. Mei, J., Scheinkman, J., Xiong, W., 2009. Speculative trading and stock prices: evidence from Chinese A-B share premia. Annals of Economics and Finance 10 (2), 225–255. Nenova, T., 2003. The value of corporate voting rights and control: a cross-country analysis. Journal of Financial Economics 68, 325–351.