The E ects of Market Development on Controlling Shareholders Participation in Rights O erings 1

Transcription

1 The E ects of Market Development on Controlling Shareholders Participation in Rights O erings 1 Qi Chen Katherine Schipper Kun Wang Xing Xiao This draft: May We received helpful comments from Robert Bushman, Wei Jiang, Ross Levine, Ross Watts, and conference and workshop participants at 2011 China International Conference in Finance, Duke University, the Stockholm School of Economics, and Tsinghua Univerisity. Qi Chen (qc2@duke.edu) and Katherine Schipper (schipper@duke.edu) are at the Fuqua School of Business, Duke University. Kun Wang (wangk@sem.tsinghua.edu.cn) and Xing Xiao (xiaox@sem.tsinghua.edu.cn) are at the School of Economics and Management, Tsinghua University. This project was begun during Qi Chen s visit as a special-term professor at Tsinghua University. Kun Wang acknowledges funding from the National Natural Science Foundation of China (Project ).

2 The E ects of Market Development on Controlling Shareholders Participation in Rights O erings Abstract We examine the relation between the level of market development (the extent to which resource allocations are guided by market mechanisms as opposed to government decisions) across regions in China and controlling shareholders decisions to participate in Chinese public companies rights o erings. We nd signi cant positive relations between measures of market development and controlling shareholders participation, and evidence that the decision to participate bene ts minority shareholders. These results are consistent with the hypothesis that better market development creates a channel to deliver de facto protection for minority shareholders by creating implicit incentives for controlling shareholders to act in the interests of minority shareholders, holding constant minority shareholders de jure rights. Our results support arguments that macro-level institutions, speci cally, the level of market development, a ect rm-level governance through a channel that is distinct from the direct channel of explicitly granting de jure rights to minority shareholders. JEL classi cation: G30, G32, G38 Keywords: market development, corporate governance, rights o ering

3 1 Introduction We provide evidence on the link between cross-province di erences in China in the degree of market development, a key feature of institutional development, and minority shareholder protection, a key factor in supporting stock market development. Speci cally, we analyze how the level of market development in rms operating environments, de ned as the extent to which resource allocations are guided by market mechanisms as opposed to government decisions, a ects the decisions of controlling shareholders of Chinese listed rms to participate in rights o erings during We present evidence consistent with the view that the participation decision represents an improvement in corporate governance, establishing a speci c channel between institutional (market) development and, ultimately, stock market development. Our analysis builds on previous research that posits the protection of minority shareholders from expropriation by insiders (controlling shareholders and/or management) as an important condition supporting the development of a stock market. For example, La Porta et al. (1997, 1998) and Acemoglu and Johnson (2005) document a positive country-level association between institutional development and stock market development, consistent with the view that strong institutions o er protections to minority shareholders. However, the mechanism that connects institutional development with minority shareholder protection is not speci ed, so the country level correlations are "still something of a blackbox" (Acemoglu and Johnson (2005)). We posit, and provide evidence, that one such connection is from the degree of market development, a key feature of overall institutional development, to de facto minority shareholder protection, operating through the e ect of market-based resource allocation decisions on corporate governance. When resources are allocated by a properly functioning market mechanism, rms are subject to market discipline as they compete for resources. This discipline forces rms to organize and conduct their operations fairly and e ciently, so as to be better positioned to attract equity and debt capital, customers and suppliers. A strong governance mechanism instills con dence in outside stakeholders that insiders will act fairly, that is, they will not expropriate rm wealth for their own bene t. This stakeholder con dence is a necessary condition for repeat transactions and therefore for long-term survival (Klein and Le er (1981), Hermalin (2008)). The need to compete 1

4 for resources and trading partners creates incentives for insiders to take voluntary actions in the interests of minority shareholders, resulting in de facto protection, as opposed to strictly legal or de jure protection, for minority shareholders. This de facto protection both encourages good (that is, e cient) behaviors and discourages bad (that is, ine cient and/or expropriative) behaviors. In contrast, legal protection limits expropriative behaviors by insiders but, by design, does not provide incentives to increase e ciency. In contrast to the theoretical simplicity of a link between the degree of market development and rm-level governance arrangements, the task of empirically isolating this e ect is complex. In our setting, the empirical task involves three steps: identifying a governance practice with unambiguous welfare consequences for minority shareholders; quantifying the extent to which the market mechanism is at work across di erent economies at di erent times; and separating the e ect of market development from the e ect of investors de jure rights. Our research design is intended to address each of these steps in a way that increases the power of our tests for identifying and quantifying the link between market development and governance arrangements. Speci cally, we analyze whether and how variation in the level of market development across regions in a single country, China, a ects controlling shareholders decisions to participate in Chinese public companies rights o erings. 1 This setting has two distinct features that increase the power of our tests. The rst feature is a clear con ict of interest between minority and controlling shareholders, and the opportunity for controlling shareholders to reduce that con ict, at a potentially considerable cost. The cost arises because, unlike minority shareholders, controlling shareholders in Chinese public companies were not permitted to trade their shares freely during our sample period. Thus, they would bear potentially substantial economic costs from participation in a rights o ering, because they could not bene t from the discount typically embedded in the o ering price. At the same time, participation better aligns controlling shareholders interests with those of minority shareholders as it narrows the gap between their control rights and cash ow rights. The second distinctive feature of our setting is that we consider the provinces (regions) of a 1 Guiso, et al. (2004a, 2004b) and Musacchio (2008) also use a single country design to control for cross-country di erences. 2

5 single country, China, where there is considerable across-province variation in the level of market development but all public companies are subject to the same national laws that are directly relevant for shareholder protection; that is, we hold constant de jure rights. As the largest economy transitioning from a centrally planned economic system to a market-oriented one, China o ers a meaningful setting to study the e ect of market development on corporate governance outcomes. It experienced signi cant improvements in its overall level of market development in a relatively short period, starting with almost no institutional infrastructure to support a market mechanism. The transition has been gradual, controlled and uneven across provinces as the government releases direct control of resource allocation decisions and focuses instead on establishing conditions that support the proper functioning of the market mechanism. The rate of transition is related to policies pursued by the government at di erent times and in di erent provinces. These policies result in signi cant cross-region and overtime di erences in the level of market development. 2 We proxy for the level of market development in China using year- and province- speci c measures of the degree to which resources are allocated by the market forces as opposed to by the government and the degree to which the government focuses on law and order. 3 We nd a strong positive relation between controlling shareholders participation in rights o erings and measures of the degree of market development. In analyses of cross-sectional di erences in the e ect of market development at the rm level, we hypothesize and nd that the e ect of market development is stronger when the marginal bene ts of controlling shareholder participation (or the marginal costs of non-participation) are higher. We nd that the e ect of market development on controlling shareholders participation in rights o erings is stronger among state-owned enterprises than privately-owned rms; among state-owned rms whose controlling shareholders are more likely to pursue objectives other than maximizing total shareholder value; and among rms with high 2 For example, a policy of protecting certain domestic industries (e.g., defense and energy) has resulted in less competition in the inland provinces where these industries are concentrated while the pursuit of export-oriented policies in the coastal regions has resulted in competitive economic conditions in those regions. The policies in recent years have shifted to become more favorable to the inland and rural provinces in an attempt to narrow the income gap between them and the coastal areas. 3 Section 3.1 describes these measures in detail. 3

6 debt levels. These results are consistent with the idea that greater market development disciplines corporate insiders to take actions to improve governance. We also provide evidence on the welfare consequences, for minority shareholders, of controlling shareholders participation in rights o erings. We nd that more controlling shareholder participation is associated with higher participation by other shareholders, more positive stock price reactions to the rights o ering announcements and better operating performance up to two years following the o ering. We also nd that when controlling shareholders participate more, rms are less likely to conduct tunneling activities, and less likely to pay dividends to all shareholders immediately after the o erings, including the non-participating controlling shareholders, which clearly harms minority shareholders. These results support the view that controlling shareholder participation in rights o erings bene ts minority shareholders, and the de facto minority shareholder protection arising from market development both discourages bad behaviors and encourages good behaviors. Our study provides empirical evidence that market development a ects corporate governance. This evidence has policy implications as it suggests a spillover e ect from establishing a well functioning market mechanism for guiding resource allocation decisions to the development of a stock market. This spillover e ect might be used for evaluating proposals aimed at improving stock market development, especially in emerging markets where the stock market itself is at an early stage of development. Speci cally, our ndings indicate that market forces contribute to rm-level governance improvements and minority shareholder protection, a key condition for stock market development, suggesting that government e orts should support the proper functioning of the market mechanism for allocating resources, as opposed to, for example, making resource allocation decisions directly. Our paper can also shed light on the current debate on the merits of the Beijing model or state capitalism. 4 Our ndings indicate that the government s biggest role in increasing the success of state capitalism may lie in building structures that support the proper functioning of the market mechanism. 4 The term "state capitalism" has a long history and more than one meaning. Usually, state capitalism refers to pro t-seeking activity undertaken by the state, taking a capitalist approach to managing corporate resources. Bremmer (2010) provides a recent discussion of state capitalism that focuses on China. 4

7 We contribute to the broad literature that provides micro-level evidence of the governance e ects of macro-institutions. Micro-level evidence can help explain the country-level correlation between institutions and stock market development (La Porta et al. (1997, 1998), Acemoglu and Johnson (2005)), by documenting the channels through which speci ed macro-institutions a ect the development of stock markets. Much of this literature has followed La Porta et al. (1997, 1998) and focused on the e ects of legal institutions, reporting mixed evidence on the relation between country-level de jure shareholder protection and rm-level governance practices. 5 Most existing literature uses third-party ratings of governance practices that tend to be subjective and that implicitly assume a single set of practices is equally appropriate for all rms. However, this assumption may not be valid. For example, as Doidge et al. (2007) point out, in countries with higher risk of government expropriation, more information disclosure may not be desirable for shareholders if it increases expropriation risk. We extend this literature by providing a withincountry analysis of how province-level market development is related to minority shareholders de facto protection, and, more speci cally, by focusing on a decision by controlling shareholders that has both clear costs for those shareholders and clear welfare consequences for minority shareholders. 6 Our study is related to and more general than studies on the governance e ect of product market competition, one indicator of market development. The theoretical argument for the e ect of product market competition is ambiguous, depending on the type of competition (Aggrawal and Samwick (1999)); 7 related empirical studies have found weak or mixed evidence on how product 5 Empirical evidence suggests that rms compensate, or substitute, for a lack of de jure protection for minority shareholders by adopting voluntary actions such as corporate charters or bylaws restricting controlling shareholders behavior, or by listing on foreign exchanges with stringent governance requirements (see, e.g., La Porta, et al. (2000), Durnev and Kim (2005), Doidge, et al. (2007)). Doidge, et al. (2007) o er a theoretical model in which rm-level and country-level de jure protection can be complements or substitutes. They also provide evidence that rms nd it too costly to adopt rm-level government mechanism when de jure protection is very low. 6 A related stream of research examines the channel between legal arrangements and contracts, i.e., how legal institutions a ect the availability and nature of nancial contracts, mostly debt contracts and bank loans, signed at the beginning of the relationship (Lerner and Schoar (2005), Qian and Strahan (2007), Bae and Goyal (2009), and Miller and Reisel (2011)). We focus on minority shareholders (not venture capitalists or creditors) and on the e ect of the implicit incentives provided by the market mechanism in disciplining controlling shareholders. 7 See Giroud and Mueller (2010) for a summary of the theoretical literature. 5

8 market competition a ects governance arrangements. For example, Giroud and Mueller (2011) report a near zero correlation between the Her ndhal-hirschman index of a rm s industry and its governance, as measured by the Gompers, et al. (2003) index. Cremers, et al. (2008) nd that rms are more likely to adopt takeover defenses in competitive industries, but only those characterized by long-term customer-supplier relationships. Giroud and Mueller (2010, 2011) focus on how competition a ects the strength of the relation between governance and performance. Several recent studies examine the e ect of market development on Chinese rms behavior. Speci cally, Wang, et al. (2008) nd that state-owned rms in less-developed provinces are more likely to hire smaller, local auditors. Li, et al. (2009) nd a negative association between rms access to long term debt and the level of market development. Fan, et al. (2007) nd that stateowned rms in more developed regions are more likely to form extensive pyramid structures. Our study contributes to this literature by documenting a positive relation between the level of market development and controlling shareholders participation in rights o erings. Relative to previous research on rm behaviors, we assess the welfare consequences of the controlling shareholders behavior, not rm behavior, thus establishing a direct, positive link between the level of market development and welfare-improving rm-speci c governance arrangements. The rest of the paper is organized as follows. Section 2 discusses the institutional background of rights o erings in China and develops testable hypotheses. Section 3 describes the construction of the main test variables and our sample construction. Section 4 presents the main results on the e ects of market development on participation. Additional evidence on the consequences of participation is discussed in Section 5 and conclusions in Section 6. 2 Institutional background and hypothesis development 2.1 Institutional background of rights o erings in China The rights issues in our sample are available only to holders of A shares of Chinese public companies. These shares are issued as part of an IPO, are denominated in the Chinese currency (RMB) and 6

9 can be held only by domestic investors. 8 All shares carry the same cash ow rights and voting rights, but di er as to tradability. Tradable shares, which can be freely exchanged, are issued to investors through IPO subscription at the government-approved IPO price, usually about times the issuing rm s previous three-year earnings average. Non-tradable shares are issued to the government or the parent and a liate companies of the listed company, often for a nominal price of RMB1 per share. These shares may be exchanged at negotiated prices, often with government approval, but they cannot be freely traded during our sample period ( ). 9 All controlling shareholders in our sample rms hold non-tradable shares, which account for the majority of the shares. However, investors in tradable shares (minority investors) are the main providers of external equity nancing to listed companies. In a rights o ering, existing shareholders have the option to purchase additional shares, usually up to a xed percentage of their existing holdings, at an o er price that is typically discounted from the current market price. All shareholders pay the same price. Because additional shares purchased by holders of non-tradable shares are also not tradable, holders of non-tradable shares do not bene t from the discount typically embedded in the o er price. The Chinese Security Regulatory Committee (CSRC) requires companies to obtain approval from their shareholders and the CSRC and to meet certain performance and disclosure criteria to qualify for a rights o ering. To qualify, companies (1) must meet a return on equity (ROE) threshold in the three years prior to the proposed o ering year (Chen and Yuan (2004)); 10 (2) must wait at least one year after the previous o ering; and (3) can issue at most 30% of the existing 8 Unless otherwise noted, shares or stock refer to A shares. From 2002, quali ed foreign institutional investors (QFIIs) can also hold A shares, subject to government-approved quotas. Most QFIIs are large investment banks or brokerage rms (e.g., Goldman Sachs, Nomura Securities). Other foreign investors can hold the dollar-denominated B shares, issued by a subset of Chinese public companies. Detailed discussions of Chinese stock markets can be found in Neftci and Menager-Xu (2007). 9 Starting in 2005, the government allowed shareholders to negotiate the conversion of nontradable shares into tradable shares, usually with the nontradable shareholders giving concessions to the tradable shareholders such as a one-time cash dividend or less than 1-to-1 conversion rate. Most companies did not nish the conversion until We exclude from our sample ve rights o erings approved after 2005; results are nearly identical to those reported if these ve observations are included. 10 The ROE threshold varies each year and is lower for some industries such as the energy industry. 7

10 shares outstanding. The majority of our sample rms issued the maximum number of shares, because of general constraints on Chinese companies ability to access the stock market during our sample period. A typical rights o ering starts with an announcement by the board of directors, usually describing the potential range of the o er price, and whether the controlling shareholders of the company plan to participate, although the extent of participation may not be stated. The proposal is submitted to shareholders at the annual meeting, during which some large shareholders may reveal whether they intend to participate. Upon shareholder approval, the company submits an application to the CSRC; if the application is approved, the next step is the issuance of a nal prospectus and the announcement of the o er price. Through 2001, right o erings were an important source of equity capital for Chinese public companies. Table 1 shows the number of rights o erings in our sample by year of completion, amount (in RMB) raised in the o erings, and the proceeds from rights o erings as a percentage of all equity nancing raised by public companies. This percentage reached nearly 75% in 1998, with proceeds totaling 33.5 billion RMB (about 4.2 billion USD). The number of rights o erings declined starting in 2002 when seasoned equity o erings became the primary mode of raising equity capital, consistent with patterns observed in other countries (Eckbo (2008)). 2.2 Hypothesis development One feature of weak governance arrangements is a misalignment of interests between corporate insiders (controlling shareholders) and outside investors (minority shareholders). The misalignment of interests provides controlling shareholders with incentives to pursue private bene ts at the expense of minority shareholders as insiders bear only a portion of the cost. Weak governance makes outsiders reluctant to invest, which in turn hurts rms long run prospects for growth. This e ect will be more pronounced in more developed markets, where resources are allocated by the market mechanism instead of government decisions. The market-based allocation process subjects all market participants to the discipline of market forces, so that poorly governed and ine cient rms will falter or fail while well governed and e cient rms will prosper. As a result, market 8

11 development provides incentives to adopt e cient organizational and operational structures. We predict that better market development increases the opportunity cost of weak governance and provides an incentive to controlling shareholders to improve corporate governance when they can. Because controlling shareholder participation in rights o erings improves governance, relative to no participation, we predict a positive association between the degree of market development and participation. This leads to our rst hypothesis, stated in alternative form: H1: Controlling shareholders participation in rights o erings is positively related to the level of market development. We expect the e ect of market development on controlling shareholders decision to participate in the rights o erings to vary cross-sectionally with the expected bene ts and costs of participation. With regard to costs, participation in a rights o ering is costly to all controlling shareholders, who hold non-tradable shares and therefore cannot bene t from the discounted rights o ering price. However, bene ts of participation will vary, increasing as pre-rights-o ering governance arrangements are less suitable for survival in the market place. Applied to our setting, we hypothesize that the e ect of market development is stronger for state-owned enterprises (SOEs) than for private rms because controlling shareholders of SOEs are more motivated by nonprice (e.g., political) considerations than controlling shareholders of private rms, and therefore SOEs are less well positioned to compete for resources allocated by the market mechanism. Similarly, within SOEs, we expect that the e ect of market development on controlling shareholders participation is stronger when the controlling shareholders are more closely related to government agencies. This reasoning lead to our second hypothesis, stated in alternative form: H2: The e ect of market development on controlling shareholders participation in rights offerings is stronger for state-owned rms (SOEs) than for private rms; further, among the SOEs, the e ect is stronger for controlling shareholders whose objectives are more aligned with those of government agencies. Lastly, we expect the e ect of market development on controlling shareholder participation in rights o erings to be stronger for highly levered rms; other things equal, these rms face a greater need for external nancing because they have used up more of their debt capacity. On the one 9

12 hand, in a well functioning market, creditors and equity investors seek to maximize the returns on their investments and are reluctant to invest in rms with weak governance. Yet in such a market it is precisely investors willingness to invest, as opposed to government mandates or regulations, that determines rms access to external capital. This implies that controlling shareholders of rms that rely more on external nancing have stronger incentives to participate in rights o erings, so as to demonstrate that their interests are aligned with those of minority shareholders. If the level of market development is low, investment decisions, especially those of creditors, which are primarily state-owned banks in China, are not necessarily motivated by pro t considerations. In this case, corporate governance plays a relatively minor role in access to capital, reducing the need for controlling shareholders to participate in rights o erings. This leads to our third hypothesis, stated in alternative form: H3: The e ect of market development on controlling shareholders participation in rights o erings is stronger for rms with high leverage. 3 Market development measures and sample description 3.1 Measures of market development in China Conceptually, a market mechanism is more developed when more resources are allocated by decisions of market participants following well-de ned, understood and enforced rules, guided by market forces. Since by de nition a well-developed market mechanism is one that allocates resources e ciently, that mechanism may take di erent forms depending on the types of resources to be allocated as well as the conditions under which they are allocated. This problem of observation and measurement is likely exacerbated in cross-country settings. While there is no single way to support a market mechanism for allocating resources, there is one way to hinder its development, namely, arbitrary resource allocations by the government with no concern for markets, as was the case in China prior to the reforms that began in By de nition, more resources under the government s control mean fewer resources to be allocated by a market mechanism. As previously discussed, China s transition to a market-oriented economy has 10

13 been both controlled by changes in government policy and uneven across provinces. We measure the development of a market mechanism for allocating resources at the province-year level, along two dimensions: (1) the degree to which the government is not directly involved in the economy, which we denote by GOV; and (2) the extent to which the government provides rules and arrangements that support a competitive market for resources, which we denote by LEGAL. The combination of these two dimensions is our composite measure of market development, M KT: We obtain measures of market development from the National Economic Research Institute (NERI) Index of Marketization of China s provinces constructed by Fan and Wang (2001, 2004, 2007). The NERI indices include 19 components in ve subcategories, namely, the relation between government and the market, the development of non-state economies, factor markets development, product markets development, and development of market intermediaries and legal environment. Information to construct each component is acquired either from National and provincial Bureaus of Statistics or by surveying randomly selected rms in each province. Based on the value of each component in province i in year t (V it ), Fan and Wang assign a relative score to the province-year. If a higher value of V indicates better market development, the score is calculated as 10 (V it V min;t )=jv max;t V min;t j where V max;t and V min;t are the maximum and minimum value of V among all provinces in a base year T. If a higher value of V indicates less market development, the numerator becomes 10 (V max;t V it ). In years other than the base year, the indices could be negative or greater than 10. The score captures both cross-sectional variation at any given year and over-time variation. Data are available from 1997 to We use the 1997 value for years prior to Our results are not sensitive to removing rights o erings before Although all components of the NERI index are meant to capture various aspects of market development, they may di er in reliability; for example, the survey-based components are more subjective than those from the Bureaus of Statistics. Some components also lack continuity over time either due to missing data or due to changes in how the data are collected and compiled at the Bureaus of Statistics. To address these concerns, we use the data in two ways. First, we use all 19 components, the associated ve subindexes and the aggregate composite index as presented. Second, we choose a subset of components based on reliability and data continuity. Results are 11

14 qualitatively similar under both approaches; in the interest of simplicity, we present and discuss only the results based on the second approach. Speci cally, we proxy for the degree of government involvement in the economy with three NERI indices, all measured at the province-year level: (1) the percentage of the labor force employed by the private sector; (2) the percentage of bank loans to non-state-owned companies; (3) an index based on a survey of business owners indicating how many hours they spend dealing with government related issues. Our measure of government involvement in the economy (GOV ) is the average of these three measures; larger values of GOV indicate greater market development (less government involvement). To capture the extent to which the government provides rules and arrangements that support a market mechanism, we calculate an index of legal enforcement (LEGAL) as the average of two measures: (1) the number of lawyers and accountants as a percentage of the provincial population; (2) the degree of intellectual property protection, measured by the number of patent applications and approvals scaled by local GDP. Larger values of LEGAL indicate greater market development. Our composite index of market development (M KT ) is the average of the ve components for each province-year. Both GOV and LEGAL are input-based. An outcome indicator of well-functioning markets is higher long term economic growth, which should appear as larger GDP per capita. On a country level, GDP per capita is often endogenous to other factors that are di cult to quantify and control (e.g., geographic endowments, social, cultural and religious di erences, accidents of history). This type of endogeneity is less of a concern in our single-country design than in a cross-country design, especially since Chinese GDP per capita prior to the inception of the transition into a marketoriented economy did not vary much across regions (Chen and Fleisher (1996)). For the main test of Hypothesis 1, we also use GDP per capita at the province-year level as a proxy for market development and nd results (not tabulated) that are qualitatively similar to those reported. We acknowledge that because it is an outcome indicator, GDP per capita cannot provide much insight into the factors that foster market development. Table 2 summarizes the GOV and LEGAL indices and their components by region (Panel A) and by year (Panel B) using the province-year as the unit of observation (the data are not 12

15 weighted by the number of sample rms in each province). Shanghai has the highest level of market development, with M KT = 12.71, followed by Guangdong at The least developed region is Tibet, with MKT = The mean value of MKT across regions is There is also variation across regions in measures of government involvement and legal protection. For example, GOV ranges from 3.94 in Xinjiang to in Guangdong, and LEGAL ranges from 0.04 in Tibet to in Shanghai. Similar variation can be observed in the components of GOV and LEGAL, and the components are not necessarily homogenous within a region. For example, Beijing ranks the highest on the measure of legal enforcement at 21 (as proxied by lawyers and accountants as a percent of the population), but ranks only the tenth lowest of 31 provinces on the measure capturing government-controlled credit (4.18). With regard to over-time changes, Panel B of Table 2 shows that the proxies for market development throughout China have improved over time; the country average M KT increases monotonically from 4.91 in 1997 to in The overtime improvement is present in all components, although it is not always monotonic. Our research design assumes that the province-level (local) economic environment matters for our sample rms. We believe this is a reasonable assumption because a majority of our sample rms were originally owned by provincial governments, and their operations continue to be in uenced by these governments, particularly after the 1980s scal reforms when the central government delegated signi cant decision rights to the provincial governments. This decentralized structure, combined with competition among provincial governments to achieve high growth, led to local protectionism (Qian (2003), Jin, et al. (2005)), increasing the in uence of the local environment. The local in uence also extends to the e ciency of legal enforcement; for example, The World Bank Survey, Doing Business in China 2008, shows that the number of days required to settle a commercial dispute can range from 120 days in Guangdong province to 432 days in Qinghai province. To the extent that local environments do not matter for our sample rms, our province-speci c measures of market development will contain measurement error that should bias against nding any results. 13

16 3.2 Sample description Our sample contains 812 rights o erings by non- nancial rms approved by the CSRC during 1993 to 2005, retrieved from the CSMAR (China Stock Market and Accounting Research Database; also available from WRDS). Our sample includes 18 rms that did not complete approved rights o erings, because of unfavorable market conditions or the share reform that began in We include these 18 rms because we are interested in controlling shareholders commitment to participate; results are qualitatively similar if we exclude them. The 812 approvals apply to 587 di erent rms, 399 of which had one issuance/approval, 155 of which had two, 29 had three and 4 had four issuance/approvals. Column 1 of Table 2 shows the distribution of our sample rms by province. All provinces are represented, with concentrations in Shanghai (115 observations) and Guangdong (106 observations), where the two stock exchanges are located. The CSRC allows holders of non-tradable shares to use non-cash assets to pay for shares in the rights o erings, whereas holders of tradable shares must pay cash. We focus on participation rates in o erings in which controlling shareholders pay cash, CSH_P AR, calculated as I(Cash) P ART _RAT E, where I(cash) equals 1 if the controlling shareholders paid cash in the rights o ering and P ART _RAT E is the controlling shareholders participation rate in rights offerings. 11 We manually collect participation data for holders of tradable and non-tradable shares if the information is available from the issuing company. Otherwise, we impute the participation rate as the change in each shareholder group s holdings from before to after the rights o ering, divided by the number of rights the group was allocated. For example, in 2001 Shanghai Tunnel Engineering Co. Ltd. (stock code ) issued rights to its shareholders at a ratio of 10 to 2.3, that is, 23%. The controlling shareholder Shanghai City Construction Group Ltd. could subscribe million shares, or 23% of its existing ownership of million shares (54.5% of outstanding shares). Its ownership increased by 2.8 million shares after the rights o ering, implying a participation rate of 4.52% (=2.8/61.93), lower than the overall average participation rate of 43%, reducing its percentage ownership to 50.1%. 11 When we repeat our tests using the unadjusted participation rate, results (not tabulated) are qualitatively similar to those reported. 14

17 Table 3 provides summary statistics for the test and control variables; all variable de nitions are in the Appendix. The cash participation rate of controlling shareholders averages 20.17%, ranging from zero to 100%; just over half the controlling shareholders did not participate at all. There is also variation in market development, M KT; among sample rms, with mean (standard deviation) of 5.48 (2.04); variation is similar for the components of the market development measure. Not surprisingly, given China s history as a centrally planned economy, 82% of the sample rms are stateowned (in Table 3, P rivate, an indicator variable for whether the rm is privately-owned, averages 18%). Ownership is in general concentrated; the mean (median) of the controlling shareholder s ownership is 46.34% (45.74%). The mean (median) leverage ratio is 41.52% (41.27%), slightly lower than the average leverage of 44% in Chinese listed rms (calculated using all Chinese listed rms in our sample period). Because of the CSRC s requirements to qualify for rights o erings, our sample rms on average perform well, with mean values of GrossMargin (the ratio of total revenue minus total cost of goods sold to total revenue) and ROA (the ratio of EBIT to assets) of 31.75% and 8.11%, respectively. 83% of the sample o erings were underwritten, with guaranteed subscriptions of tradable shares by investment banks, as captured by Stand_by. The mean and median for Offer_P rice (the ratio of the o ering price to the share price one week before the rights o ering announcement) are both 0.62, suggesting the rights o ering price is discounted about 38% on average. This discount is a measure of the cost of controlling shareholders participation, because their shares were not freely tradable in stock markets during our sample period. 4 Analysis and results 4.1 Empirical speci cation If there is an optimal ownership structure, it is unlikely to be observed in Chinese rms whose initial ownership structures were based largely on non-market factors. This is especially the case for stateowned rms, 82% of our sample observations. Rights o erings provide controlling shareholders a rare opportunity to adjust their share ownership, constrained by the 30% cap imposed by the CSRC on the shares that may be o ered. Because of this constraint, the observed participation rate 15

18 (between 0 and 100%) may not re ect the controlling shareholders actual participation preferences, which may be negative (if they want to reduce their ownership percentages) or more than 100% (if they want to increase their ownership percentages). This implies that the observed participation variable (CSH_P AR i;j;t ) is a censored version of the true rate (CSH_P ARi;j;t ) in that (Latent) CSH_P ARi;j;t = MKT j;t 1 + Z i;j;t 1 + D t + " i;j;t (Observed) CSH_P AR i;j;t = CSH_P ARi;j;t if 0 < CSH_P ARi;j;t < 100% = 0 if CSH_P ARi;j;t 0 = 100% if CSH_P ARi;j;t 100%; where CSH_P AR i;j;t is the controlling shareholders participation rate for rm i in region j for a rights o ering in year t. MKT j;t 1 measures the degree of market development in region j in year t 1. Z i;j;t 1 is a vector of rm-level control variables. D t is a year-indicator and " i;j;t is the error term. The majority of our analyses of shareholders participation decisions apply Tobit estimation with truncation points at both lower and upper bounds. For comparison and as a robustness check, we also present results from a Logit estimation (where the dependent variable is an indicator variable for whether the controlling shareholders participate) and an OLS estimation in Table 4 when we establish our main results. In all regressions, year indicators are included to control for year- xed e ects, and standard error estimates are adjusted for heteroskedasticity and correlations among rms from the same region (Peterson (2009)). All unbounded continuous variables are winsorized at the extreme 1% of values. Our main variable of interest is M KT. We include variables for rm-speci c characteristics and variables speci c to each rights o ering to control for other factors that may a ect controlling shareholders participation decisions. The rm-speci c characteristics include GrossM argin, size (Size, the natural log of total assets), and leverage ratio (Leverage, the ratio of total liabilities to total assets). To the extent that a higher gross margin suggests higher future pro tability and higher returns to equity investors, we expect a positive coe cient for GrossM argin. Our results are qualitatively similar when we use return on assets instead (results not tabulated). We do not 16

19 have predictions for the signs of Size and Leverage. To control for di erences in controlling shareholders characteristics we include P rivate and Large_SH% (the percentage of shares owned by the controlling shareholder). To the extent that governance concerns are more severe for state-owned enterprises, perhaps because their controlling shareholders are government agencies with objectives that include substantial nonprice considerations, for example, a wish to meet certain political and social welfare objectives, we expect more participation by controlling shareholders of SOEs, and a negative coe cient estimate for P rivate. Similarly, to the extent governance concerns are less severe in rms where controlling shareholders own more shares and therefore would garner smaller bene ts from participation, we expect a negative coe cient estimate for Large_SH%: These variables are calculated in the year prior to the o ering. We also include two variables speci c to each rights o ering: (Of f er_p rice) and (Stand_by)). We expect a negative coe cient for Offer_P rice as higher o er prices mean higher costs of participation. Prior research shows that rms use costly stand-by arrangements when they expect low participation rates (Eckbo (2008)). To the extent controlling shareholders view their participation in a rights o ering as a costly signal to attract other shareholders participation, we expect a negative relation between participation and Stand_by The e ect of market development on controlling shareholders participation in rights o erings Panel A of Table 4 presents results of estimating several speci cations of the relation between controlling shareholders participation in rights o erings and the level of market development. Column 1 of Table 4, Panel A presents results of a Tobit regression of controlling shareholders participation rates (CSH_P AR) on the composite index of market development, M KT; and control variables. Consistent with Hypothesis 1, the e ect of market development on controlling shareholder partic- 12 If the o er price, stand-by arrangements and participation decisions are jointly determined, including the o er price and the stand-by dummy in the regression may be problematic. We repeat our analyses without these variables, and nd similar results for the market development measures (results not tabulated). The coe cients for other independent variables remain qualitatively similar, with the signi cance level varying with the regression speci cation. 17

20 ipation is positive, reliably di erent from zero at less than the 5% level, and economically meaningful. The average partial e ect in the observed data is 3:40% and the e ect of an inter-quartile change in the market development index is 4:58%. 13 To put these numbers in perspective, consider two regions, Ningxia, an inland, rural province at the 25th percentile of market development, and Liaoning, a coastal province at the 75th percentile level of market development. The estimates suggest that if a rm with average rm-level characteristics moved from Ningxia to Liaoning, its controlling shareholders participation rate would increase by 4:58%, a more than 40% increase over the unconditional sample average participation rate of 21:4%. Column 2 of Table 4 Panel A shows the result of a Logit regression where the dependent variable is an indicator that takes the value 1 when the controlling shareholders participate by cash. The coe cient estimate on M KT is 0.107, signi cant at less than the 5% level, indicating that controlling shareholders are more likely to participate in rights o erings when the level of market development is higher. The average partial e ect is 2.528%, and the e ect of an interquartile change in MKT is 7:54%, representing a 15% increase relative to the average unconditional probability of participation (49:9%). Finally, column 3 presents results of an OLS regression, for comparison purposes. The results show that the e ect of market development on the participation decision is positive and weakly signi cant at approximately the 13% level. Panel A also shows results for rm-speci c characteristics and control variables. The results from all three estimations are qualitatively similar; we discuss mainly results in column 1, the Tobit regression. These results show that controlling shareholders participate more in more profitable rms: the coe cient estimate for GrossM argin is positive and signi cant at less than the P AR MKT +Z The average partial e ect is the empirical analogue of E = where (; ) are the coe cient estimates for the covariates (MKT; Z). The alternative (not tabulated), the marginal e ect (at sample mean), shows qualitatively similar results (and is available upon request). Bartus (2005) shows that the average partial e ect is typically more well behaved than the marginal e ect at means. The e ect of the inter-quartile change in MKT is calculated as the di erences in the predicted values of CSH_P AR evaluated at the 25th and 75th percentile of MKT, holding all other covariates at their sample mean values (Wooldridge (2002)). The e ects of inter-quartile changes are discussed in the text, not tabulated. Throughout the paper, the marginal e ect of a dummy variable is calculated similarly as the di erences in the predicted values evaluated when the dummy variable equals 1 and 0, holding all other covariates at their sample mean values. 18

21 5% level. The participation level is low when the controlling shareholders existing ownership is high: the coe cient estimate for Large_SH% is negative and signi cant at less than the 5% level. The coe cient estimate for P rivate is negative and signi cant at less than the 10% level, suggesting that controlling shareholders of private rms participate less than controlling shareholders of SOEs. Firm size and leverage have no signi cant e ects on controlling shareholders participation as the coe cient estimates for Size and Leverage are statistically indistinguishable from zero at conventional levels. As expected, controlling shareholders participate less when the o er price is high. Finally, Stand_by has a negative coe cient estimate, signi cant at less than the 1% level. The partial e ect estimate indicates that the average controlling shareholders participation rate is reduced by 17.2% for an underwritten o ering. Panel B of Table 4 shows the Tobit estimation results for the two components of the overall market development index, GOV and LEGAL, and their subcomponents. Results from the Logit and OLS estimations are qualitatively similar and are not tabulated. Column 1 shows that the measure of government (non)involvement of the economy (GOV ) has a positive coe cient estimate of 3.511, signi cant at less than the 5% level. Column 2 shows that the legal environment (LEGAL) does not have a statistically reliable e ect on participation rates. Columns 3-5 show that among the subcomponents of GOV, the index capturing the time spent dealing with government issues and the index capturing the percent of loans to private rms have signi cantly positive e ects on controlling shareholders participation, whereas the index capturing the percent of private employment has a positive relation that is not statistically signi cant at conventional levels. Columns 6-7 show that among the two components for the LEGAL index, the index capturing the number of lawyers and accountants has a positive e ect on controlling shareholders participation, weakly signi cant at approximately the 13% level (t-statistic = 1.57). The coe cient for the index capturing intellectual property protection is statistically indistinguishable from zero. To provide intuition for these estimates, we consider the e ects of inter-quartile changes in each of the subcomponents. Speci cally, an inter-quartile improvement in GOV and LEGAL would increase controlling shareholders participation rate by 4:24% and 1:22%, respectively. Among the subcomponents, an inter-quartile improvement would increase controlling shareholders participa- 19