Journal of Accounting and Economics 33 (2002) 401425

Corporate ownership structure and the informativeness of accounting earnings in East Asia$
Joseph P.H. Fana, T.J. Wongb,*
Department of Finance, School of Business and Management, The Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong b Department of Accounting, School of Business and Management, The Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong Received 23 February 2000; received in revised form 12 January 2002
a

Abstract This study examines the relations between earnings informativeness, measured by the earningsreturn relation, and the ownership structure of 977 companies in seven East Asian economies. Our results are consistent with two complementary explanations. First, concentrated ownership and the associated pyramidal and cross-holding structures create agency conicts between controlling owners and outside investors. Consequently, controlling owners are perceived to report accounting information for self-interested purposes, causing the reported earnings to lose credibility to outside investors. Second, concentrated ownership is associated with low earnings informativeness as ownership concentration prevents leakage

$ We appreciate helpful comments from Ray Ball, Gary Biddle, Charles Chen (the referee), Ellen Engel, Masaharu Hanazaki, Hasung Jang, Steve Matsunaga, Jevons Lee, Jody Magliolo, Randall Morck, Suil Pae, Enrico Perotti, Terry Shevlin, Megumi Suto, Sheridan Titman, John Wei, Joanna Wu, Takeshi Yamada, Ross Watts (the editor), Jerold Zimmerman, an anonymous referee, and workshop participants at the Asian Development Bank Institute, the Chinese University of Hong Kong, Hitotsubashi University, The Hong Kong University of Science and Technology, the Polytechnic University of Hong Kong, National Chengchi University of Taiwan, Nanyang Technological University at Singapore, and Shanghai University of Finance and Economics, and conference participants at the 2000 Conference on Accounting in Transition Economies at the William Davidson Institute, the 2000 HKUST Accounting Symposium, the 2000 AAANZ Conference, the 2000 Shanghai APFA Conference, the 2000 London EFA Conference, the 2001 AFA annual meeting at New Orleans, the 2001 AAA annual meeting at Atlanta, and the 2001 Conference on Designing Financial Systems in East Asia and Japan organized by IMF and Hitotsubashi University. T.J. Wong acknowledges the nancial support of the Wei Lun Fellowship. *Corresponding author. Tel.: +852-2358-7574; fax: +852-2358-1693. E-mail address: [email protected] (T.J. Wong).

0165-4101/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 5 - 4 1 0 1 ( 0 2 ) 0 0 0 4 7 - 2

402

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

of proprietary information about the rms rent-seeking activities. r 2002 Elsevier Science B.V. All rights reserved.
JEL classication: G32; M41 Keywords: Ownership concentration; Transparency; Earnings informativeness; Emerging market

1. Introduction Public corporations in East Asia typically have low levels of transparency and disclosure quality. Some commentators and policy advisors believe that a closer adherence to international disclosure rules and the adoption of international accounting standards are essential for improving corporate transparency in the region (World Bank, 1998). Despite efforts to impose stricter reporting rules and standards, the general perception is that corporate transparency has been declining (Asian Wall Street Journal, 1999). While the new accounting rules may have increased the quantity of accounting information, investors have reservations about the quality of the reported numbers.1 Therefore, it is important for regulators and policy makers to understand the causes of the low quality of reported accounting information in the region. This paper focuses on the relations between corporate ownership structure and the quality of accounting information in seven East Asian economies excluding Japan. More specically, we use the informativeness of accounting earnings to investors as a measure of the quality of accounting information. We develop two complementary arguments pertaining to the relations between ownership structure and earnings informativeness. The rst argument is related to the entrenchment effect of ownership concentration (Morck et al., 1988). Corporate ownership is highly concentrated in East Asia. As the controlling owners are entrenched by their effective control of the rms, their decisions that deprive the rights of minority shareholders are often uncontestable in the weak legal systems in the region and by ineffective corporate governance mechanisms such as boards of directors and the market for corporate control (Shleifer and Vishny, 1997; La Porta et al., 1999; Johnson et al., 2000b). Moreover, due to the complicated pyramidal and crossholding ownership structures typical in East Asian companies, a signicant number of controlling owners in the region actually possess more control than their equity ownership indicates, which further exacerbates the entrenchment effect.2 The
This view of low information quality was shared among business professionals at the recent World Bank Meeting. For example, a local lawyer from Thailand remarked that the major difference (in accounting disclosure) between the past and today is that statements of accounts now carry more qualications, not better information. See the report by Sender (1999). 2 Claessens et al. (2002) report that the concentrated control and the divergence between ownership and control in public corporations in eight East Asian economies diminish rm value, indicating the economic signicance of the agency problem associated with ownership structures. Consistent evidence is also found in several other studies. La Porta et al. (2002) examine over 300 rms from 27 wealthy economies and
1

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

403

entrenchment effect of the ownership structure potentially affects rms nancial reporting. Because the controlling owner oversees the accounting reporting policies and is perceived to have strong opportunistic incentives to hold up minority shareholders, the market expects that the owner will not report high-quality accounting information. This market perception will reduce the credibility of accounting earnings reports and consequently the informativeness of those earnings. The second argument is related to proprietary information and specic human capital. By concentrating ownership, decision rights can be given to individuals who possess specic knowledge (Jensen and Meckling, 1992; Christie et al., 2002). One benet of co-locating decision rights with specic knowledge is that the leakage of the specic knowledge to competitors is prevented and the transferring cost of the specic knowledge is avoided. This benet is great in East Asia where political lobbying activities are common and lucrative. As concentrating ownership limits information ows to the public, political rent-seekers are able to avoid potential competition and social sanctions. Thus, this information effect argument predicts that concentrated ownership is associated with opacity and low informativeness of accounting earnings. Our empirical evidence is broadly consistent with the predictions of the entrenchment and information effects arguments. We nd that earnings informativeness, measured by the earningsreturn relation, is signicantly negatively related to the ultimate owners control level, conditional on the owner having gained effective control. This evidence is consistent with the information effect. We also nd that earnings informativeness is signicantly negatively related to the degree of divergence between the ultimate owners control and the equity ownership level. This lends support to the entrenchment effect argument. The result is also consistent with the information effect argument, provided that controlling owners who want to protect proprietary information use stock pyramids or cross-shareholdings to leverage their control, thus creating divergence between ownership and control. These empirical results are robust to controls for rm size, market-to-book assets, leverage, the number of industry segments operated by the rm and to varying the starting and ending dates of the stock return window. This analysis of East Asian corporations allows us to study the subject of earnings informativeness in a different ownership context from that of the research on U.S. corporations. Our research results are also rich in policy implications. In general, our results support Ball et al. (2000) by nding that policy makers should consider a countrys overall institutional environment before prescribing a comprehensive set of rules and regulations for corporate reporting. Also, it is important for policy makers and regulators to understand how the concentrated share ownership structure in East Asia is associated with incentives for rms to reduce accounting information quality. Blindly adopting international accounting standards and disclosure rules
( footnote continued) report that rms with higher ownership by controlling owners have higher valuation. Johnson et al. (2000a) document that levels of shareholder protection explain the extent of stock market decline in many emerging markets during the Asian nancial crisis.

404

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

without considering the institutional environment in East Asia will not improve the corporate transparency in this region. Last, the paper illustrates that it would be fruitful for future research to focus on how ownership structures shape accounting policies in emerging markets and transition economies. The paper proceeds as follows. In Section 2, we discuss the causes and effects of ownership structures in East Asia and develop our hypothesis on the relation between ownership structure and earnings informativeness. In Section 3, we describe our sample and data, report statistics on the ownership structures of East Asian rms, and present our empirical analyses. We conclude this paper in Section 4.

2. Development of hypothesis The ownership of listed companies in East Asia is typically concentrated in the hands of large shareholders. This concentrated control is achieved through complicated ownership arrangements, i.e., stock pyramids and cross-shareholdings.3 In this section, we discuss the forces that shape the ownership structure. We then discuss how the ownership structure shapes the rms agency problems, through its entrenchment and incentive alignment effects on controlling owners. We nally discuss the entrenchment and the information arguments, which lead to a hypothesis pertaining to the relation between ownership structure and earnings informativeness. 2.1. Causes of concentrated ownership The body of property rights literature provides a general framework for analyzing the determinants of corporate share ownership structures.4 The literature emphasizes the roles of customs, social norms, and law and legal systems in shaping the structure of property rights and their governance systems. Corporate share ownership can be viewed as a property rights arrangement through which the owner of the share is entitled to three categories of property rights. First, the owner has the decision right of deploying corporate assets, i.e., the control or voting right. Second, the owner has the right to earn income, i.e., the cash ow right. Third, the owner has the right to transfer the share and the associated control and cash ow rights to another party. The value of the share depends on how well its property rights are enforced. The enforcement of property rights is usually undertaken by both individual owners and the state. In economies where the state does not effectively enforce property rights, the enforcement by individual owners plays a relatively more important role. The structure of share ownership affects the degree to which corporate contracts are
Dual-class shares are rare in East Asia. Among the seven economies that we investigate, only South Korea allows dual-class listings. They are completely prohibited in Hong Kong and Singapore. The remaining economies prohibit dual-class shares, but allow certain preferred shares to have dual-class characteristics. See Nenova (1999). 4 The literature was pioneered by Coase (1960), Demsetz (1964), Alchian (1965, 1977), and Cheung (1970, 1983). Interested readers are referred to Eggertsson (1990) for a survey of the literature.
3

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

405

enforced, because it affects the owners abilities and incentives to enforce the property rights delineated by the contracts. One prediction from the property rights framework is that concentrated ownership will be observed in economies where property rights are not well enforced by the state. Controlling owners obtain the power (through high voting rights) and the incentives (through high cash ow rights) to negotiate and enforce corporate contracts with various stakeholders, including minority shareholders, managers, laborers, material suppliers, customers, debtholders, and governments. The various parties in the nexus of corporate contracts share the benets of trade as a result.5 Shleifer and Vishny (1997) elaborate on this point and suggest that the benets from concentrated ownership are relatively larger in countries that are generally less developed, where property rights are not well dened and/or protected by judicial systems. To test this proposition, La Porta et al. (1999) investigate the ownership concentration by the three largest shareholders of the largest corporations in countries around the world and nd that weak legal and institutional environments (laws and enforcement) are associated with the highly concentrated share ownership of listed companies. The private enforcement of property rights is a probable reason for the concentrated ownership of East Asian corporations, which often confront weak legal systems, poor law enforcement, and corruption. 2.2. Incentive effects of ownership concentration The degree of ownership concentration affects the nature of contracting, creating agency problems between managers and outside shareholders. When ownership is diffuse, as is typical in the U.S. and the U.K., agency problems stem from the conicts of interest between outside shareholders and managers who own an insignicant amount of equity in the rm (Berle and Means, 1932; Jensen and Meckling, 1976; Roe, 1994). On the other hand, when ownership is concentrated to a level at which an owner obtains effective control of the rm, as is the case in East Asia and most other locations outside the U.S. and the U.K., the nature of the agency problem shifts away from managershareholder conicts to conicts between the controlling owner (who is also the manager) and minority shareholders. 2.2.1. The entrenchment effect Gaining effective control of a corporation enables the controlling owner to determine how prots are shared among shareholders. Although the minority shareholders are entitled to the cash ow rights corresponding to their share investments, they face the uncertainty that the entrenched controlling owner may
Large owners can be benecial in diffusely held rms, too. The existence of large owners mitigates the free-rider problem associated with the diffuse ownership structure in monitoring managers. Demsetz and Lehn (1985) provide evidence that ownership concentration in the U.S. is positively related to the control potential of rms, among other factors. They argue that distortions in the market for corporate control along with the managerial labor market increase the control potential of shareholders, which leads to increases in ownership concentration. Shleifer and Vishny (1986) argue that large shareholders monitor managers, which in turn increases rm value. This argument is supported by U.S. evidence (Holderness and Sheehan, 1988; Barclay and Holderness, 1989).
5

406

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

opportunistically deprive them of their rights. The effects of entrenchment by the controlling shareholder include outright expropriation, i.e., the controlling shareholder benets from self-dealing transactions in which prots are transferred to other companies he/she controls.6 The controlling shareholder can also exercise de facto expropriation through the pursuit of objectives that are not prot-maximizing in return for personal utilities. The entrenchment problem created by the controlling owner is similar to the managerial entrenchment problem discussed by Morck et al. (1988). Increasing managerial ownership may entrench managers, as they are increasingly less subject to governance by boards of directors and to discipline by the market for corporate control. 2.2.2. The alignment effect One way to mitigate the problem of controlling owner entrenchment is to increase further the controlling shareholders ownership stake, or even to go private if the problem is sufciently severe. A higher ownership stake gives a controlling owner stronger voting and cash ow rights in the rm. Once the controlling owner obtains effective control of the rm, any increase in voting rights does not further entrench the controlling owner, but his/her higher cash ow rights in the rm mean that it will cost more to divert the rms cash ows for private gain. The high ownership concentration can also serve as a credible commitment that the controlling owner is willing to build a reputation for not expropriating minority shareholders (Gomes, 2000). The commitment is credible because minority shareholders know that if the controlling owner unexpectedly extracts high levels of private benets when he/she still holds a substantial amount of shares, they will discount the stock price accordingly, and the majority owners share value will be reduced. In equilibrium, the majority shareholder will hold a large ownership stake and the stock price of the company will be higher. Thus, ownership concentration has an incentive alignment effect: increasing an owners share ownership beyond the minimum level needed for effective control improves the alignment of interests between the controlling owner and the minority shareholders and reduces the effects of entrenchment. 2.2.3. Entrenchment effect when control exceeds ownership. In addition to the characteristic concentrated ownership, the ownership arrangements of East Asian corporations are further complicated by pyramidal and cross-holding structures. These ownership arrangements allow controlling owners to commit low equity investment while maintaining tight control of the rm, creating a separation in control (voting rights) and ownership (cash ow rights).7
6 Scott (1999) studies the role of corporate governance in four Asian countries that were in nancial crisis: Indonesia, Malaysia, South Korea and Thailand. He concludes by recommending that strengthening the effective limits on self-dealing transactions of controlling owners would be the priority task for these countries. 7 Separation between cash ow and voting rights is common among public corporations around the world (La Porta et al., 1999). In the context of diffuse ownership, Stulz (1988) suggests that there exist various contractual arrangements that allow managers to increase their voting power to a degree beyond

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

407

One consequence of the divergence between voting and cash ow rights is that the controlling owner becomes entrenched with high levels of control, while the low equity ownership level provides only a low degree of alignment between the controlling owner and minority shareholders.8 A controlling owner in this situation could extract wealth from the rm, receive the entire benet, but only bear a fraction of the cost.9 We offer a simple pyramidal structure to illustrate this point. An entrepreneur owns 25% of the stock in publicly traded Firm A, which in turn owns 32% of the stock in Firm B. In the most modest scenario, we note that the entrepreneur controls 25% of Firm Bthe weakest link in the chain of voting rights. At the same time, the entrepreneur owns about 8% of the cash ow rights of Firm B, the product of the two ownership stakes along the chain. Given this ownership structure, it costs the entrepreneur only $8 for every $100 expropriated from Firm B. Clearly, if stock pyramids or cross-shareholdings were used to consolidate control, they would also result in a separation between ownership and control, which exacerbates the entrenchment problem of controlling owners. In summary, once controlling owners achieve effective control, their ownership concentration has two incentive effects on them: entrenchment and alignment. When there is no separation between voting and cash ow rights, concentrating ownership beyond the minimum level for effective control enhances the alignment of interest and hence mitigates the entrenchment effect. When the voting rights and cash ow rights diverge, however, the lower cash ow rights may fail to provide sufcient incentive alignment to curtail the entrenchment effect. 2.3. Ownership structure and earnings informativeness We now discuss the relations between ownership structure and earnings informativeness in East Asia. We provide two potential arguments that may explain the relations. The rst argument is based on the entrenchment effect discussed above. The second argument is related to the rms proprietary information and specic human capital effect, which will be detailed below.
( footnote continued) their equity ownership. Such arrangements may lead to changes in capital structures or differential voting rights, which in turn provide managers more control than what they are entitled to by their equity ownership. What causes the separation between voting and cash ow rights is a subject not adequately addressed in the literature. It is potentially related to both controlling owners nancing constraints and their desires to maintain control. Owners may not have sufcient personal wealth or loans to nance the investment projects that they desire to control. With limited amounts of cash, owners could leverage their control by way of stock pyramids or cross-shareholdings. The ownership arrangements, which help in the formation of business groups, also allow the controlling owners to create internal capital and factor markets and hence to bypass underdeveloped external markets. 8 Controlling owner entrenchment as an agency cost of the separation of cash ow rights from voting rights plays a key role in the theoretical models of Burkart et al. (1997, 1998), Bebchuk et al. (2000), and Wolfenzon (1999). 9 Although minority shareholders appear to have been expropriated by the controlling owner, they may be price-protected when they bought the shares. That is, if the minority shareholders purchased the shares after control existed, the stock price should be discounted for this action and the discount would on average pay for the apparent transfer.

408

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

2.3.1. The entrenchment argument Just as the share ownership structure delineates a rms agency problem, it also impacts the rms reporting. When an owner effectively controls a rm, he/she also controls the production of the rms accounting information and reporting policies. When the controlling owner is entrenched by his/her voting power and there is a large separation of the voting and cash ow rights, the credibility of the accounting information is reduced. That is, outside investors pay less attention to the reported accounting numbers, because they expect that the controlling owner reports accounting information out of self-interest rather than as a reection of the rms true underlying economic transactions. In particular, outside investors may not trust the rms reported earnings because the controlling owner may manipulate earnings for outright expropriation. In addition, outside investors know that the controlling owner has an incentive to avoid reporting accounting information that would attract close monitoring by outside shareholders. This does not always mean that there is outright earnings manipulation to cover up possible earnings effects of wealth extraction. The controlling owner may simply bury the wealth effects of his/her expropriation activities in the aggregate earnings numbers without reporting them as separate income statement items. The loss of credibility in earnings reports lowers the stock price informativeness of the earnings. Prior studies have noted the importance of the effects of earnings credibility. Teoh and Wong (1993) report that the market perception of the quality of accounting earnings, as proxied by the size of the rms auditor, positively affects the stock price informativeness of earnings. 2.3.2. The information argument Concentrating ownership allows rms to limit their information disclosure to the public. Opacity is a good strategy because it prevents leakage of proprietary information to competitors and allows rms to avoid unwanted political or social scrutiny. Firms with proprietary knowledge and specic human capital tend to concentrate their ownership and decision rights in the individuals who possess the specic knowledge (Jensen and Meckling, 1992; Christie et al., 2002). Assigning control to individuals without specic knowledge can lead to suboptimal decisions or a high cost of transferring knowledge to necessary individuals. Moreover, if rms with proprietary knowledge give more individuals decision rights, they also have to give those individuals the proprietary knowledge to allow them to make informed decisions. The larger the set of informed individuals, the larger the likelihood that the proprietary information leaks to the public and potential competitors. This scenario is common among rms that engage in political rent-seeking activities, which are prevalent and highly protable in East Asia.10 Morck (1996) argues that there are two reasons why closely held rms are better able to engage in political lobbying than widely held rms. First, the concentrated decision rights within the
Fisman (2001) conducted an event study on the stock price effects of the news announcements of Suhartos illness. He analyzed the value drops in the rms connected to Suharto and reported that the proportion of these rms share values attributed to Suharto connections was very largeabout a quarter of each rms share value. Political connections were valued by investors in this case.
10

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

409

rms allow them to operate in greater secrecy. A politician may desire more secrecy in order to maintain a reputation of integrity. The rms also prefer operating in secret in order to discourage entry by competitors. Second, compared to a hired manager in a widely held rm, a controlling owner is more secure in his position in the rm and thus has more credibility on which to trade favors with the government. In order to trade favors with politicians and bureaucrats in secret, rms generally have highly concentrated ownership that allows them to have tight control of information ows to the public, which in turn reduces corporate transparency. In this business environment, it is in the interest of both the controlling owners and the minority shareholders to release as little accounting information to the public as possible. This information effect argument suggests that high ownership concentration is associated with low earnings informativeness.

2.3.3. Predicted relations The relation between ownership concentration and earnings informativness is affected by the several factors discussed earlier. On the one hand, there are two incentive effects of ownership concentration that impact earnings informativeness. First, the entrenchment effect predicts that concentrated ownership leads to weak earnings informativeness. However, as the degree of ownership concentration exceeds the minimum level needed for effective control, the alignment effect may mitigate the entrenchment effect. At such level, further increasing the ownership concentration can improve earnings informativeness. On the other hand, the information argument suggests a negative relation between ownership concentration and earnings informativeness. As these two incentive effects have opposing predictions and could coexist with the information effect, the relation between ownership concentration and earnings informativeness is ambiguous and needs to be addressed empirically. Attempting to disentangle the two incentive effects of ownership concentration on earnings informativeness, we employ the degree of divergence between voting and cash ow rights as a proxy for controlling owner entrenchment. We expect that the credibility of a rms accounting information and consequently the informativeness of this information to outside investors decreases with an increase in the degree to which the level of voting rights exceeds the associated level of cash ow rights. Formally, our hypothesis is: an increase in the degree of divergence between the controlling owners cash ow rights and voting rights decreases the informativeness of the rms earnings. We acknowledge that the divergence measure may also reect the information effect. That is, a controlling owner who has proprietary information to protect may use stock pyramids and cross-shareholdings to leverage his/her control concentration; and the divergence of these rights may tend to increase with the rms desired opacity. Finding a negative relation between earnings informativeness and the level of divergence between the two rights, although consistent with the entrenchment argument, is also consistent with the information argument.

410

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

3. Empirical analysis In this section, we describe the sample, data sources, and the ownership structures of the sample rms. We then test the hypothesis developed in the previous section by analyzing the relations between the corporate ownership structure and earnings informativeness. 3.1. Sample and data We select our sample rms from seven East Asian economiesHong Kong, Indonesia, Malaysia, Singapore, South Korea, Taiwan and Thailand. We include rms that have sufcient ownership, stock returns, earnings and other nancial data for empirical analysis. Below is a description of the sample and data sources. 3.1.1. Ownership data Most prior studies of ownership structures focus on immediate ownership common shares directly owned by individuals or institutions. Immediate ownership is not sufcient for characterizing the ownership and control structure of East Asian rms, as these rms are generally associated with complicated indirect ownership. As a departure from these prior studies, we focus on ultimate ownership. We use data assembled by Claessens et al. (2000), which identies the ultimate owners of 2,980 rms in nine East Asian economies, along with their shares of cash ow and voting rights. The ownership data is available for 1996. The procedure of identifying ultimate owners is similar to the one used in La Porta et al. (1999). An ultimate owner is dened as the shareholder who has the determining voting rights of the company and who is not controlled by anybody else. If a company does not have an ultimate owner, it is classied as widely held. To economize on the data collection task, the ultimate owners voting right level is set at 50% and not traced any further once that level exceeds 50%. Although a company can have more than one ultimate owner, we focus on the largest ultimate owner. As our denition of ownership relies on both cash ow and voting control rights, the cash ow rights that support the control by ultimate owners are further identied. Firm-specic information on pyramid structures and cross-holdings are used to make the distinction between cash ow and voting rights. To facilitate the measurement of the separation of cash ow and voting rights, the maximum cash ow rights level associated with any ultimate owner is also set at 50%. However, there is no minimum cutoff level for cash ow rights. From the 2,980 rms, we exclude 1,240 Japanese rms from our analysis because Japans institutional environment and its rms ownership structures are quite different from the other East Asian economies.11 We further exclude 319 rms whose
Different from the East Asian rms that are typically family controlled, the dominant ultimate owners of Japanese rms are institutions, typically the main banks of industrial groups. Japanese rms ownership structures are also quite different from those of the East Asian rms in both the degree of control and cash-vote divergence.
11

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

411

largest ultimate owners have less than 20% of voting rights. This restriction allows us to focus on rms with controlling shareholders and is expected to increase the power of our test since the entrenchment and information arguments are more applicable to ultimate owners that have already secured effective control. La Porta et al. (1999) also use the 20% cutoff level to dene control. Bradley and Kim (1985) report that tender offers rarely occur in rms with control at the 20% level. 3.1.2. Stock return and nancial data We merge the ownership data with the PACAP electronic database, which is commercially distributed by the University of Rhode Island. PACAP contains the nancial and stock return data of publicly traded companies of the seven East Asian economies analyzed here. We select 1991 through 1995 as the period of analysis and retrieve the stock return and nancial data for that period. An exception is Korea, for which we have data up to 1994 only. We do not include 1996 because the data are not available to us for that year. We also exclude pre-1991 data because we are concerned that the ownership structures earlier than 1991 may differ too much from the structures documented in 1996. Although we have ownership data for the Philippines, we do not include rms from that country because they are not covered by the PACAP database. The merging of the 1996 ownership data and the 1991 1995 stock return and nancial data requires us to assume that the ownership and control structures of the rms did not change substantially during that period. This is a reasonable assumption since the economic and political conditions were relatively stable at the time. The nal sample has 977 rms with a total of 3,572 rmyears.12 3.2. Basic statistics of ownership structures Table 1 presents the descriptive statistics of the cash ow and voting rights of the largest ultimate owners of the nal sample. Broken down by economies, the sample covers 282 Hong Kong rms, 91 Indonesian rms, 177 Malaysian rms, 133 Singaporean rms, 95 South Korean rms, 66 Taiwanese rms, and 133 Thai rms. The sample covers 30% of all publicly traded rms in the region.13 Panel A of the Table shows that the mean voting rights of the East Asian corporations is 30.44%. In a quarter of the East Asian companies, more than 40% of the voting rights are in the hands of the largest ultimate owner. Thai rms display the most concentrated voting rights, 36.32% on average, followed by Indonesian rms (34.51%), Malaysian rms (30.73%), Hong Kong rms (29.68%), Singaporean rms (28.95%), South Korean rms (26.11%), and Taiwanese rms (24.70%). The high control concentration is not surprising, given the 20% voting rights restriction imposed on the sample. However, the control concentration remains high when the
The two extreme percentiles of rm-year observations of annual stock returns and net earnings over market value of equity (see Section 3.3 for the two variable denitions) are eliminated from the sample. 13 As of December 1996, the numbers of listed rms in these economies were: 583 in Hong Kong, 267 in Indonesia, 760 in Korea, 621 in Malaysia, 266 in Singapore, 382 in Taiwan, and 454 in Thailand.
12

412

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

Table 1 Cash ow and voting rights of the largest ultimate owners of East Asian corporationsa Firm no. Panel A: voting rights (%) Hong Kong 282 Indonesia 91 Malaysia 177 Singapore 133 South Korea 95 Taiwan 66 Thailand 133 East Asia 977 Mean 29.68 34.51 30.73 28.95 26.11 24.70 36.32 30.44 Std. err. 10.91 9.92 10.17 9.40 7.48 7.49 11.18 10.53 First quartile 20.00 30.00 20.00 20.00 20.00 20.00 30.00 20.00 Median 30.00 30.00 30.00 30.00 20.00 20.00 40.00 30.00 Third quartile 40.00 40.00 40.00 40.00 30.00 30.00 50.00 40.00 Minimum 20.00 20.00 20.00 20.00 20.00 20.00 20.00 20.00b Maximum 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00c

Panel B: cash ow rights (%) Hong Kong 282 25.60 Indonesia 91 26.08 Malaysia 177 26.03 Singapore 133 22.96 South Korea 95 21.36 Taiwan 66 21.46 Thailand 133 34.23 East Asia 977 25.84

11.19 10.81 11.45 10.48 8.16 8.76 11.68 11.37

20.00 18.00 20.00 16.00 18.00 16.00 24.00 20.00

20.00 25.00 22.00 20.00 20.00 20.00 30.00 21.00

30.00 32.00 30.00 30.00 30.00 24.00 40.00 30.00

3.00 4.00 5.00 6.00 4.00 8.00 4.00 3.00

50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00c

Panel C: the ratio of cash ow rights over voting rightsd Hong Kong 282 0.87 0.22 0.80 Indonesia 91 0.77 0.24 0.60 Malaysia 177 0.84 0.22 0.70 Singapore 133 0.78 0.22 0.60 South Korea 95 0.83 0.25 0.60 Taiwan 66 0.86 0.17 0.73 Thailand 133 0.95 0.16 1.00 East Asia 977 0.85 0.22 0.70

1.00 0.80 1.00 0.80 1.00 1.00 1.00 1.00

1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

0.13 0.13 0.17 0.27 0.13 0.40 0.20 0.13

1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

Sample: The sample consists of 977 rms from Hong Kong, Indonesia, Malaysia, Singapore, South Korea, Taiwan, and Thailand. To be included in the sample, a rm must have at least 1 year of stock returns and nancial data from the PACAP database between 1991 and 1995, its ultimate ownership data must be available from Claessens et al. (2000), and its largest ultimate owner must have at least 20% of voting rights. a An ultimate owner is dened as an owner that is not controlled by anybody else. Although a rm can have more than one ultimate owner, only the one with the largest voting rights is reported. b Sample rms are restricted to have a largest ultimate owner with at least 20% voting rights. c The voting rights and cash ow rights level of an ultimate owner is set at 50% and not traced any further once that level exceeds 50%. d The ratio of cash ow rights over voting rights is dened as the level of cash ows rights divided by the level of voting rights. The value of this ratio ranges from zero to one.

restriction is relaxed. Claessens et al. (2000) employ a lower minimum voting rights cutoff, 5% (instead of 20%), and report that the mean voting rights of the seven economies range from 35.25% (Thailand) to 17.78% (South Korea).

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

413

Panel B reports the basic statistics for levels of cash ow rights. The cash ow rights patterns are similar to the voting rights patterns in Panel A. The overall average concentration is 25.84%. Note particularly that the mean levels of cash ow rights are lower than the corresponding levels for voting rights in Panel A, indicating the divergence between cash ow and voting rights. In Panel C, we report the basic statistics of the ratio of cash ow rights over voting rights (CV). The ratio, by denition, ranges between zero and one. If a rm is widely held, i.e., it has zero cash ow and voting rights, its CV ratio is set to one. The CV ratio indicates the degree of divergence between cash ow and voting rights. The closer the ratio is to zero, the larger the divergence. In East Asia, the mean CV ratio is 0.85. The mean CV ratios are rather similar across the seven East Asian economies, ranging between 0.77 (Indonesia) and 0.95 (Thailand). Over a quarter of the East Asian rms display cashvote divergence (CVo1). The minimum CV ratio is 0.13. The Pearson correlation coefcient between the voting rights and CV is only 0.07, suggesting that CV is not simply a proxy for voting rights. The actual degree of ownership concentration in East Asia should be higher than the statistics reported here for several reasons. First, the 50% ceiling for the ownership data and the inability to trace some hidden control chains would bias our statistics downward. In addition, small rms tend to have more concentrated ownership, but our sample mainly consists of larger rms due to the limited availability of ownership data of small rms. However, we expect that the understatement of ownership data and the large rm bias in our sample would weaken but not systematically bias the data in favor of our hypothesis. Notwithstanding the data limitation, it is sufcient to conclude from Table 1 that the typical ownership and control structure of the East Asian rms is highly concentrated, in contrast to the diffuse ownership in the U.S., as documented in prior research. The East Asian rms also differ from U.S. rms in that they are characterized by a separation of ownership and control resulting from the controlling owners possession of more voting power than cash investment. 3.3. Regression analysis We next perform regression analysis to examine the relations between corporate ownership structure and earnings informativeness in East Asia. 3.3.1. Basic relations between returns and earnings Before we focus on the role of the ownership structure, we perform a set of ordinary least squares regressions to determine the basic relations between stock returns and earnings in East Asia: CARit a0 a1 NIit Fixed effects uit ; 1

where, for sample rm i; CARit is the cumulative net-of-market 12-month stock returns at year t; NIit the net earnings at year t divided by the market value of equity at the beginning of year t; xed effects the dummy variables controlling for xed effects of calendar years and/or economies, and uit the error term at year t:

414

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

The regressions are performed year by year, economy by economy, and pooling all of the years and economies. The results are reported in Table 2. Because we generally nd heteroskedasticity problems in the regressions, we report White-adjusted tstatistics for all the coefcients. Fixed effects of calendar years and/or economies, where appropriate, are included as dummy intercepts in the regressions. For simplicity, they are not reported in the table. The estimated coefcients of earnings (NI) are positive and statistically signicant across all the years and economies, suggesting that earnings have an information role in East Asia. 3.3.2. The effects of ownership structure We next test the informativeness of earnings conditional on ownership structure using the following pooled time-series cross-sectional regression model: CARit a0 a1 NIit a2 NIit SIZEit a3 NIit Qit a4 NIit LEVit a5 NIit SEGi a6 NIit Vi a7 NIit CVi Fixed effects uit ; 2 where, for sample rm i; CARit is the cumulative net-of-market 12-month stock returns at year t; NIit the net earnings at year t divided by the market value of equity at the beginning of year t; SIZEit the natural logarithm of the market value of equity in millions of U.S. dollars at the beginning of year t; Qit the market value of equity divided by the book value of total assets at the beginning of year t; LEVit the total liability divided by total assets at the beginning of year t; SEGi the number of industry segment(s) in which the rm operates, Vi the voting rights level of the largest ultimate owner, CVi the ratio of cash ow rights over voting rights of the largest ultimate owner, xed effects the dummy variables controlling for xed effects of calendar years and economies, and uit the error term at year t: We include a set of variables to control for observed variations in the earnings return relation that are likely due to causes other than voting rights or divergence of cash ow and voting rights. The ratio of market value of equity to the book value of total assets (Q) is included to control for the effects of growth on the earningsreturn relation.14 Growth opportunities are likely to be positively associated with future earnings levels and/or earnings persistence (Collins and Kothari, 1989). The higher the market-to-book assets, the larger the expected earnings growth and/or earnings persistence, the stronger the earningsreturn relation.15 On the other hand, the market-to-book ratio may also be affected by rm risk. High growth rms may be more risky, which weakens the earningsreturn relation. Also, fast growing rms are likely to be young rms with less informative earnings. Given these countervailing effects, the net effect of growth on the earningsreturn relation is therefore an empirical issue. We also incorporate leverage (LEV) in the regression. Leverage could be a proxy for the riskiness of debt or default risk (Dhaliwal et al., 1991). Highly levered rms are associated with high risk and hence their earningsreturn relation is weakened.
The use of the market-to-book value of equity produces qualitatively similar results in our regressions. We do not include a separate control for earnings persistence because the earnings history is inadequate for its empirical estimation in our sample.
15 14

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425 Table 2 Simple regressions of stock returns on earningsa Intercept 1991 1992 1993 1994 1995 Hong Kong Indonesia Malaysia Singapore South Korea Taiwan Thailand Pooled 0.23*** (6.90)b 0.05 (1.39) 0.25*** (8.19) 0.25*** (12.12) 0.14*** (8.25) 0.17*** (6.57) 0.07 (1.26) 0.12*** (4.52) 0.04 (1.58) 0.00 (0.08) 0.14*** (3.87) 0.13*** (3.16) 0.18*** (10.79) NI 1.26*** (6.19) 0.76*** (3.77) 1.09*** (6.08) 1.06*** (7.51) 0.76*** (6.20) 0.73*** (7.36) 0.73*** (3.85) 0.80*** (3.25) 0.64*** (2.89) 1.67*** (6.62) 1.65** (2.56) 2.62*** (7.99) 0.99*** (13.01) Adjusted R2 0.15 0.04 0.10 0.17 0.11 0.10 0.06 0.04 0.07 0.16 0.23 0.17 0.08

415

Number of observations 652 727 784 817 772 1045 259 684 561 346 311 546 3752

Model specication: CARit a0 a1 NIit Fixed effects uit : Variable denitions: CARit =the cumulative net-of-market 12-month stock returns for rm i at year t: The annual returns are continuously compounded from monthly stock returns starting from 12 months before the latest date, as required by law or listing rules, that the rm discloses its annual report. NIit =the net earnings at year t divided by the market value of equity at the beginning of year t for rm i: Sample: The sample includes 3,752 rm-year observations, spanning between 1991 and 1995 and covering 977 rms from Hong Kong, Indonesia, Malaysia, Singapore, South Korea, Taiwan, and Thailand. To be included in the sample, a rm must have at least 1 year of stock returns and nancial data from the PACAP database between 1991 and 1995, its ultimate ownership data must be available from Claessens et al. (2000), and its largest ultimate owner must have at least 20% of voting rights. Observations with largest and smallest 1% values of CAR and NI are excluded from the sample. ***Indicates signicant at 1% (two-tailed). **Indicates signicant at 5% (two-tailed). a Simple OLS regressions of annual stock returns on earnings are run year by year, economy by economy and pooling across years and economies (pooled). Fixed-effects of calendar years and/or economies are included for each regression when appropriate, but not reported. b White-adjusted t-statistics are in parentheses.

On the other hand, Smith and Watts (1992) suggest that leverage can proxy for a rms investment opportunity set. Mature rms with low growth opportunities generally have high leverage and are likely to have informative earnings. Hence rms

416

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

with high leverage may have higher earningsreturn sensitivity than rms with low leverage. Taking the risk and the growth effects together, the net effect of leverage on the earningsreturn relation is to be determined empirically. In addition, we include the number of industry segments (SEG) in which each sample rm operates as another control. Conglomerate rms, due to their relatively more complex earnings-generating process, may have weaker earningsreturn relations than rms operating in a single industry.16 Finally, we include rm size (SIZE) as a control for other missing factors that affect the earningsreturn relation. For example, prior literature on the U.S. case (Atiase, 1985; Freeman, 1987) has documented that public disclosure and private development of non-earnings information are increasing functions of rm size. The summary statistics of the dependent and independent variables of the regression are reported in Table 3. The mean CAR is 1.12% and the mean NI is 7.27%, but they are statistically insignicantly different from zero due to their large standard errors. Both CAR and NI display large dispersion in value in the sample. The summary statistics of V and CV are consistent with those reported in Table 1. The owners on average possess high voting control: 30% of voting rights. They also control signicantly more votes than those determined by their ownership positions. Specically, the mean CV is 0.85, indicating that the controlling owners levels of cash ow rights are on average 15% lower than their corresponding levels of voting rights. Also shown in the basic statistics, the values of the control variables are mostly distributed within reasonable ranges, suggesting that the regression results are unlikely to be plagued by outliers. We employ the ordinary least-squares method to regress CAR on voting rights level (V), the degree of separation between cash ow and voting rights (CV), and the control variables.17 The regression results are presented in Eq. (1) of Table 4. We report White-adjusted t-statistics for all the coefcients due to heteroskedasticity. For simplicity, we omit reporting xed effects of calendar years and economies. Larger rms earnings are more informative, as indicated by the signicantly positive estimated coefcient of NI * SIZE. The coefcient of NI * Q is insignicant, suggesting that the risk and the growth effects are offset by each other. The estimated coefcient of NI * LEV is signicantly positive, consistent with the view that highly levered rms tend to be mature rms that have more informative earnings. The coefcient of NI * SEG is signicantly negative, suggesting that conglomerate rms report less informative earnings than more focused rms. The coefcient of NI is insignicantly negative. It does not suggest that earnings are uninformative because the simple regressions in Table 2 show that stock returns

16 The 1996 company segment data were collected from Worldscope and supplemented with additional data from the Asian Company Handbook. Since companies report their segment data with different degrees of detail, the companies segments are grouped according to the two-digit Standard Industry Classication system. 17 We also try adjusting V by subtracting the median voting rights of the economy in which it operates because rms in different economies are associated with different levels of ultimate voting control and the results are qualitatively similar. For simplicity, we report only the results with the unadjusted V.

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425 Table 3 Summary statistics of regression variables Variables CAR (%) NI (%) SIZE Q LEV (%) SEG V (%) CV Mean 1.12 7.27 12.03 1.09 46.83 2.55 29.93 0.85 Std. err. 33.46 8.32 1.37 0.95 23.57 1.61 10.37 0.22 First quartile 23.35 3.53 11.07 0.41 28.52 1.00 20.00 0.70 Median 6.96 6.31 11.99 0.83 44.16 2.00 30.00 1.00 Third quartile 14.02 10.40 12.87 1.47 62.59 4.00 40.00 1.00 Minimum 62.53 45.20 7.13 0.02 0.08 1.00 20.00 0.13

417

Maximum 147.49 461.70 17.05 7.98 259.95 9.00 50.00 1.00

Variable denitions: CAR=the cumulative net-of-market 12-month stock returns. The annual returns are continuously compounded from monthly stock returns starting from 12 months before the latest date, as required by law or listing rules, that the rm discloses its annual report. NI=the net earnings divided by the market value of equity at the beginning of the year. SIZE=the natural logarithm of the market value of equity in millions of U.S. dollar at the beginning of the year. Q=the market value of equity divided by the book value of total assets at the beginning of the year. LEV=the total liability divided by total assets at the beginning of the year. SEG=the number of industry segment(s). V=the voting rights level of the largest ultimate owner. CV=the ratio of cash ow rights over voting rights of the largest ultimate owner. Sample: The sample includes 3,752 rm-year observations, spanning between 1991 and 1995 and covering 977 rms from Hong Kong, Indonesia, Malaysia, Singapore, South Korea, Taiwan, and Thailand. To be included in the sample, a rm must have at least 1 year of stock returns and nancial data from the PACAP database between 1991 and 1995, its ultimate ownership data must be available from Claessens et al. (2000), and its largest ultimate owner must have at least 20% of voting rights. Observations with largest and smallest 1% values of CAR and NI are excluded from the sample.

are signicantly positively associated with earnings. Their relation becomes insignicant in Eq. (1) because the inclusion of the additional independent variables may have controlled for most of the variation of NI and lowered its explanatory power. The intercept remains signicantly negative as in the regression results in Table 2. The negative intercept could be caused by the omitted expected earnings component. When we include lagged earnings as expected earnings by replacing NI with the change in earnings (current year earnings minus lagged earnings all divided by the lagged market value of equity) in the regression, the magnitude of the intercept drops by more than half. The focus of Table 4 is the role of the ownership structure. The result in Eq. (1) shows that the estimated coefcient of NI * V is negative and statistically signicant at the 5% level. This result is consistent with the information effect that high voting rights are associated with secrecy and low earnings informativeness. The result also suggests that the information effect dominates the incentive alignment effect, which predicts that additional ownership concentration beyond the minimum level of effective control increases earnings informativeness. To gain understanding of the economic importance of the result, we use the estimated regression of Eq. (1), set all of the independent variables at their mean, and then estimate how much CAR would change given that V increases by a magnitude of one standard deviation. We nd

418

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

Table 4 Pooled time-series cross-sectional regressions with interactions for corporate ownership structurea (1)b Intercept NI NI * SIZE NI * Q NI * LEV NI * SEG NI * V NI * CV Adjusted R2 Number of observations 0.18*** (10.39)d 0.59 (1.17) 0.10*** (2.67) 0.00 (0.05) 1.03*** (4.09) 0.09*** (2.71) 1.20** (2.47) 0.69*** (3.49) 0.10 3752 (2)c 0.17*** (9.09) 0.99* (1.79) 0.12*** (2.89) 0.03 (0.34) 0.99*** (3.71) 0.11*** (2.83) 0.21 (0.29) 0.74*** (3.44) 0.10 3284

Model specication: CARit a0 a1 NIit a2 NIi * SIZEit a3 NIit * Qit a4 NIit * LEVit a5 NIi * SEGi a6 NIit * Vi a7 NIit * CVi Fixed effects uit : Variable denitions: CARit =the cumulative net-of-market 12-month stock returns for rm i at year t: The annual returns are continuously compounded from monthly stock returns starting from 12 months before the latest date, as required by law or listing rules, that the rm discloses its annual report. NIit =the net earnings at year t divided by the market value of equity at the beginning of year t for rm i: SIZEit =the natural logarithm of the market value of equity in millions of U.S. dollar at the beginning of year t for rm i: Qit =the market value of equity divided by the book value of total assets at the beginning of year t for rm i: LEVit =the total liability divided by total assets at the beginning of year t for rm i: SEGi =the number of industry segment(s) of rm i. Vi =the voting rights level of the largest ultimate owner of rm i: CVi =the ratio of cash ow rights over voting rights of the largest ultimate owner of rm i: Sample: The sample includes 3,752 rm-year observations, spanning between 1991 and 1995 and covering 977 rms from Hong Kong, Indonesia, Malaysia, Singapore, South Korea, Taiwan, and Thailand. To be included in the sample, a rm must have at least 1 year of stock returns and nancial data from the PACAP database between 1991 and 1995, its ultimate ownership data must be available from Claessens et al. (2000), and its largest ultimate owner must have at least 20% of voting rights. Observations with largest and smallest 1% values of CAR and NI are excluded from the sample. ***Indicates signicant at 1% (two-tailed). **Indicates signicant at 5% (two-tailed). *Indicates signicant at 10% (two-tailed). a Two OLS regressions are run pooling across years and economies. Fixed-effects of calendar years and economies are included in the regressions but not reported. b Eq. (1) uses the nal sample that excludes observations with V o20%: c Eq. (2) uses the sub-sample that excludes both V o20% and V X50%: d White-adjusted t-statistics are in parentheses.

that when V increases from its mean, 30%, by one standard deviation to 40%, the level of CAR decreases by 1%, representing a 9% change relative to its previous level.

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

419

We now turn to investigate the effect of separation of cash ow and voting rights on earnings informativeness. CV, by denition, is inversely related to cash-vote divergence. To be consistent with the entrenchment effect and/or the information effect, we should observe a signicantly positive estimated coefcient of CV. Consistent with our conjecture, the coefcient of CV is positive and statistically signicant at the 1% level. In terms of economic signicance, when all of the independent variables are set at their mean, decreasing CV from its mean (0.85) by one standard deviation to 0.63 is associated with a 1% decrease in the level of CAR, representing an 11% drop relative to its previous level.18 In summary, when ultimate owners effectively control their rms, their voting rights levels are negatively related to the informativeness of the rms reported earnings. This suggests that the information effect dominates the incentive alignment effect. We also nd evidence supporting our hypothesis that after controlling for the level of voting rights, cash-vote divergence signicantly weakens earnings informativeness. This evidence is consistent with the entrenchment effect. However, the result is also consistent with the information effect, provided that controlling owners tend to employ pyramidal or cross-holding ownership structures to protect the information about their rent-seeking activities. 3.3.3. Checks of robustness Since all voting and cash ow rights that exceed 50% are capped, the effects of any variation in voting and cash ow rights of these rms would not be captured by our measure. Moreover, if actual voting and cash ow rights both exceed 50%, their divergence would not be captured by the CV measure, as it would be recorded as one, which indicates no divergence. As a sensitivity test of any possible bias in our results, we rerun the regression by excluding observations associated with voting rights equal to or more than 50% from the full sample. As reported in Eq. (2) of Table 4, the sign of the coefcient of NI * V becomes statistically insignicant. On the other hand, the sign, magnitude, and signicance level of the coefcient of NI * CV remains qualitatively similar to results in Eq. (1). We thus conrm that the 50% ceiling for our ownership data would not have biased the results in favor of our hypothesis. As further diagnostic checks,19 we have estimated the regression models using cumulative abnormal returns calculated from two xed annual windows: nine (six) months prior to and three (six) months after the current scal year end. We have also used cumulative raw returns, instead of net-of-market returns, and used the two-year cumulative net-of-market returns, starting 21 months before to three months after the scal year-end, as an alternative dependent variable. The two-year returns, which
18 We have also run the same regression by including rms with the largest ultimate owners having o20% voting rights. The results support our hypothesis that rms with a larger separation between their cash ow and voting rights have less informative earnings. Consistent with our conjecture that the information and entrenchment effects are weaker among rms without owners commanding effective control, the magnitude (t-statistics) of the coefcients of NI * V and NI * CV in this regression has (have) dropped from 1.2 to 0.60 (2.47 to 1.43) and 0.69 to 0.47 (3.49 to 2.46), respectively. 19 The results of the diagnostic checks are not reported in the tables but are available upon request.

420

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

include both current and lagged-year returns, attempt to adjust for any differences in price efciency in capturing future earnings between highly concentrated (high V) and less concentrated ownership (low V) rms (Jacobson and Aaker, 1993; Ali and Hwang, 2000). In addition, instead of using NI in our regression model, we have also used DNI, change in earnings (current earnings minus lagged earnings all divided by lagged market value of equity). Our results for V and CV remain qualitatively similar after using DNI or these alternative measurements of CAR. We provide a further test of whether or not the effects of voting control and cashvote divergence on East Asian rms cluster in time and/or economies. Table 5 presents the results of a set of year-by-year regressions. These regressions include NI * V and NI * CV, in addition to the control variables. We nd that the coefcients of NI * V and NI * CV are mostly of the expected signs, and NI * CV is statistically signicant in all years but 1995, while NI * V is signicant in 1991 and 1994. Table 6 presents the results of a set of economy-by-economy regressions using the same model. We nd ownership effects in several economies. The effect of V on earnings informativeness is negative and signicant in Hong Kong, Malaysia, and Taiwan. The effect of CV is positive and signicant in Hong Kong, Malaysia, Taiwan and Thailand. From the year-by-year and economy-by-economy results, we nd that the effects of the ownership variables are not concentrated in any given year or economy. The above diagnostic checks have demonstrated that our empirical results are robust to the measurement bias in the ownership variables and to the various specications of cumulative stock returns and earnings. In addition, the ownership effects are generally found in our sample, not just in any single year or economy.

4. Conclusion The Asian nancial crisis has caused East Asian economies to re-examine the adequacy of their corporate nancial reporting. Despite efforts to improve corporate transparency by imposing new accounting and disclosure rules in East Asia, the perception is that the nancial reporting quality of corporations remains low. We hypothesize that the high ownership concentration and the large separation of ownership and control, which are common in East Asia, weaken the informativeness of reported earnings to outside investors. We provide two explanations for this relation. The rst explanation is based on the controlling owners entrenchment. The earnings credibility is weakened because minority shareholders anticipate that the ownership structure gives the controlling owners both the ability and incentive to manipulate earnings for outright expropriation or to report uninformative earnings to avoid detection of their expropriation activities. The second explanation is related to proprietary information. As rent-seeking activities are prevalent and highly protable in this region, it is in the best interest of rent-seekers to concentrate ownership and decision rights, so that the specic knowledge about their activities can be concealed from potential competitors and the public. This information effect

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425 Table 5 Year-by-year regressions with interactions for corporate ownership structurea 1991 Intercept NI NI * SIZE NI * Q NI * LEV NI * SEG NI * V NI * CV Adjusted R2 Number of observations 0.26*** (7.51)b 1.28 (1.03) 0.29*** (3.39) 0.32 (1.43) 0.44 (1.09) 0.14* (1.87) 3.14*** (3.26) 0.99** (2.17) 0.20 652 1992 0.01 (0.28) 1.25 (1.05) 0.18* (1.94) 0.13 (0.61) 1.46** (2.13) 0.06 (0.67) 0.07 (0.06) 1.13** (2.19) 0.06 727 1993 0.27*** (8.35) 2.61** (2.08) 0.33*** (3.61) 0.22 (1.22) 0.10 (0.19) 0.12* (1.77) 0.85 (0.76) 0.88* (1.89) 0.12 784 1994 0.25*** (10.97) 0.24 (0.23) 0.01 (0.12) 0.26* (1.79) 1.54*** (3.42) 0.04 (0.55) 2.30** (2.42) 0.77* (1.82) 0.19 817 1995

421

0.13*** (6.34) 0.45 (0.48) 0.03 (0.38) 0.34* (1.80) 2.03*** (4.70) 0.11* (1.84) 0.05 (0.06) 0.01 (0.03) 0.13 772

Model specication: CARit a0 a1 NIit a2 NIit * SIZEit a3 NIit * Qit a4 NIit * LEVit a5 NIit * SEGi a6 NIit * Vi a7 NIit * CVi Fixed effects uit : Variable denitions: CARit =the cumulative net-of-market 12-month stock returns for rm i at year t: The annual returns are continuously compounded from monthly stock returns starting from 12 months before the latest date, as required by law or listing rules, that the rm discloses its annual report. NIit =the net earnings at year t divided by the market value of equity at the beginning of year t for rm i: SIZEit =the natural logarithm of the market value of equity in millions of U.S. dollar at the beginning of year t for rm i: Qit =the market value of equity divided by the book value of total assets at the beginning of year t for rm i: LEVit =the total liability divided by total assets at the beginning of year t for rm i: SEGi =the number of industry segment(s) of rm i. Vi =the voting rights level of the largest ultimate owner of rm i: CVi =the ratio of cash ow rights over voting rights of the largest ultimate owner of rm i: Sample: The sample includes 3,752 rm-year observations, spanning between 1991 and 1995 and covering 977 rms from Hong Kong, Indonesia, Malaysia, Singapore, South Korea, Taiwan, and Thailand. To be included in the sample, a rm must have at least 1 year of stock returns and nancial data from the PACAP database between 1991 and 1995, its ultimate ownership data must be available from Claessens et al. (2000), and its largest ultimate owner must have at least 20% of voting rights. Observations with largest and smallest 1% values of CAR and NI are excluded from the sample. ***Indicates signicant at 1% (two-tailed). **Indicates signicant at 5% (two-tailed). *Indicates signicant at 10% (two-tailed). a OLS regressions are run separately for each of the sample year from 19911995. Fixed-effects of economies are included in the regressions but not reported. b White-adjusted t-statistics are in parentheses.

argument predicts that high ownership concentration is associated with low earnings informativeness. Our empirical results are generally consistent with the entrenchment and the information effect arguments. Our results are robust to controls for rm size,

422

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

Table 6 Economy-by-economy regressions with interactions for corporate ownership structurea Hong Kong Intercept NI NI * SIZE NI * Q NI * LEV NI * SEG NI * V NI * CV Adjusted R2 Number of observations 0.19*** (7.00)b 0.93 (1.15) 0.14*** (2.61) 0.19 (1.46) 0.44 (1.31) 0.03 (0.72) 1.57*** (2.60) 0.45* (1.65) 0.11 1045 Indonesia 0.13** (1.96) 0.65 (0.62) 0.13 (1.35) 0.41 (0.86) 0.67 (0.96) 0.14 (1.51) 1.70 (1.37) 0.16 (0.35) 0.07 259 Malaysia 0.10*** (3.49) 5.50*** (2.77) 0.38*** (2.61) 0.56*** (2.58) 0.08 (0.16) 0.02 (0.21) 2.75* (1.70) 1.48** (2.17) 0.07 684 Singapore 0.04 (1.39) 1.36 (0.74) 0.09 (0.68) 0.03 (0.13) 0.49 (0.73) 0.14 (1.50) 1.83 (1.02) 0.07 (0.11) 0.07 561 South Korea 0.00 (0.08) 7.12*** (2.80) 0.10 (0.59) 2.82*** (2.82) 3.13** (2.46) 0.41** (2.48) 3.40 (1.08) 0.38 (0.47) 0.17 346 Taiwan 0.17*** (5.13) 11.83** (2.35) 0.40 (1.18) 1.85*** (3.71) 5.01** (2.15) 0.42 (1.55) 12.77** (2.28) 4.73** (2.33) 0.26 311 Thailand 0.17*** (3.95) 7.33*** (3.10) 0.36** (2.05) 0.66** (2.21) 4.65*** (4.24) 0.00 (0.01) 1.20 (0.67) 2.99** (2.31) 0.24 546

Model specication: CARit a0 a1 NIit a2 NIit * SIZEit a3 NIit * Qit a4 NIit * LEVit a5 NIit * SEGi a6 NIit * Vi a7 NIit * CVi Fixed effects uit : Variable denitions: CARit =the cumulative net-of-market 12-month stock returns for rm i at year t: The annual returns are continuously compounded from monthly stock returns starting from 12 months before the latest date, as required by law or listing rules, that the rm discloses its annual report. NIit =the net earnings at year t divided by the market value of equity at the beginning of year t for rm i: SIZEit =the natural logarithm of the market value of equity in millions of U.S. dollar at the beginning of year t for rm i: Qit =the market value of equity divided by the book value of total assets at the beginning of year t for rm i: LEVit =the total liability divided by total assets at the beginning of year t for rm i: SEGi =the number of industry segment(s) of rm i: Vi =the voting rights level of the largest ultimate owner of rm i: CVi =the ratio of cash ow rights over voting rights of the largest ultimate owner of rm i: Sample: The sample includes 3,752 rm-year observations, spanning between 1991 and 1995 and covering 977 rms from Hong Kong, Indonesia, Malaysia, Singapore, South Korea, Taiwan, and Thailand. To be included in the sample, a rm must have at least 1 year of stock returns and nancial data from the PACAP database between 1991 and 1995, its ultimate ownership data must be available from Claessens et al. (2000), and its largest ultimate owner must have at least 20% of voting rights. Observations with largest and smallest 1% values of CAR and NI are excluded from the sample. ***Indicates signicant at 1% (two-tailed). **Indicates signicant at 5% (two-tailed). *Indicates signicant at 10% (two-tailed). a OLS regressions are run separately for each of the seven economies. Fixed-effects of calendar years are included in the regressions but not reported. b White-adjusted t-statistics are in parentheses.

market-to-book assets, leverage, number of industry segments operated by the rm, and to various specications of earnings and stock returns. This paper makes several contributions. First, it provides an analysis that contrasts with those in prior research focusing on U.S. and U.K. corporations. The

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

423

prior research documents that an increase in managerial ownership (Wareld et al., 1995) or institutional ownership (Rajgopal et al., 1999) would reduce the principal agent problem between managers and shareholders, which would in turn lower the incentives and opportunities for managers to control earnings while raising the price informativeness of earnings. However, this relation between share ownership and accounting information is not applicable to East Asian corporations due to differences in the degree of ownership concentration and in the associated type of agency problems. Also, compared with the results from this body of research, our results are more generalizable to other parts of the world because concentrated corporate ownership in East Asia as compared to diffuse corporate ownership in the U.S. and the U.K. is a more representative corporate share structure throughout the world. Second, several recent accounting studies (Ball et al., 2000, 1999; Ali and Hwang, 2000) have provided evidence that in addition to accounting standards, features of the institutional environment such as corporate governance as well as legal and nancial systems can also explain the differences in the properties of accounting information across countries. We extend their work by examining ownership structure as one of the channels through which a countrys institutional environment inuences each individual rms reporting quality. Third, this research may have implications for East Asian economic reformers and regulators who are striving to improve corporate governance and transparency in their countries.

References
Alchian, A., 1965. Some economics of property rights. Il Politico 30, 816829. (Originally published in 1961 by the Rand Corporation. Reprinted in Alchian, A., 1977. Economic Forces at Work. Liberty Press, Indianapolis, IN.) Alchian, A., 1977. Economic Forces at Work. Liberty Press, Indianapolis, IN. Ali, A., Hwang, L.S., 2000. Country-specic factors related to nancial reporting and the value relevance of accounting data. Journal of Accounting Research 38, 121. Asian Wall Street Journal, 1999. Business transparency in region has worsened, survey shows, November 24, 5. Atiase, R.K., 1985. Predisclosure information, rm capitalization and security price behavior around earnings announcements. Journal of Accounting Research 23, 2136. Ball, R., Robin, A., Wu, J., 1999. Properties of accounting earnings in four East Asian countries. Working paper, University of Rochester and Rochester Institute of Technology. Ball, R., Kothari, S.P., Robin, A., 2000. The effect of institutional factors on properties of accounting earnings: International evidence. Journal of Accounting and Economics 29, 152. Barclay, M., Holderness, C., 1989. Private benets from control of public corporations. Journal of Financial Economics 25, 371395. Bebchuk, L., Kraakman, R., Triantis, G., 2000. Stock pyramids, cross-ownership, and dual class equity: the creation and agency costs of separating control from cash ow rights. In: Morck, R.K. (Ed.), Concentrated Corporate Ownership. University of Chicago Press, Chicago, IL. Berle, A., Means, G., 1932. The Modern Corporation & Private Property. Hartcourt, Brace & World. (Reprinted in 1991. Transaction Publishers, NJ.) Bradley, M., Kim, E.H., 1985. The tender offer as a takeover device: its evolution, the free rider problem and the prisoners dilemma. Working paper, University of Michigan.

424

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

Burkart, M., Gromb, D., Panunzi, F., 1997. Large shareholders, monitoring, and the value of the rm. Quarterly Journal of Economics 112, 693728. Burkart, M., Gromb, D., Panunzi, F., 1998. Why high takeover premia protect minority shareholders. Journal of Political Economy 106, 172204. Cheung, S., 1970. The structure of a contract and the theory of a non-exclusive resource. Journal of Law and Economics 13, 4970. Cheung, S., 1983. The contractual nature of the rm. Journal of Law and Economics 26, 121. Christie, A., Joye, M., Watts, R., 2002. Decentralization of the rm: theory and evidence. Journal of Corporate Finance, forthcoming. Claessens, S., Djankov, S., Lang, L.H.P., 2000. The separation of ownership and control in East Asian corporations. Journal of Financial Economics 58, 81112. Claessens, S., Djankov, S., Fan, J.P.H., Lang, L.H.P., 2002. Disentangling the incentive and entrenchment effects of large shareholdings. Journal of Finance, forthcoming. Coase, R.C., 1960. The problem of social cost. Journal of Law and Economics 3, 144. Collins, D.W., Kothari, S.P., 1989. An analysis of intertemporal and cross-sectional determinants of earnings response coefcient. Journal of Accounting and Economics 11, 143181. Demsetz, H., 1964. The exchange and enforcement of property rights. Journal of Law and Economics 3, 144. Demsetz, H., Lehn, K., 1985. The structure of ownership: causes and consequences. Journal of Political Economy 93, 11551177. Dhaliwal, D., Lee, K., Fargher, N., 1991. The association between unexpected earnings and abnormal security returns in the presence of nancial leverage. Contemporary Accounting Research 8, 2041. Eggertsson, T., 1990. Economic Behavior and Institutions. Cambridge University Press, Cambridge. Fisman, R., 2001. Estimating the value of political connections. American Economic Review 91, 1095 1102. Freeman, R., 1987. The association between accounting earnings and security returns for large and small rms. Journal of Accounting and Economics 9, 195228. Gomes, A., 2000. Going public without governance: managerial reputation effects. Journal of Finance 55, 615646. Holderness, C., Sheehan, D., 1988. The role of majority shareholders in publicly held companies. Journal of Financial Economics 20, 317346. Jacobson, R., Aaker, D., 1993. Myopic management behavior with efcient, but imperfect, nancial markets. Journal of Accounting and Economics 16, 383405. Jensen, M.C., Meckling, W.H., 1976. Theory of the rm: managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305360. Jensen, M.C., Meckling, W.H., 1992. Specic and general knowledge, and organizational structure. In: Werin, L., Wijkander, H. (Eds.), Contract Economics. Blackwell, Oxford, pp. 251274. Johnson, S., Boone, P., Breach, A., Friedman, E., 2000a. Corporate governance in the Asian nancial crisis. Journal of Financial Economics 58, 141186. Johnson, S., La Porta, R., Lopez de Silanes, F., Shleifer, A., 2000b. Tunnelling. American Economics Review 90, 2227. La Porta, R., Lopez-De-Silanes, F., Shleifer, A., 1999. Corporate ownership around the world. Journal of Finance 54, 471518. La Porta, R., Lopez-De-Silanes, F., Shleifer, A., 2002. Investor protection and corporate valuation. Journal of Finance, forthcoming. Morck, R., 1996. On the economics of concentrated ownership. Canadian Business Law Journal 26, 6375. Morck, R., Shleifer, A., Vishny, R.W., 1988. Management ownership and market valuation: an empirical analysis. Journal of Financial Economics 20, 293315. Nenova, T., 1999. The value of a corporate vote and private benets: cross-country analysis. Working paper, Harvard University. Rajgopal, S., Venkatachalam, M., Jiambalvo, J., 1999. Is institutional ownership associated with earnings management and the extent to which stock prices reect future earnings? Working paper, University and Washington and Stanford University.

J.P.H. Fan, T.J. Wong / Journal of Accounting and Economics 33 (2002) 401425

425

Roe, M.J., 1994. Strong Managers, Weak OwnerThe Political Roots of American Corporate Finance. Princeton University Press, Princeton, NJ. Scott, K., 1999. Corporate governance and East Asia. Working paper No. 176, Stanford Law School. Sender, H., 1999. Smoke and mirrors. Far Eastern Economic Review 30, 3438. Shleifer, A., Vishny, R., 1986. Large shareholders and corporate control. Journal of Political Economy 94, 461488. Shleifer, A., Vishny, R., 1997. A survey of corporate governance. Journal of Finance 52, 737783. Smith, C., Watts, R., 1992. The investment opportunity set and corporate nancing, dividend and compensation policies. Journal of Financial Economics 32, 263292. Stulz, R.M., 1988. Managerial control of voting rights, nancing policies and the market for corporate control. Journal of Financial Economics 20, 2554. Teoh, S.H., Wong, T.J., 1993. Perceived auditor quality and the earnings response coefcients. The Accounting Review 68, 346367. Wareld, T., Wild, J.J., Wild, K., 1995. Managerial ownership, accounting choices, and informativeness of earnings. Journal of Accounting and Economics 20, 6191. Wolfenzon, D., 1999. A theory of pyramidal structures. Working paper, Harvard University. World Bank, 1998. East Asia: The Road to Recovery. World Bank, Washington DC.