OWNERSHIP STRUCTURE, EXPROPRIATION, AND PERFORMANCE OF GROUP- AFFILIATED COMPANIES IN KOREA

Transcription

1 OWNERSHIP STRUCTURE, EXPROPRIATION, AND PERFORMANCE OF GROUP- AFFILIATED COMPANIES IN KOREA SEA JIN CHANG School of Business Administration Korea University Sungbukku, Anamdong Seoul, Korea Tel: Fax: June 2002 I thank Jaebum Hong for providing the KIS database and Panseop Lee for able research assistance. I also thank Hasung Jang, and Myeonghyun Cho, Woochan Kim, John Lafkas and three anonymous reviewers for helpful comments and suggestions, and Jongwha Lee for making his algorithm to calculate Tobin s Q available. Financial Support from the Korea Research Foundation (KRF 2001-C00350) is gratefully acknowledged. Additional support from Korea University Business School through SK Distinguished Research Award is also gratefully

2 TABLE 2: Models of Profitability and Ownership Structure With industry effects With firm effects 2 SLS with firm effects (just- 2 SLS with firm effects (over- (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Profitabilit y (ROIC) Intercept 5.48 (1.32)*** Profitability (ROIC) Inside Ownership Family Portion 0.03 (0.01)*** 1.01 (0.23)*** Firm Sales 8.40*10-5 (7.83*10-5 ) Firm Equity 2.41*10-4 (8.37*10-4 ) Group Equity -1.62*10-5 (5.37*10-5 ) Years of Listing 0.01 (0.02) Vo latility 0.06 (0.10) Sales Growth 0.00 Market Share 0.01 (0.01) R&D Intensity Advertising Intensity 0.24 (0.10)* 0.22 (0.04)*** Leverage (0.05)*** Sales to 0.44 Affiliates (0.52) Purchases From affiliates Equity investment to affiliates Supply of Debt Guarantees Reception of Debt guarantees (0.76)*** (0.43) (0.02) 0.00 (0.01) Inside Ownership (2.75)*** 0.22 (0.08)** -4.40*10-4 (2.38*10-4 )+ Family Portion 0.99 (0.04)*** 0.01 *** 6.30*10-6 (6.74*10-6 ) -5.64* *10-5 (1.08*10-3 )*** (4.84*10-5 ) 2.18*10-4 (1.98*10-4 ) (0.04)*** 0.37 (0.09)*** *** 0.03 (0.02) (0.35)** 0.30 (0.10)** 0.02 (0.06) 1.00 (1.46) (2.23)* 0.31 (0.91) 0.01 (0.04) (0.02) -6.92*10-6 (5.62*10-6 )** 0.01*10-1 * -5.76*10-6 (8.24*10-5 ) (0.01)** (0.04)*** (0.06) 0.03 (0.04) -0.04*10-1 ** Profitabilit y (ROIC) 7.64 (1.35)*** 0.03 (0.02) (0.55) 1.99*10-4 ((1.10*10-4 )+ 1.07*10-3 ((1.15*10-3 ) -4.57*10-5 ((9.18*10-5 ) (0.07)* 0.02 (0.08) 0.02 (0.01)* 0.05 (0.02)* (0.11) (0.11) (0.05)** 1.65 (0.86) (1.17)*** (0.78) 0.01 (0.02) Inside Ownership (2.14)*** 0.10 (0.05)* -1.17*10-4 (3.61*10-4 ) Family Portion 0.59 (0.04)*** *10-6 ((9.39*10-6 ) 3.92* *10-5 (1.76*10-3 )* ((3.54*10-5 ) -8.89*10-4 (3.50*10-4 )* (0.17)*** 0.14 (0.08) (0.01) (0.05) (0.28) 0.10 (0.20) 0.06 (0.06) (1.70) (2.07) (0.66) 0.01 (0.04) (0.02)* -1.24*10-5 (4.43*10-6 )** 0.03 *** * *10-1 (0.01*10-1 )** * (0.02)*** (0.05) (0.01) Profitability (ROIC) (5.53)** (0.24) (22.10) 0.30 (1.08) 0.03 (0.02)+ Inside Ownership (6.22)*** 0.79 (0.37)* -7.38*10-4 (2.70*10-4 )** (0.28) 0.12 (0.09) Family Portion 1.03 (0.08)*** -7.21*10-6 (4.07*10-6 ) * -0.02*10-1 (0.01*10-1 )+ Profitability (ROIC) (21.31) 0.15 (0.19) (18.75) Inside Ownership (3.37)*** 0.64 (0.34)+ ) 2.50*10-3 ((1.80*10-3 ) (0.90) 0.02 (0.01) 0.11 (0.06) *10-4 (3.08*10-4 )** (0.18)*** R-squared [Note] Numbers in parentheses are the White s heteroscedasticity-consistent robust standard errors. *** p<0.001, ** p<0.01 * p<0.05, + p<0.10. N=3, (0.09) Family Portion 0.94 (0.08)*** *10-5 ((2.84*10-5 ) -6.98*10-6 (4.49*10-6 ) *10-1 (0.01*10-1 )+

3 TABLE 3: Models of Firm Value (Tobin s Q) and Ownership Structure With industry effects With firm effects 2 SLS with firm effects (just-identified) 2 SLS with firm effects (over-identified) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Firm Value Intercept 3.28 (0.23)*** Firm Value (Tobin s Q) Inside Ownership Family Portion 0.02*10-1 (0.01*10-1 )** (0.03)** Firm Sales 1.06*10-5 ((6.23*10-6 )+ Firm Equity -7.19*10-6 ((3.45*10-5 ) Group Equity -5.89*10-6 (7.40*10-6 ) Years of Listing *** Vo latility Sales 0.00 Growth Market Share R&D Intensity Advertising Intensity 0.01 (0.01) Leverages 0.06*10-1 Sales to Affiliates Purchases From affiliates Equity investment to affiliates Supply of Debt Guarantees Reception of Debt guarantees (0.02*10-1 )*** (0.08) 0.05 (0.07) (0.02)* + Inside Ownership (3.04)*** 1.32 (0.47)** -3.75*10-4 ((2.34*10-4 ) Family Portion 1.21 (0.06)*** (0.01)*** 7.54*10-6 ((6.59*10-6 ) -5.88* *10-5 ((1.08*10-3 )*** ((5.07*10-5 ) 1.40*10-4 (1.90*10-4 ) (0.04)*** 0.38 (0.10)*** ** 0.04 (0.02) (0.34)** 0.36 (0.10)*** (0.07) 0.79 (1.45) (2.21)* 0.26 (0.91) 0.02 (0.04) (0.02) -7.00*10-5 (5.63*10-6 )*** (0.01)** *10-1 (0.02*10-1 ) (0.04)*** (0.06) 0.02 (0.04) Firm Value 4.12 (0.30)*** 0.02*10-1 (0.01*10-1 ) (0.06) 6.17*10-6 ((2.40*10-5 ) -1.03*10-4 ((8.67*10-5 ) -4.39*10-5 (3.47*10-5 ) (0.01)*** 0.03*10-1 (0.01*10-1 )** (0.01) (0.01) 0.01 ** 0.07 (0.16) (0.11) (0.03) -0.04*10-1 (0.01*10-1 )** + Inside Ownership (4.08)*** 0.87 (0.60) -1.18*10-4 ((3.56*10-4 ) Family Portion 0.51 (0.04)*** (0.01) 9.78*10-6 ((9.36*10-6 ) 3.99* *10-5 ((1.75*10-3 )* ((3.54*10-5 ) -8.22*10-4 (3.51*10-4 )* (0.23) (0.08)* 0.01 (0.01) (0.05) (0.28) 0.10 (0.20) 0.04 (0.06) (1.71) (2.07) (0.67) 0.01 (0.04) (0.02)* -1.27*10-5 (4.50*10-5 )** 0.04 *** 0.02*10-1 (0.01*10-1 )** * (0.02)*** (0.05) (0.01) 0.00 Firm Value 3.27 (1.83) (0.02) 4.99 (3.44) Inside Ownership (8.23)*** 5.44 (2.73)* Family Portion 0.66 (0.10)*** (0.03) Firm Value 0.81 (2.01) 0.01 (0.03) 5.19 (3.37) Inside Ownership 4.42 (16.75) 5.19 (2.71)+ ) ) ) ) 3.78* (0.15)** 0.03*10-1 (0.01*10-1 )* -3.56*10-4 (2.90*10-4 ) (0.14)*** 0.20 (0.08)** -8.69*10-6 (4.61*10-6 ) *** -0.02*10-1 (0.01*10-1 ) (0.15) 0.04*10-1 (0.02*10-1 )* (0.01) ((1.68*10-3 )* Family Portion 0.49 (0.21)* (0.03) -9.06*10-6 ((2.64*10-5 ) -6.59* *10-6 (3.19*10-4 )*(4.90*10-6 ) 0.93 (0.80) 0.18 (0.08)* 0.04 (0.01)*** -0.02*10-1 (0.01*10-1 )+

4 R-squared [Note] Numbers in parentheses are the White s heteroscedasticity-consistent robust standard errors. *** p<0.001, ** p<0.01 * p<0.05, + p<0.10. N=3,086 FIGURE 1 Ownership structure of SK Group SK C&C Chairman & his family SK Distribution 49.4 Sheraton Walker-Hill SK Securities SK Capital SK Global 4.1 SK Telecom Daehan Gas SK Shipping SK Gas Kookil Gas SK Chemical SK Corporation SK Energy Sales SKC SK Teletech Koomi Gas Pohang Gas SK Oxy Chemical SK Construction [Note] a. We show equity ownership of major affiliates that is more than 1% as of b. Dark circle for companies with more than 3 trillion won in assets. Dark arrow represents cross shareholding through circular equity investment. c. Companies in bold face are listed in Korean Stock Exchange. [ Source] Korea Investor s Service

5 acknowledged.

6 OWNERSHIP STRUCTURE, EXPROPRIATION, AND PERFORMANCE OF GROUP- AFFILIATED COMPANIES IN KOREA This study uses a sample of group-affiliated public firms in Korea to examine the simultaneous nature of the causal relationships between ownership structure and performance. The results show that performance determines ownership structure but not vice versa, and provides strong evidence that controlling shareholders use insider information to take higher direct and indirect equity stakes in more profitable or more promising firms and transfer profits to other affiliates through intra-group trade. These findings highlight the importance of studying further the agency problems that controlling shareholders present for minority shareholders, especially in business groups.

7 Business groups, which are gatherings of formally independent firms under common administrative and financial control, are prevalent worldwide (Kester, 1992; Granovetter, 1995; Ghemawat & Khanna 1998). Some are centered on banks and large industrial corporations such as Daimler-Benz Group in Germany and Daiichi-Kangyo Bank Group in Japan. Others are centered on wealthy families, including the Wallenberg Group in Sweden, the Krupp Group in Germany, the Tata Group in India, the Angelini Group in Chile, and the SK Group in Korea. La Porta, Lopez-de-Silanes & Shleifer (1999) found that large corporations in most wealthy countries have controlling shareholders who enjoy control in excess of their equity holdings through a hierarchical chain of ownership and participation in management. Since Berle and Means (1932), both finance and strategy researchers have focused on the agency costs of professional managers. They found that concentration of ownership, which enables owners to reduce managers discretion, and increased ownership by managers, which aligns managers interests with those of shareholders, improve firm performance (Jensen & Meckling, 1976; Shleifer & Vishny, 1986; Rajagopalan, 1997; Gedajlovic & Shapiro, 1998; Thomsen & Pedersen, 2000). These findings are currently being challenged on three theoretical and practical fronts. First, recent studies emphasize that the greatest source of agency problems stems from controlling shareholders, who expropriate value from minority shareholders (La Porta, et al., 1999; Dharwadkar, George & Brandes, 2000). Second, researchers often fail to observe that major shareholders often control several firms and form business groups through pyramids and crossshareholding. Few studies have examined what determines affiliate ownership and its impact on individual firm performance in business groups. A third line of research suggests, contrary to

8 conventional wisdom, firm performance may influence ownership structure but not vice versa (Demsetz, 1983; Demsetz & Lehn, 1985; Kole, 1996; Cho, 1998; Loderer & Martin, 1997). Korea offers an interesting setting for examining linkages between ownership and firm performance in business groups. Descendants of the individuals who founded Korean business groups, termed chaebols, control affiliated firms through both direct equity stakes and pyramids even though these affiliates are legally independent companies with their own shareholders and boards of directors. This ownership structure enabled chaebols to develop a governance structure like that of the multidivisional organization, under which individual affiliates function as operating divisions that are controlled by group-level staff (Chang & Choi, 1988; Chang & Hong, 2000). Using a sample of publicly traded chaebol affiliates, this paper explores three broadly defined research questions. First, it examines what factors determine the ownership structure of group affiliates. Second, it assesses the causal relationship between ownership and performance by considering whether family owners use insider information to increase the stakes they and other affiliates own in successful companies and decrease their stakes in unsuccessful companies. Third, it considers how intra-group business transactions may affect the relationship between ownership and firm performance. By furthering our understanding of the ownership-performance relationship, this study has both theoretical and practical implications. OWNERSHIP STRUCTURE AND PERFORMANCE OF GROUP AFFILIATED FIRMS The Determinants of Firm Performance Most agency theorists (e.g., Jensen & Meckling, 1976) posit that professional managers with little equity in the firms they run pursue their own interests at the expense of shareholders. Related work extends this perspective by examining whether diversification reflects managers

9 self-interested pursuit (Bethel & Liebeskind, 1993; Hoskisson, Johnson, & Moesel, 1994). Still other research assesses whether governance mechanisms such as boards of directors can monitor managers more effectively (Dalton, Daly, Johnson, & Ellstrand, 1999) and if control mechanisms such as stock options induce these managers to align their interests with those of shareholders (Rajagopalan, 1997). By this logic, large shareholders have both a strong incentive to monitor managers and the power to discipline them. Several studies have found that firms with concentrated ownership have higher performance (Shleifer & Vishny, 1986; Short, 1994; Gedajlovic & Shapiro, 1998; Thomsen & Pederson, 2000; Kang and Shivadasani, 1995). La Porta et al. (1999) challenge the standard agency theory argument with their finding that, even in wealthy countries, controlling shareholders -- usually families or the state are present, and that many family members are both owners and managers. Accordingly, the positive relationship between ownership concentration and performance may generalize only to countries that have efficient capital markets and well-developed governance mechanisms such as the US. For most countries, the separation of ownership and management has yet to take place. Other studies focus on not only the ownership concentration per se but also the identities of concentrated owners (McConnell & Servaes, 1990; Thomsen & Pedersen, 2000). They argue that many owners (e.g., institutional investors, banks, and other companies) are intermediate agents for final owners and have quite different incentives and risk preferences. According to them, concentration of ownership by institutional investors, which are subject to scrutiny over their own performance, leads to better firm performance. There is disagreement on the role of family owners. Thomsen & Petersen (2000) argue that family owners are more risk averse and thereby destroy firm value, while Fama and Jensen (1983) and Morck, Shleifer, &

10 Vishny (1988) argue that family owners reduce agency costs and thereby increase firm value. Also, cross-shareholding may also not be an effective monitoring tool. Gedajlovic and Shapiro (1998) found that the impact of concentrated ownership on firm performance is weaker in France, where high ownership concentrations are more common and most of the dominant owners are banks, non-financial corporations, or the government. Although these studies included affiliate ownership as an ownership category explaining firm performance, they treated group-affiliates as if these firms were independent and did not address the possibility that family owners might use affiliate ownership to maintain control. Figure 1 shows the equity stakes held by the Chey family and major affiliates in SK Group, the 5 th largest group in Korea as of The ownership structure of SK Group shows signs of both pyramid and cross-shareholding structures. La Porta, et al. (1999) define pyramid as a hierarchical chain by which a family controls a firm, and cross-shareholding as a structure through which a controlled firm owns any shares in its controlling shareholder or in the firms along that chain of control. The Chey family owns equity in core affiliates (i.e., SK Global, SK Chemical, and SKC, which are all public companies). When SK Group acquired an oil refining business from the government in 1980, SK Global and SK Chemical took 13.8% and 2.7% equity stakes, respectively, in this venture, which was later renamed SK Corporation. SK Corporation later owned large equity holdings in Daehan City Gas, SK Gas, SK Energy Sales, and SK Oxy- Chemical, and also helped purchase SK Telecom from the government in SK Group s affiliates also use cross-shareholding through circular investment. In Korea, cross-shareholding between two firms is forbidden, but it is legal when it involves more than two firms. For instance, SK Global owns 13.8% of SK Co., while SK Co. holds 18.5% in SK Telecom. SK Telecom in turn

11 holds 4.1% of SK Global. In sum, major affiliates control smaller affiliates by owning them singly or jointly, and the chairman controls all these firms with only 14.1% of the SK Group s total equity. Affiliates owned 30.1% of each others equity. Institutional investors, such as securities firms, investment trusts, and insurance firms, and minority shareholders own the remainder of these affiliates. These institutional investors are often affiliated with the same business groups Insert Figure 1 about here For business groups such as chaebols, lower levels of family and affiliate ownership may not mean there is less monitoring since affiliates function as do business divisions in a diversified corporation that are controlled by chaebol chairmen (Chang & Choi, 1988). Any observed relationship between concentrated ownership and performance may be attributable to tunneling, which Bertrand, Mehta, and Mullainathan (2000) define as transferring resources from firms in which a controlling family has few cash flow rights to ones in which it has substantial cash flow rights. For example, SK Telecom, a public firm, transferred profits to its privately owned affiliates via intra-group transactions (e.g., providing loans or buying and selling from each other above or below market rates). 1 As La Porta et al noted, large shareholders who are also managers are able to use pyramid and cross-shareholding mechanisms to tunnel. Accordingly, they can expropriate value from minority shareholders. (Shleifer and Vishny, 1997; Wolfenzen, 1999) Such instances occur even in well-developed countries: Bergstrom and Rydqvist (1990), Barclay and Holderness (1989), Zingales (1994), and Weinstein and Yafeh (1998) found evidence of shareholder expropriation in Sweden, US, Italy, and Japan, respectively. Such expropriation may be easier in countries where there are few rules and procedures to

12 protect minority shareholders (La Porta et al., 1998, 2000; Coffee, 2001). For instance, Korean securities law restricts the voting rights of institutional investors. Also, chaebol affiliates boards of directors are generally filled with insiders and friends of chaebol families. Class action lawsuits and minority shareholders rights to open up accounting books are limited. Claessens, Djankov, Fan, and Lang (1999) used data from firms in nine Asian countries to show that expropriation of minority shareholders by large shareholders is the rule rather than the exception. They found that concentration of control via cross shareholding, the use of pyramids, and deviations from oneshare-one-vote rule are negatively associated with market value. 2 Based on this discussion, we formulate hypotheses for group affiliates. Hypothesis 1(a) reflects agency theorists argument that more concentrated ownership leads to better firm performance. We need to control, however, for intra-group business transactions, which may be used to siphon profits from one affiliate to another. Hypothesis 1(a): The higher the level of inside ownership is, the higher the firm performance is. Hypothesis 1(b) is a stronger form of hypothesis 1(a), as it focuses on the family portion of inside ownership. It posits that a higher portion of inside ownership by a chaebol family leads to higher firm performance, since family owners have a higher level of cash flow rights in a company that they directly own shares than they do in a company in which they own an indirect share through pyramids or cross-shareholding, and thereby have a stronger incentive to monitor managers. Here, the alternative hypotheses would be that, after controlling for intra-group trade, ownership would not affect performance since a family would have an incentive to monitor an affiliate s performance, regardless of its level of ownership in the affiliate.

13 Hypothesis 1(b): The higher the family portion of inside ownership is, the higher the firm performance is. The Determinants of Ownership Structure Although researchers conventionally treated ownership structure as an exogenous variable in explaining firm performance, Demsetz (1983) pointed out that the ownership structure might be an outcome influenced by the equilibrium of various cost advantages and disadvantages. Demsetz and Lehn (1985) found that ownership concentration is determined by a firm s risk level, regulations, and industry-specific factors, and that the hypothesized relationship from ownership concentration to firm performance became insignificant after controlling for these factors. Kole (1996) provided related evidence for this conjecture by showing that managers prefer equity compensation only when they expect their firms to perform well, suggesting that managerial ownership might be endogenous to compensation contracting practices. Similarly, Rajagopalan (1997) showed the relationship between executive compensation and performance is contingent upon the firm s strategic context. More recently, Cho (1998) used the simultaneous equations estimation technique to show, for his sample, that corporate value affected ownership structure, while the reverse relationship did not hold. Loderer and Martin (1997) found that acquisition performance and firm value affected the size of managers stockholdings but not vice versa in their sample of acquisitions. When controlling shareholders directly or indirectly manage a firm, the possibility of reverse causality can be much greater because they can use insider information for their personal gain, just as managers prefer stock options when they think their firms will perform well. From the perspective of family owners, the ownership decision could boil down to two questions:

14 First, given a firm s performance, how much inside ownership, including both direct (family) and indirect (affiliate) ownership, do they desire above the minimum they need to ensure their control? Second, which mix of direct versus indirect ownership will be best for them, given their wealth constraints and individual firm s performance and risk profile? Inside ownership reflects both direct and indirect ownership. A family may use insider information to increase its shares in an affiliate that is likely to develop a breakthrough technology or to win a large contract and decrease its shares in less successful affiliates. Intra-group business transactions can also bias the ownership decision. Chaebol families may wish to decrease their ownership of affiliates that have potential liabilities. In Korea, banks ask financially strong firms to guarantee their weaker affiliates debts as a condition for providing loans to these weaker firms. Firms that honor this request are thus fully exposed to the liabilities of these affiliates. Chaebol families and affiliates may decrease their shares in those affiliates that provide or receive such debt guarantees in order to reduce their own downside risk. The following hypotheses thus examine whether the ownership structures of groupaffiliated companies are endogenously determined by controlling for shareholders use of insider information. Hypothesis 2 (a) posits that inside ownership, which includes both family and affiliate ownership, is higher for more profitable or more promising companies, indicating that inside owners use the information they have to expropriate profits. Thus, it argues for reversed causality between the variables used in hypothesis 1(a). Hypothesis 2 (a): The higher the firm performance is, the higher the level of inside ownership is. Hypothesis 2 (b) is a stronger form of hypothesis 2(a), as it argues that the family portion

15 of inside ownership is disproportionately higher for more profitable or more promising companies, since family owners have greater cash flow rights in a company that they directly own than they do in a company that they own indirectly. The alternative hypothesis is that the composition of inside ownership is unaffected by performance. Sample Hypothesis 2 (b): The higher the firm performance is, the higher the family portion of inside ownership is. METHODS This study used the database developed by the Korea Investors Services (KIS). The KIS has company profiles and financial information data, equivalent to the Compustat database, for all Korean public firms from the early 1980s onward. Its ownership database was developed mainly for chaebol affiliates and is available from 1986 onward. For this study, we restricted our sample of firms to publicly traded chaebol affiliates. We excluded financial services firms from our sample since their accounting scheme is not compatible with that of firms in other industries. We further restricted our sample to firms associated with business groups since we are interested mainly in the mix of family and affiliate ownership and its relationship with firm performance. Since many firms went bankrupt or underwent restructuring as a result of the Asian Crisis in 1997, we restricted our analysis to The Korean Fair Trade Commission (KFTC) legally defines a business group as a group of companies, more than 30% of whose shares are owned by some individuals or by companies controlled by those individuals, or those that are practically controlled by them despite lower ownership control. Because ownership is widely dispersed, chaebol families can control some

16 public firms even when the combined family and affiliate ownership is below 30%. In such cases, the KFTC identifies a firm as a chaebol affiliate even though the criterion of 30% ownership is not met. Using this definition, the KIS database identified a total of 461 business groups as of We then identified the public companies that are members of those business groups. The number of public firms and group affiliates increased from 1986 to In 1986, 242 out of 277 non-financial public firms were group affiliates. By 1996, 557 out of 641 non-financial public firms were group affiliates. Our initial database has 4,747 observations of chaebol affiliates during , and ownership information was available for 3,215 of this sample. Several companies either went bankrupt or showed negative equity due to accumulated losses during this period. Since we needed to calculate sales growth and volatility, we lost the first observations of firms that were newly listed between Our final sample consists of 3,086 observations of 419 chaebol affiliates during Measurements Firm Performance. This study uses both accounting- and stock market-based measures for firm performance since it is difficult to identify a single indicator for firm performance. PROFITABILITY is measured by the return on invested capital (ROIC), defined as the sum of net income before tax plus interest payments, deflated by total assets, measured at time t. This measure of performance should measure operating efficiency without being biased by the relatively high debt/equity ratios common in Korean firms. Although chaebols typically pursued growth by venturing into new businesses, their affiliates were expected to maximize profits, which the chaebols headquarters then invested in new ventures or investment opportunities in other affiliates. This study also calculates the TOBIN s Q, measured at time t, as a proxy for firm

17 value using the algorithm that was originally proposed by Lindenberg and Ross (1981) and adapted by Kim, Kim, Park, and Chang (1996) for Korean firms. Ownership Structure. The KIS developed its ownership database by establishing information on the family membership of controlling shareholders, who are often founders and their descendants, and calculated the family ownership. Identifying affiliate ownership is straightforward and follows the definition of business groups by the KFTC. INSIDE OWNERSHIP is defined as the sum of the percentage of equity shares owned by family members identified as the controlling shareholders of a business group and by other non-financial affiliates in the same group, measured at time t. FAMILY PORTION denotes what portion of INSIDE OWNERSHIP is held directly by the family. Higher values of FAMILY PORTION reflect the greater cash flow rights given the same level of inside ownership. The simple averages of ownership for our sample firms in 1996 are 18% by families; 10% by affiliates; 2% by institutional investors; 0.2% by managers; 0.03% by employees; 1.3% by foreign investors; 0.2% by the Korean government; and the rest by minority shareholders. Chaebols tend to own financial service firms such as insurance firms, securities firms, and merchant banks, and these affiliates own significant shares of other affiliates in the same group. Since it is hard to distinguish whether such equity holdings are for maintaining control of affiliates or for pure financial investment, the KIS database classifies them as the shares of institutional investors. Firm characteristics. It is necessary to control for both firm and business group size, as each may affect the ownership decision and firm performance. This study includes SALES of a firm in billion won as a control for firm size without any a priori expectation for its coefficient. In the equations for ownership structure, this study incorporates both FIRM EQUITY and GROUP

18 EQUITY in billion won. We expect that the larger the firm s equity is, the smaller the portions of inside ownership will be, given wealth constraints. The same logic holds for the total level of equity at the group level since it would be hard for a family or an affiliate to maintain a higher level of ownership for every affiliate. YEARS OF LISTING is defined as the number of calendar years since a firm was listed. We expect that both family and affiliate ownership will be dispersed over time. This study also incorporates a measure of uncertainty faced by the firm. Demsetz and Lehn (1985) argued that the more uncertain the firm s environment is, the greater is the firm s need to maintain tight ownership control. This study measures VOLATILITY by the standard deviation of profitability during the past five years (Cho, 1998). We include market share and sales growth as determinants for firm performance. SALES GROWTH was measured by the annual growth of sales. MARKET SHARE was measured by dividing a firm s sales by the industry shipments as measured at the 2 digit Korean Standard Industry Classification (KSIC) level. Previous research has used R&D and ADVERTISING intensities as proxies for intangible knowledge-based resources (Chatterjee & Wernerfelt, 1991). Both variables were calculated as the percentage of those expenditures relative to total sales. LEVERAGE was defined as a firm s debt to equity ratio. A high debt-equity ratio will increase the likelihood of bankruptcy and financial distress and thereby limit the firm s ability to finance its investment by borrowing (Myers & Majluf, 1984; Froot, Scharfstein, & Stein, 1994). All firm characteristics variables were measured at time t. Intra-group business transactions. Various forms of intra-group business transactions are reported in Korea as footnotes in the financial statements of group affiliated companies (Chang & Hong, 2000). We measure SALES TO AFFILIATES and PURCHASES FROM AFFILIATES as the

19 ratio of a firm s sales to and purchases from affiliates divided by the firm s total sales. We also measured the extent of SUPPLY OF DEBT GUARANTEES as the level of debt guarantees supplied to group members divided by the equity base of the providing firm and RECEPTION OF DEBT GUARANTEES as the level of debt guarantees received from group members divided by the equity base of the recipient. EQUITY INVESTMENT TO AFFILIATES is defined as an affiliate s equity participation into other affiliates divided by the equity base of that affiliate. This variable reflects how much this affiliate invests in other affiliates, which may lead to higher levels of inside ownership. All intra-group transactions variables were measured at time t. Methodology This study uses the panel data analysis technique and the two stage least squared (2SLS) regression methods to estimate the profitability and ownership structure with panel data for public group affiliates for We initially formulated individual equations for firm i at time t as follows: Performance it = f(inside ownership it, family portion it, firm characteristics it, intra-group trade it ) (1) Inside ownership it = f (performance it, firm characteristics it, intra-group trade it ) (2) Family portion it = f (performance it, firm characteristics it, intra-group trade it ) (3) Initially, this study estimates individual equations with the OLS method while incorporating only the year and industry effects variables, based on the 2 digit KSIC level. Himmelberg, Hubbard, and Palia (1999) observed, however, that there could be spurious causal relationships between ownership and firm performance due to unobserved firm heterogeneity. They demonstrated that the relationship between ownership concentration and performance vanished after controlling for unobserved firm heterogeneity. This study adopts the panel data technique of Himmelberg et al. by including a set of firm and year dummy variables, thus assuming fixed effects (Hsiao, 1986). 3 The Durbin-Watson

20 statistics showed no evidence for autocorrelation. We used White s heteroscedasticity consistent standard errors in order to improve the efficiency of estimators and reduce other possible heteroscedasticity problems. Nonetheless, the panel analysis technique does not resolve the endogeneity problems between dependent and independent variables. We thus use the simultaneous equation technique to explore this endogeneity as Cho (1998) and Loderer and Martin (1997) did in studying ownership concentration and performance for US firms. In equations (1)-(3), three variables, firm performance, inside ownership, and family portion of inside ownership, are identified as endogenous variables. The two stage least squared method first estimates individual endogenous variables with instruments that are independent of these variables. It then replaces the endogenous variables with the estimates from the first stage regressions. It is not easy to find suitable instruments that are both conceptually and empirically exogenous to the system. We identify firm characteristics such as firm sales, firm equity, group equity, years of listing, volatility, sales growth and market share as potential instruments because there is no a priori reason why they are endogenous to ownership or firm performance. Variables for firm investment (R&D, advertising, and leverage) and intra-group transaction variables may be endogenous to firm performance and ownership (Cho, 1998; Himmelberg et al. 1997; Chang & Hong, 2000) and are not used as instruments. RESULT S Table 1 shows the descriptive statistics, and Tables 2 and 3 present the regression results using profitability and firm value, respectively, as measures for firm performance. Models (1)-(3) of Tables 2 and 3 show the results without accounting for firm heterogeneity or endogeneity. We present these results in order to compare them with models (4)-(6), which control only for firm

21 heterogeneity, and two stage models (7)-(12), which handle both firm heterogeneity and endogeneity Insert Tables 1, 2, & 3 about here In model (1) of Table 2, both inside ownership and the family portion of inside ownership are associated with higher profitability, which supports the agency theory s hypothesis that concentrated ownership ensures more effective monitoring of managers. In models (2) and (3), profitability is positively associated with inside ownership and family portions of inside ownership, which suggests that inside ownership and family portion are higher in more profitable firms. We explored the possibility of a non-linear relationship between ownership concentration and performance by experimenting with a squared term of inside ownership and the piecewise regression technique (Morck et al, 1988; McConnell & Servaes, 1990). We found no evidence for such a relationship. Models (4)-(6), which control for unobserved firm heterogeneity, provide an interesting contrast with models (1)-(3). In models (4) and (5), the effect of inside ownership to profitability and the effect of profitability to inside ownership remain significant, but the effect of profitability to family portion in model (6) vanishes after controlling for firm heterogeneity. In addition, some firm investment and intragroup transaction variables become insignificant after controlling for firm heterogeneity. For instance, R&D and advertising variables become insignificant in models (4) and (5). Other variables, such as intra-group sales, turn significant in model (4). This result suggests that many observed significant relationships in models (1)-(3) are spurious, and merely reflect unobserved firm heterogeneity. Even after controlling for firm heterogeneity, intra-group sales are found to

22 lead to higher profitability, but intra-group purchases contribute to lower profitability in model (4), suggesting that intra-group trade might be used to transfer profits among affiliates. Internal ownership is negatively associated with the reception of debt guarantees, suggesting that both family and affiliates reduce their ownership of affiliates with poor financial credit that depend upon debt guarantees from other affiliates. The family portion of inside ownership is lower in affiliates that depend more on intra-group sales, as shown in model (6). 4 Models (1)-(6) in Table 3, which use Tobin s Q, show patterns similar to those exhibited in Table 2. The effect of inside ownership to firm value as measured by Tobin s Q remain significant from model (1) to model (4). All the effect of the family portion of inside ownership to Tobin s Q as in model (1), the effect of Tobin s Q to inside ownership as in model (2), and the effect of Tobin s Q to the family portion of inside ownership as in model (3) become insignificant in models (4), (5), and (6), respectively. Also, most of the firm investment and intra-group transactions variables become insignificant in equations using Tobin s Q after accounting for firm heterogeneity. The negative impacts of both the receipt of debt guarantees to inside ownership and intra-group sales to Tobin s Q, however, are sustained even after controlling for firm heterogeneity. These findings are consistent with the results based upon accounting profitability. We then identified instrument variables to be used in two stage regression models. The previous analysis suggests it is critical to control for unobserved firm heterogeneity to avoid any spurious causal relations. We initially selected only a partial set of potential instruments to define the system as just-identified (Johnston, 1984). Sales growth is a significant indicator of both profitability and Tobin s Q in the firm performance model. Market share is significant only in the profitability model and is strongly significant in the equation for the family portion of inside

23 ownership, which leads us to prefer sales growth (SALGR) to market share. As for the ownership models, group equity (GEQ) and volatility (VAR), which are not significant in either the profitability or Tobin s Q model but are significant in the ownership models, become our primary candidates. Firm equity is not included in just-identified models since it is significant only in the inside ownership equation but not in the family portion equation. Firm sales, which is highly correlated with firm equity, is dropped to avoid any multicollinearity problem. Years of listing (YRL) becomes significant in models (4)-(6), and we thus decided to include it as a control in all models. With our choices, we re-formulated our system of equations (1)-(3) to (1) -(3) between firm performance (PEFORM) and inside ownership (INSIDE) and family portion (FAMILY): PERFORM it = β 12 INSIDE it +β 13 FAMILY it +γ 11 YRL it +γ 12 SALGR it +firm and year dummies+ε 1it (1) INSIDE it = β 21 PERFORM it + γ 21 YRL it + γ 23 GEQ it + γ 24 VAR it + firm and year dummies +ε 2it (2) FAMILY it = β 31 PERFORM it + γ 31 YRL it + γ 33 GEQ it + γ 34 VAR it + firm and year dummies +ε 3it (3) After controlling for endogeneity, the effect of inside ownership to performance vanishes, as in model (7) in Tables 2 and 3. In contrast, the relationship between performance and inside ownership is positively significant in model (8) in Tables 2 and 3. The impact of performance to the family portion of inside ownership remains insignificant. The results suggest that firm profitability or firm value affects the degree of inside ownership but not vice versa, supporting hypothesis 2(a) but not 1(a). Hypotheses 1(b) and 2(b), which posit a relationship between family portion of inside ownership and firm performance, are not supported. To demonstrate the robustness of our results, we re-estimated the system of equations including all potential instruments (i.e., firm equity, group equity, years of listing, volatility, sales growth, and market share). These over-identified models are shown in models (10)-(12). Results from these models

24 are generally consistent with those from just-identified models. In general, our instrument variables are consistent with our expectations. Sales growth and market share are important predictors of firm performance. The group equity variable is negative for inside ownership and family portion, reflecting wealth constraints. Volatility is positive for inside ownership, consistent with prior results (Demsetz & Lehn, 1985; Cho, 1998). It is negative for family portion, suggesting that families avoid direct ownership in firms with higher volatility and instead use indirect ownership to control these firms. CONCLUSIONS AND IMPLICATIONS This study examined the simultaneous relationships between ownership structure and firm performance for chaebol affiliates. Our results show chaebol families and their affiliates take higher ownership stakes in more profitable or more promising firms and lower stakes in less profitable or less promising ones. We find no evidence for the conventional argument linking concentrated ownership to better firm performance. Contrary to agency theorists arguments, the results suggest that the observed positive relationship between ownership concentration and performance is an artifact of endogeneity; it is not evidence of effective monitoring of and control over managers. This study has several important implications. First, it suggests that strategy researchers should turn their attention from agency problems stemming from managers per se, to the agency problems prevalent in the rest of the world. Our results indicate that chaebol families capitalize on inside information by taking higher direct and indirect equity stakes in more profitable or promising affiliates, thus indicating that family-based chaebols pose severe agency problems. The various machinations large shareholders use to expropriate value from minority shareholders is a

25 serious, ongoing problem that is perhaps more prevalent in emerging economies, where corporate governance systems are not well developed and legal infrastructures to enforce laws are not strong enough to discourage such behavior (Coffee, 2001). Nonetheless, similar forms of expropriation also occur in comparatively well-developed economies. Second, this study systematically examined the determinants of family and affiliate ownership in family-controlled business groups. Prior studies that examined the relationship between concentrated ownership and firm performance did not explicitly consider the existence of business groups and the possibility that legally independent firms are actually controlled by outsiders. This study suggests that in family controlled business groups, affiliate ownership is an explicit substitute for family ownership to artificially inflate family owners control rights relative to their cash flow rights. Third, this study supports prior work positing that firm profitability determines concentrate ownership but not vice versa (Demsetz & Lehn, 1985; Kole, 1996; Cho, 1998; Loderer and Martin, 1997). Fourth, this study not only shows evidence of questionable intragroup business transactions but also highlights the governance mechanisms that let chaebol families perform these actions. Researchers should pursue direct evidence of these and similar phenomena in other countries. This study also offers some practical guidance for restructuring business groups in emerging economies. During the Asian Crisis, foreign investors sold stocks en masse and withdrew their capital from the region, causing the stock market to plummet and exchange rates to skyrocket. Many business groups went bankrupt since they could not repay their debts. Johnson, Boone, Breach and Friedman (2000) argued that the weakness of legal institutions for

26 supporting corporate governance, which could have prevented the expropriation of minority shareholders, might have caused the Asian crisis. Chaebols cross-shareholding lets chaebol chairmen make decisions to the detriment of minority shareholders while remaining totally unaccountable for their actions. Such governance failures caused the flight of foreign capital from the stock market and the failure to roll over foreign loans. In turn, these events caused Korea s banking system to collapse. This study suggests it is necessary to improve governance mechanisms further to restructure business groups properly. Governments should tighten disclosure requirements and protect the rights of minority shareholders by allowing class action lawsuits. It should also ferret out abuse of intra-group transactions. Nonetheless, this study has several limitations. It developed theories and hypotheses in the setting of family controlled business groups. Thus, its results may not be applicable either to independent firms or to business groups with no obvious family ownership. Although chaebollike business groups may be prevalent worldwide (La Porta et al., 1999) this study s theories and hypotheses may be more applicable to firms in emerging market economies where there are weak institutional mechanisms. In addition, we need to further develop empirical instrument variables to parcel out causal relationships between ownership and performance, especially to explore the relationship between the family portion of inside ownership and firm performance. In sum, this study suggests researchers should be more cautious about positing and testing causal relationships between ownership and performance. Our evidence indicates agency theory should pay more attention to the differences in power and interests between large and minority shareholders and the gap between control rights and cash flow rights created by pyramids and cross-shareholding. Finally, we highlight the need for more effective monitoring

27 devices and remedies that limit the power of large shareholders relative to minority shareholders. ENDNOTES 1 See Korea Fair Trade Commission case number 9712 for the official investigation of this case. 2 Weiss and Nikinkin (1998) and Bertrand et al. (2000) showed evidence of tunneling in Czechoslovakia and India, respectively. 3 We used the fixed effects model rather than the random effects model since the Hausman s specification test rejected the assumption that the firm-level random effects are not correlated with the other regressors. 4 In order to test the robustness of our models, we re-estimated the models with family and affiliate ownership themselves as endogenous variables instead of inside ownership and the family portion. The results are consistent with those reported in this study.

28 REFERENCES Barclay, M. & Holderness, C Private benefits from control of public corporation. Journal of Financial Economics, 25: Bergstrom, C. & Rydqvist, K Ownership of equity in dual-class firms. Journal of Banking and Finance, 14: Berle, A. & Means, G The modern corporation and private property, New York: Harcourt, Brace, and World. Bertrand, M., Mehta, P. & Mullainathan, S Ferreting out tunneling: an application to Indian business group. Mimeo. Bethel, J. & Liebeskind, J The effect of ownership structure on corporate restructuring. Strategic Management Journal, Summer Special Issue 14: Chang, S. & Choi, U Strategy, structure and performance of Korean business groups. Journal of Industrial Economics, 37: Chang, S. & Hong, J Economic performance of group-affiliated companies in Korea: intragroup resource sharing and internal business transaction. Academy of Management Journal, 43(3): Cho, M Ownership structure, investment, and the corporate value: an empirical analysis. Journal of Financial Economics, 47: Chatterjee, S. & Wernerfelt, B The link between resources and type of diversification: theory and evidence. Strategic M anagement Journal, 12: Claessens, S., Djankov, S. & Lang, L The separation of ownership and control in East Asian Corporations. Journal of Financial Economics, 58:

29 Claessens, S., Djankov, S. Fan, J. & Lang, L Expropriation of minority shareholders: evidence from East Asia. The IMF Working Paper. Coffee, J The Rise of Dispersed Ownership: Identifying the Precondition to the Separation of Ownership and Control. Yale Law Journal, 111(1), Dalton, D., Daily, C., & Johnson, J. & Ellstrand, A Number of directors and financial performance: a meta-analysis. Academy of Management Journal, 42(6): Demsetz, H The structure of ownership and the theory of the firm. Journal of Law and Economics, 26: Demsetz, H. & Lehn, K The structure of corporate ownership: cause and consequences. Journal of Political Economy, 93 (6): Dharwadkar, R., George, G. & Brandes, P Privatization in emerging economies: an agency theory perspective. Academy of Management Review, 25(3): Fama, E. and Jensen, M Separation of ownership and control. Journal of Law and Economics, 26: Froot, K., D. Scharfstein, & Stein, J A framework for risk management. Harvard Business Review, November-December, 72: Gedajlovic, E., & Shapiro, D Management and ownership effects: evidence from five countries. Strategic Management Journal, 19: Ghemawat, P., and Khanna, T The Nature of Diversified Business Groups: A Research Design and Two Case Studies. Journal of Industrial Economics 46: Granovetter, M Coase Revisited: Business Groups in the M odern Economy. Industrial and Corporate Change 4(1): Greene, W., 1993, Econometric Analysis. 3 rd edition, Prentice Hill.