EUROPEAN CORPORATE GOVERNANCE INSTITUTE (ECGI)

Finance Working Paper No. 51/2004

and

KDI SCHOOL OF PUBLIC POLICY & MANAGEMENT

Working Paper No. 04-20

What Determines the Ownership Structure of

Business Conglomerates?: On the Cash
Flow Rights of Korea's Chaebol

Woochan Kim
KDI School of Public Policy & Management, Korea

Youngjae Lim
Korea Development Institute

Taeyoon Sung
Korea Advanced Institute of Science and Technology (KAIST), Korea

Woochan Kim, Youngjae Lim and Taeyoon Sung 2004. All rights reserved.
Short sections of text, not to exceed two paragraphs, may be quoted without
explicit permission provided that full credit, includ-ing notice, is given to the
source.

This paper can be downloaded without charge

from:http://ssrn.com/abstract=594741

www.ecgi.org/wp
 Abstract
We examine a number of rm- and group-level factors that shape the ownership structure
of business conglomerates, which can include both public and non-public rms. Using an
exclusive set of 1997-2002 data on the intra-group shareholdings of 46 of Koreas largest
conglomerates, or chaebols, we show that the contribution of individual rms to group
control and protability are important determinants of ownership structure in Koreas
chaebol: the controlling shareholders cash ow rights in a group-afliated rm increases
with the rms contribution to group control and protability. We also nd that the level
of disparity between voting and cash ow rights is signicantly higher than the levels
previously reported in the literature on Korean rms, which makes use of only public
rms, indicating that non-public rms play a substantial role in increasing the disparity.

Keywords: ownership structure, business conglomerate, chaebol, voting rights, cash ow

rights, disparity

JEL Classifications: G32, G34

Woochan Kim
KDI School of Public Policy and Management
Chongryangri-Dong, Dongdaemun-Ku
Seoul 130-868
Korea
phone: (+82-2) 3299-1030, fax: (+82-2) 968-5072
e-mail: wc_kim@kdischool.ac.kr

Youngjae Lim
Korea Development Institute
Chongryangri-Dong, Dongdaemun-Ku
Seoul, Korea 130-012
email: yjlim@kdi.re.kr

Taeyoon Sung
KAIST - Graduate School of Management
Chongryangri-Dong, Dongdaemun-Ku
Seoul 130-722
Korea
phone: (+82-2) 958-3535, fax: (+82-2) 958-3604
e-mail: econsung@kgsm.kaist.ac.kr
 1. Introduction

Empirical research on corporate ownership is dominated by papers that treat

ownership as given. Early papers that study the link between corporate ownership and

firm value treat ownership as exogenous (Fama and Jensen, 1983 and Morck et al., 1988).

Even recent works that study the disparity between voting and cash flow rights and their

implications on other variables treat the disparity as given (LLSV, 2002, Claessens et al.,

2002, Mitton, 2002, Fan and Wong, 2002, Joh, 2003, Haw et al., 2003, Lins, 2003, Lemmon

and Lins, 2003, and Baek et al., 2004).

This pattern of research is surprising given that scholars knew early on that corporate

ownership can be endogenously determined. For example, Demsetz and Lehn (1985) see

the structure of corporate ownership as an equilibrium outcome influenced by various cost

advantages and disadvantages. They suggest a number of factors that determine

ownership structure, providing empirical support gleaned from a sample of 511 U.S.

corporations in the second half of 1970s. The primary factors they highlight are a firms

size, a firms risk, regulation, and amenities.

Acknowledging the endogenous nature of corporate ownership, a couple of recent

works study corporate ownership in a simultaneous regression framework. For example,

using the 1991 Fortune 500 firms, Cho (1998) examines the relationship among ownership

structure, investment, and corporate value using two-stage least squares (2SLS). He finds

that investment affects corporate value, which in turn affects ownership structure. He

also finds that corporate value affects ownership structure, but not vice versa. Another

example is Chang (2003), who investigates the relationship between ownership structure

and firm performance using the same 2SLS framework. Using a sample of group-

affiliated public firms in Korea during 1986-96, he finds that performance determines

ownership, but not vice versa.

-1-
 Despite such efforts, research on the determinants of corporate ownership is still

limited, and this paper hopes to fill the gap by studying the ownership structure of large

business conglomerates in Korea also known as the chaebol. A number of factors make

Korea a particularly interesting country to study. First, Korea is dominated by chaebol

groups, which makes Korea an attractive place to study the ownership structure of

business conglomerates. In fact, in a typical chaebol group, there is a great deal of intra-

group shareholdings among the affiliated firms, and such shareholdings separate voting

rights from cash flow rights.1 Such separation between the two allows us to study the

determinants of disparity, which is often defined as voting rights minus cash flow rights.

Second, in case of chaebols, there exists a natural person who is the common controlling

shareholder for multiple companies. This makes Korea an interesting country to study

how the ownership structures of group-affiliated companies are shaped to maximize the

controlling shareholders interest. Third, many papers have already shown that disparity

has a material implication on firm value, profitability, and share return in Korea (Joh, 2003,

Baek et al., 2004, and Black, Jang, and Kim, 2004). This further makes our study on

Korean chaebols relevant.

In this paper, we make use of a unique dataset from the Korea Fair Trade Commission

(KFTC) that contains comprehensive and detailed information on the intra-group

shareholdings of 46 large business conglomerates over a six-year period (1997-2002).

Using this dataset, we show that the ownership structures of group-affiliated firms are

influenced by the firms contribution to their respective groups control and profitability.

Specifically, we show that the controlling shareholders of chaebol groups tend to increase

their cash flow rights in firms that have higher contribution to group control and in firms

1
Among the three sources of separation between the cash flow right and the voting right suggested by
Bebchuk et al. (2000), intra-group shareholding and pyramidal ownership structure are the two main sources
in Korea. Dual class shares are not permitted.
2
Joh (2003) is an exception, which includes non-public firms in her data. But, the data set is not from
KFTC, but of her own construction, which is likely to be incomplete and prone to measurement errors.

-2-
that have higher profitability.

This study makes a number of contributions to the existing literature. First, it studies

the ownership structure of unlisted non-public firms as well as public firms. Given the

difficulty of obtaining ownership data of unlisted firms, previous research solely focused

on public firms (Demsetz, 1998, Cho, 1998, and Chang, 2003).2 This can be problematic

when one is studying the ownership structure of the chaebol, which often include many

non-public firms. It should also be noted that, in Korea, non-listed firms in which the

controlling shareholder possesses a high direct ownership stake often serve as the de facto

holding company for the group. Also, by omitting the control chains containing a non-

listed company, the measures of voting and cash flow rights of a listed company may also

be biased. In fact, our measure of disparity turns out to be significantly higher than those

reported in the existing literature, which makes use of only public firms.

Second, we adopt the KFTCs highly flexible concept of control. This allows us to

more accurately measure the true extent of control. According to the KFTCs definition,

share ownership, whether direct or indirect, is not a necessary condition for a person (or a

company) to wield control over firms. In addition to share ownership, other indicators of

control, such as the power to appointment directors, make personnel exchanges, conduct

transactions above a normal range, or using similar trademarks across firms in the group,

are taken into account when identifying which firms are under the controlling

shareholders influence. Overly restrictive condition of share ownership can result in a

downward bias of the voting right measure.

Third, we introduce an algorithm for computing the cash flow rights of firms that

share a common controlling shareholder. Using an (nn) matrix of intra-group

shareholdings and an (n1) vector of controlling shareholders direct ownership, we can

compute the cash flow rights of a common controlling shareholder for all the n-affiliated

firms simultaneously.

-3-
 Fourth, we study the group control motivation of a controlling shareholder and its

implications on the controlling shareholders ownership pattern. We emphasize in this

paper that concern over group control is particularly important when examining the

ownership structure of a business conglomerate. To test this, we propose an algorithm

that allows us to measure how much contribution each company makes to group control.

This paper is organized as follows. Section 2 describes the data. Section 3 explains

how voting rights, cash flow rights, and the disparity are computed. Section 4 discusses

hypotheses and methodology. Empirical results are discussed in Section 5. Section 6 is

the conclusion.

2. Ownership Data

2.1. Source

In this paper, we make use of a unique dataset exclusively obtained from KFTC that

contains detailed information on the intra-group shareholdings of 46 large business

conglomerates over a six-year period (1997 2002). For each of the firms in the 46 large

business conglomerates, we have data on the number of shares (common and preferred)

held by the controlling shareholder and by his related parities. Related parties include

relatives, senior managers of the firm, affiliated not-for-profit organizations, and affiliated

firms. In our data set, such information is available even for non-listed companies.

Table 1 shows the intra-group shareholding matrix of Samsung Group in 2002.

KFTC compiles such a database to monitor and enforce compliance of its regulations

by large business conglomerates. Each year in April, KFTC announces the top 30

-4-
business conglomerates, known as chaebol, in terms of their total asset size, and imposes

regulations, including equity investment ceilings, bans on cross-shareholdings, and

restrictions on related-party transactions.3 The regulations are explained in greater detail

in Appendix 3.

2.2. Main Features

This dataset has a number of nice features that cannot be found in the datasets used in

the existing literature. First, it contains data on non-listed firms, as well as listed firms.

This is an important feature in two ways: it allows us to measure the disparity of non-

listed firms and also allows us to correctly measure the disparity of listed firms.

To measure the precise voting and cash flow rights, one needs the ownership data of

each and every firm in the control chain. But this was not possible in the existing

literature since no ownership information was available for non-listed firms (see Claessens

et al., 2000; LLSV, 2002; and Lins, 2003).4 Such lack of information is a concern in that it

may result in downward-biased measures of listed companies voting and cash flow rights

by unduly omitting the chain containing a non-listed company. Suppose a controlling

shareholder holds a 25% share in firm A, which is listed, and a 51% share in firm B, which

is not listed. Furthermore, suppose that firm B holds a 25% share in firm A. If we

include firm B in our computation, its control over firm As voting rights would be 50% (=

25%+25%) and the cash flow rights would be 37.8% (= 25%+51%25%). However, if we

exclude firm B from our computation, the voting and cash flow rights would be equally

25%, which is heavily downward biased.5 This problem is completely eliminated in our

3
In 2002, KFTC changed the way it designates large business conglomerates. Instead of ranking them based
on asset size (e.g. top 30), it now uses asset size thresholds (e.g. above 5 trillion won).
4
Even for the listed firms, these studies use ownership data only for a subset of companies, ranging from a
half to three-quarters of all listed firms in terms of market capitalization.
5
In this example, the disparity measure, which is defined by voting rights minus cash flow rights, is also

-5-
study because we have complete ownership data even for non-listed firms.

Second, we use a highly flexible concept of control in this study, thus allowing us to

measure control in line with reality. In most of the studies on voting-cash flow disparity,

a person can become a controlling shareholder only if he owns company shares, whether

directly or indirectly. But, we believe this can be an overly restrictive condition, which

again results in a downward bias of the voting right measure.6 Suppose a person makes

a significant donation to a not-for-profit organization (NPO), sits on its board, and thus

controls the voting rights on the company shares held by the NPO. If such cases, the

company shares held by this NPO should be included in the computation of this persons

voting rights. Yet this is not how voting rights are computed in the existing studies.

In this study, we adopt the concept of related parties and de facto control officially used

by the KTFC, thus allowing the concept of control to be flexible enough to incorporate the

example given above. As such, we define control over voting rights as the total sum of

direct share ownerships held by the controlling shareholder and its related parties, which

includes not just spouse/relatives, but also not-for-profit organizations and firms under

the de facto control of the controlling shareholder. Also, the controlling shareholder is

defined to be a person who, alone or with its related parties, has de facto control of the

company. This is in contrast with the definitions used in other studies that are somewhat

ad hoc and fail to capture the actual extent of control.7 Detailed definitions of related

parties and de facto control are provided in the next section and in Appendix 2.

One might be concerned that the concept of de facto control is overly subjective and

the decision to classify a firm to be de facto controlled can be arbitrary. Such criticism,

downward-biased.
6
It will also result in a downward-bias of the disparity measure. One exception would be Lins (2003).
See Appendix 2 for detailed explanation of how Lins (2003) computes voting rights. Note, however, that his
concept of control is still far more restrictive than ours.
7
For example, Claessens et al. (2000) defines control rights as the sum of the weakest links in the chains of
voting rights.

-6-
however, is unfounded, for no firm subject to the KFTC regulation petitioned against the

KFTCs decision. When a firm is designated as a firm de facto controlled by the

controlling shareholder, it will be subject to serious regulations, which can be binding and

sometimes costly.8 So a manger of this company would have all the reasons to find

justifications not to be classified as a firm under de facto control. Nevertheless, no

company petitioned the KFTCs decision, which implies that the KFTC designations were

made based on reality, not arbitrarily.

Our dataset has other minor improvements over those used in the existing literature.

For example, we consider even those shareholders who own less than 5% of shares

outstanding. Existing studies use datasets that contain information only on block

shareholders holding more than 5% of outstanding shares. Another improvement is that

we distinguish individual family members when identifying the controlling shareholder,

and use a clear definition of the term relatives. A relative is classified as a party with a

blood relationship of eight degrees or less (four or less if he/she has a blood relationship

with the controlling shareholders spouse) to the controlling shareholder.

2.3. Sample Selection

Our dataset contains data on 65 conglomerates. We eliminated the 18 conglomerates

controlled by a state-owned enterprise, where the controlling shareholder is a company

and not a natural person, or by a newly privatized firm.9,10 This means that we only

8
See Appendix 3 for the detailed regulations.
9
These conglomerates have only a small number of affiliated firms, which makes the disparity measure close
to zero throughout the sample period, and thus makes it inappropriate to study the determinants of disparity,
one of this papers main topics.
10
18 large business conglomerates, the controlling shareholder of which is not a natural person: Daewoo
Electronics, Daewoo Motors, Daewoo Shipbuilding & Marine Engineering, Hanaro Telecom, Hyundai Oil,
KARICO, KEPCO, Kia, KOGAS, Korea Highway Corporation, Korea Land Corporation, Korea National
Housing Corporation, KOWACO, KT, KT&G, MBC, POSCO, S-Oil. We also drop Kohap group in year

-7-
investigate those conglomerates controlled by an individual, in line with the concept of

chaebol, and thus focus on the decision of a controlling shareholder as an individual We

do not include the Lotte Group, since KFTC does not have its complete intra-group

shareholding information. This is because the affiliated firms of Lotte Group are

established not only in Korea, but also in Japan, and KFTC does not have ownership data

for the firms located in Japan. Thus, we use 46 business conglomerates. The total

number of firm-years is 5,202 (see Table 4 for summary statistics).

2.4. Other Variables

To test the hypotheses in this paper, we merge the ownership data with other firm-

level variables, which we obtained from the National Information and Credit Evaluation,

Inc. (NICE). Since disparity is used as a dependent variable in our regression analyses,

we make sure that other firm-level variables are measured prior to the disparity variable.

As such, they are measured during or on the last day of the fiscal year, which ends before

April. When the fiscal year changes during the sample year, we keep only those years

that cover twelve full months. Table 3 shows the list of variables along with their

definitions. Table 4 provides some summary statistics for each variable.

3. Computation of Voting Rights and Cash Flow Rights

3.1. Controlling Shareholder

2001, the year of which its controlling shareholder is no longer a natural person.
11
Suppose firm A holds 10% of firm Bs total outstanding common shares, which includes treasury stocks.
If the fraction of treasury stocks is 5% out of the total outstanding common shares, the adjusted fraction of
voting right is (0.1)/(1-0.05)=0.105 (10.5%).

-8-
 The first step in computing voting-cash flow disparity is to identify the controlling

shareholder and the firms under his de facto control. In this study, following the method

adopted by KFTC, a controlling shareholder is defined as a person who, alone or with his

related parties, has de facto control of the company. Here we explain in detail the concepts

of related parties and de facto control.

Related parties include (i) relatives, (ii) not-for-profit organizations where the

controlling shareholder, alone or with related parities, contributed 30% of total donations,

(iv) not-for-profit organizations where the controlling shareholder, directly or through

related parities, has a controlling influence over the appointment of directors or business

activities, (v) any company whose business is controlled de facto by the controlling

shareholder, and (vi) agents of the controlling shareholder or his related parties, including

senior managers.

The controlling shareholder has de facto control of a particular company if any of the

following conditions are met: (i) the controlling shareholder, alone or with his related

parities, owns 30% of voting shares issued and is the largest shareholder, (ii) the controlling

shareholder appoints the representative director or at least half of the directors, (iii) the

controlling shareholder directly or through related parties has a controlling influence over

corporate strategy decisions, (iv) the company concerned and the company controlled de

facto have a personnel exchange system in place, (v) the company and the controlling

shareholder or its related parties conduct transactions of funds, assets, goods, services, or

debt guarantees above a normal level, (vi) the company can be reasonably considered

under social norms to be an affiliate of the business group controlled by the controlling

shareholder (e.g. using similar trademarks). More detailed definitions of related parties

and de facto control are outlined in Appendix 2.

3.2. Control Over Voting Rights

-9-
 In this paper, control over voting rights (hereafter voting rights) is defined as the sum

of direct share ownership held by the controlling shareholder and its related parties.

Assume that di is the direct share ownership held by the controlling shareholder in firm i.

Assume also that ri is the direct share ownership held by the related parties, including

relatives, not-for-profit organizations, and senior managers under the controlling

shareholders influence. Lastly, assume that sij is the direct share ownership in firm i held

by firm j, which is under the controlling shareholders influence. Then, a voting right for

firm i can be defined by equation (1).

n
vri = d i + ri + s ij (1)
j =1

n is the number of for-profit-firms under the controlling shareholders influence. When

computing the fraction of shares, we use common shares only, and also adjust for treasury

stocks, which do not have any voting rights, in accordance with the Commercial Code.11

Two points should be mentioned here. First, we want to differentiate between the

concept of control and control over voting rights. The concept of control, as explained

earlier, takes into account not only share ownership, but also other routes of control, such

as the appointment of directors, personnel exchange, abnormal transaction levels, and so

on. Control over voting rights, on the other hand, considers only share ownership.

Thus, it is a narrower concept. Although the controlling shareholder controls de facto all

the affiliated firms in the sample, it does not mean he has 100% control over the votes.

When computing disparity, we use the concept of voting rights. Otherwise, disparity

would be always 1 cash flow rights.

Second, our measure does not give special treatment to voting rights over 50 percent.

It is true that once a controlling shareholder holds more than 50 percent of the votes, he

- 10 -
would be able to block or pass any resolution item at the shareholders meeting. And

some might suggest that we should therefore grant special treatment to voting rights over

50 percent. However, we have decided against this on two grounds. First, depending

upon the ownership structure of outside shareholders and how quickly they can

coordinate, the threshold can be below 50 percent. Second, such information cannot be

known ex ante.

3.3. Cash-Flow Rights

Cash-flow rights are defined as sum of the products of ownership stakes held by the

controlling shareholder and his family members along the voting right chain. Assume

that fi is the direct share ownership held by the controlling shareholders family members

in firm i. Family members include spouse and relatives that are within certain degrees of

kinship.12 Shares held by senior mangers or not-for-profit organizations are excluded

from the computation of cash-flow rights. Thus, cash flow rights in firm i can be

computed by equation (2).13

n n n
cfri = d i + f i + s
j =1
ij (d j + f j ) + s s
j =1
ij
k =1
jk (d k + f k ) + (2)

The first two terms are direct ownerships levels of the controlling shareholder and his

family members. The subsequent terms are the indirect ownership levels of the

controlling shareholder and his family members through affiliated for-profit firms. To be

specific, the third term is indirect ownership in firm i through firm j (j can take values

from 1 to n). The fourth term is indirect ownership in firm i through firm k and firm j (k

12
See Appendix 2 for detailed explanation of relatives.
13
The cash-flow rights algorithm is taken from Kim (2000).

- 11 -
can also take values from 1 to n).

To simplify, we can express the cash-flow rights of all for-profit firms in matrix form.

Let d and f be (n1) vectors of direct ownerships held by the controlling shareholder and

his family members. Let S be (nn) matrix of share ownership of for-profit firms in other

for-profit firms.14 Then, equation (3) computes the cash-flow rights of all for-profit firms

under the controlling shareholders influence.

cfr = ( d + f ) + S (d + f ) + S 2 (d + f ) + S 3 (d + f ) + (3)

Equation (3) can be further simplified by using an inverse matrix.

cfr = ( I S ) 1 (d + f ) (4)

3.4. Disparity

In this paper we use the difference between voting and cash flow rights as our measure

of disparity. Though there are studies that use the ratio, instead of the difference

between the two, or the difference scaled by voting rights,15 but there are two reasons why

we prefer the simple difference. First, in our regression specification examining the

disparity between voting and cash flow rights, we include voting rights on the right-hand

side, which makes it unnecessary to scale the difference by voting rights. Second, since

we do not necessarily require our voting or cash flow rights to be non-zero, it is sometimes

impossible to divide by voting rights or cash flow rights. (Note that in our study, a

14
Elements in the diagonal are zero, since fractions of shares are already adjusted for treasury stocks.
15
LLSV (1999) and Joh (2003) use [voting rights cash-flow rights]; Claessens et al. (2000) and Mitton
(2002) use [cash-flow rights / voting rights]; Lins (2003) uses [voting rights / cash-flow rights]; and Fan and
Wong (2002) and Haw et al. (2003) use [voting rights cash-flow rights]/[voting rights].

- 12 -
controlling shareholder does not necessarily have to own company shares, whether

directly or indirectly, to become a controlling shareholder.) As for group-level disparity,

it can be easily computed by a weighted average of firm-level disparities. Book equity

values are used as weights. We treat firms with negative book equity as missing values.

Table 2 shows the voting rights, cash-flow rights, and the disparity of each firm in

Samsung Group as of April 2002. For simplicity of presentation, of the 63 firms in our

sample classified by the KFTC as a Samsung Group affiliate, we show only the 27 major

firms that appear in the 2002 Samsung Group Annual Report. The firms are ranked in

terms of their degree of disparity. One can see that there is a plenty of variance in the

disparity measure across firms, ranging from 8% to 97.6%. Samsung Electronics, the

largest company in Korea, has a disparity of 11.7%. The controlling shareholder, Mr.

Kun-Hee Lee, has cash flow rights of 5.3% and voting rights of 17.0%. On the other hand,

Samsung Card, which experienced a business failure in 2003 and was acquired by

Samsung Life Insurance, has a disparity of 84.8%. While the controlling shareholder, Mr.

Kun-Hee Lee, has a cash-flow right of only 3.6%, he controls 88.4% of the voting rights.

4. Hypotheses and Methodologies

In this paper, we try to identify a number of factors that influence firm-level disparity.

We focus mainly on three factors: profitability, risk and contribution to group control.

We hypothesize that the controlling shareholder would increase his disparity in a firm

with low profitability or high risk, and increase her cash flow rights (or direct ownership)

in a firm that heavily contributes to group control.

The intuition is simple. If a firm demonstrates poor prospects by showing low

- 13 -
profitability or high risk, the controlling shareholder may change the ownership structure

of the firm to minimize any further loss.16 That is, he may sell her direct ownership stake

to an affiliated firm that he also controls. This will increase the ailing companys

disparity. Furthermore, if possible, he may try to sell the stake at a price higher than its

true value. If the controlling shareholders cash flow rights are relatively low in the

acquiring firm, there will be an illegal transfer of wealth from outside minority

shareholders to the controlling shareholder.

Disparity can increase even when the controlling shareholder does not sell any of her

directly owned shares to the affiliated firms. Often, poorly performing firms need

injections of new equity capital. But if a controlling shareholder is not confident of the

firms survival prospects, he would not participate in purchasing the newly issued shares.

Instead, he would instruct other affiliated firms under his control to participate. This

would dilute his equity stake while increasing the equity stakes of other affiliated firms.

It is not hard to find actual cases of such phenomena in Korea. In 1998, SK Securities

was at the brink of liquidation after several consecutive years of negative income, largely

attributable to a US$200 million loss in 1997 from its trades in Indonesian derivatives. To

rescue SK Securities, affiliated firms within the SK Group, several affiliates, including SK

Trading and SK Energy Sales, injected new equity capital into the firm. This increased

the equity stake of affiliated firms, but decreased that of the Choi family, the controlling

shareholder.

We also attempt to provide empirical evidence that a controlling shareholder would

tend to hold greater cash flow rights (or direct ownership) in firms that are essential in the

control chain of a group and that allow him to wield de facto control over the whole

16
With regard to firm risk, one may suggest an opposite hypothesis. According to Demsetz and Lehn (1985),
firm-specific uncertainty increases the monitoring cost of managerial performance, which thus increases the
benefit of higher ownership concentration.

- 14 -
group.17 In any conglomerate, some firms will have relatively large cash flow or hold

voting rights in other affiliates, whether directly or indirectly. If the controlling

shareholder loses control over the companies, she may lose control over the entire group,

or at least a significant portion of it. To prevent this from happening, the controlling

shareholder has a strong incentive to hold greater cash flow rights (or direct ownership) in

such companies.

Anecdotal evidence of this group control incentive by the controlling shareholder can be

easily found in Korea. On one occasion, a chaebol group eagerly requested the

government to allow financial institutions to vote on shares issued by their industrial

affiliates. On another, a chaebol group chairman was prosecuted for malfeasance when he

masterminded a series of equity transactions among the affiliated firms that would have

strengthened his group control.

To investigate whether a firms choice of disparity (or cash flow rights) varies

according to its profitability, risk, and contribution to group control, we run pooled OLS

regressions with fixed year and group effects as well as year-by-year regressions.

4.1. Measures of Profitability, Risk and Contribution to Group Control

In our basic model, we measured firm profitability by (EBIT / Asset). For this reason,

we eliminate firms with no accounting data from our sample.18 To check for robustness,

we try alternative measures of firm profitability: (Ordinary income / Assets) and (Net

income / Assets). Ordinary income is an income statement item unique to Korea; it is

defined as earnings before taxes and extraordinary items, but after interest payments.

17
Alternatively, the controlling shareholder would change the group ownership structure so that it increases
the degree of group control contribution of a firm in which he has a high level of cash flow rights (or direct
ownership).
18
That is, firms not subject to the Act on External Audit of Stock Companies, i.e. those that are not required to
receive an external audit, are dropped from the sample.

- 15 -
We also try a three-year average in addition to a single-year measure of profitability. A

greater coefficient on the three-year average indicates that the controlling shareholder

responds with a lag to changes in profitability.

On the other hand, firm risk is measured by beta, which is estimated from a market

model where the KOSPI return is used as a proxy for market return. KOSPI is a value-

weighted market index comprising all of the listed companies in the Korea Stock Exchange

(KSE). For each year, we estimate the model using data during the past three-years with

a monthly frequency. Note that beta is replaced by a missing value if the underlying

regression uses less than 20 observations. When we use beta as our measure of firm risk,

firms not listed in KSE are dropped from the sample. One advantage of using beta as our

measure of risk is that it assumes that the controlling shareholder of a business group

holds a well-diversified portfolio of companies, which is in fact true in the case of Korea.

The greatest challenge in testing our last hypothesis (contribution to group control

and the choice of cash flow rights) is finding an objective algorithm to measure how each

firm is important in the group control chain. This task can be done by making use of our

unique intra-group shareholding matrix. Here we quantify the contribution to group

control of firm j by the amount of additional cash flow rights the controlling shareholder

would gain in other companies by having company j under his control as a fraction of

company js book equity value. Let this measure be named contribution index 1 (or con1).

Equation (5) shows the formula of firm js contribution index 1:

n n

i =1,i j
Ei cfri E cfr
i =1,i j
i i
j

con1 j = (5)
Ej

Ei is firm is book value of equity. cfri is the cash flow rights computed for firm i when all

affiliated firms of each chaebol group are included in the group-ownership structure. This

- 16 -
means that cfri is the cash flow rights computed for firm i when firm j is included in the

group-ownership structure. On the other hand, cfri-j is the cash flow rights computed for

firm i when firm j is excluded from the group-ownership structure. Ej is the book value

of equity of firm j. The first term in the numerator measures the total cash flow rights the

controlling shareholder would receive from other companies i ( i j ) when firm j is

included in the chaebol group. On the other hand, the second term in the numerator

captures the total cash flow rights the controlling shareholder would receive from other

companies i ( i j ) if firm j were excluded from the chaebol group. We divide the

difference by the companys book equity value to control for a size effect, since larger

firms could have greater contributions to group control.

Contribution to group control can also be measured using the voting rights difference,

instead of the cash flow rights difference. That is, the amount of additional voting rights

the controlling shareholder would gain in other companies by having company j under his

control as a fraction of the companys book equity value. Let this measure be named

contribution index 2 (or con2). Equation (6) shows the formula for firm js contribution

index 2:

n n

i =1, j i
Ei vri E vr
i =1,i j
i i
j

con2 j = (6)
Ej

The indices can have a value equal to zero. This happens when firm j does not have any

equity investment in other affiliated firms.19 It should also be noted that the indices have

19
Contribution index 1 (computed by cash flow rights) can be zero for another reason. Say a controlling
shareholder controls company A without holding any of its shares directly, while company A exclusively owns
company B. Also assume company B has equity stakes in companies C and D. In such a situation, the
contribution index 1 for company B is zero (since the controlling shareholder has no share of company A),
while the contribution index 2 for company B is positive (since company B holds equity stakes in companies
C and D).

- 17 -
no upper bounds. If there is no restriction on leverage or the length of the equity

investment chain, the indices can be well above 1.

4.2. Control Variables

Besides voting rights, a number of other control variables are used in this paper. At

the firm level, we include years-of-operation, a public company dummy, firm size (book

equity value), a financial institution dummy, and leverage (debt-to-asset ratio). At the

group level, we add group size, number of affiliated firms within the group, and a dummy

variable that identifies groups with a financial institution. Table 3 provides definitions for

each of these control variables.

Years-of-operation can be negatively (positively) related to disparity (or cash flow

rights). 20 One explanation for this is that old firms tend to have large outside

shareholders, thus lowering their level of voting rights and, consequently, the degree of

disparity. Another explanation is that controlling shareholders have an incentive to

establish new firms as a subsidiary of an existing company, since by doing so they can

reduce the risk of entering a new business. Since it is a subsidiary with no (or little) direct

ownership, it will have a high level of disparity. Similar logic can also be applied to the

public companies dummy and firm size (book equity value). By using book equity

20
In this paper, we use different left-hand side variables, depending upon which determinant we are
examining. When we examine firm profitability or risk, disparity is our dependent variable (with voting
right as an additional control variable). But, when contribution to control is our variable of interest, we put
cash flow right on the left-hand side (without controlling for voting right). Two points should be noted here.
First, when we control for voting rights in the right-hand side, the coefficient on our variable of interest (e.g.
profitability or risk) will have the same absolute value (with opposite sign), regardless of which left-hand side
variable (e.g. disparity or cash-flow right) we use. This is because, when the level of voting right is
controlled (thus, fixed), an increase in disparity is exactly offset by a decrease in cash flow rights. Second,
when examining contribution to control, we do not put voting rights on the right-hand side. This is because
we are interested in the absolute level of cash-flow rights, not in the relative magnitude of cash-flow rights to
voting rights.

- 18 -
value as a proxy for firm size, we retain as many observations as possible.

Financial institutions may have a greater degree of disparity in Korea. In the past,

many chaebol-controlled financial institutions were used as a lending vehicle to support

other industrial firms within the chaebol group. They were not regarded as a separate

profit-making entity. Thus, controlling shareholders had an incentive to keep the level of

disparity, in such financial institutions, high. Otherwise, their losses would be

transferred to the controlling shareholder.

Leverage (the debt-to-asset ratio) can be either positively or negatively related to

disparity. A controlling shareholder may want to lower his direct ownership in highly

levered firms, where the risk to equity holders is high. Thus, there is a positive

relationship with the disparity. Conversely, leverage may also be negatively related to

disparity. Notice that there are two ways for a controlling shareholder to externally

finance a project without diluting his voting rights in the firm. One is issuing debt and

the other is issuing equity to affiliated firms. If a firm is highly leveraged, it means that

the controlling shareholder has been relying less on the latter method, thus resulting in a

lower disparity.

Group size is the sum of each affiliated firms book equity value. Again, book equity

is used so as to maximize the number of observations in the sample. We conjecture that

firm-level disparity will increase with group size, for if there is a large-sized affiliated firm

within the group, there will be a greater amount of equity investment in other affiliated

firms. Similar logic can be applied to the number of affiliated firms. The greater the

number of affiliated firms, the greater the amount of equity investment to other affiliated

firms, thus increasing firm-level disparity. Firms that are part of a chaebol group that

includes a financial institution may also have a higher degree of firm-level disparity.

21
As can be seen in Appendix 3 (Regulation of Large Business Conglomerates in Korea), financial
institutions are not subject to the 25% upper ceiling on equity investment.

- 19 -
This is because a financial institution is an efficient vehicle for a controlling shareholder to

strengthen his control over the group. There are two reasons for this. First, a financial

institution is usually highly leveraged, and thus only a small amount of capital is required

to acquire a controlling stake. Second, financial institutions can use their creditors

money to finance its equity investment in other affiliated firms.21

4.3. Econometric Issues

There are two major challenges when estimating OLS regressions of firm-level

disparity on its determinants. One is the issue of reverse causality. That is, disparity

can cause firm profitability (risk), rather than the other way around. For example, Joh

(2003) demonstrates that Korean firms with a high disparity between voting and cash-flow

rights tended to have low profitability during the pre-crisis period (1993-1997).22 Black,

Jang, and Kim (2004) also show that firms with high disparity tended to have low market

values, as measured by Tobins q, in 2001.

As a partial remedy to this problem, we make sure that our profitability measure is

pre-determined. A three-year average of past profitability is used to make the righthand-

side variable pre-determined. Even when a single-year measure is used, we compute

profitability during the fiscal year, which ends before April, the month in which disparity

is measured.

In particular, the causality issue between group control and the controlling

shareholders cash flow right (or direct ownership) is less problematic. Even if there is a

reverse causality issue the controlling shareholder increasing the level of group control

contribution of a firm in which he has high cash flow rights the ownership structure

22
One potential underlying cause could be illegal value transfers (also known as tunneling) from companies
with high disparity to those with low disparity.

- 20 -
must still be endogenously determined by the group control motivation.

The other econometric challenge is Korea-specific. As can be seen in Appendix 3, the

KFTC regulations changed during the sample period, and this could have influenced the

share ownership behavior by the controlling shareholders. Among the five regulations

listed, two (ban on new debt guarantees and board approval and disclosure of related

parity transactions) are not directly related to share ownership. Of the remaining three,

two other regulations (ban on cross-shareholdings and no voting rights for financial

institutions on shares issued by affiliated firms) changed only in the very last year of our

sample. Thus, the only regulation of concern to us is the upper ceiling on equity

investment.

Until February 1998, a 25% upper ceiling was applied to all the firms within the top 30

chaebols, except for financial institutions. One of the major reasons to impose such a

regulation was to limit equity investment among affiliated firms, and thus reduce

disparity between voting and cash-flow rights. In February 1998, however, this

regulation was unexpectedly lifted in the name of facilitating the corporate restructuring

of crisis-hit chaebol firms. Then, in April 2001, the regulation was restored, and in January

2002 additional moderate changes were made.23

Given this information, once can identify a period that is not contaminated by

government regulation: a period between February 1998 and March 2001. Since disparity

is measured each year in April, yearly regressions in 1999 and 2000 should be free from

any regulatory influence. Yearly regressions in 1998 and 2001, however, would be partly

contaminated. With regard to the 1998 regression, firms were subject to regulation

during a 10-month period from April 1997 and January 1998. In case of the 2001

regression, it may be influenced because the bill to restore the upper ceiling on equity

23
The legislation to restore the regulation passed in December 1999 and a 15-month grace period was
allowed.

- 21 -
investment was passed in December 1999, and firms knew that the regulation would be

binding from April 2001. Thus, the controlling shareholder must have changed her

shareholdings from the second half of 2000 at the latest to comply with the regulation soon

to be effective. To see if our results are robust to such regulatory changes, we run year-

by-year regressions and see if the coefficients are greater in year 1999 and 2000.

However, it should also be noted here that the upper ceiling on equity investment

tends to weaken, not strengthen, the link between our proposed determinants and

disparity. This means that the coefficients we obtain are downward biased in sample

years other than 1999 and 2000. Thus, any control of the regulation effect, will strengthen,

not weaken, the coefficients.

5. Results

5.1. Summary Statistics

Panel A of Table 4 reports summary statistics of each variable used in this paper. The

median voting rights measure is 74.59 percent, while the median cash flow rights measure

is only 12.95 percent. The median disparity is 45.04 percent. Table 4 also shows that the

levels of voting rights, cash flow rights, and disparity are higher for private firms. For

example, the median disparity is 50.69 percent for private firms, but only 27.83 percent for

public firms.

Another important result from Panel A is that the median values of both contribution

indices 1 and 2 are zero. In fact, more than half of 4,114 firm-years have zero contribution

indices. This led us to run two sets of regressions, one including such firms with zero

- 22 -
contribution index values, and another excluding them. Also, as can be seen in Panel A,

the maximum values of winsorized contribution indices 1 and 2 are 3.45 and 3.57,

respectively. These actual contribution index figures are consistent with reality. The

firms with the highest contribution index 1 in each group in 2002 are Samsung Everland

(Samsung Group), LG Corp (LG Group), and SK C&C (SK Group), and each of these firms

is commonly regarded by the investment community in Korea as the de facto holding

company of their respective groups. This strongly suggests that our measure is reliable.24

Panel A also shows that 31.91 percent of the firms in our sample are firms either listed

on the Korea Stock Exchange (KSE) or registered on the KOSDAQ stock market.

Financial institutions comprised 12.9 percent of our sample. The median number of

affiliates for each chaebol group was 29 (=exp (3.3673)).

Panel B of Table 4 compares the level of disparity with the existing literature. The first

column computes disparity as a difference between voting and cash flow rights. The

second column calculates disparity as a ratio of voting over cash flow rights. Two

observations can be made. First, the figures computed in previous studies are generally

lower than those computed in this paper. For example, Claessens et al. (2000) and Joh

(2003) report that the average difference between voting and cash flow rights in Korea are

only 4 percent and 23 percent, respectively. 25 Claessens (2000), Chang (2003), and

Lemmon and Lins (2003) also report that average cash flow rights leverages voting rights

over cash flow rights in Korea are only 1.27, 1.47, and 2.37, respectively.26 In contrast,

24
In eight out of top 10 chaebol groups, the firm with the highest index 1 is identical to the firm with the
highest index 2, which further makes our measure reliable.
25
Joh (2003), which includes private firms in the sample, reports that the average level of disparity is 23.47
percent in 1997. This figure is much lower than what we have. The difference comes from two sources:
First, Johs data includes not only chaebol-affiliated firms but non-affiliated firms as well. Second, Joh
(2003) uses a database not from KFTC, but of her own construction, which can thus be incomplete.
26
Chang (2003) originally computes fraction of inside ownership (equivalent to voting rights) and family
portion (equivalent to cash flow rights over voting rights). Average inside ownership and family portion are
29.9 percent and 68 percent, respectively. Inverse of family portion gives the cash flow rights leverage.

- 23 -
our mean figures are 47 percent for the difference between the voting and cash flow rights,

and 3.19 for the ratio of voting over cash flow rights. Second, even when comparing for a

restricted sample of public chaebol firms, one can see that our figure of 3.25 is significantly

greater than the figure of 1.47 computed by Chang (2003). Such a difference can be due

to two factors: (i) incorporating every control chains that involve unlisted firms and (ii)

using a flexible concept of control. Detailed explanations of our dataset and the concept

of control can be found in Section 2B and Section 3A, respectively.

5.2. Profitability, Risk, and Choice of Disparity

Panels A and B of Table 5 report the OLS regression results of voting-cash flow rights

disparity on profitability with additional control variables. Profitability is measured as

EBIT over assets. In Panel A, we use a single-year measure, while in Panel B we use a

three-year average. Extreme values of (EBIT/Asset) are winsorized. That is, as for

observations outside the 1st and the 99th percentiles, the (EBIT/Asset) values are replaced

by the 1st and the 99th percentile values.

Equation (5) in Panel A, which uses a full set of control variables, including group and

year dummies, shows that the coefficient on profitability is statistically significant and has

the expected sign. The t-value is 3.42. The magnitude of the coefficient, however, is not

economically large. A one-standard deviation increase in profitability decreases disparity

by 1.4 percentage points (= 0.1063 0.1310 = 0.0139).27 When we use a three-year average,

rather than a single-year measure of profitability, the coefficient and the t-value increase.

As can be seen in Panel B of Equation (5), the t-value is now 4.43, and a one-standard

deviation increase in profitability decreases disparity by 1.7 percentage points (= 0.0989

27
A change of profitability, from worst to best, decreases disparity by 10 percentage points.

- 24 -
0.1747 = 0.0173).28

From Table 5, one can also see that the coefficient size and sign on profitability is

robust to the control variables we include. For Equations (1) to (5), the coefficients on

profitability remain stable, lying within a certain range. Most of the control variables

have the expected signs. The coefficient on voting rights is positive and highly significant.

Leverage has a negative coefficient when group or year dummies are included. As

already mentioned, one possible explanation is the substitution effect between equity

issued to affiliated firms and debt. A one standard deviation increase in leverage

decreases disparity by 1.7 percentage points (= 0.0660 0.216 = 0.0173). Table 5 also

shows that even when the level of voting rights is controlled for, years-of-operation and

disparity are negatively correlated. This can happen when new firms are established as

subsidiaries of an existing company, and in this way, the controlling shareholder can

minimize the risk of entering a new business. A one-standard deviation increase in

years-of-operation decreases disparity by 4 percentage points (= 0.0437 0.9308 = 0.0406).

Group-level variables turn out to be significant when group dummies are not

included. Group size, group with financial institution dummy, and number of affiliates

are significant at 1%, 5%, and 10% levels, respectively. Disparity increase by 7.16

percentage points with a one-standard deviation increase in group size (= 0.0277 2.5839

= 0.0716). Disparity increases by 1.74 percentage points with a one-standard deviation

increase in the number of affiliates (= 0.0221 0.7871 = 0.0174). Lastly, disparity

increases by 4.61 percentage points when a chaebol group acquires or establishes a financial

institution.

In Table 6, we conduct robustness checks. We estimate equation (5) in Table 5 with

three different measures of profitability (EBIT, ordinary income, and net income) for each

year in the sample period (1997-2002). In Panel A, we use a one-year measure of

28
A change of profitability, from worst to best, decreases disparity by 13.4 percentage points.

- 25 -
profitability and in Panel B we use a three-year average. The cells are shaded if the

coefficient on profitability is significant at the 5% level. Three observations can be made.

First, the coefficient on profitability tends to be largest when net income is used. The last

column in Panel A shows that the coefficients are -0.1310 for the EBIT/Asset measure, -

0.1739 for OI/Assets, and -0.1919 for NI/Assets. Second, the coefficient on profitability

tends to be greater when we use the three-year average. The coefficients on NI/Assets

are -0.1919 when a single-year measure is used (last column in Panel A), and -0.2402 when

a three-year average is used (last column in Panel B). This indicates that the controlling

shareholders tend to respond to profitability with a lag. Third, as expected, the

coefficients tend to be larger in 1999 and 2000, when the upper ceiling on equity

investment was absent. In 1999, the coefficient on a single-year OI/Assets reaches -0.3692.

This means that a one-standard deviation increase in OI/Assets decreases the level of

disparity by 4.7 percentage points (= 0.1279 0.3692 = 0.0472).29

Notice that in Tables 5 and 6, we do not include industry dummies in our regressions.

When we did include 4-digit industry dummies, the results of which we do not report here,

the coefficient on profitability and its t-value slightly increased. We also tried a

regression model using the first differences to verify that our result is not entirely from

cross-sectional variations. The results show that our main finding is still preserved. The

coefficient on the first difference of profitability and its t-value turn out to be -0.0734 and -

1.84, respectively. This finding is actually very surprising given that our disparity

variable changes very slowly over time.

Table 7 reports OLS regression results of disparity on firm volatility with control

variables. Beta is used as our measure of volatility. We use the same set of control

29
This figure is similar to that estimated by Chang (2003), which shows that a one standard deviation increase
in profitability increases family portion of inside ownership by 5 percentage points.
30
A one-standard deviation increase in the index translates into a 9.84 percentage point increase in the cash
flow right.

- 26 -
variables as in Table 6 and 7. The table shows that the coefficients on beta are not

significant for most of the specifications. Although the coefficient is negative and

significant at the 5% level in equation (2), the magnitude is economically meaningless. A

one-standard deviation increase in beta decreases the level of disparity by 0.7 percentage

points (= 0.463 0.015 = 0.007). In year-by-year regressions, the result of which we do

not report here, we find that the weak link between firm risk and disparity remains intact.

Also, our result remains the same even when a three-year average beta is used, instead of a

single-year measure.

One possible explanation can be the control potential argument suggested by Demsetz

and Lehn (1985), which predicts that firms with high volatility tend to have high

ownership concentration. According to this argument, there is a certain wealth gain

achievable through more effective monitoring of managerial performance by a firms

owner. This is particularly so when the market for corporate control does not exist, as is

the case in Korea. When a firm operates in an uncertain environment, it becomes more

costly to monitor managerial performance. Under this situation, greater ownership

concentration becomes an effective substitute to market for corporate control. This

substitution effect may have offset the risk minimization effect, which we originally had in

mind.

5.3. Contribution to Group Control and Choice of Cash Flow Right

In Table 8 Panel A, contribution to group control indices are grouped in quintiles.

Before grouping into quintiles, observations are dropped if the index values are zero. For

each quintile, we show the level of cash flow rights. The last column shows the results of

difference-in-mean tests we conduct between the 1st and the 5th quintiles. When using

contribution index 1, the 1st quintile (small index value) shows average cash flow rights of 8

- 27 -
percent, while the 5th quintile (large index value) shows average cash flow rights of 47

percent. The difference between the two, 38 percent, is statistically significant at the 1

percent level. When contribution index 2 is used, the difference is somewhat smaller.

The 1st quintile (small index) shows average cash flow rights of 17 percent, while the 5th

quintile (large index) shows average cash flow rights of 30 percent. The difference of 13

percentage points is statistically significant at the 1 percent level.

Panels B and C show OLS regression results of cash flow right on contribution indices 1

and 2 with control variables. Panel B uses cash flow right to compute the contribution

index, while Panel C uses voting right to compute the index. We use the same set of

control variables as in the previous regressions, except that we replace voting right with

profitability. We include profitability in the regression since we know from Tables 5 and

6 that profitability is a non-trivial factor determining the level of disparity (or equivalently

the cash-flow right). We exclude voting right since here we are not interested in the

relative disparity between cash flow rights and voting rights. In this subsection, instead

we are interested in the absolute level of cash flow rights.

The tables show that our measures of contribution are not only statistically significant,

but also economically meaningful. In Panel B equation (5), the coefficient on contribution

index 1 is 0.1683 and statistically significant at 1 percent level. A one-standard deviation

increase in the index increase cash flow right by 7.11 percentage points (= 0.4227 0.1683 =

0.07114). It should also be noted that the coefficient on profitability also remains to be

significant. The impact of contribution index 2, however, turns out to be relatively modest.

In Panel C equation (5), the coefficient on contribution index 2 is 0.0689 and statistically

significant at 1 percent level. A one-standard deviation increase in the index increase

cash flow right by 3.55 percentage points (= 0.5159 0.0689 = 0.0355). These results do

not change when we drop observations with zero index values.

Table 9 conducts some robustness checks. It reports OLS regressions results for Table 8,

- 28 -
equation (5) for each year during the sample period (1997-2002). The coefficients on

contribution indices 1 and 2 are always positive and significant at 1% level. An interesting

observation is that the coefficient on contribution index 1 peaks in 1997 at 0.2328 and

gradually decreases.30 On the other hand, the coefficient on contribution index 2 increases

over time and peaks in 2002 at 0.0930.31 One possible explanation may be that, in recent

years, controlling shareholders in Korea are increasingly less concerned about total cash

flow rights (index 1), but more concerned about direct voting rights (index 2) in other

companies. This explanation is consistent with the recent changes in the Korean capital

market. One key change is the significant growth of foreign ownership during the

sample period.32 Consequently, concerns over potential hostile takeovers or institutional

shareholder activism may have triggered the controlling shareholders to tighten their

group control by increasing their cash flow rights in companies with high levels of index 2.

5.4. Sub-Sample Results: Listed versus Non-Listed

In Table 10, OLS regressions of cash flow rights are estimated with all three

determinants of our interest: profitability, risk, and contribution to group control. To

preserve sample size and to make comparison between listed and non-listed firms possible,

we use the 5-year standard deviation of (EBIT/Assets) as our measure of risk instead of

beta. Equation (1) shows the result for our full sample and equations (2) and (3) show

results for sub-samples of non-listed firms and listed firms, respectively. Voting right is

included in all three equations as an extra control variable. Public company dummy is

naturally dropped in equations (2) and (3).

As can be seen from equation (1), when all three determinants are in the regression

31
A one-standard deviation increase in the index translates into a 4.8 percentage point increase in the cash
flow right.
32
Foreign ownership in KSE was above 40 percent as of January 2004.

- 29 -
equation, contribution to group control and profitability remain to be statistically

significant, while risk is not significant. The coefficient on profitability is slightly higher

than that found in Table 5 equation (5) and the coefficient on Contribution Index 1 is slightly

lower than that found in Table 8 Panel B Equation (5).

Comparison between equations (2) and (3) gives us three observations. First,

profitability explains cash flow right in non-listed firms, but not in listed firms. This is

intuitive given that the ownership structure of public firms is more difficult to change than

that of private firms. Second, firm risk has explanatory power neither in the public firms

nor in the private firms. Third, contribution to group control has strong explanatory

power in both types of firms, but the coefficient is stronger in case of listed firms. The

coefficients are 0.1752 in public firms and 0.1238 in private firms. One explanation is that

the threat of hostile acquisition is present in public firm, but absent in private firms. Thus,

given everything else equal, there exists a stronger control motive in public firms.

6. Concluding Remarks

Many believe that corporate ownership structure evolves slowly over time. However,

most academic works also treat ownership as exogenous as if they do not support such

view. But, when it comes to group-affiliated firms with a common controlling

shareholder, the reality can be different. In this paper, we provide empirical evidence

that the ownership structure of a business conglomerate can be deliberately shaped by its

controlling shareholder. By using an exclusive data set of 46 chaebol groups on their intra-

group shareholdings in Korea during 1997-2002, we find that controlling shareholders

concentrate their cash flow rights in firms that serve as de facto holding companies those

with the greatest contribution to group control and those with high profitability.

The strong group control motive by the controlling shareholder found in this paper

- 30 -
has a number of implications. First, it suggests that the controlling shareholders of

chaebol groups see group-affiliated firms, including financial institutions, not only as

profit-making entities, but also as means to control other group-affiliated firms. Such

motive can be especially strong in case of financial institutions, which are able to control

many other firms with only a small amount of book equity. Second, our finding implies

that there still might be significant private gains from group control. Otherwise, the

controlling shareholder would not structure ownership in a way that maximizes his group

control. This is consistent with the recent finding by Dyck and Zingales (2004), which

reports premiums from controlling block transactions. According to this paper, the

median value of block premium in Korea is 17 percent, which is above the 39-country

mean of 11 percent. Recent corporate governance scandals in Korean chaebols also show

that the controlling shareholders even engage in illegal transactions to preserve their

group control, which further supports our reasoning that private gains can be significant.

- 31 -
 References

Bebchuk, Lucian Aye, Reinier Kraakman, and George Triants (2000), Stock Pyramids, Cross-
Ownership and Dual Class Equity: The Mechanisms and Agency Costs of Separating Control
From Cash-Flow Rights," in Randall K. Morck, ed., Concentrated Corporate Ownership, pp. 295-
318.

Baek, Jae-Seung., Kang Jun-Koo, and Park Kyung Suh (2004), "Corporate Governance and Firm
Value: Evidence from the Korean Financial Crisis," Journal of Financial Economics Vol.71, pp. 265-
313

Black, Bernard S, Hasung Jang, and Woochan Kim (2004), "Does Corporate Governance Affect
Firms' Market Values? Evidence from Korea, John M. Olin Program on Law and Economics,
Working Paper No. 237

Chang, Sea Jin (2003), Ownership Structure, Expropriation, and Performance of Group-Affiliated
Companies in Korea, Academy of Management Journal Vol.46 No.2, pp.238-253

Cho, Myeong-Hyeon (1998), Ownership Structure, Investment, and the Corporate Value: An
Empirical Analysis, Journal of Financial Economics 47, pp.103-121

Claessens, Stijn, Djankov Simeon, and Larry H. P. Lang (2000), "The Separation of Ownership and
Control in East Asian Corporations," Journal of Financial Economics 58, pp. 81-112.

Claessens, Stijn, Djankov Simeon, Joseph P. H. Fan, and Larry H. P. Lang (2002), "Disentangling the
Incentive and Entrenchment Effects of Large Shareholdings," The Journal of Finance 57, pp. 2741-
2771

Demsetz, Harold and Kenneth Lehn (1985), The Structure of Corporate Ownership: Causes and
Consequences, The Journal of Political Economy, Vol.93 No.6, pp.1155-1177

Dyck, Alexander and Luigi Zingales (2004), Private Benefits of Control: An International
Comparison, The Journal of Finance. 59, pp. 537-599

Fama, Eugene and Michael Jensen (1983), Separation of Ownership and Control, Journal of Law
and Economics, 26: 301-325

Fan, J.P.H. and T.J. Wong (2002), Corporate Ownership Structure and the Informativeness of

- 32 -
 Accounting Earnings in East Asia Journal of Accounting and Economics 33, pp. 401-425

Haw, In-Mu, Bingbing Hu, Lee-Seok Hwang and Woody Wu (2003), Ultimate Ownership, Income
Management, and Legal and Extra-Legal Institutions, manuscript

Joh, Sung Wook (2003), "Corporate Governance and Firm Profitability," Journal of Financial
Economics 68, pp. 287-322.

Kim, Jin Bang (2000), Ownership Structure of Chaebols: Statistics, Concepts, and Analysis,
Kyungaehak YeunkooVol.48 No.2, pp. 57-93

La Porta, Rafael, Florencio Lopez-de-Silanes, and Andrei Shleifer A. (1999), "Corporate Ownership
around the World," Journal of Finance 54, pp. 471-518.

La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny (2002), Investor
Protection and Corporate Valuation, The Journal of Finance 57, pp. 1147-1170

Lemmon, Michael L. and Karl V. Lins (2003), "Ownership Structure, Corporate Governance, and
Firm Value: Evidence from East Asian Financial Crisis," The Journal of Finance 58, pp. 1445-1468

Lins, Karl (2003), Equity Ownership and Firm Value in Emerging Markets, Journal of Financial
and Quantitative Analysis, pp.159-184

Mitton, Tod (2002), "A Cross-Firm Analysis of the Impact of Corporate Governance on the East
Asian Financial Crisis," Journal of Financial Economics 64, pp. 215-241

Morck, Randall, Andrei Shleifer, and Robert W. Vishny (1988), Management Ownership and
Market Valuation: An Empirical Analysis, Journal of Financial Economics 20, pp.293-315

- 33 -
 Appendix 1:
Computation of Disparity: Distinctions from the Existing Literature

Claessens et al. (2000) LLSV (2002) Lins (2003) This Paper

Non-listed companies Ownership data not Ownership data not Ownership data not Ownership data available
available available available for all non-listed firms
under the controlling
shareholders influence

Listed companies Ownership data available Ownership data available Ownership data available Ownership data available
for only three-quarters of for only largest 20 firms in for only half of listed firms for all listed firms under the
listed firms in each country each country by market in each country by market controlling shareholders
by market capitalization capitalization capitalization influence

Financial institutions Included in the sample Dropped from the sample Dropped from the sample Included in the sample

Shareholders Only block holders with - Only block holders with All shareholders are
equity stakes at or above a equity stakes at or above a considered regardless of the
5% threshold are considered 5% threshold are considered size of its equity stake

Distinction among No distinction made - - No distinction made

individual family members

Not-for-profit organizations Considered as shareholders Considered as shareholders Considered as shareholders Considered as shareholders

Controlling Shareholder A shareholder that A shareholder with the A shareholder that A person who, alone or
ultimately owns the direct largest control right among ultimately owns the direct with related parties, has de
block holding; if there are those with at least 10% block holding, and is a facto control of the company
multiple ultimate owners, control right (defined member of the management (also known as the same
the one with the largest below) group (mangers and their person in the Monopoly
control rights is chosen families of the company Regulation and Fair Trade
(defined below) concerned) Act)

Related Parties Not considered Not considered Companies managed by the Spouse and relatives; not-
members of the for-profit organizations
management group where the same person, alone
or with related parties,
contributed 30% of its total

- 34 -
 donation; not-for-profit
organizations where the
same person, directly or
through its related parities,
has a controlling influence
over the appointment of
directors or its business
activities; any company
whose business is de facto
controlled by the same
person; agents of the same
person or its related parities

De facto controlled Not considered Not considered Not considered See Appendix 2
companies

Voting Rights Sum of the weakest links in Controlling shareholders Sum of direct block Sum of direct ownership
the chains of voting rights direct and indirect voting holdings ultimately held by held by the controlling
rights in the firm; indirect the management group and shareholder and its related
voting right over the firm its related parities parties
concerned is x percent if a
sequence of firms leading to
this firm forms a control
chain (e.g. each of which
has control over the next
one), and the last firm in the
chain directly controls x
percent of the voting rights

Cash-flow rights Sum of the products of the Sum of the products of the Sum of the products of the Sum of the products of the
ownership stakes held by ownership stakes held by ownership stakes held by ownership stakes held by
the controlling shareholder the controlling shareholder the management group the controlling shareholder
along the chains of voting along the chains of voting along the chains of voting along the chains of voting
rights rights rights rights

Disparity Formula Cash-flow Rights / Voting Voting Rights Voting Rights / Voting Rights
Rights Cash-flow Rights Cash-flow Rights Cash-flow Rights

- 35 -
 Appendix 2:
The Concept of Related Parties and de Facto Control

Business Conglomerates

According to the Monopoly Regulation and Fair Trade Act, a business conglomerate is a
group of companies whose businesses are controlled de facto by the same person pursuant
to the standards prescribed by the Presidential Decree.33

There can be two different types of business conglomerates:

(a) Where the same person is a company, a business conglomerate is a group composed of
said company and one or more companies over which the same person holds de facto
control; and
(b) Where the same person is not a company, a business conglomerate is a group composed
of two or more companies controlled de facto by the same person.

The Concept of Control

The Presidential Decree defines companies whose businesses are controlled de facto by the
same person in the following two ways:

(a) A company where the same person, alone or with its related parities, owns 30% of
voting shares issued, and where the same person is the largest shareholder. Related
parities can refer to the following:

Spouse and relatives (if the relative has a blood relationship, the degree of kinship
must be eight or less; if the relative has a blood relationship with the spouse, the
degree of kinship must be four or less)
Not-for-profit organization where the same person, alone or with its related parities,
contributed 30% of its total donation. The same person must be the largest donor,
or either the same person or anyone among the related parities must be the founder.
Not-for-profit organization where the same person, directly or through his related
parties, has a controlling influence over the appointment of directors or its
business activities.
Any company whose business is controlled de facto by the same person according to
(a) and (b).
Any agent of the same person or its related parties.

(b) The same person is considered to have de facto control of the following companies even

33
KFTC does not officially use the term chaebol. Instead, they use the term business conglomerates.

- 36 -
 without any share ownership.

A company whose representative director or at least half of whose directors are

appointed or can be appointed by the same person via a contract with the major
shareholders or by a mutual agreement.34
A company where the same person, directly or through its related parities, has a
controlling influence over its major decisions such as organizational change or
new business entry, or its business decisions.
A company that participates in any of the following personnel exchanges with a
company that is controlled de facto by the same person (includes the same person if it
is a company):
i. A company, the director of which has a joint appointment at a company that
is controlled de facto by the same person.
ii. A company, the director or the employee of which was previously employed
by a company controlled de facto by the same person, later employed by the
company concerned, and lastly returned to the same company he/she was
previously employed or to another company controlled de facto by the same
person.
iii. A company in which the director or employee of which was previously
employed by the company concerned, later employed by the company
controlled de facto by the same person, and lastly returned to the company
concerned or to its affiliated company.
A company that has transactions of funds, assets, goods, services, or debt
guarantees with the same person or with its related parties above a normal range; a
company that can be recognized as an affiliated company of the business group
controlled by the same person according to social norms (e.g. using similar
trademarks).

34
A representative director refers to senior directors including the CEO.

- 37 -
 Appendix 3:
Regulation on Large Business Conglomerates in Korea

Since 1987, the Korea Fair Trade Commission (KFTC) has been categorizing, each year in April,
certain large business conglomerates to be subject to a number of restrictions, which we explain in
the table below. From 1987 to 2001, KFTC designated the top 30 conglomerates in terms of their
total asset size. Since 2002, KFTC changed the way it designates the conglomerates. Instead of
using asset size ranks, it uses asset size thresholds. That is, KFTC regulations are imposed only
when the total asset size of a conglomerate is above certain asset size thresholds (e.g. 2 trillion or 5
trillion Korean won). Following is the list of regulations imposed on each of the affiliated firms
comprising the large business conglomerates designated by the KFTC. Dates are effective dates,
not the dates on which the relevant bills were passed.

Regulations Description
Ban on cross- Affiliated firms in the designated large business conglomerates
shareholdings cannot have cross shareholdings with other affiliated firms in the
same conglomerate (cross-shareholding refers to firm A holding
shares of firm B, and firm B holding shares of firm A; circular
shareholding is allowed)

! (April 1987-March 1991) Applied to all the firms in top 30

conglomerates, with the exception of financial institutions
! (April 1991-March 2002) Applied to all the firms in top 30
conglomerates, including financial institutions
! (April 2002-Present) Applied to all the firms in conglomerates
above 2 trillion won

Upper ceiling on equity Affiliated firms in the designated large business conglomerates can
investment make equity investments in other domestic companies in amounts
only up to 25% of net assets (= assets book equity invested by other
affiliates)

! (April 1987-March 1990) 40% upper ceiling applied to all the firms
in the top 30 conglomerates
! (April 1990-Dec.1994) 40% upper ceiling applied to all the firms in
the top 30 conglomerates, with the exception of financial
institutions
! (Dec.1994-Feb.1998) 25% upper ceiling applied to all the firms in
the top 30 conglomerates, with the exception of financial
institutions
! (Feb.1998-March 2001) No upper ceiling (regulation lifted to
facilitate corporate restructuring)
! (April 2001-Jan 2002) 25% upper ceiling applied to all the firms in
top 30 conglomerates, with the exception of financial institutions
(exemptions allowed on certain conditions)
! (Jan 2002-Present) Limit voting rights on shares above the 25%
upper ceiling applied to all the firms in conglomerates above 5
trillion won, with the exception of financial institutions

- 38 -
 (exemptions allowed on a variety of reasons)

Ban on new debt Affiliated firms in the designated large business conglomerates
guarantees cannot provide any new debt guarantees to domestic affiliates

! (April 1993-March 1996) Debt guarantee cannot be more than 200%

of book equity (if above the upper ceiling should reduce it by
March 1996)
! (April 1997-March 1999) Debt guarantee cannot be more than 100%
of book equity (if above the upper ceiling, should be reduced by
March 1998)
! (April 1999-Present) Ban on new debt guarantees

No voting rights for Financial institutions in the designated large business conglomerates
financial institutions on cannot exercise their voting rights on shares issued by their affiliated
shares issued by affiliated firms
firms
! (April 1993-Jan.2002) Applied to all the financial institutions in top
30 business conglomerates
! (Jan.2002-Present) Voting rights of financial institutions allowed up
to 30% of shares issued by an affiliated public firm on voting items
such as revision of AOI, appointment/removal of directors, and
mergers(the controlling shareholder cannot directly or indirectly
exercise his/her voting rights above 30%)

Board approval and Related party transactions above 10 billion won or 10% of book equity
disclosure of related party should be approved by the board and be disclosed to the public
transactions
! (April 2000-March 2001) Applied to all the firms in top 10 business
conglomerates
! (April 2001-March 2002) Applied to all the firms in top 30 business
conglomerates
! (April 2002-Present) Applied to all the firms in business
conglomerates above 2 trillion won

- 39 -
 Table 1: Intra-Group Shareholding Matrix of Samsung Group in 2002

The complex intra-group shareholding structure of Samsung Group can be effectively presented in a matrix format. For the
convenience of presentation, among the 63 firms classified by KFTC as Samsung affiliates, we show in this table only the 27 major firms
that appear in the 2002 Samsung Group Annual Report. The fractions of shares are computed out of total outstanding common shares,
including treasury stocks.

Same Person

Managers

Sub-Total
Relatives

Treasury
Stocks
NPO
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

[1] Samsung Corporation 1.42 0.01 0.23 0.18 0.42 1.96 - - - 4.66 - - - - - -
[2] Cheil Industries - - 2.58 0.33 2.91 4.97 - - - - - - - - - -
[3] Samsung Electronics 2.00 1.55 0.08 0.64 2.27 4.23 3.87 - - - - - - - - -
[4] Samsung SDI - 0.00 0.74 0.29 1.03 2.55 - - 20.01 - - - - - - -
[5] Samsung Corning - - - 0.00 0.00 - - - 48.36 - - - - - - -
[6] Samsung Electro-Mechanics - 0.12 - 0.84 0.96 0.50 - - 23.69 - - - - - - -
[7] Samsung Petrochemical - - - - 0.00 - 10.00 16.39 9.93 - - - - - - -
[8] Samsung Heavy Industries - - - 0.02 0.02 - - 0.42 17.62 - - 2.39 - - - 0.13
[9] The Shilla Hotels & Resorts - - - 0.08 0.08 3.06 - - 5.11 - - - - - - -
[10] Samsung Engineering - - - 0.28 0.28 0.03 - 13.10 - 5.09 - - - - - -
[11] Cheil Communications - - - 0.20 0.20 8.26 12.64 - 2.61 - - - - - - -
[12] Samsung Lions 2.50 - - - 0.00 - 7.50 15.00 27.50 - - 12.50 - - - -
[13] Samsung Atofina 0.44 - - 0.18 0.18 2.44 37.45 0.85 3.78 10.32 - 10.19 - - - -
[14] Samsung Economic Research Institute - - - - 0.00 - 1.00 1.00 29.80 28.60 - 23.80 - 1.00 - -
[15] Samsung Fine Chemicals - - - 0.25 0.25 - 5.59 3.16 8.39 11.49 - 0.26 - - 2.24 0.85
[16] Samsung Corning Precision Glass - - - - 0.00 - - - 42.57 - - - - - - -
[17] S1 - - - 0.01 0.01 0.07 - - - 11.03 - - - - - -
[18] Samsung Everland 3.72 50.77 0.88 - 51.65 - 1.48 4.00 - 4.00 - 4.00 - - - -
[19] Samsung SDS - 22.82 - 6.78 29.60 - 17.96 - 21.27 - - 8.29 - - - -
[20] Samsung Techwin - - - 0.03 0.03 12.75 3.85 0.10 22.93 - - - - - - -
[21] Samsung Life Insurance 4.54 4.68 4.68 2.50 11.86 - - - - - - 0.60 - - - -
[22] Samsung Fire & Marine Insurance 0.31 - 3.58 0.05 3.63 8.47 - - - - - - - - - -
[23] Samsung Card - 0.10 - - 0.10 - 9.44 - 58.59 - - 22.31 - - - -
[24] Samsung Securities 0.10 0.01 0.29 0.46 0.76 3.63 0.27 - - - - - - - - -
[25] Samsung Investment Trust Management - 17.95 - 0.38 18.33 - - - - - - - - 3.89 - -
[26] Samsung Venture Investment - - - - 0.00 - - - 16.33 16.33 - 17.00 - 17.00 - -
[27] Samsung Networks - 23.26 - 6.29 29.55 - 19.47 - 23.07 - - 8.99 - - - -

- 41 -
 [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27]

[1] Samsung Corporation - - - - - - - - - - 4.81 - - 0.00 0.01 - -

[2] Cheil Industries - - - - - - - - - - 0.01 - 4.00 0.00 - - -
[3] Samsung Electronics - - - - - - - - - - 6.94 1.21 - 0.02 - - -
[4] Samsung SDI - - - - - - - - - - 0.02 - - 0.00 0.17 - -
[5] Samsung Corning - - - - - - - - - - 1.00 - - - - - -
[6] Samsung Electro-Mechanics - - - - - - - - - - - - - 0.00 0.05 - -
[7] Samsung Petrochemical - - - - - - - - - - - - - - - - -
[8] Samsung Heavy Industries 0.13 - - - - - - 0.13 - 0.07 3.91 - - 0.00 - - -
[9] The Shilla Hotels & Resorts - - - - - - - - - - 7.30 - 0.52 3.06 - - -
[10] Samsung Engineering - - - - - - - 1.07 - - - - - - - -
[11] Cheil Communications - - - - - - - - - - - - 3.04 - - - -
[12] Samsung Lions 3.00 - - - - - - 2.00 - - - - - - - - -
[13] Samsung Atofina 0.32 - - - 3.45 - - - - 25.6 - - - - - - -
[14] Samsung Economic Research Institute - - - - - - - - - - 14.8 - - - - - -
[15] Samsung Fine Chemicals - - - - - - - - - - - - - - - - -
[16] Samsung Corning Precision Glass - - - - - - - - - - - - - - - - -
[17] S1 - - - - - - - - - - 5.34 0.97 - 1.32 - - -
[18] Samsung Everland - - - - - - - - - - - - 14.0 - - - -
[19] Samsung SDS - - - - - - - - - - - - - - - - -
[20] Samsung Techwin - - - - - - - 0.28 - - 1.21 - - 1.76 - - -
[21] Samsung Life Insurance - - - - 0.47 - - 19.3 0.35 - - - - - - - -
[22] Samsung Fire & Marine Insurance - - - - - - - - - - 9.89 - 3.15 0.00 - - -
[23] Samsung Card - - - - - - - - - - - - - - - - -
[24] Samsung Securities - - - - - - - - - - 11.5 5.47 2.62 - - - -
[25] Samsung Investment Trust Management - - - - - - - - - - 3.56 1.19 - 65.4 - - -
[26] Samsung Venture Investment - - - - - - - - - 16.7 - - - 16.7 - - -
[27] Samsung Networks - - - - - - - - - - - - - - - - -
Note: - indicates zero shares, whereas 0.00 stands for small positive figures below 1/500 (figures above 1/500 are rounded to be 0.01).

- 42 -
 Table 2: Firm-Level Disparity of Samsung Group in 2002

Among the 63 firms classified by the Fair Trade Commission as affiliated firms, we only show in
this table the 27 major firms that appear in the 2002 Samsung Group Annual Report.

Rank Name Control Right Cash-Flow Right Disparity

1 Cheil Industries 0.082 0.002 0.080
2 Samsung Corporation 0.115 0.025 0.090
3 Samsung Electronics 0.170 0.053 0.117
4 The Shilla Hotels & Resorts 0.174 0.019 0.155
5 Samsung Fire & Marine Insurance 0.202 0.027 0.175
6 Samsung Life Insurance 0.380 0.201 0.179
7 S1 Corporation 0.206 0.013 0.192
8 Cheil Communications 0.202 0.006 0.195
9 Samsung Securities 0.227 0.028 0.199
10 Samsung Engineering 0.213 0.008 0.206
11 Samsung SDI 0.218 0.011 0.207
12 Samsung Heavy Industries 0.248 0.018 0.230
13 Samsung Electro-Mechanics 0.248 0.014 0.234
14 Samsung Techwin 0.346 0.020 0.325
15 Samsung Fine Chemicals 0.353 0.009 0.344
16 Samsung Petrochemical 0.363 0.008 0.355
17 Samsung Everland 0.945 0.557 0.388
18 Samsung Corning Precision Glass 0.426 0.023 0.403
19 Samsung Corning 0.494 0.027 0.466
20 Samsung SDS 0.771 0.245 0.526
21 Samsung Networks 0.811 0.251 0.560
22 Samsung Lions 0.700 0.055 0.645
23 Samsung Investment Trust Management 0.924 0.206 0.718
24 Samsung Card 0.884 0.036 0.848
25 Samsung Atofina 0.949 0.025 0.925
26 Samsung Economic Research Institute 1.000 0.053 0.947
27 Samsung Venture Investment 1.000 0.024 0.976

- 43 -
 Table 3: Definition of Other Variables

All accounting measures are from the National Information and Credit Evaluation, Inc. (NICE).
Since disparity is used as a dependent variable in our regression analyses, we make sure that other
firm-level variables are measured prior to the disparity variable. As such, they are measured
during or on the last day of the fiscal year, which ends before April. When the fiscal year changes
during the sample year, we only keep those years which cover a full 12 months.

Variable Name Definition

EBIT/Assets EBIT divided by book value of assets (measured at previous fiscal
year end), winsorized at the 1st and the 99th percentile values.
OI / Assets Ordinary income divided by book value of asset (measured at
previous fiscal year end), winsorized at the 1st and the 99th percentile
values. Ordinary income is earnings before taxes and
extraordinary items.
NI / Assets Net income divided by book value of assets (measured at previous
fiscal year end), winsorized at the 1st and the 99th percentile values.
Beta Beta is estimated from a market model where the KOSPI return is
used as a proxy for market return. KOSPI is a value-weighted
market index comprising all the listed companies on the Korea Stock
Exchange (KSE). For each year, we estimate Beta using monthly
data over the past three-years. Note that Beta is replaced by a
missing value if the underlying regression uses less than 20
observations. Beta is winsorized at the 1st and the 99th percentile
values.
SD (EBIT/Assets) Standard deviation of (EBIT/Assets) over the past 5-year period. It
is winsorized at the 1st and the 99th percentile values.
Contribution Index 1 Amount of additional cash flow rights a controlling shareholder
would gain in other companies by having company j under her
control as a fraction of the company js book equity value. See
Section 4-A, Equation (5). The index is winsorized at the 1st and the
99th percentile values.
Contribution Index 2 Amount of additional voting right a controlling shareholder would
gain in other companies by having company j under her control as a
fraction of the company js book equity value. See Section 4-A,
Equation (6). The index is winsorized at the 1st and the 99th
percentile values.
Leverage Book value of debt divided by the book value of assets. This
measure is first winsorized at the 1st and the 99th percentile values,
and then ln[(debt/asset)+1] is computed. 1 is added since there
0 values, even after winsorization.
Years of Operation Number of years since the companys establishment. This measure
is logged.
Public Company 1 if the company is listed either in KSE or KOSDAQ; 0 otherwise
Firm Size Book value asset minus book value of debt (unit: billion won).
Negative book values are treated as missing values. This measure

- 44 -
 is logged after adding 1. This is to make this measure non-
negative after taking the log.
Financial Institution 1 if the company is a financial institution; 0 otherwise
Group Size Book value of equity summed across all affiliated firms (unit: billion
won). This measure is logged after adding 1. This is to make
this measure non-negative after taking the log.
Number of Affiliates Number of firms controlled by a chaebol the company is affiliated to.
This measure is logged.
Group with Fin. Inst. 1 if the company is affiliated to a group with a financial institution; 0
otherwise

- 45 -
 Table 4: Summary Statistics

Panel A: Summary Statistics

Variable Name # of Obs. Mean Median Std. Dev. Min Max
Voting Right (All) 5,202 0.6830 0.7459 0.3079 0.0000 1.0000
Cash Flow Right (All) 5,202 0.2144 0.1295 0.2436 0.0000 1.0000
Disparity (All) 5,202 0.4685 0.4504 0.3121 0.0000 1.0000
Voting Right (Private) 3,542 0.7591 0.9000 0.2816 0.0000 1.0000
Cash Flow Right (Private) 3,542 0.2398 0.1405 0.2690 0.0000 1.0000
Disparity (Private) 3,542 0.5192 0.5069 0.3109 0.0000 1.0000
Voting Right (Public) 1,660 0.5206 0.4779 0.2986 0.0000 1.0000
Cash Flow Right (Public) 1,660 0.1603 0.1121 0.1653 0.0000 1.0000
Disparity (Public) 1,660 0.3603 0.2783 0.2862 0.0000 1.0000
1yr (EBIT/Asset) 3,743 0.0476 0.0497 0.1091 -0.3394 0.4288
1yr (OI/Asset) 3,771 0.0065 0.0124 0.1332 -0.5350 0.4292
1yr (NI/Asset) 3,772 -0.0070 0.0084 0.1301 -0.5884 0.3595
3yr (EBIT/Asset) 4,088 0.0458 0.0493 0.1009 -0.3394 0.4288
3yr (OI/Asset) 4,121 0.0070 0.0122 0.1193 -0.5350 0.4292
3yr (NI/Asset) 4,123 -0.0062 0.0075 0.1131 -0.5884 0.3595
1yr Beta 796 1.0761 1.0470 0.4436 -0.0345 2.2510
3yr Beta 895 1.0480 1.0337 0.3826 -0.0345 2.2428
5yr SD(EBIT/Asset) 3,818 0.0615 0.0360 0.0794 0.0022 0.5330
Contribution Index 1 4,114 0.0971 0.0000 0.4227 0.0000 3.4467
Contribution Index 2 4,114 0.1982 0.0000 0.5159 0.0000 3.5684
Leverage 4,158 0.5392 0.5511 0.2221 0.0000 3.4568
Years of Operation 3,745 2.6921 2.7726 0.8373 0.6931 4.4188
Public Company 5,202 0.3191 0.0000 0.4662 0.0000 1.0000
Firm Size 4,466 2.6644 2.4925 2.3359 0.0000 10.0987
Financial Institution 5,202 0.1292 0.0000 0.3354 0.0000 1.0000
Group Size 5,202 7.1837 7.6057 2.6817 0.0013 10.9098
Number of Affiliates 5,202 3.4428 3.3673 0.5846 1.7918 4.4543
Group with Fin. Inst. 5,202 0.9208 1.0000 0.2701 0.0000 1.0000

Panel B: Comparison with the Existing Literature (Using Average Figures)

Voting Right Voting Right /
Coverage Period
Cash Flow Right Cash Flow Right
This Paper (All) 0.47 3.19 All Chaebols Firms 1997-2002
This Paper (Private) 0.52 3.17 Private Chaebol Firms 1997-2002
This Paper (Public) 0.36 3.25 Public Chaebol Firms 1997-2002
Claessens et al. (2000) 0.04 1.27 211 Korean Public Firms 1997
Chang (2003) - 1.47 Public Chaebol Firms 1986-1996
Lemmon and Lins (2003) - 2.37 188 Korean Public Firms 1997
Joh (2003) 0.23 - All Korean Firms 1997

- 46 -
 Table 5: Disparity and Profitability

OLS regressions of disparity on profitability (measured as EBIT over assets) with additional control
variables are shown in Panel A and B. The extreme values of (EBIT/Assets) are winsorized. That
is, as for the observations outside the 1st and the 99th percentiles, the (EBIT/Assets) values are
replaced by the 1st and the 99th percentile values. In Panel A, we use one-year past (EBIT/Asset)
and, in Panel B, we use three-year average of past (EBIT/Asset). t-values, based on Whites
heteroskedasticity-consistent standard errors, are reported in parentheses. *, **, and ***, indicate
significance at 10, 5, and 1% levels, respectively.

Panel A: One-Year Profitability

Disparity
(1) (2) (3) (4) (5)
1yr (EBIT/Assets) -0.1224*** -0.1009*** -0.1314*** -0.1426*** -0.1310***
(3.41) (3.26) (3.44) (3.46) (3.42)
Voting Rights 0.7037*** 0.7570*** 0.7251*** 0.6825*** 0.7254***
(62.69) (78.44) (52.92) (46.71) (52.74)
Leverage -0.0662** -0.0180 -0.0660**
(2.54) (0.63) (2.53)
Years of Operation -0.0434*** -0.0458*** -0.0437***
(6.61) (7.09) (6.62)
Public Company -0.0040 0.0051 -0.0040
(0.50) (0.58) (0.50)
Firm Size 0.0037 0.0005 0.0039
(1.50) (0.20) (1.54)
Financial Institution 0.0133 0.0121 0.0136
(1.52) (1.23) (1.55)
Group Size 0.0277*** -0.0054
(5.57) (0.56)
Number of Affiliates 0.0221* 0.0232
(1.89) (0.70)
Group w/ Fin. Inst. 0.0461** -0.0269
(2.34) (0.69)
Group Dummies No Yes Yes No Yes
Year Dummies No Yes Yes No Yes
Observations 3743 3743 2715 2715 2715
Adjusted R-squared 0.5160 0.6632 0.6778 0.5937 0.6776

- 47 -
Panel B: Three-Year Profitability
Disparity
(1) (2) (3) (4) (5)
3yr (EBIT/Asset) -0.1634*** -0.1241*** -0.1750*** -0.1941*** -0.1747***
(4.26) (4.03) (4.44) (4.36) (4.43)
Voting Right 0.7022*** 0.7601*** 0.7281*** 0.6854*** 0.7285***
(65.92) (82.87) (54.25) (47.90) (54.10)
Leverage -0.0686*** -0.0186 -0.0682***
(2.69) (0.66) (2.67)
Years of Operation -0.0414*** -0.0442*** -0.0418***
(6.35) (6.86) (6.38)
Public Company -0.0057 0.0035 -0.0058
(0.72) (0.40) (0.73)
Firm Size 0.0034 0.0001 0.0037
(1.42) (0.03) (1.50)
Financial Institution 0.0084 0.0065 0.0088
(0.97) (0.66) (1.01)
Group Size 0.0270*** -0.0081
(5.45) (0.84)
Number of Affiliates 0.0233** 0.0293
(2.01) (0.90)
Group w/ Fin. Inst. 0.0472** -0.0299
(2.41) (0.77)
Group Dummies No Yes Yes No Yes
Year Dummies No Yes Yes No Yes
Observations 4088 4088 2776 2776 2776
Adjusted R-squared 0.5114 0.6666 0.6806 0.5969 0.6804

- 48 -
 Table 6: Disparity and Profitability: Robustness Check

OLS regression of Table 6 equation (5) is estimated with three different measures of profitability
(EBIT, ordinary income, and net income) for each year in the sample period. In Panel A, we use a
one-year measure of profitability, and in Panel B we use a three-year average. The cells are
shaded if the profitability coefficient is significant at the 5% level. t-values, based on Whites
heteroskedasticity-consistent standard errors, are reported in parentheses. *, **, and ***, indicate
significance at 10, 5, and 1% levels, respectively.

Panel A: One-Year Profitability

Disparity
1996 1997 1998 1999 2000 2001 2002 All Years
EBIT / Asset -0.1413 -0.2831** 0.0665 -0.1340 -0.2782*** -0.1859** -0.0475 -0.1310***
(1.10) (1.97) (0.65) (1.17) (3.40) (2.47) (0.57) (3.42)
Obs. 376 469 433 385 326 336 382 2715
Adj. R2 0.6902 0.6441 0.6167 0.6651 0.7159 0.7540 0.7212 0.6776
OI / Asset 0.0017 -0.2895** -0.0996 -0.3692*** -0.2613*** -0.2146*** -0.0239 -0.1739***
(0.02) (2.23) (0.97) (3.97) (2.97) (2.65) (0.35) (4.92)
Obs. 378 473 439 391 328 337 382 2736
Adj. R2 0.6859 0.6506 0.6078 0.6805 0.7185 0.7564 0.7210 0.6791
NI / Asset -0.0528 -0.3124** -0.1785** -0.3235*** -0.2197** -0.2346*** -0.0196 -0.1919***
(0.45) (2.23) (2.04) (3.95) (2.35) (2.64) (0.25) (5.50)
Obs. 378 474 439 391 328 337 382 2737
Adj. R2 0.6862 0.6372 0.6106 0.6778 0.7145 0.7559 0.7209 0.6771

Panel B: Three-Year Profitability

Disparity
1996 1997 1998 1999 2000 2001 2002 All Years
EBIT / Asset -0.0930 -0.3616*** 0.0248 -0.1182 -0.2921*** -0.2694*** -0.1909** -0.1747***
(0.95) (2.71) (0.22) (1.10) (2.66) (3.05) (2.20) (4.43)
Obs. 386 488 444 395 334 339 382 2776
Adj. R2 0.6925 0.6454 0.6236 0.6673 0.7159 0.7596 0.7254 0.6804
OI / Asset 0.0260 -0.2244** -0.0450 -0.3564*** -0.3295*** -0.2771*** -0.1277 -0.1970***
(0.25) (2.00) (0.48) (3.81) (3.24) (3.18) (1.50) (5.36)
Obs. 387 493 451 402 336 341 382 2800
Adj. R2 0.6919 0.6478 0.6172 0.6795 0.7217 0.7613 0.7235 0.6822
NI / Asset -0.0436 -0.2808** -0.1254 -0.3627*** -0.3592*** -0.3224*** -0.1185 -0.2402***
(0.39) (2.32) (1.36) (4.03) (3.25) (3.28) (1.19) (6.18)
Obs. 387 494 451 402 336 341 382 2801
Adj. R2 0.6919 0.6361 0.6184 0.6799 0.7209 0.7616 0.7227 0.6808

- 49 -
 Table 7: Disparity and Volatility (Beta)

OLS regressions of volatility (measured by beta) with additional control variables are shown in this
table. Betas are estimated using a market model with KOSPI return as the market return, and with
monthly frequency. KOSPI is a value-weighted market index comprising all the listed companies
in the Korea Stock Exchange (KSE). The sample includes only those companies listed in KSE. t-
values, based on Whites heteroskedasticity-consistent standard errors, are reported in parentheses.
*, **, and ***, indicate significance at 10, 5, and 1% levels, respectively.

Disparity
(1) (2) (3) (4) (5)
Beta 0.0110 -0.0153** -0.0110 0.0080 -0.0126
(1.26) (2.34) (1.45) (0.88) (1.63)
Voting Right 0.6323*** 0.8124*** 0.7809*** 0.6017*** 0.7837***
(21.48) (34.73) (31.73) (21.61) (32.15)
Leverage -0.0412 -0.0135 -0.0445
(1.02) (0.33) (1.09)
Years of Operation -0.0469*** -0.0629*** -0.0481***
(4.75) (6.12) (4.86)
Firm Size -0.0044* -0.0112*** -0.0034
(1.67) (3.73) (1.27)
Financial Institution -0.0107 -0.0215** -0.0095
(1.29) (2.07) (1.11)
Group Size 0.0320*** -0.0181**
(6.11) (2.09)
Number of Affiliates 0.0109 0.0290
(0.91) (0.87)
Group w/ Fin. Inst. 0.0202 -0.0107
(1.06) (0.29)
Group Dummies No Yes Yes No Yes
Year Dummies No Yes Yes No Yes
Observations 790 790 729 729 729
Adjusted R-squared 0.4760 0.7193 0.7313 0.5875 0.7315

- 50 -
 Table 8: Contribution to Group Control and Cash Flow Rights

In Panel A, contribution to group control indices are grouped in quintiles. Before being grouped
into quintiles, observations are dropped if the index values are zero. For each quintile, we show
the level of cash flow rights. In the last column, we conduct difference-in-mean tests between the
1st and the 5th quintiles. In Panel B and C, OLS regressions of cash flow right on contribution index
with additional control variables are shown in this table. t-values, based on Whites
heteroskedasticity-consistent standard errors, are reported in parentheses. *, **, and ***, indicate
significance at 10, 5, and 1% levels, respectively.

Panel A: Difference-in-Mean Test

Q1 (Small) Q2 Q3 Q4 Q5 (Large) Q5 Q1
Contribution Index 1 0.08 0.13 0.19 0.22 0.47 0.384***
Contribution Index 2 0.17 0.19 0.18 0.23 0.30 0.127***

Panel B: Contribution Index 1 (Using Cash Flow Rights)

Cash Flow Right

(1) (2) (3) (4) (5)
Contribution Index 1 0.1706*** 0.1312*** 0.1686*** 0.1876*** 0.1683***
(15.19) (13.44) (11.44) (11.42) (11.42)
Leverage 0.0905*** 0.0015 0.0903***
(2.98) (0.04) (2.98)
Profitability 0.1251*** 0.1418*** 0.1232***
(2.90) (3.02) (2.85)
Years of Operation 0.0152** 0.0160** 0.0158**
(2.19) (2.34) (2.27)
Public Company -0.0110 -0.0213** -0.0105
(1.21) (2.14) (1.15)
Firm Size -0.0204*** -0.0207*** -0.0208***
(7.79) (7.19) (7.88)
Financial Institution -0.0100 -0.0071 -0.0105
(1.02) (0.66) (1.07)
Group Size -0.0103* 0.0144
(1.79) (1.36)
Number of Affiliates -0.0440*** -0.0195
(3.34) (0.57)
Group w/ Fin. Inst. -0.0198 0.0129
(0.91) (0.28)
Group Dummies No Yes Yes No Yes
Year Dummies No Yes Yes No Yes
Observations 4114 4114 2262 2262 2262
Adjusted R-squared 0.0927 0.3299 0.3874 0.2226 0.3870

- 51 -
Panel C: Contribution Index 2 (Using Voting Rights)

Cash Flow Right

(1) (2) (3) (4) (5)
Contribution Index 2 0.0684*** 0.0563*** 0.0693*** 0.0736*** 0.0689***
(7.72) (7.54) (6.45) (6.19) (6.39)
Leverage 0.1011*** -0.0024 0.1009***
(3.21) (0.07) (3.20)
Profitability 0.1321*** 0.1537*** 0.1302***
(2.94) (3.17) (2.89)
Years of Operation 0.0133* 0.0166** 0.0140*
(1.82) (2.31) (1.90)
Public Company -0.0195** -0.0310*** -0.0189*
(2.03) (2.92) (1.96)
Firm Size -0.0219*** -0.0227*** -0.0223***
(7.89) (7.44) (7.97)
Financial Institution -0.0121 -0.0079 -0.0125
(1.22) (0.71) (1.26)
Group Size -0.0076 0.0147
(1.27) (1.36)
Number of Affiliates -0.0550*** -0.0112
(3.99) (0.31)
Group w/ Fin. Inst. -0.0116 0.0160
(0.49) (0.35)
Group Dummies No Yes Yes No Yes
Year Dummies No Yes Yes No Yes
Observations 4114 4114 2262 2262 2262
Adjusted R-squared 0.0220 0.2938 0.3278 0.1413 0.3274

- 52 -
 Table 9: Contribution Index and Cash Flow Right: Robustness Check

OLS regression of Table 9 Panel B Equation (5) and Panel C Equation (5) are estimated for each year in
the sample period. t-values, based on Whites heteroskedasticity-consistent standard errors, are
reported in parentheses. *, **, and ***, indicate significance at 10, 5, and 1% levels, respectively.

Cash Flow Right

1997 1998 1999 2000 2001 2002 All Years
Panel B, eq. (5) 0.2328*** 0.2024*** 0.1935*** 0.1420*** 0.1450*** 0.1297*** 0.1683***
(4.70) (5.08) (4.43) (5.36) (5.24) (5.88) (11.42)
Obs. 461 401 368 317 329 378 2262
Adj. R2 0.2923 0.3554 0.3542 0.3728 0.5666 0.5003 0.3870
Panel C, eq. (5) 0.0238 0.0791** 0.0832*** 0.0880*** 0.0821*** 0.0930*** 0.0689***
(1.04) (2.36) (2.77) (3.67) (3.26) (4.61) (6.39)
Obs. 461 401 368 317 329 378 2262
Adj. R2 0.1966 0.2916 0.2976 0.3265 0.5136 0.4763 0.3274

- 53 -
 Table 10: Sub-Sample Results: Listed versus Non-Listed

OLS regressions are estimated with all three determinants of our interest: profitability, risk, and
contribution to group control. In columns (2) and (3), we run the same equation for non-listed
firms and listed firms. t-values, based on Whites heteroskedasticity-consistent standard errors,
are reported in parentheses. *, **, and ***, indicate significance at 10, 5, and 1% levels, respectively.

Cash Flow Right

(1) (2) (3)
All Firms Non-listed Firms Listed Firms
1yr (EBIT/Asset) 0.1356*** 0.1669** 0.0539
(3.01) (2.47) (0.97)
5yr SD(EBIT/Asset) 0.1320 0.2023 -0.0009
(1.61) (1.54) (0.01)
Contribution Index 1 0.1462*** 0.1238*** 0.1752***
(11.02) (9.36) (4.39)
Voting Rights 0.2747*** 0.3281*** 0.2149***
(18.56) (13.98) (12.49)
Leverage 0.0509** 0.0955*** -0.0184
(2.47) (3.31) (0.66)
Years of Operation 0.0019*** 0.0014** 0.0022***
(5.66) (2.02) (7.17)
Public Company 0.0002 - -
(0.03) - -
Firm Size 0.0000*** 0.0000 0.0000
(2.58) (0.29) (0.69)
Financial Institution -0.0174* -0.0174 -0.0020
(1.67) (0.94) (0.18)
Group Size -0.0000*** -0.0000*** -0.0000*
(3.45) (2.83) (1.89)
Number of Affiliates 0.0017** 0.0014 0.0020***
(2.34) (0.83) (3.15)
Group w/ Fin. Inst. 0.0149 0.0263 -0.0040
(0.38) (0.50) (0.09)
Group Dummies Yes Yes Yes
Year Dummies Yes Yes Yes
Observations 2117 939 1178
Adjusted R-squared 0.7085 0.5855 0.8242

- 54 -
 about ECGI

The European Corporate Governance Institute has been established to improve corpo-
rate governance through fostering independent scientific research and related activities.

The ECGI will produce and disseminate high quality research while remaining close to
the concerns and interests of corporate, financial and public policy makers. It will draw on
the expertise of scholars from numerous countries and bring together a critical mass of
expertise and interest to bear on this important subject.

The views expressed in this working paper are those of the authors, not those of the ECGI
or its members.

www.ecgi.org
ECGI Working Paper Series in Finance

Editorial Board

Editor Paolo Fulghieri, Professor of Finance, University of North

Carolina, INSEAD & CEPR
Consulting Editors Franklin Allen, Nippon Life Professor of Finance, Professor of
Economics, The Wharton School of the University of
Pennsylvania
Patrick Bolton, John H. Scully 66 Professor of Finance and
Economics, Princeton University, ECGI & CEPR
Marco Pagano, Professor of Economics, Universit di Salerno,
ECGI & CEPR
Luigi Zingales, Robert C. McCormack Professor of
Entrepreneurship and Finance, University of Chicago & CEPR
Julian Franks, Corporation of London Professor of Finance,
London Business School & CEPR
Xavier Vives, Professor of Economics and Finance,
INSEAD & CEPR
Editorial Assistant : Cristina Vespro, ECARES, Universit Libre De Bruxelles

Financial assistance for the services of the editorial assistant of these series is provided
by the European Commission through its RTN Programme on Understanding Financial
Architecture: Legal and Political Frameworks and Economic Efficiency (Contract no.
HPRN-CT-2000-00064).

www.ecgi.org\wp
Electronic Access to the Working Paper Series

The full set of ECGI working papers can be accessed through the Institutes Web-site
(www.ecgi.org/wp) or SSRN:

Finance Paper Series http://www.ssrn.com/link/ECGI-Fin.html

Law Paper Series http://www.ssrn.com/link/ECGI-Law.html

www.ecgi.org\wp