Diversification, derivative usage, and firm value

MISSING IMAGE

Material Information

Title:
Diversification, derivative usage, and firm value
Physical Description:
Book
Creator:
Fauver, Larry A., 1969-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 24888350
oclc - 45294731
System ID:
AA00020430:00001

Full Text










DIVERSIFICATION, DERIVATIVE USAGE, AND FIRM VALUE


By

LARRY A. FAUVER












A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2000































Copyright 2000



by


Larry A. Fauver
























To my wife (Deborah), and my parents (Larry and Marcella)















ACKNOWLEDGMENTS

I am grateful to my chairman, Joel Houston, Mark Flannery, and P.J. van Blokland for

their helpful comments and suggestions. I am especially grateful to my co-chair, Andy Naranjo,

for his belief in my ability and continued patience and guidance in this work. I would also like to

thank June Nogle, who accepted question after question when I was learning SAS. I thank my

wife, Deborah, for her encouragement and motivation throughout this process. She has been very

patient and supportive, and I am fortunate to call her my wife. Finally, I would like to thank my

parents, Larry and Marcella. It is a credit to them that I am here at this point in my life. They

have given me so many things throughout the years. I express a sincere thank you to them.















TABLE OF CONTENTS

pag=

ACKNOW LEDGM ENTS ............................................................................................................. iv

ABSTRACT .................................................................................................................................. vi

CHAPTERS

1 INTRODUCTION ............................................................................................................. 1

2 FIRM VALUE AND CAPITAL MARKET DEVELOPMENT ....................................... 4

Introduction ...................................................................................................................... 4
Data ................................................................................................................................ 12
Sum m ary Statistics ......................................................................................................... 13
M methodology .................................................................................................................. 14
Results ............................................................................................................................ 17
Conclusion ...................................................................................................................... 32

3 FIRM VALUE AND INTERNATIONAL DIVERSIFCATION .................................... 54

Introduction .................................................................................................................... 54
Data and M ethodology ................................................................................................... 57
Results ............................................................................................................................ 61
Conclusion ...................................................................................................................... 72

4 FIRM VALUE AND DERIVATIVE USAGE ................................................................ 93

Introduction .................................................................................................................... 93
Data and M ethodology ................................................................................................... 96
Results ............................................................................................................................ 98
Conclusion .................................................................................................................... 105

5 DISCU SSION AND CONCLU SION ........................................................................... 116

REFEREN CES ........................................................................................................................... 117

BIOG RAPHICAL SKETCH ...................................................................................................... 123















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

DIVERSIFICATION, DERIVATIVE USAGE, AND FIRM VALUE


By

Larry A. Fauver

August 2000
Chairman: Joel F. Houston
Major Department: Finance, Insurance, and Real Estate

This dissertation investigates the connection between firm diversification, derivative

usage and their effects on firm excess value. First, we calculate the value of corporate

diversification on the level of capital market development. With a sample of more than 8,000

firms from 35 countries during 1991-1995, we find that the value of corporate diversification is

negatively related to the level of capital market development. We also find that the value of

corporate diversification varies with the legal system.

Second, we provide evidence on the value of product and geographic diversification and

their impact on firm excess value. We collect data on more than 4,000 firms from four countries

(Germany, Japan, the U.K., and the U.S.), from 1991-1995, and find that U.S. and Japanese

multinationals trade at a higher value relative to a domestic firm operating with the same product

mix. However, we find that multinationals do no better than a comparable international portfolio

of firms in their respective domestic market.

Finally, we evaluate the effects of derivative usage on firm excess value as well as the

interactions among derivative usage, product diversification, geographic diversification, and firm

excess value. With a sample of more than 1,600 firms from the U.S., from 1991-1995, we find a








value loss for industrially diversified firms that use derivatives, with the greatest loss occurring

for large diversified firms. Results that differentiate between expected and unexpected derivative

usage and control for individual firm characteristics suggest that the value loss is associated with

unexpected derivative usage by industrially diversified firms.














CHAPTER 1
INTRODUCTION

The issue of corporate diversification and firm value has generated substantial interest in

recent years. Potential benefits may include operating synergies, and the ability to effectively

allocate scarce capital through an internal capital market. Costs may include increased agency

conflicts, which may lead to increased costs. The empirical evidence, however, appears to

indicate that firms that industrially diversify, at least in the U.S., perform worse relative to firms

that are concentrated in one line of business. The main reason attributed to the empirical findings

is that diversified firms may encounter greater agency conflicts and consequently more costs than

a firm with a focused organizational structure. The connection between corporate diversification

and firm value is less obvious outside the U.S. because it is unclear that these potential agency

costs are similar across countries and legal systems.

There is also the issue of product and geographic diversification and their effect on firm

value. International diversification may allow firms to capitalize on operating and financial

synergies, and capture possible growth opportunities. The potential benefits may also include

portfolio diversification for shareholders. However, this benefit may not accrue to the

multinational itself. The potential costs may include additional costs from managing more

resources, increased political and exchange rate risks, and foreign government intervention. The

empirical evidence is mixed and inconclusive. Therefore, the effect of international

diversification on firm value is uncertain.

The third issue is derivative usage, diversification, and their effect on firm value.

Derivative instruments may allow firms to control for the costs and the risks associated with

diversification. The theoretical evidence determines firms may use derivative instruments for tax

motives, reduction in bankruptcy costs, and leverage, asymmetric information and moral hazard








stories among others. The empirical evidence mainly supports the theoretical evidence for

derivative usage by firms. Evidence regarding the direct effect of derivative usage and firm value

is not as comprehensive. This can be attributed to the lack of detailed historical reporting of

derivative instruments by firms. The potential benefits of derivative usage may include cash flow

smoothing, increased debt capacity, and risk transfer, among others. Potential costs may include

increased agency conflicts when using derivatives, and increased risks. Some empirical evidence

indicates an increase in firm value when using derivative instruments. Although it appears that

more research is needed before resolving that derivative usage is positively related to firm value.

This dissertation explores the connection between firm excess value, diversification, and

derivative usage. First, we calculate the value of product diversification on the level of capital

market development. We examine a sample of more than 8,000 firms from 35 countries during

1991-1995. We observe that the value of corporate diversification is inversely related to the level

of capital market development. We also uncover that the value of product diversification varies

with the legal system. These results suggest that the financial, legal, and regulatory environment

all have a significant influence on the value of product diversification, and that the optimal

organizational structure for firms operating in emerging markets may be unrelated to those firms

operating in more developed economies.

Second, we provide evidence on the value of product and geographic diversification and

its impact on firm excess value. The sample consists of more than 4,000 firms from four

countries (Germany, Japan, the U.K., and the U.S.), from 1991-1995. Our regression analysis

controls for the firm's size, profitability, capital intensity, and ownership structure. We find that

the U.S. and Japanese multinationals outperform the domestic firms operating within the same

line of business. However, we find that multinationals in each country do not outperform an

international portfolio of firms in each of their domestic markets. These results are consistent

with the international investments literature which suggest that shareholders can earn an








equivalent risk-adjusted return from a multinational by holding a portfolio of domestic firms in

each international market.

Finally, we evaluate the effects of derivative usage on firm excess' value as well as the

interactions among derivative usage, product diversification, geographic diversification, and firm

excess value. With a sample of more than 1,600 firms from the U.S., from 1991-1995, we find

that focused firms that use derivative instruments have higher unconditional average excess

values than diversified firms that do not use them. After using regression procedures that control

for firm characteristics including firm profitability, growth opportunities, size, leverage, dividend

structure, and ownership concentration, we also find that the value loss is larger for industrially

diversified firms that use derivatives, with the greatest loss occurring for large diversified firms.

These results are consistent with greater agency costs in large, diversified firms. Interestingly,

results that differentiate between expected and unexpected derivative usage and control for

individual firm characteristics suggest that the value loss is associated with unexpected derivative

usage by industrially diversified firms. These findings also suggest that when firms use

derivatives as expected, there are no valuation effects.














CHAPTER 2
FIRM VALUE AND CAPITAL MARKET DEVELOPMENT

Introduction

The connection between corporate diversification and firm value continues to generate

substantial interest among financial theorists and practitioners. Recent evidence suggests that

diversified U.S. firms trade at discounts compared to firms that are more focused [e.g., Lang and

Stulz (1994), Berger and Ofek (1995), John and Ofek (1995), and Comment and Jarrell (1995)].'

One explanation for these findings is that diversified firms face higher agency costs as a

consequence of their organizational form. For example, recent papers have argued that intra-firm

coordination problems are likely to be more extensive for diversified firms, because of their need

to allocate capital among their various disparate activities [e.g., Rajan and Zingales (1996b)

Scharfstein and Stein (1997), and Scharfstein (1998)].2

Despite the observed costs arising from corporate diversification, there is theoretical

work that suggests that there may also be benefits from diversification. In particular, work by

Williamson (1975), Gertner, Scharfstein and Stein (1994), Harris and Raviv (1996), and Stein

(1997) suggests that capital constrained firms may establish internal capital markets that are able

to effectively allocate scarce capital within the firm.3 Recent empirical evidence documents that

there are systematic patterns in the internal allocation of capital in diversified firms [e.g., Shin


1 These results are also consistent with the evidence that corporate spin-offs generally enhance
shareholder value [see, for example, Hite and Owers (1983), Schipper and Smith (1983), and
Kaplan and Weisbach (1992)].
2 Denis, Denis and Sarin (1997) argue that value-reducing diversification strategies are sustained
over time because they benefit managers (at the expense of shareholders), but that a competitive
corporate control market may spur many firms to increase their focus.
3 It is interesting to note, however, that Stein's model actually implies that internal capital
markets may work best among firms that are more focused.

4








and Stulz (1997), Lamont (1997), Houston, James and Marcus (1997), and Scharfstein (1998)],

but it remains an open question whether this allocation works to increase or decrease shareholder

value.

It also remains an open question whether or not the extant empirical evidence extends

beyond the results reported for U.S. firms. On one level, the agency costs accompanying

diversification may vary systematically across countries and legal systems. At the same time,

Khanna and Palepu (1997) argue that the relative costs and benefits of corporate diversification

depend critically on the "institutional context" in which the firm operates. The institutional

context includes the financial, legal, and regulatory environment. In a similar vein, LaPorta,

Lopez-De-Silanes, Shleifer, and Vishny (1997) show that different legal systems provide

investors with varying degrees of protection which, in turn, affect the level of economic and

capital market development.4 These results also suggest that the value of corporate

diversification is related to the legal system. While diversification may have limited value in a

developed economy such as the U.S. where the institutional context enables smaller, stand-alone

firms to raise capital, it may be more valuable for firms who find it costly or impossible to raise

external capital, either because of imperfect information or incomplete capital markets.5

A firm's access to external capital depends on its ability to obtain domestic and/or foreign

capital. Consequently, the extent to which capital markets are developed within the country

where the firm operates, and the extent to which that country is able and/or willing to attract

foreign capital, will both have a strong influence on a firm's ability to raise capital. We would

expect that internal capital markets are most valuable among firms and economies where it is


4 Demirgdc-Kunt, and Maksimovic (1996) also find that legal systems affect growth rates and the
ability to enter into long-term financial contracts. Desai (1997), moreover, finds that
multinational firms employ internal capital markets to overcome the higher costs of external
finance associated with weaker creditor rights in lesser developed markets.

5 The economic and legal environment in less developed markets may also make it more difficult
to contract with other firms, and therefore, may provide an additional benefit to diversification.
Another potentially important benefit from diversification is the relatively high level of political
influence that conglomerates and business groups wield in less developed markets. These
political connections can create differential access to resources and markets.








costly to obtain external capital. Therefore, unless the agency costs accompanying diversification

are significantly higher in these countries, we would expect that the benefits from diversification

would be higher in countries where capital markets are less developed and where legal systems

provide limited protection to investors. If this conjecture is correct, it raises the possibility that

the results indicating a diversification discount for the U.S. do not generalize to other countries

where external capital markets are less developed.6 In particular, we would expect to see smaller

diversification discounts, and perhaps even diversification premiums, among firms that operate in

less developed markets.

To date, the international evidence regarding corporate diversification has been limited.

One notable exception is the recent work by Lins and Servaes (1999). Looking at a sample of

firms from Germany, Japan, and the United Kingdom in 1992 and 1994, they report valuation

discounts that are of similar magnitude to those reported for U.S. firms. Moreover, their

estimated diversification discounts remain statistically significant for Japan and the United

Kingdom even after controlling for firm characteristics. In Germany, after controlling for firm

characteristics, they also report a diversification discount, but it is not statistically different from

zero.7

Also notable is the recent work by Khanna and Palepu (1997). They argue that

diversification may be more valuable in emerging markets than in more developed economies.

Khanna and Palepu's analysis focuses on diversified business groups within India. They find that

larger diversified groups that are in a better position to tap external capital outperform smaller


6 This argument might also suggest that the value of diversification within a given country may
decline over time as the country's capital markets become more developed. Servaes (1996),
Klein (1998), and Hubbard and Palia (1998) have examined this issue by considering the value of
diversification in the U.S. during the conglomerate wave of the 1960s.

7 In a more recent paper, Lins and Servaes (1998) use data from 1995 to investigate the value of
corporate diversification for Hong Kong, India, Indonesia, Malaysia, Singapore, South Korea,
and Thailand. They find that for six of their seven countries, there is no statistically significant
diversification discount -- only for South Korea did they find a diversification discount that was
statistically different from zero.








unaffiliated firms. Khanna and Palepu's study provides some indirect support for our hypothesis

that the value of diversification depends critically on the level of capital market development.

In this paper, we investigate the link between capital market development and the value

of corporate diversification. To address this issue more extensively, we have assembled a large

data set that consists of more than 8,000 firms from 35 countries over a five-year period between

1991 and 1995. Using the methodology employed by Berger and Ofek (1995) and Lins and

Servaes (1999), we calculate the implied value gain or loss from diversification. In addition, we

test whether the gain or loss that results from diversification is systematically related to the level

of capital market development.

Our results provide evidence that the value of diversification is related to the degree of

capital market development. In particular, after controlling for the legal environment in which

the firm operates and firm-specific factors such as firm size, capital structure, profitability, and

ownership structure, we find that the value of diversification varies with the degree of capital

market development. Among high-income countries, where capital markets are well developed,

we find a statistically significant diversification discount. This finding is consistent with the U.S.

evidence and the international evidence presented by Lins and Servaes (1999). By contrast, for

the lower income countries, we find that there is either a significant diversification premium or no

diversification discount. For these firms, the benefits of diversification appear to offset the

agency costs of diversification. These results are consistent with Khanna and Palepu's evidence

from Indian business groups.

We also find that the diversification discount systematically varies with the legal system.

LaPorta, Lopez-De-Silanes, Shleifer, and Vishny (LLSV, 1997) document that the English legal

system provides the most protection to capital providers. If this protection results in better access

to external capital, the benefits of internal capital markets and corporate diversification may

arguably be smaller in countries that operate under a legal system with English origin. Consistent

with this argument, we find that diversification discounts are largest among countries where the








legal system is of English origin. We find smaller diversification discounts in countries where the

legal system is of a German, Scandinavian, or French origin.

Lastly, we find that our results are robust with respect to controlling for the agency costs

associated with concentrated ownership, differences in accounting rules across countries, and

various measures of capital market development and the legal environment.

The rest of the paper proceeds as follows. The next section reviews the connection

between capital market development, economic development, and legal systems. We also

describe the various economic development classifications and legal systems for each of the 35

countries in our sample. The third section describes our data and the methodology used to

calculate the value of corporate diversification. The cross-country mean estimates of the value of

corporate diversification are presented in the fourth section. Regression results regarding the

value of diversification after controlling for firm-specific characteristics are presented in section

five. In section five, we also provide a number of robustness tests, including the effects of

controlling for cross-firm and cross-country differences in accounting practices. Section six

examines the links between the value of diversification and ownership structure, while section

seven provides a conclusion.

Corporate Diversification and Capital Market Development

One clear drawback of corporate diversification is that it creates another layer of potential

agency problems within the firm. Internal politics and imperfect information within the firm may

complicate the ability of senior managers to effectively allocate capital among the various lines of

business that exist within a conglomerate [see, for instance, Rajan and Zingales (1996b),

Scharfstein and Stein (1997), and Scharfstein (1998)]. Despite these costs, corporate

diversification may still be beneficial. In some cases, combining different lines of business

within the same organization may generate value-creating operating synergies. Diversification

may also create financial synergies to the extent it reduces the cost of obtaining capital [see, for








instance, Lewellen (1971), Stein (1997), Williamson (1975) and Hadlock, Ryngaert and Thomas

(1998)].

The financial synergies arising from diversification are likely to vary with the level of

capital market development. For example, Rajan and Zingales (1996a) suggest that there are

important cross-country differences in access to capital markets. They demonstrate that the

development of a country's financial sector reduces the cost of external finance. In

demonstrating a link between financial development and economic growth, they show that firms

operating in industries which are generally more reliant on external finance grow faster if they are

established in a country that has a more developed financial system. These results are consistent

with our main hypothesis that the value of diversification is greater in countries where capital

markets are less developed.

At the same time, the agency costs of diversification are also likely to vary across firms

and across countries. While it is difficult to directly measure these agency costs, a long-standing

literature suggests that these costs (and therefore ultimately firm value) may be correlated with

ownership structure [see, for example, Demsetz and Lehn (1985), Morek, Shleifer and Vishny

(1988), Holdemess and Sheehan (1988), and McConnell and Servaes (1990)]. Moreover, recent

work by La Porta, Lopez-de-Silanes, and Shleifer (1999) and by Claessens, Djankov, Fan and

Lang (1998) indicate that ownership structure as well as the correlation between ownership

structure and firm value, vary across countries and legal systems. While ownership concentration

is likely to affect firm value, it remains an open question whether it also has an effect on the value

of corporate diversification. In Section VI, we address the agency costs of diversification by

explicitly controlling for ownership concentration among the subset of firms where these data are

available.

Another factor that may attenuate the link between the value of corporate diversification

and the degree of capital market development is the increased integration of global capital

markets in recent years. Indeed, if capital markets are perfectly integrated, we would expect that








firms would be able to access external capital at the global cost of capital, even if the financial

sector is less developed in the country in which they operate. Empirical studies on the degree of

capital market integration performed for various markets and under varying assumptions have

yielded mixed results [see, for instance, Jorion and Schwartz (1986), Cho, Eun and Senbet

(1986), Wheatley (1988), Gultekin, Gultekin and Penati (1989), Mittoo (1992), Chen and Knez

(1995), Bekaert and Harvey (1995), Naranjo and Protopapadakis (1997), and Stulz (1999)].

Given these mixed results, the link between capital market development and the value of

corporate diversification is ultimately an empirical question.

In order to test our main hypothesis, we need to measure capital market development

across countries. Capital market development can be measured in a variety of ways including

per-capita GNP, equity market capitalization relative to GNP, the number and dollar amount of

per-capita initial public offerings, the ratio of public and private debt to GNP, and the relative size

of the banking system.8 In our analysis, we rely on recent research which demonstrates that there

is a strong link between capital market development and economic development [see, for

example, Levine (1997), King and Levine (1993a, 1993b) and Rajan and Zingales (1996a)].

While the causation may be unclear, countries with higher levels of economic development (on

the basis of traditional measures such as per-capita GNP) are likely to have a more extensive

domestic capital markets and are also more likely or willing to obtain foreign capital.9

We primarily use two proxies to test whether capital market development influences the

value of corporate diversification. First, relying on the link between capital market development

8 King and Levine (1993a), Rajan and Zingales (1996a), and LaPorta, Lopez-De-Silanes, Shleifer,
and Vishny (1997) provide a more detailed discussion of these capital market development
measures. The problem with many of these measures is a lack of comprehensive data.
Furthermore, some of these other measures may provide a misleading depiction of the
accessibility of external capital. For example, measures of equity market capitalization relative to
GNP are typically low for many European countries, but most would argue that European firms
have good access to external financial markets.

9 One potential problem with using per-capita GNP as a measure of capital market development is
that some countries with vast natural resources may demonstrate high per-capita GNP, even
though firms that operate in these markets have limited accessibility to external capital. None of
the countries in our sample, however, fall into this category.








and economic development, we use the World Bank's classification of economic development as

a proxy for capital market development. Each year, the World Bank classifies countries into four

categories: high income, upper-middle income, lower-middle income, and low income. This

classification is largely based on the country's per-capita GNP. With this in mind, we also

employ the country's per-capita GNP itself as a proxy for capital market development.'0

In addition to these proxies, we also control for the country's legal system to take into

account the evidence by LaPorta, Lopez-De-Silanes, Shleifer, and Vishny (1997, 1998), which

documents a link between legal systems and capital market development. LLSV classify

countries into four different legal systems: those with English, French, German, and Scandinavian

origin." Their evidence suggests that a country's legal system significantly affects the level of

protection that is given to investors, which in turn affects the availability of external capital. In

particular, they find that the English system, with its common law origin, provides investors with

the strongest legal protection, while the French legal system provides the least protection. They

also argue that countries whose legal system is of German or Scandinavian origin have a

moderate level of investor protection, falling somewhere between the English and French

systems. Controlling for economic development, we would therefore expect that diversification

discounts would be largest among countries with an English legal system, since firms in these

countries are likely to have better access to external capital markets.

Table 2-1 summarizes the economic development and legal system classifications for

each of the 35 countries in our sample. We use the legal classifications reported in LLSV. The

average per-capita GNP is the five-year arithmetic average over our sample period, 1991-1995.


0 As additional measures of capital market development, we also use for each country the ratio of
the stock market capitalization held by minorities plus the sum of bank debt of the private sector
and outstanding non-financial bonds to GNP (MKTCAP + Debt/GNP), the ratio of the number of
domestic firms listed in a given country to its population (Domestic Firms/Pop), and the ratio of
the number of the initial public offerings of equity in a given country to its population
(IPOs/Pop). These data are obtained from LaPorta, Lopez-De-Silanes, Shleifer, and Vishny
(1997). See Section V, sub-section C.

" From LLSV, we also obtain the law and order tradition (Rule of Law) in each country. See
Section V, sub-section C.








This measure ranges from $316 in India to $36,800 in Switzerland. As indicated above, the

World Bank classification largely coincides with per-capita GNP.

Data

Sample Construction

Our main data source is the Worldscope database.'2 Worldscope has complete financial

data and business segment data for more than 8,000 companies, located in 49 countries. The

firms in the databank represent 86 percent of global market capitalization. The business segment

data start in 1991. For this reason, our sample period begins in 1991 and extends through 1995.13

We use the reported business segment data to classify the publicly traded firms as either single-

segment (focused) or multi-segment (diversified). We classify firms as single-segment firms if

they operate in only one two-digit SIC code industry. Firms are classified as multi-segment if

they have more than one reported segment, and the largest segment has less than 90 percent of the

total sales for the company.

Within each country, we exclude multi-segment firms from the sample if the company

does not report sales at the individual segment level. However, in cases where individual

segment sales are not reported and there is only one primary reported SIC, we classify the firm as

a single-segment firm and use the firm's total sales.14 We also exclude firms whose primary

business is financial services (i.e., where more than fifty percent of firm sales come from SICs in

the 6000-6999 range). These firms are excluded because sales figures are irregularly reported

and are difficult to interpret for financial institutions. Finally, we exclude firms where there are

no pure play matches and corresponding segment sales exceed 25 percent of total sales. For 14 of



12 This databank is also used by LaPorta, Lopez-De-Silanes, Shleifer, and Vishny (1997, 1998),
Lins and Servaes (1999), Claessens, Djankov, Fan and Lang (1998), and LaPorta, Lopez-De-
Silanes and Shleifer (1999).

" We wish to thank Worldscope for providing us with machine-readable access to their databank.

14 Due to data limitations, we are unable to disentangle firms that may be diversified, but only
report one line of business.








the 49 countries, there were insufficient data to calculate the estimated value of diversification,

leaving 35 countries with sufficient data.s'516

Summary Statistics

Table 2-2 reports firm level summary statistics broken down by the level of economic

development and the legal system in which the firm is headquartered. Panel A divides the firms

according to their country's World Bank classification. Across the four classifications,

diversified firms have a mean number of segments varying from just over 2.5 segments in the

high-income countries to just under 3 segments in the upper-middle income countries. In

virtually all cases, diversified firms are significantly larger than the focused firms in terms of both

total assets and total capital. We also find that there is no consistent distinction in the leverage

ratios between the single and multi-segment firms in our sample.'7

Looking at the firm level characteristics for the high-income country group, we find that

single-segment firms have a higher average market-to-sales ratio. This evidence is consistent

with the results of Lang and Stulz (1994), Berger and Ofek (1995), and Lins and Servaes (1999),

and provides broad evidence suggesting that single-segment firms are valued more highly than


15 We also exclude firms where the actual value (imputed value) is more than four (one-fourth)
times the imputed value (actual value) see Section IV, sub-section A. Firms are primarily
excluded from our sample according to the following two screens: firms whose primary business
is financial services and firms where the actual value (imputed value) is more than four (one
fourth) times the imputed value (actual value). These two screens account for 87 percent of the
firms eliminated from our sample, while only 2 percent of the firms are excluded from our sample
due to multi-segment firms that do not report sales at the individual segment level.
16 Our sampling procedure differs from Lins and Servaes' (1999) in three ways. First, they
exclude service firms the reason being that were relatively few service firms in Germany, and
they wanted to control for industry differences across the three countries that they were
investigating. In our study, we have chosen to include the broadest possible sample of firms and
countries. Second, Lins and Servaes also exclude firms that do not trade on the country's main
exchange. Third, to keep the data collection process manageable, Lins and Servaes only use a
random sample of 450 firms from Japan and the United Kingdom in 1992 and 1994, whereas we
use all firms in the databank that meet our screens. While our sampling procedure is somewhat
different, the estimated diversification discounts that we find for Japan, Germany, and the United
Kingdom are quite similar to those reported by Lins and Servaes (1999).
' Lins and Servaes (1999) also find no distinction in leverage ratios between focused and
diversified firms, while Lewellen (1971), Kim and McConnell (1977), Comment and Jarrell
(1995), and Berger and Ofek (1995) find that diversified U.S. firms have higher debt ratios.








diversified firms. However, this result does not generalize to the lesser-developed countries. In

two of the other three classifications (upper-middle and low-income), the diversified firms have a

median market-to-sales ratio that is higher than that found for the focused firms. For the other

two ratios, operating income-to-sales and capital expenditure-to-sales, there is no significant

distinction between the single and multi-segment firms.

In Panel B, the firms are divided according to the legal system of the country in which

they are headquartered. Once again, the results indicate that diversified firms are generally

larger, although this difference does not appear to be significant for countries with a French legal

system. Consistent with the results reported earlier, diversified firms generally have a lower

market-to-sales ratio, which again provides indirect evidence that diversified firms trade at a

discount relative to focused firms. We address this issue more completely in the next section

where we directly estimate the value of diversification.

Methodology

Estimating the Value of Corporate Diversification

To estimate the value of corporate diversification, we modify the approach originally

used by Berger and Ofek (1995). In our analysis, we use the ratio of total-capital-to-sales to

measure corporate performance, where total capital is calculated by adding the market value of

equity to the book value of debt. Along with this measure, Berger and Ofek (1995) also consider

two other ratios to measure performance: the ratio of total-capital-to-assets and the ratio of total-

capital-to-earnings. Their qualitative results are similar for each of the three performance

measures. We are unable to use these alternative measures because there is very little business

segment data regarding assets or earnings for the non-U.S. firms."8

We calculate the excess value of each firm by taking the difference between the firm's

actual performance and its imputed performance. Actual performance is measured by the

consolidated firm's capital-to-sales ratio. For single-segment firms, imputed value is calculated


"8 For similar reasons, Lins and Servaes (1999) also use the capital-to-sales-ratio as their sole
measure of performance.








as the median capital-to-sales ratio among all pure-play (single-segment firms) within the same

industry and same country. For multi-segment firms, imputed value is calculated by taking a

weighted-average of the imputed values for each of the firm's segments, where the weights

reflect the proportion of the overall firm's sales that come from each segment. Multi-segment

firms have a positive excess value (i.e., a premium) if the overall company's value is greater than

the "sum of the parts." By contrast, multi-segment firms have a negative excess value if their

value is less than the imputed value that would be obtained by taking a portfolio of pure-play

firms that operate in the same industries and country as the diversified firm.19

We define industries at the two-digit SIC code level.20 In cases where there are no other

two-digit pure-plays firms to match from, we calculate the imputed market capital-to-sales ratio

using broader industry classifications defined by Campbell (1996).21 Finally, to avoid having the

results driven by extreme values, we exclude firms where the actual value is more than four times

the imputed value, or where the imputed value is more than four times the actual value.22

The Value of Diversification

Table 2-3 reports the excess value estimates for the single and multi-segment

firms in our sample. Once again, the firms are classified according to each country's legal system

and the World Bank's classification of economic development for each country.



19 The average number of pure-plays ranges from 1.30 in New Zealand to 29.44 in the U.S., while
the average number of pure-plays in the less developed markets is 3.02. To further insure that our
results are robust with respect to the control groups, we also increased the required minimum
number of pure-play matches to three firms and obtained similar results, but with a considerably
smaller sample.
20 While this two-digit classification is somewhat coarse, it provides us with a larger number of
pure play firms. Increasing the number of pure-plays is particularly important in the less
developed markets. Lins and Servaes (1999) and others also use a similar approach.
21 The reported results are essentially the same if we eliminate firms from the sample that do not
have a two-digit pure-play match.
22 Berger and Ofek (1995) and Lins and Servaes (1999) also use this screen. When we use a more
conservative screen of excluding firms where the actual value (imputed value) exceeds the
imputed value (actual value) by a factor of three, we obtain similar results.








The results in Panel A, where the firms are divided according to the World Bank

classification, strongly suggest that the value of diversification is negatively correlated with the

degree of economic development. Diversified firms in the most developed nations trade at a

significant discount relative to focused firms. The median discount for the high-income group is

5.76 percent. By contrast, diversified firms in the low-income group trade at a significant

premium of 3.80 percent relative to focused firms. This finding suggests that diversification may

create net benefits among firms that operate in countries whose capital markets are not fully

developed, which is consistent with the evidence from Indian business groups reported by

Khanna and Palepu (1997).

One potential concern with the World Bank classification is that there are relatively few

firms (particularly diversified firms) within the lower income groupings, and these firms come

from a relatively small number of countries. For example, there are only three countries in our

sample that are in the low-income group China, India, and Pakistan. A concern that arises is

that it may be difficult to sort out whether any demonstrated effects for this group are due to its

low-level development, or to other country-specific factors. While we control for these effects

more completely in the subsequent regression analysis, another way to get at this issue is to

broaden the categories of economic development. Thus, in Panel B, we report similar excess

values, but the countries are divided more broadly according to their per-capita GNP. In this

classification, the lowest grouping also includes firms operating in Indonesia and the Philippines.

In Panel B, the mean and median excess values, using the broader per-capita GNP

groupings, are very similar to statistics reported in Panel A using the World Bank classifications.

Once again, the value of diversification varies with the level of economic development. Firms

that operate in countries with a per-capita GNP in excess of $15,000 have a mean diversification

discount of 5.79 percent and a median discount of 5.78 percent. The results are also strikingly

different for firms headquartered in the emerging market countries. Among these firms, we find a

mean diversification premium of 8.41 percent and a median premium of 5.41 percent. The








similarity between Panel A and Panel B confirms that the World Bank classifications are largely

driven by differences in per-capita GNP.

In Panel C, we classify firms according to their country's legal system. The results

indicate that diversified firms trade at substantial discounts if they operate in a country with a

legal system of English origin. Among these countries, the median discount is 8.57 percent.

Among the other countries in our sample with French, German, and Scandinavian legal origin, we

find no evidence of either a diversification discount or premium. These results complement the

evidence reported by LLSV (1997). Their results suggest that the English legal system provides

the most protection to external investors which generally leads to more developed capital

markets. Our results suggest that the value of internal capital markets is smallest when capital

markets are most developed.

Results

The results reported in Table 2-3 suggest that the degree of capital market

development affects the value of corporate diversification. While these results provide an overall

depiction of the value of diversification among various countries, they do not control for

individual firm characteristics, which may also affect the firm's market-to-sales ratio. These

other characteristics include the firm's size, profitability, and future growth opportunities. To

control for these factors, we estimate the following regression model for each of the thirty-five

individual countries in our sample:23

(1) Excess Value = a + flA (Diversification Dummy) + /82 (Log Assets)
+ f13 (Operating Income/Sales) + /8 (Capital Expenditures / Sales) + e.

Excess value is defined to be the natural log of the ratio of the firm's market value to its

imputed value. The diversification dummy (SEG) is equal to one for multi-segment firms and is

otherwise zero. The log of assets controls for potential firm size effects. The ratio of operating

income-to-sales (OIS) provides a measure of firm profitability, while the ratio of capital

23 Lang and Stulz (1994), Berger and Ofek (1995), and Lins and Servaes (1999) also estimate
similar models.








expenditures-to-sales (CES) proxies for the level of growth opportunities. Controlling for the

other factors, we would expect to see a positive link between excess value and both OIS and CES.

Since our data cover five years (1991-1995), we also include separate year dummies in the

regressions to control for intertemporal variations in market or economic conditions that may also

affect the firm's market-to-sales ratio.

Regression Results for the Individual Countries

The regression results for the individual countries are reported in Table 2-4. In 23 of the

35 countries, the estimated coefficient on the diversification dummy variable is negative. In 11 of

these 23 countries, the coefficient is statistically significant, suggesting a diversification discount.

In 12 of the 35 countries the coefficient is positive. In 4 of the 12 cases (Hong Kong, Norway,

Pakistan and Singapore), this coefficient is positive and statistically significant, suggesting that

there is a diversification premium for these countries, after controlling for the firm-level

characteristics.

As expected, we find that the estimated coefficients on the OIS (Operating Income/Sales)

and CES (Capital Expenditures/Sales) variables are generally positive and frequently significant.

These results confirm that firms that are more profitable and that have greater growth

opportunities typically have higher market-to-sales ratios. The signs on the estimated coefficients

for the log of asset variable vary considerably across the different countries. The previous

evidence on this variable is also mixed Berger and Ofek (1995) find a positive link between

firm size and firm value, while Lang and Stulz (1994) and Lins and Servaes (1999) find a

negative relation. Although not reported, the annual dummy coefficients indicate that there is

little time variation in the excess values, after controlling for firm characteristics.

The estimated coefficients on the diversification dummy appear to be reasonable and are

generally well within the ranges found in earlier studies. Among U.S. firms, we find a

diversification discount of 13.2 percent, which is similar to the 14.4 percent found by Berger and

Ofek (1995) over an earlier time period 1986-1991. For Germany, we find no evidence of a








statistically significant diversification discount or premium, confirming the conclusions reached

by Lins and Servaes (1999). Lins and Servaes also found a diversification discount for Japan of

roughly 10 percent for both 1992 and 1994. Looking at a broader set of firms, we find a

statistically significant diversification discount for Japan of 4 percent, which is smaller than their

estimate.24 Likewise, for the United Kingdom, Lins and Servaes (1999) found a 15 percent

discount. Looking at a significantly larger sample, we also find a discount for the United

Kingdom. Our estimated discount of 7 percent is smaller, but it remains highly significant.

As indicated above, most of the other diversification coefficients appear to be of a similar

magnitude to those reported for the United States, Japan, Germany and the United Kingdom.

However, the point estimates for a couple of countries do stand out. For example, the

diversification discount in Turkey is relatively large and marginally statistically significant, while

in Spain the discount is both relatively large and significant at the 1 percent level. At the other

extreme, we find a large diversification premium in both Pakistan and the Philippines, although

the premium for the Philippines is not statistically different from zero. While in each of these

cases the magnitude of the estimates appears to be large, the existence of a diversification

discount or premium is generally consistent with our predictions.

When we pool the firms in our sample along two dimensions related to the capital market

development of the country in which the firms are headquartered (the World Bank's classification

of development and the country's legal system), we find that there is a significant diversification

discount of 8.2 percent among the high-income countries.25 Interestingly, however, there is no

evidence of a significant diversification discount or premium for the firms that are not

headquartered in a high-income country. For these firms, it appears that the benefits of

24 Lins and Servaes' (1999) estimates for Japan did not include CES because many Japanese firms
did not report CES. We obtain results that are more similar to Lins and Servaes when we also
eliminate the CES criterion.
25 This result is consistent with the findings of Berger and Ofek (1995), Lang and Stulz (1994),
and Lins and Servaes (1999), and also reaffirms the summary statistics reported in Table 2-3.
This coefficient is highly significant with a t-statistic of-10.726.








diversification (operating synergies and the establishment of internal capital markets) roughly

offset the costs of diversification. These findings suggest that while corporate focus generally

makes sense in highly developed countries, its value may not extend worldwide in cases where

external capital markets are less developed. In this regard, our results lend support to the

conclusions reached by Khanna and Palepu (1997).

From the pooled legal system results (also not reported), we find that there is a strong

relation between the legal system and the value of corporate diversification. In particular, the

observed relations are consistent with our priors and are also consistent with the evidence found

by LLSV (1997, 1998). We find that diversification significantly reduces value in countries that

have a legal system with English, French, or Scandinavian origin. As expected, the value of

diversification is most negative for firms that operate in countries with an English legal system.

Finally, controlling for OIS, CES, size, and annual variations, we find neither a diversification

discount nor premium among the firms that operate in markets with a German legal system.

Firm-level Regression Results

To further test the link between capital market development and the value of

diversification, we also estimate firm-level regressions that include all of the firms from each

country and for each year of our sample period. In each case, the dependent variable is the firm's

excess value. These regressions, reported in Table 2-5, control for the firm-level characteristics

outlined above (OIS, CES, and firm size). The regressions also include variables reflecting (1)

the level of economic development of the country in which the firm is headquartered as measured

by the country's World Bank classification or per-capita GNP; (2) the country's legal system; (3)

year dummies to take into account time variation in the value of diversification.

The OLS regression estimates reported in columns (1) (3) of Table 2-5 and the fixed-

effects estimates reported in column (4) are for the full sample of firms (single-segment and

multi-segment firms), where the dummy variable, SEG, equals 1 if the firm has multiple

segments and equals 0 otherwise. The coefficient on SEG, therefore, indicates the value of








diversification after controlling for the firm-specific, time-specific, and country-specific factors.

The regression specification reported in the first column only controls for the firm-specific and

time-specific factors. This specification is the same one estimated for the country-level

regressions reported in Table 2-4. The second specification, reported in column (2), also includes

dummy variables corresponding to the World Bank classification of economic development and

the legal system of the country in which the firm is headquartered. In column (3), the regression

specification includes the legal system dummy variables and the level of the country's per-capita

GNP as a continuous variable alternative to the discrete World Bank classification dummy

variables. Column (4) provides fixed-effects estimates of the third specification.

The results indicate that across all firms, diversification has a negative impact on firm

value.26 In column (1), the estimated diversification discount for the full sample of firms is 7.8

percent. When we control for economic development and the legal system with dummy

variables, in column (2), the diversification discount for high-income countries with an English

legal system is 9.6 percent. Looking at the estimates in column (2), we also see that excess value

is significantly higher (at the 5 percent level) if the firm is from a country that is classified as low-

income by the World Bank (GI *SEG). In column (3), we also see that the value of

diversification is negatively related to per-capita GNP, in that there is a statistically significant

negative relation (at the 1 percent level) between excess value and the variable which interacts

per-capita GNP with the diversification dummy, SEG. In terms of economic significance, the

estimated per-capita GNP coefficient in column (3) implies a discount for the U.S. of 10.5 percent

(-0.426 x 105 x 24,758).27

26 In each case, the adjusted R2's are somewhat lower than those of the individual country
estimates in Table 2-4. While there are clear benefits to pooling the countries, there is also more
noise introduced.
27 As an additional robustness check, we also estimated the regression models corresponding to
columns (1) (4) using only the multi-segment firms. For the multi-segment firm regressions,
we included as our measure of diversification the number of segments, SEGN, as an additional
explanatory variable in place of SEG. Similarly, in these regressions, each of the interacted
variables was interacted with the number of segments (as opposed to interacting with SEG). The
results were very similar to those reported for the entire sample. In particular, in all cases there








The legal system dummies are also significantly different from zero, and the estimated

coefficients have the predicted signs. In particular, we find that the estimated coefficients are

positive for the French, German, and Scandinavian legal dummy variables, indicating that

diversification provides greater benefits and/or fewer costs relative to firms that operate in a

country with a legal system of English origin. Looking more closely at the estimated coefficient

for the legal system dummy variables, we also see that the coefficient for the German legal

system is the most positive. This result suggests that after controlling for the other relevant

factors, the net costs of diversification are the smallest for firms that operate under the German

legal system.

As a robustness check, we also estimate the third specification using fixed-effects.28

These results are reported in column (4). Similar to the OLS estimates, we find that for the

diversified firms there is a statistically significant negative link (at the 1 percent level) between

per-capita GNP and excess value. We also find that the German legal system provides the

smallest diversification costs. As a further robustness check, we also estimated columns (1)-(3)

on a year-by-year basis. Once again, the estimates (not reported) confirm the negative link

between per-capita GNP and excess value and the variation of excess value across legal

systems.29




was a significant negative correlation between excess value and the number of segments. We
also found that the value of diversification was negatively related to per-capita GNP and that the
estimated coefficients interacting the diversification variable with the World Bank income group
dummies and with the legal system dummies had the same signs, and were generally even more
significant than the results reported for the entire sample.

2 The fixed-effects estimates for the first specification in column (1) are similar to those reported
for the OLS estimates. For the second specification in column (2), several of the coefficients can
not be estimated using fixed-effects due to singularity of the data. The singularity arises from the
inclusion of discrete dummy variables for development and the legal system that persist over
time.
29 The development and legal system results are significant in 1992-1994. In 1995, the results are
marginally significant. In 1991, the results are largely insignificant because there are too few
low-income country firm observations to get precise estimates.








Additional Proxies for Capital Market Development and the Legal Environment

Up until now, we have primarily used per-capita GNP and legal origin indicator variables

as proxies for capital market development and the legal environment. However, it is important

that we also employ additional measures in order to insure that our results are robust. LaPorta,

Lopez-De-Silanes, Shleifer, and Vishny (1997) analyze several measures of capital market

development and the legal environment across 49 countries. In particular, as measures of capital

market development for each country, they consider the ratio of the stock market capitalization

held by minorities to GNP (External Cap/GNP), the ratio of the sum of bank debt of the private

sector and outstanding non-financial bonds to GNP (Debt/GNP), the ratio of the number of

domestic firms listed in a given country to its population (Domestic Firms/Pop), and the ratio of

the number of the initial public offerings of equity in a given country to its population

(IPOs/Pop). LLSV also find that the law and order tradition (Rule of Law) in each country is an

important determinant of external finance.

In our regression analysis, we also employ the capital market development and legal

environment proxies used by LLSV.30 These results are shown in Table 2-6. In the first column,

we provide firm level OLS regression estimates using the additional proxies, while the second

column provides fixed-effects estimates. Interestingly, we find that the coefficient estimates on

per-capita GNP, external market capitalization plus debt to GNP, and domestic firms to

population are all negative and statistically significant, whereas the coefficient on IPOs to

population is not statistically different from zero. We also find that the coefficient estimates on

the legal origin indicator variables remain significant, while the coefficient on the Rule of Law

variable is not statistically different from zero. It is also interesting to note that the fixed-effects

estimates reported in the second column are consistent with the OLS results shown in the first

column.


30 Due to a lack of debt, IPO, and /or Rule of Law data, we lose Australia, China, Hong Kong,
Pakistan, Switzerland, and Taiwan from the analysis. If we set the missing observations equal to
zero, we obtain similar conclusions.








Due to the high correlation between per-capita GNP and Rule of Law (0.76), we

eliminate per-capita GNP from the specification shown in the third column. In this instance, the

Rule of Law becomes highly significant, while the remaining coefficient estimates are similar to

those reported in the first column. The fixed-effects estimates reported in the fourth column are

also consistent with the OLS results where per-capita GNP is eliminated from the specification.

All in all, we find that value of corporate diversification varies with the level of capital market

development and legal environment.

Accounting Issues

Throughout our analysis, we have used the market-to-sales ratio as a proxy for firm

value. One concern is that our results may be biased by cross-country differences in the

accounting practices that firms employ when they hold either a majority or minority stake in

another firm.31

Whenever a parent company owns a majority stake in another firm, the market value of

the consolidated firm includes the value of its ownership stake in the subsidiary. However,

depending on the accounting practices employed, the sales of the subsidiary may or may not be

fully included as part of the company's consolidated sales. For firms that have a controlling stake

in another firm, there are two basic methods of preparing consolidated financial statements.

Under the proportional method, consolidated sales include only that portion of the subsidiary's

sales that reflects the parent's ownership percentage in the subsidiary.32 In this case, the market-

to-sales ratio is not biased. Alternatively, under the full consolidation method, consolidated sales

include all of the subsidiary's sales, regardless of the parent's ownership percentage. Clearly, this

accounting practice biases downward the market-to-sales ratio. In these circumstances, the net

income earned by the minority shareholders is subtracted out of the consolidated firm's total

3' For examples of the various accounting methods employed across countries, see International
Accounting and Auditing Trends by the Center for International Financial Analysis & Research,
Inc.
32 When this approach is used, the remainder of the subsidiary's sales is attributed to the minority
interest shareholders.








income in order to arrive at consolidated net income. Consequently, whenever the minority

shareholders' share of subsidiary sales is a significant portion of consolidated sales, we would

expect that the market-to-sales ratio would be biased downward under the full consolidation

method.

Another potential problem arises when a company (Company A) owns a minority interest

in another company (Company B), but does not choose to include its proportion of Company B's

sales on its (Company A's) income statement. In these circumstances, Company A's market-to-

sales ratio would be biased upward, since the effects of its ownership in Company B would be

included in its market value but not in its sales. In this situation, Company A's income from

Company B would show up as investment income from unconsolidated affiliates. Therefore,

whenever investment income from unconsolidated affiliates is a significant portion of net income,

the market-to-sales ratio is likely to be upward biased.

For our purposes, these accounting biases are particularly important if the magnitude of

the biases vary across countries and vary between focused and diversified firms. We find that for

5 of the 35 countries (Denmark, Hong Kong, Indonesia, Italy and Malaysia), diversified firms

have a significantly higher proportion of minority interest income as a percentage of sales. The

market-to-sales ratios for these countries tend to be biased downward more often for diversified

firms, which would bias us towards finding a diversification discount in these countries. For 2 of

the 35 countries (France and Switzerland), we find that focused firms have a significantly higher

proportion of income from unconsolidated affiliates as a percentage of sales. The market-to-sales

ratios for these countries tend to be biased upward more often for single segment firms, which

would also bias us towards finding a diversification discount in these countries.


33 When a company owns a 20%-50% stake in another company, it may have the option to
include its proportion of the sales on its income statements. This approach is referred to as the
"proportional method." Alternatively, under the "equity method," the company does not include
the sales on its income statement and instead treats it as an investment in an unconsolidated
affiliate. The "cost method" is generally used when a company has a stake that is less than 20%.
The ability to select a particular accounting treatment varies across countries and across
industries. We thank the referee and Chuck McDonald for bringing these issues to our attention.








To insure that our results are not driven by these accounting biases, we eliminated from

our sample firms where minority interest income is greater than 2% of sales and firms where

investment income from unconsolidated affiliates is greater than 2% of sales. After eliminating

these firms, the link between per-capita GNP and excess value is somewhat stronger and

statistically more significant. Moreover, there still remains a strong link between the legal system

dummies and excess value, although the dummy corresponding to the French legal system is

marginally significant and the Scandinavian legal system dummy is no longer significant.

Ownership and the Value of Corporate Diversification

The results discussed in Section V suggest that corporate diversification is less

costly/more beneficial for firms that are headquartered in countries where capital markets are less

developed. A potential problem with this conclusion is that, so far, we have not explicitly

controlled for agency costs associated with ownership concentration. Indeed, several studies

suggest that firm value is correlated with ownership structure [e.g., Demsetz and Lehn (1985),

Morck, Shleifer and Vishny (1988), Holdemess and Sheehan (1988), and McConnell and Servaes

(1990)] and that ownership structure varies across countries and legal systems [e.g., La Porta,

Lopez-de-Silanes and Shleifer (1997, 1998), LaPorta, Lopez-De-Silanes and Shleifer (1999), and

Claessens, Djankov, Fan and Lang (1998)]. To the extent that ownership concentration affects

firm value, it may also affect the estimated value of corporate diversification. This concern may

be particularly relevant if there is a strong link between ownership concentration and firm value

and if focused and diversified firms have significantly different levels of ownership

concentration.

The exact link between ownership structure and firm value, however, is not entirely clear.

On one hand, it is widely acknowledged that concentrated ownership is likely to reduce the

conflicts that arise when there is a separation between managers and stockholders. This link

suggests a positive relation between firm value and ownership concentration. On the other hand,

concentrated ownership provides large investors with opportunities to exploit minority








shareholders, thereby suggesting at least for some range of values a negative relation between

firm value and ownership concentration. In a recent study, Holdemrness and Sheehan (1998)

conclude that in the United States, legal constraints often effectively limit the actions of majority

shareholders but it is not clear to what extent their conclusions extend outside the U.S. Indeed,

La Porta, Lopez-de-Silanes and Shleifer (1999) suggest that the costs of concentrated ownership

may be particularly meaningfutil in less developed countries where the legal protection provided to

minority shareholders is often quite limited.

An additional concern is that even if ownership concentration levels are similar for both

focused and diversified firms, ownership concentration may still be important if it has a

differential effect on the value of focused and diversified firms. This concern is particularly

relevant if the costs associated with ownership concentration are lower for diversified firms in

less developed capital markets. If this scenario is correct, it raises the possibility that cross-

country variations in the value of corporate diversification can be explained by differences in

capital market development as well as by differences in ownership structure. For example,

smaller diversification discounts (or premiums) in less developed countries may be due to the fact

that diversification is more beneficial in these markets because capital markets are less developed,

enhancing the value of internal capital markets. Alternatively, smaller diversification discounts

(or premiums) in less developed countries may reflect the fact that ownership concentration is

generally higher in these countries, resulting in potentially lower agency costs associated with

corporate diversification. Clearly, these two interpretations are not necessarily mutually

exclusive, but they do again suggest the need to control for ownership concentration when

calculating the value of corporate diversification.

Ownership Data

Worldscope provides firm level ownership data that consists of reported cases where an

individual or institution holds at least five percent of a company's common stock. Summing up

these reported holdings across all shareholders, we obtain a measure of ownership concentration








for each firm.34 While ownership data are available for a subset of firms in our sample, an

important concern arises when using this data. In many cases, there is no clear distinction

between firms where no individual or institution holds a five percent stake and firms that choose

not to report any ownership data. This reporting bias also appears to be systematic in that

ownership data is reported much less regularly among firms headquartered in less developed

countries.35 To insure that this reporting bias does not affect the qualitative nature of our results,

we use two different methods to classify the unreported ownership data. In the first method, we

treat the unreported observations as missing values. Since many of these missing observations

are likely to be for firms without significant ownership concentration, this approach creates an

upward bias in the level of ownership concentration. In the second method, we assign a zero

value to the unreported observations. Using this method, the reported levels of ownership

concentration are downward biased.

The descriptive statistics on ownership concentration are summarized in Table 2-7. The

results in Panel A treat the unreported observations as missing values, while the results in Panel B

treat the unreported observations as zero values. It follows that the average ownership

concentration levels reported in Panel B are systematically lower than those reported in Panel A.

Three major conclusions emerge from Table 2-7. First, there does appear to be an

ownership reporting bias in the Worldscope data. For example, in the low-income countries,

concentrated ownership is reported for only 14% of the firms, whereas this number is 65% for the


4 In addition to total ownership concentration, Lins and Servaes (1998) also separate ownership
holdings into various detailed ownership categories and find their reported conclusions to be
largely similar across the various measures of ownership concentration.

35 Another potentially important problem with the reported ownership data is that in some
countries, cross-ownership holdings and ownership pyramids are fairly common. La Porta,
Lopez-de-Silanes and Shleifer (1999) study ownership concentration structures in considerable
detail and estimate the magnitude of cross-holdings for the twenty largest publicly traded firms in
various countries. As they point out, "the data on corporate ownership are often difficult to
assemble." Since following their approach for all of the firms in our sample is prohibitive, we are
forced to rely on the numbers reported by Worldscope. In this regard, we follow the approach
used by Lins and Servaes (1998) and Claessens, Djankov, Fan and Lang (1998). However, it is
important to note that Worldscope provides only limited ownership data for several countries in
our sample.








firms in the high-income countries and 78% for the firms in the upper-middle income countries.

Second, consistent with previous papers, we do find that average ownership concentration does

vary across countries and legal systems. Generally, we find ownership concentration is higher in

less developed markets and in markets where the legal system tends to provide less protection to

investors.36 Third, we find that while ownership concentration varies across regions and legal

systems, within each region and legal system, unconditional ownership concentration levels are

similar for diversified and focused firms.37 This result tends to suggest that our earlier results on

the effects of capital market development on the value of corporate diversification are not driven

solely by differences in ownership concentration. Nevertheless, in order to more clearly

disentangle the corresponding sources of any diversification discounts or premiums, we need to

control for ownership concentration in our regression analysis.

Regression Results Controlling for Ownership Concentration

Similar to Morck, Shleifer and Vishny (1988) and others, we also account for the

nonlinear relation between ownership structure and firm value by creating three separate

ownership concentration variables:38

OWNOtol0 = total ownership if total ownership <0.10,

= 0.10 if total ownership > 0.10;


36 Putting these two conclusions together also leads us to suspect that among the 35% of firms in
the high income category where ownership data is not reported, a relatively high percentage of
these firms may truly have disparate ownership and that ownership data is truly missing for only a
small subset of these firms. Alternatively, when we consider the 86% of low income firms with
no reported ownership data, we would suspect that a higher percentage of these observations are
truly missing.
7 Statistical tests for differences in the average level of ownership concentration between the
focused and diversified firms are not statistically significant from zero for any of the groups.
38 Morck, Shleifer and Vishny (MSV, 1988) use 5 percent and 25 percent as their breakpoints.
Given that the Worldscope databank does not generally provide firm level ownership
concentration values below 5 percent (aside from the unreported values), we use a 10 percent cut-
off for the first breakpoint and 30 percent as the next breakpoint to be consistent with MSV's
ownership ranges. As additional robustness checks, we also tried other breakpoints and used
ownership concentration dummy variables for each of the breakpoints in place of the MSV
variables. In both cases, we found that the reported conclusions were qualitatively unchanged.








OWN10to30 =0 if total ownership < 0.10,

= total ownership minus 0.10 if 0.10 < total ownership < 0.30,

=0.20 if total ownership > 0.30;

OWNover30 = 0 if total ownership < 0.30,

= total ownership minus 0.30 if total ownership > 0.30.

This classification suggests that the marginal impact of increased ownership

concentration varies depending on whether ownership concentration is less than 10 percent,

between 10 and 30 percent, and greater than 30 percent. We also interact OWN10to30 and

OWNover30 with the dummy variable SEG, which equals one if the firm has multiple segments,

to assess the impact of ownership concentration on the value of corporate diversification.39

Generally, we would expect a positive link between firm value and OWNOto 10. Within this

range, increases in ownership concentration are likely to improve managerial incentives without

dramatically increasing the risks of managerial entrenchment and expropriation. For ownership

concentration levels beyond ten percent, the expected results are less clear. For these firms, the

benefits of increased ownership may be more than offset by the costs resulting from increased

managerial entrenchment and by the potential for the expropriation of minority shareholders.

Consequently, the link between OWN10to30 and OWNover30 and firm value is less clear.

The firm level regression estimates that control for ownership concentration are reported

in Table 2-8. The first three columns [(1) (3)] contain the results where the unreported

observations for ownership concentration are treated as missing. In the last three columns [(4) -

(6)], these observations are treated as zeros. The most striking conclusion that emerges from the

results in Table 2-8 is that even after controlling for ownership concentration, there is still a

strong link between the value of corporate diversification and both the legal system dummies and

per-capita GNP. Moreover, the sign and magnitudes of the estimated coefficients are quite

similar to those reported earlier in Tables 2-5 and 2-6.



39 Note that due to singularity, we do not include OWNOtolO*SEG in our specification.








While it is not the primary focus of our analysis, the estimated coefficients for the

ownership concentration are still of considerable interest. The estimated coefficients vary

somewhat depending on the treatment of the unreported ownership observations. Nevertheless, a

few basic conclusions emerge. First, for low levels of ownership concentration, there is a positive

link between ownership concentration and excess value, although this link is significant only for

the cases where we treat the unreported ownership observations as zeros. Second, for ownership

concentration levels beyond ten percent, we generally find that increases in ownership

concentration lead to a reduction in value for both focused and diversified firms. This result

confirms the fact that there are both costs and benefits associated with increased ownership

concentration.

Finally, in columns (5) and (6), we see from the coefficients on the ownership

concentration variables that are interacted with the diversification dummy (OWN10to30*SEG

and OWNover30*SEG), that the effects of ownership concentration are significantly different for

focused and diversified firms. For ownership concentration levels between 10 and 30 percent,

excess value is significantly lower for the diversified firms, suggesting that entrenchment

problems and expropriation of minority shareholders is more of a concern for diversified firms.

However, beyond 30 percent, excess value is significantly higher for diversified firms. It is

notable, however, that these results do not hold up in columns (2) and (3), where the unreported

ownership observations are treated as zeros.

All in all, the results suggest that there is a link between ownership concentration and

excess value, and that this link may be somewhat different for focused and diversified firms.

However, the exact nature of these links depends critically on the specification and on the

treatment of the unreported ownership observations. It is also important to reiterate that

regardless of the specification, there is strong evidence that the value of corporate diversification

varies depending on the legal system and the level of capital market development.








Conclusion

Using a large database of more than 8,000 companies from 35 countries, we analyze the

link between capital market development and the value of corporate diversification. We find

evidence that the value of corporate diversification is negatively related to the level of capital

market development. Among high-income countries where capital markets are well developed,

we find that diversified firms trade at a significant discount relative to focused firms. This

evidence is consistent with previous studies (Lang and Stulz (1994) and Berger and Ofek (1995))

that have documented a diversification discount for U.S. firms. In contrast, we find that there is

either no diversification discount, or in some cases, a significant diversification premium, in

countries whose capital markets are less developed. Consistent with the recent findings of

LaPorta, Lopez-De-Silanes, Shleifer, and Vishny (1997, 1998), we also find that the value of

diversification depends in an important way on the legal system of the country in which the firm

is established.

Overall, our results suggest that the financial, legal, and regulatory environment all have

an important influence on the value of diversification, and that the optimal organizational

structure for firms operating in emerging markets may be very different than that for firms

operating in more developed countries. In this regard, our results provide support for the

arguments made by Khanna and Palepu (1997), who find that diversified industry groups in India

often outperform their stand-alone counterparts. Our results are also consistent with Lins and

Servaes (1999) who find that diversified firms in Japan and the United Kingdom (countries that

are considered to be developed) generally trade at discounts relative to focused firms.

While we have argued that cross-country variations in the value of diversification vary

with the level of capital market development, our results can be interpreted more broadly. In

addition to providing better access to capital markets, or limiting the need to access these

markets, diversification may provide other important benefits particularly in countries where

the economic and legal system are less developed. If the economic and legal environments make








it more difficult to contract with other firms, it may be more beneficial to merge related

enterprises within the same organization than it is to have them operate on a separate, stand-alone

basis. Diversified firms in these countries may also be better able to attract quality employees

and better able to lobby or influence the political and regulatory process. Ultimately, each of

these explanations may be applicable.

Finally, while we do not address this issue directly, our results indirectly suggest that

global capital markets are not perfectly integrated. Firms in countries that have less developed

capital markets appear to face a higher cost of external capital. One way to mitigate these higher

costs is to adjust the optimal organizational structure. More specifically, for these firms, the

establishment of an internal capital market within a diversified firm may more than offset the

costs of corporate diversification. Clearly, however, there may be other ways to address these

distortions. For example, Lins and Servaes (1999) stress the importance of concentrated

ownership. Other alternatives may include the establishment of private banking relationships

and/or the establishment of the type of interconnected business groups described by Khanna and

Palepu (1997). These issues await future research.








Table 2-1
Economic Development and Legal System Measures by Country: 1991 1995


Average World Bank Market Legal System
Country Per-Capita GNP (US $) Classification Classification
Australia 17,808 High Income English Origin
Austria 23,666 High Income German Origin
Brazil 3,134 Upper-Middle Income French Origin
Canada 20,098 High Income English Origin
Chile 3,206 Upper-Middle Income* French Origin
China 498 Low Income Other
Denmark 26,936 High Income Scandinavian Origin
Finland 21,090 High Income Scandinavian Origin
France 22,808 High Income French Origin
Germany 24,188 High Income German Origin
Hong Kong 18,588 High Income English Origin
India 316 Low Income English Origin
Indonesia 792 Lower-Middle Income" French Origin
Ireland 13,070 High Income English Origin
Italy 19,500 High Income French Origin
Japan 32,232 High Income German Origin
South Korea 7,830 Upper-Middle Income German Origin
Malaysia 3,180 Upper-Middle Income' English Origin
Mexico 3,530 Upper-Middle Income French Origin
Netherlands 21,322 High Income French Origin
New Zealand 13,030 High Income English Origin
Norway 26,812 High Income Scandinavian Origin
Pakistan 432 Low Income English Origin
Philippines 878 Lower-Middle Income French Origin
Portugal 8,350 High Income' French Origin
Singapore 20,266 High Income English Origin
South Africa 2,890 Upper-Middle Income' English Origin
Spain 13,430 High Income French Origin
Sweden 24,960 High Income Scandinavian Origin
Switzerland 36,800 High Income German Origin
Taiwan 10,874 High Income German Origin
Thailand 2,110 Lower-Middle Income English Origin
Turkey 2,404 Lower-Middle Income' French Origin
United Kingdom 17,974 High Income English Origin
United States 24,758 High Income English Origin
Average per-capita GNP (US $) is the five year arithmetic average of per-capita GNP from 1991-
1995. The World Bank income classifications are obtained from the World Tables. The legal
system classification identifies the legal origin of the Company Law or Commercial Code of each
country. The legal system classifications are obtained from La Porta, Lopez-de-Silanes, Shleifer,
and Vishny (1997).

' The World Bank income classifications varied across years for the following countries: Chile
(lower-middle income in 1991), Indonesia (low income in 1991), Malaysia (lower-middle income
in 1991), Portugal (upper-middle income in 1991 and 1992), South Africa (lower-middle income
in 1991), Turkey (upper-middle income in 1993).








Table 2-2
Firm Level Summary Statistics by Development Classifications and Legal System for Single-
Segment and Multi-Segment Firms: 1991 1995

Panel A: Firm Level Characteristics by World Bank Market Classifications
Firm Level Single-Segment Firms Multi-Segment Firms Statistical Differences
Characteristics by (p-values)
Development Median Mean Median Mean Median Mean
Classifications______
High Income
Number of Assets 1.000 1.000 2.000 2.554 0.000 0.000
Total Assets 276 1755 641 1906 0.000 0.015
(mil $)_____________________________
Total Capital 180 1727 380 1844 0.000 0.126
(mil $)_________ ______
Leverage Ratio 0.265 0.326 0.287 0.374 0.714 0.633
Operating 0.117 0.134 0.104 0.115 0.608 0.527
Income/Sales
Capital 0.049 0.111 0.046 0.078 0.872 0.389
Expenditure/Sales
Market/Sales 1.042 1.738 0.844 1.211 0.073 0.054
Observations 17,366 17,366 8,159 8,159_
Upper-Middle
Income
Number of 1.000 1.000 3.000 2.958 0.000 0.000
Segments
Total Assets 435 1620 931 2513 0.000 0.000
(mil $)_________
Total Capital 460 1453 664 1776 0.047 0.000
(mil $)__________________________
Leverage Ratio 0.149 0.217 0.180 0.225 0.526 0.782
Operating 0.148 0.167 0.145 0.166 0.813 0.938
Income/Sales
Capital 0.083 0.172 0.090 0.167 0.726 0.739
Expenditure/Sales
Market/Sales 1.385 2.348 1.575 1.732 0.107 0.061
Observations 1,209 1,209 336 336








Panel A--continued
Firm Level Single-Segment Firms Multi-Segment Firms Statistical Differences
Characteristics by (p-values)
Development Median Mean Median Mean Median Mean
Classifications _____
Lower-Middle
Income
Number of 1.000 1.000 3.000 2.684 0.000 0.000
Segments _______
Total Assets 329 1769 610 2571 0.013 0.000
(mil $)_________
Total Capital 209 1571 412 1531 0.000 0.562
(mil $)
Leverage Ratio 0.174 0.268 0.199 0.231 0.732 0.824
Operating 0.185 0.195 0.190 0.205 0.824 0.879
Income/Sales ______
Capital 0.101 0.227 0.067 0.255 0.213 0.307
Expenditure/Sales ______
Market/Sales 1.595 2.459 1.295 2.371 0.114 0.331
Observations 937 937 79 79
Low Income
Number of 1.000 1.000 3.000 2.833 0.000 0.000
Segments_ __
Total Assets 284 1452 838 2480 0.000 0.000
(mil $)______
Total Capital 174 1101 545 1659 0.000 0.000
(mil $)________
Leverage Ratio 0.352 0.388 0.425 0.393 0.431 0.917
Operating 0.149 0.172 0.127 0.142 0.267 0.169
Income/Sales
Capital 0.075 0.168 0.084 0.188 0.698 0.544
Expenditure/Sales I
Market/Sales 1.402 1.948 1.414 1.711 0.329 0.122
Observations 710 710 90 90








Panel B: Firm Level Characteristics by Legal Systems
Firm Level Single-Segment Firms Multi-Segment Firms Statistical Differences
Characteristics by (p-values)
Legal Systems Median Mean Median Mean Median Mean
English Origin
Number of Assets 1.000 1.000 2.000 2.554 0.000 0.000
Total Assets 276 1755 641 1906 0.000 0.015
(mil $) ____________________________
Total Capital 180 1727 380 1844 0.000 0.126
(mil $)___________________
Leverage Ratio 0.265 0.326 0.287 0.374 0.714 0.633
Operating 0.117 0.134 0.104 0.115 0.608 0.527
Income/Sales
Capital 0.049 0.111 0.046 0.078 0.872 0.389
Expenditure/Sales ____________________
Market/Sales 1.042 1.738 0.844 1.211 0.073 0.054
Observations 17,366 17,366 8,159 8,159______
French Origin
Number of 1.000 1.000 3.000 2.958 0.000 0.000
Segments
Total Assets 435 1620 931 2513 0.000 0.000
(mil $)__________________
Total Capital 460 1453 664 1776 0.047 0.000
(mil $)__________ ______
Leverage Ratio 0.149 0.217 0.180 0.225 0.526 0.782
Operating 0.148 0.167 0.145 0.166 0.813 0.938
Income/Sales
Capital 0.083 0.172 0.090 0.167 0.726 0.739
Expenditure/Sales
Market/Sales 1.385 2.348 1.575 1.732 0.107 0.061
Observations 1,209 1,209 336 336








Panel B--continued
Firm Level Single-Segment Firms Multi-Segment Firms Statistical Differences
Characteristics by (p-values)
Legal Systems Median Mean Median Mean Median Mean
German Origin
Number of 1.000 1.000 3.000 2.684 0.000 0.000
Segments
Total Assets 329 1769 610 2571 0.013 0.000
(mil $)
Total Capital 209 1571 412 1531 0.000 0.562
(mil $)
Leverage Ratio 0.174 0.268 0.199 0.231 0.732 0.824
Operating 0.185 0.195 0.190 0.205 0.824 0.879
Income/Sales
Capital 0.101 0.227 0.067 0.255 0.213 0.307
Expenditure/Sales
Market/Sales 1.595 2.459 1.295 2.371 0.114 0.331
Observations 937 937 79 79
Scandinavian
Origin
Number of 1.000 1.000 3.000 2.833 0.000 0.000
Segments
Total Assets 284 1452 838 2480 0.000 0.000
(mil $)____
Total Capital 174 1101 545 1659 0.000 0.000
(mil $)______
Leverage Ratio 0.352 0.388 0.425 0.393 0.431 0.917
Operating 0.149 0.172 0.127 0.142 0.267 0.169
Income/Sales
Capital 0.075 0.168 0.084 0.188 0.698 0.544
Expenditure/Sales
Market/Sales 1.402 1.948 1.414 1.711 0.329 0.122
Observations 710 710 90 90
In Panel A, firms are classified each year by their country's World Bank market classification,
while in Panel B firms are classified by their country's legal system. Single-segment firms are
firms that operate in only one two-digit SIC code industry. Multi-segment firms are defined as
firms that operate in two or more two-digit SIC code industries and no firm segment sales exceed
90% of total firm sales. The leverage ratio is defined as book value of debt divided by total
assets. Market-to-sales is defined as the ratio of a firm's market value of equity plus book value
of debt to its total sales.








Table 2-3
Excess Values by Development Groups, Broader Per-Capita GNP Groups and Legal Systems for
Single-Segment and Multi-Segment Firms: 1991 1995

Panel A: Excess Values by World Bank Market Classification for Single-Segment and Multi-
Segment Firms
Firm Level Single-Segment Multi-Segment Firms Statistical
Characteristics by Firms Differences
Development (p-values)
Classification Median Mean Median Mean Median Mean
High Income 0.0000 0.0199 -0.0576 -0.0584 0.000 0.000
Upper-Middle Income 0.0000 0.0070 -0.0722 -0.0181 0.051 0.398
Lower-Middle Income 0.0000 0.0330 0.0863 0.0543 0.032 0.721
Low Income 0.0000 0.0100 0.0380 0.0945 0.161 0.005
Observations
High Income 17,366 17,366 8,159 8,159
Upper-Middle Income 1,209 1,209 336 336
Lower-Middle Income 937 937 79 79
Low Income 710 710 90 90


Panel B: Excess Values by Per-Capita GNP for Single-Segment and Multi-Segment Firms
Single-Segment Multi-Segment Firms Statistical
Excess Values Firms Differences
by Per-Capita GNP (p-values)
Median Mean Median Mean Median Mean
Per-Capita GNP > 0.0000 0.0211 -0.0578 -0.0579 0.000 0.000
$15,000
$15,000 > Per-Capita 0.0000 -0.0026 -0.0542 -0.0281 0.136 0.488
GNP > $5,000________
Lo$5,000 > Per-Capita 0.0000 0.0260 -0.0400 -0.0264 0.148 0.112
GNP > $1,000
$1,000 > Per-Capita 0.0000 0.0068 0.0541 0.0841 0.101 0.014
GNP
Observations
Per-Capita GNP > 16,543 16,543 8,072 8,072
$15,000
$15,000 > Per-Capita 1,069 1,069 164 164
GNP > $5,000_____
Lo$5,000 > Per-Capita 1,643 1,643 306 306
GNP > $1,000_ _
$1,000 > Per-Capita 967 967 122 122
GNPIIII








Panel C: Excess Values by Legal Systems for Single-Segment and Multi-Segment Firms
Single-Segment Multi-Segment Firms Statistical
Excess Values Firms Differences
by Legal Systems (p-values)
Median Mean Median Mean Median Mean
English Origin 0.0000 0.0088 -0.0576 -0.0584 0.000 0.000
French Origin 0.0000 0.0287 -0.0722 -0.0181 0.051 0.398
German Origin 0.0000 0.0322 0.0863 0.0543 0.032 0.721
Scandinavian Origin 0.0000 0.0050 0.0380 0.0945 0.161 0.005
Observations
English Origin 14,931 14,931 6,207 6,207 ______
French Origin 2,378 2,378 843 843
German Origin 2,108 2,108 1,290 1,290
Scandinavian Origin 683 683 368 368
In Panel A, firms are classified each year by their country's World Bank market classification,
while in Panel B firms are classified by broader per-capita GNP groups. In Panel C, firms are
classified by their country's legal system. Excess value is defined as the natural logarithm of the
ratio of a firm's market-to-sales ratio to its imputed market-to-sales ratio. Firms with excess
values that are greater than four or less than one-fourth are eliminated from the sample. Single-
segment firms are firms that operate in only one two-digit SIC code industry. Multi-segment
firms are defined as firms that operate in two or more two-digit SIC code industries and no firm
segment sales exceed 90% of total firm sales.








Table 2-4
Country Level Regression Estimates of Excess Values: 1991 1995


Country Constant SEG OIS CES ASSETS Adj R2 Obs
Australia -0.520* -0.152*** 0.192* 0.328*** 0.022 059 9
(-1.810) (-3.021) (1.859) (4.236) (1.464) 059 596
Austria 0.666 -0.211** 0.661** 0.074 -0.035
______ (0.942) (-1.970) (2.163) (0.252) (-1.063) 005 129
Brazil -1.447*** -0.075 0.141 -0.031 0.124*** 0.080 245
(-4.417) (-0.700) (0.730) (-1.511) (4.601)
Canada -1.468*** -0.059* 0.027 0.192*** 0.073***
(-7.616) (-1.665) (0.430) (5.701) (7.413) 0062 15
Chile -1.128* -0.289 -0.016 0.048*** 0.053** 0.091 118
(-1.830) (-1.300) (-0.534) (2.533) (2.225) 091 11
China 0.392 0.221 0.065 0.275 -0.019
(0.307) (0.447) (0.158) (1.366) (-0.323) 0000 7
Denmark -0.165 -0.063 2.016*** 0.619*** -0.003
(-0.398) (-0.937) (5.432) (3.925) (-0.163) 0.137 270
Finland -0.356 -0.016 0.881*** 0.448** 0.012
(-1.291) (-0.282) (2.939) (2.119) (0.828) 0.075 209
France -0.953*** -0.085*** -0.004 0.052 0.052*** 0.038 1,131
(-4.924) (-2.502) (-0.539) (1.378) (5.821) 0.038 31
Germany 0.212 -0.050 0.807*** 0.223*** -0.009
(1.270) (-1.568) (7.054) (3.370) (-1.094) 0.039 1,296
Hong -0.235 0.145*** 0.214 0.065 0.005 -- -
Kong (-0.574) (2.786) (1.176) (0.838) (0.271) 0.021 374
India -1.940*** -0.011 2.184*** 0.260*** 0.080*** 0.235 553
S(-5.355) (-0.175) (9.863) (3.491) (5.335) 0235 553
Indonesia -1.927*** -0.104 0.400 0.236*** 0.082*** -
(-2.715) (-1.011) (1.381) (2.892) (2.896) 0.117 218
Ireland -1.646*** -0.003 0.377*** 0.074*** 0.084*** 0.116 179
______ (-4.148) (-0.037) (2.543) (2.585) (3.858) 0116 179
Italy 1.101* 0.073 0.322 0.293 -0.041* --
(1.917) (1.099) (0.948) (0.978) (-1.935) 0.015 259
Japan 2.658*** -0.039* 2.908*** -0.697*** -0.108***
(11.772) (-1.635) (12.204) (-2.760) (-12.588) 0.237 1,137
Malaysia 0.864** 0.063 1.172*** 0.121* -0.046*** 0 -5
______ (2.484) (1.274) (6.343) (1.690) (-2.583) 0.103 527
Mexico -0.342 -0.102 0.168 0.412 0.007
______(-0.439) (-1.029) (0.530) (1.289) (0.201) 0.021 108








Table 2-4--continued

Country Constant SEG OIS CES ASSETS Adj Obs
___R'__________2
Netherlands -0.787*** -0.040 1.406*** -0.165 0.039***
(-2.819) (-0.798) (4.446) (-0.668) (2.839) 0 3
New 0.773 -0.214* 0.756** -0.584 -0.446
Zealand (1.408) (-1.942) (2.511) (-0.549) (-1.579) 0.
Norway 0.377 0.172** 0.544*** 0.347*** -0.023
(0.757) (2.145) (2.793) (4.889) (-0.945) 0137
Pakistan -0.269 0.606*** 0.868*** 0.021 0.013
(-0.380) (3.188) (2.694) (0.083) (0.406) 0153 134
Philippines -0.478 0.430 0.064 0.116 0.021
(-0.598) (1.297) (0.430) (1.068) (0.556) 00
Portugal -1.205 0.031 0.774* 0.353 0.044
(-1.166) (0.094) (1.902) (0.886) (1.057) 013 4
Singapore -0.323 0.132*** 0.561*** 0.110 0.004
(-0.906) (2.570) (2.987) (1.019) (0.224) 06 3
South Korea 1.349** 0.066 0.324 0.643*** -0.052** 0.053 264
(2.011) (1.057) (1.000) (3.613) (-2.052) ___
South 0.178 -0.072 1.860*** 0.411*** -0.030 0184 305
Africa (0.423) (-1.032) (7.221) (4.710) (-1.452) 014 305
Spain -0.393 -0.307*** 0.327** 0.104 0.013
______(-0.800) (-3.206) (2.171) (0.654) (0.620) 0.043 320
Sweden -0.963*** -0.174*** 1.327*** -0.174 0.048***
______(-2.784) (-3.066) (3.509) (-1.040) (3.093) 0.111 337
Switzerland -1.067*** 0.016 1.180*** 0.239 0.048***
_______(-3.104) (0.313) (4.623) (1.144) (2.881) 0.107 358
Taiwan 0.697 0.170 0.743*** 0.096 -0.032
______(1.104) (1.239) (3.521) (0.566) (-1.208) 0.064 214
Thailand -0.325 -0.094 0.069 0.126** 0.021
(-0.774) (-0.659) (1.245) (2.506) (1.055) 0.022 460
Turkey -1.114 -0.688* -0.201 1.520*** 0.033
(-0.901) (-1.653) (-0.531) (4.037) (0.772) 0.247 67
United -0.572*** -0.067*** 0.251*** 0.575*** 0.029*** 0
Kingdom (-6.744) (-3.734) (6.470) (10.742) (6.221) 0.056 4,951
United -0.250*** -0.132*** 0.393*** 0.596*** 0.008** 05 1
States (-3.827) (-10.548) (13.216) (15.383) (2.475) 0.059 11461
Significant at 1 percent (***), 5 percent (**), and 10 percent (*) levels.

We estimate the following regression model from 1991-1995 for each of the thirty-five individual
countries in our sample:
Excess Value = o + 31 (Diversification Dummy) + 1N (Log Assets) + 03 (Operating Income /
Sales) + N4 (Capital Expenditures / Sales) + e.

Excess value is defined as the natural logarithm of the ratio of a firm's market-to-sales ratio to its
imputed market-to-sales ratio. Firms with excess values that are greater than four or less than
one-fourth are eliminated from the sample. The diversification dummy, SEG, is equal to one for
multi-segment firms and zero otherwise. Multi-segment firms are defined as firms that operate in
two or more two-digit SIC code industries and no firm segment sales exceed 90% of total firm
sales. OIS is defined as the firm's operating income-to-sales, while CES is the firm's capital





43


expenditures-to-sales. Assets are defined as the natural logarithm of the firm's total assets. The
regressions also include year dummies for 1992-1995.








Table 2-5
Firm Level Regression Estimates of Excess


Values: 1991


Variables OLS OLS OLS Fixed Effects
(1) (2) (3) (4)
Constant -0.270*** -0.232*** -0.240*** ___
~____~___ (-10.677) (-8.726) (-9.040)______
Multi-Segment -0.078*** -0.096*** -0.004 0.037
Dummy (SEG) (-10.748) (-11.584) (-0.201) (1.048)
Operating Income- 0.042*** 0.043*** 0.043*** -0.006
to-Sales (OIS) (6.540) (6.702) (6.701) (-0.372)
Capital 0.226*** 0.226*** 0.225*** 0.144***
Expenditures-to- (18.982) (18.982) (18.898) (11.733)
Sales (CES)_______
Log of Total 0.014*** 0.012*** 0.012*** -0.027***
Assets (ASSETS) (11.424) (9.343) (9.565) (-5.571)
Per-Capita GNP -___ ___ 0.426*** -0.682***
(GNPCAP*SEG)a (-4.252) (4.423)
Low Income ___ 0.142** ___
Dummy (2.398)
(G 1 *SEG)_______
Lower-Middle ___ 0.036 _
Income Dummy (0.564)
(G2*SEG)________
Upper-Middle 0.002 _
Income Dummy (0.063)
(G3*SEG)
French Legal ___0.052*** 0.047** -0.033
System Dummy (2.509) (2.268) (-0.703)
(FRENCH*SEG)
German Legal ___ 0.072*** 0.099*** 0.lll***
System Dummy (4.030) (5.170) (2.763)
(GERMAN*SEG)
Scandinavian 0.044 0.058* 0.013
Legal System (1.449) (1.919) (0.226)
Dummy
(SCAND*SEG)
Adjusted R2 0.025 0.025 0.026 0.027
NumObserv ofs 28,886 28,886 28,886 28,886
Observations
Significant at 1 percent (***), 5 percent (**), and 10 percent (*) levels.
a coefficient estimate x 10-5

Regression estimates are from 1991-1995. Excess value is defined as the natural logarithm of the
ratio of a firm's market-to-sales ratio to its imputed market-to-sales ratio. Firms with excess
values that are greater than four or less than one-fourth are eliminated from the sample. The
diversification dummy, SEG, is equal to one for multi-segment firms and zero otherwise. Multi-
segment firms are defined as firms that operate in two or more two-digit SIC code industries and
no firm segment sales exceed 90% of total firm sales. SEGN is the number of firm segments
defined at the two-digit SIC code level. GNPCAP is the annual per-capita GNP of the country
where the firm is headquartered. G l-G3 are dummy variables corresponding to each of the


1995





45


World Bank income groups. French, German, and Scandinavian are dummy variables
corresponding to each legal system. The dummy variables are equal to one for each
corresponding classification and zero otherwise. Per-capita GNP, the World Bank income group
dummies, and the legal system dummies are interacted with the multi-segment dummy (SEG) for
the all firms panel and the number of segments (SEGN) for the multi-segment firms panel.
Models 1-3 are estimated over 1991-1995 using ordinary least squares. Column (4) provides
fixed-effects estimates (within-firm estimates) of Model 3. Each model specification also
includes year dummies for 1992-1995.








Table 2-6
Firm Level Regression Estimates of Excess Values using Additional Proxies for Capital Market
Development and the Legal Environment: 1991 1995


Variables OLS Fixed Effects OLS Fixed Effects
(1) (2) (3) (4)
Constant -0.237*** -___ 0.226***
________(-8.665) _______ (-8.435)
Multi-Segment -0.164** -0.156 -0.063** -0.103*
Dummy (SEG) (-1.942) (-1.404) (-2.099) (-1.725)
Operating Income- 0.042*** 0.011* 0.042*** 0.010*
to-Sales (OIS) (6.636) (1.688) (6.645) (1.657)
Capital 0.223*** 0.260*** 0.224*** 0.260***
Expenditures-to- (18.769) (11.673) (18.804) (11.664)
Sales (CES)_____________
Log of Total 0.012*** 0.007*** 0.012*** 0.007***
Assets (ASSETS) (9.279) (3.136) (9.086) (3.234)
Per-Capita GNP -0.420** -0.467** _
(GNPCAP*SEG)a (-2.131) (-2.099)
[(MKTCAP + -0.054** -0.069* -0.054** -0.056**
Debt)/GNP]*SEG (-1.976) (-1.684) (-2.352) (-2.031)
(Domestic -0.002*** -0.002** -0.002*** -0.003**
Firms/Pop)*SEG (-3.150) (-2.404) (-2.743) (-1.983)
(IPOs/Pop)*SEG 0.013 0.005 0.016 0.003
(1.083) (0.639) (1.345) (0.537)
French Legal 0.066*** 0.035** 0.058*** 0.040**
System Dummy (4.779) (2.341) (4.574) (2.185)
(FRENCH*SEG)_______________________
German Legal 0.082*** 0.042** 0.061*** 0.059***
System Dummy (5.284) (2.347) (4.893) (2.991)
(GERMAN*SEG)__________
Scandinavian 0.034*** 0.041** 0.033*** 0.031**
Legal System (3.679) (2.254) (3.633) (2.010)
Dummy
(SCAND*SEG)
(Rule of Law)*SEG -0.005 -0.024 -0.025*** -0.016**
(-0.464) (-1.422) (-4.017) (-2.322)
Adjusted R2 0.027 0.040 0.026 0.038
Number of
Observations 27,132 27,132 27,132 27,132


oigniilcant at i percent -*-)j, 3 percent
a coefficient estimate x 105


(* ), and 10 percent (*) levels.


Regression estimates are from 1991-1995. Excess value is defined as the natural logarithm of the
ratio of a firm's market-to-sales ratio to its imputed market-to-sales ratio. Firms with excess
values that are greater than four or less than one-fourth are eliminated from the sample. The
diversification dummy, SEG, is equal to one for multi-segment firms and zero otherwise. Multi-
segment firms are defined as firms that operate in two or more two-digit SIC code industries and
no firm segment sales exceed 90% of total firm sales. GNPCAP is the annual per-capita GNP of
the country where the firm is headquartered. French, German, and Scandinavian are dummy





47


variables corresponding to each legal system. The dummy variables are equal to one for each
corresponding classification and zero otherwise. From LaPorta, Lopez-De-Silanes, Shleifer, and
Vishny (1997), we obtain for each country the ratio of the stock market capitalization held by
minorities plus the sum of bank debt of the private sector and outstanding non-financial bonds to
GNP (MKTCAP + Debt/GNP), the ratio of the number of domestic firms listed in a given country
to its population (Domestic Firms/Pop), and the ratio of the number of the initial public offerings
of equity in a given country to its population (IPOs/Pop). From LLSV, we also obtain the law
and order tradition (Rule of Law) in each country. Columns 1 and 3 provide OLS estimates over
1991-1995, while columns 2 and 4 provide fixed-effects estimates (within-firm estimates). Each
specification also includes year dummies for 1992-1995.








Table 2-7
Descriptive Statistics on Ownership Concentration by Economic Development and Legal System:
1991 -1995

Panel A: Ownership Concentration for Subset of Firms Reporting Ownership Holdings Greater or
Equal to 5%


Groups Total Single-Segment Firms
Mean Median Mean Median
Economic
Development:
High Income 0.40 0.38 0.41 0.39

Upper-Middle 0.52 0.54 0.54 0.57
Income
Lower-Middle 0.63 0.65 0.64 0.66
Income
Low Income 0.59 0.60 0.58 0.60

Legal System:

English Origin 0.39 0.36 0.40 0.38

French Origin 0.58 0.60 0.59 0.61

German Origin 0.43 0.38 0.44 0.42

Scandinavian 0.44 0.44 0.45 0.46
Origin








Panel A--continued

% of Total Sample Reporting
Groups Multi-Segment Firms Ownership
Mean Median
Economic
Development:
High Income 0.38 0.34 65%

Upper-Middle 0.46 0.46 78%
Income
Lower-Middle 0.61 0.61 34%
Income
Low Income 0.65 0.69 14%

Legal System:

English Origin 0.36 0.32 61%

French Origin 0.55 0.56 64%

German Origin 0.40 0.35 72%

Scandinavian Origin 0.42 0.40 73%








Panel B: Ownership Concentration Set Equal to Zero where Not Reported


Groups Total Single-Segment Firms Multi-Segment Firms
Mean Median Mean Median Mean Median
Economic
Development:
High Income 0.32 0.29 0.33 0.31 0.30 0.26

Upper-Middle 0.43 0.48 0.44 0.51 0.41 0.44
Income
Lower-Middle 0.22 0.00 0.21 0.00 0.41 0.00
Income
Low Income 0.08 0.00 0.08 0.00 0.11 0.00

Legal System:

English Origin 0.29 0.27 0.30 0.28 0.28 0.24

French Origin 0.44 0.51 0.43 0.51 0.47 0.51

German Origin 0.33 0.24 0.33 0.22 0.32 0.24

Scandinavian 0.39 0.39 0.41 0.42 0.37 0.36
Origin
Worldscope provides firm level ownership data that consists of reported cases where an
individual or institution holds at least five percent of a company's common stock. Summing up
these reported holdings, we obtain ownership concentration. We use two different methods to
classify the unreported ownership data. In the first method (Panel A), we treat the unreported
observations as missing values. In the second method (Panel B), we treat the unreported
observations as zero values.








Table 2-8
Firm Level Regression Estimates of Excess Values Controlling for Ownership Concentration:
1991 -1995


Subset of Firms Reporting Ownership Concentration Greater or Equal
to 5%
Variables OLS OLS Fixed Effects
(1) (2) (3)
Constant -0.073 -0.065 ___
~~____~___(-1.223) (-1.088)
Multi-Segment -0.041 -0.085** -0.084**
Dummy (SEG) (-1.504) (-2.392) (-2.400)
Operating Income-to- 0.350*** 0.349*** 0.351***
Sales (OIS) (16.700) (16.662) (16.718)
Capital Expenditures- 0.213*** 0.213*** 0.212***
to-Sales (CES) (13.287) (13.302) (13.309)
Log of Total Assets 0.005*** 0.006*** 0.006***
(ASSETS) (3.438) (3.632) (3.646)
Per-Capita GNP -0.378*** -0.327*** -0.329***
(GNPCAP*SEG)' (-3.072) (-2.619) (-2.644)
French Legal System 0.086*** 0.072*** 0.072***
Dummy (3.358) (2.747) (2.735)
(FRENCH*SEG)________________________
German Legal System 0.136*** 0.128*** 0.129***
Dummy (5.935) (5.528) (5.573)
(GERMAN*SEG)_______
Scandinavian Legal 0.097** 0.089*** 0.088***
System Dummy (2.712) (2.492) (2.493)
(SCAND*SEG)
Ownership 0.079 0.055 0.051
Concentration < 10 (0.145) (0.102) (0.094)
(OWNOtolO)
Ownership -0.289*** -0.336*** -0.344***
Concentration 10-30 (-3.387) (-3.416) (-3.500)
(OWN10to30)
Ownership -0.054** -0.077** -0.068**
Concentration > 30 (-1.968) (-2.373) (-2.129)
(OWNover30)
Ownership 0.142 0.155
Concentration 10-30 (0.950) (1.040)
interacted with SEG
(OWNl0to30*SEG)
Ownership ___ 0.089 0.081
Concentration > 30 (1.468) (1.343)
interacted with SEG
(OWNover30*SEG)
Adjusted R2 0.040 0.041 0.042
Number of 5 1825 1825
Observations 18,225 18,225 18,225








Table 2-8-continued
Ownership Concentration Set Equal to Zero where Not Reported

Variables OLS OLS Fixed Effects
__________________(4) (5) (6)
Constant -0.126*** -0.134*** ___
(-4.078) (-4.341)
Multi-Segment -0.026 -0.007 -0.006
Dummy (SEG) (-1.137) (-0.258) (-0.249)
Operating Income-to- 0.303*** 0.303*** 0.304***
Sales (OIS) (17.138) (17.142) (17.185)
Capital Expenditures- 0.160*** 0.160*** 0.161***
to-Sales (CES) (12.224) (12.254) (12.250)
Log of Total Assets 0.006*** 0.006*** 0.006***
(ASSETS) (4.368) (4.450) (4.466)
Per-Capita GNP -0.336*** -0.331*** -0.331***
(GNPCAP*SEG)a (-3.237) (-3.175) (-3.200)
French Legal System 0.067*** 0.063*** 0.063***
Dummy (2.887) (2.670) (2.656)
(FRENCH*SEG)
German Legal System 0.107*** 0.102*** 0.103***
Dummy (5.292) (4.999) (5.039)
(GERMAN*SEG)
Scandinavian Legal 0.079** 0.080*** 0.080***
System Dummy (2.439) (2.469) (2.468)
(SCAND*SEG)
Ownership 0.550*** 0.537*** 0.533***
Concentration < 10 (3.915) (3.815) (3.788)
(OWNOtol0)
Ownership -0.313*** -0.213*** -0.219***
Concentration 10-30 (-4.033) (-2.479) (-2.556)
(OWN10to30)
Ownership -0.043 -0.083*** -0.074**
Concentration > 30 (-1.590) (-2.613) (-2.364)
(OWNover30)
Ownership -_ 0.300*** -0.289***
Concentration 10-30 (-2.689) (-2.599)
interacted with SEG
(OWN10to30*SEG)
Ownership 0___ .132** 0.124**
Concentration > 30 (2.248) (2.129)
interacted with SEG
(OWNover30*SEG)
Adjusted R2 0.035 0.035 0.036
Number of 2, 2,8
Observations 28,886 28,886 28,886
.ionifinant .t 1 +,-,,+ t***\ C ..C **x j n n......u e 1) e ,oe e m t ...


," k..... i ^, ,ii t ), j pc cint k ), and 10 percent k ) levels, coefilcient estimate x 10
Regression estimates are from 1991-1995. Excess value is defined as the natural logarithm of the
ratio of a firm's market-to-sales ratio to its imputed market-to-sales ratio. Firms with excess
values that are greater than four or less than one-fourth are eliminated from the sample. The





53


diversification dummy, SEG, is equal to one for multi-segment firms and zero otherwise. Multi-
segment firms are defined as firms that operate in two or more two-digit SIC code industries and
no firm segment sales exceed 90% of total firm sales. GNPCAP is the annual per-capita GNP of
the country where the firm is headquartered. French, German, and Scandinavian are dummy
variables corresponding to each legal system. The dummy variables are equal to one for each
corresponding classification and zero otherwise. We use two different methods to classify
unreported the ownership data. In the first method (columns 1-3), we treat the unreported
observations as missing values. In the second method (columns 4-6), we treat the unreported
observations as zero values. OWNOto 10: = total ownership if total ownership < 0.10, = 0.10 if
total ownership > 0.10; OWN10to30: = 0 if total ownership < 0.10, = total ownership minus
0.10 if 0.10 < total ownership < 0.30, = 0.20 if total ownership > 0.30; OWNover30: = 0 if
total ownership < 0.30, = total ownership minus 0.30 if total ownership > 0.30. Column (3) and
(6) provide fixed-effects estimates (within-firm estimates) of Model 2. Each model specification
also includes year dummies for 1992-1995.














CHAPTER 3
FIRM VALUE AND INTERNATIONAL DIVERSIFICATION

Introduction

Over the past twenty-five years, foreign investment by corporations in the industrialized

nations has grown dramatically. Specifically, net foreign investment by firms in OECD countries

has grown from around $9 billion U.S. dollars in 1975 to $154 billion U.S. dollars in 1997

(Global Development Finance, 1999). Indeed, for many companies today, foreign investment

now represents a considerable portion of their overall sales and profits.

This increase in foreign investment has occurred for a number of reasons. Most notably,

lower transaction costs, improved communications, and increasingly integrated capital markets

have lowered the cost of doing business in foreign markets. To the extent that firms are able to

leverage their operations worldwide, international investment may enable them to capture

valuable operating synergies. International diversification may also provide important financial

synergies to the extent it is efficient for multinational firms to raise external capital and then

allocate it among their various global operations using internal capital markets. Multinational

firms also provide investors with a vehicle to diversify their investments internationally without

having to directly invest in foreign markets, although it is unclear if any of this benefit accrues to

the multinational itself.

At the same time, firms often incur additional costs and risks when investing in foreign

markets. While geographic diversification may reduce corporate risk, multinational firms also

have to contend with exchange rate risk, political risk, and the costs incurred when managing

resources over a larger geographic area. So while it is widely acknowledged that companies face

additional benefits and costs when investing overseas, it is unclear whether geographically








diversified firms trade at a premium or discount relative to firms that operate within a single

market.

Recently, several studies have found that U.S. firms that diversify along product lines

trade at a discount relative to focused firms [e.g., Lang and Stulz (1994) and Berger and Ofek

(1995)]. Other studies [e.g., Lins and Servaes (1999) and Fauver, Houston, and Naranjo (1999)]

have found that this result extends to other industrialized countries.' However, many multi-

product firms are also geographically diversified. Thus, it is difficult to disentangle the effects of

product market diversification and geographic diversification without simultaneously controlling

for the two effects. In particular, since the level of product market diversification influences firm

value, it is important to also control for the level of product market diversification when

evaluating the benefits of geographic diversification.

The existing evidence regarding the corporate benefits of international diversification has

so far yielded mixed and inconclusive results.2 Bodnar, Tang, and Weintrop (1998), for example,

use a large sample of U.S. firms over the time period from 1987 to 1993, and find that

geographically diversified firms have higher values relative to comparable single-product

domestic firms. Interestingly, they also find that the product market diversification discount

becomes less pronounced after controlling for whether or not the firm is geographically

diversified. Using Tobin's Q and a sample of U.S. firms, Morck and Yeung (1991, 1998) also

find a positive relation with geographic diversification and a negative relation with industrial

diversification. Similarly, Errunza and Senbet (1981, 1984) determine a positive relation between

the degree of international involvement and excess value. On the other hand, Denis, Denis, and




'Fauver, Houston, and Naranjo (1999), however, find that these results do not necessarily extend
to countries that have less developed capital markets.
2 Many of the earlier studies, moreover, do not simultaneously control for the effects of product
market and geographic diversification. For instance, Kim and Lyn (1986) find there exists a
positive relation between the excess value of a multinational corporation and the degree of
international involvement, but they do not simultaneously control for industry effects.








Yost (1999) find that geographic and industrial diversification, separately, lowers value, but that

firms that diversify both industrially and geographically experience an increase in value.

In this paper, we further investigate the connection between product and geographic

diversification by examining firms that are headquartered in Germany, Japan, the U.K., and the

U.S.3 We also extend earlier studies by controlling for the firm's ownership structure, since

previous studies have found that a firm's ownership structure plays a significant role in affecting

the value of corporate diversification (e.g., Denis, Denis, and Sarin (1997), Claessens, Djankov,

Fan and Lang (1998) and Fauver, Houston, and Naranjo (1999)). Lastly, we also employ both

domestic and international benchmarks in assessing the value of geographic diversification.

In our analysis, we have collected data for more than 4,000 firms located in four

industrialized countries (Germany, Japan, the U.K. and U.S.) over the time period 1991-1995.

Using the methodology adopted by Berger and Ofek (1995), we calculate the implied value of

geographic and product diversification. Our regression analysis also controls for the firm's size,

profitability, capital intensity, and ownership structure.

Our results suggest that product market diversification has a negative effect on firm value

in three of the four countries (Japan, the U.K., and the U.S.). These results directly parallel the

results reported by Lins and Servaes (1999) and Fauver, Houston, and Naranjo (1999), each of

whom found that focused firms performed better in Japan, the U.K., and the U.S., while product

market diversification had no significant effect on German firms. Moreover, after controlling for

firm characteristics and the level of product market diversification, we find that multinational

firms trade at a premium relative to firms that operate in a single domestic market in two of the

four countries (Japan and the U.S.). These results confirm Bodnar, Tang, and Weintrop's

evidence regarding U.S. firms, but they also suggest that the observed value of geographic

diversification may not extend to multinationals throughout the world.



3 Although we limit our study to these four countries, the firms within them operate in many
countries throughout the world.








In one respect, these results are not surprising. When firms are ranked worldwide,

Japanese and U.S. multinationals are typically among the largest and most profitable. At first

glance, these results suggest that the average Japanese and U.S. multinational firm is able to

generate valuable operating synergies, which leads them to be worth more than the simple sum of

their individual parts.

Interestingly, however, we also employ additional tests which indicate that while

multinationals may dominate firms in their own domestic markets, they do no better than a

matched portfolio of international firms. These results are consistent with the findings of several

studies in the international investments literature which find that the risk-adjusted returns of

multinationals are similar to that of a portfolio of individual stocks that have the same

characteristics as the individual parts of the multinational (e.g., Heston and Rouwenhorst (1994),

Griffin and Karolyi (1998), and Rowland and Tesar (1998)).

The rest of this chapter proceeds as follows. The next section describes the data and

methodology. Section three presents the unconditional results regarding the impact of product-

market diversification and geographic diversification. Section four presents the regression results

that control for various firm characteristics, including ownership concentration. In these

regressions, firm value is calculated relative to a domestic benchmark; this enables us to test

whether multinational firms outperform domestic firms that operate in the country in which they

are headquartered. Section five reports similar regression results using an alternative

international benchmark; this enables us to test whether multinational firms outperform an

international portfolio of domestic firms. Section six describes the impact of excluding

conglomerate firms from the sample, while section seven provides a conclusion.

Data and Methodology

Description of Data

Our primary data source is the Worldscope database. Worldscope has product and

geographic segment data on more than 8,000 firms, covering 49 countries. The firms in this








database represent more than 85% of the world's total market capitalization. The Worldscope

database disaggregates sales along two dimensions: business segments and geographic regions.

The business segment data breaks down the company's sales according to product markets, and

the geographic region data breaks down sales according to the country and/or region where the

products are sold. The business segment data starts in 1991. The geographic region data,

however, is primarily available for only the most developed markets. Given these constraints, our

analysis focuses on firms in four developed countries (Germany, Japan, the U.K., and the U.S.),

over the time period from 1991 to 1995.

We use a combination of business segment and geographic region data to categorize

firms (in each of the four countries) into the four categories outlined below:

Single Region (GEO=0)Multiple Region (GEO= 1)
Single Segment (SEG=0) Domestic/Focused Multinational/Focused

Multiple Segment (SEG= 1) Domestic/Conglomerate Multinational/Conglomerate


Following Lins and Servaes (1999) and Fauver, Houston, and Naranjo (1999), we use two-digit

SIC codes to classify firms along product lines. In order to be considered geographically

diversified, the firm must have more than 10 percent of their total sales outside of their home

country. Furthermore, we remove firms front the sample whose primary activity is financial

services (firms with more than fifty percent of their total sales allocated to SIC codes 6000-6999)

because their sales are irregularly disclosed. Finally, private companies are omitted because the

computation of market capitalization requires market prices.

Description of Methodology

We estimate the value of corporate geographic diversification by modifying the

techniques used by Berger and Ofek (1995). Specifically, we utilize the firm's market capital-to-

sales ratio as a measure of corporate profitability. The market value of equity plus the book value

of debt is used as an estimate of the total market value of the firm's capital. Berger and Ofek

(1995) and Bodnar, Tang, and Weintrop (1998) use the same ratio to determine profitability, but








they also consider two other ratios: the ratio of total capital (or price)-to-eamrnings and the ratio of

total capital-to-assets. In their analysis, the three measures yield similar results. Because of the

limited reporting of segment assets and earnings for firms outside the United States, we are forced

to use the total market capital-to-sales ratio as the sole measure of firm value.

All else equal, if diversification increases firm value, we would expect diversified firms

to have higher market-to-sales ratios. Lamont and Polk (1999) show that higher market-to-sales

ratios may arise for two reasons: (1) product and geographic diversification may produce

valuable operating synergies which lead to higher cash flows and higher market values for a

given level of sales; (2) diversification may reduce the firm's risk and cost of capital, in which

case the higher market value stems from a lower discount rate.

Because of differences in capital intensity, growth opportunities, and other factors, we

would expect the market-to-sales ratio to vary considerably among firms in different industries.

Therefore, we need to control for industry effects when estimating the impact that diversification

has on firm value. To control for industry effects, we calculate the excess value of each firm by

taking the difference between the firm's actual performance and its imputed performance. Actual

performance is measured by the consolidated firm's capital-to-sales ratio. For single-segment

firms, imputed value is calculated as the median market-to-sales ratio among all pure-play

(single-segment firms) within the same industry and same country. For multi-segment firms,

imputed value is calculated by taking a weighted-average of the imputed values for each of the

firm's segments, where the weights reflect the proportion of the overall firm's sales that come

from each segment. Multi-segment firms have a positive excess value (i.e., a premium) if the

overall company's value is greater than the "sum of the parts." By contrast, multi-segment firms

have a negative excess value if their value is less than the imputed value that would be obtained

by taking a portfolio of pure-play firms that operate in the same industries and country as the

diversified firm.








We consider two approaches for estimating imputed performance. The first approach

(the domestic benchmark) compares the firm's performance to firms that operate in the same

industry(s) and within the country that the firm is headquartered. This benchmark, which is also

employed by Bodnar, Tang, and Weintrop (1998) and Denis, Denis, and Yost (1999), indicates

whether the average multinational firm trades at a premium or discount relative to the domestic

firms in its home country. In effect, this approach addresses the issue of whether geographic

diversification increases firm value.

The second approach (the international benchmark) also compares the firm's

performance to firms that operate in the same industry(s), but here imputed performance is based

on a weighted average of the imputed values for the various countries in which the firm operates.

So, for example, if a computer firm has 70% of its sales in the U.S. and 30% in Canada, the

imputed value, using the international benchmark would be:

(.7)(the value of the median pure-play U.S. computer firm)

+ (.3)(the value of the median pure-play Canadian computer firm).

Once again, we use the market-to-sales ratio to measure value. This international benchmark

enables us to examine whether multinational firms trade at a premium relative to a portfolio of

firms from each of the different countries. In effect, this approach addresses the question of

whether investors are better off investing in multinational firms or investing in a portfolio of

domestic firms from different countries. One drawback of using the international benchmark is

that while we have product market segments and geographic regions, we do not have the

breakdown among both. So, for example, if a multinational's product segments are 60%

computers and 40% shoes, we are forced to assume that the multinational has the same product

mix throughout its various geographic segments, even though this is unlikely to be the case. In

Section VI of the paper, we address this potential deficiency by excluding conglomerate firms

from the sample.








Finally, for each of the two benchmarks, we remove firms where the actual value is more

than four times the imputed value, or when the imputed value is more than four times the actual

value. This removal is done to avoid biases associated with unrealistic outliers and is similar to

the approach used in previous studies.

Results

Table 3-1 displays the summary statistics for the firms in our sample broken down by the

four types of firms: single-industry -- domestic and multinational and multi-industry -- domestic

and multinational. This table is also separated into four panels, with Panels A-D providing the

descriptive statistics for German, Japanese, U.K., and U.S. firms respectively. For each variable,

the top figure in each panel displays the mean value, below which we provide the median value in

parentheses. On the right side of each panel, we provide statistical tests for differences in the

mean and median value for each variable across the four firm types.

A couple of interesting patterns emerge when we compare the data for the four countries.

In each of the four countries, the average conglomerate has roughly 2.5 business segments.

However, the average number of geographic segments varies considerably Japanese

multinationals have the fewest geographic segments (1.6), while German multinationals have the

most segments (4.0). In each country, domestic firms outnumber multinational firms the

percentage of domestic firms ranges from 66% in Germany to 85% in the United States. The

percentage of single-industry firms ranges from 47% in Japan to 71% in the United States.

While the average firm characteristics are fairly similar across the four countries, some

notable differences do emerge. U.S. firms generally have the highest leverage ratios, the highest

market-to-sales ratios, and the highest profitability (as measured by operating income/sales).

German firms, on the other hand, tend to have the lowest market-to-sales ratio but the highest

level of ownership concentration. The Japanese firms in our sample are the largest on average -

where size is measured in terms of both total assets and total capital.








It is also interesting to note that for each of the four countries, focused/ multinational

firms have the highest market-to-sales ratios, and that the second highest ratios are found for

focused/domestic firms. Moreover, this pattern (not reported) holds for each year of the sample

period. These unconditional results suggest that focused firms consistently trade at higher

multiples relative to conglomerates and that among focused firms, multinationals trade at higher

multiples relative to domestic firms. While these results seem to indicate that geographic

diversification creates value, and that product market diversification reduces value, it remains

unclear whether diversification itself affects value, or whether there are other factors affecting

firm value that are correlated with the level of diversification.

To get at this issue, we estimate regressions below that control for various firm

characteristics. As shown in Table 3-1, within each country, there are also some significant

differences between the focused and conglomerate firms and between the domestic and

multinational firms. While the exact nature of these differences varies considerably, their

presence suggests that it is important that our subsequent regression analyses control for these

firm characteristics when analyzing the effects of industrial and geographical diversification.

Excess Value Created by Product Market and Geographic Diversification

Table 3-2 reports the mean and median excess value estimates by country for the four

types of firms: single-industry -- domestic and multinational and multi-industry -- domestic and

multinational. Panel A reports the results using the standard domestic benchmark to calculate

imputed value, whereas Panel B reports the results using the international benchmark.

Looking first at Panel A, we see that among domestic firms in each of the four countries,

focused firms significantly outperform conglomerates (see columns 1 and 3). This result is

strongly consistent with earlier studies that find a product diversification discount among firms in

the leading industrialized countries. However, among multinational firms, conglomerates

perform significantly worse than single-industry firms in Japan and the United States (see

columns 2 and 4).








Interestingly, for both Japan and the U.S., single-industry firms that are geographically

diversified (single-industry, multinational) trade at premiums relative to comparable

geographically focused firms (single-industry, domestic) (see columns 1 and 2). The results

indicate a premium of 14.1 percent for Japanese firms and 7.8 percent for U.S. firms. By

contrast, for Germany, we find that single-industry firms that are geographically diversified

(multinationals) trade at a discount of 10.5 percent relative to comparable geographically focused

firms (single-industry, domestic). For single-industry firms in the U.K., there is no significant

difference between the value of domestic and multinational firms. Finally, geographic

diversification also appears to have no significant effect on the firm value of multi-industry firms

in any of the four countries (see columns 3 and 4). Overall, the results in Panel A suggest that

industry diversification reduces firm value while geographic diversification potentially adds value

for firms that operate in a single industry.

Turning our attention to Panel B where we use corresponding international firms as the

benchmark, we see that once again industrial diversification reduces value for Japanese and U.S.

firms. However, when we use the international benchmark to calculate imputed value, we find

that that geographic diversification has a significant effect on excess value only for German

firms. This result indicates that benchmark considerations are important in the determination and

interpretation of the value associated with geographic diversification.

Looking jointly at the results from Panel A and Panel B, it appears that multinationals

have higher multiples than their domestic counterparts in the same industry, but that the

multinationals do not generally outperform a portfolio of domestic firms from each of the

countries where they have operations.

Do Multinational Firms Outperform their Domestic Counterparts?

While the results in Table 3-2 provide an overall depiction of the value of geographic and

product market diversification among the four countries, they do not control for individual firm

characteristics that are also likely to affect the firm's market-to-sales ratio. These other








characteristics include the firm's size, profitability, future growth opportunities, and ownership

structure. As mentioned above, previous studies have found important links between ownership

concentration and firm value and between ownership concentration and the value of product

market diversification. However, one drawback of incorporating ownership is that this data is

available for only a subset of firms in the sample. Consequently, we have chosen to report the

results both with and without ownership concentration as a control variable.

Regressions that Omit Ownership Concentration as a Control Variable

Our first set of regressions includes indicator variables corresponding to product and

geographic diversification along with firm characteristics, excluding ownership concentration.

Specifically, we estimate the following regression model for each of the four countries in our

sample:

(1) Excess Value = a + fl (Industry Diversification Dummy) + ,8, (Geographic Diversification Dummy)
+ ,3 (Relative Log Assets) + f4 (Relative Operating Income/ Sales)
+ /l(Relative Capital Expenditures / Sales) + e.

Excess value is defined to be the natural log of the ratio of the firm's market value to its imputed

value. The product market diversification dummy (SEG) is equal to one for multi-segment firms

and is set to zero for focused (single product) firms. The geographic diversification dummy

(GEO) is equal to one for multinational firms and equals zero for domestic firms. The log of the

relative assets controls for potential firm size effects. The ratio of operating income-to-sales

(OIS) provides a measure of firm profitability, while the ratio of capital expenditures-to-sales

(CES) proxies for the level of growth opportunities. Controlling for the other factors, we would

expect to see a positive link between excess value and both OIS and CES.4 Since our data covers

five years (1991-1995), we also include separate year dummies in the regressions to control for

intertemporal variations in market or economic conditions that may also affect the firm's market-

to-sales ratio. Lastly, since the dependent variable is measured in relative terms, we also measure


4 For Japan, CES is irregularly reported, and we therefore exclude it from the regression
specification. When we include CES for Japan, we obtain similar results, although the sample is
much smaller.








the independent variables in relative terms. In particular, the independent variables are all

measured relative to the value of the weighted-average multiplier firms that form the basis for the

excess value measure.

The regression results for the individual countries (not including ownership) are reported

in the first four columns of Table 3-3. As expected, we find that the estimated coefficients on

OIS (Relative Operating Income/Sales) and CES (Relative Capital Expenditures/Sales) are

positive and frequently significant. These results confirm that firms that are more profitable and

that have greater growth opportunities typically have higher market-to-sales ratios. The estimated

coefficient for the log of the relative size variable is significant and negative for firms in Germany

and Japan, but is significant and positive for firms in the U.K. and U.S. Although not reported,

the annual dummy coefficients indicate that there is little time variation in the excess values after

controlling for firm characteristics.

The estimated coefficients on the product market diversification dummy appear to be

reasonable and are generally well within the ranges found in earlier studies. Among U.S. firms,

we find a diversification discount of 18.4 percent, which is similar to the 14.4 percent discount

found by Berger and Ofek (1995) over an earlier time period 1986-1991. Moreover, our

estimated diversification discount for U.S. firms is also similar to those reported by Bodnar,

Tang, and Weintrop (1998) and Denis, Denis, and Yost (1999), who also control for geographic

diversification. For Japan and the U.K., we find statistically significant product market

diversification discounts of 7.1 percent and 10.5 percent respectively. These diversification

discounts are generally similar to those found by Lins and Servaes (1999) and Fauver, Houston,

and Naranjo (1999), but neither of those studies controlled for the effects of geographic

diversification. All in all, our results strongly confirm earlier findings and suggest that focused

firms outperform conglomerate firms in the most developed markets.

The results also indicate that multinational firms in the United States and Japan are

valued more highly than their domestic counterparts. Controlling for other factors, Japanese








multinationals trade at a 7 percent premium the coefficient on the geographic diversification

dummy (GEO) is 0.070, which is significant at the one percent level. Likewise, U.S.

multinationals trade at a 5.5 percent premium which is significant at the one percent level. This

result is consistent with the findings of Bodnar, Tang, and Weintrop (1998), who also find that

U.S. multinationals trade at a significant premium relative to domestic U.S. firms. However,

Bodnar, Tang, and Weintrop's estimated premium of 2.2 percent is somewhat smaller than the 5.5

percent premium that we find for the firms in our sample. By contrast, German multinationals

trade at a discount of 8 percent relative to German domestic firms (this difference is significant at

the one percent level). Finally, geographic diversification does not appear to have a significant

effect on firm value in the U.K.5

Regression Results Controlling for Ownership Concentration

The results discussed above suggest that corporate product diversification is costly and

geographic diversification may be beneficial. A potential problem with this conclusion is that, so

far, we have not explicitly controlled for agency costs associated with ownership concentration.

Indeed, several studies suggest that firm value is correlated with ownership structure [e.g.,

Demsetz and Lehn (1985), Morck, Shleifer and Vishny (1988), Holderness and Sheehan (1998),

and McConnell and Servaes (1990)] and that ownership structure varies across countries and

legal systems [e.g., La Porta, Lopez-de-Silanes, Shleifer and Vishny (1997, 1998), LaPorta,

Lopez-De-Silanes and Shleifer (1999), and Claessens, Djankov, Fan and Lang (1998)]. To the

extent that ownership concentration affects firm value, it may also affect the estimated value of

corporate diversification. This concern may be particularly relevant if there is a strong link

between ownership concentration and firm value and if focused and diversified firms have

significantly different levels of ownership concentration. An additional concern is that even if

ownership concentration levels are similar for both focused and diversified firms, ownership


SAs a robustness check, we also included an interaction indicator variable between the product
(SEG) and geographic (GEO) diversification dummies as an explanatory variable. In each case,
the interactive term was insignificantly different from zero.








concentration may still be important if it has a differential effect on the value of focused and

diversified firms.

The exact link between ownership structure and firm value, however, is not entirely clear.

On one hand, it is widely acknowledged that concentrated ownership is likely to reduce the

conflicts that arise when there is a separation between managers and stockholders. This link

suggests a positive relation between firm value and ownership concentration. On the other hand,

concentrated ownership provides large investors with opportunities to exploit minority

shareholders, thereby suggesting at least for some range of values a negative relation between

firm value and ownership concentration. In a recent study, Holderness and Sheehan (1998)

conclude that in the United States, legal constraints often effectively limit the actions of majority

shareholders but it is not clear to what extent their conclusions extend outside the U.S.

As reported in the summary statistics in Table 3-1, German firms have a significantly

higher level of average ownership concentration. We also find that within each country,

ownership concentration levels differ across the four firm types, although no clear patterns

consistently emerge among the four countries. However, given the clear correlation between

organizational structure and ownership structure, it is important that we control for ownership

concentration when estimating the sources of any diversification discounts or premiums.

Similar to Morck, Shleifer, and Vishny (1988) and others, we account for the nonlinear

relation between ownership structure and firm value by creating three separate ownership

concentration variables:6

OWNOto 10 = total ownership if total ownership < 0.10,

=0.10 if total ownership 0.10;

OWN10to30 =0 if total ownership < 0.10,

6 Morck, Shleifer and Vishny (MSV, 1988) use 5 percent and 25 percent as their breakpoints.
Given that the Worldscope databank does not generally provide firm level ownership
concentration values below 5 percent (aside from the unreported values), we use a 10 percent cut-
off for the first breakpoint and 30 percent as the next breakpoint to be consistent with MSV's
ownership ranges.








= total ownership minus 0.10 if 0.10 total ownership < 0.30,

=0.20 if total ownership 0.30;

OWNover30 = 0 if total ownership < 0.30,

= total ownership minus 0.30 if total ownership 0.30.

This classification suggests that the marginal impact of increased ownership concentration varies

depending on whether ownership concentration is less than 10 percent, between 10 and 30

percent, and greater than 30 percent. To assess the impact of ownership concentration on the

value of corporate diversification, we also interact OWN10 Oto30 and OWNover30 with the

dummy variables SEG and GEO, which equal one if the firm has multiple industry segments and

has sales in more than one country.7 Generally, we would expect a positive link between firm

value and OWNOto 10. Within this range, increases in ownership concentration are likely to

improve managerial incentives without dramatically increasing the risks of managerial

entrenchment and expropriation. For ownership concentration levels beyond ten percent, the

expected results are less clear. For these firms, the benefits of increased ownership may be more

than offset by the costs resulting from increased managerial entrenchment and by the potential for

the expropriation of minority shareholders. Consequently, the link between OWN 10to30 and

OWNover30 and firm value is less clear.

The firm level regression estimates that control for ownership concentration are reported

in columns 5-8 of Table 3-3. Looking at the multi-industry dummy coefficients (SEG), we find

that firms that diversify along product lines in Japan, the U.S., and the U.K. continue to trade at a

discount relative to focused firms. Moreover, the magnitude of the estimated coefficient

increases in Japan and slightly decreases in the U.S. after controlling for ownership concentration.

With the augmented specification, we still find that German multi-industry firms do no worse (or

better) relative to focused firms.


7 Note that due to singularity, we do not include OWNOto 10*SEG in our specification.








Controlling for ownership concentration also has an effect on the estimated coefficients

for the multi-country segment dummy (GEO). Specifically, the estimated coefficients for

Japanese and U.S. firms are higher after controlling for ownership once again confirming that

Japanese and U.S. multinationals are valued more highly than their domestic counterparts. The

geographic diversification coefficient for Germany is now insignificant after controlling for

ownership. Finally, the estimated geographic diversification coefficient for the U.K. firms

remains insignificant.

While it is not the primary focus of our analysis, the estimated coefficients for the

ownership concentration are still of considerable interest. First, for low levels of ownership

concentration, there is a positive link between ownership concentration and excess value for U.S.

firms, while the relation is insignificant for firms in the other three countries. Second, for

ownership concentration levels beyond ten percent, we generally find that increases in ownership

concentration lead to a reduction in value for both focused and diversified firms. This result

confirms the fact that there are both costs and benefits associated with increased ownership

concentration. Finally, from the coefficients on the ownership concentration variables that are

interacted with the diversification dummy (OWN 10to30*SEG, OWNover30*SEG,

OWN10to30*GEO, and OWNover30*GEO), we see that the effects of ownership concentration

are significantly different for focused and diversified firms. For ownership concentration levels

between 10 and 30 percent, excess value is significantly lower for the diversified firms,

suggesting that entrenchment problems and expropriation of minority shareholders is more of a

concern for diversified firms. However, beyond 30 percent, excess value is significantly higher

for diversified firms. For example, a geographically diversified firm in the United States with a

concentrated ownership of 35 percent would be valued 8 percent more (0.232*0.35) relative to a

domestic focused firm with concentrated ownership below 10 percent. All in all, the results

suggest that there is a link between ownership concentration and excess value, and that this link

may be somewhat different for focused and diversified firms.








Do Multinationals Outperform a Portfolio of Domestic Firms From Different Countries?

The results in the previous section indicate that U.S. and Japanese multinationals are

valued, on average, more highly than their domestic counterparts. While these results suggest

that geographic diversification enhances firm value in these countries, it is unclear whether

multinationals outperform a portfolio of domestic firms in the various countries in which they

operate. To get at this issue, we compared multinational firms to a weighted average of firms that

have the same geographic and product mix. These results are reported in Table 3-4. Once again,

the regression results excluding ownership are reported in columns 1-4, while the results

including ownership concentration are reported in columns 5-8.

After controlling for ownership concentration, we find that the coefficient on the

geographic diversification dummy (GEO) is not significantly different from zero in each of the

four countries. These results suggest that while multinational firms in the U.S. and Japan

outperform their domestic counterparts, multinational firms do not outperform a simulated

portfolio of international firms that mimic their overall product mix. Interestingly, these results

parallel some recent findings in the international investments literature. Heston and Rouwenhorst

(1994) and Griffin and Karolyi (1998), for example, suggest that while multinational firms have

higher risk-adjusted returns, shareholders can duplicate these same risk-adjusted returns by

holding a portfolio of domestic firms in each international market. Moreover, similar to the

above studies, we also find that the value of geographic diversification largely arises from

differences in performance across countries, not from differences in industry composition or

clustering across countries. Interestingly, these results parallel Heston and Rouwenhorst (1994)

and Griffin and Karolyi (1998) who use international indice return data, whereas we use firm

level corporate data in our analysis. In another recent investments study that also parallels our

results, Rowland and Tesar (1998) examine the return mean-variance efficient frontier with

domestic firms, multinational firms, and international equity indices. They first find that

multinational firms add risk-adjusted value to a portfolio of domestic firms. However, when the








domestic portfolio is augmented with international indices, the multinational firms do not add any

additional value.

In a related stream of the international investments literature that also parallels our

findings, several studies find that investors overweight their portfolios with domestic securities

relative to international securities. In particular, French and Poterba (1991) and Cooper and

Kaplanis (1994) among others find that there is a "home bias" towards investment in domestic

securities. Given that investors are reluctant to purchase overseas investments, domestically

headquartered multinational firms often serve as a method to obtain some international exposure,

resulting in a potentially higher valuation of multinational firms. Therefore, one might expect

that countries with relatively greater home bias would likely value multinational firms more than

in countries where there is less home bias, all else equal. Given that French and Poterba (1991)

find that domestic portfolio dedication (home bias) is greatest in Japan and the U.S. for the

countries in our sample, we might expect that multinationals in these countries would be valued

more highly. Consistent with this conjecture, we find that multinationals in Japan and the U.S.

trade a premium when using the domestic benchmark. However, when using the international

benchmark, the valuation premiums disappear.

Taking a Closer Look at the Single-Segment Firms

As a final test of the value of geographic diversification, we re-estimate the results for the

sub-sample of focused firms that have sales in a single segment. There are two benefits that arise

from eliminating the conglomerate firms from the sample. First, as we have seen, there may be

important interaction effects between the value of product-market diversification and the value of

geographic diversification. In this regard, eliminating the conglomerate firms removes the effects

of product-market diversification, thereby potentially providing us with a cleaner test of the value

of geographic diversification. Second, we indicated above that one problem with the

international benchmark results reported in Table 3-4 is that data limitations forced us to assume

that multinationals have the same product mix in each of their geographic regions segments, even








though this was unlikely to be the case. Once again, this concern disappears if we eliminate

conglomerate firms from the sample.

The regression results for the sub-sample of single-segment firms are reported in Table 3-

5. These regressions include ownership concentration as a control variable, and are estimated

using both the domestic and international benchmarks for computing imputed value. These

results directly parallel the domestic and international benchmark results reported earlier in

Tables 3-3 and 3-4. In Table 3-5, Panel A, with the domestic benchmark, we find a geographic

diversification premium in the U.S. and Japan and a geographic diversification discount in

Germany. Using the international benchmark, Panel B, we find that the geographic

diversification premiums and discount disappear similar to Table 3-4. Once again, this confirms

the conclusion that for U.S. and Japanese firms, geographic diversification increases value, but

that multinationals do not trade at a premium relative to an international portfolio of domestic

firms.

Conclusion

While in recent years a large literature has examined the links between product

diversification and firm value, considerably fewer studies have examined the value of geographic

diversification. The lack of work in this area is surprising given the dramatic growth in foreign

investment among firms in the leading industrialized countries over the past twenty-five years.

In this paper, we investigate the connection between product and geographic

diversification and its impact on firm value. We gather data on more than 4,000 firms from four

highly industrialized countries (Germany, Japan, the U.K., and the U.S.). On average, we find

that the geographic diversification neither enhances nor reduces the value of multinationals

located in Germany and the United Kingdom. However, our results suggest that geographic

diversification does significantly enhance the value of multinational firms in Japan and the United

States. These results suggest that multinationals in these countries are able to capture valuable

operating synergies or generate benefits from risk reduction. However, we also find that in all








four countries, multinationals typically do not outperform an international portfolio of domestic

firms.

Interestingly, our results parallel some recent findings in the international investments

literature. Recent studies by Heston and Rouwenhorst (1994), Griffin and Karolyi (1998), and

Rowland and Tesar (1998), for instance, find that while multinational firms have higher risk-

adjusted returns, shareholders can duplicate these same risk-adjusted returns by holding a

portfolio of domestic firms in each international market. Overall, our results suggest that there

are important interactions between the value of product market diversification and geographic

diversification and that future studies need to consider both forms of diversification when

investigating the links between diversification and firm value.








Table 3-1
Summary Statistics by Industrial and Geographical Diversification: 1991 1995

Panel A: German Firms


Single-Industry Firms Multi-Industry Firms
Firm Level Characteristics by
Industrial and Geographical
Diversification Domestic Multinational Domestic Multinational
(1) (2) (3) (4)
Number of Industrial 1 1 2.713 2.550
Segments (1) (1) (2) (2)
Number of Geographical 1 4.037 1 3.759
Segments (1) (4) (1) (3)
Total Assets (mil $) 1,550 564 4,420 1,400
(220) (153) (683) (365)
Total Capital (mil $) 531 259 1,460 450
(89) (57) (273) (147)
Leverage Ratio 0.193 0.203 0.185 0.227
(0.168) (0.181) (0.144) (0.157)
Operating Income/Sales 0.053 0.048 0.044 0.043
(0.059) (0.061) (0.046) (0.046)
Capital Expenditure/Sales 0.082 0.078 0.075 0.065
__________(0.055) (0.056) (0.055) (0.050)
Ownership Concentration 0.612 0.619 0.505 0.574
_____ ____(0.650) (0.700) (0.514) (0.660)
Market/Sales 0.785 0.842 0.656 0.702
__________(0.597) (0.557) (0.463) (0.540)
Observations 538 272 366 191








Panel A--continued

Test of Statistical Differences
p-values
Firm Level Characteristics by
Industrial and Geographical
Diversification (1)-(2) (1)-(3) (1)-(4) (2)-(3) (2)-(4) (3)-(4)
Number of Industrial
Segments

Number of Geographical
Segments

Total Assets (mil $) 0.000 0.000 0.707 0.000 0.008 0.000
(0.001) (0.000) (0.000) (0.000) (0.000) (0.000)
Total Capital (mil $) 0.002 0.000 0.534 0.000 0.087 0.000
(0.005) (0.000) (0.001) (0.000) (0.000) (0.001)
Leverage Ratio 0.412 0.407 0.093 0.143 0.263 0.038
___________(0.452) (0.809) (0.229) (0.346) (0.592) (0.224)
Operating Income/Sales 0.641 0.120 0.182 0.732 0.699 0.908
(0.882) (0.021) (0.081) (0.078) (0.080) (0.766)
Capital Expenditure/Sales 0.601 0.313 0.005 0.719 0.080 0.131
______ ____(0.766) (1.000) (0.159) (0.873) (0.080) (0.028)
Ownership Concentration 0.761 0.000 0.131 0.000 0.122 0.013
(0.040) (0.004) (0.773) (0.001) (0.093) (0.018)
Market/Sales 0.529 0.000 0.157 0.035 0.139 0.412
(0.457) (0.000) (0.215) (0.006) (0.807) (0.024)
Observations








Panel B: Japanese Firms


Single-Industry Firms Multi-Industry Firms
Firm Level Characteristics by
Industrial and Geographical
Diversification Domestic Multinational Domestic Multinational
(1) (2) (3) (4)
Number of Industrial 1 1 2.536 2.569
Segments (1) (1) (2) (2)
Number of Geographical 1 1.689 1 1.658
Segments (1) (2) (1) (2)
Total Assets (mil $) 3,010 2,070 3,450 1,230
(406) (297) (401) (299)
Total Capital (mil $) 1,850 992 1,540 530
(214) (195) (202) (146)
Leverage Ratio 0.265 0.241 0.282 0.264
(0.242) (0.216) (0.270) (0.250)
Operating Income/Sales 0.084 0.078 0.075 0.079
(0.076) (0.073) (0.067) (0.069)
Capital Expenditure/Sales

Ownership Concentration 0.266 0.251 0.267 0.293
(0.226) (0.214) (0.236) (0.252)
Market/Sales 1.176 1.241 1.054 1.034
___________(0.944) (1.063) (0.841) (0.857)
Observations 1,585 225 1,802 260








Panel B--continued

Test of Statistical Differences
p-values
Firm Level Characteristics by
Industrial and Geographical
Diversification (1)-(2) (1)-(3) (1)-(4) (2)-(3) (2)-(4) (3)-(4)
Number of Industrial
Segments

Number of Geographical
Segments

Total Assets (mil $) 0.028 0.231 0.000 0.001 0.039 0.000
(0.005) (0.851) (0.062) (0.006) (0.818) (0.063)
Total Capital (mil $) 0.000 0.144 0.000 0.001 0.003 0.000
(0.354) (0.547) (0.016) (0.530) (0.221) (0.024)
Leverage Ratio 0.060 0.014 0.930 0.002 0.164 0.159
______ ____(0.078) (0.007) (0.875) (0.001) (0.105) (0.201)
Operating Income/Sales 0.215 0.000 0.217 0.466 0.970 0.419
(0.354) (0.000) (0.084) (0.040) (0.297) (0.596)
Capital Expenditure/Sales

Ownership Concentration 0.260 0.881 0.038 0.228 0.015 0.045
(0.571) (0.521) (0.136) (0.567) (0.270) (0.422)
Market/Sales 0.268 0.000 0.002 0.001 0.003 0.670
(0.075) (0.000) (0.144) (0.002) (0.020) (0.791)
Observations








Panel C: U.K. Firms


Single-Industry Firms Multi-Industry Firms
Firm Level Characteristics by
Industrial and Geographical
Diversification Domestic Multinational Domestic Multinational
(1) (2) (3) (4)
Number of Industrial 1 1 2.657 2.553
Segments (1) (1) (2) (2)
Number of Geographical 1 3.651 1 3.790
Segments (1) (4) (1) (4)
Total Assets (mil $) 847 579 1,480 556
(68) (54) (169) (90)
Total Capital (mil$) 514 385 844 337
(34) (27) (90) (54)
Leverage Ratio 0.198 0.208 0.222 0.214
(0.165) (0.167) (0.199) (0.184)
Operating Income/Sales 0.121 0.117 0.102 0.112
(0.101) (0.101) (0.095) (0.101)
Capital Expenditure/Sales 0.084 0.109 0.057 0.072
(0.039) (0.035) (0.038) (0.037)
Ownership Concentration 0.359 0.397 0.266 0.303
(0.351) (0.377) (0.239) (0.294)
Market/Sales 1.151 1.235 0.936 1.137
__________(0.800) (0.804) (0.749) (0.751)
Observations 2,048 789 1,412 476








Panel C--continued

Test of Statistical Differences
p-values
Firm Level Characteristics by
Industrial and Geographical
Diversification (l)-(2) (1)-(3) (1)-(4) (2)-(3) (2)-(4) (3)-(4)
Number of Industrial
Segments

Number of Geographical
Segments

Total Assets (mil $) 0.063 0.000 0.029 0.000 0.883 0.000
(0.009) (0.000) (0.003) (0.000) (0.000) (0.000)
Total Capital (mil $) 0.184 0.000 0.050 0.000 0.674 0.000
(0.001) (0.000) (0.000) (0.000) (0.000) (0.000)
Leverage Ratio 0.317 0.014 0.186 0.241 0.676 0.543
(0.395) (0.000) (0.004) (0.000) (0.067) (0.118)
Operating Income/Sales 0.497 0.000 0.235 0.004 0.512 0.237
(0.976) (0.009) (0.919) (0.192) (0.890) (0.138)
Capital Expenditure/Sales 0.007 0.000 0.061 0.000 0.000 0.005
______ ____(0.124) (0.368) (0.309) (0.203) (0.342) (0.525)
Ownership Concentration 0.000 0.000 0.000 0.000 0.000 0.003
______ ____(0.029) (0.000) (0.000) (0.000) (0.000) (0.000)
Market/Sales 0.117 0.000 0.813 0.000 0.168 0.001
__________(0.957) (0.019) (0.019) (0.054) (0.051) (0.915)
Observations








Panel D: U.S. Firms


Single-Industry Firms Multi-Industry Firms
Firm Level Characteristics by
Industrial and Geographical
Diversification Domestic Multinational Domestic Multinational
(1) (2) (3) (4)
Number of Industrial 1 1 2.428 2.373
Segments (1) (1) (2) (2)
Number of Geographical 1 2.668 1 2.688
Segments (1) (3) (1) (3)
Total Assets (mil $) 1,470 807 3,350 971
__________(247) (163) (559) (265)
Total Capital (mil $) 909 531 1,740 616
(171) (120) (358) (180)
Leverage Ratio 0.253 0.247 0.278 0.279
(0.234) (0.207) (0.260) (0.276)
Operating Income/Sales 0.144 0.157 0.134 0.131
(0.132) (0.140) (0.125) (0.123)
Capital Expenditure/Sales 0.091 0.107 0.077 0.084
(0.049) (0.049) (0.045) (0.046)
Ownership Concentration 0.283 0.295 0.241 0.285
(0.248) (0.271) (0.189) (0.245)
Market/Sales 1.730 1.980 1.343 1.387
__________(1.230) (1.305) (0.994) (0.954)
Observations 6,891 1,313 2,840 426








Panel D--continued

Test of Statistical Differences
p-values
Firm Level Characteristics by
Industrial and Geographical
Diversification (l1)-(2) (1)-(3) (1)-(4) (2)-(3) (2)-(4) (3)-(4)
Number of Industrial
Segments

Number of Geographical
Segments

Total Assets (mil $) 0.000 0.000 0.000 0.000 0.201 0.000
______ ____(0.000) (0.000) (0.616) (0.000) (0.000) (0.000)
Total Capital (mil $) 0.000 0.000 0.000 0.000 0.280 0.000
(0.000) (0.000) (0.840) (0.000) (0.001) (0.000)
Leverage Ratio 0.398 0.000 0.011 0.000 0.006 0.892
______ ____(0.026) (0.000) (0.002) (0.000) (0.000) (0.642)
Operating Income/Sales 0.009 0.021 0.039 0.000 0.000 0.631
(0.028) (0.002) (0.073) (0.000) (0.005) (0.603)
Capital Expenditure/Sales 0.008 0.000 0.258 0.000 0.005 0.238
(0.833) (0.006) (0.058) (0.114) (0.097) (0.917)
Ownership Concentration 0.127 0.000 0.852 0.000 0.516 0.002
(0.034) (0.000) (0.742) (0.000) (0.128) (0.001)
Market/Sales 0.000 0.000 0.000 0.000 0.000 0.478
(0.024) (0.000) (0.000) (0.000) (0.000) (0.467)
Observations

Panels A-D provide firm level descriptive statistics for German, Japanese, U.K., and U.S. firms
respectively. The upper number in each cell reports the mean value for each variable, while the
lower number in parentheses reports the median value for each variable. T-tests are used to test
for differences in each respective mean value, while Wilcoxon rank-sum tests are used to test for
differences in the median values. Single-industry firms are firms that operate in only one two-
digit SIC code industry, while multi-industry firms are defined as firms that operate in two or
more two-digit SIC code industries and no firm segment sales exceed 90% of total firm sales.
Domestic firms are defined as firms that have over 90% of their total firm sales in their home
market, while multinational firms are defined as firms that have more than 10% of their total sales
outside their home market. The leverage ratio is defined as book value of debt divided by total
assets. Ownership concentration is defined as the sum of individual and/or institutional
ownership holdings that are equal to or exceed five percent of a firm's common stock. Due to
missing ownership concentration data, the number of observations is slightly less than that for the
other reported variables. Market-to-sales is defined as the ratio of a firm's market value of equity
plus book value of debt to its total sales.








Table 3-2
Excess Values by Industrial and Geographical Diversification: 1991 1995


Panel A: Domestic Benchmark


Single-Industry Firms Multi-Industry Firms


Excess Value by Country Domestic Multinational Domestic Multinational
(1) (2) (3) (4)
German 0.027 -0.078 -0.057 -0.020
(0.000) (-0.070) (-0.078) (-0.007)
Japanese 0.000 0.141 -0.051 -0.038
(0.000) (0.118) (-0.074) (0.018)
U.K. 0.001 -0.009 -0.085 -0.055
______ ____(0.000) (-0.027) (-0.093) (-0.086)
U.S. -0.020 0.058 -0.172 -0.151
__________(-0.014) (0.060) (-0.174) (-0.200)

Panel A--continued

Test of Statistical Differences
p-values

Excess Value by Country
(1)-(2) (1)-(3) (1)-(4) (2)-(3) (2)-(4) (3)-(4)
German 0.008 0.019 0.324 0.631 0.292 0.484
_____ _____(0.011) (0.003) (0.816) (0.749) (0.146) (0.122)
Japanese 0.000 0.007 0.314 0.000 0.001 0.738
____________(0.003) (0.000) (0.781) (0.000) (0.160) (0.289)
U.K. 0.712 0.000 0.059 0.006 0.181 0.333
___________(0.229) (0.000) (0.084) (0.044) (0.365) (0.672)
U.S. 0.000 0.000 0.000 0.000 0.000 0.492
_______ _(0.003) (0.000) (0.000) (0.000) (0.000) (0.592)








Panel B: International Benchmark


Single-Industry Firms Multi-Industry Firms


Excess Value by Country Domestic Multinational Domestic Multinational
__________ (1) (2) (3) (4)
German 0.024 -0.017 0.028 0.282
(0.000) (0.044) (0.045) (0.244)
Japanese -0.127 -0.105 -0.173 -0.031
_______ ____(-0.133) (-0.213) (-0.176) (0.021)
U.K. -0.002 -0.123 0.020 -0.107
______ _____(0.000) (-0.135) (0.028) (-0.164)
U.S. -0.028 -0.038 -0.060 -0.007
______ ______(-0.024) (-0.047) (-0.060) (-0.074)

Panel B--continued

Test of Statistical Differences
p-values

Excess Value by Country
(1)-(2) (1)-(3) (1)-(4) (2)-(3) (2)-(4) (3)-(4)
German 0.583 0.943 0.007 0.592 0.010 0.011
(0.901) (0.956) (0.005) (0.827) (0.010) (0.027)
Japanese 0.748 0.063 0.273 0.340 0.504 0.112
(0.809) (0.052) (0.237) (0.383) (0.471) (0.107)
U.K. 0.015 0.454 0.082 0.009 0.833 0.048
_________ _(0.014) (0.317) (0.080) (0.007) (0.841) (0.044)
U.S. 0.790 0.096 0.817 0.612 0.750 0.568
___________(0.809) (0.140) (0.767) (0.654) (0.755) (0.566)
Panel A contains the excess values using the domestic firm benchmarks, while Panel B contains
the excess values from using the corresponding international firm benchmarks. With the
international benchmark, the firm's imputed performance is based on a weighted average of the
imputed values for the various countries in which the firm operates. The upper number in each
cell reports the mean value for each variable, while the lower number in parentheses reports the
median value for each variable. T-tests are used to test for differences in each respective mean
value, while Wilcoxon rank-sum tests are used to test for differences in the median values.
Excess value is defined as the natural logarithm of the ratio of a firm's market-to-sales ratio to its
imputed market-to-sales ratio. Firms with excess values that are greater than four or less than
one-fourth are eliminated from the sample. Single-industry firms are firms that operate in only
one two-digit SIC code industry, while multi-industry firms are defined as firms that operate in
two or more two-digit SIC code industries and no firm segment sales exceed 90% of total firm
sales. Domestic firms are defined as firms that have over 90% of their total firm sales in their
home market, while multinational firms are defined as firms that have more than 10% of their
total sales outside their home market.








Table 3-3
Multivariate Regression Estimates of Excess Values Using the Domestic Benchmark:
1991 1995

German Japanese U.K. U.S.
Variables (1) (2) (3) (4)
Constant 0.096*** 0.132*** -0.344*** -0.236***
(-2.59) (5.64) (-9.37) (-13.20)
Multi-Industry Segment 0.000 -0.071 *** -0.105*"** -0.184***
Dummy (SEG) (-0.01) (-4.07) (-6.11) (-15.28)
Multi-Country Segment -0.080*** 0.070*** 0.019 0.055***
Dummy (GEO) (-2.58) (2.58) (1.03) (3.49)
Relative Operating 0.030*** 0.024*** 0.200*** 0.020**
Income-to-Sales (OIS) (4.56) (2.67) (6.34) (2.93)
Relative Capital 0.052*** ___ 0.006 0.082***
Expenditures-to-Sales (4.85) (1.19) (10.44)
(CES)
Relative Total Assets 0.112*** -0.065*** 0.048*** 0.037***
(ASSETS) (-5.71) (-5.97) (3.86) (4.29)
Adjusted R2 0.09 0.06 0.15 0.08
Number of Observations 1,367 3,872 4,725 11,470








Table 3-3--continued
German Japanese U.K. U.S.
Variables (5) (6) (7) (8)
Constant -0.095 0.176*** -0.225*** -0.194***
(-1.38) (5.06) (-4.78) (-8.24)
Multi-Industry Segment -0.033 -0.168*** -0.076** -0.170***
Dummy (SEG) (-0.42) (-6.21) (-2.34) (-8.25)
Multi-Country Segment -0.130 0.116*** 0.018 0.122***
Dummy (GEO) (-1.52) (2.73) (0.43) (4.28)
Relative Operating 0.032** 0.025*** 0.201*** 0.022***
Income-to-Sales (OIS) (4.71) (2.63) (5.45) (2.80)
Relative Capital 0.046** ___ 0.011* 0.074***
Expenditures-to-Sales (4.71) (1.88) (8.95)
(CES)
Relative Total Assets 0.118** -0.067*** 0.004 0.024**
(ASSETS) (-5.85) (-5.73) (0.29) (2.39)
Ownership Concentration -0.457 0.234 -0.168 0.393*
< 10 (OWNOtolO) (-0.46) (0.68) (-0.46) (1.80)
Ownership Concentration 0.851 -0.194 -0.389* -0.419***
10-30 (OWNO10to30) (1.35) (-0.93) (-1.83) (-3.06)
Ownership Concentration 0.240** -0.631*** -0.230** -0.083
> 30 (OWNover30) (-2.37) (-3.45) (-2.15) (-1.22)
Ownership Concentration 1.004* 0.651** -0.390 -0.220
10-30 interacted with (-1.86) (2.41) (-1.49) (-1.20)
SEG (OWN 10to30*SEG)__________
Ownership Concentration 0.673*** 0.453* -0.104 0.123
> 30 interacted with SEG (4.14) (1.80) (-0.63) (1.08)
(OWNover30*SEG)
Ownership Concentration 0.277 -0.463 -0.161 -0.609**
10-30 interacted with (0.46) (-1.12) (-0.52) (-2.54)
GEO
(OWN10to30*GEO)
Ownership Concentration -0.056 -0.034 0.204 0.232*
> 30 interacted with GEO (-0.32) (-0.09) (1.15) (1.66)
(OWNover30*GEO)
Adjusted R2 0.11 0.07 0.17 0.08
Number of Observations 1,255 3,817 3,823 8,803


aigniiicani at i percent
in parentheses.


r"), 5 percent (*-), and 10 percent (*) levels, Robust-White t-statistics


Regression estimates are from 1991-1995. Excess value is defined as the natural logarithm of the
ratio of a firm's market-to-sales ratio to its imputed market-to-sales ratio. With the domestic
benchmark, the firm's imputed performance is based on a weighted average of the corresponding
pure plays within the domestic market. Firms with excess values that are greater than four or less
than one-fourth are eliminated from the sample. The industry diversification dummy, SEG, is
equal to one for firms who operate in more than one industry and zero otherwise. Multi-industry
firms are defined as firms that operate in two or more two-digit SIC code industries and no firm
segment sales exceed 90% of total firm sales. The multinational diversification dummy, GEO, is
equal to one for firms who operate in more than one country and zero otherwise. Multinational
firms are defined as firms that operate in two or more countries and no firm segment sales in a
particular country exceed 90% of total firm sales. OIS is defined as the firm's operating income-





86


to-sales, while CES is the firm's capital expenditures-to-sales. For Japan, we omit the CES
variable from the specification due to infrequently reported figures. Assets are defined as the
natural logarithm of the firm's total assets. The independent variables OIS, CES, and ASSETS
are all measured relative to the value of the weighted-average multiplier firms that form the basis
for the excess value measure. Ownership concentration is defined as the sum of individual and/or
institutional ownership holdings that are equal to or exceed five percent of a firm's common
stock. OWNOto 10: = total ownership if total ownership < 0.10, = 0.10 if total ownership >
0.10; OWN10to30: = 0 if total ownership < 0.10, = total ownership minus 0.10 if 0.10 < total
ownership < 0.30, = 0.20 if total ownership > 0.30; OWNover30: = 0 if total ownership < 0.30,
= total ownership minus 0.30 if total ownership > 0.30. Each model specification also includes
year dummies for 1992-1995.








Table 3-4
Multivariate Regression Estimates of Excess Values Using the International Benchmark:
1991 1995

German Japanese U.K. U.S.
Variables (1) (2) (3) (4)
Constant -0.071 -0.160*** -0.421*** -0.259***
_________(-1.52) (-3.75) (-7.64) (-12.29)
Multi-Industry Segment 0.052 -0.047** -0.009 -0.048***
Dummy (SEG) (1.15) (-2.09) (-0.34) (-2.71)
Multi-Country Segment 0.033 0.051 -0.124*** -0.015
Dummy (GEO) (0.54) (1.04) (-3.59) (-0.44)
Relative Operating 0.013* 0.219*** 0.283*** 0.017**
Income-to-Sales (OIS) (1.72) (6.38) (6.67) (2.29)
Relative Capital 0.046*** 0.010 0.098***
Expenditures-to-Sales (2.98) (0.72) (10.25)
(CES)
Relative Total Assets -0.081*** -0.052*** 0.054*** 0.035***
(ASSETS) (-3.27) (-4.11) (3.33) (3.32)
Adjusted R2 0.07 0.12 0.22 0.08
Number of Observations 778 2,509 2,739 7,901








Table 3-4--continued______
German Japanese U.K. U.S.
Variables (5) (6) (7) (8)
Constant -0.074 -0.126** -0.321*** -0.210***
(-0.89) (-2.14) (-4.61) (-7.61)
Multi-Industry Segment -0.047 -0.124*** -0.035 -0.028
Dummy (SEG) (-0.40) (-3.48) (-0.76) (-0.94)
Multi-Country Segment -0.106 0.040 0.006 0.053
Dummy (GEO) (-1.28) (0.45) (0.08) (1.01)
Relative Operating 0.017* 0.248*** 0.278*** 0.013*
Income-to-Sales (OIS) (1.82) (6.62) (5.51) (1.82)
Relative Capital 0.043*** ___ 0.030*** 0.097***
Expenditures-to-Sales (2.85) (3.54) (8.63)
(CES)
Relative Total Assets -0.097*** -0.057*** 0.010 0.023*
(ASSETS) (-3.69) (-4.12) (0.52) (1.92)
Ownership Concentration -0.023 0.111 -0.267 0.459*
<10 (OWNOto lO) (-0.02) (0.26) (-0.61) (1.74)
Ownership Concentration 0.358 -0.276 -0.350 -0.482***
10-30 (OWN10to30) (0.48) (-1.22) (-1.50) (-3.14)
Ownership Concentration -0.171 -0.447** -0.208 -0.040
> 30 (OWNover30) (-1.62) (-2.54) (-1.72) (-0.55)
Ownership Concentration -0.522 0.519 -0.308 -0.337
10-30 interacted with SEG (-0.63) (1.51) (-0.80) (-1.23)
(OWN10to30*SEG)
Ownership Concentration 0.672*** 0.241 0.298 0.352**
> 30 interacted with SEG (2.72) (0.76) (1.03) (2.19)
(OWNover30*SEG)
Ownership Concentration 0.132 0.410 -0.638 -0.537
10-30 interacted with GEO (0.13) (0.45) (-1.12) (-1.14)
(OWN10to30*GEO)____
Ownership Concentration 0.104 -0.387 -0.078 0.136
> 30 interacted with GEO (0.31) (-0.53) (-0.26) (0.48)
(OWNover30*GEO)
Adjusted R2 0.09 0.14 0.24 0.08
Number of Observations 703 2,480 2,224 6,076


igiiiicant at i percent ( ),
in parentheses.


percent ( -), and iU percent (") levels, Robust-White t-statistics


Regression estimates are from 1991-1995. Excess value is defined as the natural logarithm of the
ratio of a firm's market-to-sales ratio to its imputed market-to-sales ratio. With the international
benchmark, the firm's imputed performance is based on a weighted average of the imputed values
for the various countries in which the firm operates. Firms with excess values that are greater
than four or less than one-fourth are eliminated from the sample. The industry diversification
dummy, SEG, is equal to one for firms who operate in more than one industry and zero otherwise.
Multi-industry firms are defined as firms that operate in two or more two-digit SIC code
industries and no firm segment sales exceed 90% of total firm sales. The multi-country
diversification dummy, GEO, is equal to one for firms who operate in more than one country and
zero otherwise. Multi-country firms are defined as firms that operate in two or more countries
and no firm segment sales in a particular country exceed 90% of total firm sales. OIS is defined
as the firm's operating income-to-sales, while CES is the firm's capital expenditures-to-sales.





89


For Japan, we omit the CES variable from the specification due to infrequently reported figures.
Assets are defined as the natural logarithm of the firm's total assets. The independent variables
OIS, CES, and ASSETS are all measured relative to the value of the weighted-average multiplier
firms that form the basis for the excess value measure. Ownership concentration is defined as the
sum of individual and/or institutional ownership holdings that are equal to or exceed five percent
of a firm's common stock. OWNOtolO: =total ownership if total ownership < 0.10, = 0.10 if
total ownership > 0.10; OWN10Oto30: = 0 if total ownership < 0.10, = total ownership minus
0.10 if 0.10 <: total ownership < 0.30, = 0.20 if total ownership > 0.30; OWNover30: = 0 if
total ownership < 0.30, = total ownership minus 0.30 if total ownership > 0.30. Each model
specification also includes year dummies for 1992-1995.








Table 3-5
Multivariate Regression Estimates of Excess Values Using Only Pure-Play (SEG=0) Firms:
1991 1995

Panel A: Domestic Benchmark
German Japanese U.K. U.S.
Variables (1) (2) (3) (4)
Constant -0.083 0.194*** -0.348*** -0.211***
(-1.02) (3.74) (-5.85) (-7.35)
Multi-Industry Segment ___ ___ ___
Dummy (SEG)_ _
Multi-Country Segment -0.235** 0.249*** -0.024 0.122***
Dummy (GEO) (-2.27) (4.61) (-0.40) (3.77)
Relative Operating 0.036*** 0.070*** 0.226*** 0.018**
Income-to-Sales (OIS) (3.73) (2.65) (5.82) (2.13)
Relative Capital 0.042*** ___ 0.057*** 0.083***
Expenditures-to-Sales (3.45) (6.42) (7.04)
(CES)
Relative Total Assets -0.128*** -0.053*** -0.021 0.024*
(ASSETS) (-4.63) (-2.59) (-1.08) (1.88)
Ownership 3.504** -1.326*** 0.424 0.993***
Concentration < 10 (2.34) (-2.90) (0.88) (3.74)
(OWNOto 10)_ _
Ownership -1.328* 0.314 -0.545** -0.624***
Concentration 10-30 (-1.73) (1.37) (-2.29) (-4.16)
(OWN10to30)
Ownership -0.131 -0.735*** -0.257*** -0.078
Concentration > 30 (-1.22) (-3.93) (-2.27) (-1.10)
(OWNover30)
Ownership --
Concentration 10-30
interacted with SEG
(OWN10to30*SEG)
Ownership ~ --
Concentration > 30
interacted with SEG
(OWNover30*SEG)
Ownership 1.229 -1.533** -0.273 -0.627**
Concentration 10-30 (1.59) (-2.52) (-0.65) (-2.21)
interacted with GEO
(OWN10to30*GEO)
Ownership -0.281 0.636 0.374* 0.340**
Concentration > 30 (-1.24) (0.95) (1.74) (2.01)
interacted with GEO
(OWNover30*GEO)
Adjusted R2 0.12 0.09 0.22 0.07
Number of Observations 759 1,785 2,269 6,271








Panel B: International Benchmark __ ___
German Japanese U.K. U.S.
Variables (5) (6) (7) (8)
Constant -0.069 -0.091 -0.384*** -0.207***
(-0.80) (-1.38) (-4.93) (-6.97)
Multi-Industry Segment ___ ___
Dummy (SEG) __________
Multi-Country Segment -0.106 0.087 0.099 0.026
Dummy (GEO) (-1.18) (0.82) (1.05) (0.46)
Relative Operating 0.016* 0.251*** 0.255*** 0.010
Income-to-Sales (OIS) (1.66) (6.60) (4.42) (1.54)
Relative Capital 0.030** 0.057*** 0.094***
Expenditures-to-Sales (2.04) (5.55) (7.73)
(CES)
Relative Total Assets -0.134*** -0.046*** 0.002 0.030**
(ASSETS) (-4.31) (-2.81) (0.09) (2.14)
Ownership Concentration 2.714* -0.589 0.456 0.821***
<10(OWNOtolO) (1.73) (-1.07) (0.91) (2.83)
Ownership Concentration -0.929 -0.053 -0.520** -0.588***
10-30 (OWN10to30) (-1.25) (-0.22) (-2.12) (-3.65)
Ownership Concentration -0.147 -0.505*** -0.216* -0.041
> 30 (OWNover30) (-1.39) (-2.84) (-1.77) (-0.57)
Ownership Concentration ___
10-30 interacted with SEG
(OWN10to30*SEG)
Ownership Concentration __
> 30 interacted with SEG
(OWNover30*SEG)______
Ownership Concentration 0.902 -0.767 -1.411** -0.664
10-30 interacted with GEO (0.71) (-0.73) (-2.12) (-1.29)
(OWN10to30*GEO)___________
Ownership Concentration -0.199 0.495 0.080 0.393
> 30 interacted with GEO (-0.46) (0.55) (0.24) (1.32)
(OWNover30*GEO)
Adjusted R2 0.10 0.14 0.25 0.08
Number of Observations 517 1,537 1,684 5,131


ignmicant at I percent
in parentheses.


(-'*), 5 percent (*"*), and 10 percent (*) levels, Robust-White t-statistics


Regression estimates are from 1991-1995. Industrially diversified firms (SEG=1) are excluded
from the sample. Excess value is defined as the natural logarithm of the ratio of a firm's market-
to-sales ratio to its imputed market-to-sales ratio. With the domestic benchmark, the firm's
imputed performance is based on a weighted average of the corresponding pure plays within the
domestic market. The firm's imputed performance with the international benchmark is based on
a weighted average of the imputed values for the various countries in which the firm operates.
Firms with excess values that are greater than four or less than one-fourth are eliminated from the
sample. The industry diversification dummy, SEG, is equal to one for firms who operate in more
than one industry and zero otherwise. Multi-industry firms are defined as firms that operate in
two or more two-digit SIC code industries and no firm segment sales exceed 90% of total firm
sales. The multi-country diversification dummy, GEO, is equal to one for firms who operate in
more than one country and zero otherwise. Multi-country firms are defined as firms that operate





92


in two or more countries and no firm segment sales in a particular country exceed 90% of total
firm sales. OIS is defined as the firm's operating income-to-sales, while CES is the firm's capital
expenditures-to-sales. For Japan, we omit the CES variable from the specification due to
infrequently reported figures. Assets are defined as the natural logarithm of the firm's total
assets. The independent variables OIS, CES, and ASSETS are all measured relative to the value
of the weighted-average multiplier firms that form the basis for the excess value measure.
Ownership concentration is defined as the sum of individual and/or institutional ownership
holdings that are equal to or exceed five percent of a firm's common stock. OWNOto 10: = total
ownership if total ownership < 0.10, = 0.10 if total ownership > 0.10; OWN10to30: = 0 if total
ownership < 0.10, = total ownership minus 0.10 if 0.10 total ownership < 0.30, = 0.20 if total
ownership > 0.30; OWNover30: = 0 if total ownership < 0.30, = total ownership minus 0.30 if
total ownership > 0.30. Each model specification also includes year dummies for 1992-1995.














CHAPTER 4
FIRM VALUE AND DERIVATIVE USAGE

Introduction

Several recent studies find that focused firms are generally more valuable than firms that

are diversified along product lines. Berger and Ofek (1995), for instance, find that U.S. firms

trade at discounts ranging from 13 to 15 percent during 1986-1991. Extending this evidence

internationally, Lins and Servaes (1999) and Fauver, Houston and Naranjo (1999) also find that

diversified firms in developed economies generally trade at valuation discounts relative to

focused firms in those markets. While there are potential benefits to diversification, such as the

ability to effectively use internal capital markets and other firm resources as well as potential tax

benefits from leverage, the empirical evidence largely suggests that the costs of diversification

generally outweigh these benefits.1 Berger and Ofek (1995), Scharfstein and Stein (1997), Stein

(1998), and Rajan, Servaes and Zingales (1997) among other researchers show that the key costs

of diversification arise from agency costs associated with intra-firm coordination problems that

result in inefficient investment and cross-subsidization.

The valuation effects associated with derivative usage by diversified firms also presents

some potentially interesting insights into the magnitude of agency costs as a consequence of firm

organizational form. That is, while the organizational structure of diversified firms provides

some insights into the magnitude of potentially hedgeable risks that the firm may be exposed to, it

also provides some additional insights into the magnitude of agency costs associated with

derivative usage and corporate diversification. The agency costs associated with derivative usage

are particularly relevant in light of several well-publicized cases of losses incurred by firms as a


' An exception to this evidence is Fauver, Houston, and Naranjo (1999) who find that the benefits
from diversification outweigh the costs for firms in less-developed capital markets.

93