« j
m State
crsx'ty

  
   

 

 

 

 

 

ABSTRACT

THE EFFECTS OF A CORPORATION'S MULTINATIONAL
DIVERSIFICATION ON ITS COMMON STOCK SHARE-
HOLDERS' RETURN AND ON THEIR RISK EXPOSURE

BY

Larry Ross Lang

This dissertation is an effort to supply an answer
to a simple but important question: Does multinational
Operation benefit the common stock shareholder? Specifi-
cally, do the shareholders of a multinational corporation
benefit from the corporation's diversification of its
Operations on an international scale? That is, if a corpo-
ration diversifies into external markets to the extent that
it qualifies as being multinational, will its shareholders
benefit by achieving higher returns, by having a lower
exposure to risk or through some combination of these two?

The literature of finance is replete with many
injunctions about the Optimum balance between risk and
return. In practice, however, many of the questions that

operating Officers must face in deciding the tradeoff

Larry Ross Lang
between these two are left inadequately or ambivalently
answered. Deduction cannot answer such questions. What
is needed is a basis of fact and experience so that the
tradeoff can be decided with full knowledge of such facts
and experience.

To this end, the returns and risk exposure for share—
holders in multinational firms were compared with the
returns and risk eXposure for shareholders in comparable,
but essentially domestic corporations. This research
covered the period 1963 through 1971 and concentrated on
corporations that were headquartered in the United States.

The first corporate sample for this study con—
tained one hundred and forty-six firms that were equally
divided between multinational and domestic. Based on
three-digit SIC codes, these firms were grouped into nine
different industries. The second corporate sample was
merely a further stratification and realignment of the one
hundred and forty-six firms contained in the original sam-
ple. Because a less aggregative classification technique
was used in this second sample, only one hundred forty
firms were retained out of the original one hundred and
forty-six. Those firms that remained were grouped into

ten industry categories, again using three—digit SIC codes.

Larry Ross Lang
Shareholder return was measured using a quarterly price
relative that included both changes in market price and
dividends received. Four surrogate measures were used to
capture the different aspects of shareholder risk expo—
sure. These measures included: (a) total variability in
shareholder returns, (b) downside risk in shareholder
returns, (c) the general market segment of risk contained
in shareholder returns, and (d) the residual element of
risk unique to each security.

The major finding from the total nine-year period
was that the risk exposure on the common stocks of multi-
national corporations was significantly less than it was
for comparable domestic corporations. The Mann-Whitney U
statistics associated with the comparisons for the four
risk surrogates were statistically significant at the
.014, .021, .035 and .047 levels, respectively. A
secondary finding from the multivariate statistical
analysis for this same period was that the majority of
the difference in risk exposure between the two groups
centered on three aspects of risk: total variability in
shareholder returns, downside risk in those returns, and
the general market element of risk. A third finding was

that, overall, shareholder returns were not significantly

Larry Ross Lang
different between the multinational and domestic corporate
groups. The conclusion drawn from these findings was that
over a long-term time horizon the shareholders of a multi-
national corporation, while they do not receive substan—
tially higher returns, do have significantly less risk
exposure than do their counterparts in comparable domestic
firms.

For the second portion of the study, the nine-year
period was divided into three equal—sized sub—periods.

Each of these sub-periods correspond to a particular phase
of the United States domestic economy cycle. They included
a period of strongly expansive growth (1963 to 1965), a
period of limited growth (1966 to 1968), and one of sub-
stantial contraction (1969 to 1971).

The first finding from this part of the study was
that shareholder return for multinationals was significantly
greater than those for comparable domestic firms during the
period of contraction in the United States economy. The
MannAWhitney U statistic associated with these differences
was statistically significant at the .006 level. A second
finding was that during each of the three sub-periods
shareholder risk exposure on the multinational firms was

lower than it was for the comparable domestic firm. But

Larry Ross Lang
' more importantly, the risk differential that separated the
two corporate groups tended to become much larger when the
domestic economy entered a recession period. Two conclu-
sions were drawn from the findings of the shorter-term
sub-periods. First, there were potential benefits to the
shareholders of multinational corporations in the form of
reduced risk exposure during each of the different phases
of the domestic economic cycle, regardless of whether that
phase was characterized by a strongly expansive period, a
limited growth period, or a period of sustained economic
contraction. Second, and the more important of the two,
this reduction in risk exposure was most pronounced when
there was a substantial contraction of economic activity
in the domestic economy.

A fitting concluding question for this study would
be: ‘What answers were supplied for the initial research
question of whether multinational Operation benefits the
shareholder? First, the largest benefits have accrued to
the shareholder in the form of reduced risk exposure. This
benefit was largest when the domestic economy was experi-
encing an economic recession. This reduced risk, was not
accompanied by a lower return than was received by his

cGunterpart in a comparable domestic firm. In fact, during

Larry Ross Lang
a period of economic contraction there was considerable
evidence that the shareholders of multinational firms not
only had less risk exposure, but also had significantly
higher returns than did their domestic counterparts.
Quite clearly then, the overall answer would be that the
shareholder has benefited from a corporation's multi-
national diversification.

Several caveats need to be attached to this final
conclusion. First, while the corporate sample was felt to
be fully adequate, it encompassed only a limited number of
manufacturing industry classifications. Second, the time
period covered by the study contained certain eXpansion
patterns and special circumstances that will not, in all
likelihood, be repeated in future years. Last, changes in
trade and commercial policies of foreign countries could
substantially alter the future performance of multinational
corporations. The conclusions and implications from this

study must therefore be interpreted with care.

THE EFFECTS OF A CORPORATION'S MULTINATIONAL
DIVERSIFICATION ON ITS COMMON STOCK SHARE-
HOLDERS' RETURN AND ON THEIR RISK EXPOSURE

BY

Larry Ross Lang

A THESIS

submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Accounting and Financial Administration

1973

ACKNOWLEDGMENTS

5} .475 4/75

I wish to eXpress my appreciation to the members
of my dissertation committee, Dr. Roland I. Robinson
(Chairman), Dr. Reed M. Moyer and Dr. William H. Schmidt,
for their guidance, encouragement, and aid in completing
this dissertation.

To Dr. James Don Edwards and Dr. Gardner Jones,
who served as Chairman of the Department of Accounting and
Financial Administration during the period of my doctoral
program, I express my gratitude for their advice and
assistance. I take this Opportunity to acknowledge the
support I received as a National Defense Act Title IV
fellow and the financial assistance from the General
Electric Corporation.

My thanks to Dr. Mordechai Kreinin, for it was in
his international finance course that I gained much of
the background for this research. To my fellow doctoral
candidate and friend Charles L. McDonald, I express my
appreciation.

Mrs. Jo McKenzie deserves to be commended for her

ii

..
”0". '

3
“All "

 

 

 

efforts to reduce my nearly illegible drafts to type.
Finally, I owe the biggest debt to my wife, Pamela,

not only for her patience and encouragement during the

writing of my dissertation, but also for her editing

assistance in the preparation of this manuscript.

iii

TABLE OF CONTENTS

Page
ACKNOWLEDGMENTS. . . . . . . . . . . . . ii
LIST OF TABLES O O C O O O O O C O O O 0 Vi i
LIST OF FIGURES. . . . . . . . . . . . . ix
LIST OF APPENDICES . . . . . . . . . . . X
Chapter
I O IMRODUCTION O O O C O O O C O O 1
Purpose. . . . . . . . . . . . 1
Multinational Corporations . . 2
Definition 0 O O O O O O O O 2
Expansion Pattern. . . . . . 4
Justification. . . . . . . . . 5
Previous Studies . . . . . . . ll
Approach to This Research. . . 17
Summary. . . . . . . . . . . . 20
II. THEORETICAL CONSIDERATIONS AND
RESEARCH HYPOTIE SES 0 O O O O O O O O O C O 2 2
IntIOduction O O O O O O O O O O O O O O O 2 2
Basic Investor Valuation Model . . . . . . 23
Implications for the Expected
Earnings Stream 0 O I O O O O O O O O O 24
Implications of the Required Rate
Of Return 0 O O C O O O O O O O O O O O 25
Determinants of an Investment Project's
Earnings Stream . . . . . . . . . . . . . 27
Combining a Group of Capital Invest-
ment Projects . . . . . . . . . . . . . . 31
International Diversification in the
Array of Available Investment
Opportunities 0 O O O O O O O O O O O O O 34

iv

Chapter

III.

IV.

Assumption of an Unchanged Capital
Structure . . . . . . . . . . . . . . . .
Limitations on Changes in Dividend Policy.
Effects of Diversification into External
Markets on the Wealth of the Common
Stock Shareholder . . . . . . . . . . . .
Research Hypotheses. . . . . . . . . . . .
Rate of Return Differential Between
Multinational and Domestic Corpo-
rations . . . . . . . . . . . . . . . .
Short-Term Time Horizon. . . . . . .
Risk Exposure Differential Between
Multinational and Domestic Corpo-
rations . . . . . . . . . . . . . . . .
Long-Term Time Horizon . . . . . . .
Short-Term Time Horizon. . . . . . .
Summary. . . . . . . . . . . . . . . . . .

METHODOLOGY AND RESEARCH DESIGN. . . . . . .

Introduction . . . . . . . . . . . . . . .
Nature of This Research Study. . . . . . .
Description of the Corporate POpulation
and Samples . . . . . . . . . . . . . . .
Population Defined . . . . . . . . . . .
Sample Selection Criteria. . . . . . . .
Selection of Corporate Subjects -
Initial Sample. . . . . . . . . . . . .
Selection of Corporate Subjects —
Final Sample. . . . . . . . . . . . . .
Specification of the Empirical Response
Measures. . . . . . . . . . . . . . . . .
Introduction . . . . . . . . . . . . . .
Rate of Return Measure . . . . . . . . .
Surrogate Measures for Risk Exposure . .
Time Period Covered. . . . . . . . . . . .
Data Sources . . . . . . . . . . . . . . .
Specification of the Statistical Models. .
Significance Levels for Statistical Tests.
Limitations of the Methodology . . . . . .
Summary. . . . . . . . . . . . . . . . . .

RESULTS OF PRELIMINARY STATISTICAL TESTS . .

IntrOduction O O O O O O O O O O O O O O O

Page

39
4O

41
43

44
44

48
49
50
52

54

54
54

56
56
60

62

64

67
67
69
70
81
84
87
92
93
94

96

96

Chapter

Rate Of Return Differential —
Short-Term Time Horizon . . . . . .
Difference in Risk Exposure —
Long-Term Time Horizon. . . . . . .
Difference in Risk Exposure -
Short-Term Time Horizon . . . . . .
Summary. . . . . . . . . . . . . . .

V. RESULTS OF FINAL STATISTICAL TESTS . .

Introduction . . . . . . . . . . . .
Difference in Risk Exposure over a
Long-Term Time Horizon. . . . . . .
Difference in Risk Exposure for
Short—Term Time Intervals . . . . .
Summary. . . . . . . . . . . . . . .

VI. SUMMARY, CONCLUSIONS AND IMPLICATIONS,
AND SUGGESTIONS FOR FUTURE RESEARCH .

Summary of This Study. . . . . . . .
Conclusions and Implications from
the Results . . . . . . . . . . . .
Findings for the Long—Term Time
Horizon . . . . . . . . . . . . .
Findings for the Short-Term
Time Intervals. . . . . . . . . .
Results for the Individual
Industry Categories . . . . . . .
Suggestions for Future Research. . .
Summary and Outlook. . . . . . . . .

B IBLIOGMPI” O O O O O O O O O O O I O O O O 0

vi

Page

100
122

150
154

158
158
160
178
206
210
210
218
218
224
234
240

242

253

Table

LIST OF TABLES
Page
List of Qualifying Industry Groups . . . . . 62
Initial Sample of Corporations . . . . . . . 64
Final Sample of Corporations . . . . . . . . 67

Results of the Mann-Whitney U Test of

Research Hypothesis H1: Rate of Return
Differential During a Period of Substantial
Contraction in the Domestic Economy. . . . . 102

Results of the Mann-Whitney U Test of

Research Hypothesis H2: Rate of Return
Differential During a Period of Rapid

Growth or a Period Of Limited Growth to Near

Zero Growth in the Domestic Economy. . . . . 114

Results of the Mann—Whitney U Test of

Research Hypothesis H3: Risk Exposure
Differential Over a Long Term Time

Horizon: Total Multinational and

Domestic Sample. . . . . . . . . . . . . . . 126

Results of the Mann-Whitney U Tests of

Research Hypothesis H3: Risk Exposure
Differential Over a Long Term Time

Horizon: Individual Industry Groups . . . . 128

Results of the Two-Way Multivariate

Analysis of Variance Test for Research

Hypothesis H3: Difference in Risk Exposure

Over a Long Term Time Horizon. . . . . . . . 163

Results from the Two-Way Multivariate

Model: Lease Squares Estimates of
Main-Class Effect: Type of Corporation. . . 166

vii

Table Page

10 Results from the Two-Way Multivariate
Model: Discriminate Analysis for Main—
Class Effect: Type of Corporation . . . . . 169

11 Results from the Two-Way Multivariate
Model Univariate F Tests for Main-Class
Effect: Type of Corporation . . . . . . . 171

12 Results from the Two-Way Multivariate
Model Step—Down F Tests for Main-Class
Effect: Type of Corporation . . . . . . . . 173

13 Results from the Multivariate Repeated
Measures Model for Research Hypothesis
H4: Difference in Risk Exposure Over a
Short Term Time Horizon. . . . . . . . . . . 182

viii

’1

"an.

 

Figure

LIST OF FIGURES
Page

Significance Level for the Mann-Whitney

U Tests Associated With the Differences

in Risk Exposure Between the Common Stocks

Of the Combined Multinational Sample and

Those of the Domestic Sample . . . . . . . . 152

Difference in Mean Score on the Overall
Variability Risk Measure and the Downside

Risk Measure: Estimated Combined Domestic

Group Score - Estimated Combined Multi-

national Group Score . . . . . . . . . . . . 184

Difference in Mean Score on the Overall
Variability Measure and the Downside

Risk Measure for the Electrical Equipment

and Supplies Industry: Domestic Group

Score - Multinational Group Score. . . . . . 186

Difference in Mean Score on the Overall
Variability Measure and the Downside

Risk Measure for the Food and Beverage

Industry: Domestic Group Score -

Multinational Group Score. . . . . . . . . . 193

Difference in Mean Score on the Overall
Variability Measure and Downside Risk

Measure for the Industrial Equipment

Industry: Domestic Group Score -

Multinational Group Score. . . . . . . . . . . 197

ix

LIST OF APPENDICES

Appendix Page
A Alternative Empirical Response Measures. . . 259
B Evaluation of the Relationship Among the

Five Surrogate Risk Measures . . . . . . . . 265

B-l Correlation Matrix for the Five Surrogate
RiSR Measures. I O O O O O O O O O O O O 0 O 267

B-2 Principal Components for the Correlation
Matrix in Table B-1. . . . . . . . . . . . . 269

B-3 Maximum Likelihood Factor Analysis:
Coefficients for the Two Factors . . . . . . 269

n... .

v
.4...“

 

b..

:fils

«‘~-

I
7“
"l:

 

»

‘:
‘\
‘~

Q].
l D

CHAPTER I
INTRODUCTION

Purpose
This research is an attempt to evaluate the effect
of multinational operation on common stock performance.
Multinational Operation exists when a firm derives a sig-
nificant percentage Of its total business activity in
countries external to the one where it is headquartered.
This evaluation was made by comparing the market perform-
ance of the common stock for a sample of corporations
which do have extensive foreign content in their total
°Perations with that of a comparable sample of corpora-
tiJOns with little or no foreign content. For purposes of
taidis study market performance included not only the rate
(3f? return achieved by the shareholders but also an allow-
a1ice fOr differing degrees of risk exposure. Additionally,
tirlis research will investigate whether the common stocks
(315 corporations with extensive foreign Operations are less
affected by cyclical swings in the overall economic

aQtivity of the country where they are headquartered.

1

2

Because the empirical portion of this study is
based on ex post data, the results and the conclusions
drawn from those results will be primarily positive in
that they describe the historical relationship in the mar-
ket performance of the two groups of corporations. Cer-
tainly, from the results of this study, it would be
possible to draw some tentative normative conclusions,
but this was not the purpose of the study. Consequently,
in presenting the results there was no attempt made to
formulate normative conclusions.

This research was intended to provide empirical
evidence as to whether diversification by a corporation
across national boundaries has benefited the firm's common
stock shareholder. The results should also add to the
bOdy of current knowledge concerning the effects of diver-

sification in the capital investment decision process on
tile shareholder when this diversification is achieved by
itlle corporation selecting investment Opportunities in

‘3C>untries external to its headquarters.
Multinational Corporations
lassginition
This study adOpted the present convention of
(afiafining a corporation with extensive foreign content in

jvts total business activity as being a multinational

 

 

 

u

 

‘9

I...

'l.-

0 Was.

'5...

1'...
t

NJ

“A.
x."

‘-

.v.
.5"

 

 

corporation. While the term multinational corporation
appears Often in the literature, there is no consensus as
to exactly which corporations can prOperly be considered
multinationals. There is not even total agreement as to
what criterion is most important in discriminating between
domestic and multinational corporations. For these rea-
sons, one of the initial steps in this research was the
development of a working definition for "multinational
corporation."

Like a number of other sources in the literature,1
the principal criterion used to distinguish a multinational
from its domestic counterpart was the geographical loca-
tion of sales and manufacturing activity. The following
Standards were used: (a) that the firm derive a minimum
of 20% of its total business sales in countries external

t<> the one where it is headquartered, (b) that such ex-

teirnal sales come from markets in at least 5 countries,

1See, for example, Michael Z. BroOke and H. Lee
Iifiinmers, The Strategy of Multinational Enterprise-
SEEEganization and Finance (New YOrk: American Elsenier
I-:’1:l'b1ishing Company, Inc., 1969): John Fayerweather,
£E£3§ernational Business Management (New York: McGraw-Hill
(DOk Company, 1969): RObert Warren Stevens, "Scanning the
Ialtinational Firm," Business Horizons, Vol. XIV, No. 3
(CTune, 1971), pp. 47-54. And the above list is by no
l"‘leans exhaustive, it suggests that the use of such a
QI‘iterion is well founded.

O
u.-
a

 

r-n

HA.

I" .

...‘

a"
\

I...

 

 

i.:‘
ll.

in.

 

.-
I

 

 

4

and (c) that the firm have manufacturing locations in at
least 3 external countries. As a point of information,
the pOpulation of multinational firms used for this study
covered various types of industry, especially, those in
the manufacturing sectors. Chapter III provides a further
description of the population classification method and
the justification for its usage. Also contained in that
chapter is an evaluation of the possible limitations to

using this classification method.

Expansion Pattern
Growth of the multinational corporation in the

Past 50 years, and especially during the past two decades,

has been very rapid. Corporations with well established

multinational Operations have continued and in many cases

accelerated their expansion in international markets
during this period. At the same time many small to medium-
sized corporations, which had previously concentrated their
GaiffOrt in the domestic market, turned their attention to
inacreasing their participation in markets external to the
<>r1e where they were headquartered. This expansion into
E3Utternal markets, however, has not been solely a function
(31? increased exports; much of it has simultaneously in-

vC>lved an increase in manufacturing activities within the

5

external market. As a result of this rapid growth, the
composition of firms within certain industries is approach—
ing the point where 40 to 50% of them can prOperly be clas—
sified as multinational, based on the previously stated
criteria. Rolfe2 found that 75 to 85 of the 200 largest
United States corporations — 1967 Fortune listing - had
25% or more foreign content in their total activity.

Last, growth of the multinational corporation is
by no means limited to those which are headquartered in
the United States. A similar expansion has character-
ized many EurOpean based firms as well. So, while the
empirical data for this research was drawn from U.S. based
corporations, it is hOped that the results can be general-
ized to foreign-based firms as well. Certainly, the re-
8ults can be utilized to suggest the underlying pattern

and some of the basic relationships that exist.

Justification
While the concept of diversifying amongst a set of
Irj~sky assets has long had intuitive appeal, it was the

\

2Sidney E. Rolfe and Walter Damm (editors), The
gzslltinational Corporation in the World Economy (New York:
¥?l?aeger Publishers, 1970), p. 17. The authors defined
' foreign content" according to any of these four criteria:
Bales, employment, production or investment.

 

I‘d-4
.

sun
‘

lob -

can.‘

to-

 

0
th-

 

.44.‘
I

'9.‘

"A.

“a.

a.“
s"

s

 

 

 

 

6

pioneering work of Markowitz3 which first provided a rigor—

ous mathematical explanation as to how and why risky assets

should be combined into a portfolio. Markowitz's contri—

bution was to show, by use of the mean-variance two para—

meter model, that a definite mathematical solution existed

for how a portfolio should be diversified among the uni-

verse Of risky assets. And he further showed that norma-

tive rules could be derived for diversifying amongst a

group of risky assets. Much of the early impetus to the

develOpment of portfolio theory, and its initial applica-

tions, concerned the construction of portfolios containing

corporate common stock as the principal risky asset.
Later, however, the possibilities for constructing a port-

fOIiO from the corporation's available capital investment

oPportunities received increased attention.
As indicated in a recent study by Mao,4 the use of

portfolio construction techniques in the selection of

Qa‘Pital investment Opportunities has gained fairly wide

a‘chptance in the theory of finance. In that study Mao

\

3Harry M. Markowitz, "Portfolio Selection,"
1 (March, 1952), pp. 77-

wnal of Finance, Vol. VII, No.

:1: 4James C. T. Mao, "Survey of capital Budgeting:
11 . ,, .
C eory and Practice, Journal of Finance, Vol. XXV, No. 2

stated:

. . .most financial writers7 argue that firms
should choose portfolios rather than [single
capital investment] projects. . .

7For an exception, see Steward
C. Myers, "Procedure for Capital
Budgeting under Uncertainty," Indus—
trial Management Review, 9 (Spring,
1968), pp. 1-20.

Several recent books on corporate financial management
provide further evidence for the acceptance of the useful-
ruess of diversification in the selection decision: and in
skirticular the use portfolio techniques in that process.
11118 is best illustrated by the following direct quotes

“fliich are offered without comment:

In the present section, however, our primary con—
cern is to see how the concepts and methods
developed by Markowitz can help the financial
management of a firm to Optimize its real
investment decisions. .

The firm should be cognizant of the potential
benefits of diversifying investment projects,
to achieve the best combination of expected
net-present value and risk. An Optimal com-
bination would be the one that maximized the
value of the firm to its shareholders.6

\

5James C. T. Mao, Qpantitative Analysis of Finan—
Cgaeglg‘l Decisions (New York: The Macmillan Company, 1969),
“ 289.
1;: 6James C. Van Horne, Financial Management and
Q - . ,
:t:1b1:lac , lst ed. (Englewood Cliffs, N.J.: Prentice-Hall,
Q , 1968), p. 91.

8

In much of what follows the portfolio problem
will be developed in a context in which the
differences between investments in stocks and
bonds, on the one hand, and business capital
expenditures, on the other, are assumed to be

of negligible importance.7

The previously noted books along with recent jour-
nal articles suggest that one of the principal purposes in

constructing such a portfolio is to minimize fluctuations

1J1 the overall earnings stream accruing to the corporation.
Thus, by judicious selection from among its available in-
Vestment Opportunities the corporation is able to Signifi-

<Bantly reduce fluctuations in the stream of earnings

accruing to each common stock shareholder. And, to the

eJ‘Ktent that there is less variability in the market price
<>f5 those shares, the risk to the shareholder will be less.

It is interesting to compare and contrast these

pronouncements from the theory with the current beliefs

E‘lfixi practices of corporate executives. In his study Ma08

”13(31xnd that executives look upon diversification as an
attempt to combine major business actiVities or pursuits

£14111:O a balanced portfolio, where each one of these

\

7Harold Bierman, Jr. and Seymour Smidt, The
The Mac-

E“:~JEzital Budgeting Decision, 3rd ed. (New York:
hi llan Company, 1971), p. 345.

"Survey of Capital Budgeting.

8
Mao,

 

n...~
-

.gn‘.

Do!‘
. a

 

 

 

5'7;

n..-

 

-
o u,

n.‘

:v’:

‘5
1"
a

fit

~-‘

‘e

‘1

 

 

9

activities will likely involve a series of investment

suggests that the port-

projects. The theory, however,

folio be constructed using the individual investment

projects. Also, executives indicated that they were more

concerned with the size of the growth in earnings rather

than just stability in the earnings stream. In summary
then, what this study suggests is that executives recog-
nize diversification among major business activities as a
Viable means of lowering the corporation's risk exposure.
The study also suggests that by combining a series of
these major business activities into a portfolio, the

c3C>:::‘poration, or more prOperly its executives, may be able

to attain or more closely approximate their established

goal for earnings growth.

Certainly, one of the available avenues for this

diversification would be the selection of investment
QPportunities which involve expansion into markets ex-
ternal to the one where the firm is headquartered. In
15>Eilrticular one plausible motive for the recent expansion
:i”111:o external markets by United States based corporations
was their desire to effect greater diversification in
.tzlnlfiair portfolio of capital investment Opportunities.

veral aspects of this expan31on lend credence to this

a:'
blanation. First, as was noted in the previous section,

 

-¢b-.

.
I'll!
o

I..'I

 

on.

'cas
a

O
l'l'

 

D
5"
‘1‘

"Q. ~
84..
a

\.: ‘

u
a.
. c

.Kl.-

'J

 

I A‘
\v.

ll.
«1'
I I

dun
c ‘1
III

10
much of this expansion has involved the commitment Of con-

siderable corporate resources in the form of direct foreign

investment. Of course, even an expansion in exports re-

quires some investment of corporate resources, although to

a lesser degree. Second, the decision to expand into ex-

ternal markets would qualify as a "major business activity"

for purposes of the corporate executive's perspective of

Last, since the rate of

What constitutes diversi fication .

growth and the cyclical patterns in that growth would be
unique to the country where the firm is presently head-
quartered, then a decision to develOp a significant foreign

Content in the corporation's total business activity might
Well be dictated by the corporate executives' attempts to

atitain his goal of earnings growth.

Accepting for the moment the above explanation for

a Corporation's expansion into external markets, a sequel

question is: What are the effects of this diversification

9“ the corporation's conunon stock shareholder?

This research attempts to examine the effects of
bI'm-is diversification on the rate of return achieved by the
$11acreholder and On the degree of his risk exposure. A
3QQond, but closely related, question which it explores is
”1.1% ther the corporation by diversifying on an international

s Q
file is less vulnerable to the effects of the business

 

 

11

cycle in the country where it is headquartered. Specifi-

cally, the central point is whether the rate of return to
the shareholder in a multinational corporation has less
pronounced shifts, and therefore less risk exposure, when
the total economic activity of the country where it is

headquartered gOes through an expansion—contraction cycle.

Previous Studies

To date the two parameter portfolio model, as

Originally develOped by Markowitz, has seen only limited

application to the problem of diversifying on an inter-

national scale. Earlier studies have concentrated on the

finstification for diversifying amongst the respective

e"~I'uity markets for a particular sample of countries. Thus

9
C-“'3l'-"‘l.1bel in an early study analyzed the benefits to an in-

veator when the portfolio consisted of couuuon stocks

<31: . . . 10
awn from the equity markets of 11 different countries.

As a surrogate for individual securities, he utilized a

sliars price market index from each of the 11 countries.

 

\
9 9Herbert G. Grubel, "Internationally Diversified
t>~"’:tfolios: Welfare Gains and capital Flows," American

&\nomic Review, Vol. LVIII, No. 5 (December, 1968),
- 1299-1314. Hereafter referred to as the Grubel study.

:11 ”All of the 11 countries of his sample could be
firacterized as having well develOped industrial
afi Qnomies .

12

The two parameters for the model were the g§_post realized
annual rate of return on each of the indices for the

period 1959 through 1966,and the standard deviation in

those rates of return as the risk surrogate. After com-

puting the return and risk parameter for each country,

these were then combined into portfolios using the methods

of quadratic programming. His principal conclusion was

‘Ehat.a series of diversified portfolios, which contained
Varying mixes of the indices from the 11 countries, not
Gully achieved higher returns than a portfolio comprised

s<>lely of the United States market index, but did so with

-ltiss exposure to risk.11 From this Grubel concluded that

i3n investor was justified in holding such a mixed port-

fC>1io. By doing so he could achieve superior returns with
less risk exposure.

Levy and Sarnat12 followed with a greatly expanded

Study that included 28 countries 3 and covered the period

11~“_
11That is, the standard deviation for the diversi-

fied portfolios was smaller.

12Haim Levy and Marshall Sarnat, "International

:I2’5L\Jersification of Investment Portfolios," American

EgstSEggnomic Review, Vol. LX, NO. 4 (September,1970), pp. 668—
)5
. Hereafter referred to as the Levy and Sarnat study.

13The expanded sample includes Grubel's original
:1; :l‘ . _
i‘ countries plus others, some of which would be categor-
§§= Ed as having developing rather than developed industrial

(anomies.

13

1951 through 1967. Their conclusions paralleled Grubel's

earlier work in that they also concluded that a diversi-

fied portfolio consisting of some mix of share price

indices from the 28 equity markets achieved a higher return

with less risk exposure than did the one security port—

folio: the United States stock market index. This study,

like Grubel's, relied upon a share price market index as

‘fhe applicable "risky asset" in constructing the diversi-

Ified.portfolios. They also used the standard deviation in

the realized rate of return on the respective market in-

dices as the second parameter for the model: the surrogate

for risk .

A research study by Lessardl4 addressed the ques-

tIiJon of diversification among countries in a different

manner. First, the securities, which formed the basis of

his portfolios consisted of individual common stocks
ESEElected from the equity markets in four develOping

Q0untries. By comparing the various portfolios, he con-

lczildaded that diversification among the securities of the

;tE<31xr countries was superior to that achieved by diversify-

:l”t‘sg within a single country. As a further conclusion, he

\
14Donald Roy Lessard, "Multinational Portfolio

:i~‘Jersification for DevelOping Countries," Unpublished
assis, Stanford University (1970).

”U

 

 

l4
determined that much of the variability for each of the

four equity markets was accounted for by a common market

effect. But that after removal of this common element the

resulting residual for each of the markets tended to be

independent of each other. This latter finding parallels

that of Grubel and Fadner.l They found that the _e_x post

returns on a series of stock market sub-indices from the

United States, British, and West German equity markets

Were positively correlated. Although they made no attempt

to estimate a common element between the three markets,

the positive correlation they observed suggests there is

a common factor between the three markets. The residual

factors that are unique to each of the markets would be a
Plausible explanation as to why both Grubel plus Levy and
Sarnat were able to achieve superior results from their
r'eSpective internationally diversified portfolios.
Clearly, there are some gaps in the previous re-
search studies in the area of international diversifica-
tion. To date most of the work has centered on analyzing

the benefits to be gained when an investor diversifies his

1moldings by purchasing the common stocks of a group of

 

\
15Herbert G. Grubel and Kenneth Fadner, "The Inde-

:?ndence of International Equity Markets," Journal of

\lnance, Vol. XXVI, No. 1 (March, 1971), pp. 89-94.

 

15

foreign based corporations. One question which immediately

comes to mind is whether some of the same benefits that can

be gained from international diversification among equity

markets might also be captured, to some extent, by an in—

dividual corporation electing to diversify its business

activity on an international scale. Thus, rather than

expecting the investor to diversify his holdings on an
international scale, the corporation might be able to
accomplish somewhat the same benefits by diversifying its

buSiness activities into a number of external markets.

In a journal article Myers16 stated that, given

certain conditions, the corporation should rely upon port—
folio diversification by the investor rather than attempt-
ing this diversification through the selection of certain
ca‘I>ital investment projects. A similar position was

a.
annced by Bierman and Smidtl7 in their book. But when

E C I O O I I O
iQed With the problem of international diver81fication
hh a o o u o
Gare are definite limitations on the extent of the port-

EQlio diversification that can be accomplished by the
lhfiividual investor.

\

 

E 16Steward C. Myers, "Procedures for Capital
Rudgeting under Uncertainty," Industrial Management
W, Vol. 9, No. 3 (Spring,1968), pp. 1—19.

17Bierman and Smidt, op. cit., p. 365.

V.

o... I

a
all:

'0'-

'IA.

we...

 

‘9‘“.

I...

I...

IV
lb.

I..

u
v‘

‘-

 

 

 

16

First, requiring the inVestor to purchase the com-
mon stocks of foreign-based firms, that will in all likeli-
hood be denominated in the local currency, directly sub-
jects him to changes in currency parity plus any restric—
tions on convertibility. Second, there is generally a
dearth of financial information on most foreign based
firms, making the investor's selection process little
better than a naive draw. Last, due to the non-existence
of broadly based mutual funds in many equity markets,
81“all investors would be hard pressed to attain the
desired degree of diversification, especially when there
were a number of different markets.

It was not the intent of this discussion to suggest
that diversification on an international scale is unprofit-
able. Rather it was meant to highlight some of the prob-
lems . . . . .

that could arise if an investor is faced with the
p“llrchase of common stock in a number of foreign corpora-
tiQns. One question which this research will attempt to
butQvide a partial answer to is whether multinational cor-
erations offer an alternative vehicle for an investor to
Qfi~IEDture some of the benefits of international diversifica-
t 2LOn. One reason for the above reservation as to providing
thy a partial answer stems from the fact that this re-

3%
a~rch does not attempt to compare the results for a group

17
of multinational corporations to some internationally
diversified portfolio. Consequently, no conclusions can
be made as to the efficiency of a portfolio of multi—
national corporations versus that of a portfolio drawn

from the respective equity markets in a group of countries.

Approach to this Research

Chapter I has primarily dealt with a statement of
the purpose of this research and the justification for its
undertaking. Chapter II begins with a brief presentation
of the basic investor valuation model, the parameters
which comprise it, and how investor expectations enter the
mOdel. Next, using the parameters of the valuation model,
a breakdown is given to show how the effects of the indi-
Vidual capital investment projects enter the model. And,
in particular, how the parameters are affected when a cor-
lporation combines a series of investment projects into a
Single portfolio. A broad review of the major groups of
‘economic factors that influence a particular capital in-
‘Vefitment project is then given. Following this, the dif-
ferences between the factors which affect an investment
p”Qject in the domestic market and those influencing a
project in an external market are analyzed. Implications

are then drawn as to how this would affect the parameters

18
of the valuation model.

The results from the analysis of each of the para-
meters Of the first sections of Chapter II are then drawn
together to provide some implications as to how the valu-
ation model can be expected to reSpond when a corporation
derives a significant portion of its business activity
from capital investment projects that are concentrated in
external markets. From these hypothesized responses of
the model certain test implications were drawn. Using
these test implications, Chapter II concludes with a state-
ment of the principal research hypotheses of this study.

Chapter III begins with a description of the pOpu-
lation of corporations from.which the sample was chosen
for use in the empirical portions of this research. The
aCtual sample selection criteria are given with a justifi-
cation for their use. The chapter continues with the
deveIOpment and explanation of the response measures used
in the empirical validation of the test implications
‘deVeIOped in Chapter II. The chapter concludes with a
:presentation of the statistical models which were utilized
to test for differences and shifts in the response measures.

Chapter IV presents the preliminary results from
the statistical models using the initial corporate sample.

1”Ollowing this, Chapter V presents further results from

19
different statistical models and a slightly altered
sample. Both chapters contain a discussion of the results
and provide interpretations as to the significance of the
findings.

Chapter VI concludes with a summary of the study
and the major findings and implications of this research.
Some recommendations are then made as to future research
in the area.

To some extent, the research presented parallels
the Grubel study and the Levy and Sarnat study in that it
attempts to evaluate the effects of international diversi-
fication on the common stock shareholder. But, whereas
b0th Of the above studies used the international equity
markets as their principal vehicle for diversification,
this research concentrates on the effects of that diversi-
fication when it is undertaken by the individual corpora-
tion through eXpansion of its activities into external
markets. By concentrating on the individual corporation,
and in particular United States based corporations, this
research overcomes many of the potential and real prOblems
of eJitpecting an investor to purchase common stock in a
group of foreign based firms. Also, by concentrating on
the most recent historical experience this research avoids

many of the currency controls, direct investment

20
regulations and governmental controls that could have sig-
nificantly biased the two earlier studies. Last, the re-
sults should be of interest to a broader group within the
financial community. These would include:

(a) common stock shareholders in multinational
corporations,

(b) an investor who is attempting to achieve
international diversification in his equity
holdings,

(c) the financial executives of corporations
which are currently multinational in their
scope of Operations, and

(d) financial executives in a corporation which
is presently considering its initial expan—

sion into external markets.

Summary
Chapter I has stated that the purpose of this
reSearch was to provide empirical evidence on the effects of
‘jiVersification by the corporation into external markets
c”"the common stock shareholder. For purposes of this
study a multinational corporation is defined as any corpo-
ration that derives a significant portion of its activity

frrnn markets external to the one where it is headquartered.

21
Further, it must meet certain criteria as to the extent of
this diversification as well as certain conditions for
manufacturing locations.

Using several studies from the financial literature,
it was shown that diversification in the selection of capi-
tal investment projects, and in particular a portfolio
approach to that diversification, is a generally accepted
part of the theory of finance. Further, it was shown that
the recent expansion by corporations into external markets
Could be conceived of as one aspect of this diversification
Process. This lead to the following sequel question: What
have been the effects of this form of diversification on
the common stock shareholder? It was then pointed out that
this research addresses itself to examining and evaluating
the effects of such diversification on the shareholder.

The next section of the chapter reviewed some of
the previous studies that are directly related to this
research,and summarized their findings. Next, it was shown
that there are some serious shortcomings to an investor
diversifying his holdings using the various foreign equity
markets. The chapter concluded with a brief statement of
the approach to be used in this research and a comparison

of this research to the earlier work in the field.

CHAPTER II
THEORETICAL CONSIDERATIONS AND RESEARCH HYPOTHESES

Introduction

This chapter begins with the Specification of the
basic investor valuation model and a brief review of the
Parameters which comprise it. Using this model, and in
Particular its parameters, the effects of an individual
capital investment project on the valuation of a corpora—
tion's common stock was analyzed. Initially, some simplify-
ing assumptions are made and certain constraints are placed
on the parameters to allow a piecemeal approach in the
analysis. Following this, the analysis considers the
effects of a corporation diversifying its portfolio of
caPital investment projects to include not only those from
the domestic market where it is headquartered, but also
other projects from external markets. Finally, the piece-
mealtechnique of the early sections is dr0pped and we
attempt to draw some broad conclusions as to how such a
diversification by the corporation might be expected to
affect the valuation of its common stock. These

22

23
implications from the theory form the basis for the re-
search hypotheses that are developed in the final section

of the chapter.

Basic Investor Valuation Model
This section begins with a brief statement of the
basic investor valuation model. As of any particular

point in time, the price of a security can be written as:

P0 = tgl E[EPSt(l-bt)]
(1 + re)t

 

where P0 = the current price of the security
when t = O

E[ = the ex ected value of the random
variable within the brackets [ ]

bt = the percentage of the corporate
earnings which are to be retained
in period t for reinvestment pur-
poses, this for t = l, 2, ...m

t = the time increment

re = the required rate of return appropriate
to the corporation's particular risk class

EPSt = the earnings per share in time period t,

this for t = l, 2, ... m

‘A8 it is formulated, the present price of a security, Po'
is the summation of investor expectations as to the future
stream of earnings, these discounted back to the present

tiMe at the apprOpriate rate, r An implicit assumption

e0
Of this model, as specified, is that the discount rate, re,
13 invariant as to time horizon. The model also implies
that investors can formulate expectations as to future

reinvestment Opportunities for the corporation and there-

fore can anticipate the future retention rate, bt' The

 

-.
9!

.n.

I
tn.

na~

 

 

'.
do

fl.

(1,
All

24
actual amount within the brackets becomes the dividends
which accrue to the investor over time.

As given the time horizon is infinite, this could
just as well be specified as some finite time interval.
Then the price would have two components: the stream of
dividends that would accrue over this finite time span and
the terminal share price that would be received at the end
of the time span.

To simplify and permit a piecemeal analysis for
each of the parameters of the valuation model, several
assumptions are necessary: first, the capital structure
of the firm is held constant: second, the dividend policy
of the firm is assumed to be unchanged; and third, the
business riSk of the firm, at the outset, is assumed to

be constant.

Implications for the Expected Earnings Stream

As of a point in time, investor expectations, for
the future earnings stream, will include not only the
earnings stream from the firm's present investments, but
also those streams which will arise from future investment
Opportunities as well. If the earnings streams for present
and future investment projects were characterized by cer-

tainty, then finding the expected value would merely

25
entail summing these flows. But the earnings stream for
the majority of capital investment projects has some degree
of uncertainty. Consequently, the actual earnings stream
will be a random variable whose value in any particular
time period will depend upon a host of factors. These
will influence the overall size of the earnings stream,
the timing as to how that stream will occur over time, and
last, the overall variability in the array of possible

flows during the life of the project.

Implications of the Required Rate of Return

The future stream of earnings which accrue to a
share of common stock is characterized by a continuum of
possible outcomes. Since this stream is not known with
certainty by an investor, then the required rate of return
parameter, re, of the basic valuation model includes not
only a risk free interest rate, but also some provision for
the risk inherent in the earnings stream. So an investor

at the margin will demand a rate of return, r which in-

e'
corporates his attitudes toward the time value of money
Plus an additional increment attributable to the risk he
Perceives in the future expected stream of earnings.

The components for the required rate of return can

be shown as:

26
r =i+BR+FR

the risk free rate of return, or
the rate for the pure time value
of money

BR = a provision for business risk

FR a provision for financial risk.

§
0
H
(D
H
II

The single-most important determinant of a corpo-
ration's perceived business risk, BR, is the variability
in the firm's operating income. And, since the variability
of the total stream is a function of the variability of its
components, then it follows that the perceived business
risk of a corporation depends on its existing capital
investment projects plus its anticipated future capital
investment opportunities. While a subsequent section of
this chapter investigates the possible effects that an in-
vestment project can have on a corporation's perceived
business risk, BR, it is nearly an impossible task to
single out the contribution that a single investment
project makes to the overall business risk of the firm.
If, however, a series of projects are grouped into broader
categories, according to type of business activity, it is
generally possible to estimate whether their contribution
to the corporation's perceived business risk is positive
or negative. But the magnitude of the resulting incre-
ment, despite this simplification, is still nearly im-

Possible to estimate .

an;

....

....

'Ovl

CK,“

u§.,

:- .
lu'
"a.

 

 

:‘n
'u‘

‘1‘
s“

.A

n,‘

27

The investor's attitude toward the perceived finan-
cial risk, PR, of the corporation is the final component of
the required rate of return. The single-most important
determinant of financial risk is the sources of capital
which the corporation has utilized to finance its Opera-
tions. That is, to the extent that these other capital
sources have preference to the firm's earnings, then the
earnings available to the common stock shareholder will

have an additional source of variability.

Determinants of an Investment
Project's Earnings Stream

To attempt anything approaching a complete enumer-
ation of the myriad of economic factors which could influ—
ence the earnings stream, and therefore the ultimate out—
come of a single capital investment project, would be a

formidable task. One plausible solution is to collapse the
multitude of individual factors into a limited number of
broad general categories. By doing this, the following
model could be specified:

E[Rij] = f(GB, IDj.Ui)

where R-- = the earnings stream for capital

1] investment project i, within
industry class j.

GB = an index of general business
activity

IDj = an index of economic factors
attributable to industry j
Ui = unique factors attributable to

project i

28
What this model implies is that the factors affecting the
outcome, and in particular the variability in the array of
possible outcomes for an investment project, can be attri—
buted to three sources: the level of general business
activity within the economy, the level of activity and
collection of events which stem from the industry where the
project is located, and the residual factors not accounted
for by the above two categories and therefore considered
unique to the particular project.

The exact form of the functional relationship will
not be estimated by this research but some preliminary ob—
servations can be offered. First, for most investment
projects the correlation between its earnings stream and
the level of general business activity will be positive.
Likewise, the correlation to the series of economic events
which arise within the industry where the project is
located will also be positive. That is, an event which
benefits the industry will also be favorable to the out—
come of the investment project, and conversely, an event
with a negative effect in the industry will likewise have
a negative effect in the outcome of the project. Conse-
Quently, a change in either the general business index,
GB, or the particular industry index, 1., will cause a

J

(:hange in the same direction for the project's earning

29

stream. But the exact magnitude of the resultant change
remains unspecified. One study which lends some weak evi—
dence as to the nature and estimated magnitude of the
underlying relationship is the work by King.1 The results
of his study suggests that approximately 50% of the change
in the market price of a security can be attributed to a
general market effect, approximately 10% can be considered
due to an industry effect, and the balance is due to resid—
ual factors unique to the security in question. Although
the parallel here is a weak one, it appears that the level
of general business activity is the single-most important
group of factors influencing the outcome of an investment
project. And in all likelihood a significant portion of
the variability in the possible outcomes from an invest-
ment project can be attributed to swings in the level of
total business activity. This has further implications
when a series of these projects are combined into a port-
folio by a corporation: but further discussion of these
implications is postponed until the next section.

Up to this point the discussion has been limited

to considering all investment projects as if they arose

1Benjamin F. King, "Market and Industry Factors in
Stock Price Behavior," Journal of Business,‘Vol. XXXIX,
ISO. 1, Part II (January,1966--Supplement), pp. 139-190.

 

30

within a single economic market. But it is an oversimpli-
fication to speak in terms of a single market since the
world is characterized by a series of fragmented national
markets. Thus, there will not be a single index of general
business activity: instead there will be a separate one for
each national market. To the extent that each of the na—
tional markets is influenced by different business cycles,
natural catostrOphes and government policies, movements

and shifts within the general business indices for these
different markets will be independent of one another.

There will be a separate set of industry indices for each
of these national markets as well.

Within a particular national market, however, it
would appear that changes in the level of general business
activity will have the single-most important influence on
the array of possible outcomes for a capital investment
project. Consequently, if a corporation undertakes a
capital investment project that is located in an external
market the apprOpriate indices for general business
activity and the industry factors are those for that par-
ticular external market. As before, the general business
activity category would be expected to be, by far, the

more important of the two.

31
Combining a Group of Capital Investment Projects

The preceding section examined the principal cate-
gories of economic factors which influence the possible
outcome of an individual capital investment project. This
one addresses the question: What are the implications to
the corporation from combining a group of these projects
into a portfolio?

Since the outcome for a particular investment
project is not characterized by certainty, the stream of
earnings from a project will be a random variable which
has a range of possible values it can assume. For pur-
poses of the immediate discussion it is assumed that this
random variable will be approximately normally distributed.
Consequently, its central tendency can be described by the
mean of the possible outcomes and its dispersion by the
variance of the possible outcomes.

When a group of capital investment projects are
combined into a single portfolio, the resulting expected
value for that portfolio is merely the simple sum of the
individual parts. The dispersion of the portfolio, how-
ever, is not that straightforward for it involves not only
the components attributable to each of the individual in-
vestment projects, but also an additional component due to

the interrelationship between the projects. Thus. for a

32
group of investment projects, n in number, the total vari-
ability is described by:
n

n
= Z z rok Ujok
j=l k=1 3

the variance for the portfolio of n

investment projects

rjk = the correlation coefficient between
project j and project k: when j=k then
rjk=1, as this is the correlation of a
project with itself.

O'Uk = when j=k: the variance of the par-

ticular project in question

when j#k: the cross product of the

standard deviation of project j with

the standard deviation of project k

Since the correlation coefficient, r, can assume any value
between -1 and +1, then the variance for the combination
of n projects may be greater than, equal to, or less than
the sum of the individual variances from the n investment
projects. The increment to an existing group of n projects
from adding the n + 1 project will be a function of its own
variance 02 n+1 plus its covariances:

n

.2 rn+1j °j°n+l

j=l
with the original n projects.

Ideally, when a corporation must select from an

array of available investment Opportunities, assuming for

the moment that they are roughly equivalent in terms of

their rates of return and the dispersion in their possible

5...!

 

lpfi

unv‘

....

 

e
.0;

Nu

 

0‘-

"w
~y.

.UI

...:

ya

33
outcomes, the corporation would be most interested in
those which have negative correlations with its existing
investment projects. But the number of qualifying proj-
ects -— those with negative r's -- is likely to be small
and the corporation will instead concentrate on investment
projects which have only low positive correlations -- r>0
but significantly <1 -- with its existing ones. Neverthe-
less, by diversifying into investments which are not highly
correlated, the corporation can reduce the dispersion in
its earnings stream. By relaxing the previous assumption
that all the investment Opportunities have comparable
rates of return and variances, a model can be develOped
which specifies the trade Off between differences in rate
Of return and variance. But for purposes of this research
such a model is not required, and therefore it will not be
develOped in detail here. The present model has shown,
other things being equal, that a corporation will actively
seek capital investment Opportunities which have low
positive correlations to their existing investment proj-
ects. Even if an investment Opportunity has a lower rate
of return, or a higher variance, than one of the competing
projects, it may still be in the best interest of the
total corporation to select this alternative because of

its lower correlation to the firm's existing holdings.

34

International Diversification in the Array
Of Available Investment Opportunities

Whenever a capital investment project is added to
a corporation's existing holdings, the effects Of that
addition will enter both the stream of earnings parameter
of the valuation model plus the required rate of return
parameter of that model. In terms Of the variability in
the stream of earnings to the firm the previous section
showed that between two projects with equal variances, the
project with the lowest correlation to the firm's existing
holdings would make the smallest contribution to the over-
all variance Of the total stream. Because almost all in-
vestment projects within a particular market tend to be
rather highly correlated to that market's index of general
business activity, they in turn tend to be highly corre-
lated with each other. Thus, if a corporation limits its
capital investments to those concentrated in a single mar-
ket the available investment Opportunities will generally
be positively correlated with its existing holdings. And
these correlations will, in all likelihood, be substantial.
Furthermore, even if the corporation aggressively diversi-
fies its capital investments, it will be unable to diversify
away that portion of the variability in expected returns

which is attributable to changes in the level of general

35
business activity. And, as was pointed out in an earlier
section, it is this group of economic factors that will
likely have the greatest influence on the outcome of most
investment projects.

But if the corporation expands its array of capital
investment Opportunities to include ones that concentrate
on external markets, it may be able to diversify away a
portion of the variability which arises from swings and
shifts in the level of general business activity within a
particular market. To be sure, these new investment Oppor-
tunities are likewise correlated to a general business in-
dex, but it will be the index for the external market
where the project is located. While the indices of gen-
eral business activity in the various national markets are
positively correlated, the degree of correlation is only
moderately high.2 Because of this, the correlation among
investment projects arising in different national markets

should, as a general rule, also be moderate.

 

2During the period 1967 through 1971, the average
correlation (Spearman Rank Coefficient) between the indus-
trial production index of the United States and the re-
spective index from a sample Of ten other highly indus-
trialized countries was approximately rs = .47. Even
among the three largest members of the original EurOpean
Economic Community - France, Germany and Italy - the

average correlation between their respective industrial
production indices was only rs = .84, during the same time
span.

36

What does this suggest? It suggests that a corpo-
ration can significantly reduce the intradependence in its
holdings of capital investment projects by selecting in-
vestment projects located in different national markets.
The resulting portfolio of investment projects, then, will
have diversified away a portion of the variability that is
attributable to swings in the level of general business
activity of the country where the corporation is head-
quartered.

A reduction in the total variability of the firm's
stream of expected earnings will have a concomitant affect
on the perceived business risk, BR, of that corporation.
However, an exact specification as to how this risk
premium can be expected to respond is difficult to make.
On the one hand, the diversification into external markets
should reduce the variability of the corporation's Oper-
ating income, and thereby lower its business risk as per-
ceived by the investor at the margin. Yet, by expanding
into external markets the corporation adds several new
dimensions to its overall business risk. Beside the mar-
ket factors that were previously discussed, there will be
two additional categories of factors influencing the out-
come Of the investment projects concentrated in external

markets. Accordingly, the functional notation for busines

37
risk, BR, becomes:

where MK = market category
MN = monetary category
G = governmental regulation and control

As the earlier discussion of the domestic market category
noted, each of these general categories would comprise a
host of individual factors.

The brief summary that follows makes no attempt
to formulate an exhaustive list of the myriad of individual
items comprising each group. But it is informative to note
the general types of items that would be found in each
group. While the factors within the market category, MK,
pertain to a different national market, they would be com-
parable tO the types of things discussed in the earlier
section: Determinants of an Investment Project's Earning
Stream. The second category, that of monetary factors,
MN, would include such things as changes in the exchange
rate of the local currency where the investment project is
located, restrictions on convertibility of the local cur-
rency into foreign exchange, limitations on the transmital
of local currency out of the country, and restrictions on
the holding of currency or bank balances by a non-resident
‘COrporation. It is very difficult to make a clean dis-

tinction between the items of this and the final category.

a! I 1
mi.

' .:
‘n’ 1

04;"

it u v
,

-
'VM
.u~‘..

a p
r I .
I O

u ‘p ,‘

u. ._

a...

h

.
.-

n..‘

he
‘I

38
that Of governmental regulations and controls, G. But in
general terms this final category would include those rules
and regulations which are Specifically intended to control
or limit the Operations of the non-resident corporation.
Depending on the individual situation, these could range
from simple restrictions and limitations on certain phases
of the corporation's Operations, at one end of the spec-
trum, to the more involved controls which more or less
dictate how the corporation is to conduct its overall
Operations. Examples of the first would be restrictions
as to the types of industry a foreign firm can enter and
restrictions on the degree Of foreign ownership. Within
the latter group would be such things as specifying the
type and quantity of product the firm must produce and
requiring a certain level of capital investment.

The previous discussion suggests, that, while a
corporation may reduce the variability in its total earn-
ings stream by diversifying its selection of capital in-
vestment projects to include those from external markets,
this is not gained without injecting several new dimen-
sions to the corporation's perceived business risk, BR.
Thus, the eXpected future earnings continue to be a random
variable. But it is now subject to additional fluctuations

and shifts arising from factors within the two additional

39
categories: monetary and governmental regulation and

control.

Assumption of an Unchanged Capital Structure

An initial assumption for the first part of this
chapter was that the capital structure of the firm be held
constant: this can now be relaxed. One of the principal
purposes for doing this was to hold financial risk, FR,
unchanged until the analysis of the effects of diversifying
into external markets could be completed. Given that a
corporation has elected in some previous period to diver-
sify into external markets, there appears to be no reason
to expect this expansion would reduce the amounts avail—
able from current sources of capital. Nor would the avail-
able sources Of capital be necessarily curtailed. In fact
the corporation may, as a consequence of its expansion
into external markets, be in a position to obtain funds in
some Of the international capital markets: Euro-bonds,
Eurodollar loans, etc. Last, this expansion into external
markets would not require some perceptible shift in the
corporation's capital structure. In conclusion, then,
it appears that a corporation's expansion into external
markets would not, in and of itself, cause the perceived

:Einancial risk, FR. of the corporation to rise perceptibly.

40

Limitations on Changes in Dividend Policy

To simplify the earlier analysis, it was assumed
that dividend policy was held constant; the consequences
of relaxing this are now examined. Most corporations can
be characterized as having stable dividend policies. When
changes occur they are typically small relative to the
previous dividend level. Generally, when large or abrupt
changes do occur they are the consequence of some major
unfavorable event. Only infrequently is such a sharp re—
duction dictated by a sudden shift in the corporation's
demand for internally generated funds to meet its capital
investment Opportunities. Rather, by its previous actions
a corporation will have established a dividend policy that
investors at the margin can utilize to form reasonable
expectations as to future dividend payouts. Consequently,
there is little reason to expect the dividend policy of
multinational corporations will be subject to larger or
more frequent change than the dividend policy of their
domestic counterparts. While differences in the level of
dividend payout are very much the rule among corporations,
there is little reason to expect some systematic bias in
the payout level of multinational versus domestic corpo-

rations.

41
Effects of Diversification into External
Markets on the Wealth of the Common
Stock Shareholder

The preceding sections have attempted to analyze
how the effects of diversification by a corporation into
external markets would enter the parameters of the basic
investor valuation model. By utilizing the results from
this analysis, we can now attempt to draw some plausible
conclusions as to how the whole valuation model might be
expected to respond, when a corporation diversifies its
capital investment projects to include those that are
concentrated in external markets. More specifically, we
attempt to draw some plausible conclusions as to how the
market price Of a multinational corporation's common stock
can be expected to respond. In turn, since a change in
the common stock's market price is the principal deter-
minant Of the current shareholder's wealth, we can like-
wise make some statements about how shareholder wealth
will be affected.

For purposes of this study shareholder wealth is
defined to include not only the returns accruing to the
shareholder, whether that be in the form of a dividend
payment or a change in the market price of the share, but

also the risk that the shareholder was exposed to in Ob-

taining these returns. Adapting the standard convention,

42
investors are taken to be return maximizers -- prefer more
to less —- and risk averters -- prefer less risk exposure
to more. Risk exposure for a particular common stock was
defined as being the variability in the observed rates of
return on that common stock over some specified interval
of time. A discussion of the actual empirical measures
which were utilized as risk surrogates for this study
follows iJIChapter III. In the presentation of materials
for the balance of this chapter, both components Of share—
holder wealth —- rate Of return and degree of risk expo-
sure -- are considered.

Up to this point the effects of a corporation's
decision to diversify into external markets has been
analyzed on a piecemeal partial equilibrium basis, but
before any conclusions can be drawn about how shareholder
wealth will be affected all aspects of this diversifica-
tion must be considered jointly. That is, given that a
corporation has diversified on an international scale,
then what is needed is an ex 2252 prediction of the mar-
ket price of the shares. Once these are known, the rate
of return and risk exposure computations would easily
follow. But the underlying relationships for this type
of predictive model are complex and nearly impossible to

estimate, as the discussion of earlier sections has

43
pointed up. Even though all of the implications from the
earlier analysis cannot be integrated into a single predic-
tive model, they can be used to postulate what would be the
most likely outcome for shareholder wealth under a particular
set of circumstances. By utilizing just this sort of tech-
nique the research hypotheses of the next section were
develOped.
Research Hypotheses

Because the empirical portion of this research was
done as a comparative study, the research hypotheses which
follow are couched in terms Of the study's two groups:
multinational corporations that are headquartered in the
United States and domestic corporations that are likewise
based in the United States. Subsequent references tO
either group implies a corporation based in the United
States. In the research hypotheses which follow, the joint
aspects of shareholder wealth -- rate of return and risk
exposure -- are considered independently.

The general procedure for develOping the research

hypotheses was as follows:
1. After specifying the particular aspect --
rate Of return or risk exposure -- Of

shareholder wealth that was to be examined,

a given set of circumstances was hypothesized.

44
2. Drawing upon the implications that were
developed in the earlier sections of this
chapter, the most prObable Outcome for this
aspect of shareholder wealth was postulated.
3. Then, based on the implications from the
theory, the research hypothesis was develOped.
Rate Of Return Differential Between
Multinational and Domestic Corporations
Short Term Time Horizon. Given a significant
downturn in the level of general business activity within
a national market, all common stock investors at the mar-
gin will revise downward their expectations as to the
future stream of earnings, and concomitantly increase
their premium for business risk. As a result of these
combined actions, the market price of all common stocks
will decline. This is the typical pattern for stock
prices shortly before an economic recession begins and
during its early stages. The actual size of the overall
price declines will depend upon how severe investors ex—
pect the recession will be. Because some common stocks
will be affected by an economic recession more than
others, the magnitude of their respective price declines
will also vary. In particular, because domestic corpo-

rations have their capital investments concentrated in

45
the domestic market, their market price will not only
suffer a larger decline relative to those of comparable
multinational corporations, but also their recovery in the
later stages of the recession will be slower. First, be—
cause the stream of earnings from the domestic corpora-
tion's capital investments is heavily dependent upon the
general level of business activity in the domestic economy,
investors will sharply revise their expectations as to
future earnings flow. Second, because the array of
possible outcomes is characterized by greater variability,
the increment to the premium for business risk will be
larger for the common stocks of domestic corporations than
it would be on comparable multinational firms. Conse-
quently, the market price on the common stocks of domestic
corporations will be more extensively depressed during a
recession than those of comparable multinational corpora-
tions. And, as a consequence of the larger price decline
and its longer duration, the rates of return on the common
stock of domestic corporations will be less than those of
multinationals during a recession in the domestic economy.
That is, if rates of return are computed during a period
when the domestic economy is in a recession, then the
rates for the common stocks of domestic corporations will

be less than those Of comparable multinationals. The

46
research hypothesis for this is:
Research Hypothesis H1

Given two groups Of corporations, having a com-

parable business size and being located within the
same industry, and if the members within one group
have extensive foreign Operations such that they are
considered multinational and the members of the other
group have limited or nil foreign Operations such
that they are considered domestic, the rate of return
on the common stocks of the multinational group will
be significantly greater than the rate of return on
the common stocks of the domestic group, when the
comparison is made over a time interval that spans a
significant recession in the economic activity of the
country where both groups are headquartered.

A sequel to the above question is how rates Of
return would compare during a short term time interval
that spanned either a period of rapid growth in the
domestic economy or a period of limited growth or near
zero growth. Given either of these conditions, there is
little reason to expect that investors would necessarily
revise their expectations as to future earnings or busi-
ness risk for domestic corporations more frequently, or

by a larger amount, than they would for multinational

47

firms. Nor is there much reason to expect the converse
situation would be found. Given this, the changes in mar-
ket price for the common stocks of the two groups -- multi—
national and domestic -- would also be roughly comparable.
Consequently, the rates of return on the common stocks of
multinational corporations would not be significantly
greater than those for domestic corporations when these
rates correspond to a short term time interval that either
spans a period of rapid growth in the domestic economy or
a period of limited growth or near zero growth. An im-
plicit assumption of this conclusion is that both groups
have roughly comparable risk exposure. But implications
of the valuation model suggest that such an assumption
would be valid. Support for such a conclusion centers
around the idea that the volatility in market price for
the common stocks of the two groups will not be different
from each other. Therefore, both groups would have roughly
the same risk exposure. The research hypothesis for this
section is:

Research Hypothesis H2

Given two groups of corporations, having compar-
able business size and being located within the same
industry, and if the one group consists of multi-

national firms, and the other is comprised Of domestic

48
corporations and if the risk exposure on the common

stocks of the two groups are not significantly dif-

ferent from each other, then the rate of return on

the common stocks of the multinational group will

not be significantly different from the rate of

return on the domestic group when the comparison is

made over a short term time interval that Spans

either a period of rapid growth or

limited growth or near zero growth

a period of

in the domestic

economy of the country where the groups are head-

quartered.

Risk Exposure Differential Between

Multinational and Domestic Corporations

 

As was previously pointed out,

risk exposure for

a given common stock is considered to be the variability

in its returns over some specified time span. That is,

had an investor held this common stock
fied time span, then his risk exposure
ability in the rates of return for the
period time intervals that spanned the

period. To the extent that a security

during the speci-
would be the vari-
series of sub-
specified time

is characterized

by high variability in its array of sub-period rates of

return, any investor who must liquidate his holdings faces

the possibility of having to do this during a period when

49
the price, and therefore the sub-period return, is unusu-
ally depressed. Because an investor is subject to the
possibility of this type of financial loss, the risk
exposure for a security with high variability must itself
be considered high.

Long Term Time Horizon. To begin this analysis,
recall that in an earlier section of this chapter it was
shown that the market price of a corporation's common
stock is a function of its present plus future capital
investment projects. And it was further shown that the
outcome Of an investment project was likely to be highly
correlated with the level of general business activity for
the country where it is located. Thus, because their
investment projects are almost totally concentrated in a
single market, the market price of the common stocks of
domestic corporations will be subject to significant fluc-
tuations due to the cyclical swings in the level of general
economic activity for the country where it is headquar-
tered. The earlier sections also established that by
diversifying into external markets, a multinational cor—
poration would have less variability in its earnings
stream since all of the components in the overall stream
are no longer dependent on a single national market. As

such, the common stocks of multinational corporations will

50
have smaller fluctuations in their market price than will
the common stocks of domestic corporations. It immediately
follows that with smaller fluctuations in price there
should be less variability in the rates of return for a
series of sub—period time intervals. And therefore, the
risk exposure on the common stocks of multinational corpo—
rations is less. The research hypothesis for this is:

Research Hypothesis H3

 

Given two groups of corporations, having compar-
able business size and being located within the same
industry, and if the one group consists of multi-
nationals while the other comprises domestic corpo-
rations, then the risk exposure on the common stocks
of the multinational group will be significantly less
than the risk exposure on the common stocks of the
domestic group, when that comparison is made over a
long term time interval.

Short Term Time Horizon. If the total long term

 

time interval of the previous section is divided into a
set of sub-intervals where each approximates one phase of
the domestic country's economic cycle, one can analyze the
change in the Observed risk exposure differential between
multinational and domestic corporations. TO do this, we

first note that changes in investor expectations about

51
future earnings streams, and the required premium for busi-
ness risk inherent in these streams, will be greatest
during the rapid expansion phase and the contraction phase
of the typical economic cycle. And, as was established in
a preceding section, the resulting change in market price
for the common stocks of domestic corporations will be
larger and more frequent than these for multinational cor-
porations. If this is indeed the case, then the pattern
of larger and more frequent price changes will cause a
similar pattern in the series of sub-period rates of
return. This will in turn cause the risk exposure for the
common stocks of domestic corporations to be greater than
multinationals. In addition, if this differential in risk
exposure between the two is compared for the different
phases of the typical economic cycle it should decrease
during a period Of rapid growth and increase during a
period of significant contraction in the level of general
economic activity for the particular country where the two
groups of firms are headquartered. What this suggests,
then, is that over a series of these short term time
intervals the risk exposure differential will widen and
contract. Or, alternatively, there will be an interaction
between this differential and the typical pattern of

economic cycles. The resulting research hypothesis is:

52

Research Hypothesis H4

 

Given two groups of corporations, having compar—
able business size and being located within the same
industry, and if the one group consists Of multi-
nationals while the other is comprised of domestic
firms, then not only will the risk exposure on the
common stocks of multinational corporations be sig-
nificantly less than the risk exposure on the domestic
group, but the differential between the two groups
will increase and contract as the economy of the
country, where both groups are headquartered, goes
through the distinct phases of the typical economic

cycle.

W

In the Opening sections of this chapter it was
shown that the market price of a common stock was a func-
tion of investor expectations as to the future stream of
earnings and their attitudes toward the risk inherent in
the future stream of earnings. Next, it was shown that
the market price of the corporation's common stock could,
alternatively, be considered to be a function of the
possible outcomes on the corporation's present plus

future capital investment Opportunities. Then, using the

53
basic investor valuation model, it was established that
given certain conditions the corporation could reduce the
variability in its expected stream of earnings by diver—
sifying into external markets. Using these conclusions,
it was shown that the perceived business risk of the cor-
poration would also be likely to decline with such an
expansion. This was qualified, however, by the observa-
tion that the corporation's diversification into external
markets would add two new dimensions to its perceived
business risk.

The final section of the chapter began by defining
shareholder wealth to include not only the rate of return
achieved, but also the risk exposure inherent in achieving
that return. Following this, some implications were postu-
lated as to what would be the expected relationship
between the rate of return and risk exposure for the common
stocks of multinational corporations versus those of
domestic corporations. The chapter concluded with the
develOpment of these implications into a set of research

hypotheses.

CHAPTER III

METHODOLOGY AND RESEARCH DESIGN

Introduction

Having developed and specified the research hypoth-
eses in Chapter II, this chapter continues with a descrip-
tion of the empirical evidence that was collected and
analyzed to support or deny these research hypotheses. As
such, the chapter has four principal sections.

The first begins by describing the general nature
Of the research study and the methodology that was used in
gathering the empirical evidence. The next section begins
by describing the pOpulation of corporations which was the
source for the two corporate samples used in this study.
Following this, is an explanation of the criteria that
were used in selecting corporations for the samples. And,
finally, the details Of the actual samples are given. The
third section contains an explanation of the empirical
response measures that were employed along with the
rationale for their usage. The time periods covered by
the study and the sources of data are presented in the

54

55
closing parts of this section. Finally, the three statis-
tical models that were used in this study are presented in

the fourth section.

Nature of This Research Study

The empirical portion of this research study is a
comparative study between two groups of corporations:
multinationals and domestics. Since the empirical evi-
dence was drawn from gx‘pgsg data, it was necessary to use
existing naturally occurring groups. The next section
gives the criteria used to classify corporations within
the pOpulation into one of the groups. An equal number of
multinational and domestic corporations was used in the
two corporate samples. Partially, this was done to
simplify subsequent statistical analysis, but more impor-
tantly there was no real justification for having unequal
sized groups. The response for an individual corporation
was considered to be sufficiently independent so that the
unit of analysis within each group could be the individual
corporation.

Because it was a comparative study, the principal
analytical technique was to compare the responses of the
two groups on the different empirical measures to see if

there was a significant difference between them. From the

56
results of this comparison, then, certain inferences were
made as to the nature of the underlying relationships.
Description of the Corporate
Population and Samples
Population Defined
The initial universe of corporations included all

those listed in the 1970 volume of the Directory of Com-

 

panies Filing Annual Reports, with the Securities and
Exchange Commission.1 Using this list as the primary
source, the total universe of qualifying corporations was
reduced to arrive at an initial unstratified pOpulation of
corporations which met the following criteria:

1. The principal headquarters for the corporation
had to be located within the United States.

2. The corporation had to be publicly held and
its common stock widely traded. That is,
either the common stock had to be listed on
the New York or American Stock Exchange or
if not listed on one of these exchanges, then
its trading in the Over The Counter market had

to be widespread and adequately publicized so

 

1Securities and Exchange Commission, Directory of
Companies Filing Annual Reports with the Securities and
Exchange Commission (Washington, D.C.: U. S. Government
Printing Office, 1970).

 

57
as to ensure a viable and realistic market
price.
3. The shares of common stock had to be continu-
ously traded during the period of 1963 through
1971.
4. The annual sales in each of the above 9 years
had to be in excess of 20 million dollars.
One extremely useful aSpect of utilizing this SEC
listing was that all corporations had been classified
according to their most representative 3 digit SIC Enter?
prise code. Given this, the initial corporate pOpulation
was easily stratified according to industry code. Since,
as was previously pointed out, a balanced sample - equal
numbers of multinational and domestic corporations was
specified, many of these codes were eliminated by a pre-
liminary review of the types of corporations comprising
each 3 digit code. In almost all cases this elimination
was caused by the lack of adequate multinational repre-
sentation within the particular industry code. Examples
of the general types Of codes excluded at this point were
service industries, financial and insurance related acti-
vities and retailing activities. For those industry codes
that remained, a more detailed analysis was made to

identify the multinational corporations within an industry.

58
Results from previous research and the preliminary findings
of this study suggested that this would be the more con—
straining of the two groups.

Those corporations which were classified as multi—

national had to meet the following criteria:

1. At least 20% of the corporation's annual sales
for 7 of the 9 years -- 1963 through 1971 -—
had to have been in national markets outside
the United States and Canada. The foreign
sales for the remaining 2 years had to be in
excess of 15% of the yearly total.

2. These foreign sales had to have been made in
a minimum of 5 external markets.

3. The corporation had to have at least 3 manu—
facturing locations outside the continental
United States and Canada.

Whenever possible this information was obtained from the
corporation's annual report to its shareholders. The
alternative data sources are detailed in a later section
of this chapter.

For this study Canada was not considered an ex—

ternal market either for purposes of foreign content in
sales or for manufacturing locations. Because of its

close proximity to the United States, the economic cycles

59
for the two countries show a much higher degree of associ-
ation than is the case between the United States and other
industrialized countries. For this reason it was felt
that diversification into the Canadian market Offered only
limited possibilities for reducing the variability in the
corporation's earnings stream.

To improve the precision of the study, corporations
that had sizable foreign sales but were still short of the
20% hurdle level, were eliminated from consideration for
either group in the final samples. Likewise, any corpo-
ration which had undergone a significant transition,
either toward becoming multinational in its sc0pe of Oper-
ations or away from being a multinational firm, was also
eliminated: again to improve the precision of the study.

Last, those industry groups which are primarily
engaged in the extraction of natural resources -- basic
and precious metals, petroleum, coal, etc. —- were not
considered for the final corporate pOpulation. In the
first place, many of these groups would have been elimin—
ated in the preliminary analysis because they lacked an
adequate number of multinational or domestic corporations.
But even for those groups which contained an adequate
sample, their exclusion was justified on two points. First,

because the corporations within these industry groups tend

60
to have their foreign Operations concentrated in less
develOped nations, where these resources are most preva—
lent, the nature of their Operations in terms of risk,
variability, price stability, and other characteristics
are totally different from the manufacturing type of
multinational corporation: the largest element in the
initial population. To the extent that the production
from these foreign locations is but a sourcing activity
for uses in the domestic market, the corporation will not
have truly diversified into external markets because its
level of Operation would still be heavily dependent on the
general business activity in the country where it is head-

quartered.

Sample Selection Criteria

Selection of Industry_Groups. After removal of
the deficient industry groups in the preliminary review,
the groups that remained were concentrated in the manu—
facturing types of business activity. From these, the
industry groups for the two corporate samples were chosen
on a non-random basis. By using a explicit selection
technique it was possible to concentrate on those industry
groups with the largest number of multinational and domes-

tic corporations. Furthermore, it ensured a broad coverage

61
of the qualifying industry codes. Consequently, the
minimum number of corporations within an industry category
was set at twelve: a minimum of six multinational and six
domestic corporations. At this point, the rigid adherence
to a single 3 digit code within an industry category was
dropped. Whenever possible a category was limited to one
3 digit code, but when a 3 digit industry code contained
less than the prescribed minimum, it was evaluated to see
if the code could be consolidated with some other incom—
plete 3 digit code. The principal criterion being the
need to maintain homogeneity within the industry category.
For several corporations there was some question as to the
correctness of their industry classification, these were

cross—referenced to two additional sources: Standard &

 

Poor's Register of Corporations, Directors, and Executives,

 

2 3

1971 and the Million Dollar Directory 1971.

 

The industry groups that were selected for the

initial corporate sample are given in Table l.

 

2Standard and Poor's Register of Corporations,
Directors and Executives, 1971 (New York: Standard &
Poor's Corporation, 1971).

3Dun & Bradstreet, Million Dollar Directory 1971
(New York: Dun & Bradstreet Inc., 1970).

 

C.
.

i.

\.

o"

-i

.--

fl.‘

62

TABLE 1

LIST OF QUALIFYING INDUSTRY GROUPS

 

 

Number SIC Enterprise
Industry Description of Codes Contained
Corporations in the Group

Food and Beverage Products 30 20.2, 20.3, 20.4
Basic Chemical Products 27 28.1, 28.9
Drug and Pharmaceutical Products 18 28.3
Soaps, Cosmetics, Cleaners and

Paint Products 27 28.4, 28.5
Machine Tools and Related Products 12 35.4
General and Specialized Industrial

Equipment 12 35.5, 35.6
Typewriters, Electronic Computing

Equipment and Office Equipment 16 35.7
Electrical Machinery, Equipment

and Supplies 23 36.6
Professional, Scientific and Con—

trolling Instruments; Photographic

and Optical Goods 20 38.1, 38.3, 38.4

Total 185

 

Selection of Corporate Subjects - Initial Sample
Multinational Corporate Group. Seven of the
industries listed in Table 1 contained more domestic than
multinational corporations, while the remaining two indus-
tries had an equal number of each. For this reason, all
qualifying multinational corporations from each of the

selected industries were chosen for the sample.

63

Domestic Corporate Group. For those seven indus-
try groups where the number Of domestic firms exceeded the
required sample, a random selection technique was used to
select the domestic contingent. There was one exception
to this procedure, however, that was the initial placement
of the Radio Corporation of America in the domestic sample
for the Electrical Machinery, Equipment, and Supplies indus-
try group. Because of this company's size and dominance
within the industry, it was considered essential that it
be included in the sample. One initial reservation to the
use Of a random selection procedure was that it might be
necessary to select the domestic corporations based on
their sales volume in order to get approximately equal-
sized firms within both groups. But a preliminary compari-
son Of the potential domestic candidates to their multi-
national counterparts indicated that such a reservation
was not supported by the data. As finally selected, the
domestic sample did not have a significant upward or
downward bias from the multinational group in terms of
corporate size.

From the initial corporate pOpulation shown in
Table l, 146 corporations were selected for this sample.
Table 2 gives their distribution by industry group and by

the nature of their corporate Operations.

64

TABLE 2

INITIAL SAMPLE OF CORPORATIONS

 

 

 

Frequency

Industry Description Multinational Domestic
Food and Beverage Products 9 9
Basic Chemical Products 8 8
Drug and Pharmaceutical Products 9 9
Soaps, Cosmetics, Cleaners, and

Paint Products 9 9
Machine Tools and Related Products 6 6
General and Specialized Industrial

Equipment 6 6
Typewriters, Electronic Computing

Equipment, and Office Equipment 8 8
Electrical Machinery, Equipment,

and Supplies 9 9
Professional, Scientific and Controlling

Instruments; Photographic and Optical

Goods .2. .12

Total 73 73

 

Selection of Corporate Subjects - Final Sample

 

Before this sample was specified, the preliminary

analysis was completed using the corporate sample shown

in Table 2. The results of this preliminary review sug-

gested several revisions that should be made to the

"initial sample." Furthermore, the three statistical

models, which had been hypothesized for the empirical

65
portions of this study, entailed significantly different
assumptions about the nature of the empirical data. This
also suggested several further revisions to the original
"initial sample."
TO gain better intragroup homogeneity among the
corporations within the industry groups the following revi-
sions were made:
1. The two corporations from SIC code 28.9 that
had been included in the Basic Chemical
Products group, were drOpped from the sample.
As a consequence, the revised group contained
only a single 3 digit SIC code: 28.1.

2. The four corporations from SIC code 28.5,
which had been consolidated with the Soaps,
Cosmetics, Cleaners, and Paint Products group,
were excluded for this sample. After doing
this, only SIC code 28.4 was included in this
group. And to reflect this refinement, the
title was altered slightly to Soaps, Cosmetics,
and Clearners.

3. The final industry group in Table 2 contained

three SIC groups: 38.1, 38.3, 38.4. SO it
was decided that these should be divided into

two separate groups. The first revised group

66
contained SIC codes: 38.1, 38.4: while the

second contained only the corporate members

of code 38.3.

Although these revisions were not major, each one signifi-
cantly improved the homogeneity of the affected industry
groups. Also, it was anticipated that these changes in
the corporate sample would improve the precision Of the
final results.

The final corporate sample contained 140 corpora-
tions after the indicated revisions. Table 3 shows their

distribution by industry group and by sc0pe of operations.

67

TABLE 3

FINAL SAMPLE OF CORPORATIONS

 

Industry Description

 

 

Food and Beverage Products

Basic Chemical Products

Drug and Pharmaceutical Products
Soaps, Cosmetics, and Cleaners
Machine Tools and Related Products

General and Specialized Industrial
Equipment

Typewriters, Electronic Computing
Equipment, and Office Equipment

Electrical Machinery, Equipment,
and Supplies

Professional, Scientific and Controlling
Instruments

Photographic and Optical Goods

Total

Freguency
Multinational Domestic
9 9
7 7
9 9
7 7
6 6
6 6
8 8
9 9
5 5
__4_ .3
70 70

 

Specification of the Empirical Response Measures

Introduction

 

Before presenting the results from the empirical

tests of the research hypotheses of Chapter II, this sec-

tion describes the response measures which were utilized

to establish the existence or nonexistence of a predicted

relationship. Each of the sections which follow'begins

68
with a description of the particular empirical response
measure, its computational form, and the justification for
its usage in this study. The measures themselves are
grouped according to the two elements of shareholder
wealth: rate of return achieved and the risk exposure in
attaining those returns. There is a single measure for
the rate of return category and five different surrogate
measures for risk exposure.

The reason for using more than one surrogate for
risk exposure is that there are several different ways of
conceptualizing risk exposure. Because each of these
different definitions concentrates on a distinct aspect
of risk exposure, a separate empirical measure was needed
to capture and highlight each one. In addition to the
measures actually reported here, four alternative measures
were also evaluated during the preliminary stages of this
study. These are discussed in Appendix A together with a
brief summary as to why they were not selected for use in
this study.

It should also be noted that the computational
equations as given are for estimating pOpulation para-
meters when the sample size is large. Since sample sizes
were small for this study, the necessary correction factor

was incorporated in the actual computations to give

69

unbiased estimates of the pOpulation parameters.

Rate of Return Measure
Price Relatives. The market performance for the
common stocks of the corporations in the samples was mea-
sured by their quarterly price relatives. These price
relatives included not only dividend payouts for the
quarter but also any change in market price. The justifi—
cation for using a quarterly holding period is given in a
subsequent section. The computational form for a single
share of common stock was:
PRt = Pt + Dt
Pt-l

where
PRt = the quarterly price relative for period t

Pt = the price of the share at the end of period t

Dt = any dividends received on the share during
period t

the price of the share at the end of period
t-l, which is the same as the beginning of
period t

'6
fl
l
H
II

t = the quarterly time interval

An overall rate for the entire period was computed
by Obtaining the geometric mean of a security's quarterly

price relatives. This was done as follows:

_. n /n

PR III PRt

70

where

PR = the average price relative and where n is the
number of quarterly time periods encompassed by this aver-
age. Last, PRt and t are as defined previously. The
geometric average was used because it is not subject to an
upward bias when there are wide fluctuations in the price
relatives.4 The quarterly price relatives or the average
for the period can easily be converted to a fractional
rate of return by simply deducting —l.OO from each one.
Where the closing price plus applicable dividends is less
than the Opening price - PR<1.00 - the rate of return

would be negative.

Surrogate Measures for Risk Exposure

Having defined the return measure to be the quar-
terly price relatives for each of the common stocks, a
narrower but more explicit definition for risk exposure

is now possible. Accordingly, risk exposure is now

 

4The following sources provide a more detailed
discussion of the geometric average, Haim Ben Shahar and
Marshall Sarnat, "Reinvestment and the Rate of Return on
Common Stocks," Journal of Finance, Vol. XXI, No. 4
(December,l966), pp. 737-742: John L. Evans and Stephen
H. Archer, "Diversification and the Reduction of Disper-
sion: An Empirical Analysis," Journal of Finance,
Vol. XXIII, NO. 5 (December,l968), pp. 761-767: RObert.A.
Levy, "Measurement of Investment Performance," Journal of
Finance andgguantitative Analysis, Vol. 3, NO. 1 (March,

1968), pp. 35-57.

 

71
defined as the variability in a security's quarterly price
relatives over some specified time interval. But the term
variability can be conceptualized in several different
ways and each of these concentrates on a different aspect
of risk exposure. Consequently, in the sections that
follow,different empirical surrogate measures are used to
capture and highlight these different aspects of risk
exposure.

Overall Variability in Returns. The surrogate
measure for this section was intended to capture the total
variability within the series of quarterly price relatives
over a specified time interval. High variability, regard-
less of whether the array of pg Egg; price relatives con-
tains primarily positive returns, PR's >l.O, negative
returns, PR's<1.0, or some combination of the two, repre-
sents high risk exposure for this concept of risk.

The actual statistic used to describe the total
variability of a series of price relatives was Gini's5
mean difference. It represents the average absolute
difference between each and every pair Of observed price

relatives within the series. Because the pOpulation Of

 

5Corrado Gini, Memorie di metodologia statistica,
2nd. ed. Ernesto Pizzetti and T. Salremini (Rome:
Libreria Emedi Virgdio Veshi, 1955).

72
price relatives for each common stock was not large, a
complete iteration was done for all pairs. The computa-

tional equation was:

n n
= Z Z .. -
GMD [1;1 t'=l ABS (PRt PRt'El / n(n 1)

where
GMD = Gini's mean difference

ABS = the absolute value of the quantity within
the parenthesis

t,t' = the series of quarterly time intervals,
which Span the total interval n.

and PRt and n are the same as defined previously. Because
the study involved comparing this statistic for the dif-
ferent members from the corporate groups, a relative mea-

sure was chosen. This was computed according to:

GINCC GMD / (2 x PR)

where

GINCC = Gini's coefficient of concentration

GMD = Gini's mean difference
PR = the geometric mean Of the price relatives

AS such, this measure Satisfied the two principal
requisites for this concept of risk eXposure: first, that
it capture the total variability for the series of price
relatives and second, that it was a relative instead of
an absolute measure. In addition the measure is nonpara-
metric. Consequently it requires no assumptions about the

distribution of the underlying pOpulation: nor does it

73

require estimates of population parameters in its compu-
tations. Finally, the measure gives greater weight to
extreme values without being unduly influenced by any
Single extreme Observation.6

Downside Risk Exposure. This aspect of risk
attempts to highlight the investor's exposure to the pos—
sibility of a financial loss during one or more of the
quarterly Sub-periods. PrOponentS of this concept of risk
claim that variability, in and Of itself, is not true risk
to the investor unless some of the quarterly price rela-
tives have been less than an investor specified minimum.
That is, if all the values in the array of price relatives
for a particular common stock were arranged in ascending
order, then only those values which fell below the in-
vestor's required minimum should really be considered as
part of the risk eXposure inherent in that security. But,
because every investor would have a different required
return, it is difficult to Specify a single rate. The
solution to this problem was to set the minimum at a zero
rate of return -- a price relative of 1.00, since all in-

vestors would consider a negative rate of return as being

unfavorable.

h

. 6A further discussion of this measure can be found
111.Maurice G. Kendall and Alan Stuart, Advanced Theory of
éflEgtistics in Three Volumes, 2nd ed. (New York: Hofner
Publishing CO., 1963), Vol. 1.

74

The quadratic mean, modified as noted above, was

n 2
til (Loge PRE) /n

the modified quadratic mean

computed as:

 

QUADMN

where

QUADMN
2
*
(Loge PRt)

the square of the natural logarithm of
the adjusted price relative for a given
period t, where its value is defined as:
. * =
(i) Loge PRt LogePRt

(ii) Loge PR: = 0 for all LogePR

when LogePRt<0
t'SZLO
and where n and t are the Same as previously defined. To
compute this measure, the previous quarterly price rela-
tives were transformed into natural logarithm values.
When the PR was 1.00 I .15, the resulting log value
approximates a quarterly rate of return for that particular
common stock, assuming continuous compounding. Because of
this transformation, the array of 35 pggp price relatives
better approximates a normal distribution: also, the re-
sulting measure will be a relative one.

By concentrating on those price relatives which
suggest a financial loss for a Sub-period -- natural log
value of less than zero -- this measure captures the in-

vestor's eXposure to the possibility of a financial loss.7

 

7For a further discussion of this measure see,
Robert A. Levy, Op. cit.

75
And, as a consequence Of the individual price relatives
being squared, the measure gives greater weight to the
extreme values.

DichotomizipijiSk Exposure into Its Market and
Residual Components. A third alternative is to consider
the risk exposure on a particular common stock as being
composed of two distinct elements: a market component and
a residual component. The first stems from the host of
economic and general factors that affect all common stocks
in the market. The residual element stems from a second
group of factors that tend to be unique to each particular
company; the effects of these factors, however, are re-
flected in the market action of only that company's common

stock. The general functional form would be:

R*j = f(Rm, Rj)
Where
R*j = the overall risk exposure for security j
Rm = the general market component of risk that
affects all securities in the market
Rj = the residual risk component that is unique

to security j

The risk exposure of a common stock continues to
be described in terms of the variability of its returns:
or for purposes of this study, the variability in the com-

mon stock's quarterly price relatives. While it was not

76
possible to establish a direct cause and effect relation-
ship between the previously noted general market factors
and the variability of a common stock's price relatives,
one possible estimation technique is to partition the
variability of a common stock's price relatives into two
elements. By comparing the variability of its price rela-
tives to the comparable price relatives for some general
market index of all common stocks, it is possible to esti-
mate to what extent the variability of the two series
parallel each other. Then by inference, that portion of
the common stock's variability that parallels the general
market index can be attributed to the same factors that
caused the changes in the general market index.

To Operationalize this concept of risk exposure
several assumptions and Simplifications were made. First,
it was assumed that the relationship between the vari-
ability Of an individual security's price relatives and
those Of the general market index was approximately
linear.8 Second, the Standard and Poor's 425 Industrial
Stock Index was used as the surrogate for the general

market index. This moderately specialized index was

 

8This linear model was first suggested by William
F. Sharpe, "A Simplified Model for Portfolio Analysis,"
Management Science, Vol. 9, NO. 2 (January, 1963), pp. 277-
293.

77
chosen because the corporations in both samples were pre-
dominately industrial firms. Second, this index contains
firms drawn from many different industries, and also
within each industry it contains corporations that range
from smaller, regional firms to the large industry giants.
Actual computation of the quarterly price relatives, like
those for the individual corporations, included not only
market price but dividend payments as well.

To estimate the relationship between these two
variables, a linear regression model was used. The
dependent variable for that model was the quarterly price
relative of the individual common stocks, while the com-
parable price relative of the Standard and Poor's 425
Index served as the independent variable. The functional
form was:

Loge PRjt = a. + BjLogePRm + e

3 t jt
where
a. = the intercept Of the estimating equation
3 for security j
Bj = the market volatility of security j relative

to that Of the general market index

LogePRmt.LogePRjt = the natural logarithm of the price
relative for the general market index
and for security j, respectively:
this for sub-period t

and where t is defined as before. The actual computational

technique used to estimate Gj, B- and the e. was least

3 jt

78

squares.

As Specified in this model, the coefficient, Bj,
served as a measure of the volatility of security j's
price relatives compared to those for the general market
index. The individual B coefficient for each of the com-
mon stocks in the two corporate samples served as the sur-
rogate for market risk, Rn' Since the estimation technique
and the periods covered were identical for all common
stocks, the individual B's could easily be compared be~
tween groups. That portion of security j's variability
which was not associated with a Similar pattern in the

general market index was contained in the error terms,

ejt's. In particular, the standard error of the estimate
n
s.=E

served as the surrogate for residual risk. And because
this was a relative measure, a direct comparison between
different common stocks was valid. One further output of
this model was the estimated aj for each common stock.
The general interpretation attached to this parameter was
that it represented a measure of the superior performance
of a common stock relative to the general market index.
As estimated for this study, however, many of the a-‘s

3

were small and somewhat volatile from one sub-period to

79

the next. Consequently, it was decided that they were not
of further use in this study.9

Last, it Should be noted that this concept of risk
is currently receiving much attention in the literature of
finance. Not only has there been extensive research at
the academic level, but also many practitioners within the
securities industry have utilized use dichotomous treat-
ment of risk exposure.10

Exposure to Major Financial Loss. This interpre-
tation of risk exposure attempts to highlight the possi-
bility of an investor suffering a major financial loss

when there is a large decline in the price relatives of a

corporation's common stock. All investors would be

 

9Probably the Single-most important reason for
dropping these estimates from further analysis was that
many were near zero. In particular, in testing the co-
efficients of this linear regression model the null
hypothesis--Ho: a- - O--could not, in many cases, be
rejected: at least not at an acceptable confidence level.

10Some examples Of the work in this area are,
Michael C. Jensen, "The Performance of Mutual Funds in the
Period 1945—1964," Journal of Finance, Vol. XXIII, NO. 2
(May, 1968), pp. 389-416; Jack L. Treynor, "How to Rate
Management of Investment Funds," Harvard Business Review,
January-February, 1965, pp. 63-75: J. L. Treynor, W. W.
Priest, Jr., L. Fisher, C. A. Higgins, "Using Portfolio
Composition to Estimate Risk," Financial Analysts Journal,
Vol. 24, NO. 5 (September-October, 1968), pp. 93-100:
Chris Welles, "The Beta Revolution, Learning to Live with
Risk," Institutional Investor, Vol. V, NO. 9 (September,
1971), pp. 21-27.

 

 

80

reluctant to undertake any investment which offered the
possibility-~however small—-of a large loss, yet offered
only a limited gain. According to this interpretation, a
security has high risk exposure if its array Of Ex Egg;
price relatives contains large losses--PR'S Significantly
less than l.O-—and only limited gains--PR'S only slightly
greater than 1.0.

The relative skewness of the gxflpggp quarterly
price relatives was used to capture this aspect of risk

exposure. For a share of common stock this was:

n .— 3
[g (LogePRt-LogePRl]
t=l

SK=L n J/SD

3

 

where
SK = the relative skewness of the natural log
price relatives
LogéPR = the natural logarithm of the geometric mean

of the quarterly price relatives

SD = the standard deviation of the natural log
price relatives

and where LogePRt, t and n are the same as previously
defined.

AS a consequence of the cube factor, the above
formulation gives high weight to extreme values yet still
retains the Sign for each one. AS it is given, if the

statistic has a large negative value —— a distribution

81
that is skewed to the left -- this would be indicative of
a security with high risk exposure. AS used in this study,
this measure was intended as a supplement to the previous
empirical measures. In their study Fisher and Hall11 used
skewness as their second risk surrOgate. Arditti12 has

also suggested that skewness can be utilized to capture an

additional aspect of risk eXposure.

Time Period Covered

For all of the common stocks in the two Sample
groups, a price relative was computed for each three month
quarter during the time period. The market price used was
as of the last business day for the months of March, June,
September, and December. One justification for using a
quarterly holding period was that it eliminated the need
to allocate dividend payments between different sub-
periods since most corporations also make their dividend
payments quarterly. But a more important justification

for the use of a three month quarter was that it was

 

111. N. Fisher and G. R. Hall, "Risk and Corporate
Rates of Return," The Quarterly Journal of Economics, Vol.
LXXXIII, No. 1 (February, 1969), pp. 79-92.

 

12Fred P. Arditti, "Risk and the Required Rate of
Return on Equity," Journal of Finance, Vol. XXII, No. l

82
deemed to be frequent enough to capture major trends in
the market price of each common stock. Yet the interval
was not so short that the quarterly price relatives would
be unduly influenced by minor recurring price fluctuation,
nor would they be buffeted by noise within the system.
Last, the quarterly holding period would appear to roughly
approximate the points at which a long term investor would
evaluate his holdings.

The time period for this study began with the
first quarter of 1963 and extended through the fourth
quarter of 1971: a total of 36 quarterly reference
points. An additional feature of this nine year Span was
that it included an example of the major phases of the
United States economic cycle. This was important because
it permits generalizations to be made from the results of
this study to a much broader time horizon covering all
periods. Also, to eliminate the possibility of biasing
the results for the multinational group, whether favorably
or unfavorably, it was desirable that the total period
include different phases of the economic cycles for the
various external markets as well. The more recent periods
were emphasized both to minimize prOblems of maintaining
consistency within the two corporate groups and to avoid

the prOblems of currency and investment restrictions that

83

characterized the late 1950's.

Another desirable feature of the nine year period
-- 1963 through 1971 -- was that, by dividing it into
three equal sub-periods, each of these sub—periods reason—
ably approximated a single distinct phase of the United
States economic cycle. Thus, the first Sub—period, 1963
through 1965, represented a period Of rapid growth in the
general United States economy. During most of the quar—
terly periods within this three year Span, there was
rapid eXpansion in almost all sectors of the economy. The
second sub-period, 1966 through 1968, represented a period
of limited growth to near zero growth in the United States
economy. Within this period, the early segment had fairly
steady growth followed by a mild downturn during the
middle portion and then strong grOwth in the closing quar-
ters of the period. While the internal consistency for
this sub-period is not as good as the first, it was still
felt to be adequate to allow a general classification of
limited growth to near zero growth. Finally, the third
period, covering 1969 through 1971, was classified as a
period of substantial downturn in the United States
economy. Although the early quarters were strongly expan-
sive, the middle and latter quarters were periods of con-

traction within most sectors of the economy. Again,

84
although there are several distinct patterns of economic
activity within the period, the overall conclusion was
that the period represented a significant downturn in the
United States economy.

To summarize then, the three sub-periods for this

study were as follows:

1. The quarterly periods Of 1963 through 1965:
classified as being a period of rapid growth
in the United States economy.

2. The quarterly periods of 1966 through 1968:
classified as being a period of limited to
near zero growth in the United States economy.

3. The quarterly periods of 1969 through 1971:
classified as being a period of Substantial

downturn in the United States economy.

Data Sources
The primary source for determining the foreign
content of a corporation's sales, and therefore whether
it was classed with the multinational or domestic group,
was the firm's annual report to its Shareholders. The
same source was also utilized to determine the extent of
external market diversification in these foreign sales and

the number Of manufacturing locations outside the United

85

States and Canada. Whenever possible, these criteria
were evaluated for three reference years: 1965, 1968 and
1970. When the required data was not available in this
source, several secondary sources were used: the annual
report —- Form 10K -- that the corporation filed with the
Securities and Exchange Commission for the apprOpriate
years and the various publications from the two investor
advisory services -- Moody's and Standard & Poor's. In
cases where adequate information could not be found in
any Of these sources the corporation was excluded from the
sample.

For those corporations whose common stock was
listed on either the American Stock Exchange (ASE) or the
New YOrk Stock Exchange (NYSE), the quarterly market

prices were gathered from the various volumes of ISL Daily

13

 

Stock Price Index. The closing price on the selected
day was the one that was used. The data as to cash divi-
dends, stock dividends and stock Splits were obtained from
the same source as the prices. Partially, as a test of

this source, and whenever there was data missing the

following secondary sources were used: Wall Street

 

 

13Standard & Poor's Corporation, ISL Daily Stock
Price Index (New York: Standard & Poor's Corporation,
various dates of publication).

 

86
Journal14 for market prices and the various annual issues
of Moody's Dividend Record15 for the dividends and stock
splits.
For those corporations whose stock was traded
over the counter, the closing bid price was Obtained from

the various issues of the Bank and Quotation Record.l6

 

Information as to cash dividends, stock dividends and

stock splits was Obtained from the various annual issues

of Moody's Dividend Record.17
The market price data and the dividend series for

the Standard & Poor's 425 Industrial Stock Index were Ob-

tained from issues of The Analysts Handbook - Monthly
18

 

Sppplement for the most recent years. The others were
obtained directly from the New York office of Standard &

Poor's corporation.

 

14Dow Jones & Company, Inc., The Wall Street
Journal, various issues during the years 1963-1971.

 

15Moody's Investors Service, Inc., Moody's
Dividend Record (New York: Moody's Investors Service,
Inc., various dates of publication).

 

16William B. Dana Company, Bank and Quotation
ReCOld (New YOrk: William B. Dana Company, Publishers,
various dates of publication).

 

17Moody's Op. cit.

18Standard & Poor's Corporation, The Analysts
Handbook — Monthly Sppplement (New YOrk: Standard &
Poor's Corporation, various dates of publication).

 

87
Specification Of the Statistical Models

This section contains a brief explanation of the
three statistical models that were used for the empirical
portions of this study. Following the eXplanation of each
model, there is a short summary of the rationale which
lead to its selection. The first of these models used the
empirical evidence gained from the initial corporate
sample, Table 2. The second and third of these models
utilized the evidence obtained from the final corporate
Sample, Table 3.

Univariate Statistical Model - Initial Sample.
Since the research design for this part of the study was
a comparative analysis between the multinational and
domestic corporate groups, a model was needed that would
test whether the two independent groups had indeed been
drawn from the same pOpulation. The actual test for dif-
ferences was made using the six empirical response mea-
sures presented in an earlier section of this chapter.
A distribution free, or nonparametric, statistical model
was used in this portion of the study. The reasons for
using this class of model over the more common parametric
type were several. Probably, the Single-most important
one was that it eliminated the need for any assumptions

as to whether the pOpulation distribution underlying the

88

different empirical measures approximated a normal one, or
whether the variances for each of the two groups were
necessarily equal. The actual statistical test employed
was a ManneWhitney U test, hereafter called the Mew U test.
The sacrifice in power and efficiency is small19 when this
nonparametric test is used rather than a parametric one
like the t test. Consequently, by sacrificing a small
amount Of power and efficiency, the assumptions required
for this model were much weaker than those required for
its parametric counterpart. Because there were less
qualifications as to initial conditions, the conclusions
from the results were more general.

For this study a series of independent univariate
MAW U tests were performed; one for each of the six empirical
measures taken over the four different time intervals.
One very real limitation to using this type of analysis
is that the overall probability of making a Type I error --

rejecting the null hypothesis when in fact it is true --

becomes a function of the prObability levels associated

 

19Even in moderate-sized samples the power-
efficiency coefficient between the Mann-Whitney U and the
t test, assuming a valid population for use of the t test,
approaches 95%. For a more complete discussion of this
see, A. M. Mood, "On the Asymptotic Efficiency of Certain
Non-Parametric Two-Sample Tests," Annals of Mathematical
Statistics, Vol. 25 (1954), pp. 514-522.

 

89
with each of the independent tests. Furthermore, the
results of these individual tests cannot be linked to-
gether to draw an overall conclusion as to the total
results. These shortcomings were, however, recognized
and care was taken in interpreting the results from this
analysis so as not to violate these restrictions.

The purpose and principal points of concentration
for the model were twofold: first, to examine the rela-
tionship between the combined multinational and domestic
samples in regard to the six specified response measures,
and second, to examine the same response measures when the
corporate samples were the nine individual industry cate-
gories.

Multivariate Design Model - Final Sample. A two-
way factorial design was used in this part of the study.
The first factor was the previously discussed classifi-
cation by 3 digit industry codes. There were ten levels
-- ten distinct industry codes -- for this factor. The
second factor for the design was the classification of the
corporations as either being multinational or domestic.
Since each level of the first factor was crossed with
each and every level of the second factor, it was a com-
pletely crossed design.

The actual model that was hypothesized for this

90

part of the study was a two-way, main class multivariate
design model. The design model was suggested because the
levels within the two factors were actually discrete as
Opposed to continuous intervals. The two—way and factorial
portions Of the model were a direct result of the research
design that was used. Since there was no.3 priori reason
to suspect interaction between the factors, a main class
model was hypothesized. There was, however, replication
within each of the cells of the model which permitted a
test of whether the fit of the main class model was ade—
quate. Finally, the use of a multivariate model was
called for because Six empirical response measures were
used as dependent variables. Because each of these mea-
sures was correlated to some extent, it was desirable to
test them jointly rather than piecemeal.

A multivariate two-way analysis of variance tech-
nique was used to test the equality of the mean reSponses

20 That is, the mean

between the different sample groups.
response on the Six previously specified empirical mea-

sures for the multinational group within each of the

 

20A complete discussion of this technique can be
found in Donald F. Morrison, Multivariate Statistical
Methods (New York: McGraw-Hill Book Company, 1967) espe-
cially Chapter 5 or T. W. Anderson, An Introduction to
Multivariate Statistical Analysis (New York: John Wiley
& Sons, Inc., 1958).

 

91
industry categories was compared to those for the compar-
able domestic groups. In addition, the number of subjects
between cells was purposely kept prOportional: this to
simplify the testing for main class effects.

In contrast to the univariate model of the last
section, the model in this and the next section concen-
trateson comparing the mean response for the combined
multinational and domestic samples. But, whereas the
last model treated each of the risk response measures as
being independent, the multivariate models of this section
considered all of the risk measures jointly. And by
analyzing the results from these models one could attempt
to identify which of the risk measures accounted for the
major differences between the two corporate groups.

Multivariate Repeated Measures Design Model -—

Final Sample. Recall that the final research hypothesis

 

of Chapter II not only addressed the question of whether
the two groups Of corporations had different scores on the
various empirical response measures, but also, whether

the differential between the two would change with time.
For purposes of this section, the scores that were com-
puted on each Of the Six response measures for the three
equal-sized sub-periods were considered as a repeated

measurement On these Six response variables. This

92
suggested that a repeated measuresresearch design Should
be used with the 18 scores for each corporation from the
sample serving as the repeated measurements.

A multivariate repeated measures design model was
selected for this portion of the study.21 The principal
purpose for using this model was to test whether there
was an interaction between the repeated measurements on
the response variables and the main class effects of the
model. Because there were three measurement points it
was possible to test for three types of trends over time:

a constant, a linear trend, and a quadratic trend.

Significance Levels for Statistical Tests

The actual significance levels will be reported
for each of the statistical tests so that the reader can
make an independent judgment as to the significance of a
reported result. But when reporting conclusions on the
results from the statistical tests, a level of signifi-
cance, a = .05, will be Observed. That is, if a statisti-
cal test yields a value whose probability of occurrence is

.05 or less then the null hypothesis Ho corresponding to

 

21For a further discussion of this technique see,

Darrel R. Bock, "Multivariate Analysis of Variance of
Repeated Measurements," C. W. Harris (ed.), PrOblems in
Measuring Change (Madison, WI.: University of Wisconsin
Press, 1963), pp. 85—103.

 

93

that test will be rejected and the alternative H will be

1
accepted. Given the Specified a level, then the prob—
ability of committing a Type I error -— to reject Ho when
in fact it is true —- will also be equal to .05.

The decision to set a at .05 was based on two
criteria. First, this level appeared to be full justi-
fied given the nature of this study, the depth of the
empirical sample, the statistical models employed and the
prior research in this area. Secondly, a similar level
has been widely used in other areas of social science
research. Although the level of Significance was set at
.05, when reporting the results of the statistical tests
in the next two chapters a flexible approach to inter-
preting these results was taken. That is, regardless
of whether the significance level is exactly .05 or less
the principal thrust will be to document the results and
to interpret these results for possible underlying rela-

tionships.

Limitations of the Methodology
Because of the dual requirements for a balanced
Sample Of multinational and domestic firms within an
ithdustry group and the need to maintain a reasonable sample

Si2e, the number of different industry codes included in

 

94
the final stratified pOpulation was necessarily restricted.
TO the extent that this group was not representative of
the total pOpulation, generalizations from the results of
the study to this total population would be subject to
some degree of error.

Strictly Speaking the results of this study are
applicable only for the nine years, 1963 through 1971.
But, to the extent that they are typical of the years
before 1963 and future years following 1971, this would
not be a Significant limitation.

Because the criterion variables used to classify
corporations as either multinational or domestic could
only be measured on a nominal scale, the range of appro-
priated statistical models was constrained. At the Same
time, the lack of a continuous variable to measure the
degree of multinational expansion also entailed some

sacrifice in the precision of the results.

Summary
The first section of Chapter III provided a
description of the stratified corporate population from
which the two samples for this study were ultimately
chosen. This section continued by describing how the

corporations in this population were classified into the

95
two corporate groups: multinationals and domestics. AS
selected, the initial sample contained 146 corporations
while the final sample had 140 firms.

The next section described the Single empirical
response measure for the shareholder's rate of return and
the five surrogate measures for the shareholder's risk
exposure. Following this, the time periods covered were
specified as the nine year period, 1963 through 1971, and
the three equal sub-periods: 1963 through 1965, 1966
through 1968, and 1969 through 1971.

The final sections Of the chapter contained an
explanation of the statistical models that were used for
the empirical portions of this study: the univariate
model, the multivariate design model and the multi-

variate repeated measures design model.

CHAPTER IV

RESULTS OF PRELIMINARY STATISTICAL TESTS

Introduction

The order of presentation for the preliminary
results of this chapter and the final results in the next
chapter will be the same as the research hypotheses that
were given in Chapter II. Each of these sections opens
with a brief statement of the research hypothesis under
review. In the interest of brevity these summaries are less
formal than was the presentation in Chapter II. Follow-
ing this, there is a Short explanation as to how this
particular research hypothesis was Operationalized into
an empirically testable statistical hypothesis. Next,
the statistical hypotheses are presented. The results
of the statistical tests that were employedzne then given
together with an interpretation of these results. On some
of the research hypotheses additional analytical work was
performed. When this was done, the findings of that work
are presented next. A summary of the results and any

accompanying conclusions are presented in the closing

96

97
portion of each section.

This chapter presents the outcomes from the series
of independent statistical tests which were performed
using the univariate statistical model presented in
Chapter III. The particular statistical test used was
the Mann—Whitney U test for two independent samples.
Later references to this test used the Shortened title
MAW U test. The corporate sample that was used in this
part of the study was the "initial" sample presented in
Table 2.

Before proceeding with the details for the first
hypothesis, some further comments and qualifications
need to be made about the results presented in this
chapter. First, the principal test used was the Mann-
Whitney U test which is a univariate test between two
independent samples. When a series of these tests is
performed, as was the case for the individual industry
groups, the overall Significance level is not the prob—
ability level associated with each of the tests, but
instead is a function of the probability levels from the
series of individual tests. Thus in interpreting these
results the reader Should be aware Of the implications
that necessarily follow when a series of these MAW U

tests is employed. It should also be pointed out that

98
the data for the individual industry groups are a sub-
population from the larger overall multinational and
domestic corporate groups. Second, the five response
measures that were Specified for the different aspects of
risk exposure are not independent of each other. This is
clearly indicated by the results presented in Appendix B
where it is shown that there was a sizable common element
between many of the risk surrogates. Again, in interpret-
ing the results from the various tests associated with
the different risk measures the reader Should be aware
that there was a definite correlation between those mea-
sures, and therefore the individual tests are not totally
independent of each other.

Last, the work undertaken in this study was pri-
marily of an exploratory nature with emphasis on identify—
ing possible underlying relationships. As such, when
analyzing the results from the different statistical tests
it was recognized that there could be a sizable number of
outside factors that would tend to obscure the actual
relationships. And given the aggregative nature of the
study, the action and influence of these factors could
substantially reduce the significance on some of the
observed results. For this reason, supplemental work was

undertaken on many of the individual industry results.

99
The purpose of this analysis was to attempt to identify
and explain some of the probable causes when an industry's
results were Opposite from those for the overall corporate
Sample, or where the probability levels associated with
those results was very high. The reader is cautioned,
however, that the explanations that are offered represent
just that: possible explanations. They should not be
interpreted as being the only explanation, nor Should
they be considered as definite answers as to the exact
causes of the observed results. Nevertheless, it was
felt that a considerable amount of additional information
and insight could be gained from the supplemental analysis
of individual industry results. The explanations provided
were those that were supported by the data and appeared
to be viable. Certainly these brief summaries present
only the major factors underlying the observed results.
Consideration of additional factors not encompassed by
this analysis could substantially alter these probable

explanations and possibly suggest other alternatives.

100

Rate of Return Differential -
Short Term Time Horizon

Research Hypothesis H1: Rate of Return

Differential During_a Period of Substan-
tial Contraction in the Domestic Economy

Research hypothesis H1 stated that if the rates
of return on the common stocks of domestic corporations
were compared to those of multinational corporations
during a period of substantial contraction, a recession,
in the domestic economy, then the rate of return achieved
by common stock shareholders in the multinational group
would be significantly greater than those achieved by
their counterparts in domestic corporations. To Oper—
ationalize this hypothesis, the quarterly price relatives
for the common stocks of the multinational corporations in
the "initial" sample were compared with those Of the
domestic corporations in that sample. The period of com-
parison was the three year Sub-period, 1969 through 1971.

The statistical hypotheses were:

Hla:

0 The price relatives for the common stocks

of multinational corporations will not
differ from the price relatives for the
common stocks of domestic corporations,
when that comparison is made during a
period of substantial contraction in the
domestic economy.

Hla: The price relatives for the common stocks
of multinational corporations will be
greater than the price relatives for the
common stocks of domestic corporations,

101
when that comparison is made during a
period of substantial contraction in
the domestic economy.

Table 4 presents the average price relatives for
the common stocks of the total multinational and domestic
corporate samples as well as the details for the individual
industry groups. The second half of this table presents
the statistics and significance levels associated with
the differences between the price relatives for the multi-
national corporate sample and those for the domestic
Sample. These are given for the combined sample that in-
cluded all industry groups, as well as for the individual
groups.

A preliminary indication Of the direction of the
differences between the two groups can be gained from
comparing the average price relatives, columns 3 and 4 of
Table 4. For the combined industry sample and for eight
out of the nine individual industry categories the multi—
national contingent was higher than their domestic
counterparts. This was in accord with g priori predic—
tions. The one exception, Industrial Equipment, is ex-
plored further in a subsequent paragraph.

The results of the statistical test for the com-
bined sample, Shown in the last column of Table 4, suggest

that the common stocks of the multinational group had

102

museum zhumSpsfi mnu use

.umou pmameumco m homes

.Hmoausopa mum

.N wanna a“ so>aw wmosu Bonn pmsouuonw hauzwfiam mum mOHuHu m>auawuumaaa

 

 

 

 

 

How. o.Hm .auaaz Hoo.H ooo.~ we mvooo Hmuauao saw
canamquOOsm .oqmaus0fiom

“as. o.a~ .Huasz mma.o oma.o we mmaaaasm cam
Osmaaasvm Hoowuuuwau

H2. 0.2 .332 3:. some 3 “853:5
nousaaou pom wouwmo
com. o.qs ummsoa coo.a Hmm.o NH uawsaasem HmauumseaH
omm. o.mH .aussz oom.o Nma.o NH mecca magnum:
MSN. 8.4m .auasz oma.o moo.a ma muaumamoo saw mamom
mac. o.H~ .auanz mHo.H omo.a we HmuauamUQSHmnm new ways
soc. o.m .Huaaz Nmm.o aao.H ea «amusemeo gamma
2:. 8.8 .31.: 884 884 3 $338 as soon
coo. o.m~o~ .ausaz mam.o Hoo.a sea (macaw suumaeaH Haa

D HmHHmBm UHumflumum QDOHO .ummEOQ .fiuHflz OHQEGW
no mans mo 2 undamam moouu mamamm mo «soauafiuommn
««Suaaanmaoum you m>aumaom muam stumsunu
owes D 31: spasm mwmum><

 

H

wzozoom UHHmmzoo Nae zH onHU<mHzoo
A<HHZ<HmmDm mo Qonmm < ozHMDQ AdHHzmmmmmHn szHmm ho ma<¢
" m mHmmmHomwm mom<Mmmm mo Emma D MMZHHmSIzz<Z mmB mo mHADmmm

c mum<9

103

Significantly higher price relatives, and therefore higher
rates of return, than did their domestic counterparts.
But the results for the individual industry categories
present a mixed pattern. Specifically, in five out of
the nine industry categories the U statistics associated
with the differences between the two groups were Signifi-
cant only at the .15 level or above. And, as was previ-
ously pointed out, the domestic group within the Indus-
trial Equipment industry had higher price relatives than
did the multinationals. The next section analyzes the
results from the individual industry groups and offers
possible explanations of the factors and circumstances
underlying the Observed results. While it was felt these
wenaplausible and viable explanations, it was recognized
that there could be a number of alternative eXplanations.

The reduction in the size of Sample when going
from the overall corporate group to the individual in-
dustry category affected the probability levels of the
individual U statistics in two distinct ways. First, the
power of the applicable MAW U test would necessarily be
lowered by the reduction in sample size. Consequently,
a shift of several positions within the rank ordering of
the sample corporations could change the prObability

levels quite dramatically. Second, because of the small

104
sample sizes, several atypical firms within the multi-
national or domestic group could perceptibly alter the
outcome of any given statistical test. This latter Situ-
ation appeared to be a major contributing factor for the
high probability levels in three out of the five industry
groups with levels of .15 or above: Soap and Cosmetics,
Industrial Equipment, and Scientific, Photographic and
Optical Goods.

For the Soap and Cosmetics industry the three
atypical firms were concentrated in the domestic group.
In particular, unlike the other members of this domestic
contingent, these three firms were less affected by the
recession in the U.S. economy than were the others. One
firm, from among the three, was clearly the "giant" of
the industry: nearly three times the sales Of its nearest
competitor. Because of its size, this firm had diversi-
fied into a larger number of product lines which probably
reducedthe effects of a recession on its overall Opera-
tions. For the other two firms, there was no apparent
single cause or factor which could account for their
performance. Since it was not the purpose of this study
to do an in-depth analysis of each firm, this was not
pursued further. None of the firms was so unique as to

require its removal from the sample. While the

105
difference for this industry was in the same direction as
the combined sample, the probability level associated with
this difference was so high that the possibility that this
difference arose by chance could not be rejected even after
the above review. Consequently, the results from this in-
dustry category offer little, if any, support for the
initial conclusion of this section: given certain pre-
scribed conditions in the domestic economy, shareholders
in multinational corporations achieve significantly higher
returns than do their domestic counterparts.

The converse result from the Industrial Equipment
group appeared to be due to several things. First, the
intragroup homogeneity appeared to be considerably less
than other industry categories. That is, the product
lines of the various firms within both the multinational
and domestic groups suffered from a lack of comparability
not only within the group but especially between the two
groups. Furthermore, there were two domestic and one
multinational firm whose performance was atypical of the
other members within their respective group. In large
measure this could be attributed to their unique product
line which did not parallel the other firms. Although
it was not pursued in this study, a different industry

classification scheme might have yielded different

106
results: however, this cannot be stated with certainty.
It should also be pointed out that this same industry
group reappears in subsequent sections, again because its
results are generally Opposite of those for the other in—
dustry categories. What this analysis suggests is not
that this group represents an exception or denial of the
original overall conclusion, but that there were some
prOblems with the Specification of this industry category.
In particular it is questionable whether the firms within
this category were sufficiently homogenouS to permit
direct comparison Of their results.

The last industry category where atypical firms
within the group appeared to be the major factor was the
Scientific, Photographic and Optical Goods group. This
might possibly have been anticipated, Since this category
was a coalition of three SIC industry codes. Consequently,
there would likely be less homogeneity among the member's
product lines. While this no doubt reduced the precision
when comparing the two groups, the principal cause of the
high probability levels was the existence of two moder-
ately atypical firms: this time in the multinational
group. Both were small firms with somewhat specialized
product lines and both had an exceptionally high incidence

of poor returns during the twelve quarters, 1969-1971.

107

Although the exact causes of the poor performance for the
two firms was not identified, the pattern for each one
was such that it appeared to stem from internal factors
unique to each firm. AS a result of this poor perform-
ance, both firms ranked at the very bottom of the combined
sample. This low ranking did not alter the direction of
the difference between the groups: the multinational
firms were higher than their domestic counterpart. Yet
it did, in all likelihood, cause the high probability level
for this particular MPW U test statistic. A further con-
sequence of the low relative ranking by these two multi—
national firms was that the average price relative Shown
in columns 3 and 4 of Table 4 do not accurately reflect
the Spread between the two groups. The respective group
medians are a superior statistic for this comparison.
That statistic for the multinational group was 1.03,
while the domestic group's was .99. After considering
the findings from this analysis, the conclusion was that
the difference for this industry was in the prOper direc-
tion, but the lack of significance suggests that these
results offer only limited support for the initial con-
clusion of this section.

The remaining two industry categories where the

probability levels were greater than .15, Office and

108
Computer Equipment and Machine Tools, had no single factor
that stood out as being the principal cause for the lack
of significance. Rather it was probably the combined
effects Of several factors that caused the observed condi-
tions. The lack of intragroup homogeneity among the
product lines of the firms within both industry categories,
but especially in the case of the Office and Computer
Equipment group, appeared to be a contributing factor.
Whether an altered industry classification scheme would
have given better results cannot readily be determined.
In the case of Machine Tools, the specialized nature of
its products, and the fact that the business cycle for
that industry tends to lag the general business cycle,
may have influenced the results for that category. In
the case of the Office and Computer Equipment industry,
the results suggest that there was a difference between
the two groups and that it was in the expected direction.
Yet, any conclusion as to whether this supports the
initial conclusion must be tempered by the high prOb-
ability levels for the NEW U test statistic from this
industry. Thus, the conclusion was that this industry
Offered, at best, limited support for that initial conclu-
Sion. As to the Machine Tool industry, the conclusion

‘Was that the difference was small and therefore could

109

have been a chance occurrence. Thus, the results from
this industry really offered no support for the initial
conclusion that multinational Shareholders achieved Sig-
nificantly higher returns than did their domestic counter-
parts, given certain prescribed conditions in the domestic
economy.

What conclusions can be drawn from this interpreta-
tion of individual industry results? First, except for
the Opposite result found in the Industrial Equipment
industry, the differences within the other eight industries
were in the predicted direction: the multinational group
was higher than the domestic group. But the size of these
differences, and therefore the statistical Significance,
was not consistent across the eight industries. The
results for three industries clearly offer substantial
support, as did the outcome from the test using the com-
bined samples, for the conclusion that the multinational
group was significantly higher than its domestic counter-
part. The differences associated with two other industries,
were the Significance levels were .117 and .201, could at
best be classified as Providing moderate support for the
above contention. Because of their very limited statisti-
cal significance, the results for the remaining three in-

dustries really added little or nothing to the position

110
that the multinational group was the higher of the two.
An analysis of the individual industry results for those
with very low significance levels suggested that, for the
most part, the lack of Significance could be traced to
several atypical firms. These were usually located in
the multinational group, but on several occasions they
were from the domestic contingent. A further review of
the individual industry results suggested that for the
majority of multinational firms the spread between their
average price relative and those for their domestic
counterpart was substantially larger than the aggregate
results in Table 4 would imply. But when the unique firms
were introduced the spread between the two groups narrowed
substantially. The conclusion from this analysis was
that, while the differences were in the predicted direc-
tion, the high probability levels associated with their
respective Mew U test statistics made it difficult to
reject the possibility that the difference was a product
of chance. As such, these results offered little, if any,
support for the initial conclusion. Because of the quali-
fications and special circumstances surrounding the Indus-
trial and Equipment industry, it was felt that the con-
verse results from this one industry did not present a

strong case for denial of the initial conclusion.

111

In summary, the results from the tests of these
statistical hypotheses suggest that the price relatives
for the multinational group were higher than those for
the domestic group. Specifically, the results from the
statistical test for the combined sample was significant
at the .006 level. Since this was less than the Specified
level of .05, the null hypothesis H;3 was rejected and the
alternative Hia was accepted. It was concluded that price
relatives for the common stocks of multinationals were
greater than those for the common stocks of domestic cor-
porations. Since this alternative hypothesis Hia is but
an Operational statement of the first research hypothesis
H1, the outcome from this test provides confirming evi-
dence for that hypothesis. While the results from the
individual industry categories were not considered direct
tests of the research hypothesis, the findings for the
individual categories generally lend further support to
research hypothesis H1. There were, however, several
anomalies within the individual results which were dis-
cussed previously. By inference, the conclusion is:
given a period of substantial contraction in the domestic
economy where the corporations are headquartered, the rate

of return on the common stocks of multinational corpora-

tions will be significantly greater than those on the

112

common stocks of domestic corporations.
Research Hypothesis H2: Rate of Return
Differential During a Period of Rapid Growth or
a Period of Limited Growth to Near Zero
Growth in the Domestic Economy

Research hypothesis H2 stated that if the rates of
return on the common stocks of multinational corporations
were compared to those of domestic corporations during
either a period of rapid growth or a period of limited
growth to near zero growth in the domestic economy
then the returns achieved by common stock Shareholders in
the multinational group would not differ from those
achieved by their counterparts in domestic corporations.
This research hypothesis was Operationalized by comparing
the quarterly price relatives for the common stocks of
the multinational corporations in the "initial" sample to
those of the domestic corporations in the same sample.
Separate empirical tests were made for this hypothesis
using two different sub-periods from the total study
period. The first included the years 1963 through 1965,
while the second included the years 1966 through 1968.
The discussions in Chapter III pointed out that the first
sub-period represented a period of rapid growth, while
the second represented a period of limited growth to near

zero growth in the U.S. economy. The statistical

113
hypotheses were:

H23: The price relatives for the common stocks

0 of multinational corporations will not
differ from the price relatives for the
common stocks Of domestic corporations,
when that comparison is made during
either a period of rapid growth or a
period of limited growth to near zero
growth in the domestic economy.

a . .
H2 : The price relatives for the common stocks

of multinational corporations will differ
from the price relatives for the common
stocks of domestic corporations, when
that comparison is made during either a
period of rapid growth or a period of
limited growth to near zero growth in
the domestic economy.

AS Specified, there was a separate empirical test
for each of the patterns of economy activity that was

. . . . . 2a 2a

speCified in statistical hypotheses, HO and H1 . Table
5 exhibits the results for both of these tests: the first
half pertains to the sub-period, 1963 through 1965, the
second half 'tothesub-period, 1966 through 1968. The aver-
age price relatives are given for the combined multi-
national and domestic samples as well as the individual
industry groups. Also presented are the statistics and
Significance levels associated with the differences be-
tween the price relatives for the common stocks of the
combined multinational group and those for the common

stocks of the domestic group. Similar statistics are

presented for the multinational and the domestic contingent

114

.kuwuaapw one masouw muumaps«

as“ use .~ «Hams

.uouu voaanuuoau a uomcc

sq =o>wm omo:u scum megaphonm haunmaau ous uoauau u>uuaauuuoa¢

 

 

 

 

 

 

Ono. O.wa .umosoo sno.a omo.a mon. o.nm .Huas: mco.a ooo.a apooo «nuwuno pan
ownauquuozm .quuusuuom
«mm. 0.0m .«uanz nHo.H «NO.H mma. o.~m .ummsoo moo.a Hmo.H nowaaasm and
ucoaawsau Hauuuuuuau
SNm. o.o~ .auasz ~no.a moo.a mac. o.- .umoaoa Hao.a «no.a unoaaaaau
umu:afioo can ouauuo
«ma. o.m .umoaoo amo.H H~o.a oqw. 0.0H .«uanz Hmo.H ~wo.a unaaawavu Hawuunspcu
nmq. O.MH .«uaaz Nmo.H umO.H «on. o.~H .umwaoo ooO.H o~o.~ «H009 unusual
ohm. o.m~ .auasx aqo.a moo.H Noe. o.am .«uasz oao.a Hmo.~ muauuanoo use mason
oou. o.o~ .umosoa omO.H OHO.H mac. 0.0m .«uaaz omO.H Hmo.a Hauwusouuauosm pan mane
Noe. o.m~ .ummaon wwo.fl o~o.H ooo. o.- .«uasz mco.a mno.H manuaamso canon
nmu. o.- .«uasz oao.a oNo.H nmm. o.mn .«uasx ono.u cmo.~ wwnuo>om pas poem
mwo. o.Hem~ .umumdo uno.H mmo.H one. o.mqo~ .Huasz Hmo.a nno.H nmsouo ”Mumspcu HH<
a uoaawsm uuuufiunum asouo .umoaoo .Huasx : uoaaqam caumauaum nacho .umoson .wuHs:
no mass no a unusmam msouo «Assam no mash we a unonwam macho seamen
caowuawuonoo
«sxuaaabmnoum qu o>auaaum saxuaaabmboum yew o>auuaou spun: a
game 5 3|: ouwum wwwua>< once 3 3n: uuwum owmuo>< p H

 

 

powuomansm womancoma

poHuomIbom womanhood

 

UZHHDQ A<HHZH¢Mhth zmbhmu mo Nh<m

N

n uAn<H

Mzozoum UHHmmzoa may zH mh3omo ome m<mz OH mHzomo QmHHZHA mo aonmm < mo mhsomu nHm<M mo OOHmmm 4
n m mummmhomhm mum<umux mo memH D VHZHHm3IZZ<= nah ho wHADmHu

115
within each of the nine industry groups. Because two
separate sub-periods were used in the empirical test of
this hypothesis, and Since each contained several special
situations that required further explanation, the analysis
of their results are presented separately.

For the first sub-period, 1963--1965, the average
price relative for the combined multinational group at
1.055 was little different from that of the domestic group
at 1.051. This initial conclusion of limited difference
is further substantiated by the lack of statistical Sig-
nificance of the U statistic associated with these differ—
ences: significant at the .94 level. Continuing with the
pattern used in the preceding section, the results for the
individual industry categories are examined next. A com-
parison of the average price relative for the multinational
group with those for the domestic grOup within each industry
presents a mixed pattern. The multinational group was
higher in six industries while the domestics were higher
in three industries. The statistics associated with these
differences, however, suggest that in eight out of the nine
industries the differences between the groups were not
Significant at the .05 level. Consequently, the results
from these eight industry categories appear to support

the earlier conclusion that the differences between

116
multinational and domestic firms were small.

The Office and Computer Equipment industry was the
exception to the above general statement. Within that
industry the domestic group was the higher of the two,
and differences between the two groups were significant
at the .038 level. As was pointed out in the preceding
section, the existence of severe atypical firms within a
particular industry, especially where the sample of firms
within that industry was small, can substantially alter
the results. In the case of this industry, there were two
domestic firms which had unusually high price relatives
compared with the other firms of the domestic group.

Also, there was one multinational firm whose average price
relative was substantially less than those of the other
multinationals. In all likelihood these three firms were
the principal cause for the domestic group's average
being significantly higher than that for the multi-
nationals. Regarding the first domestic firm, the market
price of its stock rose Sharply during 1965, but this
pattern was completely reversed in 1966 and the subsequent
decline nearly matched the 1965 run-up. Had the first
sub—period encompassed both years, the average price
relative would have been only moderately positive. But

because the 1963-~1965 sub-period encompassed only the

117

price run-up, the average for that period had an upward
bias. At the same time the average for the 1966—-l968
sub-period suffered a downward bias. One possible explan—
ation for this observed pattern was that during 1965 and
several later years the firm in question was pursuing an
aggressive acquisition program not only within this in-
dustry, but into other product lines as well. Whether
this was a major contributing factor to the substantial
price run-up cannot be stated with certainty. The common
stock of the second atypical domestic firm exhibited a
Similar pattern of unusually large price increases during
1965, causing a similar substantial upward bias in its
average price relative for the l963—-l965 period. While
its market price also turned down during the early part
of the l966-—l968 sub—period, the subsequent decline did
not completely reverse the 1965 price run-up. Consequently,
its average for the 1966-1968 period had less of a down-
ward bias. A limited review of this case revealed no
single factor that would explain these observed results.

The multinational firm that was singled out had
very low price relative averages for each of the three
sub-periods compared with the other firms in the multi-
national group. In particular its average for the 1963—-

1965 period was more than 1.5 standard deviations below

118

the mean for the combined multinational group. In large
measure this poor performance appeared to stem from the
firm's inability to increase its overall profits despite

a continual growth in sales. Whether this was a conse-
quence of its heavy dependence on a limited product line
cannot be ascertained with certainty, but it appeared to
be a significant factor in this case. What was concluded
from the review of this apparent "exception" industry?
First, while the results for this industry group do not
support the initial conclusion of no difference between
the two groups, they do not present a very strong case

for denial of that conclusion either. After analyzing the
three firms whose results were judged atypical, it appears
that the significant differences Observed in this industry
prObably stemmed from special circumstances within these
three firms. Therefore, the results did not suggest that
there was some systematic difference between the typical
multinational firm and its domestic counterpart.

Turning next to the second of the two periods,
l966--1968, the differential between the combined domestic
group at 1.037 and the multinational group at 1.035 was
again very small. The statistical significance, as indi-
cated by the U statistic associated with these differences,

was small and not significant at the .05 level. The

119

comparison by individual industry category revealed an
even more mixed pattern than that for the 1963--1965 sub-
period. By direction of difference, the multinationals
were higher in five industries while the domestic group
was higher in the remaining four. For eight of these
industries the corresponding U statistic clearly indicated
that these differences were not Significant at the .05
level. The actual prObability levels ranged from .132 to
.529. Certainly, the results from these eight industries
support the earlier conclusion of no difference between
the multinational and domestic groups.

The differences for the single remaining industry,
Scientific, Photographic and Optical Goods, were signifi-
cant at the .05 level. As was previously pointed out,
this industry was a consolidation of three SIC industry
codes and therefore there was likely to be less intragroup
homogeneity. Consequently, any comparison between groups
within this industry was likely to be less meaningful. A
review Of the results for the domestic group revealed no
truly atypical firms. In fact, the individual price rela-
tives for the domestic firms within this industry were all
closely clustered about the domestic group's overall mean.
Instead, the large Observed differential between the two

groups appeared to be due to the unusually low average

120
price relatives within the multinational group. Specifi-
cally, there were three multinational firms whose relative
ranking among the eighteen firms of this industry skidded
from sixth, seventh, and ninth during the l963--l965 Sub-
period to positions of sixteenth, seventeenth, and
eighteenth, respectively, for the 1966-1968 sub-period.
No single common factor was identified which would account
for the substandard performance of the three during the
l966--l968 period. Instead several unique factors were
identified for each of the firms which probably contributed
to the Observed market results.

This analysis, plus the one from the preceding
section, points up the inherent prOblem of attempting to
compare two corporate groups using a Single criterion
variable. At first glance, the results associated with
the "exception industries" present a case for denial of
the initial conclusion of no difference between the multi-
national and domestic groups. After a more detailed
analysis, however, it appeared that many of the major
differences within each industry were caused by special
circumstances that have little, if any, relationship to
whether the firm in question was a multinational or a
domestic. Consequently, the conclusion was that the

results for these exception industries provide only

121
limited support for the position that the price relatives,
and therefore market performance, of the multinational
group were necessarily higher or lower than those for
their domestic counterpart. Rather this and the previous
analysis suggested that special Situations within a few
firms from a particular industry had probably caused the

differences between the two groups to be statistically

significant at the .05 level or less.
Clearly, the results from the M-W U tests on the

combined samples from the l963--l965 sub-period, and from
the 1966—1968 sub-period, suggest that the differences
between the multinational group and their domestic

counterparts were not large. And that they were not

statistically significant at the .05 level. At the same

time, the results for eight of the individual industry

categories from the l963--l965 sub—period, and eight
industries from the l966--l968 sub-period, suggest there

was no difference between the multinational group and
their domestic counterpart. The differences for the
single remaining industry category from the 1963-~1965
Sub—period, and the single industry from the 1966-1968
sub-period, were larger and significant at the .05 level.
But the general conclusion after reviewing the individual
results from both industries was that neither industry

presented a strong case for denial of the above conclusion

122
of no difference. In summary, based on the above findings

the null hypothesis H2a was not rejected. Since the null

0
H2a

hypothesis 0

was but an Operational statement of the
second research hypothesis H2, the results from these
tests provide confirming evidence for that hypothesis. By
inference, the conclusion is: the rate of return on the
common stocks of multinational corporations will not be
significantly different from the rate of return on the
common stocks of domestic corporations, when that compari-
son is made either during a period of rapid growth or
during periods of limited growth or near zero growth in
the domestic economy.

Difference in Risk Exposure -

Long Term Time Horizon

Research Hypothesis H3: Difference in Risk
Exposure When Measured Over a Long Term Time Horizon

Research hypothesis H3 stated that, when measured
over a long term time horizon, the risk exposure on the
common stocks of multinational corporations will be sig-
nificantly less than the risk exposure on the common
stocks of domestic corporations. Five distinct surrOgate
measures were employed to capture the different aspects
of risk exposure. In Operationalizing this hypothesis,

the risk measurements for the common stocks of the

123

multinational corporate sample were compared with the risk
measurements for the domestic corporate sample. That is,
the scores On the five separate risk measures for the
multinational group were compared with the scores on
Similar measures for the domestic group. The corporate
sample used in this empirical test was the "initial"
sample exhibited in Table 2. Since the hypothesis Speci-
fied a long term time horizon, the comparison was made
over the total nine year period, 1963 through 1971.

One further point of information about the five
risk measures: for the first four measures - Gini'sl
coefficient of concentration, modified quadratic mean,
beta coefficient and error of the estimate - the larger
the positive score on these measures, the greater the risk
exposure. But on the fifth measure — relative skewness --
a large negative score indicates high negative skewness
in the distribution of price relatives, and therefore the
possibility of a major financial loss. Thus, for the
first four measures the group with the highest scores
would likewise have the greater risk exposure. Conversely,
for the final measure, the group with the highest score —
largest positive values - would have the lesser risk

exposure.

 

1Gini, Op. cit.

124

Since there were five distinct measures, a separate
univariate statistical test was performed on each one. But
the results for the fifth measure, skewness, were treated
differently. Partially, this was because relative Skew-
ness was considered a supplement to the other four and not
a separate independent measure. AS such, the results from
the statistical tests on this measure were not considered
a direct test of the research hypothesis. Instead they
are reported and interpreted in the supplemental analysis
portion of this section. The statistical hypotheses for

the four remaining risk measures are:
H3a: The overall variability in returns for

0 the common stocks of multinational cor-
porations will not differ from the over-
all variability in returns for the com-
mon stocks Of domestic corporations,
when that comparison is made over a
long term time horizon.

3a . . . .

H : The overall variability in returns for the
common stocks of multinational corporations
will be less than the overall variability
in returns for the common stocks Of domestic
corporations, when that comparison is made
over a long term time horizon.

3b
H

The downside risk exposure on the common
stocks of multinational corporations will
not differ from the downside risk exposure
on the common stocks of domestic corpora-
tions, when that comparison is made over a
long term time horizon.

H : The downside risk exposure on the common
stocks of multinational corporations will
be less than the downside risk exposure on

125
the common stocks of domestic corporations,
when that comparison is made over a long

term time horizon.

H : The market risk for the common stocks of

0 multinational corporations will not differ
from the market risk for the common stocks
of domestic corporations, when that compari—
son is made over a long term time horizon.

Hie: The market risk for the common stocks of
multinational corporations will be less
than the market risk for the common stocks
of domestic corporations, when that compari-
son is made over a long term time horizon.
Had: The residual risk exposure in the common

stocks of multinational corporations will

not differ from the residual risk exposure
in the common stocks of domestic corpora-

tions, when that comparison is made over a
long term time horizon.

H3d: The residual risk exposure in the common

stocks of multinational corporations will
be less than the residual risk exposure in
the common stocks of domestic corporations,
when that comparison is made over a long
term time horizon.

The results for the individual univariate statis-
tical test associated with each of the four risk hypotheses
are presented in Table 6. These results are listed in the
same order as the hypotheses were given. The table also
gives the results for the fifth risk exposure measure;
this was the measure that was not accorded a separate
statistical hypothesis. The description column explains

the particular aspect of risk exposure that was investi—

gated and gives the surrogate measure that was used. The

126

.ummu VOHHQHIUQO a Mom!

 

 

 

 

mqo. o.HmNN .uwmaoo na.0| cm.o1 mumsaoxm o>fiumaom
”mmoa anacondam wane:
mac. o.om- .ummaoo and. awe. Oumawumm one no uouum
nxwfim Happened
Hmo. o.q¢H~ .umwaoo mom.a «Nm.a usowoawwooo muom
”seam segue:
«mo. o.~o- .ummsoa cos. «so. cams chanteuse vuamaeoz
”swam opamcaon
«Ho. o.qu~ .ummaon moo. mmo. coaumuusmosoo mo powwowmwooo m.«owo
"monoumm s“ mafiafibmaum> Hamum>o
D HOHHmSm Oeumfiumum asouu .umoaoo .auasz
no menu mo D uwonwam ozone mdmwmm was muonmmz
«zuHHHbmboum you owsoammm whomoaxm xwfim
mummy : 3L2 mwmum><

 

m

mamz<m UHHmmzoo 92¢ A<ZOHH<ZHHADZ A4908
”ZONHmom NZHH Emma 020A < mm>o A<HBZHMMhmHG anamomxm MmHm
" m mHmmmHomwm wom<mmmm mo Hmmh D MNZBHm3lzz<z Huh ho mfigbmmm

o mummfi

127

average response on the five empirical measures for the
combined multinational and domestic groups are given to-
gether with the U statistics associated with the differ-
ences between the two groups. Results for the nine in-
dustry categories on the five different empirical measures
are presented in Table 7. The discussion and interpreta-
tion of the results which follows is divided into four
sections: the first covers overall variability in returns:
the second covers downside risk: the third includes both
market risk and residual risk: and the fourth covers major
financial loss.

The results for the first aspect of risk exposure,
overall variability in returns, indicated that the aver-
age response for the multinational group was less than
that for the combined domestic group. And the U statistic
associated with the differences between the two groups was
significant at the .014 level. The initial conclusion from
these results was that the overall variability in returns
on the common stocks of multinational corporations was
less than that for the common stocks of domestic firms.
For eight out Of the nine industry categories in Table 7
the multinational group was the lower of the two. This
suggests a similar conclusion. In seven of these indus-

tries the differences between the two groups, as reflected

128

.ueOu voawenuoue I name.

.auunusevw one uasonu annuapsu use .N ode-H an se>wm ocean lonu peanuneao hausuuao one seamen Ibuuanno-onc

 

 

«no. .uuoaoo nan. .uaoaoo nan. .uooaoo on~. .uuuaoo nan. .uuoaoo .ooao nounuao on.
unseenueuozm .euuuuauuum

«no. .umuaoo nNN. .uauaoo ooo. .uaoaoo n-. .uooaoo oon. .uuoaoo .unnao5m can
useaaasvu Haennueoam
«no. .uauaoo «SN. .uuoaoo non. .uouaoo oon. .uaoaon non. .unoaoo uuoaaasou sausages on. «cameo
moo. .nunaz moo. .nunsx non. .nunaz non. .nunsz ooo. .nunax uaoaannou nanuuuaoan
ooo. .uuoaoo own. .uooaoo ooo. .uuuaoo non. .uauaoo oon. .uuoaoo unoon magnum:
omo. .qun—OQ 55H. .UOUEOQ HON. ounun—OQ and. .HIOQOQ H00. .uuaa IUfluOB-OU Us unflom
oon. .nunax man. .uuuaoo nno. .uooaoo nan. .uuoaoo non. .u.uaoo nuunusounau-na on. mono
one. .uuoaoo one. .uaoaoo ooo. .uuoaoo «no. .uuoaoo o~n. .uauaoa anaunaoao unuan
nos. .uaoaoo can. .uooaoo nos. .nunax omo. .uuoaoo nos. .uaoaoo oouuu>on one soon

: noaasam no aseno D nouauam no asenu D nedawam no nsono D nuaanam no abono a nudflqlm no ozone

vo>noobo we uuenmnz pe>neneo we neonmwm ve>nonno we uuozwnm ve>noabo we usezwnm vo>neono we newsman nowuawneeen
awnunnnaenoua alsunnasusoua acnunnnnunoua ..nunnnnanona :.nunnnaunoua . nus-aoan

 

 

 
 

auosaexm o>wuaaua unmanunu one we nonnu emmwunwueeo seem cue: .pa: puuwnvot .aeusoo we

.uuoou o.wafiu

 

mmbomo wmhmDQZH AanH>HozH "ZONHmom mzHH Emma Uzoa < mm>o

Aduazuunhhnn nubmomxu “mud ”n: mumumhomrz mum<umux mo mammh D rMZHHmzlzz<x ”:9 ho mHADmMu

s udnda

129

in the various U statistics, were marginally significant.
The actual probabilities ranged from .061 to .147. The
fact that each of these industry categories contained a
sample of limited size was, no doubt, a major factor in
the probability levels being as high as they were.

Within the Food and Beverage industry the differ-
ences were in the predicted direction: the domestic group
being higher than the multinational contingent. But the U
statistic associated with these differences suggested only
a limited spread between the two groups: significant at
the .447 level. This lack of significance appeared to be
caused by three multinational firms with unusually high
scores on this risk measure: they occupied second, third,
and fourth position in a ranking of these eighteen firms
by decreasing risk exposure. On the other hand, the
remaining six firms from the multinational Sample were
clustered in the lower two quartiles of this relative
ranking. These findings suggest that the probable cause
of the lack of Significance stemmed from the three multi-
national firms that had abnormally high risk exposure.

But in analyzing these firms further, no consistent single
factor emerged as the probable cause of their high scores
on this risk measure. Consequently, while the difference

for this industry was in the expected direction, the

130

conclusion was that the probability level was such that
the possibility of this being a chance occurrence could
not be rejected, even after the above review. Thus, the
results really offered no support for the initial conclu-
sion of this section.

The results for the Industrial Equipment industry,
contrary to the findings for the other eight industries,
disclosed that the multinational group within this industry
had greater overall variability in the returns on their
common stocks, and therefore higher risk exposure, than
was the case for the comparable domestic group. Several
points, however, need to be made at this time. First,
this same industry category appeared in the discussion of
the results for the first two hypotheses because, like the
present Situation, this industry's results were exactly
Opposite to those for the other industry categories. The
point was made in previous sections that the product lines
for the firms within this industry tended to be quite
heterogeneous. It was also noted that the sample size,
twelve firms, for this industry was small. Second, be-
cause this industry had converse results for all five
risk measures, a single general analysis was made. For
the most part, the differences between the two groups, as

shown by the U statistics associated with each comparison,

131

lacked significance. A limited analysis of the industry
disclosed two firms, one domestic and one multinational,
whose market performance was atypical to the other firms
within their respective groups. The multinational firm
tended toward substandard price relatives on average, but
the series of quarterly price relatives within any sub-
period was highly volatile. Exactly the Opposite situation
described the domestic firm: it had above average price
relatives and there was little volatility within the
individual quarterly price relatives. What was the con-
clusion? Because this industry lacked homogeneity among
the sample firms, and because of the atypical performance
of several sample firms within the industry, its results
did not present a strong case for the position that the
common stocks of multinational firms have greater risk
exposure than did those Of comparable domestic firms.

The results from the Mew U test for the combined
multinational and domestic corporate Samples present a

strong case for rejecting the null hypothesis Hga in favor

3a
1 .

statistical significance of their results, the individual

of alternative hypothesis H Ranked according to the
industry findings were as follows: a) results from the
first industry provided substantial support for the above

position, b) results from the second through the seventh

132
industry provided only moderate to limited support for the
above position, c) results from the eighth industry really
added nothing to the initial position, and d) results from
the ninth industry were Opposite from the above conclu-
sion: but given the qualifications for that industry, it
was felt that it did not constitute a strong denial. Con-
sequently, the conclusion was that the overall variability
in returns for the common stocks of multinational corpora-

tions was less than it was for the common stocks of

3a

domestic corporations. And, because H1

was but an Opera-
tionalized statement of research hypothesis H3, these
findings lend empirical support to that research hypothe-

sis. Before attempting to draw an overall conclusion,

however, the results corresponding to the second interpre-

tation of risk exposure, downside risk, need to be examined.

The average response for the combined multinational
group on the downside risk meaSure was less than that for
their domestic counterparts. Since the U statistic asso—
ciated with the differences between the two groups was
significant at the .035 level, these results suggest an
initial conclusion that the common stocks of multinational
corporations were indeed subject to less downside risk
than were the common stocks of comparable domestic firms.

Results from the individual industry categories, Table 7,

133
suggest a similar conclusion. In eight out of the nine
industries, the respective domestic groups were higher.
And, as was pointed out in the previous section , the
ninth industry, that of Industrial Equipment, requires
certain qualifications in the interpretation of its con-
verse results. The statistical significance of the dif-
ferences within the eight remaining industries presented
a mixed situation. Two of the industries had prObability
levels which were close to being Significant at the .05
level: actual levels were .052 and .056 respectively.
Given the limited sample Size that characterized these
individual industries, their results were interpreted as
providing reasonably strong support for the initial con-
clusion of this section. The conclusion from the three
industries where the prObability levels ranged from .135
to .197 was that they provided only limited support for
the initial conclusion of this section. While the direc-
tion of their differences paralleled those of the combined
sample, the high probability levels associated with these
differences necessitated a guarded conclusion as to the
extent they support the initial conclusion. Because
there was a definite lack of significance for the last
three industries, they were subjected to further analysis

in an attempt to find a possible explanation for their

134
high probability levels.

In the case of the Electrical Equipment and Sup-
plies industry, significance at the .227 level, the multi—
national firms tended to be grouped in the lower three
quartiles, but their relative positions were such that
the differences in rankings between the multinational and
domestic groups were not that significant. In as much as
there were no multinationals in the upper quartile and
given their positions in the other three, the results
suggest that multinational firms generally were lower:
but the differential, relative to the domestic firms, was
limited. But given the high prObability level, the con-
clusion was that the difference between the two groups
could be a chance occurrence. Therefore, the results for
this industry really provide little, if any, support for
the initial conclusion.

The differences within the second of the three
industries, Scientific, Photographic, and Optical Goods,
were only significant at the .254 level. A perusal of the
results for this industry revealed two extenuating circum—
stances. First, multinational firms occupied six out of
the total nine positions in the lower two quartiles.

That is, out Of the nine corporations with the lowest risk

exposure on this measure, Six were multinationals. The

135

remaining three multinational firms, however, were quite
the Opposite for they ranked first, second, and third in
terms of risk exposure. The Spread between their scores
on this risk measure and the corporations in fourth, fifth,
and sixth was small: this suggests that while their risk
exposure was greater than the three domestic firms ranked
immediately below them, the differences were not substan-
tial. Nevertheless, the conclusion from the results for
this industry was that they really added nothing to the
initial conclusion.

The probability level for the third industry in
this group, Office and Computer Products, was even higher
than those for the previous two industries. A review of
the details for this industry disclosed that much of the
differential between multinational and domestic firms
stemmed from the high incidence of domestic firms in the
high risk quartile of the total industry sample: three
firms out of the four firms in that quartile were domestic.
Although a considerable margin separated these from the
next three lower rankings, it appears that one would only
be justified in concluding that this industry's results
were in the predicted direction but that they really
Offered no support to the initial conclusion. The con-

verse results for the Industrial Equipment industry were

136

discussed at some length in the preceding section. As

was pointed up in that discussion, these results did not

present a strong argument against the original conclusion.
In summary then, the results from the combined

multinational and domestic corporate Sample presented a

strong case for rejecting null hypothesis Hgb and accept-

ing alternative hypothesis H3b The individual results

1 O
for the eight industries whose differences were in the

Same direction as those for the combined Sample were

interpreted as follows: a) two industries provided rea-

sonably strong support for rejecting Hgb, b) three pro-

vided limited support for the rejection of Hgb, and

c) three industries really added little, if any, support

for rejecting Hgb. When these industry results were con-

sidered concurrently with those from the combined sample,

they presented a reasonably strong and unified case for

3b

rejecting null hypothesis Ho

tive Bib. As such, the conclusion drawn was that the

and accepting the alterna-

downside risk in the returns to the common stocks of
multinational corporations was less than it was on the
Shares of comparable domestic corporations. Furthermore,
since Bib was an Operationalized statement Of research

3

hypothesis H , these findings provide further empirical

support for that research hypothesis. Because the next

137

two concepts of risk exposure are closely related, the
results pertaining to both measures are discussed concur—
rently in the next section.

The average response for the combined multinational
and domestic corporate samples was in accordance with
g priori expectations. The domestic contingent was the
higher of the two. Results from the MPW U tests associated
with the comparison of the two groups disclosed that the
differences between them were significant at the .021
level for market risk and at the .047 level for residual
risk. Both findings provide confirming evidence for an
Opening conclusion that the common stocks of multinational
corporations have less market risk and less residual risk
than do the common stocks of their domestic counterparts.
For market risk, three items were of particular interest
in the individual industry results that are Shown in Table
7. First, in seven out of the nine industries the differ-
ences were in the predicted direction with the domestic
group being the higher of the two. Second, in four of
these industries not only were the differences in the pre-
dicted direction, but also these differences were signifi-
cant at the .10 level or less. The remaining three indus-
tries were still in the predicted direction: but the dif-

ferences were considerably smaller: prObability levels

138

ranged from .147 to .201. Finally, there were two indus-
tries where the multinational group, contrary to g_priori
expectations, was higher than the domestic group. These
included the Food and Beverage industry along with the
Industrial Equipment industry. This latter industry had
converse results on most of the previous empirical measures
so the existence of Similar results here was not totally
unexpected. A review of the distribution of corporations
within the various quartiles of the combined Food and
Beverage industry revealed that both groups were about
equally represented throughout the range. A further
review of the multinational group disclosed no Single
factor or Special Situation that would account for the
multinational group ranking above its domestic counterpart
on this risk measure. Any conclusion drawn from these
results, however, necessarily had to be tempered by the
lack of Significance. Thus, the conclusion was that the
difference between the two could easily have been a
chance occurrence.

In summary, the results from the combined multi-
national and domestic group presented a strong position

for the rejection Of statistical hypothesis ch

and the
acceptance of the alternative Hie. The individual

industry results were categorized as follows: a) four

139

industries provided moderate support for rejecting the

null hypothesis H3C

0 , b) two industries provided very

limited support for the rejection of ch, c) two indus-
tries really added nothing to the position that H3O Should
be rejected, finally d) the Single remaining industry
where the results were Opposite from the other industries,
and marginally significant, had several qualifications,
and therefore it did not present a strong case for
accepting ch. When the above findings were considered
jointly, the conclusion was that the market risk on the
common stocks of multinational corporations was less than
it was for the common stocks of Similar domestic corpo-
rations. Because Hic represents an Operationalized state-
ment of research hypothesis H3, these findings provide
empirical support for that research hypothesis.

While the results from the combined sample sug-
gest that the common stocks of multinationals had less
residual risk than did their domestic counterparts, there
was only weak and very mixed support for this conclusion
from the individual industry results shown in Table 7.

In eight out of the nine industries the difference was in
the predicted direction, but there was a distinct lack of

Significance in these differences. At best, three of the

industries could be categorized as providing very limited,

140
if any, support for the initial conclusion. The results
for the remaining five industries, where the differences
were in the predicted direction, were interpreted as
adding nothing to the initial conclusion. Given the high
prObability levels in these five industries, it was diffi-
cult to reject the possibility that these differences
could have occurred by chance. A similar circumstance
prevailed in the Industrial Equipment industry where the
results were Opposite from those given above: the high
probability level for that industry suggested the distinct
possibility that the difference was a chance occurrence.
Thus these results were considered to be only a weak
rejection of the initial conclusion of this section. A
review of the individual corporate responses within these
eight industries whose differences were in the predicted
direction but were of limited size, revealed several
general characteristics. First, within each industry the
individual corporate scores tended to be closely grouped
about the overall mean response for that industry.
Further, the spread between the respective scores within
the middle two quartiles of each industry was especially
small: Often times an increase or decrease of I% to 3%
would have shifted the relative rankings considerably.

This suggests that residual risk, at least residual risk

141

as it was quantified for this study, did not clearly dis-
criminate between the two groups, especially when dealing
with the small samples that made up the individual industry
categories. Whether this stemmed from the particular sur-
rogate measure not adequately capturing the difference, if
one existed, was not established. Certainly, the lack of
strong confirming evidence from the individual industry
categories did little to enhance the case for the initial
conclusion that the common stocks of multinationals had
less residual risk than did the common stocks of domestic
firms.

Since the result for the combined multinational
and domestic groups on the residual risk measure was sig-

nificant at less than the specified .05 level, statistical

hypothesis Hgd was rejected and the alternative hypothesis
3d . . . 3d . .
Hl was accepted. But this rejection of HO was qualified

by the generally limited results that characterized the
comparisons by individual industry category. While these
findings were not in conflict with those from the combined
samples, they offered only very limited support for the

. . 3d . . . 3d
rejection of H0 in favor of its alternative H1 .

such, these findings offered some, albeit rather qualified,

As

empirical support for research hypothesis H3. The conclu-

sion from these results was that the common stocks of

142
multinational corporations tend to have less exposure to
residual risk than do the common stocks of domestic cor-
porations.

In each of the four cases, whether risk exposure
was defined in terms of overall variability in returns,
downside risk, market risk, or residual risk, the decision
in all cases was that the null statistical hypothesis
should be rejected and its alternative should be accepted.
The results from the four MAW U tests associated with the
combined multinational and domestic samples strongly sup-
ported this decision. Likewise, the results from the
individual industry categories generally supported this
same decision: moderate support in the cases of overall
variability, downside risk, and market risk and weak sup—
port in the case of residual risk. There were, however,
certain qualifications and these were discussed above.
While the four sets of statistical tests were by no means
independent empirical tests of research hypothesis H3,
each of their outcomes provide some measure of confirming
evidence for that hypothesis. By inference the overall
conclusion was: when measured over a long term time
interval, the risk exposure on the common stocks of multi-
national corporations is less than the risk exposure on

the common stocks of domestic corporations.

143

The last of the risk exposure measures, major
financial loss, was not Operationalized into a separate
set of statistical hypotheses. Rather, this measure was
expressly intended to highlight a single aspect of risk
exposure: the possibility of a major financial loss.

The purpose being to single out those firms whose common
stock had drOpped sharply in price during one or more of
the quarters under review. The thought here being that
if an investor should have to liquidate his holdings
during this depressed period, he would do so at a large
financial loss. As in the case with the previous share-
holder return and risk measures, one of the central ques-
tions was whether the common stocks of multinational firms
were more prone to this type of loss than were the common
stocks of domestic firms. A second but equally interesting
question concerned the individual corporations within
those industries where the multinational group displayed
greater exposure to this type of risk. Specifically, the
question was whether the quarter or quarters in which the
multinational's common stock suffered this sharp price
retreat generally coincided with a similar pattern in the
overall stock market indices or whether the timing Of
these declines was unique to each multinational firm.

Last, if these price declines did not parallel the overall

144
market action, was there a systematic pattern among multi-
national firms?

Several points of explanation need to be made
about the empirical measure, relative skewness, used in
this part of the study. First, a large negative score,
suggesting one or more quarters where the price relative
was substantially less than 1.00, generally denoted a firm
whose common stock, or more especially its shareholders,
were subjected to the possibility of a major financial
loss. On the other hand, when the score ranged from
Slightly negative to substantially positive this was
indicative of a firm whose common stock had only limited
exposure to the possibility of a major financial loss.
Second, because a relative measure was felt to be more
meaningful, each firm's absolute skewness score was divided
by its standard deviation cubed. But the use of the rela-
tive measure was not without a cost. In particular, it
tended to rank the domestic firms lower than the multi-
nationals. Not necessarily because they had less abso-
lute skewness but because their respective individual
standard deviations were larger; this tended to give a
downward bias to their relative skewness scores. Despite
its shortcomings, use of the relative measure appeared

to be warranted in this application. Interpretation of

145

the results for this section, however, required explicit
consideration for possible biases caused by the use of
the relative measure.

For the overall results of the two groups, the
domestic group with its average score of -.17 was consider-
ably less negative than the multinational contingent at
-.34. There was, however, a lack of Significance in the
differences between the two groups. But this overall
result has certain reservations attached. First, while
they are not Shown in the tabled results, the average
standard deviation score for the combined domestic group
was substantially greater than the average for the multi-
national group. As a consequence, this injected a down-
ward bias in the relative skewness scores for the domestic
firms which partially explains why this group's average
was less than the multinationals' average. Furthermore,
the differences between the two groups, as shown by the
various U statistics in Table 7, was not consistent across
all nine industry categories. The results from the four
industries with probability levels of .40 or above were
interpreted as indicating no difference between the two
groups since there was a distinct possibility that these
differences were a chance occurrence. These included Food

and Beverage at .447, Basic Chemicals at .458, Industrial

146
Equipment at .409, and Scientific, Photographic and
Optical Goods at .482. The previously mentioned downward
bias could well be the reason the domestic group held a
slight margin over their multinational counterparts.
Because the differences within the five remaining indus-
tries were marginally significant, with probability levels
ranging from .052 to .100, these were the areas where this
analysis was concentrated.

The differences within the Office and Computer
Equipment classification were the most significant of all
the industry categories. Within that industry, multi-
national firms occupied the tOp five positions in terms
of large negative scores. But three of these firms had
large negative scores not because their absolute skewness
was large, but because the dispersion in their returns
was limited. This caused a definite upward bias in their
skewness scores relative to those of the other firms in
this industry. These three cases clearly pointed up the
problems and weaknesses of using this relative measure.
On the remaining two firms, however, the large scores in
both cases were due to Substantial price declines in a
single period: the second quarter of 1970. This exactly
coincided with a substantial downturn in the total stock

market.

147

The industry with the second most Significant
difference between groups was Machine Tools. The tOp
three positions in the ranking by negative score were
occupied by multinational firms. The large negative score
for the top ranking firm clearly stemmed from the large
price decline in that firm's common stock during the
second quarter of 1970. As previously noted, this
paralleled a similar action in the total stock market.
The probable causes of the large negative score in the
other two multinationals were twofold. First, both firms
experienced large price declines during the second quar-
ter of 1970. These declines were especially significant
when incorporated into a relative measure as was done
here. But when considered on an absolute basis, the
degree of skewness in the respective returns for these
two firms, and therefore the exposure of their investors
to major financial loss, was probably no greater than
many of the lower ranked domestic firms.

Multinational firms had the largest, second
largest and fourth largest negative scores in the third
industry: Electrical Equipment and Supplies. All three
firms were so ranked because their absolute skewness was
large. Thus an upward bias in the relative skewness mea-

sure, because the firms had only limited dispersion in

148

their returns, was not a significant factor in this cir-
cumstance. The apparent cause of the large negative skew-
ness for all three firms was a major price decline during
the second quarter of 1970. This coincided with a similar
turn in the total stock market.

Within the fourth industry, Soaps and Cosmetics,
the large negative scores, and therefore high relative
rankings, for three multinational firms appeared to be the
principal cause for that group's average score being sig-
nificantly more negative than the domestic group's score.
For the most part, large absolute skewness scores’not
limited dispersion in the returns, were the reason these
multinationals ranked as they did. Once more this skew-
ness appeared to stem from a Single period: the second
quarter of 1970 in all three cases. Again, the decrease
paralleled a major market turn in timing and direction.

The fifth and final industry, Drug and Pharma—
ceutical, was unique on several counts. First, this was
the only industry out Of the five under review'where the
multinational group's average skewness was less negative
than was the domestic group's average. And at the same
time, the difference between them was marginally Signifi-
cant at the .10 level. Second, this was prObably the

industry where the domestic group's average was least

149

affected by the previously noted downward bias. While the
average dispersion in the quarterly returns for the two
groups of firms clearly and significantly singled out the
multinationals as being lower, the Spread between the
various firms was narrow enough so that negative multi-
national risk scores were not unduly biased upward relative
to their domestic counterpart. Three domestic firms seemed
to be the principal contributing factor behind the domestic
group's large negative average. In all cases, a Single
but very Substantial decline in market price led to the
large negative skewness scores for all three firms.
Furthermore, all three firms experienced this decline in
the second quarter of 1970.

Several general Observations and conclusions
appear to be in order. First, the multinational firms
that were singled out for review did not demonstrate a
pattern of price reversals that was totally unrelated or
unique to those of the general stock market. This held
not only for their worst quarters, as the preceding analy-
sis has shown: but in general their pattern of price
changes roughly tracked those of the overall market. Nor
were they subject to a larger number of price reversals--
where the quarterly price relative was less than 1.0--

than either the general market index Or the domestic firms

150

within their respective industries. The overall conclu-
sion from this analysis was that in general the common
stocks Of multinational firms do not expose their investors
to a major financial loss to any greater extent than do the
common stocks of comparable domestic firms. But several
caveats must be added. Recognize that there were some
problems in the use, especially for comparative purposes,
of the relative empirical measure that was chosen. Also,
this was a limited multinational sample in that some manu-
facturing and all non-manufacturing industries were totally
excluded. And these could be the very industries where

the above conclusion does not hold. Last, a computation
period shorter than the three-month quarter that was used
in this study might better capture the true extent of an
investor's exposure to a large financial loss.

Difference in Risk Exposure -
Short Term Time Horizon

Recall that research hypothesis H4 had basically
two major points. First, that the risk exposure on the
common stocks of multinational corporations would be less
than it is on the common stocks Of domestic firms. But
even more important this spread between the two risk expo—

sure levels would change as the domestic economy entered

different phases of the typical economic cycle. In

151

particular, it was hypothesized that this spread would be
greater during a period of recession than it would be
during either a period of rapid growth or a period of
limited growth in the domestic economy. The research
hypothesis was not Operationalized into a set of statis-
tical hypotheses in this section because it was not
possible, using the univariate statistical techniques of
this chapter, to specify a direct test of this hypothesis.
But the outcomes from several univariate statistical tests
provide some insight as to the probable relationship be-
tween the different phrases of the typical economic cycle.

As the first step in this supplemental analysis,
three risk measures were chosen from the five that were
available. For each common stock within the combined
multinational and domestic samples separate scores were
computed on each of these measures for the three Specified
sub-periods of this study. Next, a MAW U test was per-
formed On these measures for each of the three different
periods: a total of nine tests. The outcomes from this
series of tests are shown in Figure 1. In all cases the
risk exposure for the multinational contingent was less
than its domestic counterpart. But the important thing
to notice in Figure 1 is the change in the significance

level: eSpecially When the economy enters a period of

One-Tailed Test

Level of Significance:

152

FIGURE 1

SIGNIFICANCE LEVEL FOR THE MANN-WHITNEY U TESTS ASSOCIATED
WITH THE DIFFERENCES IN RISK EXPOSURE BETWEEN THE
COMIDN STOCKS OF THE COMBINED MULTINATIONAL
SAMPLE AND THOSE OF THE DOMESTIC SAMPLE

w: .-a .... ..

.34.
.32'
.30 \
.28 ‘

.24
.22
.20
.18
.161 ‘

 

.12
.10
.08
.06
.04
.02 ‘\

1963-65 1966-68 1969-71

Sub-Period

 

 

Coeff. of Concentration
Medified Quadratic Mean --------
Beta Coefficient .........

153
contraction as was the case for the 1969-1971 sub-period.
For the first sub-period, 1963-1965, the Significance
levels ranged from .164 to .291, but by the third period,
1969-1971, this range had narrowed to a low of .01 and a
high of .029. Clearly, the differences between the two
groups had become much more pronounced during this later
period. Although the complete details are not presented
here, a similar conclusion was born out by the change in
the risk differential, that is the actual spread between
the two groups. When the multinational group's average
score on the three risk measures was compared with the
average for the domestic group the difference or Spread
between the two changed with time. In particular, the Size
of this Spread increased anywhere from 50% to 300% between
the first and third sub-period depending upon the particular
risk measure that was being examined.

What are the implications of these results? Quite
clearly they seem to suggest that the risk exposure dif-
ferential between the common stocks of multinationals and
the common stocks of domestic firms changed as the domestic
economy went through the typical economic cycle. Further,
they imply that this differential becomes more pronounced
as the domestic economy enters a recession period. But a

direct test of this research hypothesis awaits the second

154

section of Chapter V where the results from the repeated

measures model are presented.

Summary

This chapter had four major sections: one corre-
sponding to each of the four research hypotheses that were
first presented in Chapter II. For the first three
hypotheses, the Opening part of their sections presented
the set of testable statistical hypotheses that were
develOped from the original research hypothesis.

Section one presented the results from the compari-
son of the price relatives, and therefore rates of return,
between the common stocks of domestic corporations and
those for multinational corporations during a period of
economic recession in the domestic economy. The finding
from the combined corporate samples was that the multi-
national group was the higher of the two: this was in
accord with g priori expectations. Furthermore, the dif-
ferences between the two groups were significant at the
.006 level. A similar conclusion was drawn from the
results by individual industry category but with a number
of qualifications. The more significant of these qualifi-
cations and contradictions were analyzed further and

details of that analysis were presented in the section.

155

After considering all aspects of the analysis, it was
concluded that the null hypothesis should be rejected. By
inference then, the rate of return on the common stocks of
multinationals, given the prescribed contraction period in
the domestic economy, is greater than those for the common
stocks of domestic corporations.

The second section also presented the results from
a comparison of the price relatives between the two groups,
but this during a period of rapid growth in the domestic
economy and during a period when the domestic economy was
experiencing near-zero expansion. The finding from the
rapid expansion period was that the multinational group
was the higher of the two. But the differences between it
and the domestic group were not significant at the .05
level. For the second comparison period, when there was
near-zero growth in the economy, the domestic group was
the higher of the two: but the difference was not signifi-
cant at the .05 level. §_priori expectations were that
there would be essentially no difference between the two
groups: those findings paralleled that prediction. While
the results by individual industry were for the most part
in accord with the overall results, several industries
required further analysis of their results. The findings

from each of these separate industry analyses, when

156

considered jointly, clearly supported the decision not to
reject the null hypothesis. Since the null hypothesis was
not rejected, it was concluded that the rate of return on
the common stocks of multinational corporations, given the
prescribed periods in the domestic economy, do not differ
from those on the common stocks of domestic corporations.

The third section reported the findings from the
statistical tests using the four surrogate risk measures
associated with the third research hypothesis. For the
combined sample the domestic group was found to have the
higher average score for each of the four measures, and
therefore had the greater risk exposure of the two groups.
Significance levels associated with these four tests
ranged from .014 to .047. Likewise, the results from the
comparison by individual industry category generally showed
the domestic group to be the higher of the two groups. But
the significance levels associated with these tests varied
considerably depending on the particular risk surrogate
used and the particular industry category. Specifically,
it was found that the significance levels associated with
the residual risk measure were uniformly low. This led to
a conclusion that differences between the two groups using
the separate industry categories were not large for this

risk measure. Those industries where there was a distinct

157
lack of significance were analyzed further and the find—
ings of these investigations were given. Based on the
results from the combined samples, the results from the
individual industry categories, and the findings from the
above analysis of individual industry results, all four
null hypotheses were rejected. By inference it was con-
cluded that the risk exposure, whether that be defined in
terms of overall variability, downside risk, market risk
or residual risk, on the common stocks of multinational
corporations, is less than the risk exposure on the common
stocks of domestic corporations, when that comparison is
made over a long term time horizon.

Next, based on findings from further supplemental
analysis, it was concluded that the differential between
the risk exposure on multinational corporations versus the
risk exposure on domestic corporations was not constant
over time. Rather, it appeared to be largest when the
domestic economy was in a recessional period: and con-
versely, given a period of rapid expansion in the domestic

economy, the differential narrowed considerably.

CHAPTER V

RESULTS OF FINAL STATISTICAL TESTS

Introduction

 

This chapter continues with the presentation of
the findings of this study, but attention is now directed
to the results from the series of multivariate statistical
tests that were performed using the data from this study.
In particular, the first half of the chapter presents the
results from the two-way multivariate ANOVA model while
the last part of the chapter contains the results from
the multivariate ANOVA of repeated measures model. Both
of these models were discussed in Chapter III. Instead
of using the "initial" corporate sample as given in Table
2, this portion of the study used the "final" corporate
sample shown in Table 3. The "final" sample subdivided
the Scientific, Photographic, and Optical Goods industry
into two separate groups. And for three of the remaining
industries certain corporations were removed from the
initial sample: this was designed to improve homogeneity
within these industries. Exact details behind the changes

158

159

are found in Chapter III.

The principal purpose of the univariate statistical
models of Chapter IV was to establish whether or not there
was a significant difference between the multinational and
domestic corporate groups on one or more of the six depend-
ent response measures. When there was more than a single
measure connected with a particular set of hypotheses, as
was the case for risk exposure, the separate univariate
tests were performed ignoring the ramifications of possible
correlation between the different response measures. By
switching to a multivariate technique it was possible to
continue to test for differences between the groups: but
now by using all five risk surrogate measures simultane-
ously. Even more important, by shifting to a multivariate
technique it was now possible to investigate which of the
response measures appeared to have the most substantial
between-group differences. That is, one could attempt to
identify which of the response measures appeared to
account for the majority of any Observed difference.

Since the difference in the rates of return between
the domestic and multinational groups was described by a
single response measure, extension of the multivariate
technique to these hypotheses involving only rate of

return was not required. Had this been done, the ensuing

160

multivariate results would have been nearly identical to
the univariate results presented in Chapter IV except for
possible minor deviations imparted by the respecification
of the corporate sample. Consequently, the results pre-
sented in this chapter concentrate on the risk exposure
portions of the study where five different response mea-
sures were used to describe the various aspects of risk
exposure. In particular, the results pertaining to
research hypothesis H3 are reported in the first half of
this chapter, while the second half concentrates on the

findings associated with research hypothesis H4.

Difference in Risk Exposure Over a
Long Term Time Horizon

Recall that research hypothesis H3 stated that
when measured over a long term time interval, the risk
exposure on the common stocks of multinational corporations
would be less than it was for the common stocks of domes-
tic corporations. To capture selected aspects of risk
exposure, the five surrogate risk measures presented in
Chapter III were used. The research hypothesis was
Operationalized by specifying these five measures as the
dependent response measures for the two-way multivariate
model that was presented in Chapter III. The model that

was employed had two main-class treatment effects. The

161

first and most importantA—Type of CorporationA-had two
levels: multinational corporations within a particular
industry as the firSt level and domestic corporations from
the same industry as the second. The second main—class
treatment effectA-Industry Category-—had ten levels: one
corresponding to each of the different industry cate-
gories. Because the research design had intentionally
provided for within cell replication, the possibility of
an interaction between main—class effects was also in-
vestigated.

By order of their importance, the multivariate

'null hypotheses were:

H3e:

0 The mean level scores on the dependent

measures for the common stocks of the
multinational group will not differ
from the mean level scores for the
common stocks of the domestic group,
when that comparison encompasses a
long term time interval.

H : The two main-class treatment effects--
Industry Category and Type of Corpora-
tion (multinational or domestic)--will
not interact in such a way so as to
cause significantly different mean
scores within certain levels solely
because of the particular combination
of factors within a cell.

While it was not stated, there could be a third
null hypothesis added to the above two: this to test

whether the second main-class treatment effect was present.

162
Since there was a strong 3 priori expectation that the
industry categories would have different mean level re-
sponses, and further because the existence of such differ-
ences was not of central interest to the study, this third
hypothesis was not explicitly stated.

In implementing the multivariate model for this
study the single rate of return measure, quarterly price
relative, was included as a sixth dependent response mea-
sure to the multivariate model. The rationale for doing
this was twofold. First, when the analysis had progressed
from a search for the existence of a significant differ-
ence between the groups to one of identifying the measure
or measures where the Observed difference was centered,
it was felt that inclusion of the rate of return measure
would improve the precision of that identification search.
That is, by conditioning on this rate of return measure
the remaining variability would be uniquely due to differ-
ences in risk exposure between the two corporate groups.

A secondary question for this study was whether there was
a difference in the rate of return between the multi—
national and domestic groups over a long term time inter-
val. Including the rate Of return measure in the multi-
variate model provided a test of this point.

The results from the two-way multivariate analysis

163
of variance are given in Table 8. The results from the

test of hypothesis H3f

0 , Shown at the bottom of Table 8,

clearly indicate that H3f

0 should not be rejected.

TABLE 8

RESULTS OF THE TWO-WAY MULTIVARIATE ANALYSIS OF VARIANCE FOR
RESEARCH HYPOTHESIS H3: DIFFERENCE IN RISK EXPOSURE OVER
A LONG TERM TIME HORIZON

 

Probability of
Source of Degrees of This or a Larger
Variation Freedom F Statistic F Value

 

Industry Category
Effect 9 2.362 .0001

Type of Corporation
(Multinational or
Domestic) Effect 1 2.327 .037

Industry by Type
of Corporation
Interaction 9 .731 .924

 

Therefore, interaction between the two main—class treat-
ment effects was not a significant factor and the fit of
the main-class model was adequate. The results from the
test for the main-class effect of major interest—-Type of
Corporation--indicated that the mean level response for
the two treatments was significantly different. That is,
for one or more of the dependent measures there was a sig-
nificant difference between the mean score for the multi-
national group versus the mean score for the domestic

group. Since the Observed difference was Significant at

164
the .037 level, the null hypothesis H39 was rejected.
This indicated there was a difference in the mean level
response between the two groups. Having established the
existence of a difference between the two groups, further
analysis was required to establish the direction of that
difference and to identify the particular response measure
or measures where this difference was most evident. Be-
fore the results from that analysis are presented, however,
the results Shown at the tOp of Table 8 are discussed
briefly. These are for the second main-class effect:
differences among the ten individual industry results.
It was predicted that there would be substantial differ—
ences among the mean level responses for the ten industry
categories: the highly significant finding shown here
clearly substantiates that expectation.

By determining the least squares estimates of the
mean response for each of the two corporate groups, using
the results from the two-way multivariate model, it was
possible to Obtain an indication of the direction and
magnitude of the individual differences on the six response
measures. The estimates from this procedure are Shown in
Table 9. Before discussing the results, however, note that
each of the different response measures has been given a

shortened phonetic label. This was done to shorten and

165
simplify the presentation of results.

The second column of Table 9 presents the differ-
ence between the estimated mean response for the multi-
national group minus the estimated response for the
domestic group, this for each of the six response measures.
The positive difference on the rate of return measure,
GMEAN, indicates the multinational group had the higher
returns of the two. Likewise, the fact that the next four
measures have negative differences indicates the multi-
national group was lower, and therefore had less risk
exposure than did its domestic counterparts. The last
measure, SKEW, differed from the other surrogate risk mea-
sures in that a large positive value was indicative of
limited risk exposure. The negative difference between
the two groups suggests that the multinational group's
mean response was more negative than its domestic
counterpart, and therefore that it has a higher exposure
to this aspect of risk. The findings for the five risk
surrogate measures were in conformity with the results
from the univariate statistical results presented in
Chapter IV. To Obtain an indication of the significance
of the differences, univariate confidence intervals were
computed at the a - .05 level: these are shown in the last

two columns Of Table 9. While the computation technique

166

 

 

ammo. Namm.I eonH.I SNMm I mmmsaexm e>nueaem
"smog Honesecnm news:

mmoo. mmmo.I «HHo.I onmm I mumanumm one we nonnm pnepsmum
"xmnm Hoopnmem

quo.I mamm.I momH.I <Hmm I usenenmweoo seem
“xmfim nexnmz

mnoo.I Hm~0.I «mao.I zzn<no I smez ennmnpmoo penance:
"swam spamsaoa

Haoo.I omHo.I moao.I oozHo I connmnuomesoo

«0 usenenmweeo o.ns«o
"monouem an hunanbsnnsé Hamne>o

omao. maoo.I «moo. z<mzw
"me>numaem eennm manouneoo

neon: nosed omsoomem ozone onumoaon

 

moold

mas>nensH oesopnmsoo

meanne>nsa

peasanumm I ewooomem
esonu assonuoonufisz
pennanumm

oneness
emooomem

 

ZOHH<mommoo mo mmwh ”Homm&m mmdAUI

mmm<bom Hm<MA

2H4: ho mma<zHHmm

"annex MH<Hm<>HHAD= H<3IO3H mma 20mm mHADmNM

m mam<a

167
was univariate, and therefore ignored possible correlation
among the Six reSponse measures, they did, nevertheless,
point up these measures where the between—group differ-
ences were most significant. In particular the confidence
intervals for the first three risk measures, GINCC, QUADMN
and BETA, did not spam zero which suggests that the dif-
ferences between the two groups were probably centered in
one or more of these measures. On the two remaining risk
measures the confidence intervals included zero, indicating
that the between-group differences on these measures were
not significant at the .05 level. These results paralleled
the findings shown in Table 6. Those results indicated
that the differences between the multinational and domestic
groups were considerably more Significant statistically
for the GINCC measure, the QUADMN measure, and the BETA
measure than they were for either the RESID measure or the
SKEW measure.

When a univariate confidence interval was estimated
for the rate of return measure, GMEAN, it included zero.
Thus, while there appeared to be a difference in the mean
response on this measure between the two groups, that dif-
ference was not statistically Significant. A_priori expec—
tations were that, when compared over a long term time

interval, the efficiency of the securities market would be

168

such that the rate of return on the common stocks of
multinational corporations would not be Significantly
different from those of domestic corporations. Since the
difference between the two groups of this study was not
statistically significant, the results of this study
substantiate the above expectation.

Having gained some preliminary information as to
the location of the differences between the multinational
and domestic groups from the univariate confidence inter—
vals, a discriminate analysis was performed on the two
groups. The standardized coefficients of the discriminate
function are shown in Table 10. Because these coefficients
have been standardized, their relative size provides an
indication of how important a particular response measure
was in discriminating between the two groups: multi-
national corporations and domestic corporations. Quite
surprisingly, the response measure with the largest coef-
ficient, RESID, was not one of the measures the univariate
confidence intervals had singled out as having a large
between-group difference. A partial explanation for this
apparent contradiction was found in the high correlation
between this and three of the other risk measures: GINCC
(.96), QUADMN (.85), and BETA (.63): the amounts Shown in

parentheses are the respective correlation coefficients.

169)

TABLE 10

RESULTS FROM THE TWO—WAY MULTIVARIATE MODEL:
DISCRIMINATE ANALYSIS FOR MAIN—CLASS EFFECT:

TYPE OF CORPORATION

 

Response Measure

Standardized Coefficients
of the Discriminate

 

Function

Rate of Return

GMEAN — .8817
Overall Variability

GINCC -2.6379
Downside Risk

QUADMN 4.1326
Market Risk

BETA 1.4167
Residual Risk

RESID 5.0067
Major Financial LOSS

SKEW -1.3142

 

That is, despite this measure's moderate statistical sig—

nificance, the fact that it was highly correlated to these

other measures resulted in its being an important variable

in discriminating between the two groups. AS would be

expected from the confidence intervals in Table 9, the

coefficients were also Sizeable for the QUADMN, GINCC and

BETA measures.

Although the rate of return coefficient

was not large relative to those for the other response

measures, it did nevertheless suggest that this variable

170

contributed some information to the discriminate function.
Consequently, the variable was retained for the final
model.

The series of six univariate F tests Shown in
Table 11 were examined to Obtain an estimate of the rela—
tive importance of each measure. Because the results for
the RESID measure and the SKEW measure indicated only
limited differences, it was decided that they should be
ranked near the bottom of the variable list in performing
the subsequent step-down F tests. Due to the computa-
tional technique for the step-down procedure, placing
these measures in that position guaranteed they would be
conditioned by all those response measures preceding them
in the list. A somewhat similar justification led to the
placing of the rate of return measure first in the list of
response measures for the step-down F tests. It was felt
that the precision of the step-down F tests for the various
risk measures would be improved by first removing the
variability associated with differences in the rates of
return between the two groups. For the remaining three
response measures, the exact ranking pattern was less
clear cut. First, their respective significance levels
were reasonably comparable. Second, there was no strong

3 priori reason for ordering these measures in a

171

TABLE 11

RESULTS FROM THE TWO-WAY MULTIVARIATE MODEL
UNIVARIATE F TESTS FOR MAIN-CLASS EFFECT:
TYPE OF CORPORATION

 

Probability of
ReSponse This or
Measure F Value Larger F Value

 

Rate of Return

GMEAN 2.11 .148
Overall Variability

GINCC 6.58 .030
Downside Risk

QUADMN 2.47 .032
Market Risk

BETA 4.69 .012

Residual Risk
RESID 4.80 .118

Major Financial Loss
SKEW 2.03 .156

 

particular way. Consequently, Table 12 presents results
from three possible orderings: notice that each of these
orderings has a different risk measure in the crucial
second position of the step-down list. Furthermore, the
selection of risk measures was limited to three: BETA,
GINCC, and QUAMN. The intent being to provide a clear
test for the existence of a main-class effect in each of
these measures. At the same time, it was also possible

to test for a main-class effect in the remaining two risk

172

measures after they had been conditioned on the rate of
return measure and the risk measure that was placed second
in the ordering.

The F values associated with the last four mea-
sures in Order A, Table 12, indicated there was just a
marginal treatment effect in these measures after condi-
tioning on the first two measures. The F value for the
market risk measure, BETA, was significant at the .05
level. AS such, the preliminary conclusion was that there
was a main—class treatment effect-AType of Corporation--
for this risk measure. When the Same market risk measure
was preceded by either the overall variability measure,
GINCC, or the downside risk measure, QUADMN, as was the
case for Order B and Order C respectively, the statistical
Significance of the previously noted main-class effect
declined Substantially: to the .186 level for B and to
the .069 level for C. The results for the SKEW risk mea—
sure, which was ranked last in all of these orderings,
appears to warrant further comment. Although it was not
significant at the .05 level, the observed F statistic
suggests there was a marginal treatment effect for this
risk measure. Furthermore, as the discussion of the
results in Table 9 pointed out, the multinational group

had the higher exposure on this particular aspect of risk.

173

TABLE 12

RESULTS FROM THE TWO-WAY MULTIVARIATE MODEL

STEP-DOWN F TESTS FOR MAIN-CLASS EFFECT:
TYPE OF CORPORATION

 

Probability of

 

Response This or
Measure F Value Larger F Value
Order A
Rate of Return - GMEAN 2.11 .148
Market Risk - BETA 5.96 .016
Residual Risk - RESID 0.01 .901
Downside Risk - QUADMN 1.51 .221
Overall Variability - GINCC 0.64 .425
Major Financial Loss - SKEW 3.46 .065
Order B
Rate of Return - GMEAN 2.11 .148
Overall Variability — GINCC 4.16 .044
Market Risk - BETA 1.77 .186
Residual Risk _ RESID 2.00 .159
Downside Risk — QUADMN .131 .718
Major Financial Loss — SKEW 3.46 .065
Order C
Rate of Return - GMEAN 2.11 .148
Downside Risk - QUADMN 3.03 .084
Market Risk - BETA 3.35 .069
Residual Risk - RESID 1.01 .316
Overall Variability - GINCC .64 .425
Major Financial Loss - SKEW 3.46 .065

 

Since some possible explanations for these converse re-

sults were explored in considerable detail in Chapter IV,

they will not be repeated here except to note that the

final conclusion from that review was that the difference

between the two groups was prObably not as substantial as

these initial results suggested.

But more importantly

these results suggest that this risk measure had a unique

174

element from the other four risk measures. Even after the
results had been conditioned on the preceding five mea-
sures there was still an element of variability that was
unique to this particular risk measure. The implications
and Size of this unique element are explained further in
Appendix B.

A Similar but reversed Situation occurred on the
GINCC measure. When it was ranked third or fourth on the
list so that it had been conditioned on at least one other
risk measure, the statistical significance of the main-
class treatment effect was limited. Yet, when that same
measure was ordered second on the list, as in Order B, the
F value corresponding to that step-down test was Signifi-
cant at the .044 level. That finding suggested there was
a main-class treatment effect for that response measure.

Likewise, when the downside risk measure, QUADMN,
was ordered second, as in Order C, the statistical signifi-
cance of that measure's F value rose sharply. Not, how-
ever, to a point where it was significant at the .05 level.
One possible explanation for its lower significance level
relative to the other two measures was its moderate nega-
tive correlation with the rate of return measure. Thus,
by first conditioning on the return measure, as was done

in Order C, the variability that was common to both GMEAN

175

and QUADMN would have been removed and therefore, the
probability levels associated with Subsequent tests would
be raised. The residual variability that remained, how-
ever, suggested that there was in all likelihood a main—
class treatment effect, but not one that was Significant
at the .05 level. Finding a moderate negative correlation
between the two measures, GMEAN and QUADMN, was not un—
expected. Recall that the QUADMN measure weighted the
negative quarterly price relatives very heavily and the
positive ones at zero. Thus, if a common stock had pri-
marily positive quarterly price relatives, it would tend
to have a large positive score on the GMEAN measure but a
very small score on the QUADMN: conversely, a security
that had a mixed pattern with some negative price relatives
would tend towards a small GMEAN score but a large QUADMN
score. Both extremes would augment the negative correla-
tion between these measures. Given the relationship
between these two variables, it was concluded that there
was a main-class treatment effect for the QUADMN measure
despite its lower Significance level.

The findings from the three orderings of Table 12
suggested there was a Sizeable common element between the
three risk measures: BETA, GINCC and QUADMN. And because

of this, when a risk measure was first conditioned by one

176

of the other two measures, thereby removing part of the
common element, the remaining variability unique to the
particular risk measure was insufficient to conclude a
main—class treatment effect existed. As a digression
from the central issue of this study, the existence and
size of this common element for the three measures is
explored in Appendix B.

Having proceeded from the general results reported
in the Opening section through the more specific results
of the most recent sections, we now attempt to summarize

the findings. First, the F value associated with

3e

hypothesis HO

clearly indicated that that hypothesis
Should be rejected. And the conclusion was that the mean
level responses for the multinational corporate group were
different from similar mean level responses for the
domestic group. Next, based on the outcomes from a series
of step-down F tests, the between-group differences ap-
peared to be concentrated in three of the dependent re-
Sponse measures: market risk, BETA: overall variability,
GINCC: and downside risk, QUADMN. But the results from
the different orderings of the response measures for these
tests were inconclusive in singling out one measure that

somehow "best" captured the between-group risk exposure

differences. Nevertheless, the general conclusion was

177

that there was a main-class treatment effect--Type of
Corporation--for these three measures. Therefore, it was
concluded that the risk exposure for the common stocks of
the multinational corporate sample was less than on the
common stocks of comparable domestic corporations, whether
that risk exposure was described by market risk, overall
variability or downside risk.

In general, the results indicated that the
between-group differences on the three remaining measures--
rate of return, GMEAN: residual risk, RESID: and major
financial loss, SKEW--were at best marginal and not Sig-
nificant at the .05 level. A_priori expectations were
that the mean response on the GMEAN measure would not be
appreciably different between the two groups. The results
for the remaining two risk measures generally paralleled
the findings of the univariate tests in Chapter IV. These
early results indicated that both the residual risk mea-
sure, RESID, and the major financial loss measure, SKEW,
had relatively low Significance levels when compared with
those for the other three risk measures: their general
lack of significance was not unexpected. Consequently,
the findings of this chapter generally coincide with the
earlier results.

Quite clearly, the results from the step-down

178
analysis suggest that the majority of the differences in
the mean response levels between the two groups were con-

centrated in the market, downside, and overall variability

3e

risk measures. Since H0

was an Operationalized statement
of research hypothesis H3, the results from the statistical
tests and the findings from the supplemental analysis of
this section provide additional empirical support for that
hypothesis. Thus, whether risk exposure was described in
terms of market risk or total variability in returns or
downside risk, the common stocks of multinational corpo-
rations appeared to have significantly less risk exposure
than was true for their domestic counterparts. By infer-
ence the overall conclusion is that when measured over a
long term time interval, the risk exposure on the common
stocks of multinational corporations is leSS than the risk
exposure on the common stocks of domestic corporations.

Difference in Risk Exposure for
Short Term Time Intervals

As originally specified, research hypothesis H4
had two principal points. The first was whether the risk
exposure on the common stocks of multinational corpora-

tions would be Significantly different from that of com-

parable domestic corporations. But the question of

greater interest was whether this differential in risk

179

exposure would widen and contract as the domestic economy
progressed through the different phases of the typical
economic cycle. In particular, it was hypothesized that
the multinational group would have the lower risk exposure
of the two when measured over a short term time interval,
and further that the differential separating the two would
increase substantially when the domestic economy entered

a period of economic recession.

Drawing upon the results presented in the first
half of this chapter, the dependent response measures used
for this model were the three risk measures where the dif-
ferences between the two groups appeared to be most pro-
nounced: market risk, BETA; overall variability, GINCC:
and downside risk, QUADMN. These three measures became
the criterion variables for the two-way, multivariate
repeated measures model presented in Chapter III. The
"repeated measurement points" on these response measures
coincided with the three sub-periods used in the study:
1963 through 1965, 1966 through 1968, and 1969 through
1971. Thus, by suitable transformation it was possible
to test for a constant, linear or quadratic trend over
time. This model, like the one used in the first part of
this chapter, had Type of corporation as its first treat-

ment effect with two levels: multinational and domestic.

180
As before, the model's second treatment effect-—Industry
Classification--had ten levels. Since the short supple-
mental analysis section at the end of Chapter IV had sug-
gested the differential between the two groups was likely

to change with time, research hypothesis H4

was operation-
alized into a single null statistical hypothesis; this to
test for a three-way interaction between the three factors:
Type of Corporation x Industry Classification x repeated
measurement points.

H4a There will not be a significant effect

0 from the interaction of the three inde-

pendent factors--Type of Corporation x

Industry Classification x repeated

measurement points. Thus, the mean

level response within a cell will not

be significantly different from the

response within other cells solely due

to the interaction of the factors for

those cells.
Specifically, the hypothesis was intended as a test for
the existence of a possible three-way interaction between
the two treatment effects and the trend over time measure.
Barring the lack of a significant three-way interaction,
one could proceed with the specification of the three null
statistical hypotheses associated with the existence of
possible two-way interactions in the results. But, as the

findings presented in the next section show, this was not

neces sary o

181

The findings from the first section of this chapter
were inconclusive in singling out one of the three risk
measures as being most important. Consequently, all three
of the previously noted risk measures were retained as
dependent response measures for the repeated measures
model. In addition, the rate of return measure, GMEAN,
was also used in the final model. Actual testing of
hypothesis Hga was performed using only one of the risk
measures plus the rate of return measure. It was felt
such a procedure would provide a clearer indication of
possible interaction effects.

The results from the multivariate model for each
of the three different risk measures are presented in
Table 13. Consistently, the results from each of the
three measures suggested that the null statistical hypothe-

4a
sis H
0

should be rejected. The conclusion followed that
there was a significant three—way interaction, and there-
fore the risk differential between the various groups
appeared to change over time. Given the existence of an
interaction, no further hypothesis testing was done for
this model. There was, however, an interesting supple-
mental finding from the results. On the two-way inter-

actions, Type of Corporation by repeated measurement

points, the F value associated with the overall variability

182

TABLE 13

RESULTS FROM THE MULTIVARIATE REPEATED MEASURES
MODEL FOR RESEARCH HYPOTHESIS H4: .
DIFFERENCE IN RISK EXPOSURE OVER

A SHORT TERM TIME HORIZON

 

Source of

Probability of

 

F This or

Variation Statistic Larger F Value
Market Risk - BETA

Industry by Type of Corporation

by Time Interaction 2.009 .0007
Overall Variability - GINCC

Industry by Type of Corporation

by Time Interaction 2.052 .0005
Downside Risk - QUADMN

Industry by Type of Corporation

by Time Interaction 1.625 .0144

 

measure, GINCC, and the downside risk, QUADMN, were con—

siderably more significant than market risk, BETA. While

this was not considered a direct test of the relative

importance of the three risk measures, the interpretation

of these results was that the GINCC and the QUADMN risk

measures were the ones of principal interest.

A further

reason for deemphasizing the BETA measure stemmed from the

special circumstance that arose in estimating coefficients

for a small number of sample corporations during the 1963

to 1965 sub-period. The estimated coefficients for these

select cases were not significantly different from zero.

183

That is, the null hypothesis that these coefficients
equalled zero could not be rejected at the .05 level.
While their numbers were not large, certainly not enough
to seriously bias the overall results, they would distort
the graphical analysis for several of the individual in-
dustries. Consequently, the GINCC and QUADMN measures
were analyzed with only limited use of the BETA measure.

The initial step in the analysis of this inter-
action was to compare the mean score for the combined
multinational sample with the mean score for the compar-
able domestic sample at the three repeated measurement
points. The difference between the two, domestic group
mean minus multinational group mean, is plotted in
Figure 2. Several generalizations and conclusions were
drawn from the results shown in Figure 2. First, because
the difference was positive at all six points, the multi-
national group mean was less than that of the domestic
group. This suggests that the risk exposure on the com-
mon stocks of the combined domestic group was greater
than that for the multinational group. Second, the differ-
ences between the two groups became larger between the
first and third sub-periods. Such a trend suggests that
the risk exposure differential increases when the domestic

economy enters a recession period, as was the case for the

184

FIGURE 2

DIFFERENCE IN MEAN SCORE ON THE OVERALL VARIABILITY RISK MEASURE
AND THE DOWNSIDE RISK MEASURE: ESTIMATED COMBINED DOMESTIC
GROUP SCORE - ESTIMATED COMBINED MULTINATIONAL GROUP SCORE

Difference in Mean Score

(Domestic - Multinational)

.026
.024
.022
.020
.018
.016
.014
.012
.010
.008
.006
.004

.002

 

Overall Variability GINCC
Downside Risk QUADMN --------

 

4 j .1

1963—65 1966-68 1969-71
Sub-Period

185
third sub-period 1969 to 1971. But because the results
from the repeated measures model indicated there was also
an interaction by industry, the next step was an examina-
tion of risk exposure differentials by individual industry
category.

Rather than plotting the interaction relationship
for all ten industries, several representative industries
were chosen to illustrate the general findings for this
analysis. Like the univariate analysis in Chapter IV, the
results by individual industry were less clear cut. No
doubt part of this problem was caused by the small sample
size that separate industry categories dictated; at the
same time, part of this appeared to be due to the lack of
homogeneity within the individual categories. Because of
the mixed results, an attempt was made to analyze these
in some detail. Where explanations are offered they
represent one possible explanation of the underlying
causes. It cannot be stated with certainty, however, that
these are definitely the cause of the Observed results.

Figure 3 presents the results from the comparison
of the mean score of the domestic corporations in the
Electrical Equipment and Supplies Industry with the mean
score for the comparable multinational group. Continuing

the comparative techniques of Figure 2, the multinational

 

 

186

FIGURE 3

DIFFERENCE IN MEAN SCORE ON THE OVERALL VARIABILITY MEASURE AND THE
DOWNSIDE RISK MEASURE FOR THE ELECTRICAL EQUIPMENT AND SUPPLIES
INDUSTRY: DOMESTIC GROUP SCORE - MULTINATIONAL GROUP SCORE

Difference in Mean Score

(Domestic - Multinational)

O 034

.032

.030

.028

.026

.024

.022

.020

.018

.016

.014

.012

.010

.008

.006

.004

.002

0

-.002

 

Overall Variability GINCC
Downside Risk QUADMN -------- /’

 

- f 1966-68 1965-71
1963 65’ Sub-Period

187
group score was deducted from the domestic score. There-
fore, if the difference was positive, this indicated the
domestic group's score was larger; and consequently, its
risk exposure was greater. In general, except for the
1963 to 1965 sub-period on the QUADMN measure, the domestic
group's score was larger. The interpretation attached to
that very small negative difference observed in the 1963
to 1965 period was that both groups probably had nearly
equal risk exposure during that period. What was of even
greater interest, however, was the Sizeable increase in
the risk exposure differential between the 1963 through
1965 sub-period and the 1969 through 1971 sub-period.
Quite clearly the spread between the risk exposure for the
multinational group and that for the domestic group in-
creased during the recessional period of 1969 to 1971.
Five industries out of the remaining nine showed similar
results and trends over time. These included Basic
Chemicals, Drug and Pharmaceutical, Soap and Cosmetics,
Machine Tools, and Photographic and Optical Goods. Two
general characteristics emerged from the analysis of these
six industries. First, for most of the sub—periods the
mean score for the multinational contingent was lower
than that of the domestic group, and therefore its risk

exposure was less. The major exceptions to this

188
generalization occurred during the second sub-period when
several multinational groups were higher. ‘While the mean
score for the multinational group was not significantly
larger than its domestic counterpart, an attempt was made
to identify some possible causes for these Observed re-
sults; the findings are given in the next section. The
second characteristic of this group was that the spread
between the two groups was much larger during the third
sub-period, 1969 through 1971, than it was for the first
sub—period, 1963 through 1965. These findings suggested
that not only was the risk exposure generally greater for
the domestic group, but also the spread between its risk
exposure and that for the multinational group increased
substantially during a recession period.

The change in the risk differential between the
domestic and multinational groups, when going from the
1963 through 1965 sub—period to the 1966 through 1968 sub-
period, revealed two distinct but Opposite trends. For
three of the six industries the risk differential in-
creased between the first and second sub-periods and
again between the second and third sub-periods. Such a
trend was generally in accord with a priori expectations.
But for the three remaining industries this risk differen-

tial contracted between the first and second sub-periods,

189
then it rose sharply between the second and third sub-
periods. Even given the contraction, the differential
still generally favored the multinational group during the
second sub-period. A_priori eXpectations had been that
the risk differential would increase substantially during
the third sub-period, but this second sub-period contrac-
tion was unexpected. Consequently, an attempt was made to
find a plausible explanation for this decrease.

The first of the three industries, Basic Chemicals,
appeared to be going through a general industry-wide reces—
sion during the 1966 to 1968 period. While the general
stock market had four quarters in which returns were nega-
tive, the average for the firms within this industry was
closer to six. Among the sixteen firms within this in-
dustry, the price declines were especially sharp and deep
for four of the multinational firms. Although the multi-
national group's mean score continued to be less than that
for the domestic group, the results from these four firms
were sufficient to cause the between-group risk differen-
tial to narrow. The quarterly returns on the multinational
group had returned to their previous levels by late 1968,
and they remained considerably more stable than the
domestic group during the 1969 to 1971 period. Conse-

quently, the spread for the 1969 to 1971 sub-period not

190
only favored the multinational group, but by a very size—
able margin. While it cannot be stated with certainty,
the above explanation appeared to account for the converse
results for this industry.

Two potential factors emerged from the review of
the results for the second industry, Soap and Cosmetics.
First, both multinational and domestic firms within this
industry displayed much greater volatility in their quar-
terly price relatives during the 1966 to 1968 sub-period.
Price run-ups of from 20 to 30% were not uncommon, and
these were often followed by similar—sized price breaks
several quarters later. Second, when the overall market
turned down during the first quarter of 1968, the common
stocks of soap and cosmetic firms declined by even greater
amounts, especially within the multinational group. The
combined effect of these two factors was that the risk
exposure on both the multinational and domestic group rose
significantly from the first to second sub-period. But
because the multinational group was even more volatile
and suffered a larger decline in 1968 relative to its
domestic counterparts, the spread between the two narrowed.
With the onset of the economic recession in late 1969 this
pattern was reversed. The common stocks of multinational

firms were not only less volatile, but their declines were

191
smaller: consequently, the risk differential between the
groups increased substantially. Again it must be pointed
that there were a number of factors to be considered, but
these two appeared to provide a meaningful eXplanation of
the Observed results.

The results from the last of the three industries,
Photographic and Optical Goods, yielded much more mixed
patterns and trends. In part, this was because there were
only eight companies in the total sample, and in part be-
cause the product lines lacked homogeneity between firms.
The smaller spread between the groups was traced to three
multinational firms. All of them experienced much greater
volatility in their respective price relatives during the
second sub-period. Yet no single cause or group of factors
was identified in this review which might account for this
rise. Even the pattern of quarterly changes in price rela-
tives did not coincide among the firms. Consequently, it
was concluded that many of these changes were caused by
factors that tended to be unique to each of the individual
firms. The narrowing of this risk differential, however,
completely reversed itself during the 1969 to 1971 sub-
period when the domestic firms were by far the more
volatile of the two groups.

The above analysis reemphasized the problem of

192
attempting to deal with small industry groups. In all of
these cases a special circumstance within the industry, or
within particular firms from that industry, was sufficient
to distort the results from a short term time interval.
However, the general conclusion was that none of these
findings represented a strong denial of the two general—
izations drawn from Figure 3. Thus, the conclusion
remained that the multinational corporations generally had
less risk exposure than did their domestic counterparts,
and further, that the differential between the risk expo—
sure of the two groups tended to increase when the domestic
economy was in a period of economic recession.

Two general groupings, each containing two indus-
tries, emerged from the remaining four industries. Figure
4 plots the first of these two categories, and in particular
the results from the Food and Beverage industry. The re-
sults for this industry and the second industry in this
category, Office and Computer Equipment, were very mixed.
So much so that no general consensus could be drawn. The
results for the first sub-period, where the risk differen-
tial was small, were in accord with _a_ priori expectations ;
as was the second sub-period where the risk differential
increased substantially. But the reversal of the previous

upward trend in the risk differential during the third

193

FIGURE 4

DIFFERENCE IN MEAN SCORE ON THE OVERALL VARIABILITY MEASURE AND THE

Difference in Mean Score

(Domestic - Multinational)

DOWNSIDE RISK MEASURE FOR THE FOOD AND BEVERAGE INDUSTRY:

DOMESTIC GROUP SCORE - MULTINATIONAL GROUP SCORE

.034
.032
.030
.028
.026
.024
.022
.020
.018
.016
.014
.012
.010
.008
.006
.004
.002

0

-.002

 

Overall Variability GINCC
Downside Risk QUADMN --------

 

 

1963-65 1966-68 1969-71

Sub—period

194

sub-period was exactly Opposite of‘g priori expectations.

A review of the results by individual corporation
within the Food and Beverage industry pointed up several
possible contributory factors. First, this entire
industry, because of the general nature of its products,
must be considered a defensive one during an economic
recession. Since the domestic recession that began in
late 1969 had only a limited affect on the operating
results of the firms within this industry, neither corpo-
rate group experienced a substantial rise in its risk eXpo-
sure between the second and third sub-periods. In addi—
tion, the relationship between the second and third sub-
periods was further distorted by two domestic firms which,
because of their highly volatile price relatives, had
unusually high second sub-period scores on both risk mea-
sures. While this contributed to the large observed dif-
ferential in the 1966 to 1968 sub-period, their scores
during the 1969 to 1971 sub-period were much more typical
and the large differential disappeared. Last, within the
multinational group there were three firms which experi-
enced unusually large declines in the market price of
their common stocks during the second quarter of 1972.
While this paralleled a similar overall stock market

action, their respective declines were so large relative

195
to the other firms in the industry that this contributed
heavily to the limited spread between the multinational
and domestic groups. To the extent that the previous
industry was characterized by limited volatility in the
price relatives, the second industry, Office and Computer
Products, was the exact opposite. Some of the firms, both
multinational and domestic, demonstrated the expected
pattern of increasing risk exposure from the first to the
third sub-periods. Had the analysis dealt only with
these, in all likelihood the conclusion would have been
that the multinational had less risk exposure, and further,
that its margin over the domestic group increased as the
domestic economy entered an economic recession late in
1969. But within each group there were a number of
atypical firms whose mean response for the two risk mea-
sures did not parallel the previous sub-set. Furthermore,
within these atypical groups no clear pattern emerged for
their changes in risk exposure from one sub-period to the
next. Consequently, about all that could be said about
the individual corporate results for this industry was
that they were so mixed that it could not be determined
whether they supported or denied the initial two-part
conclusion. Quite likely the lack of homogeneity within

the product lines for these firms made any comparisons,

196
especially over the short term, very tenuous.

To summarize, the findings from the preceding two
industry reviews generally suggested that the difference
in risk exposure between the common stocks of multi-
national corporations and those for domestic corporations
were rather limited for the three measurement points of
this section. Consequently, the results from these two
industries were inconclusive in establishing whether the
risk exposure differed between the two groups and whether
it changed over time.

The final group of results contained the Industrial
Equipment industry and the Professional, Scientific and
Controlling industry. Figure 5 presents the results from
the first of these two industries. Two generalizations
were drawn from the results shown in Figure 5. First,
because the multinational group had the higher mean score
on both risk measures, its common stocks appeared to have
greater risk exposure than was the case for its domestic
counterparts. Second, for both measures the spread between
the two corporate groups appeared to widen during the
second sub-period, then contract during the third sub—
period. Clearly, the finding that the domestic group had
less risk exposure was directly Opposite of'a priori expec-

tations that the multinational group would be less. The

Difference in Mean Score

197

FIGURE 5

DIFFERENCE IN MEAN SCORE ON THE OVERALL VARIABILITY MEASURE AND

(Domestic - Multinational)

DOWNSIDE RISK MEASURE FOR THE INDUSTRIAL EQUIPMENT INDUSTRY:

DOMESTIC GROUP SCORE - MULTINATIONAL GROUP SCORE

.002
0
-.002
-.004
-.006
-.008
-.010
-.012
-.014
-.016
-.018
-.020
-.022
-.024
-.026
-.028
-.030
-.032
-.034
-.036
-.O38
-.O40
-.O42
-.O44

 

Overall Variability GINCC
Downside Risk QUADMN --------

 

I
\I
\I
L I ;

 

1963-65 1966-68 1969-71
Sub-Period

198

pattern of change in the differential was less surprising
in that it did contract during the third sub-period. This
would suggest that at least the multinationals were becoming
less risky as the domestic economy entered an economic
recession in late 1969. The appearance of the Industrial
Equipment industry was not totally unexpected since it had
exhibited converse results for several of the earlier
hypotheses. Because this industry was examined in some
depth in those instances, only a limited review was made
here. In general, it was concluded that the firms within
this industry lacked homogeneity of product lines and
further that the industry itself was subject to an economic
cycle that tended to lag the general economy. Whether a
different set of measurement points might have altered
these findings was not investigated. While the results
do tend to refute the original conclusion drawn from the
comparison at Figure 3, their significance was consider-
ably reduced by the prOblems and qualifications that arose
within this industry. These findings were interpreted as
being a moderately strong case for rejection of the first
half of the original two-part conclusion.

The second industry in this group, Professional,
Scientific and Controlling Instruments, was created when

the last industry in the initial sample, as given in

199
Table 2, was split into two separate industries for the
final sample shown in Table 3. While the resulting
industry contained only ten firms, there were still two
three-digit SIC codes represented: one with four firms,
the other with six. The results from the code containing
four firms exhibited the expected differential--the multi-
national group was lower. Also it exhibited the antici-
pated trend in this differential over time--the differ-
ential that favored the multinationals in the first sub—
period became even larger in the third sub-period. Conse-
quently, it was the results for the firms in the second
SIC code that caused the multinational group within this
industry to show higher risk exposure than their domestic
counterparts.

A review of the results by individual firm
revealed two multinational firms that appeared to be the
principal cause of that group's high mean score on both
risk measures. The market price of their common stocks
was extremely volatile during the 1966 to 1968 sub-period;
price rises of from 15 to 50%.were not uncommon and these
were generally followed by similarly sized price declines
one or two quarters later. While the net effect of these
price changes left both firms' average price relative for

the second sub—period near zero, they did, nevertheless,

200

produce large scores on both risk measures. Likewise,
during the third sub-period, the common stock price move-
ments for both of these firms were much larger than those
registered by the general market index. For example,
during the second quarter of 1970 the general market
index declined 20%, whereas these two firms experienced
declines of 60 and 64%” respectively. This pattern of
sharp advances and declines continued throughout the third
sub-period and led to both firms' exhibiting unusually
large risk exposure scores. A brief review of the
pattern of these price changes in relation to the overall
Operations of the two firms disclosed no single group of
factors that might have caused the observed results. Yet,
because this was a small industry, the market performance
of these two atypical firms appeared to be sufficient to
cause the multinational group's average to be higher than
the domestic group. While these results represent a
denial of the first half of the initial two-part conclu-
sion, it would seem that this finding should be qualified
because of the two atypical firms in the multinational
group.

A further finding Of this analysis was that between
the second and third sub-periods the differential between

the two corporate groups narrowed appreciably. Unlike the

201
first situation, this would lend some support, admittedly
very weak empirical evidence, to the second half of the
original two—part conclusion.

Up to this point, the analysis has concentrated
on the overall variability and the downside risk measures;
we now turn to an examination of market risk, BETA. In a
small number of instances during the first sub-period a
few of the BETA coefficients were found to be very small
and of limited significance. When this was the case, the
coefficient involved was set equal to zero. Although the
number was small, it did, nevertheless, distort the com-
parisons between the two corporate groups in several
industries. For that reason, this analysis concentrated
on the second and third sub-periods. In seven out of the
ten industries the multinational group's mean BETA score
was lower than their domestic counterparts during these
two sub-periods. Six industries had substantial differ-
ences between the two groups, while the seventh had only
a small difference. The interpretation of the results
was that the common stocks of multinational corporations
within these six industries had less risk exposure than
did their domestic counterparts. For the remaining three
industries the multinational group's mean score exceeded

that of the domestic group. But because the difference

202
for two of these industries was small, the conclusion was
that the risk exposure was nearly equal for these indus-
tries. It was interesting to note that a similar conclu-
sion was made for the same two industries on the other two
risk measures: GINCC and QUADMN.

The multinational group within the single remaining
industry, Professional, Scientific and Controlling Instru—
ments, not only had greater risk exposure on their common
stocks, but also the difference was quite substantial.
Since this finding was exactly Opposite of‘a priori expec-
tations, a review was made as to possible causes. Two
multinational firms with unusually high risk exposure
appeared to be a major contributing factor. This coupled
with the small size of the industry prObably accounted for
the Observed contrary results. It is interesting to note
that the earlier analysis for the GINCC and QUADMN risk
measures produced a similar conclusion for this industry.

TO summarize, the results from six of the industries
generally supported the first half of the original conclu-
sion: that the common stocks Of multinational corporations
had less risk exposure. The results from three Of the
industries were inconclusive in that they suggested there
was little difference between the two corporate groups.

And last, the results for one industry, albeit with some

203
qualifications, represented a denial Of the original con-
clusion.

The empirical evidence for the BETA measure was
much less useful in assessing whether the risk differential
that separated the two groups changed over time; this was
the second half of the original conclusion. Because, as
previously pointed out, first period comparisons could not
be made in all industries, two Options were Open: first,
use only those industries with undistorted first sub-
period results; or second, use the results for all indus-
tries but only for the second and third sub-periods.
Naither of these alternatives appeared entirely adequate
to draw an overall conclusion. Consequently, the only
interpretation drawn from these results was that the risk
differential did change over time; and for the majority of
individual industries the differentials appeared to widen
considerably when the domestic economy entered a recession
period late in 1969.

This closing section attempts to draw together the
results for the statistical test associated with hypothesis
H4 as well as the findings from the analysis Of the com—
bined sample groups and the individual industry categories.

Clearly, the results from the F test associated with

4a

hypothesis Ho

indicated that the null hypothesis should

204
be rejected and the alternative hypothesis should be
accepted. As such, the conclusion was that there was a
significant three-way interaction between the two main
class treatment effects and the repeated measurements
points. Drawing upon the results from the first half of
this chapter, the analysis was primarily confined to two
risk surrogates: overall variability, GINCC, and downside
risk, QUADMN. A two-part tentative conclusion was drawn
from the comparison of the combined multinational and
domestic groups. First, since the multinational group's
mean score was substantially smaller for both measures,
it appeared to have less risk exposure than did its
domestic counterpart. Second, this differential in risk
exposure between the two groups widened even further when
the domestic economy entered the recessional period of
1969 to 1971.

The results from six of the ten individual industry
groups supported the above initial conclusion. The results
for the remaining four industries were clustered into two
equal—sized, but distinctly different groups. The results
for the first group were generally inconclusive in estab-
lishing a difference between the two groups. As such,
they presented a limited case for rejecting the initial

conclusion. But this conclusion was qualified by the

205
existence of several special circumstances within each of
the industries. For the last two industries, the multi—
national group had higher scores on both risk measures,
and therefore had greater risk exposure. These results
definitely represented a denial Of the first half of the
original two-part conclusion. Despite the multinational
group's having a higher mean score in all three periods,
the spread between the two groups narrowed significantly
during the recession period of 1969 to 1971. This finding
was interpreted as providing some confirming evidence,
admittedly quite weak, for the contention that, while the
risk exposure on most common stocks will increase during
an economic recession, the increase for domestic firms
within an industry will be substantially greater than it
is for multinational firms from that industry.

Thus the results from the combined samples and
from the individual industries provided reasonably strong
confirming evidence in support of research hypothesis H4.
When these two sources of empirical evidence were con-
sidered concurrently, the conclusion was that the multi-
national group generally had less risk exposure, and
further than when the domestic economy entered a recession
period, as was the case in 1969, the risk differential

between the two groups became much larger. By inference,

206
when measured over a short term time interval, the
risk exposure on the common stocks of multinational cor-
porations is less than that for the common stocks of
domestic corporations. And secondly, the difference in
risk exposure between the two corporate groups will not
only continue to favor the multinational contingent when
the domestic economy enters a recession period, but this

difference will become much more pronounced.

Summary

This chapter had two major sections: one corre-
sponding to research hypothesis H3 and the other related
to research hypothesis H4. The results reported in this
chapter were from the multivariate statistical models that
were presented in Chapter III. As such, the emphasis of
both sections was on differences in risk exposure between
the common stocks of multinational corporations and those
of comparable domestic corporations.

The first section presented the results from the
comparison Of risk exposure levels between the two groups,
when measured over a long term time interval. The finding
from the multivariate analysis of variance was that there
was a significant difference between the two groups and

that it was significant at the .037 level. Next, a series

207

of step-down F tests on the results disclosed that the
majority of the difference appeared to be concentrated in
the overall variability risk measure, GINCC, in the down-
side risk measure, QUADMN, and in the market risk measure,
BETA. The finding from the combined multinational and
domestic samples, using these three risk measures, was
that the multinational group had less risk exposure; this
was in accord with a priori expectations. It was also
pointed out that the interrelationship between the three
risk measures was such that when the common element of
variability among the measures was removed,the remaining
residual difference generally had limited statistical
significance. A further review Of the relationship among
the risk measures is given in Appendix B. After consider-
ing all aspects of this analysis, it was concluded that
the null hypothesis should be rejected. By inference it
was concluded that risk exposure, whether that be defined
in terms of overall variability, downside risk,or market
risk, on the common stocks of multinational corporations
is less than the risk exposure on the common stocks of
domestic corporations, when that comparison is made over
a long term time horizon.

The final section of the chapter also dealt with

differences in risk exposure, but the comparison was now

208
made over a series of short term time intervals. The
finding from the multivariate repeated measures model was
that there was a significant three-way interaction, indi-
cating that the risk differential appeared to change over
time. Further, the extent of this change was different
for the various industries. In particular, it was found
that for six of the ten industries, the common stock of
the multinational group had lower risk exposure than did
their domestic counterparts; this finding was in accord
with a priori expectations. Further, it was found that
the spread separating the two corporate groups widened
substantially as the domestic economy entered a recession
in late 1969; this too was in accord with expectations.
For the two industries where the difference between the
groups was almost nonexistent, the conclusion was that
these represented a limited denial of the original con—
clusion that the multinational group had less risk expo-
sure. For the remaining two industries where the multi-
national group was found to have significantly higher
risk exposure, the interpretation was that these repre-
sented a reasonably strong denial of the previous conclu-
sion. But a further review of the results disclosed
several possible qualifications to this conclusion; the

more important of the qualifications and contradictions

209
were discussed in the chapter. The principal analytical
technique throughout the last section was to graph the
different interactions between the main-class treatment
effects. Consequently, it was not possible to attach
specific statistical confidence levels to the above con-
clusions and interpretations. Nevertheless, after consider—
ing all aspects of this analysis and review, it was con-
cluded that multinational corporations within most indus—
tries, or more specifically their common stocks, are
subject to less risk exposure than those of comparable
domestic corporations; this was found to be true regardless
of whether risk exposure was defined in terms of overall
variability, downside risk,or market risk. Furthermore,
the risk exposure differential was generally found to be
much more pronounced when the domestic economy entered a

recession period.

CHAPTER VI

SUMMARY, CONCLUSIONS AND IMPLICATIONS, AND
SUGGESTIONS FOR FUTURE RESEARCH
Summary of This Study

One of the unresolved issues in corporate finance
continues to be whether a company should employ a port-
folio approach in selecting its capital investment
projects. PrOponents Of this approach maintain that a
corporation can significantly reduce its common stock
shareholders' risk exposure by considering the portfolio
implications when adding a new capital investment project
to their present holdings. That is, by judiciously select-
ing its capital investment projects the firm can reduce
the variability in the returns on its capital investment
projects without necessarily having to accept a signifi-
cantly lower overall expected rate of return on those
investments. To do this, it would concentrate on those
investment projects which show a low correlation between
their expected returns and the returns from the firm's

existing investment projects. PrOblems of actually

210

211
implementing a portfolio approach in the capital invest-
ment decision process not withstanding, the two major
rebuttals put forth by Opponents of the portfolio tech-
nique center around two issues. First, they state that
investors will have develOped their own individual port-
folios by purchasing corporate securities in a number of
unrelated business activities. And for a firm to pursue
a diversification goal by using a portfolio technique in
its own capital investment selection process will not only
involve double counting, but it could undo what the share-
holder has previously accomplished in his own portfolio.
Second, they maintain that an investor can, by apprOpriate
purchases in the marketplace, create his own diversifica-
tion among different industry categories; this by pur-
chasing the common stocks of one or more companies engaged
in each Of the business activities the shareholder con-
siders appropriate to his particular portfolio strategy.
Thus to have the firm pursue a similar diversification
goal represents an unnecessary duplication of what the
individual shareholder could do for himself. Further,
given the diverse preferences Of the total population of
shareholders, this diversification task could probably be
accomplished more efficiently by the individual share—

holder.

212

But neither of these arguments appears to be
directly applicable when a corporation attempts to diver-
sify on an international scale. First, while several
studies1 have shown there are benefits to be gained from
holding an internationally diversified portfolio of com-
mon stocks, it is doubtful that many investors would have
formulated their own international portfolio strategy.
Therefore, the claim Of duplicate portfolio strategies
has limited validity in this circumstance. Second, the
claim that investors can provide their own diversification
by purchasing the common stocks of different corporations
does not appear feasible when the goal is international
diversification. The average investor simply does not
have access to the large volume Of foreign common stock
issues that would be required for a widely diversified
portfolio. Given the limited amount of investment infor-
mation that is generally available for foreign security
issues, it would make his selection process difficult and
not terribly SOphisticated on those few issues that were
available. Consequently, the firm is in a much better
position to pursue an international diversification goal

than is the individual investor.

 

1Grubel, Op. cit., and Levy and Sarnat, Op. cit.

213

By undertaking investment projects outside the
country where it is headquartered, the firm would substan-
tially reduce its dependence on the domestic economy, and
thereby reduce those fluctuations in its returns that are
caused by swings in the economic cycle of the domestic
economy. Further, if these external investments are them-
selves widely dispersed, their returns will have a low
positive correlation with each other. Thus, should the
domestic economy Of one of the countries where an external
investment project is located swing into a recession
period, given reasonably diversified external projects,
this would have only a minimal effect on the returns for
the balance of the firm's investment projects. To the
extent that the various investment projects have low posi-
tive correlations with each other, the corporation's share-
holders should experience less fluctuations in their
returns. As a result, their risk exposure would be reduced.

The principal thrust of this study then, was to
examine the effects on the common stock shareholder when
a corporation diversifies its capital investments on an
international scale. To do this, the study concentrated
on only those corporations which are headquartered in the
United States. While it would have been desirable to

extend the sample to all corporations regardless of where

214
they are headquartered, limitations on available financial
data precluded such an expansion. Despite its being limited
to corporations headquartered in the United States, the two
corporate samples included a broad representation of dif—
ferent industry categories: nine for the initial sample
group and ten in the final group. Generally, within each
of these industries there were more domestic than multi-
national corporations. Consequently, a random selection
procedure was employed in selecting the domestic control
group while exhaustive sampling was generally used for the
multinational test group. As specified, the initial sample
contained one hundred and forty-six corporations, while the
final sample had one hundred and forty.

This study examined the effects of a corporation's
international diversification on the common stock share-
holder's rate of return and on his exposure to risk as
well. It was considered essential that both measures be
examined jointly, this to determine the degree of trade
off between the two. That is, if the common stocks of one
Of the corporate groups offered less exposure to risk,then
an immediate sequel question was what did the shareholder
have to sacrifice in his rate of return, if anything, to
Obtain this lower level of risk. To measure and evaluate

the international diversification effects on these two

215

aspects of shareholder performance, the market performance
of the common stocks of the multinational corporations
within an industry was compared to that for the common
stocks of comparable domestic corporations. To capture
both the rate of return and risk exposure aspects of mar—
ket performance, six empirical response measures were
specified: one for shareholder rate of return and five
as surrogates for risk exposure. All of these response
measures utilized a three-month quarter as their measure-
ment period. Further, the return measure encompassed not
only changes in market price over the period but also any
dividends the shareholder received during that quarter.
Last, the risk surrogates were specified so as to capture
different aspects of the shareholders' risk exposure.
They included total variability in the rates of return,
downside risk in those returns, the segment of risk in
shareholder returns that was attributable to general
market factors, the residual element of risk unique to a
security after removal of the market segment, and the
possibility Of a major financial loss.

The study covered all quarterly time periods
beginning with the first quarter of 1963 and ending with
the final quarter Of 1971: a total of nine years or 36

quarterly periods. The total period, 1963 through 1971,

216
was considered to be characteristic of a long term time
horizon. Next, the total interval was divided into three
sub-periods Of three years each. As specified, these
three sub—periods constituted a series of short term time
intervals, each of which coincided with a particular phase
of the domestic (United States for this study) economic
cycle: 1963 to 1965, a strongly expansive period: 1966
to 1968, a period of limited growth; and 1969 to 1971, a
period of recession.

There were four principal research hypotheses for
this study. Two of these concerned possible differences
in the rates of return between the two corporate groups
over short term time intervals. Specifically, it was
hypothesized that the existence of a significant difference
would be dependent upon the particular phase of the domestic
economic cycle. The remaining two centered on possible
differences in risk exposure between the two groups: the
first over a long term time interval and the second over
' the series of short term sub-periods.

The first research hypothesis stated that, when
the domestic economy is in an economic recession, the
rates of return on the common stocks of multinational cor-
porations will exceed those of comparable domestic corpo-

rations. Research hypothesis two stated that the rates of

217
return on the common stocks of multinational corporations
will not be significantly different from those for their
domestic counterparts if the comparison is made during a
period when the domestic economy is expanding rapidly or
when it is expanding, but at a moderate pace. The third
hypothesis stated that over a long term time horizon the
risk exposure on the common stocks of multinational corpo-
rations will be less than it is on the common stocks of
comparable domestic corporations. Last, the fourth hypothe-
sis addressed two points. First, that during a series of
short term time intervals the risk exposure on the common
stocks of multinational corporations will generally be
less than it is on the common stocks of comparable
domestic corporations. But of greater interest, this dif-
ferential between the two corporate groups will change over
time as the domestic economy goes through the distinct
phases of the typical economic cycle. Specifically, it
was hypothesized that the risk exposure on the common
stocks of domestic corporations would increase substantially
more, during a recessional period in the domestic economy,
than it would for comparable multinational corporations.
Thus, the risk exposure differential between the two
would favor the multinational group by a much larger

margin during a period Of recession.

218
Conclusions and Implications from the Results

There are three major facets to the presentation
of results and conclusions. The first section presents
the findings and implications from the entire nine-year
period encompassed by this study. The results in this
section are, however, limited to those from the combined
multinational and domestic corporate samples. The second
summarizes the findings and implications from the series
of three equal—sized sub—periods within the total nine-
year span. Again, the results are from the combined cor-
porate samples which includes all industry categories.
The findings from the individual industry categories are
presented in the last section together with the conclusions

drawn from these findings.

Findings for the Long Term Time Horizon

The comparison of the combined corporate groups
for the entire nine-year period disclosed that the rate of
return achieved by the common stocks of the multinational
group was slightly higher than that for the domestic
group. But the differences were not large and the F
statistic associated with the comparison of the two groups
was only significant at the .148 level. Consequently, the

conclusion from these results was that there was no

219
significant difference in the rate of return between the
two groups. Such a finding was not totally unexpected
since it was felt that over an extended period it was un-
likely that the entire multinational group would outperform
their domestic counterparts. That is, given a reasonably
efficient market, there would have to be extenuating cir-
cumstances such that individuals would demand a large
return premium to encourage them to hold these multinational
issues. Only under these conditions would multinational
shareholders achieve a significantly higher return. Al-
though there was no formal research hypothesis specified
for possible differences in the rate of return over a long
term time interval, the findings were in accord with
.3 pgiggi expectations.

Initial expectations were that the common stocks
from the multinational group would have less risk exposure
than the domestic group. The results from the univariate
statistical analysis clearly supported this contention.

On the four principal risk surrogate measures, the multi-
national corporate group had the lower risk exposure of
the two. Furthermore, the differences between the two
were quite large; the MSW U test statistics associated
with the four risk measures were statistically significant

at levels between .014 and .047. However, the finding for

220

the fifth risk surrogate, relative skewness, was exactly
opposite from those for the other four risk measures: the
multinational group had higher risk exposure on this mea—
sure. But a further analysis of the results suggested
several qualifications to this initial conclusion. First,
as was detailed in the introductory portion of this chapter,
relative skewness was not interpreted as constituting a
separate and independent risk surrogate. Instead, it was
intended as a supplement to the other four measures.
Second, a relative measure of skewness was specified for
use in this study rather than an absolute form. In the
majority of applications this measure providedan unbiased
response; but in a limited number of cases, primarily con—
nected with domestic corporations, there was a decided
downward bias in their respective response levels.
Finally, within both corporate groups there were several
atypical corporations whose results were measurably dif-
ferent from the other members of their corporate group.
After allowance was made for the outliers that arose from
the above special situations, the final conclusion was
that there was probably little difference between the two
corporate groups on this supplemental risk measure. Be—
cause of this detailed analysis, several further Obser-

vations can be Offered as to the general usefulness of

221
this particular risk surrogate measure. First, it appeared
to be especially adept at singling out those situations
where a corporation's common stock has indeed suffered a
major price reversal. But at the same time, the measure
was much less effective when it came to making comparisons
between large groups of subjects which contain only a few
of these special situations, as was the case in this study.
On several occasions the initial result indicated one group
of corporations had substantially higher risk exposure;
whereas a more detailed review of individual results seemed
to suggest that the original conclusions were not fully
supported after adequate allowance had been made for out-
liers.

The finding from the multivariate statistical
analysis was that the majority of the difference in risk
exposure between the two groups of common stocks centered
on three particular aspects of risk: overall variability
in the returns on the respective common stocks, downside
risk in those returns, and the general market risk in those
returns. The findings from this analysis were inconclu-
sive for singling out one risk surrogate measure, and
therefore a particular aspect of risk, as "best" describing
the differences in risk exposure between the two groups.

Yet an important implication drawn from this finding was

222
that risk exposure on the common stocks of the multi-
national corporate group was lower than that for their
domestic counterparts, regardless of whether it was
described in terms of some overall variability measure,
some downside measure, or by some general market risk
measure. An equally important implication was that the
multinational group's lower risk profile was not limited
to a single aspect of the shareholders' exposure to risk.
Last, it refutes, to some extent, the possibility that
the findings of this study were due solely to the selec—
tion of a particular risk surrogate measure.

The conclusion from the combined findings of the
univariate and multivariate statistical analysis was that
the risk exposure on the common stocks of multinational
corporations was significantly less than it was for com-
parable domestic corporations. Furthermore, this conclu-
sion held regardless of whether risk exposure was described
by an overall variability risk measure, by some downside
risk measure, or by a dichotomized general market risk
measure. Clearly, the results reported above provided
empirical support for the third research hypothesis.

One implication from the previous findings is that
a corporation's diversification on a multinational scale

has only a limited affect on its common stock

223
shareholders' rate of return over the long run. That is,
any additional returns achieved by its shareholders versus
those received by their counterparts in comparable domestic
corporations will not be significant over a long term time
interval. The implications for shareholder risk exposure
were dramatically different. Shareholders in those corpo-
rations which have diversified on a multinational scale
have enjoyed significantly less exposure to risk than has
been the case with their domestic counterparts. This
suggests that a corporation's multinational diversification
while offering the shareholders only a minimal change in
their rate Of return, has nevertheless substantially
lowered their risk exposure. One question that immediately
arises is whether there is a direct relationship between
the level of multinational diversification and the degree
of reduction in shareholder risk exposure. That question
was considered tO be outside the sc0pe of this study so
its answer must wait until some further work is done in
this area. Despite the lack of a precise estimate of the
relationship between the level of diversification and the
degree of risk reduction, the evidence clearly suggests
that a corporation can, by diversifying on a multinational
scale, substantially lower the risk exposure to its share-

holders without sacrificing a perceptible reduction in

224

their rate of return. Further, it appears that this con-
clusion is not limited to a single industry category, but
instead applies to a reasonably broad spectrum of different
manufacturing categories.

Several caveats need to be made at this point.
First, these implications and conclusions are based on the
research findings from the nine-year period, 1963 to 1971.
To the extent that this period is not representative of a
typical long term time interval, then the results for
previous or subsequent time intervals may not parallel the
findings of this research. Second, the corporate sample
used in this study specifically excluded non—manufacturing
types of industries. Thus, the above conclusions and
implications may not be applicable to all industry cate-
gories. One of the suggestions contained in the final
section of this chapter is that a future study be eXpanded

to incorporate some of these excluded industry categories.

Findings for Short Term Time Intervals

The objectives in dividing the total nine-year
time period encompassed by this study into a series of
short term time intervals were twofold. Since shareholder
rate of return and risk exposure on all common stocks,

whether they be multinational or domestic corporations,

225
will be influenced to some degree by the domestic (United
States for this study) economic cycle, much additional
information could be gained not only from making compari-
sons within a particular phase of the cycle, but also from
comparisons between the different phases. For this reason,
specific time intervals within the nine-year study period
were identified as corresponding to a particular phase of
the general economic cycle. By comparing the respective
shareholder rate of return and risk exposure for the
multinational and domestic corporate groups, further con-
clusions were drawn as to differences between the two
groups. At the same time, by examining the size of these
differences for the various sub—periods, it was possible
to draw conclusions as to the existence of trends over
time.

A priggi expectations were that differences in
rates of return and risk exposure between the two groups
would be highly dependent upon the specific phase of the
United States economic cycle: an expansionary period, a
limited growth period, or a contraction period. It was
also expected that the spread, or absolute size of the
differences between the groups, would change over time.
Consequently, it would be possible to identify a definite

trend over time.

226

The nine-year period encompassed by this study
was divided into three equal-sized sub-periods. Each
three-year sub-period corresponded to a particular general
phase of the United States domestic economic cycle. Thus,
the period 1963 through 1965 was identified with strong
expansive growth, the period 1966 through 1968 a time of
limited growth, and the period 1969 through 1971 a time
of substantial contraction in the domestic economy.

Shareholder Rate of Return During Short Term Time

Intervals. A priori expectations were that when the

 

domestic economy was rapidly expanding there would be
little difference between the rate of return achieved by
the common stocks of domestic corporations versus those of
comparable multinational corporations. The finding from
the univariate statistical analysis covering the sub-period
1963 through 1965 confirmed this expectation. While the
average rate of return for the multinational group was
slightly higher, the MAW U statistic associated with the
differences between the two groups was only significant at
the .939 level. Likewise, it was anticipated that during
a period of limited growth in the domestic economy any
differences in the rates of return between the two groups
would be small. This was confirmed by the MHW U statistic

associated with the differences for the sub—period 1966

227

through 1968; it was only significant at the .636 level.
Clearly, the conclusion from both of these findings was
that the returns on the common stocks of multinational
corporations were only slightly different from those for
comparable domestic firms. Results from both of these
tests provided empirical support for research hypothesis H2.

Unlike the previous two circumstances, 2_priori
expectations were that the common stocks of multinational
corporations would achieve substantially higher returns
than those for comparable domestic corporations whenever
there was a definite contraction in the United States
economy. The finding from the comparison of the two
groups during the 1969 through 1971 sub-period was that
the multinational contingent was higher. Furthermore, the
Mew U statistic associated with these differences was sta—
tistically significant at the .006 level. The conclusion
from this result was that during a recession period in the
domestic economy the common stocks of multinational corpo-
rations will substantially out—perform those of comparable
domestic firms. The finding from this comparison provided
empirical support for research hypothesis H1.

One implication from the above findings is that
during periods when the domestic economy is expanding, the

effects of a corporation's previous multinational

228
diversification upon its shareholders' rate of return will
be minimal: neither significantly improving nor diminish-
ing those returns. But given a period of substantial con-
traction in the domestic economy, then the effects of a
corporation's diversification on its shareholders' returns
will definitely be beneficial. As such, its shareholders
will achieve significantly higher returns than their
counterparts in firms that are essentially domestic. This
suggests that the returns from the multinational's foreign
locations tend to cushion the negative aspects of a down-
turn in the domestic economy. Yet, as was brought out in
an earlier part of this section, multinational shareholders
continue to achieve returns comparable to those attained
by domestic shareholders during expansionary domestic
periods. This suggests the firm's foreign Operations do
not present a noticeable drag on the multinational's over-
all results during these periods. Such a finding is con-
sistent with the requirement that a multinational corpo-
ration, for purposes of this study, not only had to derive
a sizeable prOportion of its revenues in external markets,
but the location of these markets had to be widely diversi-
fied as well. However, as was brought out in the preceding
section, over a long term time interval the superior short-

run rate of return performance of multinational corporations

229

does not dominate the longer—term relationship. Thus when
the quarterly rates of return from this series Of short
term sub-periods were combined into a single long term
time interval, the common stocks of multinationals tended
to be higher. But the difference was not statistically
significant at the .05 level. Apparently the host of
other factors and influences that were introduced when
the time period was extended were such that they masked
this short-term superiority of multinational corporations
during recessionary periods in the domestic economy. An
interesting question is whether the shareholder rate of
return superiority of multinational corporations would be
especially pronounced during the major short term down-
turns in the overall stock market that generally precede
a domestic economic recession. Nevertheless, this question
was considered to be outside the domain Of this study. SO
it will have to be answered at a later date.

Shareholder Risk Exposure During Short Term Time

Intervals. There were two principal areas of emphasis in

 

the examination of possible shareholder risk exposure dif-
ferences between the multinational and domestic corporate
groups during the series of short term time intervals.
First, expectations were that the common stocks of multi-

national corporations would generally have lower levels of

230
risk exposure than those for comparable domestic firms,
regardless of whether the comparison was made during a
short term time interval that encompassed: (a) an expan—
sive growth period for the domestic economy, (b) a period
of limited growth in the domestic economy, or (c) a period
of substantial contraction in the domestic economy. It
was not expected, however, that the size of this differen-
tial between the two groups would necessarily be equal
among these three different phases of the domestic
economic cycle. This gave rise to the second area of
interest, mainly, was there a pronounced trend in this
differential? It was anticipated that the risk differen-
tial separating the two groups of common stocks would be
at a minimum when the domestic economy was expanding
rapidly and would be widest when the domestic economy was
in a pronounced contraction period. At the same time, the
third phase, one of limited growth, was expected to fall
somewhere in between these two extremes. Drawing upon the
results from the previous long term analysis, this exam-
ination concentrated on only three of the five available
risk surrogate measures: total variability in returns;
downside risk in those returns, and the general market
risk element in those returns.

The results from the nine univariate Mew tests--

231
one for each risk measure during the three sub-periods—-
were in accord with initial expectations: the common
stocks for the combined multinational corporate sample had
lower risk exposure in each of the three sub-periods than
did their domestic counterparts. The statistical signifi—
cance of the Observed differences was greatest during the
1969 through 1971 sub-period when the three U statistics
were significant at the .010, .013, and .029 levels. The
U statistics associated with the 1966 through 1968 sub-
period were somewhat less significant with levels
ranging from .02 to .09. Finally, the significance levels
for the 1963 through 1965 sub-period were even lower; the
most significant one being .16, while the least significant
was .29. Quite clearly, the findings suggest that while
there appeared to be a positive risk differential favoring
the multinational group in all periods, it was much more
pronounced during a recession period in the domestic
economy. Such a finding was entirely in keeping with
‘3 priori expectations. It was felt that, because a multi-
national corporation has its Operations located throughout
a number of semi-independent national economies, the com-
mon stocks of those corporations would be much less
affected by cyclical swings in the United States domestic

economy. That is, during a substantial contraction period

232
the general market performance of multinational common
stocks would be much less subject to the sharp advances
and declines that characterize the common stocks of
essentially domestic firms. The results of this section
clearly supported the position that the risk exposure on
the common stocks of multinational corporations would be
lower, especially during a contraction period in the
domestic economy.

The results from the multivariate repeated measures
model clearly indicated the existence of a trend in the
risk differential between the three sub—periods. The F
values corresponding to the three risk measures were
statistically significant at the .0005, .0007, and .0144
levels. Given the existence of this trend, a graphic
analysis was employed for the combined corporate group and
for select individual industry categories. For the most
part, the conclusion from each of these suggested that not
only did the common stocks of multinational corporations
generally have less risk exposure, but also the resulting
risk differential between the two groups widened as the
domestic economy progressed from an expansive phase to a
phase of limited growth and finally to a contraction phase.
It was during this latter phase that the differential was

especially pronounced and statistically significant.

233

The immediate implication from these results was
that the potential benefits from a corporation's multi-
national diversification will be greatest when the domestic
economy is experiencing a contraction in its overall eco-
nomic activity. For it was during this period that the
reduction in shareholders' risk exposure was most pro—
nounced. But at the same time, the implication was that
even during a strong domestic expansionary period, the
shareholders' risk exposure in multinational corporations
still appeared to be less than their domestic counter-
parts. Although the differences during this period were
limited. Certainly, there was no indication that a corpo-
ration's multinational diversification necessarily sub-
jected their shareholders to higher levels of risk expo-
sure during periods of rapid eXpansion, or periods of
limited growth in the domestic economy. The final conclu-
sion from this analysis was that the shareholders in multi-
national corporations have definitely benefited through a
lower exposure to risk because of the corporation's deci-
sion to diversify on a multinational scale.

Several caveates need to be offered at this point.
First, the short term time intervals used in this study
covered activity for a three-year period. It is entirely

possible that some adverse results associated with a much

234
shorter time period may have been averaged out by using
the three-year interval. Specifically, it could be that
immediately prior to a major international economic or
political event, and immediately following that event, the
shareholder of a multinational corporation may be subjected
to significantly higher risk exposure than is true for his
counterpart in a domestic corporation. But because the
empirical data for this study covered a three-year period,
the implications from these special situations may have
been masked by the averaging process. Along similar lines,
the use of a quarterly measurement period may have been
too long and not adequately sensitive to capture these
shorter term Special situations. Both of these possibili-
ties could be analyzed in a future study, but this was

considered to be outside the scope of the present study.

Results for the Individual Industry Categories

In addition to the previously reported results
from the combined corporate samples, comparisons were also
made using the individual corporate samples within each of
the separate industry categories. Work with the individual
industry results concentrated on three specific areas:
1) differences in rates of return between multinational

and domestic common stocks during a short term time

235
interval that encompasses a substantial contraction in the
domestic economy, 2) differences in shareholder risk expo—
sure between the multinational and domestic corporate
groups when the comparison period covers a long term time
horizon, and 3) possible differences in shareholder risk
exposure between the two corporate groups during a short
term time interval that includes a recession period in the
domestic economy. Before presenting the actual results,
there are several general explanations and qualifications
which apply to all of the work in this area. Since the
sample of firms within each of these industries was small,
the statistical significance Of the test results was gen-
erally much lower than was true for the combined corporate
samples. Further, because there was only a limited number
of firms, the final results could be substantially altered
if there were one or two atypical firms in either the
multinational or domestic group. Finally, the Industrial
Equipment category suffered from a distinct lack of homo-
geneity among the sample firms which tended to make the
results from this industry difficult to interpret except
on a piecemeal basis. Consequently, it was decided to
exclude this industry; so the results from only eight of
the industries are reported here.

During the 1969 through 1971 sub-period, the rate

236
of return on the common stocks from the multinational
group appeared to be higher than it was for the domestic
group in all eight industry categories. But the prob—
ability levels for the U statistic associated with the
Mew tests in those eight industries varied over a con-
siderable range: five were significant at the .20 level
or less, while the remaining three ranged up to the .35
level. Despite their limited significance, the results
from the first five industries generally supported the
conclusion from the earlier section: that the rate of
return, when measured during a short-term time interval
that encompasses a period of substantial contraction in
the domestic economx,for the common stocks of multinational
corporations will be significantly higher than those of
domestic firms. The implication of this finding was that
for the most part the common stock shareholder has bene-
fited from a corporation's multinational diversification.
In this case, they have achieved higher returns during a
contraction period in the domestic economy. Although
full details will not be presented, it should also be
noted that these same multinational shareholders achieved
returns approximately equivalent to their counterparts in
domestic corporations during short term periods of strong

expansion or periods of limited growth in the domestic

238
sixteen cases generally suggest, like those from the com—
bined sample, that shareholder risk exposure, whether that
be defined in terms of overall variability, downside risk,
or market risk, will be significantly lower for share-
holders in multinational corporations than it is for their
counterparts in domestic corporations. Yet, at the same
time, there appears to be little difference between the
two corporate groups in terms of rates of return over the
same long term time horizon. The combined effects of
these two results implies that in almost all cases the
shareholder has benefited through lower risk exposure; yet
to gain this lower risk eXposure he has had to make no
perceptible sacrifice in his rate of return.

The comparisons for the shorter three-year sub-
period, 1969 through 1971, focused on the same three risk
surrogate measures that were used in the preceding section.
The results from the series of twenty-four MPW tests
indicated that in twenty-two of those cases the
shareholders from the multinational group had lower risk
exposure than did their counterparts in domestic firms.
The statistical significance of the U values associated
with these twenty-two Mew tests were moderate, although
highly dispersed: sixteen were significant at the .15

level or less. One possible explanation for the higher

239
probability levels than those for the long term time
interval could have been the large number of atypical
firms and special situations that characterized this par—
ticular sub-period. Nevertheless, the results from these
sixteen cases appeared to support the earlier conclusion
that the risk exposure on the common stocks of multi-
national corporations was significantly less than it was
for comparable domestic firms during a period of substan-
tial contraction in the domestic economy.

Overall the results from the comparisons by indi-
vidual industry category generally supported the conclu-
sions and implications that were drawn from the combined
samples. Certainly, none of the findings from the indi-
vidual categories would constitute a strong denial of
those earlier results. But the analysis by industry group
did, however, point up the prOblems encountered when the
sample is restricted to a few firms. Unless each firm is
reasonably well matched as to important Operating charac-
teristics,and furthen,unless the period under review is
reasonably free from special events or situations that
effect only one or two of the sample firms, the results
from overall comparisons within the industry group are

likely to be subject to some degree of bias.

240
Suggestions for Future Research

One possibility for future work would be to take
the present study and expand the corporate sample to in-
clude additional SIC industry codes. These should include
not only additional manufacturing types of industries, as
was the central thrust of the present study, but more im—
portantly, it should include non-manufacturing industry
codes as well. Financial intermediaries such as commer—
cial banks and corporations that are engaged in the explor-
ation and develOpment of natural resources are two examples
that come immediately to mind. Certainly, the lack of an
adequate number of multinational or domestic corporations
would pose a problem with some of these industries.
Nevertheless, the preliminary work done in connection with
this study suggested that enough viable industry cate-
gories could be found to permit a substantial expansion of
the present study.

A second possibility would be to place increased
emphasis on matching the multinational and domestic firms
that are selected within a particular industry category.
In particular, the sample corporations should be more
closely aligned as to size of Operations, product lines,
plus other important Operating characteristics. The im—

proved homogeneity within the industry, even if it entailed

241
a small decrease in sample size, should yield, in many
cases, gains in precision that would more than offset loss
Of power due to the decrease in sample size.

As specified for this study, the variable used to
classify the corporation as to type had but two levels;
either the firm was classed a multinational or a domestic
depending upon certain of its Operating characteristics.
One possible extension of this study would be to substi-
tute a continuous variable to measure the level of a cor-
poration's multinational diversification, this in place
of the current discrete one. This would greatly expand
the possibilities for examining the relationships between
the response measures--shareholder returns and risk
exposure--and the degree of multinational diversification.

Last, there is a need to investigate the implica-
tions of altering the interval between measurement points
On the response measures. Potential gains from such a
change could arise in several areas. First, it would
allow a straightforward comparison between the different
measurement intervals to see whether the conclusions from
a particular response measure are reasonably consistent.
Second, by having a shorter interval, it would be possible
to scrutinize the performance of the two groups at or near

major turning points in the domestic economy. And, as was

242
pointed out earlier, by using these altered response mea-
sures it would be possible to analyze more closely the
effects that a major international economic or political

event has on the respective corporate groups.

Summary and Outlook

This research has attempted to provide answers to
several questions concerning the effect Of a corporation's
multinational diversification on the welfare of its common
stock shareholders. The overall finding was that the
shareholders have generally benefited not only in terms
of having a lower exposure to risk, but also because they
have achieved superior returns during the period when the
United States economy was experiencing a contraction in
economic activity. But several observations and qualifica-
tions need to be pointed out about the time period-~1963
through l97l--that was encompassed by this study. Specifi-
cally, these concern two different circumstances: first,
special conditions or situations which existed during the
period spanned by this study, and second, structural
changes and trends that appear to be on the horizon for
multinational corporations in the 1970's and 1980's. It
is essential that these points and their implications be
considered before any attempt is made to extend the find-

ings of this study to future periods.

243

First, it should be pointed out that during the
period of this study most United States—based multinational
corporations experienced extremely rapid growth. Much of
their external growth was achieved through acquisition of
smaller existing EurOpean corporations. But there is some
question as to whether there are an adequate number of
"attractive" acquisition candidates remaining that could
support a continuation of the past rapid multinational
expansion. The continued growth of the European common
market appears to be generating a recent and growing trend
toward consolidation and concentration within individual
industries for each of the member countries. Increasingly,
the trend appears to be for the smaller independent Euro—
pean companies to link up among themselves rather than be
acquired by some large United States-based multinational
firm. At the same time, there is the distinct possibility
that some host governments will be increasingly reluctant
to have a large segment of their domestic industry be
absorbed and placed under the control of some external
multinational corporation. Even more important, there is
increasing evidence that alteration to existing United
States tax regulations may significantly reduce the
profitability of external investments, and therefore sub-

stantially slow multinational expansion by United States

244
corporations. All three of these categories-~the matura-
tion and increased concentration in EurOpean markets, the
possibility of limiting regulation by host governments,
and the possible enactment of changes in United States tax
regulations--suggest a more restrained expansion rate in
multinationals during the 1970's and '80's. The question
to be answered during the next few years will center
around this: Given the reduced rate of expansion, will
the multinational corporation continue to offer benefits
to its shareholders over and above those available from
comparable domestic corporations; this either through
larger returns or lessened risk exposure?

Second, the period covered by the study had several
distinctly different patterns of economic activity in the
various domestic economies. For most EurOpean economies
it was a period of generally strong economic eXpansion
with sharp increases in real economic growth, and for
some, a period of moderate inflationary pressure. On the
other hand, the United States economy, especially during
the late 1960's, was plagued with unusually strong infla-
tionary pressures and recurring periods of economic con-
tractions and adjustments. As we enter the '70's, at
least one-half of this picture seems to be changing: most

EurOpean economies are experiencing much stronger

245
inflationary pressures which may necessitate restrictive
fiscal and monetary policies. In implementing these
policies, there is the possibility that they could lead
to a series of recurring contraction and expansion periods
in the various EurOpean domestic economies. If indeed this
does become the pattern for the 1970's, an immediate ques—
tion is this: How will the foreign Operations of United
States multinational corporations fare in this changing
economic climate? That is, during the late 1960's the
foreign Operations of most United States multinational
corporations tended to be a stabilizing force in their
overall corporate Operation. But if these same foreign
operations are subjected to much wider fluctuations in the
local economic activity during future periods will their
contribution continue to be a stabilizing one? It is
entirely possible that the increased volatility of their
Operating results, when combined with the overall corpo-
rate results, could totally overwhelm the reductions in
risk exposure that were achieved by the multinational
shareholders during the period of this study. Again, the
problems and experiences of the 1970's and 1980's may
necessitate a considerable modification to the observed
findings of the 1960's.

Third, during much of the period covered by this

246

study there was reasonable stability in the international
monetary system. Beginning with 1971, however, and at an
increasing tempo since then, the entire international mone—
tary system has been subject to serious strains and has
undergone considerable forced change and adjustment. As
this is being written (Spring, 1973), flexible exchange
rates are becoming the norm rather than the exception.
Furthermore, there is a definite question as to whether
the stable patterns of the 1960's will again prevail;
indications are that this is not likely to be the case in
the immediate future. Certainly, an immediate question is
how this will affect the Operations of the multinational
corporation and in particular, the welfare of its share-
holders. Not only would this influence future expansion,
but more directly it will affect the ongoing Operations
of existing foreign operations. Their profitability could
be substantially altered by sizeable swings in the pre-
vailing exchange rates. Consequently, any attempt to
extend the findings of this study to future years should
consider possible effects from forthcoming changes in the
international monetary system.

Next, the latter portion of the period covered by
this study was characterized by increasingly stringent

controls over direct external investment for all United

247
States corporations. Currently there appears to be a
definite movement towards dismantling and removing these
controls, possibly by 1974 or 1975. This gives rise to
the question: What will be the effect of their removal
on the external expansion by United States multinational
corporations? In particular, the effects Of this change
will likely be most pronounced for smaller firms. Be-
cause large corporations have minimized the effects Of
the present limitation by increasing their reliance on
local sources of capital, they have not had to seriously
restrict their external expansion. What remains to be
seen is whether, when these external investments are
removed, there will be a sharp increase in multinational
expansion activities by smaller corporations. And if so,
what will be the long run effect of their expanded foreign
Operations on the welfare of their shareholders.

During recent months there has been an increased
emphasis on studying the effects of multinational corpo—
rations On the United States domestic economy. In par—
ticular much of this work has concentrated on estimating
the effects of United States-based multinationals on em-
ployment levels in the domestic economy. From this contro-
versy two groups have emerged with widely divergent posi-

tions. One group claims that domestic-based multinationals

248
have not contributed to the general welfare of most indi-
viduals in United States because their activities have
exported jObs and therefore reduced potential employment.
The other group paints quite the Opposite picture; they
claim that multinationals have increased the level of
employment in the United States, and therefore have made
a positive contribution to aggregate welfare. The key
issue for both arguments centers on the following points.
First, have multinational corporations eXported jobs by
producing their foreign market requirements in one of
their external manufacturing locations rather than relying
on exports from the United States? Second, has domestic
employment been significantly reduced by these corpora—
tions producing products for the domestic market in one
of their foreign locations and reimporting these into the
United States? The immediate conclusion would seem to be
that this type of multinational activity would export jObs.
But there is an important qualifying question: Could the
products that were produced at the foreign locations have
been competitively produced in the United States? While
the answers are far from complete, preliminary data sug-
gests that domestic employment has not been significantly
reduced by the foreign Operations of domestic multi-

nationals. But regardless of the final outcome, the trend

249

for the future would seem to be towards increased control
over domestic multinational corporations. An indication
of this can be found in sections of the Trade Reform Act
of 1973 that is currently (Spring, 1973) before Congress.
Although the final form of the legislation may be con-
siderably altered, the message is unmistakably evident:
the executive branch is requesting broad and flexible
authority to regulate and possibly limit the foreign manu-
facturing Operations of domestic multinationals through
selective taxes on certain segments of their foreign earn-
ings. The implication from this is that the expansion of
the multinational corporation in the 70's is likely to be
slower than it has been during the last decade.

Besides the control measures contained in the Trade
Reform Act, there are additional legislative prOposals
that include more drastic measures for altering the
present tax treatment of foreign earnings. Should these
measures be enacted, the trade-off between the risk and
reward from foreign Operations would be considerably
changed from recent historical experience. And given
these changes, the eXperience of the 60's, which formed
the basis of this study, will have to be interpreted with
caution.

Up until recently, the multinational corporate

250
scene has been dominated by the large EurOpean-based or
United States-based multinational corporations. Not only
have these firms been the most aggressive in expanding
their operations, but also much of that growth has
centered on the markets in EurOpe or those in the United
States. Within recent months, however, Japanese-based
corporations have been rapidly eXpanding and broadening
their sc0pe of EurOpean Operations. While many Japanese
firms continue to rely heavily on exports to service these
markets, increasingly they have established local manu-
facturing Operations to supply these markets. Quite
clearly, the number and size of Japanese multinational
firms is likely to increase during the next decade. To
the extent that EurOpean markets become more competitive,
this could reduce the profitability of United States multi-
national foreign Operations that are located in those mar-
kets. If in the final outcome this substantially reduces
their profitability, then the positive benefits from
multinational diversification that were found in this study
may be considerably altered. ‘At the very time that the
preceding develOpment is gaining momentum, there is a
possible offsetting develOpment in the domestic Japanese
market. Although the prOposals are still in the preliminary

stages, indications are that the highly restrictive

251
controls on foreign ownership of domestic Japanese industry
may be dismantled on a selective basis in the near future.
This could lead to a substantial rush by European and
United States multinationals to establish manufacturing
locations in Japan. Clearly this would substantially aug—
ment the expansion rate of United States and EurOpean
multinational firms. It is nearly impossible to predict
exactly what the implications of this expansion are to the
multinational firm's shareholders; but it will, in all
likelihood, make the historical experience of the past
two decades an even more important input to the decision
process for the 70's.

These past few paragraphs have attempted to point
up a few of the large number of unanswered questions that
currently exist or will shortly arise in the area of the
multinational corporation and its effects on its common
stock shareholders. Furthermore, because of the rapid
changes that are taking place in the Operating environment
of the multinational corporation, any extrapolation of past
findings into future periods must necessarily be done with
extreme caution. While it is certainly possible that some
of the changes that are presently taking place, and those
yet to come, may substantially alter the findings of this

study, it is essential that answers be found for the

252
questions concerning the effects of the multinational
corporation on its common stock shareholder. The findings
of this study suggest that during the 1960's the share-
holders generally benefited from multinational diversi-
fication, but there is no guarantee that this will continue
to be the case in the 1970's and '80's. Consequently,
there is a need for continuing research in the area of the
multinational corporation and its effect on its share-
holders. HOpefully, the findings from this study will be
a step in the direction of increased knowledge about the
multinational corporation and its effects on the common
stock shareholder.

But the need for further research is not limited
to the area of the effects of the multinational corpora-
tion on its shareholders; it includes the whole range of
its possible effects on such things as the domestic
economy of the host country, international capital flows,
the domestic economy of the country where it is based,
plus numerous others. It is essential that some answers
be found in these areas if we are to solve some of the

critical problems of the 1970's and '80's.

BIBLIOGRAPHY
Books

Anderson, T. W. An Introduction to Multivariate Statis-
tical Analysis. New York: JOhn WiIey & Sons,
Inc., 1958.

Bierman, Harold, Jr. and Seymour Smidt. The Capital
Budgeting Decision. New York: The Macmillan
Company, 1971, 3rd ed.

 

Bock, Darrel R. ”Multivariate Analysis of Variance of
Repeated Measurements," in C. W. Harris (editor),
Problems in Measuring Change. Madison, Wi:
University of Wisconsin Press, 1963, pp. 85-103.

Brooke, Michael Z. and H. Lee Remmers. The Strategy of
Multinational Enterprise Organization and Finance.
New York: American Elsenier Publishing Company,
Inc., 1969.

 

Fayerweather, John. International Business Management.
New York: McGraw-Hill Book Company, 1969.

Gini, Corrado. Memorie di metodologia statistica.
Ernesto Pizzetti and T. Salremini, Rome: Libreria
Emedi Virgdio Veshi, 1955.

Kendall, Maurice G., and Alan Stuart. Advanced Theory of
Statistics in Three Volumes. New York: Hofner
Publishing Company, 1963, Vol. 1, 2nd ed.

 

Mao, James C. T. Quantitative Analysis Of Financial
Decisions. LNew York: The Macmillan Company,
1969.

 

Morrison, Donald F. Multivariate Statistical Methods.
New York: McGraw-Hill Book Company, 1967.

Rolfe, Sidney E., and walter Damm (editors). The Multi-
national Corporation in the WOrld Economy.
New York: Praeger PubliShers, 1970.

 

‘253

254

Van Horne, James C. Financial Management and Policy.
Englewood Cliffs, N.J.: Prentice-HaII, Inc.,
1968, lst ed.

 

Periodicals

 

Arditti, Fred P. "Risk and Required Rate of Return on

Equity," Journal of Finance, Vol. XXII, NO. 1
(March, 1967), pp. 19-20.

 

Ben-Shahar, Haim and Marshall Sarnat. "Reinvestment and
the Rate of Return on Common Stock," Journal of
Finance, Vol. XXI, No. 4 (December, 1966), pp.
V37—742.

 

Evans, John L. and Steven H. Archer. "Diversification
and the Reduction of Dispersion: An Empirical
Analysis," Journal of Finance, Vol. XXIII, No. 5
(December, 1968), pp. 761F767.

 

Fisher, I. N. and G. R. Hall. "Risk and Corporate Rates
of Return," The Quarterly Journal of Economics,
Vol. LXXXIII, NO. 1 (February, 1969), pp. 79-92.

 

Grubel, Herbert G. "Internationally Diversified Port-
folios: Welfare Gains and Capital Flows,"
American Economic Review, Vol. LVIII, No. 5
(December, 1968), pp. 1299-1314.

 

Grubel, Herbert G. and Kenneth Fadner. "The Independence
Of International Equity Markets," Journal of
Finance, Vol. XXVI, NO. 1 (March, 1971), pp. 89-94.

 

Jensen, Michael C. "The Performance of Mutual Funds in
the Period 1945-1964," Journal of Finance, Vol.
XXIII, No. 2 (May, 1968), pp. 389-416.

 

King, Benjamin E. "Market and Industry Factors in Stock
Price Behavior," Journal of Business, Vol. XXXIX,
NO. 1, Part II (January, 1966 - Supplement),
pp. 139-190.

 

Levy, Haim and Marshall Sarnat. "International Diversifi-
cation of Investment Portfolios,” American
Economic Review, Vol. LX, No. 4 (SeptEfiEer, 1970),
pp. 668-675.

 

Levy, Robert A. "Measurement of Investment Performance,"
Journal of Finance and Quantitative Analysis,
VOl. 3, NO. 1 (March, 19687, pp. 35-57.

 

255

Mao, James C. T. "Survey of Capital Budgeting: Theory
and Practice," Journal of Finance, Vol. XXV,
NO. 2 (May, 1970). PP. 349-360.

Markowitz, Harry M. "Portfolio Selection," Journal of
Finance, Vol. VII, NO. 1 (March, 1952), pp. 77-91.

 

Meyers, Steward C. "Procedures for Capital Budgeting
under Uncertainty," Industrial Management Review,
Vol. 9, No. 3 (Spring, 1968), pp. 1-19.

Mood, A. M. "On the Asymptotic Efficiency of Certain Non-
Parametric Two-Sample Tests," Annals of Mathematical
Statistics, Vol. 25 (1954), pp. SIX-522.

 

Sharpe, William F. "A Simplified Model for Portfolio

Analysis,” Mana ement Science, Vol. 9, NO. 2
(January, 1933), pp. 277-293.

 

Stevens, Robert warren. "Scanning the Multinational Firm,"
Business Horizons, Vol. XIV, NO. 3 (June, 1971),
pp. 47-54.

Treynor, Jack L. "How to Rate Management of Investment

Funds," Harvard Business Review, January-February,
1965’ Pp. 63-75.

Treynor, J. L., W. W. Priest, Jr., L. Fisher, and C. A.
Higgins. "Using Portfolio Composition to Estimate
Risk," Financial Analyst Journal, Vol. 24, No. 5
(September-OctOber, 1968): PP. 93-100.

wells, Chris. "The Beta Revolution, Learning to Live with

Risk," Institutional Investor, Vol. V, No. 9
(September,l§7l), pp. 21-27}

Other

 

Bank and Quotation Record. New York: William B. Dana
Company, PubIIshers, various dates of publication.

 

ISL Daily_Stock Price Index. New York: Standard and
Poorrs Corporation, various dates of publication.

Lessard, Donald Ray. "Multinational Portfolio Diversifi-
cation for Developing Countries," unpublished
thesis, Stanford University, 1970.

Million Dollar Directory 1971. New York: Dun & Bradstreet,
Inc.,21870.

256

Moody's Dividend Record. New York: Moody's Investors
Serice, Inc., various dates of publication.

Securities and Exchange Commission, Directory of Companies
Filing Annual Reports with the Securities ami
Exchange Commission. washington, D.C.: U. 8.
Government Printing Office, 1970.

Standard and Poor's Register of Corpgrations, Directors
and Executives, 1971. New York: Standard and
Poor's Corporation, 1971.

The Analysts Handbook - Monthly Supplement. New York:
Standard and Poor's Corporation, various dates of
publication.

The wall Street Journal. Dow Jones and Company, various
issues from 1983 through 1971.

Selected General References

Books

 

Johnson, Robert W. Capital Budgeting. Belmont, Cal.:
wadsworth PubliShing Company, Inc., 1970.

Markowitz, Harry M. Portfolio Selection: Efficient
Diversificationof Investments. New York: John
Wiley & Sons, Inc., 1959.

Robinson, Richard D. International Management. New York:
HOlt, Rinehart and Winston, 1987:

Salera, Virgil. International Business. Boston:
Houghton Mifflin Company, 1969}

Siegel, Sidney. Nonparametric Statistics: _For the
Behavioral Sciences. New York: McGraw-HiIl
Book Company, Inc., 1956.

Steiner, George A., and Warren M. Cannon. Multinational
Corporate Planni‘g. New York: The MacmilIan
Companyf‘1966.

 

Vaupel, James W., and Joan P. Curham. The Making of Multi-
national Enterprise. Boston: Graduate School of
Business Administration, Harvard University, 1969.

 

257

Weston, J. Fred, and Bart W. Sorge. International

Managerial Finance. Homewood, Ill: Richard D.
Irwin, Inc., 1972.

 

Zenoff, David B., and Jack Zwick. International Financial
Management. Englewood Cliffs, N. J.: Prentice—
Hall, Inc., 1969.

Periodicals

 

Bruck, Nicholas K., and Francis A. Lees. "Foreign Content
of U. S. Corporate Activities," Financial Analyst

Journal, Vol. 22, No. 5 (September-October, 1966),
pp. 127-132.

 

Culver, David M. "International Business and Its Economic
Implications," Advanced Management Journal, Vol.
35, No. 4 (October, 1970), pp. 56-58.

Fama, Eugene F. "The Behavior of Stock Market Prices,"
Journal of Business, Vol. 38, NO. 1 (January,
1965), PP. 34-105.

Fama, Eugene F. "Risk, Return and Equilibrium: Some
Clarifying Comments," Journal of Finance,
Vol. XXIII, NO. 1 (Maréh) 1968), pp. 20940.

Kendall, Donald. "Corporate Ownership: The International
Dimension," Columbia Journal of World Business,

Vol. IV, NO. 4 (July-August, 1969), pp. 59-65.

Kocan, Peter. "Geographical Distribution of Earnings and
Assets," Journal of Accounting, Vol. 115, No. 6
(June, 1963), pp. 49-54.

"Multi—national Companies," Business Management, December,
1966, pp. 53-56.

 

Perlmutter, Harry V. "The Tortuous Evolution of the
Multinational Corporation," Columbia Journal of
WOrld Business, Vol. IV, No.71'TJanuary-February,
1969), pp. 9-18.

 

Sharpe, William F. "Capital Asset Prices: A Theory Of
Market Equilibrium Under Conditions of Risk,"
Journal of Finance, Vol. XIX, NO. 3 (September,
1964). pp. 425-442.

Simmons, Kenneth. "Multinational? Well Not Quite,"
Columbia Journal of WOrld Business, Vol. 1, No. 4
(Fall, 1966), pp. 115-122.

258

Smith, Keith V., and John C. Schreiner. "A Portfolio
Analysis of Conglomerate Diversification,"
Journal of Finance, Vol. XXIV, NO. 3 (June, 1969),
pp. 413-427.

 

Wilkinson, Theodore L. "International Accounting:
Harmony or Disharmony," Columbia Journal of
World Business, Vol. IV, NO. 2 (March-April, 1969),

 

 

pp. 29-36.

   

APPENDICES

APPENDIX A

ALTERNATIVE EMPIRICAL RESPONSE MEASURES

Introduction

 

The single rate of return measure and the five
surrogate risk measures presented in Chapter III are the
final result from a series of evaluations made on an
entire array of competing empirical response measures.
Since some of the measures that were excluded from this
final list have had extensive exposure in the literature,
it was felt desirable to present the rationale behind
their being drOpped in favor of the measures adOpted at
Chapter III. As would be expected, the alternative risk
exposure measures that were available far outnumbered the
rate of return alternatives. Because of their numbers,
only the more general and widely used of these risk mea-

sures will be reviewed here.

Rate of Return Measure

 

About the only alternative that was seriously

considered for measuring the returns on the common stocks

259

260

of the sample corporations was an arithmetic mean of the
quarterly price relatives. The actual price relatives
themselves would have still embodied the combined effects
of changes in market price during the period plus any
dividends received. While the arithmetic mean has prob-
ably seen wider usage, it offered no clear cut advantages
over its rival, the geometric mean. And given the vola-
tile markets that characterized part of the study period,
this average would have been subject to some degree of up—
ward bias. All things considered, the geometric average

was adOpted instead.

Alternative Risk Exposure Measures

 

Overall Variability in Returns

 

In addition to the surrogate measure presented in

Chapter III, Gini'sl Coefficient Of Concentration, two
other surrogate measures were considered for this aspect
of risk exposure. The first of these was the standard
deviation of a security's quarterly price relatives. This
measure has probably received wider use as a surrogate for
risk exposure than any other. While the outcomes were not
reported in the results chapter, the standard deviation of

the quarterly price relatives was computed for all

 

lGini, Op. cit.

261
corporations in this study. As used in this study, the

computational form was:

n ...—
SD = )3 (loge PRt - logePR)2‘t

t l

 

i
n J
Where

logePRt = the natural logarithm of the quarterly
price relative for period t

logéPR = the geometric mean of the natural log
price relatives, this for the total
time period n
t = the particular quarterly time period
n = the total time period
As a consequence of converting all quarterly price rela-
tives to logarithms, the resulting measure was a relative
one, thereby facilitating later comparative analysis.
Although the use of this measure does not require
the quarterly price relatives to be exactly normally dis-
tributed, it does make the implicit assumption that this
statistic adequately describes the variability of the
price relatives. Further, to compute this statistic one
must estimate the mean Of the pOpulation using the sample
results. While nothing in the preliminary analysis sug-
gested that the data of this study violated these assump-

tions, the Gini measure involved none of these assumptions

or limitations. Consequently, it was adopted instead of

262

this measure.

The second of the alternative measures for this
aspect of risk exposure was the average deviation of a
security's quarterly price relatives about the geometric
mean. A Bank Administration Institute study2 dealing with
the measurement of pension fund performance recommended
this measure as a surrogate for risk exposure. As with
the last measure, the average deviation was computed for
all corporations in the study. The actual computational

form was:

 

AD = tg ABS(PRt - "151-i)
n
Where
PRt = the quarterly price relative for period t
‘PR = the geometric mean of the quarterly price

relatives for the total period n
and where n and t are as previously defined. As actually
used in this evaluation, the above statistic was converted
to a relative measure by dividing the above average devia-
tion by the geometric mean Of the price relatives, PR.

Two weaknesses led to the rejection of this measure.

 

2Bank Administration Institute, Measuring the
Investment Performance of Pension Funds (Park Ridge,
Illinois: Bank Administration Institute, 1968).

 

263

First, it failed to give any additional weight to abnor-
mally low or high price relatives. For the majority of
common stocks in the sample this was not a problem, but

on the few exceptions it made a substantial difference.
Second, its computation also required an estimate of a
population parameter: the mean of the price relatives.
But, since the Gini measure had neither of these short-
comings, it was adopted instead of this relative deviation

measure .

Downside Risk Exposure

 

The only alternative that was seriously considered
for this aspect of risk exposure was the semi-variance of
a security's quarterly price relatives. Recall that the
modified quadratic mean, as presented in Chapter III,
specified risk exposure as being any quarter in which the
price relative was less than 1.0--a negative rate of
return. This alternative measure likewise concentrated on
the downside aspects Of risk exposure but it described risk
in terms of those quarters when the price relative was less
than the mean of the price relatives for the period. The

computational form was:

264

SV =[tgl (logePRt* - logéPR)€]/n
Where

logePR; = logePRt when logePRt'<logePR, and

= 0 when logePRt zlogéPR
And where logePRt, logéPR, t and n are as previously
defined. Use of natural logarithms for the price relatives
produced a readily usable relative measure.

Unlike the two previous alternative measures, the
real question here was not whether this measure had more
restrictive assumptions or limitations, but instead it
involved what really was downside risk. First, the con-
cept that risk is really those returns of less than zero--
the cut-off point for the quadratic mean measure--appears
to be the more reasonable Of the two. Second, due to its
method of computation this alternative measure would rank
two securities as having equal risk exposure even though
the one had a series of all positive returns, yet the
other one had the identical series of returns except they
were all negative. Clearly this measure, given the prOper
circumstance, had some serious shortcomings. Thus, it was

rejected in favor of the modified quadratic mean measure.

APPENDIX B

EVALUATION OF THE RELATIONSHIP AMONG
THE FIVE SURROGATE RISK MEASURES

Recall that a part of the analysis in the first
half of Chapter V attempted to identify which of the five
risk measures had the largest difference between the multi-
national and domestic groups, and therefore contributed
most to differentiating between the two groups. One of
the problems encountered in that analysis was that in
general when a surrogate risk measure was ordered first
in the step—down F procedure it was usually found to be
statistically significant at a reasonable level. Yet, if
the same measure was ordered second or lower in the list,
and therefore it had been conditioned by at least one of
the other risk measures, its statistical significance
drOpped sharply. The interpretation of this finding was
that there was a sizeable common element of variability
among the risk measures. And once this common element
had been removed by conditioning on the first risk measure,

the remaining residual variability in the other risk

265

266
measures was not sufficient to make the between group dif-
ferences on these measures statistically significant at a
reasonable level. Several additional points require
further clarification at this time. First, the above
comments apply to all of the risk measures except relative
skewness. Recall it had a special but limited purpose:
to capture the possibility Of a major financial loss.
Second, the residual risk measure, even when it ranked
first, had only limited statistical significance.

Table B-1 presents the correlation matrix for the
five surrogate risk measures. It was estimated using the
140 firms shown in the final corporate sample at Table 3,
Chapter III. The time period covered was the total period
of the study: the first quarter of 1963 through the
fourth quarter of 1971. From that table, it appears that
four of the risk measures, GINCC, QUADMN, BETA, and RESID
were highly correlated. The slightly lower correlation
between the residual risk measure, RESID, and the market
risk measure, BETA, probably stems from two circumstances.
First if the majority of a security's risk exposure was
attributed to market factors--a high BETA coefficient--its
residual risk would likely be small. Likewise, if a secu-
rity had low market risk, in some cases its residual risk

might be quite high. Both of these cases would contribute

267

TABLE B-l

CORRELATION MATRIX FOR THE FIVE

SURROGATE RISK MEASURES

 

 

 

to the observed low correlation.

 

 

SKEW GINCC QUADMN BETA RESID
SKEW 1.000
GINCC .218 1.000
QUADMN - .103 .884 1.000
BETA - .157 .779 .831 1.000
RESID .245 .962 .853 .626 1.000
LEGEND
Symbol Response Measure Risk Exposure Aspect
SKEW Relative Skewness Major financial loss
GINCC Gini's Coefficient of Concentration Overall variability
QUADMN Modified Quadratic Mean Downside risk
BETA Beta Coefficient Market risk
RESID Standard Error of the Estimate Residual risk

The one risk measure

that appears to be unique among the five was relative skew-

ness .

Given its low correlation to the others, it appears

that this measure captures an aspect of risk exposure

that is unique to the other four.

The uniqueness of the skewness measure was further

demonstrated when a principal components analysis was done

268
on the correlation matrix in Table B-1. Since the first
two components accounted for nearly 93% of the variation,
only the standardized coefficients for those two are
shown in Table B-2. The nearly equal sized coefficients
for the last four risk measures in Component 1 clearly
suggest there was a sizeable common element between these
four measures. While the second component accounted for
a considerably smaller portion of the total variation,
it did, nevertheless, highlight the previously mentioned
uniqueness of the relative skewness measure. Based on
this finding, it appears that this measure is reasonably
independent of the other four.

As the final step of this analysis, a maximum
likelihood factor analysis was performed on the correla-
tion matrix of Table B—1. The principal purpose of this
analysis was not so much the identification of the common
element, but instead, this represented an attempt to estab-
lish a rough estimate of the statistical significance of
the common element. Drawing upon the results from the
previous section, two factors were estimated. The coef—
ficients for these, after they were rotated using a vari—
max criteria, are shown in Table B-3. The chi square
statistic for the goodness of fit for this two-factor

solution was not significant at the .05 level. And since

269)

TABLE B-2

PRINCIPAL COMPONENTS FOR THE CORRELATION

MATRIX IN TABLE B-l

 

Risk Measure*

Coefficient for Component

 

 

1 2
SKEW -.086 .977
GINCC -.979 .153
QUADMN -.953 -.186
BETA —.857 -.307
RESID -.932 .224

 

*Refer to Table 8-1 for details on the risk measures.

TABLE B-3

MAXIMUM LIKELIHOOD FACTOR ANALYSIS:
COEFFICIENTS FOR THE TWO FACTORS

 

 

Risk Measure* Coifficients for Facgor
SKEW .035 .495
GINCC .952 .294
QUADMN .921 .089
BETA .932 -.361
RESID .864 .500

 

*Refer to Table 8-1 for details on the risk measures.

. V. .Aiiahi'

270
this was in the rejection region the conclusion was that
there was a lack of fit for the two—factor solution. But
there were several extenuating circumstances that quali—
fied this conclusion. First the residual variances that
remained after removal of the first factor were very
small on all risk surrogate measures except relative skew—
ness. On several of these measures the residuals approxi-
mated zero. This suggested the existence of a definite
common element among four of the risk measures. Because
of the limited residual variances that remained after
fitting the two-factor solution, no attempt was made to
fit a three-factor solution; had there been a test for
goodness of fit on such a model it probably would have
shown the fit to be adequate. Last, the actual chi square
goodness of fit test tends to be overly sensitive to lack
of fit in this application. Given these qualifications,
it was concluded that the fit of the model was in all
likelihood adequate. Therefore, the results from this
test further substantiate the existence of a significant
common factor among the last four risk surrogate measures
shown in Table B—3. The results from the second factor
suggest that relative skewness is probably unique and
reasonably independent of the other risk measures. But

the observed results offered considerably less support for

this second conclusion.

“w I, ..-.-u-.- .t" _- :g

   

MIWITIWINIIIISIW‘I‘IW IUIWIIVIDEHIWWIMIT WES
3 1293 03015 6029