ALFRED
P.
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
DIVERSIFICATION AND PROFIT PERFORMANCE
IN THE FOOD PROCESSING INDUSTRY
Krishna Gour Palepu
WP 1218-81
May 1981
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS 02139
DIVERSIFICATION AND PROFIT PERFORMANCE
IN THE FOOD PROCESSING INDUSTRY
Krishna Gour Palepu
WP 1218-81
May 1981
M.I.T.
LIBRARIES
JUL 15
1981
RECEIVED
j
^
page
INTRODUCTION
Diversification strategy is an important
management
strategic
of
a
leads us
believe
to
diversification strategy of
Much of
empirical
the
orgainization
economic
that
a
be
theory
and
related.
literature,
the
in
nowever, fails to provide conclusive
notion
in
performance
firm should
evidence
of
Literature
firm.
management as well as industrial
component
support
this
to
.
Whether or not diversified firms are better performers,
it is clear
more
that
diversification
more
and
strategies.
Fortune
non-single businesses,
500
in
are
by
19^9 only about thirty
corporations
U.S.
1969
pursuing
provided
Data
Rumelt(1974) shows that while in
percent of the
firms
over
pursued some form of diversification.
ninety
were
percent
the light
In
of
such overwhelmingly prevelant practice, the question of
the
relationship
of
diversification
to
economic
performance is of considerable interest to students
of
management
The purpose of the present paper is
look at
the
relationship
to
take
fresh
a
between diversification and
firm performance drawing upon empirical ev
idefi.&«. "fV^^im
''"(^'^'WW'
a
%
sample of firms in the Food Products industry;.
In
the
074 J?"^'"^
A, -J.
I.
">
page
following
section,
shall
we
expectations
theoretical
profitability.
Section
related
the
to
diversification.
on
diversification
and
deals
three
section
In
four,
corporate
brief review of
a
relationship
profitability
and
in
analysis are described
the stasitical
consists
section
of
a
issues
the
of
The hypotheses examined
presented.
with
measurement
diversification
The last
the
literature dealing with the
the empirical
between
present
first
this
will
study
be
and
section five.
in
discussion
of
the
findings of the present study and the conclusions there
from
.
DIVERSIFICATION AND PROFITABILITY: THEORY
management
early
One of the
diversification was
Alfred
theorists
Chandler.
In
analysis of seventy of America's largest
Chandler
(
1
962
noted
)
that
many
firms
that
a
studied
historical
corporations,
moved into new
industries from their original product-markets in order
to
better employ their existing productive resources as
the primary markets
diversification as
to maintain
presence of
or
a
declined.
a
Thus
Chandler
viewed
means through which firms attempt
improve
their
performance
in
the
decline of their current businesses.
Economic theory, as well as management theory,
indicate
page
that
a
diversified firm should be able to attain better
economic performance than that of
non-diversified firm.
economists
less diversified or
this belief that led
is
concerned
be
to
It
a
about
many
impact
the
of
increasing conglomeration on competitive market forces.
MarkhamC
1
973
reason behind
of
and
a
his study on conglomerate enterprises
in
,
policy
public
and
)
implications,
suggests
diversified corporation is its ability to
exercise
market
subvert
able to
mechanisms like
buying and
entry.
power (pp. 24-46).
support
cross-subsidization,
revenues
a
earned
reciprocity
is
a
one
in
'subsidize'
earned
customers
suppliers.
It
to
on
some
are
is believed
larger
will
when
also
opportunities
another.
to
competition,
a
a
uses
divisions
product
to
Reciprocity in
firm's
actual
actual
or
and
potential
that the more diversified
a
be the number of markets
in
which it buys and sells and hence, the
iiklehood of
productline
mechanism, whereby
similar
buying and selling occurs
company, the
it
the lower prices of others.
potential
in
when
pertaining
strategic weapon against
revenues
be
to
predatory pricing and barriers to
activities
the
Predatory pricing
firm, as
believed
the competetive market forces through
selling,
employs the
acquire
Because of its
firm engages in cross-subsidization
A
the
expectation of higher profitability
the
market power, the diversified firm is
the
that
to
larger
practise
is
the
reciprocity.
page
Barriers to entry is
industry.
It
firm
'
s
industry
might
that
diversified
arising
sets
in
the acquired
companies
to
supernormal
the
is
Microeconomic
the existence of supernormal
profit information due
theory
new
in
by
argues that
industry
an
entry.
consolidated
to
'information
reporting
profits
induce
paribus,
diversified firms
lead
consolidated
from
diversified firms.
will, ceteris
firms
.
Another mechanism
loss'
und iversif led
concentration
rising
for
profits of
an
firms into industries previously
populated entirely by small
the stage
of
has been claimed by economists that entry
diversified
of large
phenomenon
structural
a
Loss
of
reporting
by
inhibits this process as it does not
indicate the profitability of specific productlines
barriers
raises
of
to
entry, thus,
helping sustain supernormal profits for
a
considerably
longer
in
the firm
and
time.
this
recent
The
requirements, which
will
change
be
discussed
reporting
have
later,
reduced this information loss problem considerably.
Management theorists cite several
possible subversion
other
than
of competetive market forces for
diversified firm's ability to
Synergy, economies
reasons
of
scale,
allocation, opportunity to
attain
higher
efficiency
exploit
a
profits.
in resource
particular
skills
page
management
like superior
and
development capabilities are
diversified corporations
factors
a
combined return
Ansoff(1975) refers to several
types of synergy that might result when
common
results
synergy
distribution
administration;
channels
operating
from
systematic
a
from
products
using
or
common
sales
synergy results from better
utilization of facilities
synergy occurs
takes
firm
a
care to select its products and markets in
Sales
the
on
than the sum of its parts is
greater
referred to as synergy.
manner.
enable
that
improve their performance.
to
The effect which can produce
firm's resources
superior research
strength,
joint
investment
personnel;
and
of
use
plant, common raw
materials and transfer of know-how from one product
another;
to
management
utilize
the
across products.
results from the ability
synergy
general
superior
Salter
and
Weinhold (1978) mention
might result in better economic value for
argue that
a
to apply the
partner
to
strategies
a
firm.
diversifying acquisition allows
particular skills
solve
opportunities facing
the
and
problems
other;
the
skills
management
diversification
several ways through which
to
knowledge
and
a
company
of
exploit
that
They
one
the
diversified
companies can channel cash from units with surplus cash
to
units
with
current
operating
promising future potential;
effect, the
diversified
deficit
that due to the
firm
can
but
with
portfolio
reduce its overall
page
risk and hence lower its cost of capital.
of reasoning for expecting diversified
firms to perform
better than non-diversified ones is that
expected to
undertake
Another line
a
firm can
be
new activities rather than grow
within the scope of its existing product structure only
if the former alternative promises
return
To
higher prospective
a
.
conclude,
then,
economic theory
both
suggest
management
literature
diversification
that
and
should
bring superior performance to business firms.
MEASUREMENT OF DIVERSIFICATION: STATE OF THE ART
The method used to measure diversification is the basic
determinant of
study.
the
number
A
validity
of
any
of measures of diversification, both
quantitative and qualitative, have
used by several researchers.
present
a
diversification
In
been
proposed
and
this section, we shall
brief discussion of the issues related to the
measurement
of
diversification,
and
discuss
the
relative merits of the various approaches.
Gort (1962) was one of the earliest researchers in
field
enterprise
of
diversification as
output of
a
firm,
the
diversification.
degree
of
He
the
defined
heterogeneity
of
from the point of view of the number
page
of markets
served
concept of
elasticity
output.
that
by
of
demand
Borrowing
economics, he
from
markets
specified that two products belong to seperate
cross-elasticity
if their
zero
As
of
demand
the
approximately
is
.
measure
Gorecki (1974) pointed out, any summary
of
diversification should take into account two dimensions
heterogeneity
of this
-
(i)
which the enterprise operates
importance of
of the
each
the number of markets in
and
relative
the
(ii)
of the markets in the total output
enterprise.
measures
Several
have
been
developed by researchers in an attempt to capture these
two dimensions.
Gort proposed and employed three measures in his
on
diversification
and
integration
study
American
in
count
of
industry.
The simplest of these was
number of
industries in which the company produced and
sold goods and
Another
services.
diversification proposed
by
a
the
measure
simple
of
him was the complement of
the ratio of primary industry output to total output of
the
enterprise
specialization
literature).
(which
ratio
also
is
in
known
the
as
organization
industrial
Both these measures, however, suffer from
severe limitations.
The
former
fails
to
take
into
consideration the relative magnitudes of the outputs
in
page
various industries.
does
latter
The
not
take into
account the composition (both in terms of the number of
industries and
output.
nonprimary industry
these weaknesses,
was
a
multiplying
With
Gort proposed
composite
complement
by
which the enterprise
operated.
than the
the
two measures,
other
industry
primary
number
specialization ratio
which
was derived by
It
of
the
measure
third
a
of
view to counter
a
the above two.
of
the
magnitudes)
relative
their
of industries in
Though
refined
more
Gort's composite measure
of diversification also fails to fully account for
the
composition of the firm's nonprimary output.
Berry(197l) and MGVey(1972),
independently,
weakness of Gort's measure by deriving
this
to
remedy
an
index of diversification by applying
system used
in
concentration,
attempted
the Herfindahl
measure
a
organization research.
weighting
Summary Index of industry
widely
The
the
used
industrial
in
measure
Herfindahl
was
defined as
1
where.
Pi
is the
classified in
-
.
Pi
proportion of the enterprise's
the
ith
industry
number of industries in which
The relative
Pi
the
and
N
output
is the total
industry
operates.
importance of the various products in the
page
firm's total output is thus taken into consideration by
weighting the share of each product by itself.
The Herfindahl concentration index has been widely used
and
hence historically the Bureau of
been available
for
data
Census
Due to this easy availability
it.
of data and because of its bounded properties, ease
understanding
intuitive
and
has
appeal,
of
Herfindahl
the
diversification index too has become quite popular with
researchers
.
most of the studies on diversification employing the
In
above indixes, products were identified as belonging to
seperate markets
on
basis
the
Classification
Industrial
because, as
of
practical matter, most
a
This
code.
(SIC)
Standard
the
of
the
publicly
available data was cast in the mould of SIC codes.
choice
2-
,
of
3- ,or4-digit
researcher
level
the
level
of
was
industry
detail,
however,
left
was
The
namely,
to
the
.
Jacquemin and
Berry
limitation of
the
pointed
to
Herfindahl index.
As
(1979)
one
serious
2-digit level
essentially
SIC classification
of
aggregation of
finer classification at the 4-digit
the
industries
level, one might be interested
in
diversification of
to
a
firm due
is
an
knowing the extent of
its
diversification
page
across the
to
2-digit industry groups as opposed
broader
diversification
the
10
within
2-digit
one
industry
group across some-what related 4-digit industries.
Herfindahl index,
because of the way it is calculated,
decomposed
cannot be
contribution
level
product
industry groups
Herfindahl
each
at
total
the
to
acros
index
2-digit
large manufacturing firms as
460
for
which
For example, the average
the firm.
I960 of the
elements
diversification
of
aggregation
diversification of
value in
additive
into
define the
of
The
calculated by Berry
(1975)
value of
across 4-digit industries, for the
index
the
was
same firms in the same year was 0.645.
much larger
The latter
the former not merely because
than
at a finer
level of classification;
counting
in
computing
there
latter
the
of diversification
separately,
the
what
2-digit
the
at
Herfindahl
the
level
some
consequence
a
index,
industries.
In
the scheme of
4-digit industries
more closely
relaed
industries across
are, clearly,
each
within
It
calculate
to
the
to
within
a
to
one
is not
total
purely
at
the 2-digit level
of
SIC
was
because
diversification was, of the diversification
the 4-digit
it
double
level.
contribution
was
was
diversification at the 4-digit level is
possible, using
average
The
0.379.
classification,
as
2-digit industry group are
different
diversification
another
2-digit
at
the
than
4-digit
industry
groups
4-digit
level
page
within
industry
2-digit
a
1
group
diversification'
'related
diversification
across
can
viQwd
be
contrast
in
2-digit
groups
with
which
relatively 'unrelated.' The Herfindahl index
provide measures
fails
-
of diversification
to
The entropy measure,
has
measure
its
entropy
of
origins
in
which was first
borrowed
of industry concentration.
is
from
industrial
the
organization research where, it was used as
the
measure
a
Jacqumin and Berry modified
for measuring corporate diversification.
index
-
overcome this limitation of the
developed from the concept
also
to
separately.
Jacquemin and Berry (1979) proposed an entropy
physics,
is
of the two types of diversification
related and unrelated
Herfindahl index.
as
It
defined as follows:
Consider
a
These
4-digit
N
firm
operating
industries
in
4-digit
N
industries.
turn aggregate into M
in
2-digit industry groups (where M is obviously less than
or equal
to
N)
.
Let
Oij
=
Output of the firm in the
jth
4-digit
industry
within
ith 2-digit industry group,
Oi
=SOij
=
Output of the firm in the ith 2-digit
group
page
=lOi
12
Total output of the firm,
=
i
Pij
Oij/Oi,
=
Pi
=
Oi/0,
Pj
=
Oj/0
and
The entropy measure of diversification at
level within the 2-digit industry group
Dwi
21
=
Pij ,ln
(
the
M-digit
is defined
i
as
1/Pij)
The entropy measure of diversification
across
2-digit
groups is also similary defined as
Da
=
21
i
Pi
-if!
(
1/Pi)
The total diversification of the firm is defined as
Dt
=
X_
Pj.ln
(
1/Pj)
J
It
can
easily
diversification
Dt
,
that
then,
shown,
be
the
the within diversification Dw
the across diversification Da,
defined
as
above,
total
,
and
are
related to each other as follows:
Dt
=
T:
Pi .Dwi
The first term on the right
hand
+
Da
side
of
the
above
page
expression is
13
weighted average of the firm's 4-digit
a
diversification within each 2-digit group in which
the
firm operates (each 2-digit group being weighted by its
relative proportion
The second
2-digit industry
diversification
firm's
term is the
firm's total output).
the
in
Pi
groups.
Thus
entropy
the
across
measure
enables one to decompose the total diversification into
two components -
average
weighted
a
within 2-digit
industry
groups,
and
across 2-digit
industry
groups.
If
diversification within
2-digit
related diversification DR
diversification
diversification
we
define
industry groups as the
and diversification
,
the
across
2-digit groups as unrelated diversification DU, then we
can
write
DT
DR + DU
=
where DT is the firm's total diversification and
M
As
=
-fi.
Pi.Dwi
and
DU
=
Da
diversification across 2-digit groups takes
away from
present market much 'further'
its
the diversification within
implications
of
each
corporate performance
Researchers were
a
type
might
2-digit
of
group
a
firm
than does
does,
the
diversification
for
significantly
differ.
unable to study such implications due
page
to the
inability of all indexes other than the
measure to provide such
above attempt to quantify the
of diversification of
'amount'
entropy
distinction.
a
the measure discussed
All
^^
firm.
a
different
A
approach towards diversification was proposed by Rumelt
(197M).
by
a
viewed diversification as
He
'resulting from
firm
balanced consideration of
a
opportunities
firm's skills and resources, the
in
the
consists
skills
purposes
and
old
activities.'
operational ize this
(pp.11).
definition,
se
(ii)
;
this
span
that
activities
diversity, demonstrated by the way new
related to
of
two dimensions:
of
the firm's commitment to diversity per
the strengths,
extant
definition
Rumelt's
(pp.10).
diversification strategy
'(i)
a
environment, and the personal desires
economic
of management'
strategic move
a
attempt to
an
In
adopted
Rumelt
are
the
classification system for diversified firms proposed by
Wrigly (1970)
and
further refined
Using
it.
set of
a
quantitative and qualitative criteria, he proposed nine
categories of
business,
dominant
dominant
constrained,
constrained, related
and
unrelated passive.
rather
than
being
single
strategies:
diversification
unrelated,
dominant
vertical,
dominant
linked,
related
linked, acquisitive conglomerate,
In
given
his scheme
a
of
numerical
diversification, each firm is placed
measurement,
index
in one of the
of
nine
page
diversification
categories based on its
summary of
15
strategy.
A
criteria used by Rumelt in this scheme
the
of classification is shown in the Appendix.
proposed
The method
measures of
diversification
quantitative and
pointed
(1974)
blessings
criteria.
a
As
mix of
Rumelt
brings
approach
this
other
from
that it uses
in
qualitative
out,
differs
Rumelt
by
mixed
:
While the use of this type of c lassif ic ation
of
consid eration
permits
the
system
data
the
app arent
nonqua ntif iable
catego ry bound aries
d efinite
pr ec is eness of
fact
that
the
the
must not
obscu re
plays
an
impo rtant
resear cher s jud gement
the
In
pr esent
role i n
the
tec hnique
of
s emisub je ctive
contex t)
this
type
techni que
several
stren gths
has
these
advantages are only ob ained
Of cour se
of obj ectiv ity and
by sac rificing a degree
repl ic ability.
p. 46)
,
'
.
(
,
(
DIVERSIFICATION AND PROFITABILITY: EMPIRICAL LITERATURE
Several researchers
relationship
have
between
attempted
findings of these studies have been
stud ies
.
we
shall
examine
the
and
firm
diversification
profitability at an empirical level.
this section,
to
briefly
By
and
large, the
inconclusive.
In
review some of these
page
16
One of the earlist studies in the area was conducted by
Gort (1962).
an
In
diversification
attempt to
the
,
trace
corporations
the largest U.S.
1947,
1950,
different measures
discussed in
measured as
a
same period.
using data on^
drawn
from
1954.
over
the
Gort
section.
failed to
three
which
were
Profitability was
profitability.
companies
taxes
after
19^7-1954 to average net worth for the
Measures of diversification were computed
employment
U.S.
and
Bureau
Industrials.
produce
relationship
of
manufacturing
payrolls
Census
(1958).
the
He
studied,
drawn
The results of his analysis
evidence
between
data
for
any
diversification
significant
and
found that profit rates for the
during
the
period
1947-54
correlated neither with diversification as measured
1954 nor
1954.
that
with
of
employed
Profitability measures were computed using
from Moody's
111
1929,
diversification
of
of
years
ratio of average net income
period
for the
and
earlier
the
of
profitability
Gort analyzed statistical data covering
firm,
1939,
history
causes for its development, and
the effect it had on size, growth and
a
the
change
in
d
firm
111
were
in
iver sif cat ion from 1947 to
However, commenting on his results, Gort pointed
:
interpreted
...this result must
not be
to
diversification
exerted
no
mean
that
influence upon the
profitability of firms.
page
17
factors
unrelated
....numerous
to
return
influence
the
rate
of
diversification
primary and secondary industries (of
in both
point
to
a
The results, however,
a firm).
influence
of
that
the
conclusion
diversification upon profits is not, alone,
sources of
overcome other
sufficient
to
variation in rates of return.
(1962, p. 77)
Arnould (1969)
analyzed
diversification and profitability using
U.S.
processing
food
diversification for
movement from
purpose
4-digit
one
a
firms.
the
between
relationship
the
sample of 104
defined
He
of
study
his
SIC industry into another.
Four measures of diversification were constructed
ratio of
nonprimary
output,
the
ratio
output
firm operates, summation of the share of
output
each
in
industry
the
(at
the
output
total
to
multiplied by the number of 4-digit industries
a
-
manufacturing
total
to
nonprimary
of
as
which
in
firm's
the
level)
4-digit
multiplied by the four-firm concentration ratio of that
industry, and summation of
output in
of
share
the
firm's
the
each industry multiplied by the share of the
total output of the industry accounted for by the firm.
The latter two
structure of
indexes
the
attempt
account
to
for
industries entered by the firm while,
the former two were the same as the ones used by
Profitability was
to
networth.
publicly
As
the
defined
data
available,
on
as
Gort.
the ratio of net profits
output
employment
census of manufacturers were used.
were
data
not
from
Arnould,
directly
the
based
1963
on
page
the results
of
diversified
firms
analysis, could not conclude that
his
nond iversif ied
18
were
firms
profitable
more
or
diversified
less
than
firms.
Commenting on his results, Arnould said:
The lack of any correlation
between
the
measures of diversification and profitability
conclusion
does
not
permit
the
that
diversification did not increase the level of
profitability of individual firms.
It
must
indicate that diversified
be interpreted
to
firms tend to be no more profitable, at least
explained
a
simple
as
might
be
by
mathematical
relationship,
than
lesser
or
nond iversif ied firms."
(1969, p. 79)
Markham (1973), as
part of an extensive
a
public policy implications, found that,
analysis
of
all multiple
in
regression models relating diversification with various
firm-related variables,
at
significant level,
a
sign.
diversification
a
company
the
a
different
the
in
acquisitions made during 1961-1970.
measures used
assets.
His
were
data
Harvard Business
measures
earnings
was
of
industries
1961-1970,
The
to
all
profitability
share and return on
per
drawn
period
acquisitions
diversifying
of
negative
a
the number of diversifying
1970,
company made
ratio
with
the number of 4-digit SIC
operated in
acquisitions
and
-
entered
it
three
employed
He
whenever profitability entered
form
Compustat
tapes,
School (HBS) multinational enterprise
project data, and HBS diversifying company survey data.
He
speculated that the
negative
relationship
between
page
diversification and
all
of
19
profitability may be due to any or
reasons
following
the
acquisitive
-
profitability for growth while
conglomerates sacrifice
companies earning relatively
pursuing diversification;
low rates of return in their present industries are the
most active acquirers of companies in other industries;
or,
alternatively,
may
it
simply
acquisition
diversification through
that
be
1961-1970 was
in
not generally an especially profitable activity.
The most recent study in the area was by Rumelt (197^).
All the three above
quantitative indexes.
using purely
discussed
section
the
in
diversification, Rumelt's
have already
we
As
measurement
on
approach
Using his nine
was different.
diversification
measured
studies
of
diversification
to
clasif ication
category
of firms based on their diversification strategy, which
Rumelt found that among
was discussed earlier,
Fortune
of 246
during the
500
period
U.S.
of
significantly
differed
1960-1969
profitability
He
including
employed several
earnings
price
ratio of common stock, and after tax return on
analysis
His
indicated
showed
in
the highest mean
related-linked,
that
the
firms
dominant-linked
dominant-constrained and
sample
corporations, profitability
among various categories of firms.
measures
a
equity.
in
the
categories
profitability followed by firms
single
business
and
acquisitive
page
20
Dominant-vertical
categories.
conglomerate
were
unrelated-passive firms
results
of
Rumelt concluded
that
firms
but
have
restricted
some extent
activiies to
poorest performers.
the
Summarizing the series
and
of
analysis,
his
that have diversified to
range
their
of
central skill or competence have shown
a
significantly higher rates of profitability than
other
types of firms
In
his study,
when no
Rumelt addressed the question as to
why,
between quantitative measures of
relationship
diversification and profitability was found by Gort and
profitability
connection between
classes.
argued
He
showed
analysis
own
others, his
that
represented managerially
significant
a
diversification
and
categories
strategic
his
meaningful distinctions among
corporate diversification postures and that it
the magnitude
of
diversity itself
-
not
is
which is what the
earlier measures of diversification tended to measure
but rather the
business
to
old
a
scope is
and
a
that
which
is
the
the
related
firm
the
critical
new
factor
in
he
Thus,
differentials.
classification
strategy 'based on
goals and
in
performance
explaining
commented,
way
-
system of diversification
essence
of
management's
top
concept of the corporation's purpose and
better predictor
of
performance
financial
than simple measures of diversification.'
(
Rumelt
,
1
974
page
P. 150)
21
.
THE PRESENT STUDY
from the above review, that the evidence
It can
be seen
on the
issue of diversification and
is far
None
conclusive.
from
profitability
firm
the studies which
of
employed the quantitative measurement approach
conclusive
diversification yielded
ev idence
provide
did
study, on the other hand,
results.
some
towards
Ruraelt's
concrete
.
studies
carefully conducted
diversification.
comprehensive
most
Rumelt's study was one of the
and
the area of corporate
in
The approach to and classification of
diversification strategies proposed by him
indeed
are
very valuable contributions to our understanding of the
phenomenon of
diversification.
corporate
should,
however,
be
In
spite of
noted
that
all these merits,
it
Rumelt's results
do not concern pure diversity per se
As
he,
actually
himself,
a
pointed
crude
measure that takes
addition to
approximation to
into
diversity
which
account
per
se
dimensions
various qualitative
category in
'the
out,
a
.
'
a
categories
are
multi-dimensional
several
factors
in
p. 95).
The
determine
the
(197^,
that
.
firm was classified
included the
firm's rationale for diversification, its product
mix,
page
22
the methods of entry into each business, and
to
which
potential
relationships
The information on
sources
numerous
these
was
like
annual
the degree
had been exploited.
obtained
him
by
reports,
books,
newspapers, magazine articles, and investors'
Clearly
this
type
diversification allows
all
the
information
quantitative and
of the method
is
disadvantages.
analysis
of
available
qualitative.
surveys.
corporate
of
researcher
the
to make
him
to
use of
both
-
However, this richness
bought at the expense of some
To
from
serious
quote Rumelt again,
....
this type of semisub jective
technique
has
several
strengths.
....
Of
course,
these advantages
are
only
obtained
by
sacrificing
a
degree of objectivity and
replicability.
(1974, p. 46)
Thus despite all its merits,
Rumelt's approach, besides
being heavily time consuming and expensive,
reluctant, but
and
unavoidable
replicability.
believe,
that
systematic
it
Is
is
not
and
quantitative measures
researchers are
path.
possible
of
us
discern
to
between
performance,
a
with objectivity
true, as Rumelt leads
relationship
diversification
Rumelt's
it
tradeoff
implies
to
any
corporate
using
diversification?
simple
If
so,
left with no alternative but to follow
Before
discarding
the
use
of
page
quantitative measures,
however,
examining the short comings
ability
weakened the
23
might
that
if any,
,
worthwhile
seems
it
have
of all the studies that employed
quantitative measures
discover
to
relationship
any
between diversification and profitability.
It
possible
is
weaknesses of
Markham.
and
quantitative
conducted by Gort, Arnould
studies
the
First,
all
measures
that
failed
firm's total diversification.
proposed by
Jacquemin
measure
quantitative
Berry
and
the
be
linked
different way
does.
in
to
examine
study
diversification and
other
no
(1979),
discussed
profit
the
way
the
on
in
systematically
a
unrelated diversif cation
demonstrate,
relationship
the
firm
As
related diversification seems
performance
from
a
diversification
of
Jacquemin and Berry (1979) clearly
their
of
components of diversification.
two
Ruraelt's study indicates,
to
components
diversification
of
earlier allows the reseearcher to
impact of
recognize
to
may be recalled that,
measure
entropy
the
for
It
employed
three
the
separately the related and unrelated
except
significant
three
point
pin
to
growth,
that
between
the
use
of
entropy measure enables to trace out relationships that
could not
The
be
failure
uncovered when other measures were used.
to
diversification may
segregate
the
two
types
of
conceivably be an important reason
page
wny
failed
studies
these
relattionships
significant
find
to
.
drawback
Another serious
was
studies suffered
worked
who
which
with
three
the
all
the lack of availability of data
measuring
for accurately
researchers
24
diversif icattion
in
the
Several
.
corporate
of
area
Arnould
diversification commented on this impediment.
(1969) put it succintly:
Almost ev er yone who has worke d in the are a of
prof itab ilit y h as been
d iver sif i cation and
nd
eavour s to theo r i ze and
tious
in
e
ambi
more
ects
of d iv ersi f ica tion
eff
on
the
specul ate
pi rica iiy V ali date
to
em
attempt
ing
than in
the hypot heses d evel oped from the the or i z ing
to
atte mpts
Much of t he ambi tion is lost in
t
to
iffi
c
ul
i
s
d
It
data
find
sui table
as
ind
xes
simp
le
e
emin
gly
uch
se
develop s
rent
by
d
iffe
produc
ed
duct
pro
s
1 istings o f
in
t ime
poi nts
companies at di ffer ent
ent
and
neon
sist
and
i
inc
ompl
ete
Sources a re
More
u
sed
ar
e
different classi f ica tions
such a s ou tput in each
desirable inform atio n
t he
cost of
btained
at
produc tl i ne can be o
i ons
assu
mpt
le
undesirab
incoraplet eness a nd
t
he
data
n
prese
nt
i
These sho rt comi ngs are
tent
can
onl
Their
ex
y be
used in t his stu dy.
dat
are
a
adequat e
un til
specul ati ve
forthcomi ng
969, p. 77)
.
.
,
.
.
(
.
As data
on
product
available, plant
used as
indexes.
to
the
1
line
employment
surrogates
for
output
and
was
not
directly
payroll figures were
diversification
computing
Clearly, this lead to unreliable measures due
inter-industry
differences
in
wage rates and
page
The difficulty of obtaining reliable
labour intensity.
data has been
general problem associated with much of
a
research
the empirical
25
These
sources of data emerged.
new
Recently, several
organization.
industrial
in
(Financial
include segment reporting required by
FASB
Accounting Standards
line of business
Board)
(1976),
reporting program of FTC (Federal Trade Commission) and
PIMS (Profit
Though
(1977).
lines
information
provide
P270-271
)
.
on
Also,
this
issue,
for
Scherrer
(1980,
tradeoffs associated
with
Scherrer (1979, p. 3-118).)
most easily
accessible
discussion
detailed
a
and
the segment reports required
a
complete
(For
see
in
a
standardized and authentic fashion.
discussion
set
individual
on
corporations
business
of
data
Strategy)
are by no means devoid of
data
these
limitations, they
product
Market
of
Impact
the
on
segment data reporting, see
Of all
sources,
these
the
publicly available one is
by
FASB,
as
these
are
included in the annual reports of each corporation.
A
limitation
third and perhaps most important
earlier studies
on
variables.
All
the
relationship
studies
a
cross-sectional design.
they attempted to test the
between
discussed
including the one by Rumelt, analyzed the
using
the
diversification has been the basic
approach used in researching the
the
of
above
relationship
One way or the other,
hypotheses
concerning
the
page
26
relationship between diversification and performance by
firms
across
comparing
performance achieved by
There is
of
that
firms with
group of
'improve'
diversification.
diversified
highly
group of less diversified firms.
a
in
approach.
this
variable
the
through
interest
of
point
time.
in
latter
The
is
absolute
change in profitability over time rather than
a
The
firm will
a
its performance over time
Thus
profitability at
the
time,
in
the theory seems to be that
expectation of
be able to
point
a
weakness
clear
a
a
at
is
influenced by numerous factors of which diversification
is only one.
in
A
cross-sectional comparison at
time fails to provide any control
for these
a
point
factors.
However, if we compare across firms, the growth rate in
profitability over
control for
a
period of time, we will be able to
factors
the
which
are firm specific
and
remain relatively constant over the period of analysis.
Comments of Gort and Arnould on
behind
failure
their
possible
the
significant
uncover
to
reasons
relationships, which were quoted earlier, point to this
very same reasoning.
When
we
look
the
at
possible
motives of firms wishing to diversify, there seem to be
two
classes.
broad
firm,
A
performing well, may seek
exploit exceptional
profits.
Another
primarily as
a
to
which
diversify
at
in
present
is
order
to
opportunities to earn still better
may
firm
means
to
diversification
seek
improve
its
present below
page
27
average or poor performance.
profitability of
might
former
the
absolute
The
level
far different
be
(both prior to and subsequent to diversification)
from
However, in
profitability of the latter.
the absolute
of
both the cases, diversification might result in similar
growth rate of profitability.
trace the
relationship
returns
levels of
relationship
due
fail
there
exists
however
performance
here
This
does
mean
not
between
measure
The relevant
clearly
is
any
effect.
relationship
no
diversification and performance.
of
discover
to
to
absolute
the
at
averaging
the
apparent lack of relationship
that
test that attempts
looking
by
will
to
A
change
the
in
profitability (in relation to the absolute prior level)
most important
uncover
any
reason
why
relationship
diversification and
Perhaps
itself.
and not absolute level
this
is
the
Gort and Arnould failed to
when
profitability
correlated
they
explicitly
without
considering other factors.
Hypotheses
The main aim of the present study is to
look at
the
relationship
profit performance
measures.
An
of
take
a
fresh
between diversification and
firms,
using
quantifiable
attempt has been made to remedy the three
limitations discussed
above, of the earlier studies by
page
taking advantage
of
28
diversification
measurement of
Through
availability.
hypotheses,
pairs
been made
successively
limitations discussed
logical, albeit
of
It
is
analysis
limitation
each
in
data
of
as
testing
an attempt has
eliminate
above.
elaborate,
role
understand the
well
as
sequential process of
a
three different
to
developments
recent
the
three
the
hoped that this
helps
us
to
played
in
the
reflect
the
results of earlier studies.
The first pair of hypotheses
propositions
tested
essentially
Gort,
by
Arnould
Markham.
and
These hypotheses derive their support from
of industrial organization which suggests that
diversified firm
profits.
is
in
a
studies
this respect using
in
a
highly
postion to earn super normal
replicate
This study proposes essentially to
the earlier
theory
the
a
superior
data set for measuring diversification.
The first pair of hypotheses can be stated
HI
:
as:
Level of total diversification is positively
related to the level of profitability across
firms
.
HI':
the average, the profitability of firms
with relatively high level of total diversification is greater than the profitability
of firms with relatively low level of
d iversif ication
On
.
page
It
29
should be noted that HI'
is a restatement of HI
in
a
differently testable form.
A
second set
hypotheses
of
distinction
between
seeks
introduce
to
related
unrelated
and
diversification.
The hypotheses themselves
directly on
results
the
it
diversification
way
itself
but
the
based
are
of Rumelt's study.
the thrust of the argument is that
the
Briefly,
merely
not
is
different
the
businesses that span this diversity are related to
the performance consequences.
influences
another that
one
This is demonstrated by the argument that synergy,
principal alleged
reason for the improved profitablity
of diversified
firms,
businesses are
somehow
occurs
only
the
new
related to the old ones.
Thus
when
there is strong reason to believe that the
higher related
diversification
better
do
with high unrelated diversification.
points out,
a
theme
may
performance.
Also,
looking
with
than
firms
Rumelt's study
As
in
fact
at
lead
it can be
better performing
less
firm
is
to
a
common
poorer
the relationship from
the other side of causality,
is
firms
diversification strategy without
and coherent
'away'
the
argued
likely
that
to diversify
from its existing lines of business and that
mostly
those
businesses to
unrelated
d
be
firms
that
unprofitable
find
that
a
their
seek
it
existing
highly
iver sif ication This argument leads us to the
.
:
page
30
following pair of (equivalent) hypotheses:
H2
:
Levels of related and unrelated diversification
are respectively positively and negatively
related to the average profitability across
firms
.
H2':
the average, the profitability of firms
with relatively high related diversification
is greater than the profitability of firms
with relatively high unrelated diversification
On
The above set of hypotheses effectively address two
the three
limitations discussed earlier.
third and
perhaps
using the
absolute
the growth
rate
serious
more
support of
these
profitability
in
as
performance
a
The third set of
modifies the second set to address
shall
as
not
they
repeat
the
arguments
in
have been stated at length
The hypotheses themselves can
earlier.
namely,
level of profitability rather than
hypotheses suitably
We
However, tne
limitation,
measure, still needs to be addressed.
this issue.
of
be
stated
as
follows
H3
:
H3':
Levels of related and unrelated diversification
are respectively positively and negatively
related to the growth rate of profitability
across firms.
the average, the profitability growth rate
of firms with relatively high related
diversification is greater than the growth
rate of firms with relatively high unrelated
On
diversification.
Measures
,
,
page
31
We have already discussed earlier the
in
the
measuremenc
diversification.
of
employed the entropy measure
of
discussed already,
this
that allows us
separately
'unrelated'
to
diversification into
within
a
2-digit
across different
measures
of
has
SIC
and
defined
been
as
unrelated
diversification
as
industry
2-digit
related,
'related'
group;
defined
been
As
SIC 4-digit industries
industry
diversification has
study
the purpose of this
For
different
This
measure available
measure
diversification
involved
diversification.
is the only
diversification.
study, related
issues
unrelated
groups.
and
The
total
diversification that have been used are:
DRs
=
DUs
=
21
Pis.lnC 1/Pis)
Ps
.ln( 1/Ps
Ps
DR
=
E.
DU
=
S- DUs
DT
.DRs
,
,
)
,
and
DR + DU
where
Pis is the share of the ith
4-digit
industry in the
firm s
sth 2-digit industry group's total sales,
Ps is
the
share of the sth 2-digit industry group in
the
firm's total sales,
DRs is the related diversification contributed
by the
sth
'
2-digit industry group,
DUs is the unrelated diversification contributed by the
sth
2-digit industry group,
DR,DU,DT are the overall related, unrelated, and total
diversification of the firm.
.
page
Profitability of
several ways.
32
firm can be defined
a
the
For
purpose
this
of
owners
measure chosen was the return on
one
any
in
of
study, the
equity
(ROE)
computed as the net profit (after taxes and dividend on
preferred stock
before
but
percentage of shareholders'
extraordinary items) as
a
Profitability in
a
equity.
given year is computed as the average return on
equity
over the preceding ten year period including that year.
(For
profitability
example,
the
for
defined as the average ROE over
the
year
1973
period
year
ten
is
1964-1973)
Profitability growth rate between two
computed as
the
years
from the base year to the final year.
index
the prof itibility index
been
increase in profitability
percentage
the profitability
has
for
for
Thus if
is
1973 and P79 is
the year
profitability
the
1979,
P73
growth rate has been defined as:
PG(79:73)
100
=
*
(P79
-
P73)
/
P73
Data and Computation of Measures
The present study uses
the Food
20).
Products
data
sample of
industry
Diversification
segment sales
a
indexes
as
30
companies
from
group (SIC industry group
were
reported
computed
using
by the companies and
page
33
compiled by Standard and Poor on their Business Segment
Profitability index and
Compustat tape.
growth rate
profitability
index were computed using data reported by
the companies as compiled by Standard and
Poor on their
expanded annual industrial compustat tape.
period
The
covered by the study was 1973-1979, the years for which
for which segment sales data was publicly available.
As
mentioned
briefly
Accounting Standards
FASB
earlier,
statement
Board)
that the annual financial statements
enterprises
prepared
accepted accounting
related to
enterprise's
the
industries.
principles
requires
4
business
generally
with
information
include
different
in
Included in the required information to be
identifiable assets
Segments
managment
that
such
combined
either during
factors required
products into
each
by
a
represents
at
least
profit/loss
revenue,
to
or
and
The
grouping
are the nature of the product,
and marketing methods used
15,
ten
assets
year.
while
the nature of the production process, and
reporting
firm's
or
preceding
considered
be
segments
begining December
loss,
or
defined
are
current
the
profit
of each of the busines segments of
an enterprise.
The segment
1
.
all
operations
reported are revenue, operating
percent of
no
of
conformity
in
(Financial
in
markets
the case of each product.
requirement
1976.
the
came
All companies
into
which
effect
were
page
34
registered with the SEC were required to report similar
information in their 10-K reports to the SEC.
Standard and Poor's Business Information
tape includes
Compustat
II
industry segment file containing tne
an
data reported by firms in accordance with FASB
The
14.
file contains up to seven years of information for each
company, the period covered at present being 1973-1979.
Each year for which data was available for
there are
to
up
company,
a
ten records of industry segment data,
with each industry segment record providing information
data on
up
segment.
to
Each
Whereever it
into only
one
describe the
four principal
segment
was
is
SIC
code,
activities
For each of the principal
and
sales data were
decsription of
the
referred to Standard
products of the industry
assigned
possible
not
to
and
code.
SIC
an
classify
SIC
two
codes
a
segment
that
best
of the segment were assigned.
products listed, its SIC code
For
availble.
data
contains
Each record
on a particular indstry segment.
and
a
the tape,
Poor's
more
detailed
t-he
reader is
Compustat
Services'
mannual on Business Information Compustat II.
Diversification indexes were computed for each
for each
year
using
the SIC codes and
each of the principal products of the
few cases,
when an industry segment of
Sales data for
company.
a
company
In
a
firm consisted
page
of only
to
principal product and it was not possible
one
assign
a
were split
unique SIC code to that segment, the
between
equally
for that segment.
to
35
sales
the two SIC codes asigned
The sensitivity of the final results
this allocation assumption was tested by replicating
the analysis using another rule
split in
the
proportion
ninety
of
percent between the two SIC codes
result of
this
that
test,
which
by
,
sales
percent
It
to
ten
was found, as
sensitivity
the
were
of
a
the
results to the allocation rule was insignificant.
Data for the computation of profitability measures were
drawn from the
First,
tape.
expanded
return
annual
on
Compustat
industrial
equity
for
each
of
the
companies for each year during the period 1964-1979 was
computed.
Then ten year moving average ROE for each of
the years from
computed.
was chosen
1973-1979
The moving average
for
use
indicated
(as
,
earlier)
rather than yearly ROE,
as the profitability index
view to eliminate some of the noise in
numbers. The sensitivity
of
the
was
final
the
with
a
accounting
results to the
choice of ten year period for computation of the moving
average was later tested by using five year averages in
the place of ten year averages.
to be
The results were found
robust with respect to the choice.
Profitability growth rates were computed for the
years
page
36
1973 as the base year.
1974-1979 using
Once again,
in
order to ensure that the choice of the
base
not determine
results, growth
the
pattern
of
final
rates were computed using 1974 as the base
pattern of
results
final
found
was
to
year
did
year.
The
be same with
either base year.
The statistical
confined
sections is
group.
It
should
analysis
presented
was felt that the sample
drawn
be
from
single
a
the
following
Products industry
Food
the
to
in
for
investigation
industry
that
so
industry-specific effects can be controlled for in
analysis.
random.
in
choice
The
of the food
industry itself was
order to check whether the results obtained
In
this study were purely
choice or
the
they
hold
function
a
the
industry group.
results of
this paper.
Drugs
on
a
Chemicals
and
Due to lack
industry
the
the industry analysed,
beyond
similar analysis was carried out
firms from
of
of
sample
(SIC
space,
the
of
code
that
3^
28)
complete
this parallel analysis were not reported
However, suffice it to say
a
all
in
the
results that were found in this study are were found to
equally hold
group.
well
in
the Drugs and Chemicals industry
Though this does not ensure
external
of the findings reported here, it does provide
of
confidence
to
claim
that
the
peculiar to the food industry alone.
results
valadity
a
source
are
not
page
37
Data requirements for computing the diversification and
profitability indexes for all seven years 1973
1979 limited
the sample size to 30 firms.
criterion employed in choosing the
sample is
not
should be
It
the data requirements was the only
meeting
noted that
through
sample.
Thus
the
random in the probabilistic sense.
Any
biases that may have resulted due to this could not
avoid ed
be
.
Stastical Methods
For the purpose of analysing the data and
postulated hypotheses
discussed
techniques were employed.
used
Pearson's
were
The
of
require rather
stringent
nature of
data
the
nominal, ordinal
or
(that
correlation
distributional properties
assumptions
testing
regarding
whether
is,
(that
for
the
nature),
in
is,
the
The
problem
becomes
and
here.
has
severe especially
be
The conclusions that emerge from such
analysis are thus subject to qualifications
the assumptions
its
probability
when the sample size is rather small, as happens to
the case
the
data is
distribution of the population from which the data
been drawn).
for
parametric techniques
These
interval
techniques
parametric
product-moment
means.
the
earlier, two types of
measurement of association and the t-test
the equality
testing
regarding
like,
the nature of the data,
'if
and
.
'
page
38
distribution
population
the shape of the
then we may conclude
....'
are
valid,
Nonpar ametr ic techniques,
.
on the other hand, usually require only ordinal data or
nominal data
-
clearly an easier requirement to meet
and are distribution-free.
the parametric
analysis,
were also conducted.
correlation, the
and
With
view
compliment
to
several non-parametric tests
Spearman's
include
These
median test, the Mann-Whitney
finally, the Wald-Wol fowit z
discription
excellent
a
-
of
runs
these
test.
rank
test,
U
(For
an
see
Seigel,
matrix
between
tests,
1956)
Results
Hypotheses
HI
and HI
The Pearson product moment-correlation
profitability indexes
for
the years
1973 through
(denoted as P73 through P79) and total
index
is
diversification
the same year (denoted as DT73 through DT79)
for
shown in Table
matrix for
1.
The
Spearman
the same variables is shown
can be seen that none of the
in either
1979
rank-correlation
in
correlation
table are significant.
Table
2.
It
coefficients
Thus, the results do
not support the proposition in HI.
A t-test
was carried out to examine whether proposition
page
in
The sample was split into
is true.
HI'
39
on the basis of the level of DT73.
sample
Firms with value of
were classified as the
median
DT73 above
the
'High DT73'
group and those with DT73 below the
classified
median were
as
'Low DT73'
was then carried out to see if the
either in
1973 or in
groups
two
group.
sample
A
t-test
profitability
mean
1979 was different between the two
groups, the direction of difference being as postulated
in
The results of the test are shown
HI'.
does
The test
not
hypothesis that
allow
mean
the
us
to
Table
in
reject
null
the
profitability
of
3-
the
two
significantly
groups either in 1973 or in 1979 are not
different from each other.
Three nonpar ametr ic tests were performed on
DT73'
and
Two of
'Low
these
Mann-Whitney
'High
the
groups to supplement the t-test.
DT73'
median
test
the
tests
-
the
test
-
were meant to test whether the
U
and
two groups represent populations which differ in
Wald-Wolfoowitz
central tendency.
runs
determine whether the two samples are from
which differ
in
any respect at all,
dispersion or skewness.
shown in
Table
3.
i.e.,
The results of the
test
their
was to
populations
in location,
tests
are
None the three tests supported the
proposition as postulated in HI'.
In
conclusion, then, the results do no
support
either
.
'
page
HI'.
or
HI
40
significant
No
correlation was found to
exist between total diversification in any
and
the
profitability
subsequent
the
in
same
Similarly,
years.
or
given
year
of
any
the
significant
no
correlation was found to exist between profitability in
a
given
year
total diversification in the same or
and
any of the subsequent years.
Also, when the sample was
split into the high and the low groups on the basis
their
1973
total
diversif icatio
,
significant
no
difference was found to exist between the two
of their profitability either in
of
1973 or in
terms
in
1979.
Hypotheses H2 and H2
Correlation matrices showing the Pearson product-moment
correlation and the Spearman
rank-correlation
profitability for
1973-1979 (denoted as P73
the
years
between
through P79), and related and unrelated diversification
indexes for
same
the
(denoted
period
respectively
DR73-DR79 and DU73-DU79) are shown in Table
Table
6
and
coefficients
relationships
Table
were
7.
Table
4,
5,
Once again, all the correlation
stastically
expected
in
H2
insig if ic ant
were,
thus,
The
not
substantiated by the results.
The sample was split into
four groups on the
basis
of
high and low 1973 related and unrelated diversification
.
page
as shown
Table
in
41
This table shows the number of
8.
companies, che mean 1973 profitability,
1979
profitability
of
each
for
the
'High DR73
DU73'.
other
The
observations
-
two
Low DU73'
hence, was confined only to
the
two
data,
the
'High
two
High
-
only
of
analysis
Further
each.
the
into
'Low DR73
and
contained
cells
cells.
four
Interestingly, most of the observations fell
two cells,
mean
the
and
Low
-
'
cells
In
order to examine whether H2' holds, four tests
conducted.
t-test was carried out to see whether the
A
of the 'High DR73
mean profitability
was higher than that of the
either in
were also
or
1973
Mann-Whitney
were
U
in
'Low DR73
-
High
median
The
1979-
Low DU73' group
-
In
none
test
tests was it
the
of
'
test, the
runs
test, and the Wald-Wolfowitz
conducted.
DU73
possible to reject the null hypothesis that there is no
significant differnce in the profitability of
The above
tests,
hypotheses H2
firms in
the
diversif icat-ion
once
and
sample
or
9.
again,
fail
chose
high
either
unrelated
Very few firms chose either both
support
to
indicate
Results
H2'.
two
The results of these
groups either in 1973 or in 1979tests are summarized in Table
the
or
high
the
that, the
related
diversification.
none.
Thus,
in
:
page
1973,
consists
sample
the
sub-groups
two basic
of
diversification
related
predominatly
those with
42
those with predominantly unrelated diversification.
-
or
It
seems that firms in the sample treated the two types of
strategies
diversification
However,
choices.
groups did
terms of profitability,
in
the
two
differ significantly either in 1973 or
not
seven years
exclusive
mutually
as
later,
correlation between
diversification of
in a given
profitability
kind
either
significant
no
Also,
1979.
in
year
that
in
year and
or
subsequent years, or vice versa, seems to exist.
Hypotheses H3 a"d H3
'
Hypothese H3 was first tested using 1973
computing
year in
The product moment correlations and
profit
order correlations
of
period
(denoted
1974-1979
PG(79:73)
)
with
the
rates
PG(74:73)
DR73
for
rank
the
through
and unrelated
related,
diversifications (denoted as DT73,
1973 were
Spearman
the
growth
as
total,
base
rate in profitability.
growth
the
the
as
DU73)
and
computed and are shown in Table
10
in
and Table
11.
The results
in
patterns.
The
these
two
correlation
profitability growth rate
in
tables
show
between
interesting
DT73
all the years is
with
positive
page
but insignificant.
in
to
However,
DR73 and DU73,
significant.
43
when DT73 was decomposed
correlations
the
become
highly
postulated in H3, profitability growth
As
rate is positively correlated with DR73, and negatively
correlated with DU73.
Three tests were carried out in order
validity of
H3'.
As
in
examine
to
the case of tests for H2'
DU73'
groups.
pr foitab il ity
The
-
,
analysis
sample was split into four grups and the
focussed on 'High DR73
the
Low DU73' and 'Low DR73
the
was
High
-
t-test indicated that the the mean
growth rate of the former group is higher
than that of the latter at 0.005 level of significance.
The median test and
the Mann-Whitney test
same conclusion
0.01
at
led
level of significance.
light of such strong evidence in support
Wald-Woi fowitz runs
test,
which tests
H3', was considered superflous.
a
of
the
to
In
the
the
H3',
weaker form of
The results
of
these
tests have been summarized in Table 12.
The above analysis conclsively supports H3 and H3'.
further demonstrates
found to
that
relationships,
which
It
were
be highly significant, might not be exhibited
if the analysis is carried out at the
diversification rather
level
of
total
than at the disaggregated level
of related and unrelated diversification.
:
page
44
Further Exploration of H3 and H3
An
Examination
Table
of
shows
8
'
that
mean
the
profitability in 1973 for the 'High DR73 -Low DU73' was
'Low DR73
for the
that
lower than
High DU73' group.
-
significant
Though this difference was not found to be
in our
earlier
there is
tests,
possible doubt that
a
group
the positive growth rate of the former
the
and
negative growth rate of the latter might be nothing but
the
'regression
classical
discussion on this, see
pp. 10-12.)
test
A
mean.'
the
to
(For
Campbell
and
Stanley,
carried
out
to
was
a
1963,
see if this
indeed was the case.
Table 13 shows
The
'High
recast of the data
a
DR73
Low
-
consisting
of
whose
firms
profitbility
the other consisting of firms with
'Low DU73
The
observed
were
and
above
High DR73' was similarly
-
If our
split into two subgroups.
growth rates
table.
1973 was below the sample median,
profitability in
the median.
2x2
a
group was split into two
DU73'
subgroup
subgroups, one
in
hypothesis
indeed
a
that
result
the
of the
regression to the mean phenomenon is true, then the two
subgroups with 'Low'
growth rates
each other.
which
On
associated with
1973
do
not
profitability
have
should
significantly differ form
the other hand,
if diversification
the growth rates as proposed
in
was
H3 and
page
the prof itib il ity growth rate
H3', then,
P73
-
45
High
DR73
Low
-
of
subgroup
DU73'
significanly greater than that of the 'Low
DR73
led
-
High
reject
us to
a
seems
mean
mere
a
Low
t-test
a
hypothesis
in
significance level
reversion
evidence
be
to
-
be
case
the
Thus, clearly, the growth
other two subgroups.
rates were not
there
P73
Similar result was found to hold in
.05.
of the
reversion
mean
the other hypothesis at
favour of
of
the
'Low
should
The results of
subgroup.
DU73'
the
to
phenomenon
strongly
and
support
hypotheses H3 and H3'.
DISCUSSION AND CONCLUSIONS
The results presented in the previous section
fail
support the first two sets of our hypotheses.
They do,
to
however, provide rather strong evidence in favour of H3
the logical reasoning that led to the
Given
and H3'.
development of three sets of hypotheses,
surprising.
this area
We have argued
suffered
this
is
not
that the earlier studies in
from three major limitations.
Our
first set of hypotheses addresses only one of these and
is
subject
the
to
second set
still
remaining
suffers
two
limitations.
from one of the limitations
while having dealt with the other two of them.
only the
third
limitations.
The
set that fully addressed
It
is
all the three
The results seem to indicate that all the
page
three shortcomings
need
45
significant relationships
eliminated
be
to
before
diversification and
between
profitability could be traced out.
The failure of
the
earlier studies, just like the failure of our first two
sets of hypotheses, to find such relationships does not
stem from
the fact that they used
measures' of
suggest,
diversification
as
results
do
our
for,
'simple quantitative
Rumelt
indeed
seems
to
support
the
the
relationships proposed in H3
and
H3'.
Instead,
earlier
our
first
two
studies,
also
as
hypotheses, seem to have failed because
sets
of
suffered
they
from one or more of the limitations, namely, failure to
segregate
related
simple cross
and
sectional
profitability
levels,
diversification,
unrelated
analysis
and
employing
having
absolute
of
use
to
a
not-so-reliable data.
Our results indicate that the firms with
diversification (DR)
related
high
1973 achieved, on an average,
in
sigificantly larger prof itib
il ity
growth rates than the
firms with high unrelated diversification (DU) did.
fact, the average
profitability
latter group was negative.
remaining the
same,
growth
rate
of
the former group would outperform
absolute
profitability,
sufficient time.
analysis
also
absolute profitability
the
This implies that, ail else
the latter in terms of
Our
In
of
the
two
shows
groups
given
that the
did
not
.
page
significantly differ
47
both
in
and
1973
closer look at the figures in Table
profitability
mean
of
in
shows
8
the
DU73'
group was 13-36 per cent while that of
two values were
14.25 percent.
was
group
cent
per
14.57
Thus there seems to be
respectively.
mean profitability
of
reasons
over
a
enough
strong
1979 the
cent
per
this
-
could
There
either the association
growth
between diversification and profitability
was not
'Low
two groups in the expected
the
behind
in
movement of the
a
direction though it was not significant.
be two
the
In
12.64
and
that
A
'High DR73-Low
the
1973,
DR73-High DU73'
1979-
rate
produce significant changes
to
period of seven years or else,
there
was
some
acting to dampen the changes
other counter-association
expected
In
order to
gain
issue, one
was split
a
analysis
final
into
consisting of
further
two
groups,
the sample median and the
median.
1973
mean diversification
'Low P73'
profitability
-
-total
case
of
group
P73'
(DT),
group consisting of
below
14.
both the groups.
the
sample
1973 to 1979 of the
related
for the two groups were
has been summarized in Table
in the
'High
The percentage change from
unrelated (DU)
The sample
was performed.
the
the
into
firms with 1973 profitability above
the
the firms with
understanding
(DR),
computed
and
and
The mean DT increased
However, the growth
page
rate in the case of the
twice that
of the
48
'High
took
was
about
group, indicating that far
'Low P73'
more diversification
group
P73'
place
in
more
profitable
this
change
firms than in less profitable firms.
Let us now examine the composition of
total diversification
that the
'High P73'
growth
was
pr
groups.
two
mean
edominentl y related in nature.
diversification being
the diversification
also
had
growth
notice
equal
and
mean
unrelated
the
little larger.
a
pattern
that
The
about
diversification
related
diversification,
unrelated
We notice
diversification
experienced
group, on the other hand,
in
groups, we
the
group
'Low P73'
growth that
of
in
If
between
compare
we
the
two
the growth race in mean
related diversification of the two is quite comparable;
but the growth rate in mean
of the
diversification
unrelated
more profitable group was about four times that
of the less profitable
indicates that
undertook far
the
more
profitable firms
group.
more
profitable
diversification
during
analysis,
This
the
seven
thus,
firms in general
than
the
less
year period under
study and that in particular, the former undertook
more unrelated
diversification than the latter.
combine this conclusion with our earlier
interesting conclusions seem to emerge.
far
If
we
observations,
page
49
The nature of interaction between
profitability that
analysis is represented in
1.
Briefly it
related
can
conjecture
can
we
diversification
a
diversification
is
from the above
diagramatic form
described
be
as
in
Figure
follows.
High
associated
in
due
course
prof itaoil ity levels, in turn, seem
with high
of
diversification.
unrelated
absolute
High
time.
be
to
high
with
profitability growth rate which leads to high
profitability levels
and
associated
However, as we
have seen, high unrelated diversification is negatively
associated with growth rates
process
would
dampen
then
in
profitability.
growth
the
rates
profitability resulting from the original high
diversification level.
If
this
This
related
process
complex
mutual causation occurs simul taneously ,as seems
in
to
of
be
the case, the actual differential between the levels of
profitability of
the
be smaller
could
than
high DR and high DU groups would
be
expected
diversification
association between
from
a
simple
profitability
and
growth rates.
The
interaction
between
diversification seems
be captured by
study attempted
a
to
profitability
be too complex
and dynamic to
simple static analysis.
to
the relationship of
and
The
present
take cognisance of this by studying
diversification
with
changes
profitability rather than with its absolute level.
in
The
page
results that
50
indicate
emerged
such an approach
that
could be quite fruitful.
We do
not
richness
pretend
of
the
captured
have
to
process
of
diversification and profitability.
successfully demonstrate
out relationships
in
that
this
measures of diversification.
useful beginning.
and extend
It
is
for
it
area
complete
the
interaction
between
did,
however,
We
is possible to
using
This study is
trace
quantitative
a
small but
future research to explore
the issues raised by this study.
page
51
BIBLIOGRAPHY
Ansoff,
H.I.,
York, 1965.
Corporate
Strategy,
McGraw-Hill,
New
Arnould, R.J., "Conglomerate Growth And Public Policy,"
in L.
Gordon ed
Economics of Conglomerate Growth,
Department of Agricultural Economics, Oregon State
University, Corvallis, Oregan, 1969, pp. 72-80.
.
,
Berry, C.H.,
"Corporate Growth
Journal of Law and
Economics,
371-383.
and
XIV
Diversification,"
(October
1971),
Corporate Growth and
Diversification,
.,
Princeton University Press, Princeton, 1975.
Campbell, D.T,
and
Stanley,
Experimental
J.C.
and
Quasi-Experimental Design For Research, Rand Mcnally,
Chicago, 1966, pp. 10-12.
Chandler, A.D.,Jr.
Strategy and
Structure:
Chapters
in the History
of the American Industrial Enterprise,
M.I.T.
Press, Cambridge, 1962.
FASB,
Statement of Financial Accountng Standards No. 14,
Financial Reporting for
Segments of a
Business
Enterprise, Stamford, Conn., December 1976.
Gorecki,
P.K.,
"The
Measurement
of
Enterprise
Diversification," The Review of Economics and Stastics,
56, August 1974, pp. 399-401.
Gort,M., Diversification and
Integration in American
Industry, Princeton
University Press, Princeton, N.J.,
1962.
Jacquemin, A. P., and
C.H. Berry,
"Entropy Measure of
Diversification and Corporate Growth," The Journal
of Industrial
Economics, volume XXVII, No 4 June
359-369.
1979, pp.
.
,
Markham, J.W.,
Conglomerage
Public
Enterprise
and
Policy,
Harvard
University
Press,
Cambridge,
Mass.
1973.
,
McVey,
Industral
Diversification
of
Manufacturing
Firms,"
Canadian
Stastical Review, XXXXVII (1972), 4,
112-117.
6,
J.S.,
"The
Mul tiestablishment
page
52
Study of Strategy,
"Cross-Sectional
Schoeffler, S.
the
PIMS
Aspects of
Performance:
Structure, and
Strategy
ed
+
B.
Thorelli,
Program," in
H.
=
University
Press,
Performance, Indiana
Structure
Bloomington, 1977, pp. 108-121.
,
.
,
Economic
and
Structure
Strategy,
Rumelt,
R.P.,
Harvard
Cambridge,
Performance,
University
Press,
Mass., 1974.
Salter, M.S., and
W. A. Weinhold
"Diversification via
Creating Value," Harvard Business Review,
Acquisition:
July-August 1978, pp. 166-176.
,
Market
Scherrer,
F.M.,
Industrial
Rand
Mcnally,
Economic Performance,
pp. 268-272.
and
Structure
and
Chicago,
1980,
Needs
., "Segmental Financial Reporting:
Business
Goldschimd,
ed
Tradeoffs,"
in
H.J.
need
to
know,
Disclosure:
Government's
McGraw-Hill, New York, pp. 3-118.
.
,
Seigel, S.
Nonpar ametr ic Stastics for the Behavioural
Sciences, McGraw-Hill, New York, 1956,
pp. 95-156.
,
Autonomy and Diversification,
Wrigley, L.
Divisional
Harvard
Doctoral
Dissertation,
Unpublished
Business School, 1970.
,
TABLE
1
PEARSON PRODUCT MOMENT CORRELATION
BETWEEN TOTAL DIVERSIFICATION AND PROFITABILITY
TABLE
2
SPEARMAN RANK ORDER CORRELATION
BETWEEN TOTAL DIVERSIFICATION AND PROFITABILITY
TABLE
3
RESULTS OF TESTS FOR HI'
TEST
TABLE
4
PEARSON PRODUCT MOMENT CORRELATION
BETWEEN RELATED DIVERSIFICATION AND PROFITABILITY
P73
DR73
P74
P75
P76
P77
P78
P79
TABLE
5
SPEARMAN RANK ORDER CORRELATION
BETWEEN RELATED DIVERSIFICATION AND PROFITABILITY
P73
DR73
P74
P75
P76
•
P77
P78
P79
TABLE 6
PEARSON PRODUCT MOMENT CORRELATIONS
BETWEEN UNRELATED DIVERSIFICATION AND PROFITABILITY
P73
DU73
P74
P75
P76
P77
P78
P79
TABLE
7
SPEARMAN RANK ORDER CORRELATION
BETWEEN UNRELATED DIVERSIFICATION AND PROFITABILITY
P73
DU73
P74
P75
P76
P77
P78
P79
TABLE
LOW DU73
NUMBER OF FIRMS
LOW
DR73
HIGH DR73
8
HIGH DU73
:
'
TABLE 9
RESULTS OF TESTS FOR H2
TEST
l.t.Test
STASTIC
VALUE 1973
VALUE 1979
-0.465
0.999
1.708
0.155
1.393
2.71
2.
Median Test
3
Mann- vVhitney
U Test
70
62
Wald-Wolfowitz
Runs Test
10
12
.
4.
CRITICAL VALUE*
51
NOTES
The null hypothesis in test 1 was that the mean profitability of
the 'High DR73 - Low DU73' group was equal to the mean profitability
of the 'Low DR73 - High DU73' group. The alternative hypothesis was
that the mean profitability of the former group was higher than that
of the latter. The null hypothesis could not be rejected at 0.05 level
of significance.
The null hypothesis in tests 2 and 3 was that the median profitability
of the 'High DR73 - Low DU73' group was equal to that of the 'Low DR73 High DU73' group.
The alternative hypothesis was that the median
profitability of the former group was higher than that of the latter.
The null hypothesis could not be rejected at a level of significance 0.05.
The null hypothesis in test 4 was that the distributions of profitibality
of the 'High DR73 - Low DU73' group and the 'Low DR73 - High DU73' group
did not differ at all.
The alternative hypothesis was that they differ
at least in some respect.
The null hypothesis could not be rejected
at 0.05 level of significance.
TABLE 10
PEARSON PRODUCT MOMENT CORRELATION
BETWEEN TOTAL, RELATED AND UNRELATED DIVERSIFICATION AND PROFIT GROWTH RATE
ns
*
**
***
****
*****
PG(74:73)
0.052
PG(75:73)
0.121
PG(76:73)
0.104
PG(77:73)
0.155
PG(78:73)
0.198
PG{79:73)
0.175
Not significant
Significant at 0.10
Significant at 0.05
Significant at 0.025
Significant at 0.005
Significant at 0.001
DU73
DR73
DT73
ns
ns
ns
ns
ns
ns
level
level
level
level
level
(The
(The
(The
(The
(The
0.338
-0.290
0.406
-0.292
0.398
-0.300
0.437
-0.292
0.493
-0.306
0.539
-0.374
critical
critical
critical
critical
critical
value
value
value
value
value
at
at
at
at
at
this
this
this
this
this
level
level
level
level
level
is
is
is
is
is
.260)
.330)
.388)
.453)
.496)
TABLE 11
SPEARMAN RANK ORDER CORRELATION
BETWEEN TOTAL, RELATED AND UNRELATED DIVERSIFICATION AND PROFIT GROWTH RATE
DT73
ns
PG(74:73)
0.020
PG(75:73)
0.099
PG(76:73)
0.128
PG(77:73)
0.115
PG(78:73)
0.143
PG(79:73)
0.164
Not significant
Significant at 0.10
Significant at 0.05
Significant at 0.025
Significant at 0.01
ns
ns
ns
ns
ns
ns
level
level
level
level
(The
(The
(The
(The
DR73
DU73
0.341
-0.322
0.397
-0.304
0.364
-0.244
0.451
-0.343
0.462
-0.328
0.481
-0.326
critical
critical
critical
critical
value
value
value
value
at
at
at
at
this
this
this
this
level
level
level
level
ns
is
is
is
is
0.260)
0.330)
0.388)
0.496)
:
TABLE 12
RESULTS OF STASTICAL TESTS FOR H3
TEST
STASTIC
l.t-Test
2. Median
3
Test
.Mann-Whitney
U Test
U
COMPUTED VALUE
CRITICAL VALUE SIGNIFICANCE LEV/.L
3.615
2.797
0.005
5.538
5.410
0.01
35
39
0.01
NOTES
The null hypothesis in test 1 was that the mean profit growth rate of the
'High DR73 - Low DU73' group and that of the 'Low DR73 - High DU73' group
were equals The Alternative hypothesis was that the mean profitability
growth rate of the former group was larger than that of the latter.
The null hypothesis was rejected at 0.005 level of significance.
The null hypothesis in tests 2 and 3 was that the median profitability
growth rate of the 'High DR73 - Low DU73' group was equal to that of the
The alternative hypothesis was that the
'Low DR73 - High DU73' group.
median profitability growth rate of the former group was higher than that
of the latter. The null hypothesis could be rejected at 0.01 level of
significance in both the tests.
TABLE 13
HIGH DR73-LOW DU73
Number of Firms
LOW
P73
HIGH P73
7
LOW DR73-HIGH DU73
TABLE 14
PERCENTAGE CHANGES IN MEAN DIVERSIFICATION
DURING THE PERIOD 1973-1979
GROUP
LOW
P73
DT
DR
DU
H
APPENDIX
CO
E-i
O
S
§
w
Cn
E-<
O
E-i
w
O
O
E-i
Eh
W
w
<
z
o
W
W
<:
2
O
H
2
H
Q
H
W
OQ
PQ
O
Z
o
H
Z
O
M
<:
tH
<:
H
Q
u
H
H
CO
M
Q
+
>
o
o
V
o
h:|
u
H
H
«
Pm
>
M
Q
72G7'^023
OC
5
'8
Date Due
Lib-26-G7
H028.M414 no.l218- 81
Palepu,
742713
3
Knshn/Diversif ication and pro
D*BKS
TDflD
001"""
002 000 TT7