waxwav

 

 

 

 

 

2%.. 4...:

513mm;
ra.‘

.
I

 

x ,
fink} in
53:350..)
£21553!
inf. 5.. ..
. x}! i. 37.:
I713: 3r:

355$Iau
:39 J!

a!

i
2...... .33.? 3;
graiimrréhﬂ
.1;

:3! U. “3.5!...

 

8
53.373... 5:351} .Inv

5:1... .

«(is-(5 .lul.‘_-a.zll‘l.5‘l$ull

31:21:.)

{2.2.5.2 ~

:-

.. i .izlesﬁl
. l..(!‘.£}f.5 .
$.31. 1...: I:

 

 

m...

ha
‘ :....§nﬁ§m Maw? ,.

l
(4

IFK

a; I
.31... bl.

 

5%.é}.l£: O...
whining

13.7231: .(319.
1555:5333-3
I. 2.11:1...
Ia!» ﬁt; III:
I... ’3

 

 

.. #33: 2.1..
(:64 t 6613... till!!!
515’: ’12::
{tiiizhlfinb

 

 

 

 

 

138815 LIBRARY
’5 Mkﬂnb-.u State
100“ University

 

 

 

This is to certify that the
dissertation entitled

INEFFICIENCES IN THE INTERNAL CAPITAL
ALLOCATION PROCESS - ANTECEDENTS AND
PERFORMANCE IMPLICATIONS

presented by

Mathias Arrfelt

has been accepted towards fulfillment
of the requirements for the

Ph.D degree In StrateL Management

 

 

WWW/W

Major Professor’s Signature W
3/3 / /80 0 9
’ /

Date

 

MSU is an Affirmative Action/Equal Opportunity Employer

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K1/Proj/Acc8-PreleIRC/DateDue tndd

 

INEFFICIENCIES IN THE INTERNAL CAPITAL ALLOCATION PROCESS -
ANT ECEDENTS AND PERFORMANCE IMPLICATIONS

By

Mathias Arrfelt

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Strategic Management

2009

l

ABSTRACT

INEFFICIENCIES IN THE INTERNAL CAPITAL ALLOCATION PROCESS —
ANT ECEDENTS AND PERFORMANCE IMPLICATIONS

By

Mathias Arrfelt

In this dissertation, I examine the internal allocation of investment capital across
business segments within the ﬁrm. Introducing a multilevel model that relies more on
narrower and less distant measures of ﬁrm performance, I predict that a number of
multilevel antecedents including performance aspirations, corporate governance
variables, and environmental uncertainty affect the efﬁciency of the internal capital
allocation process which, in turn, affects three important ﬁrm performance outcomes —
the diversiﬁcation discount, the corporate effect and ﬁrm ROA. Results suggest a
pervasive effect of performance aspirations on the efﬁciency of the allocation process as
well as a mainly positive effect of allocation efﬁciency on performance outcomes. Thus,
by combining four distinct literatures on capital allocation, ﬁrm performance
decomposition, diversiﬁcation, and the diversiﬁcation discount into a more integrated
framework for looking at the efﬁciency of the capital allocation process, this dissertation
should be viewed as an initial step in building knowledge about both antecedents to

allocation efﬁciency as well as performance implications of allocation efﬁciency within

the boundaries of the ﬁrm.

DEDICATION

To my patient parents and sister who provided much support on this eventful journey.

iii

 

ACKNOWLEDGEMENT

This dissertation beneﬁted from the advice and support of many. I thank the
members of my committee for their skill and hard work in steering this dissertation to a
successful conclusion. I am especially indebted to my mentor and friend, Bob Wiseman,
who not only eased my entry into the program, but also supported me every step of the
way on this journey. Great gratitude also goes out to fellow doctoral students, staff, and
other faculty members at Michigan State University who contributed to making this
difﬁcult endeavor so much more enjoyable.

Finally, none of this would have been possible without my wonderfully

supportive parents who insisted on us receiving our higher education in the Land of

Dreams.

iv

fill

ll rh-
Ll‘u'

ff.

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................... vii
LIST OF FIGURES ......................................................................................................... ix
INTRODUCTION .............................................................................................................. 1
CHAPTER 1: THEORETICAL MODEL DESCRIBING ANTECEDENTS TO
CAPITAL ALLOCATION EFFICIENCY ................................................................... 20
INTRODUCTION ................................................................................................................ 20
ANTECEDENTS TO THE CAPITAL ALLOCATION PROCESS ................................................... 25
Business segment historical and social performance comparisons ......................... 28
Agency-theoretic corporate governance variables .................................................. 33
Level of diversiﬁcation ............................................................................................. 42
Environmental uncertainty ....................................................................................... 45
Environmental uncertainty: moderation effects ....................................................... 52

CHAPTER 2: THEORETICAL MODEL DESCRIBING PERFORMANCE

IMPLICATIONS OF CAPITAL ALLOCATION EFFICIENCY ............................. 59
PERFORMANCE IMPLICATIONS OF CAPITAL ALLOCATION EFFICIENCY .............................. 59
Performance Implications: Corporate Effect ........................................................... 60
Performance Implications: Diversiﬁed ﬁrm value ................................................... 66
Performance Implications: F irm ROA ..................................................................... 68
CHAPTER 3: METHODS ............................................................................................. 71
SAMPLE ........................................................................................................................ 71
MEASURES .................................................................................................................. 76
Dependent Variables ................................................................................................ 76
Independent Variables .............................................................................................. 88
Corporate Governance ............................................................................................. 90
Environment ............................................................................................................. 92
Controls .................................................................................................................. 100
MODEL SPECIFICATION AND ESTIMATION ...................................................... 101
Antecedents to Capital Allocation Eﬂiciency ...................................................... 105
Performance Implications: Corporate Effect ........................................................ 110
Performance Implications: Diversified ﬁrm value and Firm ROA ........................ 117
CHAPTER 4: RESULTS ............................................................................................. 120
ANTECEDENTS TO CAPITAL ALLOCATION EFFICIENCY .............................. 120
Antecedents to overinvestrnent ............................................................................... 124
Antecedents to underinvestrnent ............................................................................. 128
PERFORMANCE IMPLICATIONS: CORPORATE EFFECT ................................. 132
Unconditional Variance Partitioning ..................................................................... 133
Hypotheses testing .................................................................................................. 138

PERFORMANCE IMPLICATIONS: DIVERSIFIED FIRM VALUE ....................... 145

Diversiﬁed ﬁrm value scaled by sales .................................................................... 145
Diversiﬁed ﬁrm value scaled by assets ................................................................. 148
PERFORMANCE IMPLICATIONS: FIRM ROA .................................................... 149
CHAPTER 5: DISCUSSION, LIMITATIONS, AND EXTENSIONS .................... 153
DISCUSSION: INTRODUCTION ......................................................................................... 153
Performance Implications on Diversiﬁed ﬁrm value ............................................. 154
Pen‘ormance Implications on Firm ROA ............................................................... 159
Performance Implications on Corporate Ejfects ................................................... 159
Antecedents to Capital Allocation Efﬁciency ......................................................... 162
LIMITATIONS AND EXTENSIONS ................................................................................... 165
REFERENCES ............................................................................................................... 178

vi

 

likki
Till!
lbkkl
luff?
ldkﬂ
lmki

lbkf

kaﬂ
Ilbit 3;

lbk9

Ibkl

Ink]

Tit]

Ink:

ENE

idkj
ENE

«bkl

LIST OF TABLES

Table 1: Observations by Segment and Firm Levels ....................................................... 75
Table 2: Measures ............................................................................................................ 99
Table 3: Intra-class Correlations (ICC) of Relevant Variables ...................................... 107
Table 4: Descriptive Statistics and Correlations .................................................... 121-122
Table 5: (A) Antecedents to Capital Allocation Efﬁciency — Overinvestment ............. 125
Table 5: (B) Antecedents to Capital Allocation Efﬁciency - Overinvestment .............. 126
Table 6: (A) Antecedents to Capital Allocation Efﬁciency — Underinvestment ........... 130
Table 6: (B) Antecedents to Capital Allocation Efﬁciency — Underinvestment ............ 131
Table 7: Performance Implications: Corporate Effects Variance Partitioning .............. 135
Table 8: Comparison of Effects with Selected Previous Studies ................................... 138

Table 9: Performance Implications: Corporate Effects Hypotheses Testing
(internal) ......................................................................................................... 139

Table 10: Performance Implications: Corporate Effects Hypotheses Testing
(external) ........................................................................................................ 141

Table 11: Performance Implications: Corporate Effects Hypotheses Testing
(overinvestment) ............................................................................................ 142

Table 12: Performance Implications: Corporate Effects Hypotheses Testing
(overinvestment) ............................................................................................ 143

Table 13: Performance Implications: Corporate Effects Hypotheses Testing
(underinvestment) .......................................................................................... 144

Table 14: Performance Implications: Corporate Effects Hypotheses Testing

(underinvestment) .......................................................................................... 144
Table 15: (A) Performance Implications: Diversiﬁed Firm Value (sales) ..................... 147
Table 15: (B) Performance Implications: Diversiﬁed Firm Value (assets) ................... 148
Table 16: (A) Performance Implications: Firm ROA (Underinvestment) ..................... 150

vii

 

hit 16:

libit 1'5.

Table l.':

Title l8

link 13

llhlt ll

Table 16: (B) Performance Implications: Firm ROA (Overinvestment) ....................... 151

Table 17: (A) Independent Sample T-tests (Antecedents to Overinvestment,

population) ..................................................................................................... 172
Table 17: (B) Independent Sample T-tests (Antecedents to Underinvestment,

population) ..................................................................................................... 172
Table 18: (A) Independent Sample T-tests (Antecedents to Overinvestment, limited

population) ..................................................................................................... 173
Table 18: (B) Independent Sample T-tests (Antecedents to Underinvestment, limited

population) ..................................................................................................... 173
Table 19: Independent Sample T-tests (Antecedents to Allocation Errors) .................. 174

viii

LIST OF FIGURES

Figure l: Antecedents to Capital Allocation Efﬁciency .............................................. 27
Figure 2: Performance Implications of Capital Allocation Efﬁciency ........................ 61
Figure 3: Illustration of Overinvestment Measures ..................................................... 80
Figure 4: Illustration of Underinvestment Measures ................................................... 84

ix

 

INTRODUCTION

In this dissertation, the internal capital allocation process is studied in a different
setting than what is customary in traditional ﬁnancial and economic research. I take a
behavioral perspective in re—examining the transaction cost economics (TCE) approach
(Williamson, 1975: 1985: 1996) to internal capital allocation including viewing the
capital allocation decision (i.e., the transaction itself) as the unit of analysis and
emphasizing behavioral assumptions (opportunism and bounded rationality) that have
traditionally been viewed as hindering complete market contracting and requiring ﬁrms
to use internal governance structures to mitigate inefﬁciencies in the external capital
allocation process.

According to Williamson, in order to mitigate such inefﬁciencies related to
underlying behavioral assumptions regarding both human behavior and environments,
diversiﬁed ﬁrms allocate capital using an internal capital allocation process - referred to
as a ‘miniature capital market’. That is, this internal capital allocation process substitutes
for an external process by transferring the responsibility of allocating capital from
external intermediaries (such as directly by capital markets, or by banks and other
ﬁnancial institutions) to the corporate headquarter and its agents. This will, Williamson
argues, give the multidivisional ﬁrm multiple advantages including better information to
base the capital allocation decision on, an internal competition for funds instead of funds
being automatically returned to their sources, and the opportunity to implement a more
sequential and thorough investment process relative to what may be feasible through the

capital market.

 

 

 

d~§
. .A

I...

\
an“...

‘1
3.
l ‘-
‘M‘

Acknowledging the clear delineation between the external capital allocation
process performed by external intermediaries and the internal allocation process
performed by knowledgeable internal managers made by Williamson and many ﬁnance
scholars (e.g., Scharfstein & Stein, 2000), I argue however that external and internal
markets are related in many ways and may compliment instead of substitute for each
other as much of the previous literature has assumed. That is, ﬁrms, regardless of
structure (single or multi-segment), seek investment capital from external capital markets
that they in turn reallocate to a number of potential uses within the ﬁrm, including
allocations to business segments, dividend payments, ﬁrm stock buy backs, payment for
mergers and acquisitions, as well as payment of corporate level expenses such as for
corporate headquarters and corporate salaries. Thus, this view of external and internal
markets for capital allocation conﬂicts with assumptions made by Williamson and the
ﬁnancial literature in at least two ways. First, ﬁrms are not passively allocated capital to —
they actively seek different amounts as well as types of investment capital based on, at
least to some extent, the attractiveness of their uses for such capital. Second, prior work
on capital allocation (e.g., Williamson, 1975) effectively excludes alternative uses for
investment capital (besides allocations to business segments) such as those described
above.

In fact, it is the second assumption that enables the clear delineation between
internal capital markets and diversiﬁed ﬁrms on the one hand, and external markets and
single-segment ﬁrms on the other: single-segment ﬁrms do not need an internal allocation
process because all capital investment is assumed to go to the single segment, while

diversiﬁed ﬁrms use the internal structure to allocate capital between their business

segments. Thus, with a less restrictive view of how ﬁrms use their investment capital, 1
lessen the distinction between internal and external markets and allow single-segment
ﬁrms to have internal allocation structures. That is, even single-segment ﬁrms have '
control over both the capital they seek from capital markets as well as how they use such
capital. This extends Williamson’s model of capital allocation to also include single-
segment ﬁrms previously viewed as not having an internal allocation mechanism.
Notwithstanding that the potential beneﬁts of an internal allocation process
described earlier may also apply to single-segment ﬁrms, I argue that the internal market
for capital allocation often put in place to attenuate opportunistic contracting under
conditions of bounded rationality is not itself immune to those same influences that create
inefﬁciencies in the external market for capital allocation. In fact, since ﬁrms may make
the switch from traditional U-form structures (functional structure) to M-form structures
for a number of reasons other than to achieve efﬁcient capital allocation (including
cheaper dispute resolution, more effective audits, and a beneﬁcial division of labor
between headquarters and divisions that allow each to focus on speciﬁc duties), we
cannot conclude that the internal capital allocation process is efﬁcient at all times, or
even that it is always the most efﬁcient way of allocating capital'. That is, market failures
where capital is misallocated as a direct result of both human and environmental
characteristics also applies to the internal market for capital allocation and thus also to

the internal capital allocation process in single-segment and diversiﬁed fums.

 

‘ Williamson. on the other hand, assumes that the capital allocation decision is always performed at the
highest possible level of rationality (still assuming bounded rationality) by giving agents perfect foresight
to costlessly switch governance structure (from external to internal and vice versa) once they believe one
structure may be beneﬁcial over the other in response to upcoming contracting hazards. This is what he
means by “semistrong form” (1996: 9) of rationality.

v .
fa“ .P;
dub“ .“

"“ '.

.~.

I“)-

>I‘IA

 

.‘ I.
r' a"
I
N ‘VI.
..
”,-J
“--V~

n‘.~ - .
I

L—

a"!

 

. ,,
|’(~ .
A

Therefore, to account for such human and environmental characteristics that may
Iffect the efﬁciency of the internal capital allocation process, I look to a number of well-
mown theories of human behavior and decision-making including Prospect Theory
;Kahnemann & Tversky, 1979), Behavioral Theory of the Firm (Cyert & March, 1963),
Agency Theory (Jensen & Meckling, 1976: Fama, 1980: Fama & Jensen, 1983: Jensen,
1986), as well as to theories (e.g., Duncan, 1972: Dess & Beard, 1984) describing the
:nvironment’s impact on decision making, in particular on its rationality. The listed
heories support a number of business segment, ﬁrm and industry level antecedents that
are posited to directly affect the efﬁciency and rationality of the capital allocation
decision.

Furthermore, to account for and emphasize the increased importance of the
corporate ofﬁce for ﬁrm performance that an internal allocation process varying in
efﬁciency implies, 1 model the efﬁciency of the allocation process as affecting important
ﬁrm performance outcomes including the corporate effect, the diversiﬁcation discount2 (I
will refer to the diversiﬁcation discount as ‘diversiﬁed ﬁrm value’ from here on unless
speciﬁcally referencing prior work within the diversiﬁcation discount literature) and ﬁrm
ROA. The corporate effect, in particular, should reﬂect this increasing importance of the
corporate ofﬁce by measuring how much value the corporate ofﬁce creates or destroys by
its actions including decisions to allocate capital. In sum, therefore, I argue that a number
of multilevel antecedents affect the efﬁciency of the capital allocation process which, in

turn, affects important measures of ﬁrm performance.

 

! Please note that the diversiﬁcation discount, because of its conceptual construction, can only include
effects of capital allocation efﬁciency for diversiﬁed ﬁrms, and is therefore constrained to only include
such effects.

Overall, by incorporating a number of multilevel antecedents and performance
outcomes into the study of the internal capital allocation process, this dissertation brings
together a number of distinct literatures on ﬁrm performance decomposition, on
diversiﬁcation, on the diversiﬁcation discount, and on capital allocation. Each literature
offers unique insights, opportunities and challenges as it relates to the capital allocation
process and its efﬁciency. First, the ﬁrm performance decomposition literature (e.g.,
Bowman & Helfat, 2001: Adner & Helfat, 2003: Misangyi, Elms, Greckhamer & LePine,
2006: Brush, Bromiley & Hendrickx, 1999: McGahan & Porter, 1997: Hough, 2006)
answers the question of where ﬁrm performance is generated, i.e., by the industry,
corporate ofﬁce, or by the business segments themselves (see Bowman & Helfat, 2001,
for a review of how ﬁrm performance is decomposed into industry, corporate and
business segment effects). Although estimates vary of the relative importance of industry,
corporation and business segments for ﬁrm performance, with the importance of
corporate inﬂuences on ﬁrm performance especially contentious3, a portion of that
performance is generated by the corporate ofﬁce and is therefore related to important
corporate-level processes such as the capital allocation process (Bowman & Helfat, 2001:
Misangyi er al., 2006). That is, the more efﬁciently the corporate ofﬁce allocates scarce
resources such as capital, the more value it may be able to add to the business segments,
thereby increasing the corporate contribution to ﬁrm performance as reﬂected in higher

empirical estimates of such contributions, i.e., higher corporate effects.

 

3 One group of ﬁrm performance decomposition studies points to very small corporate effects of around 4
percent (Schmalensee, 1985: Rumelt, 1991: McGahan & Porter, 1997), while another group of studies ﬁnd
substantially larger corporate effects of between 9 and 28 percent (Brush, Bromiley & Hendrickx, 1999:
Roquebert, Phillips & Westfall, 1996: Hough, 2006). This lack of agreement reﬂects disagreement over
both which methods are appropriate as well over what the resulting empirical estimates (corporate effects)
actually mean (Bowman & Helfat, 2001).

Second, the literature on diversiﬁcation discounts (e.g., Stein, 1997: Maksimovic
a; Phillips, 2002: Berger & Ofek, 1995: Servaes, 1996: Lang & Stulz, 1994) is also
related to this dissertation and to the capital allocation process. Offering a comparison
between actual ﬁrm values of diversiﬁed ﬁrms and their hypothetical values if all
business segments instead were freestanding, the diversiﬁcation discount literature has
consistently shown that business segments are worth more freestanding than under the
corporate umbrella4 (e.g., Berger & Ofek, 1995: Servaes, 1996: Lang & Stulz, 1994).
Therefore, based on the ﬁnding that ﬁrm values ﬂuctuate based on level of
diversification, a tie between capital allocation efﬁciency and the diversiﬁcation discount
may Seem likely whereby inefﬁcient allocation of capital across business segments may
help to explain observed discounts.

Finally, the literature on diversiﬁcation (e.g., Palich, Cardinal, & Miller, 2000:
Montgomery & Wemerfelt, 1988: Bowen & Wiersema, 2005) is also related to this
dissertation and the internal capital allocation process because increased diversiﬁcation
may increase the difﬁculty and salience of the capital allocation function. In other words,
by diversifying into related and/or unrelated businesses, the allocation decision becomes
more uncertain as corporate managers, limited in their ability to understand, process, and
digest information, are facing an increasing amount of diverse information to gather,
analyze and act upon. In addition, the salience of the capital allocation function may also

merease as the number of allocation decisions as well as the impact of formal allocation

7\
some argue (Campa & Kedia, 2002: Villalonga, 1999), however, that this discount may be partly due to
e endogeneity of the diversiﬁcation decision. That is. and as also discussed in more detail in the methods

secﬁon. ﬁrms may diversify in response to speciﬁc ﬁrm characteristics (such as poor performance) that
may make the opportunity cost of diversifying less.

rules often found in organizations (e.g., Bower, 1986: Shin & Stulz, 1998) both rise with
increasing levels of diversiﬁcation.

Against the background that the capital allocation function is related to, and has
much in common with each of the three described literatures, it is surprising that so
relatively little attentions has been paid to this potentially very important corporate-level
process within the ﬁeld of strategic management. In fact, the capital allocation process
becomes a strategic question of fundamental importance as capital sought directly from
capital markets as well as from banks and other ﬁnancial intermediaries is effectively
reallocated to important internal uses including allocations to business segments. Thus,
the fundamental function of effectively allocating scarce resources, which otherwise
could have been carried out by banks and other ﬁnancial intermediaries (e.g.,
Williamson, 1975), is internalized. That is, for both single-segment and diversiﬁed ﬁrms,
the burden of allocating capital to internal uses including business segments is effectively
tralISfet-red from capital markets to the corporate ofﬁce.

Also supporting the overarching importance of the allocation of ﬁnancial
resources for the ﬁrm, Donaldson (1984) argues that this allocation of scarce ﬁnancial
resources reﬂects the strategic priorities of the ﬁrm by entailing critical risky strategic
dads i011s: “The most critical choices top management makes are those that allocate
reSources among competing strategic investment opportunities” (95). As such, the
allocation decision becomes a core attribute of the diversiﬁed ﬁrm and a core activity of
its col‘porate officers reﬂecting its importance for overall ﬁrm performance: “In many

reSPeCts, this assignment of cash ﬂows to high yield uses is the most fundamental

\

s
A Search for various derivatives of ‘capital allocation’ in major US. management journals returned fewer
8 Such studies.

attribute of the M-form enterprise” (Williamson, 1975: 148). Thus, based on the
fundamental strategic importance of the allocation decision in diversiﬁed ﬁrms, the
minimal attention given to capital allocation within strategic management contrasts
sharply with the extensive coverage afforded in ﬁnance and economics. In those areas,
the degree of optimality of the capital allocation process has been shown to affect both
the performance of diversiﬁed ﬁrms and the observed differences in valuation between
freestallding business segments and similar segments held under the corporate umbrella,
Leo, tlle ‘diversiﬁcation discount’ (e.g., Stein, 1997: Maksimovic & Phillips, 2002: Berger
&. Ofek, 1995: Servaes, 1996: Lang & Stulz, 1994), referred to as ‘diversiﬁed ﬁrm value’
in this dissertation. In addition, the capital allocation process has been found to be
iIlﬁffieient (e.g., Lamont, 1997: Shin & Stulz, 1998: Rajan, Servaes, & Zingales, 2000) as
capital often is allocated relatively independent of business segments’ prospects.

This relative lack of attention to the capital allocation process within strategic
management is also a surprise because research on capital allocation may have stronger
perfol‘lrnance implications in the US. as compared to countries with institutional
atTangements that emphasize an allocation process based less on future prospects than on
available internal capital (Thomas 111 & Waring, 1999). That is, because US. ﬁrms tend
to allOcate capital based on future prospects of business segments to achieve “adequate
re“-ll‘tls on equity investments or face attacks from equity investors that can culminate in
proxy battles and hostile takeovers” (Thomas [H & Waring, 1999: 735), the allocation
1”meeSs should logically be much more strongly related to ﬁrm performance as compared

to countries such as Germany and Japan, where investments are allocated more based on

available internal capital and a broader bargaining process involving a number of long-

teml partners with power to inﬂuence allocations, including labor, banks, and

governments. Thus, the potential value of an efﬁcient capital allocation process may be

higher for ﬁrms in the US. than for ﬁrms in countries with different institutional

environments.

Considering the fundamental importance of the internal capital allocation process,
its close ties with a number of other literatures, and the lack of attention to this process
within strategic management, my model of the internal capital allocation process in ﬁrms
has the promise to offer insight and clariﬁcation into a number of issues raised (or
omitted) by existing research within the area of capital allocation as well as by research
within the other three related literatures (ﬁrm performance decomposition,
divers iﬁcation, and the diversiﬁcation discount). First, a number of antecedents to the
efﬁCiencyé of the capital allocation mechanism are investigated. Prior research on capital
allocation within ﬁnance and economics (e.g., Scharfstein & Stein, 2000: Stein, 1997:

Raj an, Servaes & Zingales, 2000) has largely ignored antecedents and looked strictly at
the allocation process as a given, by assuming (varying levels of) optimality through ad
hoc assumption. Thus, a key gap in existing research on capital allocation is how
antec<=dents to a capital allocation process varying in efﬁciency have largely been
overlooked. That is, we currently know very little about _w_hy misallocation of resources
takes place in organizations. My model, therefore, extends existing research by viewing
Camt-“=11 allocation as a decision-making process and investigating a number of

behaviorally based antecedents to its efﬁciency. In addition, by relaxing the delineation

\

e
aNotef-hat the use of ‘efﬁciency’ does not imply absolute optimality of the capital allocation process, only
aglatl‘le comparison between ﬁrms’ allocational efﬁciencies based on instances and magnitudes of over

fe umlerinvestment. In other words, a ﬁrm’s allocational efﬁciency approaches optimality as there are
We: and fewer instances of over and underinvestment.

 

between internal and external capital allocation processes and allowing single—segment
Fri-111$ to have internal capital allocation processes, I can investigate antecedents to the

internal capital allocation process also in single-segment ﬁrms.

In investigating antecedents to a capital allocation process varying in efﬁciency, I
assume that corporate headquarters regardless of allocation process (see Bower, 1986, for
a more formal discussion of the capital allocation process in organizations) have a ﬁnal
say (i-e., control rights in Stein’s (1997) language) with regards to allocations, have the
right to ‘winner-pick’ (Stein, 1997) by transferring investment funds from relatively less
well-performing segments to those performing better, as well as have the right to allocate
investment capital to other uses such as dividend payments, buybacks of stock, and to pay
for acquisitions. The assumption that the corporate ofﬁce has the right to winner pick,
apart from being fairly reasonable and often found in studies published in ﬁnance,
economics and other areas that model the mathematical relationship between the
efﬁCierIcy of the capital allocation process and ﬁrm performance (e. g., Stein, 1997:
Maks imovic & Phillips, 2002: Rajan, Servaes & Zingales, 2000), also conﬁrms that the
allocation process is at its heart a decision-making process. It follows, therefore, that the
capital allocation process, as an example of a strategic organizational decision-making
Pr0ces s, may be susceptible to various inﬂuences or drivers at different levels in the
organization.

A number of such multilevel drivers of behavior have been shown to affect
decision-making processes in organizations. Greve (1998), for example, showed that
Social and historical aspiration levels affect behaviors at the individual level. Bethel and

Liebesltind (1993) described how managerial ownership incentives affected decisions at

10

the corporate level, while Goll and Rasheed (1997) found that environmental muniﬁcence
and dynamism at the industry level also affected decisions in the organizational context.
Thus, following current research and recognizing that antecedents at business segment,
corporate, and industry levels may affect organizational behavior and decision-making, I
structured the current investigation to look at antecedents at all levels that could affect,
are thought to be inherent in, or are thought to capture the capital allocation process —
antecedents at business segment, corporate and industry levels.

Therefore, I will investigate four sets of drivers of behaviors at three different
levels : at the business segment level, I am investigating historical and social segment
Performance comparisons capturing the tendencies of loss-averse decision makers to
PTOtCCt gains and try to reverse losses. At the corporate level, I am investigating a number
9f cOrporate governance variables - block holder ownership, institutional shareholder
Omership, and contingent compensation that together capture how agency-theoretic
Prescriptions try to control self-interested and risk-averse decision makers in the context
0f the capital allocation process. Also at the corporate level, I am investigating how level
0f leersiﬁcation may capture additional complications in the allocation process
St"ﬂitting from a lack of overview and expertise from diversiﬁcation efforts into
i“Cl'eas ingly unrelated areas. In addition, environmental uncertainty at the industry level
(Dess & Beard, 1984) may have a direct effect on the efﬁciency of the allocation decision
as it Captures dimensions of the task environment that may increase its uncertainty.
Finally, environmental uncertainty may also have a moderating effect on some of the

described relationships. In particular, it may exacerbate the effect of increased levels of

11

diversiﬁcation as well as reduce the effectiveness of corporate governance variables in

their respective relationships with capital allocation efﬁciency.

In sum, by investigating multilevel antecedents to the capital allocation process
and relaxing strict assumptions of rationality and efﬁciency so often applied to research
within finance and economics (Bromiley, 2005), this dissertation identiﬁes a key gap in
prior research on capital allocation by offering unique insight into what actually accounts
for different degrees of efﬁciency of the allocation process. In addition, by relaxing
previous strict assumptions made as part of most research on capital allocation thereby
allowing single-segment ﬁrms to have internal allocation processes, antecedents to
allocation efﬁciency apply also to ﬁrms with only single segments.

Second, this study addresses the ﬁrm performance decomposition literature by
Providing guidance on what accounts for or produces corporate inﬂuences on ﬁrm
Perfortnance, i.e., the corporate contribution. Any such guidance is currently missing
from that literature. In fact, the variety of single-level models most often used in
decomposition research do not permit the investigation of antecedents to any of the three
decomposed effects as measures of business segment, corporate and industry inﬂuences
on firth performance (see Misangyi et al., 2006, for a discussion of why only multilevel
Spe<3ifications allow such an investigation). This lack of guidance implies that we are not
Sure how, or even if, an important corporate-level process, such as capital allocation,
affects ﬁrm performance. Without such knowledge, the natural connection between
antecedents to the capital allocation process, the capital allocation process itself, and ﬁrm

perfoﬁnance remain largely unknown. Thus, with a better understanding of how the

capital allocation process affects ﬁrm performance outcomes, the study of antecedents to

12

the capital allocation process as well as policy prescriptions targeting the capital
allocation process can be justiﬁed and become more meaningful.

Acknowledging this omission by the literature and responding to calls for
research into what accounts for or determines individual decomposed effects (Bowman &
Hel fat, 2001: Misangyi et al., 2006), this dissertation, therefore, will investigate how the
capital investment allocation process is related to the corporate contribution as measured
by corporate effects. Speciﬁcally, I posit that the capital investment allocation process is
directly related to the corporate effect by measuring how effective the corporate ofﬁce is
in allocating capital among its potential uses including business segments. Said
differently, this dissertation measures how much value the corporate ofﬁce adds by
allocating capital in a relatively more efﬁcient manner. This relationship is not only
intuitive, it follows logically based on the speciﬁc segment ROA decomposition process
used in most studies (e.g., Brush, Bromiley & Hendrickx, 1999: McGahan & Porter,
1997 2 Hough, 2006) where overall ﬁrm ROA is used as a proxy for the corporate effect
as individual business segment ROAs are regressed on this proxy. In other words, if fums
alloCzlte relatively more capital to segments with less favorable prospects and lower
ROA. we would expect overall ﬁrm ROA to be negatively affected, leading to a lower
col'p<>l‘ate contribution as reﬂected by a lower corporate effect. Therefore, inefﬁciencies
in the allocation process as found by a number of researchers (e.g., Berger & Ofek, 1995:
Dav id et al., 2006) may reduce the corporate contribution and the corporate effect below
what it should be with a more efﬁcient allocation of capital. Thus, this study creates the

connection between two related literatures (the capital investment allocation and the ﬁrm

performance decomposition literatures) that Brush, Bromiley and Hendrickx (1999: 542)

13

alluded to in their study. At the same time, Brush, Bromiley and Hendrickx (1999) also
sug gest inefﬁcient capital allocation as one possible explanation for the small corporate
effects often found in the latter literature.

Also addressing the ﬁrm performance decomposition literature, a third
contribution of this dissertation focuses on empirical estimates of the corporate
contribution: corporate effects. Although the logic underlying associating a corporate-
level process (the capital allocation process) with a measure of how effective the
corporate ofﬁce is in adding value is clear, a number of researchers have pointed out
methodological problems with corporate effects that may confound or distort the
relationship between corporate-level processes and the true contribution of the corporate
level to ﬁrm performance, the corporate contribution (e.g., Bowman & Helfat, 2001:
Adner & Helfat, 2003: Misangyi et al., 2006).

These concerns focus on how sensitive the attribution of variance between
bus iness segment and the corporate level is to dispersion in business segment retums.
They argue that empirical estimates of corporate contributions are likely to be biased by a
high. but inappropriate, sensitivity to the patterns of returns across a ﬁrm’s business
Sﬁglllﬁitnts. That is, ceteris paribus, at relatively higher levels of business segment return
homogeneity, some of the variance that otherwise would be attributed to the business
StagIlent effect may now instead be attributed to the corporate effect, in effect overstating
it relative to the corporate contribution. At higher levels of business segment return

disPel‘sion, the opposite is likely to happen as corporate variance instead is attributed to

the buSiness segment effect. To incorporate those concerns, multiple hypotheses are

14

offered that acknowledge that an unmeasured construct7, the variance of business
segment returns within ﬁrms, may in fact have a strongly distorting or confounding effect
resulting in a null or even a positive relationship between inefﬁciencies in the capital
allocation process and empirical estimates of the corporate contribution at low levels of
return variance. To estimate this distorting or confounding effect and achieve a more
satisfactory measure of the ‘true’ corporate contribution, I examine and measure the
variance of business segment returns within each ﬁrm.

A fourth contribution of this dissertation is how it ties the efﬁciency of the capital
allocation function to the level of ﬁrm diversiﬁcation. Notwithstanding that level of
divers iﬁcation is one antecedent to the efﬁciency of the capital allocation process
suggesting a negative impact of diversiﬁcation on capital allocation process efﬁciency,
the use of diversiﬁed ﬁrm value as a ﬁrm performance outcome provides an additional
connCCtion between capital allocation and diversiﬁcation. That is, although single-
SEgnlerit ﬁrms in my model are assumed to have internal allocation mechanisms, the
efﬁciency with which diversiﬁed ﬁrms allocate capital can still be estimated by looking
at diversiﬁed ﬁrm values as the difference in value between free-standing business
seglnents and those under the corporate umbrella.

Although many studies have looked at diversiﬁcation discounts (e.g., Berger &
Ofek. l 995: Lang & Stulz, 1994: Comment & Jarrell, 1995) and some have looked at
how eaPital allocation efﬁciency affects the diversiﬁcation discount (e. g., Berger & Ofek,
1995), few have looked at underinvestrnent and none has looked at antecedents to

mlsauoCations of capital. Thus, by more carefully measuring inefﬁciencies in the capital

\

7 A Sea:-
study thch in major management journals failed to ﬁnd even a single ﬁrm performance decomposition
in eXplicitly measures the dispersion of business segment returns.

15

 

31 location process, i.e., measuring both over and underinvestment, as well as more fully
investigating behavioral antecedents to each, a more complete picture may emerge of
how capital is allocated in ﬁrms and how inefﬁciencies in this process are tied to the level
of diversiﬁcation through diversiﬁed ﬁrm values. This puts M-form prescriptions
emphasizing the efﬁciency8 of the internal capital allocation function - diversiﬁed ﬁrms
have more ﬂexibility in accessing capital (because they can also access internal sources)
(Lang <9: Stulz, 1994: Stulz, 1990) as well as superior access to information helping to
optimize allocation decisions (Schleifer & Vishny, 1986: Servaes, 1996: Williamson,
1986) —- against current evidence starting to point to M-forrn ﬁrms allocating relatively
more capital to weak lines of business than their stand—alone counterparts (Rajan et al.,
2000), M—form ﬁrms being insensitive to the prospects of their divisions in allocating
in"CSICInent funds (Rajan et al., 2000), as well as them being sensitive to cash flows of
unrelated divisions (Scharfstein & Stein, 2000).

In sum, therefore, with the caveat that even single-segment ﬁrms may have an
inter”llal capital allocation function, this dissertation may be able to speak to the
effecti‘Ieness of M-form prescriptions related to the allocation of capital. In addition, by
inveStigating likely antecedents to this internal allocation process, I may also be able to

determine why current evidence has shown diversiﬁed ﬁrms to be poor at allocating

Capital .

\
a
miggugh it could be argued that observed discounts may be a function of costly internal capital allocation
First, mus that are not able to add value in proportion to their costs, this is unlikely for a number of reasons.
of the toe capital allocation function is often not a separate function instead performed by regular members
exmmal p management team (Bower, 1986), thus is unlikely to add substantial additional costs. Second,
on to d e se apital allocation also carries costs in the form of transaction costs as capital is sourced and passed
el‘ving free-standing ﬁrms.

16

As a ﬁfth contribution, I provide guidance to the controversy within the ﬁrm
performance decomposition literature regarding which method is most appropriate to
model the inherently multilevel phenomena of decomposed ﬁrm performance. Several
methods have been suggested and used by different researchers including nested
AN OVA and variance components analysis (VCA), as well as the more recent use of
multilevel speciﬁcations including models used by Hough (2006) and Misangyi et al.
(20()6) - Although a number of speciﬁc issues, such as for example the lack of reliability
and insensitivity to small effects in VCA, have been identiﬁed with the former two
methods (Brush & Bromiley, 1997), the main drawback of the two single-level methods
is that they assume independence between effects. That is, single-level models assume
away the endogeneity that may for example exist as corporate parents, skilled at adding
value to business segments, also are effective at selecting attractive industries in which to
comPete. Thus, this assumption not only ﬁts poorly with underlying theories and data
describing the nested structure of business segments, ﬁrms and industries, but it also
effectively makes it impossible to advance theory and empirical work by investigating
relationships between individual effects as well as investigating speciﬁc factors as
anteK-‘redents to individual effects. In other words, multilevel speciﬁcations by controlling
for, ins tead of assuming, independence enable extensions to theory by accounting for
relationships among effects as well as for relationships within those effects. Applied to
the Context of my study, a multilevel approach becomes a necessary tool to investigate

multileVel antecedents to the corporate effect.
I:inally, since the use of ﬁrm performance has been criticized as being too far

rem . . . . .
oved from corporate strategies and actions to reﬂect “distmctrve advantages at the

17

a» '9‘.
ﬁ .

-31‘
”uh m

'f.‘ -
n.

process level” (Ray, Barney & Muhanna, 2004: 23), both the use of a corporate level
process, the allocation of investment capital, as well as the use of the corporate effect as a
measure of the corporate contribution may alleviate some of that criticism. In particular,
the use of the capital allocation process (as an alternative to using ﬁrm performance as
dependent variable) is in accordance with suggestions by Ray, Barney and Muharma
(2004) to alleviate concerns with differentiating between competing “competitive
advaxltages in some business activities and competitive disadvantages in others” (Ray,
Barney & Muhanna, 2004: 24). In other words, using ﬁrm performance we may not be
able to differentiate between strategies and actions that increase ﬁrm performance and
those that do not. In addition, because if its high level of aggregation, intervening
‘POWer’ (Westphal, 1998) and other processes may be more likely to distort and add
noise to ﬁrm performance relative to closer processes. Therefore, we may have a better
Chance of picking up on their (governance variables) true effect by using a measure of
“the effectiveness of business processes” (Ray, Barney & Muhanna, 2004: 24), i.e., the
efﬁciency of the capital allocation process, which is likely to be more sensitive to ﬁrm
strategies and decisions than overall ﬁrm performance.

In addition, using the corporate effect as a measure of the corporate contribution
should enable a more precise relationship between corporate governance variables, the
Capital allocation process and ﬁrm performance with intervening industry and business
segment effects excluded from the outcome measure. In other words, increased alignment

as a 1' es ult of better corporate governance should be more likely to increase performance
at the Corporate-level, thus primarily affecting the corporate effect. In sum, the use of a

I .
ess dls tallt process (allocation of capital) and a more narrowly deﬁned ﬁrm performance

18

 

measure (the corporate effect) may increase the likelihood of ﬁnding true and precise
relationships between corporate governance variables and ﬁrm performance outcomes.

In summary, with the goal of providing a better understanding of the internal
capital allocation process, including its antecedents and effects on both diversiﬁed ﬁrm
values, the corporate effect as well as on ﬁrm ROA, this dissertation is organized as
follows. In Chapter 1, I discuss antecedents to the degree of efﬁciency of the capital
allocation process including business segment historical and social performance
comparisons, a number of corporate governance variables, environmental uncertainty,
and level of diversiﬁcation. Hypotheses are then advanced tying antecedents to the
degree of efﬁciency of the allocation process.

In Chapter 2, I discuss how the degree of efﬁciency of the capital allocation
Process affects the corporate effect, diversiﬁed ﬁrm value as well as ﬁrm ROA. Several
hypotheses are then advanced outlining possible relationships between the capital
alloCation process on the one hand, and the corporate effect, diversiﬁed ﬁrm value and
ﬁrm ROA on the other.

In Chapter 3, I describe the testing of my multilevel model. Sampling, data
conanion, measures and data analytical decisions are all discussed in detail.

In Chapter 4, I review the ﬁndings including the extent to which the hypotheses
are SuPported. In Chapter 5, ﬁnally, I discuss the ﬁndings themselves in detail including
What they may mean for the literatures on ﬁrm performance decomposition, capital
allocation, diversiﬁcation, and diversiﬁed ﬁrm value as well as the overall conclusions

that may be reached based on them.

19

 

U

Introducﬁ

”

A“!

I
u:vp'h;

5&45 u. 1.03

‘
I" ‘2'- ‘1
bu-‘rbnAN‘
o

. I
8— I." '

an- ab ..

..

 

r. ‘ .‘
mums.“ i
.

T!

u, __5 :26.

., ‘
\" .
‘
~““u.§
h
h"~v‘, I.
«a. Ml,
‘ \
II-
.t I
wk 1»
l, ‘
. W9

1'. ‘
¢~:n;\aﬂ
l.

 

E39,.

. “la‘

 

CHAPTER 1: THEORETICAL MODEL DESCRIBING ANTECEDENTS
TO CAPITAL ALLOCATION EFFICIENCY

Introduction

‘ The central argument of this dissertation is that capital allocation plays a critical
role in the performance of diversiﬁed (M-form) and single-segment ﬁrms, and that the
efﬁciency of this process is related to a variety of characteristics at the business segment,
ﬁrm and industry levels. Thus, the key construct in this study is capital allocation
efficiency, which I deﬁne as the degree of over and underinvestment in individual
business segments.

Noting that capital investment is a forward looking process that should be based
on business segment prospects, not their past results, overinvestment is characterized as
the allocation of capital to business segments with relatively worse prospects.
SPe<=ifically, if more capital is allocated to business segments with relatively worse
prospect“ than those with relatively better prospects, overinvestment is implied.

Following Jensen (1986), overinvestment is a potential source of value loss. That
is, fifths that are cash rich are likely to invest at least a part of that excess cash in
unprofitable investments thereby reducing the value of the ﬁrm. In later work, Jensen
(1989. 1991, 1993) also notes a tendency of conglomerates to cross-subsidize, i.e., an
i“creased tendency (as compared to stand-alone ﬁrms) to invest in failing business
Segments. Both are characterizations of inefﬁciencies where too much capital is being
allocated to relatively underperforming business segments resulting in a lower ﬁrm value.

Underinvestment, on the other hand, is characterized as the relative lack of

uwashnent in business segments with better prospects. Although effectively the ﬂip-side

20

\‘ ":4 7"
.t 4‘ ‘

“9"
I?“

.,.....\ ..

-a~-.\ 3"
..,.- 5..

\t.......
i

5...:

'3.
an“

‘F 9' “pct
. I .
‘ be “\-

q
" U. \‘.

‘

n“."~ IA

3-;
r. v

u“.-

C.‘.
I ~
s...‘ u

5

¢..
. .

"L

" V
15“:
5

~.
"‘1‘;.

'~ .
9-;~A
h
\a u
.H-.‘ . b
5‘. a
I
. a

 

 

of the overinvestment construct, the underinvestment construct has not received much
attention in the literature. Together, however, over and underinvestment more fully
represent the degree of efﬁciency of the capital allocation process in that inefﬁcient
investment can both be because of ‘too much’ and ‘too little’ investment in business
segxnents with relatively worse and better prospects, respectively.

The importance of the capital allocation process for diversiﬁed ﬁrms is suggested
by Williamson’s defense of M-form ﬁrms. Extending arguments by Coase (1937),
Williamson (1975) advocates the multidivisional (M-form) of organization for efﬁciency
reasons because the separation of duties between headquarters and divisions is thought to
lower transaction costs by enabling headquarters to focus on important overarching
functions such as overall strategy, control questions and allocation of capital, while
divisions can focus mainly on business level strategies and operations. Williamson points
to a number of beneﬁcial outcomes of such separation, including a more efﬁcient internal
capital market where inefﬁciencies in external markets can be overcome through internal
funding- Speciﬁcally, by internalizing transactions of capital, he argues, corporate
managers may have more complete and better information than external providers of
Capital: “division managers are subordinates: as such, both their accounting records and
backup ﬁles are appropriate subjects for review. . ..the general ofﬁce can expect
knowledgeable parties to be much more cooperative than can an outsider. . .intemal
discloSure is afﬁrmatively regarded as necessary to the integrity of the organization and is
rewarded accordingly” (Williamson, 1975: 146). Thus, in line with the logic underlying
his W1itings, it could be expected that internal managers are better able to identify and

eValuate business segment prospects, therefore making more good investments and being

21

"l ; ..I .4"
..‘5 lb ‘
4 u
.. '3- .-
.Jslt-B.‘ l
o ‘ ~ '
u .v H ‘
..‘no‘ik
.

‘1
tn! ahﬁ“

L‘- huk‘.

rn

.“ ‘q-

11'.
-

”Q1-
‘~-. -.. \t...
I

:7“ ‘- 3.-

bl..‘ ‘§“_

I

‘.'~ ,

J... '3'
u-u.

I

n-

\
I o

KN. ..t 'J

I n

. ‘~s.

‘4) a p—
' 55.5“

 

 

able to avoid bad investments to a higher extent. Because multiple uses for capital exist in
modern ﬁrms, this logic can easily be extended to single-segment ﬁrms — internal
managers with better information to base decisions on may be better also in identifying
and choosing alternative uses for capital besides allocations to business segments.
Empirical evidence, however, has not corroborated Williamson’s M-form
prescriptions regarding capital allocation efﬁciency. First, there is overwhelming
empirical evidence that internalized transactions of capital are executed in a less than
efﬁcient fashion. Berger and Ofek (1995), Scharfstein (1997), Shin and Stulz (1998),
Bromiley (1986) and Schlingemann et al, (1999) ﬁnd evidence pointing to the
inefficiency of internal markets. Shin and Stulz, for example, ﬁnd that business segment
cash ﬂows to a large extent determine future allocations of capital and that the amount of
allocated capital is relatively insensitive to investment opportunities, while Bromiley
ﬁlids a number of additional signs of allocation inefﬁciencies including stale investment
hurdle rates and positively biased forecasts of capital expenditures relative to actual
Spending. As mentioned earlier, this points in the direction of a capital allocation process
1953 driven by investment opportunities and without efﬁciency as its main goal. Instead,
allocations may also be driven by available capital as managers may have a tendency to
overinVest out of free cash ﬂow (Jensen, 1986, 1993), by reputational concerns inducing
managers to reject investments in riskier business segments and projects in favor of safer,
but Potentially less lucrative such investments (Hirscheleifer & Thakor, 1989), by
Political considerations such as herd inﬂuenced investment behaviors of adhering to

es“ablished ‘norms’ of allocation (Scharfstein & Stein, 1990). and by investment

22

.n-1¥-. -

I -
"7 .o.‘
en's. “lt‘

s;.-I.

 

 

 

behaviors driven by the inﬂuence of rent-seeking divisional managers (Meyer, Milgrom
& Roberts, 1992: Scharfstein & Stein, 2000).

Second, Berger and Ofek (1995) have suggested inefﬁciencies in capital
allocation as a main reason for the established diversiﬁcation discount. In other words,
business segments under the corporate umbrella tend to be valued less than similar free-
standing segments due to capital being misallocated at the corporate level.
Notwithstanding that there may be differences in incentives and coordination costs
between a business segment under the corporate umbrella and one that is not
(Williamson, 1975), and that single-segment ﬁrms in my model are assumed to have
internal allocation processes and thus may misallocate capital, the consistent ﬁndings of a
diversiﬁcation discount point in the direction of capital allocation inefﬁciencies at the
Corporate level.

Third, in ﬁnding small as opposed to large corporate effects, the ﬁrm performance
decomposition literature (e.g., Schmalensee, 1985: Rumelt, 1991: McGahan & Porter,
1997) also seems to point in the direction of inefﬁcient and wasteful capital allocation.
Assuming in line with Williamson (1975: 148) that the process is important: “In many
reSpfbcts, this assignment of cash ﬂows to high yield uses is the most fundamental
amibllte of the M-form enterprise”, small corporate effects may logically be a function of
assigning cash flows (less efﬁciently) away from its ‘high yield uses’. In addition, a
number of authors within the ﬁrm performance decomposition literature (e. g., Misangyi
8’ al,’ 2006: Bowman & Helfat, 2001: Adner & Helfat, 2003: Brush, Bromiley &

Hend-l‘ickx, 1999) have suggested a relationship between the efﬁciency of the capital

23

'A.-..\a

3...“... .

.-

ELM:

,... ".o

.
.100 a...

. I»
.n‘. a;

1

ul‘ ‘b-a

1' v-;.. H

L. u~u s

.‘ ﬁt
t. b',.,...
I

3"
..u‘\“‘
u

. ‘v
2., p.

“in

QA“.

...,;

 

 

M;
I; .
ll.
'5.‘ ‘
. ‘__

 

 

 

allocation process and the size of the corporate effect9 based on both theoretical and

methodological10 grounds. Brush, Bromiley and Hendrickx (1999: 524), for example,
state that “wise allocation of capital across divisions may be one mechanism underlying
how well-managed (proﬁtable) corporations should positively inﬂuence the performance
of their business segments”. In other words, they argue for a positive relationship
between efﬁcient capital allocation and the corporate contribution as reﬂected by the size
of corporate effects.

Based on described theoretical and empirical evidence, therefore, the capital
allocation process in multidivisional furns and by extension also in single-segment ﬁrms
may very well often be inefﬁcient which, if we continue to assume that managers are not
perfectly rational in choosing between governance structures, i.e., cannot perfectly judge
when hierarchies are less costly than market exchangesl 1, opens the door for M-forrn
prescriptions to internalize the capital allocation function (as well as for internal

allocation processes in single-segment ﬁrms) to negatively affect ﬁrm performance.
Firms may allow, or even promote, the allocation of investment capital away from
relatively deserving uses including business segments with better prospects in favor of

less deserving uses including business segments with less favorable prospects. If so, the

q‘

9

No“? that this relationship is likely to be asymmetrical since the variance decomposition methodology
inmates Corporate effects at the left side. In other words, poor capital allocation will not cause a negative
CO’POT ate effect because of how it is estimated within this stream of research.

10
No“? that a number of authors (e.g., Bowman & Helfat, 2001: Adner & Helfat, 2003) have emphasized
pr Oblems With the measurement of the corporate contribution suggesting that the relationship between
Gama! allOcation and corporate effects may not be clear.
11
(C(‘gsuéough contemporary organizational economics to a large extent continue to rely on the ‘Coase’
'ﬂarcil 1 937) assumption that organizations effortlessly and optimally replace market exchanges with
efﬁcie y Once the latter is less costly, this dissertation does not. In fact, the current investigation into the
mana “Cy of the capital allocation process would make little sense under the ‘Coase’ assumption since
ex chic“. under that assumption, could always achieve optimal allocation by switching from market
go to hierarchy and back.

24

use of the over and underinvestment construct implying varying levels of inefﬁciencies in
the capital allocation process is valid and may capture those inefﬁciencies. In addition,
once the capital allocation process is assumed to vary in efﬁciency, the ensuing
investigation into antecedents to the degree of efﬁciency as well as performance

implications of this process becomes meaningful.

Antecedents to the Capital Allocation Process
Notwithstanding that few studies have directly investigated antecedents to a
capital allocation process varying in efﬁciency, my selection of antecedents was based on
theoretical arguments and empirical results suggesting that business segments are
embedded in both industry and corporate parent environments (e. g., Granovetter, 1985:
McG ahan & Porter, 2002: Misyngai et al., 2006). The embeddedness of business
segnients in the two environments is important for at least two reasons. First, prior
research has shown that both industry and corporate parents (in addition to business
segment speciﬁc inﬂuences) are likely to inﬂuence business segment proﬁtability (e. g.,
McGahan & Porter, 1997, 2002: Brush, Bromiley & Hendrickx, 1999). Thus,
characteristics or factors at all three levels are likely to affect ﬁrm performance through
the efficiency of the capital allocation process, a process that Williamson (1975)
suggested may be the single most important or fundamental function of the diversiﬁed
ﬁrm. Second, prior research has also argued that industry, corporate and business-
segment effects (and by extension characteristics at those three levels) are not
independent of each other (e. g., McGahan & Porter, 1997, 2002: Hough, 2006: Misangyi
et al., 2006). McGahan and Porter (1997: 838), for example, suggest “strong covariance

between industry and corporate parent effects” implying that corporate parents may not

25

only be able to add value to their business segments by efﬁcient corporate level actions
such as control, auditing and capital allocation functions, but may also add value by
choosing to enter industries with particularly attractive characteristics. Therefore, since
the corporate capital allocation process is embedded in two additional environments,
industry and business-segment environments, its efﬁciency may also be affected by
variables at those levels. In sum, because characteristics at all three levels (business
segment, corporate and industry) are likely to inﬂuence ﬁrm proﬁtability (through the
efﬁciency of the allocation process), as well as the strong embeddedness or covariance
between characteristics at all three levels, I elected to include variables from each level as
antecedents to the capital allocation process (Figure 1).

Speciﬁcally, based on the logic of aspiration-driven organizational behaviors
(e.g., Festinger, 1954: Cyert & March, 1963: March & Simon, 1958: Levitt & March,
1988), business segment performance in relation to historical and social aspirations of
such performance may affect the allocation decision by inviting different levels of
investment depending on whether current performance happens to be above or below
such aspirations. Thus, aspirations for business segment performance may reﬂect
important business segment inﬂuences on the capital allocation process by affecting
actual capital allocated to individual business segments.

To represent important corporate level inﬂuences on the capital allocation process
and to reﬂect the importance of agency theory (Jensen & Meckling, 1976: Fama, 1980:
Fama & Jensen, 1983: Jensen, 1986) in the area of corporate level strategies and

decisions, three agency theoretical variables are suggested to affect the capital allocation

26

 

.'
I
I
I
I
l
I
I
I
I
I
l
l
I
I
I
I
I
I
I
I
I
I
I
l

 

 

 

 

 

 

 

.5. 3.58... .... . .

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

_ .838 .o .80 ..
. . 2.22:3»: .
. 26E :30 .
u 22.3 03 ..
. .v , cam accuse-20
. All, Idealdﬂadd
u. <0... :5... 8... > JL
"
. . . . .
. . REE—8.8:.
" . u... .26. .
. 83> Es“.   ._ 5:822 .
n 8522:. M .338
u
.
.
. 1+. 2-..:
r u .v .85
220900 352...; \
Ego:
.eoEuom
30525 3.53500
tea 525259
252.895
3.8%.: $582 A .................. . 25m ;
incense—.2550 A. ttttt .

aim—8:53— 3853»: Al

 

 

 

 

 

A+v be:

 

 

 

 

.350
a... 83228.

355.85.. A
, .SeoEec..>mwl

 

 

 

 

 

To 0:

 

 

. :96. 328.3. .
53:38:“.
. .o .23

 

 

 

 

 

3 YNE

 

 

 

. . 0033.3)
855260

 

 

 

 

 

A+v 3*

 

LIV

 

 

>UZBUEW ZOE—@0943. Jenn—Add OH. mBZHGm—Um—HE

m nay—DUE

 

85.8.2.3...

38» 328.33.:

 

 

27

process by decreasing the divergence between “the agent’s decisions and those decisions
that would maximize the welfare of the principal” (Jensen & Meckling, 1976: 308). That
is, agency-theoretic characteristics of the corporate environment may limit possible
inefﬁciencies in the capital allocation process. In addition, level of diversiﬁcation is also
included as an additional corporate-level inﬂuence on the efﬁciency of the capital
allocation process.

Finally, environmental dynamism and complexity have been associated with
uncertainty and difﬁculties in organizational decision making in a number of theoretical
(e.g., Duncan, 1972: Dess & Beard, 1984: March & Simon, 1958: Mintzberg, 1990:
Weick, 1979: Von Krogh, Roos & Slocum, 1994: Milliken, 1987, 1990: Child, 1972) and
empirical contexts (e. g., Henderson, Miller & Hambrick, 2006: Li & Simerly, 1998:
Simerly & Li, 2000: 6011 & Rasheed, 1997: Wiersema & Bantel, 1993: Boyd, 1995,
1990). Thus, environmental dynamism and complexity were chosen to represent

important industry characteristics inﬂuencing the capital allocation process.

Business segment historical and social performance comparisons

Performance cannot be judged as satisfactory without a point of reference. This
point of reference, i.e., goals or aspirations, is a critical part of many models describing
organizational behavior - both organizational learning models (e. g., Levitt & March,
1988) and aspiration models (e.g., March & Simon, 1958) are to a large extent driven by
goals or aspirations. In the latter models, for example, current performance in relation to
aspirations determines if outcomes are acceptable or unacceptable. Thus, judgments are
not possible without some pre-established reference against which a comparison of

current performance can be made.

28

It follows, therefore, that organizational behaviors are driven or affected by
current performance in relation to aspirations or goals — organizations likely react
differently depending on whether performance falls short or is above some aspiration
level. This assumption that organizations set goals and subsequently adjust behaviors
based on the relative performance in relation to those goals is in accordance with ﬁndings
in organizational learning positing organizations as goal directed, rule based, and
sensitive to prior performance history (Levitt & March, 1988). Thus, based on the
assumption of satisﬁcing (March & Simon, 1958: Eisenhardt & Zbaracki, 1992), without
which the optimal choice would always be chosen leaving no room for improvement,
performance feedback in relation to aspirations determines if performance is acceptable
or not which, in turn, is associated with organizational decision-making and subsequent
behavior.

Aspirations, in turn, may be in response to two different types of comparisons -
social and historical. Social comparison is based on how people establish a sense of self-
assessment based on comparison to a reference group of similar others (Festinger, 1954:
Cyert & March, 1963). Historical aspiration levels, on the other hand, are established by
the performance history of the focal fu'm or entity itself (Cyert & March, 1963), instead
using prior performance as a guide to self-assessment and subsequent perceptions of
success or failure. The latter aspiration levels have been related to a number of decisions
in the organizational context including increased managerial risk taking in response to
attainment discrepancies (Palmer & Wiseman, 1999), early entry into research and
development consortia (Bolton, 1993), changes in organization strategy under different

industry environments (Lant, Milliken & Batra, 1992), and decisions in simulated market

29

conditions (Lant, 1992). In addition, other studies (Bromiley, 1991: Lant & Montgomery,
1987) have used a more complex measure of aspirations based on the logic underlying
the Behavioral Theory of the Firm (Cyert & March, 1963) where aspirations are
represented by either past performance or the performance of competitors depending on
whether current performance is above or below average industry performance. I initially
incorporated this measure but later elected not to use this measure due to inconsistencies
when used in conjunction with my measures of capital allocation efﬁciency. This is
explained in further detail in the methods section. In sum, underlying theoretical
arguments and empirical support point to the importance and usefulness of aspiration
levels as determinants of behaviors in a number of different organizational contexts.
Translating the logic of aspiration-driven organizational behavior to the context of
capital allocation in the multidivisional ﬁrm suggests that this process may be affected by
aspirations for business segment performance. A number of researchers have indirectly
looked at aspiration driven allocation behaviors without necessarily attributing
subsequent inefﬁciencies to varying levels of aspirations - Shin and Stulz (1998), for
example, use own and comparable single-segment ﬁrm cash ﬂows as predictors of capital
investment inefﬁciencies, while Rajan et a1. (2000) predict and ﬁnd that the dispersion in
resources increases instances and magnitudes of inefﬁcient capital allocation. Thus, based
on the application of aspiration-driven behaviors to the particular context of capital
allocation, performance below either aspiration level, i.e., performance shortfalls, should
increase the urgency to change investment allocations to quickly get back above the
reference point to achieve satisfaction or success. In particular, since failing to attain

aspirations has been associated with increased risk taking both theoretically (e. g.,

30

Kahnemann & Tversky, 1979: March & Shapira, 1992) and empirically (e.g., Greve,
1998; 2003), business segment performance below aspirations implying a loss context
may mean that managers at the corporate ofﬁce may consider increasing investment
allocations up and above what may be deemed reasonable to try to reverse losses. In fact,
although corporate managers are assumed to have ‘ﬁnal’ say in allocation decisions
(Stein, 1997), business segment managers will likely also seek more investment capital in
response to performance below aspirations, thereby contributing to this relationship.
Thus, overinvestment ensues, which may ex post be viewed as risky since returns on
additional capital allocated, especially capital going to business segments with poor
prospects, are highly uncertain and may reduce ﬁrm performance further.

Managerial tendencies to escalate commitments or the reluctance of managers to
withdraw from losing courses of action, projects, and businesses (Staw, 1976, 1981,
1997) may also contribute to overinvestment in business segments performing below
aspirations. Thus, escalating commitments may reinforce and strengthen the tendency of
loss-averse managers to keep constant or even increase investment allocations to poorly
performing business segments. This tendency to keep constant or increase commitment is
partly due to a failure of managers “in coming to terms with their failure” (Hayward &
Shimizu, 2006: 541). That is, the potential costs of reversing course in the form of public
embarrassment and humiliation (e.g., McNamara, Moon & Bromiley, 2002) may be so
high that managers postpone this reckoning to a later date and continue on the chosen
path. Ross and Staw (1993) emphasize that such psychological determinants of escalation
behaviors, sometimes in conjunction with social norms emphasizing consistent and

strong leadership, high closing costs and low salvage values at exit, and political support

31

Y"'J"\ II
.LILH..-

ovja‘ \’ .
I

I
-1 'l- -l

 

 

 

 

-"‘o-a.
.n._”‘~|

 

“"910.

‘—

‘ ‘iu'. ..
.

‘P‘In

I!‘
‘I

u...“ .
. ‘

1:1

 

"u,
*‘I I”.
”‘54

for continuation, often lead to extreme strategic persistence in the face of growing
concerns with the chosen strategy. They illustrate just how powerful these psychological
drivers of escalation can be by describing how the Shoreharn nuclear power plant on
Long Island continued to receive investment funds over a 20-year period although the
viability of the project was in question almost immediately and the cost-overruns were
massive and increasing during the entire time period. Thus, based on underlying
psychological and other reasons, managerial tendencies to escalate commitments may
reinforce and increase overinvestment in poorly performing business segments once
substantial sums have been contributed in the past. Escalating commitments and
aspiration-driven loss contexts, therefore, provide complementary arguments for
explaining overinvestment in poorly performing business segments.

For well-performing segments operating above aspirations implying at least some
level of satisfaction and success, however, corporate managers may perceive less urgency
to act, instead keeping allocations constant because additional allocations may be viewed
as unnecessary and perceived as being better spent elsewhere in the organization.
Alternatively, since segment performance above aspirations implies gain contexts and
increased risk aversion on behalf of corporate managers, they may also consider reducing
allocations since additional investments may be viewed as risky because the return on
those funds is uncertain and could affect current segment returns negatively, thus
lowering returns of well-performing segments.

Thus, I propose that effects stemming from performance aspirations and

accompanying gain/loss context may lead to biased and non-optimal decision-making

32

where corporate managers overinvest in segments with less favorable prospects and
underinvest in segments with relatively better prospects. Formally stated:

Hypothesis Ia: Current business segment performance below aspirations is
positively associated with overinvestment.

Hypothesis 1 b: Current business segment perfonnance above aspirations is

positively associated with underinvestment.
Agency-theoretic corporate governance variables

Agency-theoretic corporate governance variables including block holder
ownership, institutional shareholder ownership and contingent forms of executive
compensation may impact the investment allocation process. According to agency theory
predictions (e.g., Jensen & Meckling, 1976: Fama, 1980: Fama & Jensen, 1983: Jensen,
1986), the above listed corporate governance variables will act to increase the alignment
between principal and agent, thereby increasing the efﬁciency and performance of the
ﬁrm. In other words, they will act to decrease the divergence between “the agent’s
decisions and those decisions that would maximize the welfare of the principal” (Jensen
& Meckling, 1976: 308), or what governance researchers usually refer to as agency costs.
Thus, one potentially important, yet not widely researched, avenue to increase this
alignment and decrease agency costs is to increase the efﬁciency of the capital allocation
process by allocating relatively more capital to better uses including business segments
with better prospects. This would ‘maximize the welfare of the principal’ as capital is to a
larger extent allocated to better uses, potentially raising overall ﬁnn returns and share

price”. The following three hypotheses (HZ-H4) with separate predictions for over and

 

'2 Although dividends are received by shareholders and as such should be of interest to this group of
stakeholders, shareholders are likely more interested in share price appreciation because it avoids the
‘double taxation’ often cited as a reason to avoid dividend payments. In fact, based on the tax argument,

33

underinvestment will elaborate on how block holder and institutional shareholder
ownership as well as contingent forms of executive compensation are associated with the
efﬁciency of the capital allocation process.

Large shareholders may achieve efﬁcient monitoring of managers, resulting in
enhanced corporate performance. Both block holders and institutional shareholders have
been suggested as efﬁcient monitors of corporate insiders leading to increased ﬁrm
performance through an increased alignment between the interests of corporate insiders
and their organizations’ shareholders (Hansen & Hill, 1991: Schleifer & Vishny, 1986:
Kabir, Cantrijn & Jeunink, 1997). In other words, as argued by Jensen and Meckling
(1976), Kroll, Wright, Toombs and Leavell (1996) and others, shareholder concentration
may have positive effects on ﬁrm performance and ﬁrm share price.

This positive effect on shareholder wealth from an increase in shareholder
concentration may be mainly due to its effect on decisions involving corporate risk
taking. In other words, as suggested also by Schleifer and Vishny (1986), it may be that
increased monitoring by large shareholders mainly affects ﬁnn performance through an
increased level of risk taking - large shareholders are able to force value maximization,
thereby decreasing agency costs, through their positive effect on ﬁrm risk taking. Thus,
the risk differential that Coffee (1988) and others use to characterize the divergent
interests between self-interested (risk-averse) managers and risk neutral shareholders may
decrease as a result of monitoring by shareholders large enough to be able to inﬂuence

risky ﬁrm decisions.

 

dividends not only lower share prices but also unnecessarily destroy shareholder wealth by increasing the
combined tax burden of the ﬁrm and the investor.

34

Although this relationship (between shareholder concentration and ﬁrm risk
taking) has not been studied to the same extent as the relationship between shareholder
concentration and ﬁrm performance, a number of researchers have found or suggested
relationships. Hill and Snell (1988), for example, suggested a relationship between
shareholder concentration and risk taking, implying that increased shareholder
concentration may lead to fewer risk-reducing corporate strategies such as diversiﬁcation.
Furthermore, Bethel and Liebeskind (1993) suggested a similar reduction in managerial
tendencies to engage in risk-reducing strategies, while Mikkelson and Ruback (1991)
found support for a relationship between large shareholders and managerial decisions
thought to positively affect shareholder interests. In addition, Wright, Ferris, Sarin and
Awasthi (1996) found a positive relationship between level of institutional equity
ownership and corporate risk taking for ﬁrms with growth opportunities. Their arguments
and ﬁndings are consistent with Barclay and Holdemess (1990), Brickley, Lease and
Smith (1988), McConnell and Servaes (1990), Mikkelson and Ruback (1985) and
Baysinger, Kosnik and Turk (1991), suggesting that large institutional investors are likely
to enhance corporate value by supporting and increasing “growth-oriented risk taking”
(Wright, Ferris, Sarin & Awasthi, 1996: 454). Thus, taken together, existing theoretical
and empirical research seems to point in the direction of large shareholders enhancing
ﬁrm value by not only emphasizing shareholder interests, but also by directly increasing
the level of risk in managerial strategies and decision-making, and thus decreasing the
risk differential (Coffee, 1988) between managers and shareholders.

Translating those ﬁndings and the logic underlying a positive association between

monitoring by large shareholders and ﬁrm risk taking to the current context of the

35

internal capital allocation process, increased shareholder concentration should raise the
level of efﬁciency of this process by forcing reluctant corporate managers to increase
allocations to well performing business segments”. That is, because segments currently
performing above aspirations may be perceived as operating in gain contexts requiring
protection, corporate managers may be hesitant to increase allocations to such well
performing business segments with relatively better prospects because additional
allocations may be perceived as ‘diluting’ or jeopardizing already high business segment
retums, therefore perceived as risky by corporate managers. Thus, based on a direct
relationship between managerial risk aversion and the reluctance to invest, preferring to
protect current levels of performance instead of aiming for additional performance gains,
large shareholders may be able to decrease instances and magnitudes of such
underinvestment by forcing managers to increase the level of investment to well-
perforrning business segments. Formally stated:

Hypothesis 20: Block holder ownership is negatively related to
underinvestment.

Hypothesis 2b: Institutional shareholder ownership is negatively related to

underinvestment.

Large shareholders may also be able to force value-maximization, and thereby
enhance firm value by acting to control overinvestment related to corporate governance
breaches and imperfections in less successful business segments. This is in accordance
with Pound’s (1988) efﬁcient-monitoring hypothesis suggesting that large shareholders

support managerial decisions enhancing furn value, but tend to oppose decisions

 

'3 Although the idea that managers are reluctant to invest in well performing segments may seem extreme
and improbable at ﬁrst, it closely follows theory (e.g., Prospect theory) as well as being in line with other
risk-reducing managerial actions such as for example diversiﬁcation outside the ﬁrm’s main area of
expertise.

36

I, .
l-un‘

llhah‘l . -

ml ‘3'" ‘
N. abbu— .

 

-5.- - or.
..bngv:
-

’q.— ...

“no u- -h-.
.

I“ It: a
r‘ _,

I ..h— 5.

 

.
.-, .'.
I .

"—«uu . e

um.

I...

-.-

‘i O
‘Pen .
..L

 

 

.1,”

.
.l
‘..vl

I
-.-.

Q ‘ ".
.4‘ .2

I, .I.

r. ;.-:(

detrimental to shareholders’ interests. Jensen and Meckling (1976) and Kroll, Wright,
Toombs and Leavell (1996) also argued for a theoretical association between stock holder
concentration and ﬁrm performance. Supporting that theoretical association, Brickley,
Lease, and Smith (1988) found support for large institutional shareholders opposing
managerial decisions harmful to shareholders, while Mikkelson and Ruback (1991) found
empirical evidence for large shareholders supporting managerial decisions deemed to
positively impact shareholder interests. A number of other studies have also concluded
similar relationships between shareholder concentration and shareholder interests
(McConnell & Servaes, 1990: Barclay & Holdemess, 1990: Mikkelson & Ruback, 1985).
Thus, large shareholders efﬁcient at monitoring self-interested and risk-averse corporate
managers may be able to minimize instances of this type of overinvestment referred to as
“allocational socialism” in a number of important studies (e. g., Rajan et al., 2000:
Scharfstein & Stein, 2000: Bernardo, Lou & Wang, 2006). Conceptualized as investment
capital that is presumably allocated away from better uses including business segments
with stronger prospects in favor of less favorable uses including segments with relatively
weaker prospects, allocational socialism is measured similarly to how I measure
overinvestment, i.e., as investment in business segments with relatively lower prospects.
This way of measuring overinvestment recognizes that overinvestment often is
independent of underinvestment in modern ﬁrms where capital can be used for a wide
array of other uses such as for dividend payments, for buy-backs, and to acquire other
ﬁrms (Dow & Gorton, 1997). Thus, as explained earlier, overinvestment extends to ﬁrms
with only a single business segment as such ﬁrms still need to decide how much capital

to allocate to business segments versus other potential uses.

37

To explain how large shareholders may be able to counter and control managerial
tendencies to practice allocational socialism by overinvesting, a number of behavioral
processes or reasons underlying these tendencies to overinvest need to be focused on.
Thus, I suggest three such underlying behavioral reasons or processes that are likely to
lead to overinvestment unless properly monitored — agency problems at the allocational
level resulting from interactions between self-interested corporate managers and
divisional managers (Scharfstein & Stein, 2000: Bernardo et al., 2006), the tendency of
loss-averse managers to prefer not having to realize losses whenever possible, as well as
the tendency of managers to be inﬂuenced by escalating commitments where
psychological, social and political reasons may make it difﬁcult for them to withdraw
from losing courses of actions, projects and businesses (Staw, 1976, 1981: see Staw,
1997, for a review). First, in addition to agency costs due to separation of ownership
(Berle & Means, 1932), both Scharfstein and Stein (2000) and Bernardo et al. (2006)
refer to a “second layer of agency costs”, encompassing costs generated between
corporate managers in charge of allocating investment funds and divisional managers
receiving those funds. For example, unless efﬁciently monitored, corporate managers
may view it as “less personally costly to distort investment in favor of those divisions
whose managers require extra compensation, thereby conserving on cash payments to
these managers” (Sharfstein & Stein, 2000: 2540). Thus, divisional managers running
business segments with poor prospects may be able to extract “excessively large capital
allocations” (Scharfstein & Stein, 2000: 2540), resulting in overinvestment in their
divisions because corporate managers may, unless efﬁciently monitored by large

shareholders, perceive additional allocations of capital as less costly than using

38

., '5'

bl—N‘”

«—
ys. .u

 

'VI,.‘
I b. .
'... -
4a, \

 

..y I ‘
L.\'-.5

A..‘ ‘
SCH...

K
U

“c
-' .LJ

:! \..,

0‘
Nb...

1: H"-
‘A >h.

 

“‘.
l-'

‘5.

. k
l

p”
A

discretionary funds that the manager controls and can “potentially divert to himself”
(Bernardo et al., 2006: 3).

Instances of allocational socialism resulting in overinvestment may also be in
response to managerial tendencies for loss aversion unless efﬁciently monitored. Loss
aversion, the basis of Prospect theory (Kahneman & Tversky, 1979), suggests that people
have strong tendencies to want to avoid realizing losses whenever possible. To do so,
managers have been shown to make decisions they would not normally make, and accept
elevated levels of risk that they would not normally accept (e.g., Kuhberger, 1998). Thus,
although not often cited and addressed in the capital allocation literature, loss aversion
may lead managers to overinvest to try to turn failing business segments around, thereby
avoiding having to admit failure by drastically reducing investment rates or by divesting
the particular business segment in question. That is, due to managerial loss aversion
underlying risk seeking below an acceptable level of performance, unless efﬁciently
monitored, managers may be tempted to try to increase allocations in the hope of using
the increased leverage to reverse poor performance and avoid the embarrassment and
potential threats to employment from admitting that the weakest divisions need to be
divested or put on “much-needed diets” (Scharfstein & Stein, 2000: 2539).

A third reason underlying allocational socialism and overinvestment is, as
previously discussed in H1, managerial tendencies to escalate commitments. Thus, unless
efﬁciently monitored, managers may be very reluctant to withdraw from losing courses
of action, projects, and businesses (Staw, 1976, 1981, 1997), instead preferring to

continue overinvesting in poorly performing business segments.

39

a v-

j .. nob

al.:D‘ 43
\ .
*5ukl9‘

' wan-'3‘.

h-u—«S‘Ae
U

1' '33" -. ‘1
an hu‘. ubk

I
f0 a «I... r.

5‘. b\'.au s.

'~~- '1.
“We ..‘
.

A

Is...
' 1“".

l “~QFI-y
~ ' “n.
”54,3 '1
it.“ h“.
H.
0.
H.
O"
C\
.‘w
| .
I'.N|_?\. '5
N “
.‘_

In sum and in line with suggestions by previous researchers (e.g., Hansen & Hill,
199 1 : Schleifer & Vishny, 1986: Jensen & Meckling, 1976: Kroll, Wright, Toombs &
Leavell, 1996: Pound, 1988: Brickley, Lease & Smith, 1988: McConnell & Servaes,
1990: Barclay & Holdemess, 1990: Mikkelson & Ruback, 1985, 1991), large
shareholders may be able to efﬁciently monitor inside managers, leading to an increased
alignment between the interests of corporate insiders and their organizations’
shareholders. In particular, by their inﬂuence, large shareholders may be able to counter
and control managerial tendencies to generate agency costs at the allocational level by
improperly using investment funds to reward divisional managers, as well as to control
managerial tendencies related to loss aversion and escalation of commitment also leading
to overinvestment. Thus, large shareholders may be able to contribute to a more efﬁcient
capital allocation process by efﬁciently monitoring the process. Formally stated:

Hypothesis 3a: Block holder ownership is negatively related to
overinvestment.

Hypothesis 3b: Institutional shareholder ownership is negatively related to

overinvestment.

Contingent forms of executive compensation (e. g., granted shares of stock and
oPtions) may have a similar efﬁciency increasing effect (as shareholder concentration) on
the capital allocation process. Higher levels of contingent forms of compensation increase
the alignment between the interests of shareholders and corporate managers by making
the latter group share some of the costs of managerial self-interest and risk avoidance. Put
differently, sharing ownership with corporate managers increases managers' potential
opponunity costs of not maximizing shareholder wealth. Thus, higher levels of

contingent compensation may not only reduce the risk differential (Coffee, 1988) by

40

if;

u rv.s‘
a. \ .-
0 he..‘.

.-‘ _

4..

H-
'- -.L._

R: -.
a“; )

7'37: 33

3: i1L<

 

 

“ I.
l “v
N H‘.
I.
a“!
l
-1
'19‘<-
m
*“H\‘
-
l‘u
4

F;

r Ir
..1

making managers more interested in additional risk-taking, but may also reduce the
frequency of other non-maximizing actions and decisions such as, for example, those
suggested to underlie allocational socialism (H3).

Following that logic, contingent compensation would reduce instances of under
and overinvestment alike. Underinvestment in business segments due to managerial risk
aversion as well as to other reasons would increase opportunity costs as managers now to
a larger extent are able to participate in ﬁrm success. That is, as the level of managers’
contingent compensation increases, their wealth is now more strongly correlated with
ﬁrm Share price, which should make them more interested in maximizing share price by
aVOiding protection-minded underinvestment. Scharfstein (1997 ) supports this reasoning
as he suggests that misallocation of capital is driven by a misalignment of incentives
b'itt‘W'iten outside investors and top management. In his sample of diversiﬁed ﬁrms, he
Points to an increased sensitivity of divisional investment to Tobin’s q as top managers
own increasingly larger equity stakes in their ﬁrms as evidence of misaligned incentives.
Thus, he recommends high-powered (equity-based) incentives to increase the sensitivity
of divisional investments to prospects and thereby reduce protection-minded
undeI‘investment. Overinvestment, although not necessarily based on risk
Cons iderations”, instead based on different self-interested actions such as the managerial
tendency to prefer to compensate managers with extra allocations of investment capital
instead of with cash payments, as well as the tendency to try to reverse poorly performing

.4\
also“. that the argument could be made, but seldom is, that contingent compensation may help to control
thateSSlve risk seeking below some reference point. Underlying the reluctance among scholars in pursuing
diﬁ‘ argument may be assumptions of risk aversion across gains and losses (expected utility theory) and the
equlacfllty in prescribing additional risk-taking under some circumstances, but not others. In addition,
belo tlng overinvestment with risk seeking conﬂicts with how decision-makers likely perceive their actions
the W the reference point, i.e., actions resulting in avoiding having to recognize losses and expose

Ives to employment risk may instead be perceived as risk reducing.

41

... mu?“
.~--

3.03:

-‘.,..,.v

..u.--
.

".1
(u.
I,

..1- ..
m-J.

'° Y He:

0084‘. ll

1’": V;

- g 1

H
Du. ..
l'dl &

la-

 

 

business segments by allocations over and above what may be deemed reasonable, may
also be affected by contingent compensation through a process similar to the one
whereby underinvestment is affected by contingent compensation. That is, with higher
levels of share and option ownership, self-interested motives not maximizing shareholder
wealth such as those described would increase opportunity costs, and therefore are likely
to be perceived as more costly by corporate managers. This is in line with Scharfstein and
Stein’ 3 (2000) model where “large socialist-type inefﬁciencies are especially likely to
arise - - - when the CEO has low powered incentives” (Scharfstein & Stein, 2000: 2450).
T1“18 . as a result of “more high-powered incentives”, i.e., more contingent compensation,
such actions may be avoided to a higher extent, leading to less overinvestment.

In sum, due to underlying arguments positing that opportunity costs of both risk-
avoidance and other self-interested actions increase as a direct result of higher levels of
contingent compensation, a negative relationship between executive contingent

corripensation and inefﬁciencies in the capital allocation process is likely.

FOrIl'lally stated:
Hypothesis 4a: Contingent compensation is negatively related to overinvestment.
Hypothesis 4b: Contingent compensation is negatively related to

underinvestment.

Level of Diversiﬁcation

Consistent with typical measures of diversiﬁcation based on breadth and salience
of bllsiness segments (e.g., Jacquemin & Berry, 1979), the level of diversiﬁcation can be
Vlev"ed as a continuum from one-segment focused ﬁrms to multi-segment conglomerates

"11 a number of busmesses 1n unrelated industries. Thus, two mﬂection pomts capture

42

‘- ‘9.-\: 1"

L, nun—oh '-
_ i

Q
.0‘ a!“

“by; ‘
."..._55“N

3:"‘1 1,’
a» “start s
I

1......“ .-
vnlu wk
.

-—,~.,.- 2. .
«to: L:
Q
. ‘ v
unfit? l
. .‘ . .
'I: .3“ a;

.525; 0]

g .
fly.“ I

““qude

rm.‘

4...“; S l6‘d

two dimensions — the fust dimension as an organization diversiﬁes and is no longer
Operating only in a single industry, and the second one as the ﬁrm increases its
divers ification efforts into more and increasingly unrelated segments. Both dimensions
are incorporated in this dissertation to investigate the impact of diversiﬁcation on the
capital allocation process.
Increasing levels of diversiﬁcation can negatively impact the capital allocation
Process by making the allocation decision less clear (or more uncertain). That is, an
increasing number of segments as well as their decreasing relatedness would likely
increase the amount of information that needs to be gathered, analyzed and acted upon.
cor‘POrate managers may also be limited in their knowledge and understanding to just one
or a few of the organization’s business segments, likely those relatively closer to the core
bus iness of the organization. Thus, assuming bounded rationality (March & Simon,
195 8), this increasing amount and diversity of information may quickly exceed the
limited information processing ability of the centralized group of corporate managers in
charge of allocating capital to business segments, thus making it less likely that decision-
Inaliers can gather and process the necessary information to make efﬁcient decisions. In
addition, this limited ability to process information may hamper efforts to achieve an
overView of the firm’s business segments which, as explained earlier, is necessary to
ach ieve an efﬁcient capital allocation. Thus, because of a limited ability to quickly
anal YZe vast amounts of diverse information, and because the need to gather and analyze
Such information increases as the level of diversiﬁcation increases, instances and
Inagllitudes of inefﬁciencies in the capital allocation process may likely increase as the

fix—1h . . . . .
8 level of d1versrﬁcation mcreases.

43

‘ ‘ Q
*Q'. '39.:

iv.- bl.-
h

‘I‘ ~«nl<-

a .
.— -nkao \.

1e" :2; u

iﬁ'l‘ p! '
an “.0:

“M.

sci-3:1.

.\

:m-
W”. at .
' l.
12"». .
.ulung :

Diversiﬁcation may also lead to more capital allocation inefﬁciencies, i.e., more
over and underinvestment, because higher levels exacerbate the effect of rigid
bureaucratic allocation rules often found in organizations (e.g., Bower, 1986: Shin &
Stulz, l 998). Business segments may, for example, be allocated investment funds based
on their revenue which does not take segment prospects or actual “need” into account.
Both B erger and Ofek (1995) and Lamont (1997) point to these bureaucratic tendencies
in allocation polices as one reason behind the overinvestment and cross-subsidizationls
they find when exploring underperforrnance in diversiﬁed ﬁrms.

Building on their arguments, I argue that increased diversiﬁcation likely
exacerbates the problem of adhering to rigid rules of allocation because more diversiﬁed
ﬁrnls are likely to have more business segments being allocated capital (Argyres, 1996),
thus more opportunity for inefﬁcient allocation based on rigid allocation rules. In
ad(‘litiom the rigid allocation rules themselves are likely to impact performance more
negatively as diversiﬁcation increases. The latter needs elaboration - more over and
undel'investment is likely since a corresponding divergence in business segment prospects
is 1iliely as a result of increased diversiﬁcation. That is, the more a ﬁrm diversiﬁes, the
further it gets away from its core business, thereby likely increasing the divergence in
business segment prospects and performance. Thus, rigid allocation rules disconnected

from segment prospects (such as allocating investment funds based on relative segment

evehue) are likely to ﬁt even worse in that context, leadmg to mcreased mefﬁcrencres 1n

the <tapital allocation process.

\

ls

age)! use cross-subsidization to deﬁne the practice of re-allocating investment capital from well-

go - ol‘lning business segments to failing ones. Since they only emphasize and measure the capital ﬂow
g to less well-performing segments, this def'mition is similar to my deﬁnition of overinvestment.

44

llli

'I

. .1.o.~_1- l
. . .,
bLkhL-Lu:

W1.

[minim
3.
iii con:

ll I .
f1: iii-T":

ran in

1:.‘4'1if‘
““ulﬂ

In sum, increased diversiﬁcation can lead to more inefﬁciencies in the capital
allocation process by making the allocation decision itself more complicated and
uncertain, as well as exacerbating the negative effects of rigid allocation rules. Based on

those arguments, I expect that:

Hypothesis 5a: Level of diversiﬁcation is positively associated with
overinvestment.

Hypothesis 5b: Level of diversiﬁcation is positively associated with
underinvestment.

EnVironmental uncertainty

Duncan (1972) was one of the ﬁrst to study characteristics of the environment that
could contribute to organizational members experiencing uncertainty in decision-making.
He identiﬁed two dimensions of the task environment that would likely increase decision
Inaliters’ uncertainty - the static-dynamic dimension that measures the rate of change or
instability of an environment and the simple/complex dimension that is deﬁned as the
mmlber of factors involved in the decision process. Although he uses a 2 X 2 (static-
Sinlple, static-complex, dynamic-simple and dynamic-complex) and tests for both main
and interaction effects, it is clear from his discussion that he is more interested in their
(environmental dynamism and complexity) individual main effects on the decision-
Inalﬁing process than a possible interaction between the two dimensions. In addition, he
finds no evidence of any interaction between the two dimensions and argues that one
dimension is a far more salient contributor to uncertainty than the other: “. . .the static-
dytlarnic main effect is much higher (F=30.504) than the simple-complex main effect

(F§ 1 0.43 8). The potency of the static-dynamic dimension in contributing to perceived

45

Q'370‘.’ D II
Jim-w “A

"W‘ O;

—n..«. .

“i :0‘ 3n- ,

St.\\...|1

‘.~' .‘~ H19; -
‘\-‘~\.uLO.

"If? A CL

59;! T .3

t in
“3‘s 1 . ‘
“my (I2
5
”5113s. '~

“-45

5135?. On p

"u

A
'5

l) he

"U. I
on.

K; Mo‘
wt...

5.; _
.. w!
“‘ “we '1

’5? .
.u Kaine! (

w“ ‘ _
«“enau‘ ’ 1‘

it».

. {.0ch

; N‘h.l_'rh
.‘0
n‘ ,‘
‘ b l. l?—,
IQ n
.eg .
~

uncertainty is further enhanced by the non-signiﬁcant interaction effect between the two
environmental state dimensions” (Duncan, 1972: 325).

Following Duncan, Dess and Beard (1984) build on those dimensions16 and
reaffirm the positive relationship between environmental dynamism and complexity on
the one hand, and uncertainty in decision-making on the other. Subsequent empirical
research has viewed environmental uncertainty as having both a direct relationship with a
number of performance-related outcome variables (e. g., fum performance) (e. g.,
Wiersema & Bantel, 1993; Palmer & Wiseman, 1999) as well as a moderating effect on
associations between ﬁrm level characteristics (including rationality of decision
Processes) and performance outcomes (e. g., Boyd, 1995: Henderson, Miller & Hambrick,

2006: Hough & White, 2003: 6011 & Rasheed, 1997: Simerly & Li, 2000: Li & Simerly,
1998). The difference between the two approaches of viewing environmental uncertainty
hiIlges on whether or not antecedents to performance outcomes are actually measured, or
juSt assumed. That is, even when environmental uncertainty is modeled to have a direct
effCCt on performance outcomes, the assumption is that some unmeasured antecedent
affects the outcome variable differently depending on the level of environmental
“Certainty. In effect, therefore, both direct and moderated relationships are indications of
the Same underlying relationships where decisions, strategies or actions undertaken have
outcomes of varying rationality attached to them depending on the level of environmental
uncertainty. Thus, in most cases, an investigation of direct effects is a necessary
bypl'Oduct of the fact that only decision outcomes, not the decisions themselves, are

measured and are found to vary based on the environment. Therefore, since I measure
\

'6
In addition, the authors suggest a third dimension (environmental muniﬁcence), deﬁned as the abundance
re‘Sources in an environment. Since that dimension is more related to future prospects than decision-
lng uncertainty, it will instead be used to measure future prospects as an alternative to Tobin’s q.

0f

46

OK

in v"
15:. 3.,Lhklt

‘ a
ha! 9"‘1 '7'
Hub. nub "

I. ’3‘. ‘4”
”..t-» *N
A

1.. 1,... "x I.
~6~‘..th
I

1”,... ,1
bat“‘ EL. g

‘2-(". 'x- w-
in u:

s, , .
its :J‘CKESS

mail)
1

“A
:1- '| I- .
“‘3‘ OI he

35'“ v.
in“: Jon

Walt. m

lull

5V!)
\E'L‘Tno
l

““1011
‘3‘... . _
"QC III?"

A“

M 5- and

both outcomes and factors affecting those outcomes, I provide hypotheses outlining both
direct and moderated relationships involving environmental uncertainty in H6-7 and H8-
10, respectively.

In addition, I do acknowledge a discrepancy between the level of rationality
(bounded rationality) that is accompanying my antecedents and the higher level of
rationality that is inherent in most measures of ﬁrm performance outcomes (e. g., stock
price) used in prior research within strategic management. That is, most prior research
has taken the efﬁciency of various ﬁrm performance measures for granted by assuming
that they are indeed reasonably unbiased indications of how well a ﬁrm is performing. In
effeCt, the difference in rationality stems from the implicit assumption that the level of
rationality increases as more and more people are involved in a decision. That is,
deCiSions performed by individual decision-makers or small groups of people may indeed
be biased due to a number of powerful behavioral inﬂuences, while those inﬂuences may
“0‘ have the same powerful effect, or may even out, as crowds grow into markets. Thus,
the Process underlying how, for example, stock prices are formed reﬂects collective
Perceptions that individually may be biased, but may be relatively more bias free at the
l(“’61 of the collective to be relied on as indications of performance. In fact, this
assumption of rationality is a sufﬁcient precondition for productive management
resSnitch, including the investigation of antecedents to ﬁrm performance.

In this dissertation, however, I argue based on a signaling perspective that
assul'l'tptions of rationality, although sufﬁcient, may not be necessary for my market-
related ﬁrm performance measures (diversiﬁed ﬁrm value, corporate effects, ﬁrm ROA).

That is, under conditions of information asymmetry where executive intentions and

47

has can as c

run 35:1" "‘1
321:»: “mean
D I

16520213 SEOLLC

i‘l‘x- -. .l
kaiser“

units 3111'.
a n J

fie“: on the:

actions cannot be efﬁciently monitored, investors and boards look for signals to be able
to gauge economic value (Carter & Manaster, 1990: Leland & Pyle, 1977: Spence, 1976)
as well as executive contributions to such value. In addition, executives also look for
credible signals to gain insight into what is valued by the market, including what “good”
decisions should look like and what outcomes they should result in. Thus, the rationality
of those signals may not matter, i.e., whether they are efﬁcient or not, since top
executives still are likely to pay close attention to them because their performance (as
perceived by investors and boards) and, subsequently, their pay will to a large extent
depend on them.

Translated to the current context, I argue that my measures of performance
outcomes (diversiﬁed ﬁrm value, corporate effects, and ﬁrm ROA) contain such
important signals. Following prior arguments, they are important both because executives
can affect them by allocating capital in a relatively more efﬁcient manner, and because
they are, in turn, likely to affect executive pay. Diversiﬁed ﬁrm value, for example,
°°nVeys to investors and boards how much value executives add or destroy by
diversification including the efﬁciency of the internal capital allocation process. As such,
managers may pay close attention to diversiﬁed ﬁrm value (as well as to the other two
Incasures) as important signals of what the ﬁrm is likely to reward and are therefore
likely to adjust their behaviors accordingly.

Environmental dynamism has been associated with difﬁculties in decision-making
(e.g., March & Simon, 1958: Mintzberg, 1990) both theoretically and empirically. A
nunlber of scholars (e. g., Duncan, 1972: Von Krogh, Roos & Slocum, 1994: Weick,

1 979; Milliken, 1987, 1990) have suggested that environmental dynamism increases

48

 

. 4500')“
ishvmm.‘

313135 0

“..rral't
“Sh-hon-
I

> ' I.
. ‘31"

1 intUbUA I‘
0

”h seq- - :0
bL.

“Lu-‘5‘.-

ILR ii c;
3“! o

a... ,0 :1

'ﬁ'bnt“.
~‘Elu‘ 'e e!

environmental uncertainty, which in turn affects the rationality of decision-making
outcomes of organizational members. This perspective, referred to as the “information
uncertainty perspective”, suggests that because the environment provides information for
boundedly rational organizational decision-makers, the rate of change or instability of the
environment should affect the availability and reliability of information. In other words,
under conditions of high environmental dynamism, decision-makers intending to make
rational decisions will face situations that are ambiguous in terms of both alternative
actions and criteria to evaluate those alternatives. Thus, decision-makers will face
“tremendous cognitive demands” (W iersema & Bantel, 1993: 488) as they “constantly
need to adapt their perceptions of the environment to ﬁt the current reality” (Wiersema &
Bantel, 1993: 488) when assessing both current and future states of the environment. In
Short. as environmental dynamism increases, they will face less useful information upon
which to base their decisions, thereby complicating their decision process with potentially
m3gative effects on decision outcomes.

This direct effect of environmental dynamism on decision outcomes including
ﬁn“ performance has been empirically studied by a number of scholars. Wiersema and
Baillie] (1993) investigate direct effects on a number of variables including ﬁrm
performance, strategic change, and top management team turnover, while Palmer and
vVis'ﬁman (1999) study effects of environmental dynamism on organizational risk. Both
find signiﬁcant direct effects on their respective outcome variables.

Translating the logic of an association between environmental dynamism and
decision-making uncertainty to the context of capital allocation in multidivisional ﬁrms,

e . . . . . .
11" lronmental dynamism 18 related to the investment allocation process by havmg a

49

iio;~' 3.0;“!
‘55! iliﬁ~‘

.' .. 1‘!- ﬂ“
:s:i~‘..'n.t

I
.._..l--; y-zu
, H.551»...

'1‘,“ . .°"
n-VI. “'3 IV

‘1’

u
5......“ Y“
vub‘h . \\

argue v

0....15.

I". u Tip!
‘34!
l"
L‘tH-h
. U. i ‘l
.i.‘“
I
I
.r“. h
shay”-
ML“
a

.4

u“

k‘t‘iﬂr

957383

direct effect on capital allocation efﬁciency. That is, assuming boundedly rational
decision-makers with individually ﬁxed cognitive limitations, segment prospects may
become harder to forecast as the environment becomes increasingly more dynamic and
more apt to change unexpectedly in the future. The allocation decision therefore, in
effect, becomes more uncertain since underlying information needed to make this
decision (e.g., projections of segment prospects) may not be available, may be less
reliable, and/or may be more likely to change in the near future. This increasing
uncertainty in correctly assessing future prospects is likely symmetrical in its effect on
the efﬁciency of the capital allocation process — overestimation of segment prospects
(leading to overinvestment) may potentially be as common and severe as underestimation
(leading to underinvestment). Thus, the efﬁciency of the capital allocation process may

be adversely affected by increasing environmental dynamism, resulting in more instances
and higher magnitudes of over and underinvestment.

Formally stated:

Hypothesis 6a: Environmental dynamism is positively associated with
overinvestment.

Hypothesis 6b: Environmental dynamism is positively associated with
underinvestrnent.
Environmental complexity refers to the heterogeneity of an environment
including the range and variability of factors needed to be considered in strategic
deciﬁion-making (Child, 1972). As such, increased heterogeneity leads managers to
“Perceive greater uncertainty and have greater information-processing requirements than
mamigers facing a simple environment” (Dess & Beard, 1984: 56) because they must

now Operate in an environment with, for example, a larger number and variety of

50

£32135 Lit
{iii 3:. Cl 5'

- ' l
q u-te‘q.~‘-"F
£~21Liu :AAJA‘

33.1.. Till I
A

3"de \3.’

'II; ”9109' Q‘.
duh KARL: H»:
v

"2"? r "_
l-u.‘ on .35
‘ 1

'. y _.
Ali ‘55 16:15,.
an;

. h‘ t”.
'5‘“ n:- “he
‘ b

:ﬂl’ﬂ'a Tat

no“ “a.

ids -

. W\t5l0r
l'”,

by, l :1\€\

competitors using an expanded variety of inputs and outputs. In effect therefore, to
achieve an overview (of a more complex environment) required for effective strategic
decision-making may necessitate a much higher level of information processing on behalf
of corporate managers. This was also suggested by Wiersema and Bantel (1993: 489):
“Increased variety and diversity of a ﬁrm’s business operations suggest that a wider range
and greater quantity of information be processed for effective decision making, creating
strains on the organization to achieve high quality decision making outcomes”. Thus, at
the very least, operating in a more complex environment is a much more difﬁcult task
requiring information processing resources that may not be available from existing
corporate managers, therefore affecting performance outcomes of decision-making
processes.

Although a search of major management journals returned only a few empirical
studies that have investigated the direct relationship between environmental complexity
and decision-making outcomes, Wiersema and Bantel (1993), and Palmer and Wiseman
(1999) investigate this relationship by looking at how complexity affects ﬁrm
performance/top management team tenure and organizational risk, respectively.
Wiersema and Bantel ﬁnd that environmental complexity decreases both ﬁrm
Performance and top management team tenure, the latter presumably because increasing
complexity over time creates managerial demands current managers are unable to deal

With. Palmer and Wiseman, however, ﬁnd no support for their hypothesis of a direct
Positive relationship between environmental complexity and organizational risk.
Finally, Siggelkow and Levinthal (2003) and Siggelkow and Rivkin (2005)

conclude that environmental complexity directly impacts the choice of optimal

51

. U
, . at.“ , I I
:13.“ Earl}: .
.1.

U
1
. .~-.0- '1
I
‘V‘IJ .p a

.
"’wisk‘lh us»
‘ O

' . :0.
'n r: n' a
.‘J..\uu’eut bit

I

mes :td ::

‘ .
h.:\q ‘ e‘\ if“

mmmg

h .
an. 9'05

a. n _
“bi: 31‘ Q‘s-t:
5

organizational. structures, search routines, rewards and several other decisions in the
organizational context. Thus, the described studies together suggest that environmental
complexity increases the difﬁculty of top management decision-making processes with

subsequent effects on turnover, ﬁrm performance, organizational structures, search

routines and rewards.
This need to cope with an increasingly diverse and complex environment may be

eSpecially salient in regards to the capital allocation process since the allocation decision
effectively involves all potential uses for capital including business segments in the ﬁrm.
In particular, for boundedly rational decision makers to effectively execute the allocation
decision, an overview of segments including information on both current and future
SE=glnent prospects is needed. Thus, because the capital allocation process requires inputs
from all segments and its efﬁciency is highly dependent on the quality of those inputs, it
may be especially susceptible to complex environments, making the allocation decision

increasingly difﬁcult as the level of complexity increases. This implies that the

Complexity of the environment is likely to directly affect the efﬁciency of the capital

allocation process by leading to greater inefﬁciencies. Formally stated:

Hypothesis 7a: Environmental complexity is positively associated with

overinvestment.

Hypothesis 7b: Environmental complexity is positively associated with

underinvestment.

Environmental uncertainty: moderation effects
The following three hypotheses (H8-10) will provide support for a number of

l()Cleration effects where environmental uncertainty moderates associations between

1
evel of diversiﬁcation and corporate governance variables on the one hand, and capital

52

n

innate
time com
Kenna
.‘t in: bad. it 1
manna
11311:!) :1;

at”??? 'Kﬂ-w;
‘-.kub“‘ *y““
|

hmmsh
in: time. hour
a: casino:
Wt 13 ex

“P336 and c

A5 d8:

lat? .* .
“dilemma:

 

allocation efﬁciency on the other. No hypothesis moderating the relationship between
performance comparisons and capital allocation efﬁciency will be provided, however,
because it is not clear how this relationship is affected by environmental uncertainty. On
the one hand, it may be that managers are less likely to be able to determine gain or loss
contexts, or view them as less sustainable, under conditions of high environmental
uncertainty. That is, current business segment performance in relation to social and
historical aspirations may be harder to estimate as well as likely to be perceived as less
sustainable under more dynamic and complex environments leading to a weakening of
the contexts, thus resulting in less inefﬁciencies in the capital allocation process. At the
Salne time, however, managers may view some contexts as more worthy of protection
uIlder conditions of high environmental uncertainty. For example, gain contexts may be
Perceived as even more worthy of protection if viewed as harder to sustain in a highly
dynamic and complex environment. Thus, because the moderating effect of
enV ironmental uncertainty on the associations between performance comparisons and

capital allocation efﬁciency is not clear, I am not offering any formal hypotheses

describing those relationships.
As described in hypothesis 5, the relationship between level of diversiﬁcation and

inefﬁciencies in the capital allocation process is explained by focusing on the uncertainty

of the allocation decision as well as on rigid bureaucratic allocation rules. That is,

Irlc=1‘eased diversiﬁcation stretches decision-makers' limited information processing ability
even thinner and exacerbates the problem of rigid allocational decision rules, thus,

affecting the efﬁciency of the capital allocation process by leading to more over and

Lllclerinvestment.

53

1 1'. 1
BLLILU'ZE

I
$333533 L
war": :33“ I

.nS-‘n‘b mu.
5

ms ~~

h .w k .5.
v w. “I
ﬁ’vﬁfgsfﬂih

Embim

151%

.. manner.
35:32 L. 16?
.
1mm.
3:151:33 33

“‘4”. , w

an.
.nruema
“ﬁst: 1“
."\.. .‘ , i
v Clef
10

Building on those arguments, I suggest here that environmental dynamism and
environmental complexity as previously deﬁned17 are likely to further exacerbate the
negative effect of increased diversiﬁcation on the efﬁciency of the capital allocation
process through much the same processes as those described in H5. First, increased
environmental dynamism and complexity by forcing decision-makers to face situations
that are both more ambiguous, more likely to change more often, and that require more of
an overview to make efﬁcient decisions are likely to stretch decision-makers' already
bounded information processing ability even thinner. That is, the processing of each piece
0f information in a more dynamic and complex environment is likely to take longer
1'ﬁsulting in less information processed to support each decision. Put differently, as

deCision-makers face increasing difﬁculties, their available information processing
Capabilities are likely to be exceeded even earlier, resulting in more instances of over and
Underinvestment.

In addition, as the level of environmental change and complexity increases, rigid
allocation rules observed and suggested by researchers (e. g., Bower, 1986: Shin & Stulz,
1 998) are also likely to lead to increasingly more over and underinvestment. For
e"iii-11:1ple, as the environment becomes increasingly apt to change, i.e., increasingly
dynamic, it follows that business segment prospects may also be more likely to change in

unpredictable and sudden ways. Thus, following rigid allocation rules that are not ﬂexible

.\

7
Q QEDVimnmental dynamism is the rate of change or instability of the environment, while environmental
I“lplexity refers to the heterogeneity of an environment including the range and variability of factors

t1ﬁe<1ed to be considered in strategic decision-making (Child, 1972).

 

54

I 4‘”

..‘ 91- l.

1.: Nab“:
V

t\ cut.
”it n. :
5' $¢.u&u -

\-

-
.0

1.3315165

‘ ‘
.1”:

Mil 52;:
Hypo
bf.“ n t
such

mix.

{1? ,-
“ems:

t1
magma

enough to adapt to a changing environment'8 is likely to lead to additional instances of
over and underinvestment as the level of environmental dynamism increases”.

In sum, I argue that environmental dynamism and complexity likely moderates
the relationship between level of diversiﬁcation and the capital allocation process by,
ceteris paribus, making it more difﬁcult for decision-makers to allocate capital as well as
exacerbating the negative effects of rigid allocation rules leading to increased
inefﬁciencies as the level of environmental dynamism and complexity increases.
Formally stated:

Hypothesis 8a: Environmental dynamism will moderate the positive association
between level of diversiﬁcation and ineﬂiciencies in the capital allocation process
such that a higher level of environmental dynamism will strengthen the

relationship.

Hypothesis 8b: Environmental complexity will moderate the positive association
between level of diversiﬁcation and inefﬁciencies in the capital allocation process
such that a higher level of environmental complexity will strengthen the

relationship.
Underlying the direct effect between shareholder concentration and capital allocation
efﬁciency (HZ-3) is the suggestion that large shareholders are efﬁcient monitors of top
Inanagers (e. g., Pound, 1988). That is, block holders and institutional shareholders are
able to force value maximization by controlling and reducing instances of over and

uIlderinvestment in response to managerial risk aversion, self-interest, agency costs at the

allOcational level, and escalation of commitment issues.

\
l 8
Although not described, a similar mechanism accounts for how a more complex environment moderates

1e association between diversiﬁcation and inefﬁciencies in the allocation process (see H7 for guidance).

 

l 9
I acknowledge that decision heuristics such as rules and routines often are used to economize on

co
8 - gllition. However, this lack of independence is unlikely to affect the direction of the moderation effect
e, ceteris paribus, increased use of rigid allocation rules, even as a result of bounded rationality, leads to

i . .
“creased capital allocation inefﬁcrencres.

55

3w «Man-3"
‘1..._LAL.AI§|

inﬁrm if

C, ~“"-n
J“. .ki.

V
. h
u.“ an-

on r.

ICA“\ .

"'l H: )- ‘p
\. I 9
“JR. bug“

Yiéﬂl ~,_
-‘~ 55t.tu
- 4

' .
‘“ “‘u-r. u

l
Té’a:-
“I 5‘ 3:8

I
"5 w- .
Nb§rde§

 

Their monitoring effectiveness may, however, not remain constant across various
levels of environmental uncertainty. As explained earlier when describing effects of
environmental uncertainty on capital allocation efﬁciency (H6-7), just like the allocation
decision becomes more uncertain as the level of environmental uncertainty increases, the

monitoring of that decision may also face the same increasing uncertainty. That is, the
effect on managers and large shareholders may be relatively symmetric in this respect

since each face the same increasingly uncertain projections of future prospect (i.e.,
Projections of future prospects may be less reliable, more likely to change in the near
ﬁlture, and/or may not be available at all) as the level of environmental uncertainty
increases. Thus, since efﬁcient monitoring presumes that large shareholders are able to
differentiate between good and poor prospects, thereby being able to differentiate
between efﬁcient and inefﬁcient (over and underinvestment) capital allocation, the
monitoring process becomes compromised as they lose that ability. By extension
therefore, their ability to monitor and enforce efﬁcient capital allocation decisions
becomes less pronounced in more uncertain environments. Support for this contention is
Provided by Li and Simerly (1998) and Zajac and Westphal (1994). Li and Simerly argue
that “under conditions of greater environmental dynamism, the effectiveness of
monitoring by owners of the behavior of top managers will be extremely difﬁcult if not
i111possible” (171). Thus, as environmental uncertainty increases, owners are thought to
be less able to “gauge whether the ﬁrm’s goals and strategies are consistent with owners
Obj cctives” (171). In addition, Zajac and Westphal found that ﬁrm-speciﬁc risk affected
the efﬁcacy of managerial monitoring put in place to attenuate agency costs. Together,

t1lese studies indicate that the monitoring effectiveness of large shareholders may not

56

ﬂ ‘-
': it? ‘0 n
‘05 Bu» :1,

‘-

' 9“ ‘taa aé
“ti-utter“ El

vira. “

"theft. &
T133235 set

V~
l ‘ 'I...
Khuf “5‘“
l

@515 Slim

551"

mime:

ML the 35

v
.
l.

l .
"1063:; 0n pm

55pr ['19

remain constant as the level environmental uncertainty changes. Therefore, I present the

following hypotheses:

Hypothesis 9a: Environmental dynamism will moderate the negative associations
between black holder and institutional ownership and ineﬁ‘iciencies in the capital
allocation process such that a higher level of environmental dynamism will

weaken the relationships.

Hypothesis 9b: Environmental complexity will moderate the negative associations
between black holder and institutional ownership and ineﬂ‘iciencies in the capital

allocation process such that a higher level of environmental complexity will
weaken the relationships.

The direct negative relationship between contingent compensation and capital allocation
inefﬁciencies (H4) requires that managers have access to reasonably accurate predictions
0f segments’ future prospects. They need to be able to differentiate between segments
With good versus poor prospects to be able to avoid misallocating capital. By extension
therefore, as the uncertainty of the allocation decision increases because underlying
bl-1Siness segment prospect information cannot be acquired, is more ambiguous, less
I‘eliable, and/or more likely to change in the near future, the power of contingent
Compensation to assure wealth maximization through efﬁcient capital allocation is
deCreased. That is, managers driven by high powered incentives to avoid opportunity
c'DSts stemming from inefﬁcient capital allocation may, under conditions of high
environmental uncertainty, still not be able to avoid such costs, simply because they are
1eSS able to differentiate between segments with good versus poor prospects. Thus, as a
result, the association between contingent compensation and inefﬁciencies in the capital

allocation process weakens as the level of environmental uncertainty increases. Zaj ac and

We stphal (1994) partially support this contention by ﬁnding that ﬁrm-speciﬁc risk

57

:tifilrs '

l1“: "r9

455”.»

moderates the effectiveness of managerial incentives in controlling agency problems.
Therefore, I suggest the following hypotheses:

Hypothesis 10a: Environmental dynamism will moderate the negative association
between contingent compensation and inefficiencies in the capital allocation
process such that a higher level of environmental dynamism will weaken the
relationship.

Hypothesis 10b: Environmental complexity will moderate the negative association
between contingent compensation and inefﬁciencies in the capital allocation
process such that a higher level of environmental complexity will weaken the
relationship.

58

 

CHAPTER 2: THEORETICAL MODEL DESCRIBING PERFORMANCE
IMPLICATIONS OF CAPITAL ALLOCATION EFFICIENCY

Performance implications of capital allocation efﬁciency

This dissertation has so far emphasized how an internal or miniature capital
market put in place to attenuate market contracting hazards may itself be subject to a
number of business-segment, ﬁrm, and industry characteristics that may affect its
efﬁciency by resulting in internal market failures. As such, I have investigated a number
of antecedents to internal capital allocation market failures (measured as over and
underinvestment) including stronger and weaker prior firm and competitor performance,
increased environmental uncertainty, and increased levels of diversiﬁcation. At the same
time, I have outlined arguments positing that three agency-theoretic corporate governance
variables may lead to relatively fewer instances and magnitudes of inefﬁciencies in the
process.

The preceding investigation of antecedents to an internal capital allocation
process varying in efﬁciency illustrates the increasing importance of the corporate ofﬁce
for ﬁrm performance. By extension, and assuming in line with Williamson (1975) that
the capital allocation process is important for ﬁrm performance, the importance of the
corporate ofﬁce for ﬁrm performance should increase as the efﬁciency of the capital
allocation process now becomes an important comrate-level antecedent to ﬁrm
performance. Thus, a natural next step of this dissertation is to investigate performance
implications of this internal capital allocation process including how internal market
failures, measured as over and underinvestment, are associated with the corporate effect

as a measure of how much value the corporate ofﬁce adds to business segments, i.e., a

59

measure of the corporate contribution. Hypotheses 11 through 13 outline how over and
underinvestment are associated with the corporate effect, as well as how these
associations may be moderated by the variance of a ﬁrm’s business segment returns. In
addition, associations between over and underinvestment on the one hand, and diversiﬁed
ﬁrm value and ﬁrm ROA on the other, are examined in hypotheses 14 and 15,
respectively (Figure 2).

This chapter’s investigation into the capital allocation process and its performance
implications begins with the ambiguous relationship between capital allocation efﬁciency
and the corporate contribution with the latter measured, as in previous studies, by the
corporate effect (e.g., Brush, Bromiley & Hendrickx, 1999: McGahan & Porter, 1997).
On the one hand and as described next, theory suggests that corporate-level factors in
general (e.g., Rumelt, 1974: Chandler, 1962: Hansen & Wemerfelt, 1989: Andrews,
1987: Hambrick & Mason, 1984) and the efﬁciency of the capital allocation process in
particular (Williamson, 1975, 1985), may enhance corporate-level inﬂuences on ﬁrm
performance, i.e., the corporate contribution. Thus, inefﬁciencies in the latter process can
be expected to decrease the corporate contribution. In addition, the process of ﬁrm
performance variance decomposition where corporate-level inﬂuences (such as the
efﬁciency of the capital allocation process) are assessed separately from business
segment and industry inﬂuences may also lead us to expect a negative relationship
between capital allocation inefﬁciencies and the corporate effect.

On the other hand, the corporate contribution as measured within the ﬁrm
performance decomposition research stream, i.e., the corporate effect, may be

systematically biased (e. g., Bowman & Helfat, 2001) leading one to, under some

60

ell. mmmemmm \
>. .VZU.’ Una-..nvu ZA V..-.<.. VA V-.-< .-l\.-.-a-<. V .L-AV 12A V-.-.<. Vhtinilxzh mi. VZ<E£V~AV3i=U~A~

N lea 'WVI...

 

 

 

 

 

 

 

 

 

>0“. 25:00:.— a...

 

 

 

 

 

 

 

3:53:50
can Engage
358.025
.5:0E:2_>:m

 

 

o

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

:05. 0.03903
5:02:9030
he .05..

 

 

 

 

 

0030...;
02.05030
0.0.0900

 

 

 

u. uuuuuuuuuuuu 1 .538 .o .30 .
_ 3.22:. a a m ..
. 26E 3000 , .
. 05:0... Omo .
u 3.0 =2§c35 ..
u. .1 <0: E..."— 00: 3. .la...c> .2200 x
_
_ a
_
. aces—825.01g
" use .90.
.. I .v >822=m
_ 0:_0> 5...". 5:30.?
“ $059.02.“. .8250
n f
u 3:: _ _ 3:: _
_ /
.. I .v .85 \
0.03900 005:;
£33.
508000
0005030 3:33500
use 835253
253.095
3.8:...— uaeﬁz A .................. 3585.35
emu—maemzim 3.3.30 4. IIIII

328%: 25852:: Al

 

 

 

 

 

 

 

0:029:02
00:08.25."—
_0_oom new 30.3.0.1
EoEmom

 

 

>02H—UEH ZOF—ZUOAi 44.5.35 EC mZOFH<U—HE Hogmchmm—m

N gnu-h

61

circumstances, expect a null or a positive relationship between capital allocation
inefﬁciencies and the corporate effect. Thus, to reﬂect and account for divergent
theoretical arguments and arguments pointing to systematically biased empirical
measures of the corporate contribution, multiple hypotheses are offered where the
relationship between inefﬁciencies in the capital allocation process and empirical
estimates of the corporate contribution is modeled both as a direct negative relationship
(H11) and a relationship that is moderated by the variance of business segment returns
within fums (H13). In addition, a direct negative relationship between the variance of
business segment returns and the corporate effect is also hypothesized (H12).

The ﬁrm performance decomposition framework described earlier is a way to
disentangle sources of ﬁrm performance into variance from business segment,
corporation and industry. Thus, the logic behind that framework suggests that corporate-
level factors, actions and resources such as scope of the ﬁrm (Rumelt, 1974) where ﬁrms
choose to operate in more proﬁtable industries, vertical integration and M-form structures
where ﬁrms adopt more efﬁcient integrated structures as well as internal capital markets
(Williamson, 1975, 1985), core competencies where valuable competencies span many
business segments (Chandler, 1962), better organizational climates (Hansen &
Wemerfelt, 1989), and better and more able corporate management (e. g., Andrews, 1987:
Hambrick & Mason, 1984) should enhance the corporate contribution to total ﬁrm
performance. Thus, ﬁrms sharing these corporate-level factors should logically have a
higher corporate contribution as measured by the corporate effect relative to ﬁrms that do
not share them to the same extent. Bowman and Helfat (2001) and Adner and Helfat

(2003) support this logic by suggesting a clear effect of “factors internal to the ﬁrm at the

62

corporate level” (Adner & Helfat, 2003: 1011) on ﬁrm proﬁtability as they ﬁnd that
“dynamic managerial capabilities” reﬂected in corporate strategic decisions add a
statistically signiﬁcant increment of variance in ﬁrm proﬁtability.

Applying this logic to the association between efﬁcient capital allocation and the
corporate contribution is straightforward. First, since the capital allocation decision
clearly resides at the corporate-level with a ﬁnal say in the allocation decision being
retained by headquarters (e. g., Stein, 1997), the association between the capital allocation
process and the corporate contribution is just as clear as any other action or process
emanating from the corporate ofﬁce. Second, and as described earlier, Williamson (1975)
suggests that the capital allocation decision may be one of the most important
determinants of ﬁrm performance. Thus, based on the fact that the capital allocation
decision clearly resides at the corporate-level, in addition to its theoretical importance
pointing towards it being a major determinant of ﬁrm performance, it should logically
also be an important corporate-level determinant of f'u'm performance and thus be closely
associated with the corporate contribution.

An example illustrating how ﬁrm performance is decomposed to measure
corporate-level inﬂuence on business segments may clarify this direct theoretical
association between the efﬁciency of the capital allocation process and the corporate
contribution (e. g., Williamson, 1975). That is, the common use of ﬁrm ROA to proxy for
the corporate effect by regressing business segment performance on ﬁrm ROA20 (see for
example Bromiley, Brush & Hendrickx, 1999) suggests a direct positive association

between (the efﬁciency of) the capital allocation process and the corporate contribution

 

2° In later studies using multilevel models, corporate effects are estimated in a similar way as business
segment performance is regressed on ﬁrm ROA leaving the residual to represent business segment effects.

63

c L»

 
 

xi.»
..

 

as measured by the corporate effect. Suppose, for example, that a ﬁrm has three equal
sized business segments, two that perform well with correspondingly high ROA (15 and
10 %), and a third segment that is not quite as successful showing a substantially lower
return on assets (5 %). An equal allocation between these three segments would account
for an overall ﬁrm ROA of 10 %. If, however, the corporate ofﬁce exercises its right to
allocate investment capital away from the poorly performing segment in favor of the two
seemingly better performing ones, overall ﬁrm ROA is likely to increase. For example,
allocating a more reasonable 50/50/0 split between the three segments, respectively,
implies an overall ﬁrm ROA of 12.5 %21. This would also increase the corporate effect as
business segment returns are regressed on the higher level of overall ﬁrm ROA. That is,
overall ﬁrm ROA would explain more business segment return variance resulting in a
higher corporate effect. Thus, based on theoretical insights and the way those insights
have shaped the methods used to decompose ﬁrm performance, it follows that a more
efﬁcient allocation of capital should increase the corporate effect. Therefore, I expect a
negative association between inefﬁciencies in the capital allocation process and the
corporate effect. Formally stated:

Hypothesis 1]: Ineﬁ'iciencies in the capital allocation process will be negatively
associated with the corporate eﬁect.

 

2' Note that this example would also apply to two ﬁrms with similar performing business segments (15, 10
and 5 %). After reorganization, one of the ﬁrms divests the poorly performing segment (5 %), thus will be
credited with a higher corporate effect as a direct result of more efﬁcient capital allocation. Indeed, there is
some evidence that business segments performing below some hurdle rate of return are divested (See
Bromiley (1986) or Bower (1986) for a discussion and empirical evidence on hurdle rates and divestitures).
Since these divestitures represent a natural decrease in investment funds (to zero) and a reallocation of
freed-up funds to other business segments, they are and should be reﬂected in overall ﬁrm ROA and need
not be controlled for.

-r

U.-.‘I

ﬁl'.
-

~LA§

1.».
L.\_'I
l

t: ' -
.‘ ‘

‘ -
~

'v
u‘.‘ v\

‘K

M

- n»

' a

‘h;
0...

 

Building on Bowman and Helfat’s (2001: 5) conclusion that “empirical estimates of
corporate effects may not fully and accurately capture the inﬂuence of corporate strategy
on the variance of proﬁtability”, I suggest that statistical artifacts inherent in the way
business segment performance variance is decomposed in most ﬁrm performance
decomposition studies systematically biases empirical estimates of the corporate
contribution. Subsequently, this complicates the association between capital allocation
efﬁciency and those estimates. Speciﬁcally, it is the treatment (or lack of treatment) of
the within ﬁrm variance of business segment performance that distorts the corporate
effect. Since individual business segment returns are regressed on overall ﬁrm ROA
regardless of decomposition method (e. g., variance decomposition, ANOVA, I-ILM),
more variance is explained if business segment returns are homogenous, thus resulting in
a higher corporate effect. That is, the corporate contribution as measured may not only
derive “from differences between multiple-business ﬁrms in the average of returns to
individual businesses within each corporation” (Bowman & Helfat, 2001: 3) as intended,
but also from differences in the level of variance of business segment performance within
ﬁrms where relatively more “homogenous ﬁrms” have higher corporate effects and
relatively less homogenous ﬁrms have lower corporate effects. Thus, ceteris paribus,
lower levels of business segment return variance may directly raise the corporate effect.
As one example illustrating this direct effect, consider two different ﬁrms
operating in the same industry/business segments, with similar overall ﬁrm ROA (10 %),
but with differences in the patterns of returns within each ﬁrm. The ﬁrst ﬁrm has a large
amount of dispersion (low homogeneity) in its business segment returns (5, 10 and 15 %,

respectively), while the second ﬁrm has a more homogenous pattern of returns where the

65

three segments have ROAs of 9, 10 and 11 %, respectively. Since less dispersed business
segment returns of the latter ﬁrm are regressed on same ﬁrm ROA, the latter ﬁrm would
be credited with a much higher corporate effect. Thus, the second ﬁrm’s more
homogeneous return distribution raises the corporate effect relative to the ﬁrst ﬁrm.

To more clearly see how this analytical artifact inherent in the measurement of
corporate contributions increases the corporate effect, consider that only the “lowest

average improvement”22

across business segments is reﬂected in corporate effects. That
is, although the corporate ofﬁce may contribute just as much to proﬁtability in each of
my two example ﬁrms (more dispersed versus less dispersed business segment
proﬁtability), the contribution is much more uneven for the former ﬁrm, resulting in a
lower corporate effect. That may be because the former ﬁrm (but not the latter) has
access to some very powerful corporate-level resource (such as Apple’s ability to design
exceptionally sleek MP3 players, computers, and now also phones) that the ﬁrm is able to
apply to the products sold by the third business segment (15 % ROA), but not necessarily
to the products sold by the other two lower performing segments. This particular
corporate-level resource may therefore greatly improve the performance of the third
business segment compared to competitor segments, but may only slightly (or not at all)
improve the other two segments. Thus, because empirical estimates of the corporate
contribution represent the lowest average improvement across business segments, the

strong but uneven effect from the design advantage will not be fully represented by

corporate effects. In other words, the ﬁrm with less dispersed business segment returns

 

22 Bowman and Helfat (2001) refer to this as the “average improvement” meaning that it is the lowest
common denominator improvement across business segments. This interpretation of “average
improvement” becomes clear as they subsequently suggest that the business effect soaks up the remaining
variance that should have been attributed to corporate strategy in ﬁrms with more dispersion in business
segment returns.

66

-n ‘ ‘
q 39“ ’1'
Q“, rd .

29-156 Vt

- 7a;v-' '
{gnytl's‘E‘

l3 1

2;
l-_
I
('10
(.-

H;

u ’<
OI

- ' v
.35 .EW.

I '
a ,"F
tin-xv. a...

.-

‘1—“41‘1 .’
v""“ I . Us.

‘ ",..4.J
“-‘~¢“~
l

will end up with a larger corporate effect just because it may have a corporate-level
resource with a low to moderate impact on the performance of m of its three business
segments.

In sum, therefore, I expect increasing levels of business segment return variance
to have a negative effect on the corporate effect. Stated formally:

Hypothesis 12: Business segment return variance will have a negative association

with the corporate eﬂ'ect.
The level of business segment return variance within ﬁrms may also moderate the
relationship between capital allocation efﬁciency and estimates of the corporate
contribution on ﬁrm performance. As suggested by Bowman and Helfat (2001: 3),
corporate effects may not only reﬂect the “true” corporate contribution as characterized
by “differences between multiple-business ﬁrms in the average of returns to individual
businesses with each corporation”, but may also reﬂect the level of dispersion of business
segment performance within ﬁrms, i.e., relatively more “homogenous” ﬁrms have higher
corporate effects. If that is true, business segment return variance as a previously
unmeasured artifact of how corporate contributions are measured may inﬂuence the
relationship between capital allocation inefﬁciencies and the corporate effect (H11). In
other words, the relationship between capital allocation inefﬁciencies and the corporate
effect may change depending on the level of business segment return variance,
suggesting a joint effect on the criterion variable. Thus, a moderating effect would take
place where the negative association would stay the same at higher levels of variance, but

may weaken at relatively lower levels of segment return variance.

67

In sum, based on the moderating inﬂuence of business segment return variance on
the relationship between capital allocation efﬁciency and the corporate effect, I expect

the following:

Hypothesis 13: Business segment return variance within ﬁrms will moderate the
association between capital allocation ineﬁ‘iciencies and the corporate eﬁ'ect such
that at high levels of segment return variance a negative association may be
expected, while a null or a positive association may be expected at low levels of

return variance.

Empirically, the association between inefﬁciencies in the capital allocation process and
diversiﬁed ﬁrm value is not hard to defend. As mentioned earlier, a number of
researchers have found evidence that inefﬁcient capital allocation at the corporate-level
contributes to differences in valuation between freestanding business segments and those
under the corporate umbrella, i.e., they have linked inefﬁcient capital allocation with
diversiﬁed ﬁrm value (e.g., Berger & Ofek, 1995: Shin & Stulz, 1998: Maksirnovic &
Phillips, 2002: Comment & Jarrell, 1995).

A main difference, however, between those studies and mine is how capital
allocation inefﬁciencies are treated and measured. In line with Jensen (1986, 1989, 1991,
1993), most of the above studies emphasize overinvestment, but spend little or no time
investigating instances of underinvestment in business segments with relatively better
prospects. Although one could argue that with a binding capital constraint,
overinvestment in one segment would automatically result in underinvestment in another,
this is not likely to hold in modern corporations where available investment capital
instead can, and often is, allocated to buy-backs, dividends and other uses at the

discretion of corporate managers (Dow & Gorton, 1997).

68

To address this limitation of past research, this dissertation uses a more complete
measure of inefﬁcient capital allocation where over and underinvestment are measured
separately, offering the possibility that they may be fairly independent. That is, the
presence or absence of overinvestment does not automatically preclude either the
presence or absence of underinvestment. In addition, although the inﬂuence of over and
underinvestment on ﬁrm performance outcomes are measured separately, I suggest that
they have similar symmetrical negative effects on ﬁrm performance outcomes including
the corporate effect, diversiﬁed ﬁrm value, and ﬁrm ROA.

Theoretically, the association between inefﬁcient capital allocation and
diversiﬁed ﬁrm value can be viewed as an outgrowth of the emphasis that Transaction
Cost Economics (Williamson, 1975, 1985) traditionally has put on the capital allocation
function. Williamson (1975: 148) points to this function being “the most fundamental
attribute of the M-form enterprise”, indicating its powerful effect on ﬁrm performance. In
particular, and noting that Williamson assumes that the choice between organizational
form always is made in an effortless and optimal manner, his conclusion is that the
internal capital allocation mechanism reduces transaction costs by replacing the less
efﬁcient external mechanism that single-segment ﬁrms rely on, thereby resulting in better
ﬁrm performance.

This is not likely to hold true, however, if clear inefficiencies are observed within
the internal mechanism. At the very least, as inefﬁciencies in the form of over and
underinvestment increase, performance implications for the diversiﬁed firm are likely to
become increasingly negative, leading to an increased discrepancy in valuation between

diversiﬁed and single-segment ﬁrms. In other words, inefﬁcient capital allocation in

69

diversiﬁed ﬁrms is likely to lead to a lower diversiﬁed ﬁrm value where business
segments under the corporate umbrella are valued lower than similar freestanding
segments. Thus, based loosely on transaction cost economics and its emphasis on the
capital allocation function, I expect that:

Hypothesis 14a: Overinvestment will be negatively associated with diversiﬁed
ﬁrm value.

Hypothesis 14b: Underinvestment will be negatively associated with diversiﬁed

ﬁrm value.
Inefﬁciencies in the capital allocation process will also affect a ﬁrm’s return on assets
(ROA). As argued earlier, inefﬁciencies will lower overall ﬁrm ROA as investment
capital is not allocated to business segments with the highest prospects and ROA. That is,
overinvestment where “too much” capital is allocated to business segments with
relatively lower prospects and ROA, and underinvestment where “not enough” capital is
allocated to business segments with relatively better prospects and ROA, will lower
overall ﬁrm ROA as weighted segment ROAs are aggregated to the ﬁrm level. Based on
that, I expect that both under and overinvestment will lead to lower ﬁrm ROA.
Formally stated:

Hypothesis 15a: Underinvestment will be negatively associated with ﬁrm ROA.

Hypothesis 15b: Overinvestment will be negatively associated with ﬁrm ROA.

7O

CHAPTER 3: METHODS

SAMPLE

The model requires data at business segment, ﬁrm and at industry levels. At the
business segment level and following a number of studies (e.g., McGahan & Porter,
1997: Roquebert, Phillips & Westfall, 1996: Hough, 2006), annual segment level data on
proﬁtability was collected from COMPUSTAT business segment database. This
database reports up to 10 business segments because the SEC requires the identiﬁcation
of any business segment comprising 10 % or more of consolidated ﬁrm sales. For each
business segment, COMPUSTAT provides a 4-digit SIC code (for identiﬁcation of
corporation as well as primary and secondary industries) as well as each segrnent’s
operating income, sales and identiﬁable assets. Annual business segment performance
(ROA, hereafter) was calculated as operating income23 represented by COMPUSTAT
item “OPS” divided by identifiable assets (e. g., McGahan & Porter, 1997, 2002:
Misangyi et al., 2006).

My initial sample of yearly business segment performance covers the years 1998
through 2006. The start date coincides with Financial Accounting Standards (FAS) 131
superseding PAS 14 by becoming effective for ﬁscal years after December 15, 1997. This
change in accounting standards represents a major change in how public ﬁrms account
for and report their activities by making their segment reporting more in line with how
ﬁrms actually operate their business segments. Referred to as a “management approach”

(FASB, 1997: #4) for determining reportable segments, this new standard speciﬁes that

23 Note that there are similar measures of operating income available in the COMPUSTAT data base.
Although highly correlated (0.87), I rejected operating income before depreciation (OIBD) in favor of OPS
because the latter variable offers substantially more useable observations.

71

ﬁrms further split up broader “industry” or “line of business” reporting (required by the
old standard PAS 14) into more narrow operating segments. That is, conditions for
aggregating business segments were tightened by only allowing aggregation if segments
are similar in terms of a number of characteristics including the type and nature of
products and services, type of production processes, type of customers, and, if applicable,
also the nature of the regulatory environment (FASB, 1997: #17). Thus, public ﬁrms
often end up reporting more segments under the new standard. Wal-Mart, for example,
now reports four segments (Wal-Mart Stores, Sam’ 3 Club, International, and Other)
versus only one segment previously.

Although the new segment reporting rules require ﬁrms to restate prior years’
information, the wording in FASB 131 paragraph 34 is vague: “information. . ..shall be
restated unless it is impracticable to do so”, thus giving considerable freedom to
individual fnms to determine what burden should be placed on them to furnish the
additional information. In addition, the following paragraph (35) provides ﬁrms with the
option to only restate segment information for the current period “in which the change
occurs” if “both the old basis and the new basis of segmentation” is provided for that
particular year. Thus, I carmot be sure that ﬁrms actually went back and restated segment
information based on the new reporting standard, which implies that the same unchanged
ﬁrm may have very different reported segment structures before and after the new
accounting standard took effect. Furthermore, ﬁrms that ceased to exist (mergers,
bankruptcy, dissolution) before the new rules took effect will not have restated their
information, resulting in another possible bias to data sets incorporating observations

from both before and after the reporting change in 1998. Therefore, and because the time

72

period starting in 1998 provides an ample number of observations, my data set only
includes observations from after the change of reporting standards.

The data are screened in line with criteria and steps used by McGahan and Porter
(1997) and followed by other studies (e.g., Misangyi et al., 2006). To exclude business
segments that lack direct competitors (such as for example government entities) or
economic signiﬁcance in their respective industries, and thereby avoid biases leading to
understated industry and overstated business segment effects (e. g., McGahan & Porter,
2002), I exclude all business segments not containing a primary SIC designation, all
business segments that lack competitors in their primary SIC classiﬁcations in a speciﬁc
year (i.e., monopolies), as well as all segments classiﬁed as “not elsewhere classiﬁed,”
“nonclassiﬁable establishments,” or “government excluding ﬁnance” (SIC codes in the
90003). I also exclude all ﬁnancial business segments designated as “depository
institutions” (SIC codes in the 6000s) because returns of ﬁnancial institutions cannot be
compared to those of other industries as explained in McGahan and Porter (1997). In
addition, since “single-year appearances and small segments often are anomalous because
they are created for the disposition of assets prior to exit” (McGahan & Porter, 1997: 51),
business segments with only one observation across the sample as well as segments with
sales and/or assets below $10 million are excluded.

Observations at the business segment level were then matched with ﬁrm level
ﬁnancial data including Organizational size, Cash ﬂow, R&D intensity, Cost of capital
and Firm ROA from the COMPUSTAT ﬁrm level (Industrial Annual) database, as well
as with CEO compensation, CEO tenure, and stock ownership data from the

COMPUSTAT Executive Compensation database. In addition, data on block holder and

73

‘ .l
.D.'!l.' ‘5 ‘ .
:2“~J.1L1E51‘

-‘D

rm-.. «.5-
\b

P.
f

_!;_L.-301
83“ .'bo.\lt

M J‘Jd
l.»- 15'5-

l
m .150 L-

" a..-

 

institutional ownership were calculated based on underlying data from Thomson
Financial and their CDA/Spectrum database S34. Furthermore, Level of diversiﬁcation
and Diversiﬁed ﬁrm values were calculated using COMPUSTAT business segment and
ﬁrm level databases. Finally, at the industry level, measures of environmental uncertainty
were also collected and calculated using COMPUSTAT business segment and ﬁrm level
databases.

This resulted in 65,713 observations at the business segment level, 25,037 ﬁnn-
years of observations, and a total of 5,802 ﬁrms represented in my sample (see Table 1
for an overview of the sample including a breakdown of selected variables for illustration
purposes). Diversiﬁed as well as single-business ﬁrms, deﬁned as ﬁrms that have
business segment(s) in only one unique 4-digit SIC code, are represented in my sample: a
total of 34,114 segments belong to single—business ﬁrms, while the remaining 31,599
segments belong to diversiﬁed ﬁrms with segments in more than one unique 4-digit SIC
code. Of the segments belonging to diversiﬁed ﬁrms, the vast majority of segments
(28,790) belong to ﬁrms with two unique 4-digit SIC codes, while only 14,735 segments
belong to ﬁrms with segments in at least three unique 4-digit SIC codes. Thus, a large
percentage of my business segments (77.6 %) belong to ﬁrms that are either not
diversiﬁed or have segments belonging to only two unique 4-digit SIC codes.

To incorporate divergent arguments that the inclusion of single-business ﬁrms
“masks the corporate effect” (Bowman & Helfat, 2001: 14), but that their exclusion
would artiﬁcially inﬂate corporate effects “because diversiﬁed ﬁrms tend to perform

close to the overall mean” (McGahan & Porter, 2005: 879), segments belonging to both

74

types of ﬁrms are included in the analysis where appropriate according to both applicable

theory and methods used.

Table 1. Observations by Segment and Firm levels

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Segment level Finn ear Firm level
% % %
Variable name N missing N mlsslnl N missingi

Full sample 65,713 25,037 5,802
Organization size 65,621 0.14% 24,974 0.25% 5,761 0.71%
R 8: D Intensity 36,634 44.17% 14,1 17 43.47% 3,150 45.32%
Cost of capital 56,656 13.78% 19,431 22.39% 4,611 20.53%
Component: total interest

expense: L-t and st debt 61,617 6.23% 22,608 9.70% 5,219 10.05%
Component: L-t debt 65,514 0.30% 24,986 0.20% 5,782 0.34%
Component: Debt out of

s-t liabilities 65,633 0.12% 25,005 0.13% 5,785 0.29%
Block holder ownership 15,460 76.47% 4.71 1 81.18% 942 83.76%
Institutional shareholder

Ownership 20,662 68.56% 6,449 74.24% 1 .395 75.96%
Contingent compensation 22,185 66.24% 4,881 80.50% 891 84.64%
Component: Restricted stock

grants 22,347 65.99% 4,917 80.36% 896 84.56%
Component: Options granted 22,185 66.24% 4,881 80.50% 891 84.64%
Level of Diversiﬁcation 65,713
Environmental dynamism 65,580
Environmental complexny 65,713

 

 

 

 

 

 

 

 

Finally, the resulting data set extracted from and calculated with data from

COMPUSTAT and Thomson Financial, although large, should be viewed as a sample of

ﬁrms and industries having the following characteristics. First, it covers only public ﬁrms

traded on domestic exchanges, thereby excluding many foreign ﬁrms and all private

ﬁrms. In addition, the screening procedures discussed above exclude a large number of

the original observations. According to McGahan and Porter (2002: 840), about 15 % of

observations may be excluded because their “four-digit categories do not correspond to

75

economically meaningful industries (i.e., ‘not elsewhere classiﬁed’)” alone. Thus,

statistical analysis can be used to make inferences to the population based on the sample.

MEASURES

Dependent Variables
Capital Allocation Efﬁciency: Over and underinvestment.

The measures of capital allocation efﬁciency are based on business segment
prospects which, in turn, are estimated using Tobin’s q. Following prior research (e. g.,
Berger & Ofek, 1995: Lang & Stulz, 1994: Maksirnovic & Phillips, 2002: Rajan et al.,
2000: David et al., 2006), business segment Tobin’s q offers, in contrast to stock returns
or accounting performance measures, the advantage that “no risk adjustment or
normalization is required to compare q across ﬁrms” (Lang & Stulz, 1994: 1249). In
other words, since Tobin’s q is the market value divided by the replacement cost of the
ﬁrm, q implicitly incorporates the correct risk-adjusted discount rate as well as minimizes
other “distortions due to tax laws and accounting conventions” (Montgomery &
Wernerfelt, 1988: 627). Thus, no additional adjustments are necessary to compare
business segments with different individual risk levels. In addition, since a business
segment q cannot be observed directly, it is estimated by looking at the q of the particular
industry that the business segment belongs to. Thus, in line with LeBaron and Speidell’s
(1987) approach and the empirical study by Rajan et al. (2000), business segment q’s are
approximated by the mean q of single-business fnms in the same industry at the lowest
possible SIC level. Speciﬁcally, a step wise technique following prior research (e. g.,
Berger & Ofek, 1995: Villalonga, 1999: Carnpa & Kedia, 2002) is used to estimate the

mean Tobin’s q of at least ﬁve single-business ﬁrms by taking the ratio of ﬁrm value

76

.-

'5‘.

«uh

.25

o-

(deﬁned as the value of equity plus the book value of total assets minus the book value of
equity) to total assets for the speciﬁed number of single-business ﬁrms at the lowest
possible SIC code. That is, I check if an estimate is available at the 4-digit level. If one is
available, I use that estimate. If not, I check if an estimate is available at the 3-digit level
and instead use that estimate. If neither 4-digit nor 3-digit estimates are available, I check
to see if an estimate is available at the 2-digit SIC level. Thus, an estimate of segment
Tobin’s q is calculated for the focal segment as long as at least 5 single-business ﬁrms are
available at any SIC level.

Two additional methods of estimating business segment prospects were
considered but ultimately rejected due to their inability to represent the forward-looking
element of segment prospects. That is, neither ﬁrm ROA nor environmental muniﬁcence
(the latter measured as the average industry sales over the last ﬁve years) contains a
prediction (or estimate) of future performance, thus neither satisﬁes the basic requirement
for a measure of future performance and success.

Using this method (Tobin’s q) to calculate business segment prospects,
overinvestment is the allocation of relatively more investment funds measured as scaled
(based on segment assets and sales) capital expenditures to business segments with less
favorable prospects, while underinvestment is the allocation of relative fewer investment
funds to higher prospect segments. Thus, tests determine whether segments with
favorable prospects underinvest and whether segments with less favorable prospects
overinvest.

Speciﬁcally, with the exception of the Berger and Ofek (1995) overinvestment

measure described separately, four steps are required to measure over and

77

.3
a.

l l

Vim

.~‘l

“as

underinvestment at the business segment level. As a ﬁrst step, a segment’s growth
prospect is calculated using the described Tobin’s q methodology, the stepwise
technique, and data from that segrnent’s industry. Second, the prospect of a particular
segment is compared to two reference points, one external to the ﬁrm and one internal to
the ﬁrm. Following prior research on capital allocation, the external reference point is the
median prospect of all business segments (e. g., Shin & Stulz, 1998: Berger & Ofek,
1995), while the internal reference point is ﬁrm level Tobin’s q (e.g., Rajan, Servaes, &
Zingales, 2000). For external calculations of reference points, the focal segment is
excluded. For internal calculations, however, the focal segment is not excluded because I
assume that managers use ﬁrm level Tobin’s q as their reference point and are unlikely to
go through the additional work of partialling out the focal segment from their
calculations. This approach to estimating internal reference points is also consistent with
prior research (e. g., Rajan et al., 2000). Therefore, the internal reference point is set as
ﬁrm level Tobin’s q without adjusting for the focal segment.

The described comparison to external and internal reference points determines if
the business segment is a high or low prospect business segment which, in turn,
determines whether the test is for underinvestment or overinvestment, respectively.

In the third step, the information gained in the previous steps is used to check for
over and underinvestment by comparing segment capital expenditures scaled by both
segment assets and segment sales to both internal or external reference points of capital
allocation. The internal reference point is the level of scaled allocations to business

segments within the ﬁrm, while the external reference point is measured as the average

78

investment of the segments in either the bottom half or top half (as divided by the
median“) of the rank-ordered distribution of all segments.

The ﬁnal and fourth step for calculating capital allocation efﬁciency creates both
continuous and categorical measures from the comparison between actual allocations and
internal and external reference levels of allocation. Thus, based on the logic underlying
the four steps, overinvestment is measured as scaled capital expenditures allocated to low
prospect business segments that are each allocated more investment capital than internal
or external reference allocations. This generates two continuous measures based on the
internal reference point, one for each scalar (OVER_int_asset and OVER_int_sales), as
well as two measures based on the external reference point (OVER_ext_asset and
OVER_ext_sales), again distinguished by the scalar used to control for size effects.
Similarly, 4 categorical variables are constructed based on similar logic
(OVERC_ext_asset, etc.). See Figure 3 for overinvestment measures.

In addition, I include Berger and Ofek’s (1995) measure of overinvestment in
order to compare the results of my study to their examination of overinvestment. Since
they did not examine underinvestment, no comparison of those results can be made. They
measured overinvestment as the depreciation-adjusted scaled capital expenditures
allocated to a business segment with very low prospect (Tobin’s q below 0.25) based on
the rank-ordering of all segment prospects. This generates a ﬁnal continuous measure of
overinvestment (OVER_Berger&Ofek).

Four things should be noted at this point. First, the external reference point for

capital allocation uses a more restrictive investment hurdle to test for over and

 

2‘ Dividing business segments by median prospects instead of by mean prospects may also help to protect
against non-normal data and the undue inﬂuence of outliers. In addition, I also looked at distributions and
identiﬁed outliers as much as possible.

79

' bin-unann-

mo—mw “K0 UMW>O

3.3 :8 mm>0

830.

EEme

£33 :8 umm>o .323 Em Dam—>0 £33 35 Dam—>0

30me “KO Mum—>0 mug—am HE Mum—>0 maommm HE mm>o

38$. 83m macaw/w

\

RES:—

0 pence

sco§m0>58>o

3.5982 Eo—EmgamugO we 55.26::—
n «Suﬁ

9:53:— 33.53an

Qhﬂmavs m—uQn—ﬂmuﬂanv

enemas:— uo ass—sum

.8332? ..o .32. 3:95qu

Bax—moi mesa—=23 ..e .5502

5:33:05 E5333 :5an

80

underinvestment. That is, the actual allocation to a business segment is not compared to
the average investment of the entire distribution, but instead to the average investment in
either the top (testing for overinvestment) or bottom half (testing for underinvestment) of
the distribution. Thus, this procedure effectively cuts my initial sample into two separate
non-overlapping sub-samples, one sub-sample used to test for overinvestment and the
other used to test for underinvestment, as it provides for a more stringent test. The reason
behind that stringency is that segment prospects vary around their respective industry
means, thus cannot be known. That is, segments could have either better or worse
prospects than their respective industries implying the existence of random variation.
Therefore, to minimize the impact of that random variation and the possibility that some
segments may be incorrectly classiﬁed as either over or underinvesting, the stricter
investment hurdle is used.

Second, although more restrictive, the construction of measures using the external
reference point allows for the inclusion of single-business ﬁrms. Since the inclusion of
single-business ﬁrms increases the likelihood that a reference point can be calculated, it
also increases the chance that misallocation can be recognized when external measures
are used. This is clearly reﬂected in the number of instances of over and underinvestment
reported between models differentiated only by the way reference points are calculated.

For example, over and underinvestment scaled by assets account for a total of
51,197 observations when external reference points are used, but only 21,777
observations when measures are constructed with internal reference points. Thus, I lose
almost 30,000 observations by excluding single-business ﬁrms where an internal

reference point cannot be meaningfully calculated. As also discussed later in this section,

81

it session c
v M .‘ _
12:31.11: 5
' q .t, . ,5
:1 13. a
1216-51165
: . u ,g .
:1 new} i
T'- up:
all M.
I 9 won‘t; ,L
IL} “A; L“
1121126211
5
SW“. ' 0
Q-aﬁlOﬂ DC
"15‘" 3 panic:
restarts poi
.,,, J _
it 461.. bur:

FOLK:

the inclusion of single-business segments in external measures provides a way to
differentiate between effects of capital allocation efﬁciency for diversiﬁed versus for
ﬁrms that are not diversiﬁed. Note that the external measure had to be adapted to exclude
single-business fnms in models (e.g., diversiﬁed ﬁrm value) where the inclusion of such
ﬁrms clearly would be contradictory to underlying theory.

Third, the various individual measures of over and underinvestment (described
next) make this measure fairly comprehensive in that it targets several different potential
contingencies that managers may consider when allocating capital. For example, the
distinction between an external and internal reference point suggests that managers may
view a particular segment’s prospect or allocation against either an internal or external
reference point. Since we cannot be sure which reference point, if any, is most likely to
be used, both reference points are included.

Fourth, based on the comprehensiveness of the described measures of capital
allocation efﬁciency, I argue that existing measures to some extent “substitute” for a
proximal measure of optimal investment. That is, the absence of inefﬁciencies even after
the application of all individual measures could be taken to mean that the capital
allocation process for a particular ﬁrm indeed is efﬁcient. Furthermore, the underlying
logic behind such a measure also conﬂicts with the assumption of bounded rationality at
the individual decision-making level, i.e., decision-makers could never achieve the level
of optimal investment that the proxy is intended to measure. Therefore, measuring
inefficiencies in a process that is assumed to be biased and contain error may seem more

reasonable than to measure optimality where none can be had.

82

Underinvestment at the business segment level is based on the same logic with the
exception that the asymmetry of over and underinvestment does not allow for the
depreciation-adjusted measure. That is, the measure that Berger and Ofek (1995) used
was not meant to be applied to underinvestment since the logic behind it - that capital
expenditures in excess of depreciation to very low prospect business segments represents
overinvestment — cannot easily be applied to underinvestment. Thus, underinvestment is
measured as scaled capital expenditures allocated to high prospect business segments that
are each allocated less investment capital than reference allocations. Note that the
average of the lower half of the distribution of segment capital investment is now used as
the external allocation reference, while the internal reference stays the same. This
generates two continuous measures with internal reference points scaled by assets and by
revenue (UNDER_int_asset and UNDER_int_sales) and two measures with external
reference points, one scaled by assets (UNDER_ext_asset) and one scaled by revenue
(UNDER_ext_sales). In addition, four categorical measures are created (Figure 4).

Diversiﬁed ﬁrm value (DIVERSIFIED VALUE). Diversiﬁed ﬁrm value measures
the degree of discount (or premium) of diversiﬁed ﬁrms by comparing current ﬁrm
market value with an estimated value based on a scenario where the ﬁrm’s business
segments are freestanding entities. To estimate this potential “stand-alone” value and
enable a comparison with current ﬁrm value, researchers (e.g., Berger & Ofek, 1995:
Lang & Stulz, 1994: Comment & Jarrell, 1995) use a multiplier method where single-
segment fums in the same industry are employed to calculate a potential imputed value
for each business segment. Speciﬁcally, the median ratio of total capital (market value of

common equity plus book value of debt) divided by two accounting items (assets and

83

.11.....t..>..m....-u:.u 3.v-.vm.vPP‘.v-m nun-m-u~.vas-: -m.~.~nnI-le

thine-7.39..)— --.dn-ul.v>-.-.hmd!9-:d haw unavnuaeuhslnshhh
.1 attenuatio-

333 :8 Dam—Q23

wQ—ﬁw “K0 “mm—Z:

3?m

.8583.

383 :8 UyEQZD 3:3 2: Damn—ZS £83 .5 Damn—ZS

$03 :8 MmQZD 3:3 3: ”5075 383 :3 MmQZD

m~0mm< wO—dm w~0mm<

\

EEOaE

o pesos

udOEuWQzEHOUGD

mob-582 aﬂoaumctrﬂmho—uﬂD ac 3633.53:—
4 as»...

9.33... 3m8u3e0

05:30:- mﬂéﬂﬂmaﬂéo

233... .8 83m

.8333? .8 88.— 3.8.8.8:

82—8.... 9.32.8.3 .8 38.392

3833?... .8333? .5380

sales) for at least ﬁve single-segment ﬁrms in the same industry is estimated by using the
previously described stepwise technique. That is, I check if estimates are available
starting at lower SIC levels to construct a proxy at the lowest possible SIC level. The
estimated median ratio of capital to accounting item is then multiplied by the business
segment’s level of the same accounting item as illustrated in equation 1 below. Thus,
single-segment ﬁrms in the same industry as the focal business segment are used to
provide an estimate of a ﬁrm’s potential value if all of its segment were operated as
stand-alone businesses. The current ﬁrm value is then compared to the sum of the
imputed values for all ﬁrm business segments and excess value is calculated as the
natural log of ﬁrm value to imputed value (e. g., Berger & Ofek, 1995). A positive excess
value indicates that diversiﬁcation enhances ﬁrm value while a negative excess value
indicates the opposite (Equation 2). Because this measure is constructed to identify
excess values for diversiﬁed fnms, excess values are only calculated for ﬁrms with

segments in at least two unique 4-digit SIC codes, i.e., single-business ﬁrms are

excluded.
Imputed Valuek = Z (VM/AIM) *AI,‘ (1 )
Excess valuek = In (VI/Imputed Valuek) (2)

AI,- Value of accounting item (sales and assets) for segment i
AIM Value of accounting item for median single-segment ﬁrm

VM Total capital for median single-segment ﬁrm (market value of common
equity 4» book value of debt)

Vk Total capital of ﬁrm It

Thus, based on the accounting item used, this results in two measures of

diversiﬁed firm value: DIVERSIFIED VALUE_asset and DIVERSIFIED VALUE_sales.

85

Because of the endogeneity of the diversiﬁcation decision, I acknowledge the
possibility that observed diversiﬁed ﬁrm values (e.g., Lang & Stulz, 1994: Berger &
Ofek, 1995: Servaes, 1996) may not necessarily mean that diversiﬁcation destroys value.
In other words and as argued by Campa and Kedia (2002: 1731), ﬁrms choose to
diversify in response to ﬁrm characteristics that may “make the beneﬁts of diversiﬁcation
greater than the costs of diversiﬁcation”. Poorly performing ﬁrms, or ﬁrms that believe
they will beneﬁt from an internal allocation process, may therefore be more likely to
diversify or diversify to a larger extent because their opportunity costs for doing so tend
to be much less. Villalonga (1999) and Campa and Kedia (2002) support this reasoning
by ﬁnding that diversiﬁed ﬁrms tended to trade at a discount also prior to diversifying.

I argue, however, that ﬁrms diversifying in response to particular ﬁrm
characteristics may not affect the direction and validity of my results for a number of
reasons. First, I am not interested in the absolute level of the diversiﬁed ﬁrm value per se.
Instead, I am interested in the relative change of this variable in response to a more or
less efﬁcient capital allocation process. Thus, I am actually incorporating the efﬁciency
of the capital allocation process as an antecedent to diversiﬁed ﬁrm value, thereby
acknowledging that a ﬁrm’s belief that it can implement a highly efﬁcient capital
allocation process may contribute to that fnm’s decision to diversify. This is in line with
suggestions by Campa and Kedia (2002: 1733), who argue that important “ﬁrm
characteristics that affect the diversiﬁcation decision” should be included or controlled
for in the analysis. Second, if poorly performing ﬁrms and/or ﬁrms who believe they will
have an advantage over other ﬁrms in implementing an internal capital allocation

function indeed diversify to a larger extent, this would, if anything, weaken the

86

7‘ ,4" ‘
1‘sb‘lLl‘.‘

.w- 'H,‘

:(,,.t.\

1
":1'3'

'1, '5”

d l- M

a.»
,3"
He‘s.

51:6

"“3.“
Unanﬁ.\

C3303

J’ﬂaw
bs-\\L

‘ ‘nt‘

’1'.
(fr;

ZR-‘u

relationship between inefﬁciencies and diversiﬁed ﬁrm value. That is, fewer instances of
inefﬁciencies will be associated with smaller diversiﬁed ﬁrm values, which should
weaken the hypothesized relationship between the constructs.

Comrate effect (EFFECT). Conceptually, the corporate effect is deﬁned as
“differences between multiple-business ﬁrms in the average of returns to individual
businesses within each corporation” (Bowman & Helfat, 2001: 3). Thus, the larger those
differences are, the larger the corporate effect is, implying an increasing potency of the
corporate ofﬁce to affect the overall returns of their business segments. The corporate
effect is, in turn, operationalized by decomposing business segment ROA, calculated as
the ratio of operating income to segment assets (e.g., Misangyi et al., 2006: McGahan &
Porter, 1997: 2002), into effects stemming from industry, corporation, and the business
segments themselves following the majority of ﬁrm performance decomposition research
(see Bowman & Helfat, 2001, for an overview).

As discussed earlier, a number of single-level methods (e. g., analysis of variance
and variance components analysis) have been used to decompose business segment ROA
with the main drawback that they assume independence between effects, which not only
ﬁts poorly with the nested structure of the underlying data, but also makes it impossible
to advance theory by investigating relationships between effects as well as antecedents to
such effects. Therefore, I have elected to use a more suitable multilevel modeling
technique where effects are represented at different levels starting with time (Level 1),
followed by business-segment (Level 2), and corporation (Level 3). Due to the cross-
nesting of industry effects, they are introduced at the business-segment level. Thus,

corresponding to the deﬁnition given earlier where corporate effects are deﬁned as

87

differences between ﬁrms in their mean business—segment returns, the corporate effect
can be calculated by adjusting the variance of the residual at the corporate level for the
cross-nesting of industry effects. This process of classical hypothesis testing is described
in detail in the section on model speciﬁcation and estimation (page 102).

Firm ROA (F IRMROA). Firm ROA is a more traditional measure of ﬁrm
performance in strategic management. It is deﬁned as operating income divided by total

assets and is used as an additional and complementary measure of ﬁrm performance.

Independent variables

Business segment historical and social performance comparisons (HISTPERF,

S OCPERF).

Based on March and Simon’s ( 1958) argument that both past performance and the
performance of competitors will inﬂuence aspiration levels, historical segment
performance (measured as the prior year’s segment ROA) and the performance of
competitors (measured as the current average ROA performance of at least 5 segment
competitors) are used to assess aspiration driven inﬂuences on the capital allocation
process. The stepwise methodology previously discussed to search for reference levels is
applied here as well. That is, I ﬁrst search for ﬁve competitors at the 4-digit SIC level. If
they cannot be found at that level, I renew my search at the next higher level, and ﬁnally
at the 2-digit level. This way of gathering reference levels of social performance assumes
that ﬁrms are looking for segment competitors in single-business as well as diversiﬁed
ﬁrms, i.e., across organizational structure. Evidence from the business press supports this

assumption - large ﬁrms, such as for example GE, benchmark their divisions against the

88

best competitors in their respective market segments or industries regardless of parent
company structure. Alternatively, well-performing segments of single-business ﬁrms may
use segments in large well-performing conglomerates (such as GE) as benchmarks. In
GE’s case, this may be particularly true since they often are at the forefront of their
industries due to the much talked about constraint that divisions within GE need to be
either ﬁrst or second in their respective industries or face threats to their existence. Thus,
because I cannot ex ante exclude ﬁrms looking at segments belonging to ﬁrms with
different organizational structures than their own, I decided to include both types of
segments as likely competitors and not further differentiate between them.

A single measure replacing separate historical and social inﬂuences on aspirations
by combining them (Bromiley, 1991) was initially incorporated, but later rejected for the
ﬁnal analysis because of inconsistencies when used in conjunction with measures of
capital allocation efﬁciency. This measure is based on logic from the Behavioral Theory
of the Firm (Cyert & March, 1963) where ﬁrms with business segments performing
above average segment performance are unlikely to aspire to average performance,
instead aspiring to somewhat increased future performance above previous performance
levels (e.g., Eliasson, 1976), while ﬁrms performing below average may aspire to raise
their performance to just that average level. Thus, ﬁrms are assumed to have different
aspiration levels depending on whether current business segment performance is above or
below current average ROA of its segment competitors. It is these dual aspiration levels
that create problems when used in conjunction with efﬁciency measures. When used with
overinvestment measures, the likelihood that current business segment performance is

below either social or historical aspiration level is high because only segments with

89

below median (below average) prospects are assumed to overinvest. When used with
underinvestment, however, the likelihood that current business segment performance is
above the combined aspiration level is low for two reasons. First, current business
segment performance needs to be above aspirations, thereby taking the social aspiration
level out of play. In addition, the historical aspiration level is now more restrictive since
it is set at 5 per cent higher than actual performance last year. Thus, the asymmetry
between over and underinvestment due to the construction of capital allocation

efﬁciency measures makes the Bromiley measure unsuited for use in this dissertation.

Corporate Governance

Block holder ownership (BLOCKOWNER). Ownership by block holders is a
common measure of the ability of shareholders to monitor and control executives (Davis,
1991). I measured this proxy for monitoring and control as the total proportion of
ownership held by institutional external shareholders controlling more than a standard 5
% of total outstanding equity.

Institutional shareholder ownership (INS TIT OWNER). As an alternative measure
to block holder ownership (Sanders & Carpenter, 2003), the percentage of total
outstanding equity owned by institutional shareholders such as pension funds, mutual
funds, and hedge funds may also correspond to shareholder involvement in monitoring.
Due to potential overlap in ownership between the two measures (i.e., institutional
ownership includes positions by block owners), I test shareholder monitoring with each
variable separately.

Contingent commnsation (COMPENSATION). I focus on CEO compensation for

two reasons. First, this ensures that my results not only are comparable to ﬁndings of

90

previous corporate governance research (e. g., Devers et al., 2008), but also comparable to
studies modeling the capital allocation process including its efﬁciency (e. g., Scharfstein
& Stein, 2000). Second, prior research has suggested strong relationships between both
structure and level of CEO and TMT pay due to inﬂuential CEOs being able to affect the
compensation of their top management teams (e.g., Carpenter & Sanders, 2002: Gomez-
Mejia & Wiseman, 1997). That is, CEOs may act as “gate keepers with respect to TMT
member pay” (Carpenter & Sanders, 2002: 378) which would also be consistent with
literatures concerned with CEO power (Finkelstein & Hambrick, 1996: Hayward &
Hambrick, 1997). Thus, CEO compensation may be a reasonable proxy for incentive
alignment effects not only for the CEO, but also for the remainder of the top management
team. Therefore, CEO compensation contingent on the performance of the ﬁrm’s share
price was measured by adding the following COMPUSTAT Executive Compensation
categories representing contingent compensation: restricted stock representing value of
stock-related awards (e. g., restricted stock and restricted stock units) that do not have
option like features, and options granted (3 C ompustat Black Scholes value) representing
the aggregate value of stock options granted to the CEO during the year as valued using S
& P’s Black Scholes methodology. The aggregate value of granted options and restricted
stock was used to represent contingent compensation because traditional theory (e. g.,
Jensen & Meckling, 1976) is not clear on differences between them, if any, in regards to
their effects on corporate decision-making and risk taking. In addition, the use and size
of restricted stock grants has traditionally been quite modest until FASB amended their

rules in 2004 (to take effect in 2005) requiring companies to treat all forms of share-

91

based payments to employees in a similar fashion by recognizing their expense on the

income statement.

Environment

Building on work by Aldrich (1979), Duncan (1972) and Child (1972), Dess and
Beard’s (1984) inﬂuential multidimensional characterization of organizational
environments is used to model environmental inﬂuences on the capital allocation process.
Their model has been supported by a conﬁrmatory factor analysis (Rasheed & Prescott,
1992) where 94 % of environmental variance was explained by environmental dynamism, .
complexity and muniﬁcence. Additionally, a large number of empirical studies have used
all or some of the dimensions in studies of corporate strategy (e. g., Boyd, 1990, 1995: Li
& Simerly, 1998: Simerly & Li, 2000: Henderson, Miller & Hambrick, 2006). As noted
earlier, environmental dynamism and complexity are modeled as independent variables.

Environmental dmamism (DYNAMISM). Following the popular characterization
of dynamism as the variance of muniﬁcence (e. g., Boyd, 1990, 1995: Goll & Rasheed,
1997: Li & Simerly, 1998: Palmer & Wiseman, 1999: Keats & Hitt, 1988: Misangyi et
al., 2006), environmental dynamism is operationalized as the variance of industry sales
growth over a 5-year period. Speciﬁcally, following common practice for the calculation
of muniﬁcence and dynamism (e.g., Boyd, 1990, 1995: Li & Simerly, 1998: Simerly &
Li, 2000: Palmer & Wiseman, 1999: Keats & Hitt, 1988: Misangyi et al., 2006), I am
regressing time against industry sales based on the ﬁve preceding years. The standard
error of the regression slope divided by mean sales over the period is my estimate of

environmental dynamism.

92

Writ
1:708:16:
15:33:31
1.2.151:
11.11101
Ilium: it.
'31:: 3311::
3111301 do
trace.

£1
it ruched .
Emcee:
31013 COmp]
rm are p
ER predict:
tends? Ire

68.3371de

  
  

Ule. 6‘.

With regards to the measurement of environmental dynamism, I note that Folta
and O’Brien (2004) use a very complicated generalized autoregressive conditional
heteroscedasticity (GARCH) model (Bollerslev, 1986) that is seldom used in strategic
management but often in ﬁnance and economics. They argue that more conventional
calculations of industry-speciﬁc uncertainty involving “the variance of some output or
indicator (e. g., stock price, GDP, sales)” (Folta & O’Brien, 2004: 127) may produce
biased estimates for two reasons: conventional methods do not account for trends in the
data, nor do they allow for variance that is not constant over time, i.e., heteroscedastic
variance.

Although they are correct that more conventional methods (such as for example
the method of estimating industry dynamism described earlier) do not take trend and
heteroscedasticity into consideration, I provide three counterarguments suggesting that
more complex models may also provide biased estimates. First, their assumption that
trends are predictable, and that “they may not constitute an element of uncertainty if they
are predictable” (Folta & O’Brien, 2004: 127) begs the question of how predictable are
trends? Trends can, and often do, reverse at any time since their continuation is
dependent on the continuation of all necessary conditions promoting those trends.
Clearly, even very complex models are not able to control for random events that could
reverse such conditions and derail current trends. Thus, the assumption that trends are
predictable relies on the assumption that models somehow can control the randomness of
the environment - an enormous assumption as evidenced by the generally poor track

record of forecasting in predicting future trends.

93

arc-site rem
relic: 161 c
tsetse. For i
'3} axing p:
as acting CO]
12:11 prof
it} at: me:
taped Pros;
17:51: ﬂail ‘
lit also not
:11 Knoll
Irony of
This are n

“31 prrdx‘

‘55 likelt

 

 

 

In addition, because of strong behavioral instincts and tendencies, people often
perceive trends as over long before they actually are, even though their models may
predict their continuation”. Thus, people often take active steps to beneﬁt from their
reverse. For example, people reacting to gain contexts (e. g., Kahneman & T versky, 1979)
by “taking proﬁt”, i.e., protecting their gains, after a period of strong market performance
are acting contrary to the implied logic of Folta and O’Brien that trends are predictable.
Instead, proﬁt takers would seem to be acting as if trends are unpredictable, thus as if
they are uncertain about how long the trend will continue. In addition, following S-
shaped Prospect theory predictions (Kahneman & T versky, 1979), the behaviors people
exhibit that “double down” and otherwise accept more risk to reverse failing performance
are also not in agreement with the authors’ logic that trends equal certainty. After all, if
trends implied certainty about the future and were certain to continue, why do the
majority of people in loss contexts try so hard to reverse them? In sum, therefore, if
trends are not predictable and/or if people, because of cognitive limitations, do not follow
such predictions, models that nevertheless try to predict such trends would be more, not
less, likely to bias estimates of environmental dynamism

Second, since the purpose of measures of environmental dynamism is to represent
beliefs about the future, it may be more likely that those beliefs are based on simple as
opposed to complex heuristics given our cognitive limitations. That is, the simpler the
model, the more accurately it may reﬂect the heuristics we use when making decisions.

Thus, complex models such as the one suggested by Folta and O’Brien, may not

 

25 A special case of not just prematurely perceiving trends as coming to an end, but also actively ignoring
prior trends is described by Kahneman and Lovallo (1993). This leads to ‘bold forecasts’ as each new
project is viewed as unique and not evaluated against similar projects implemented in the past. Thus, even
if the ﬁrm has been unsuccessful with similar projects in the past, they may still view the current project as
likely to succeed.

94

correspond very well with heuristics actually used to make decisions. Thus, more
complex models, because they do not reﬂect actual decision making, may actually be
more, not less likely to bias estimates.

Third, although GARCH models that account for heteroscedasticity may be
necessary to capture long economic time series (Campa, 1993), the same argument
cannot be applied to this study since my data structure is not that of an extended
economic time series. It covers the years of 1998 through 2006 only, and contains just a
few estimates of environmental dynamism since a new estimate is calculated for each 5-
year period. In sum, therefore, I have argued that measurement issues involving trend and
heteroscedasticity, if at all applicable to my measures and sample, are likely to have
rather benign effects on the calculation of environmental dynamism. Therefore, I am
electing to follow existing research practice within management and measure dynamism
as the variance of sales over preceding 5-year periods.

Environmental complexity (COMPLEX). Following Aldrich’s characterization of
the organizational environment as exuding different degrees of complexity based on the
homogeneity and concentration of ﬁrms in that particular environment or industry, I
operationalize environmental complexity using the Herfmdahl index (Herﬁndahl, 1950:
Kelly, 1951). That is, in line with suggestions by Porter (1980) and Scherer (1980), the
Herfmdahl index takes into account both the number of fums in an industry, as well as
inequalities in market shares among those ﬁrms by being measured as the sum of the
squared market shares for all ﬁrms in a particular industry. Thus, the Herﬁndahl index
adds a dimension to other measures of complexity, such as the four ﬁrm concentration

ratio (Hay & Morris, 1979) and simple count measures of industry competitors, by also

95

considering inequalities in market shares among industry competitors in addition to
industry concentration.

Note that since complexity is the least studied of Dess and Beard’s (1984)
dimensions of environmental uncertainty, no consensus or “standard” measure of
complexity has yet emerged from the literature. At the same time, because the Herfmdahl
index has been used at least as many times as other measures (e.g., Boyd, 1990, 1995),
and because it is more inclusive of, or accommodating to, underlying theoretical
arguments (e.g., Aldrich, 1979: Porter, 1980: Scherer, 1980), I have elected to use it in
this dissertation. Therefore, following Boyd (1990, 1995), the Herﬁndahl index (H) is

calculated as:

n 2
Hi = I- 2 (market sharej) (3)
1:1

where j = I, 2,...n, number of ﬁrms in the industry and market share,- is the market share
for that particular ﬁrm.

Herﬁndahl index scores are conﬁned to a range of between 0 and l where a value
close to 1 implies many competitors and no ﬁrms with dominant market shares (i.e., no
monopolistic or oligopolistic industry structure). In effect, therefore, industries with a
Herfmdahl score approaching 1 are considered more complex, while industries with
scores approaching 0 are considered relatively less complex.

Level of diversiﬁcation (DIVERSIFICA TION, DIVERSIFICA TIONdummy). To
measure differences in diversiﬁcation strategies, the entropy measure of diversiﬁcation
(J acquemin & Berry, 1979) is used. This is an objective measure of diversiﬁcation that

has shown strong construct validity (Hoskisson, Hitt, Johnson & Moesel, 1993) and has

96

been used in a number of recent studies (Sanders & Carpenter, 2003: Bowen &
Wiersema, 2005: Boyd, Gove & Hitt, 2005: Birkinshaw, Braunerhjelm, Holm &
Terjesen, 2006).

In addition, Robins and Wiersema (2003) show that of the two most widely used
measures of diversiﬁcation in contemporary research (Barker & Duhairne, 1997: Bergh,
1997), the entropy measure and the concentric measure of diversiﬁcation, only the former
is affected in the appropriate direction by two simulated changes in ﬁrm portfolios. To
show how these measures differ, Robins and Wiersema perform several tests. First, the
authors add three (related) new business segments with distinct 3-digit SIC codes, but
within the same 2-digit SIC area of business to the existing ﬁrm portfolio. The result is
that the original segments now each account for a smaller piece of total corporate activity
because of the addition of the new segments. In response to this increase in related
diversiﬁcation, the entropy measure shows an increase in its related component (DR),
while the concentric measure actually shows a decrease in related diversiﬁcation.

Next, Robins and Wiersema reduce the business segments down to the original
four, but make one of the segments far more dominating than in the original
conﬁguration. That is, the number of business segments are reduced and their
concentration is increased which should decrease related diversiﬁcation. Here, the
authors ﬁnd that the entropy measure shows a decrease in related diversiﬁcation while
the concentric measure, again, shows the opposite. Based on those ﬁndings, they
conclude that related entropy (but not the concentric measure) is positively related to the
number of business segments in the corporate portfolio, while negatively related to the

size of the dominant business segment in the corporate portfolio. In sum, because related

97

. - f
:19) >65.

. ." w; ".3

3133129 3

4|"; . .
$14.36 3h,

[ca

333531 1

5331338 1

entropy seem to measure what it should, because it has shown strong construct validity,
and because it has been used extensively in strategic management, I elect to use the
entropy measure of diversiﬁcation in this dissertation.

The entropy measure has two components - related diversiﬁcation (DR) and
unrelated diversiﬁcation (DU) - that together add up to total diversiﬁcation (DT).
Following Hoskisson et al., (1993), related diversiﬁcation is deﬁned as four-digit SIC
segments within a two-digit SIC group, while unrelated diversiﬁcation is deﬁned as ﬁrms
operating across two-digit groups and total diversiﬁcation is calculated based on its

relative share of sales according to the following formula:
N
Total entropy = 2le“ ln (1/ Pi) (4)
1:

where Pi is deﬁned as the ﬁrm’s share of sales in segment i and In (UP) is the weight of
each segment. Thus, this measure is a combination of the number of segments a ﬁrm is
active in as well as their relative sales importance.

In addition, a categorical variable (DIVERSIFICA TIONdummy) that distinguishes
corporations Operating two or more business segments (dummy = 1) from those operating
only a single segment (dummy = 0) is used to investigate the impact that multi-segment
corporations may have on performance.

Business segpent return variance (SRV). To account for homogeneity in business
segment returns within ﬁrms, this measure of dispersion was calculated as the variance of

business segment returns (ROA) for ﬁrms with at least 2 segments.

98

"
ll
n-

0;"

.\ ,

3.1.
a
b
'J

31mg

Ex"?!
,
t.
loser?
11
Eric:
‘1.“ lh’alaw\n
' "‘ \‘AL ..
f Bums s:

 

 

m- 1 M\Et\w\.\

 

 

Table 2. Measures

 

Full Name

 

Short Name

 

Dependent Variables

 

External measure of overinvestment (asset)

OVER_ext_asset

 

External measure of overinvestment (sales)

OVER_ext_sales

 

Internal measure of overinvestment (asset)

OVER_int_asset

 

Internal measure of overinvestment (sales)

OVER_int_sales

 

Berger & Ofek measure of overinvestment

OVER_Berger & Ofek

 

External measure of undernvestrnent (asset)

UNDER_ext_asset

 

External measure of underinvestment (sales)

UNDER_ext_sales

 

Internal measure of underinvestment (asset)

UNDER_int_asset

 

Internal measure of underinvestment (sales)

UNDER_int_sales

 

Diversiﬁed ﬁrm value (asset)

DIVERSIFIED VALUE_asset

 

Diversiﬁed ﬁrm value (sales)

DIVERSIFIED VALUE_sales

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Corporate effect EFFECT
Firm ROA FIRMROA
Note: Cateﬂrical measures of above exist, but not shown here
Independent Variables
Business segment historical performance comparisons HISTPERF
Business segment social performance comparisons SOCPERF
Block holder ownership BLOCKOWNER
Institutional shareholder ownership INSTITOWNER
Contingent compensation COMPENSATION
Environmental dynamism DYNAMISM
Environmental complexity COMPLEX
Level of diversiﬁcation DIVERSIFICATION
Level of diversiﬁcation (categorical measure) DIVERSIFICATIONdummy
Business segment return variance SRV
Controls
Organization size SIZE
CEO tenure TENURE
Cash ﬂow CASHFLOW
R & D intensity R&DINT
Cost of capital COST
Long-term incentive plans LTIP

 

99

 

Controls

Organization size (SIZE). The log of ﬁrm revenue is used as a measure of
organization size and was included because it may affect ﬁrm resource allocations
(Haveman, 1993).

CEO tenure (TENURE). To control for the possibility that CEOs feel varying
levels of “ownership” of business segments depending on if they acquired a particular
segment or built it “from the ground up”, CEO tenure proxies for the tendency of CEOs
to more strongly defend and be more inclined to ﬁx, rather than divest, poorly performing
segments they feel ownership in.

In addition, CEO tenure may also control for a possible interaction between the
length of time a CEO has held ofﬁce and the trend in capital allocation, i.e., controlling
for consistent over or underinvestment over long time periods.

Cash ﬂow (CASHF LOW). Cash ﬂow may also affect resource allocations within
the ﬁrm (Jensen, 1986) by representing capital available for investment. Following Gibbs
(1993), I measure ﬁrm cash ﬂow as earnings before interest, taxes, depreciation, and
amortization divided by total assets. Using assets instead of market value reduces
unnecessary correlation with Tobin’s q as a measure of investment opportunity.

R&D intensity (R&DINT). Research intensity (R & D expenditures/sales) was
included as a control because higher research intensity has been shown to increase
information asymmetries in organizations (Aboody & Lev, 2000). If this association
holds true also in the context of this study, capital allocation decisions may become more
complicated as the level of research intensity increases thus potentially leading to

additional capital allocation inefﬁciencies. Thus, because lam restricting my study to

100

333')? I116

4.33., 9
LunbaVnL

3. ‘: 1"
“I331 A

i‘
”13‘” ‘13?-
Anubis.

13331216

I l
l)" . 4..
Bull, 1‘

-

‘ y ..‘HI‘ .
~‘_,.\H‘;
5

333 ‘tea:
exaCVrv- ,

' “Cal‘s

\. ‘
UI “Tags
5

“Odel S

W «21:

investigate effects of environmental uncertainty on capital allocation efﬁciency without
investigating additional antecedents to that uncertainty, I control for R&D intensity at the
corporate level. Naturally, that would allow me to include ﬁrms in my sample with
different levels of research intensity without the possibility of biasing my results.

Cost of debt capital (COST). The weighted average cost of debt is included to
control for the possibility that ﬁrms with a lower funding cost may pursue more
inefﬁcient capital allocation strategies because of their presumably larger pools of
investment capital to draw from. This control was calculated using COMPUSTAT data as
total interest rate expense of short and long term debt (data item 15) divided by long-term
debt (data item 9) and the part of short term liabilities constituting debt (data item 34).

Long-term Incentive Plans (LTIP). The payouts of long-term incentive plans are
contingent on accounting retums and measure performance over a period of more than
one year. Although this control variable represents the actual amount paid out to the
executive and does not reﬂect an award that may grow over time, the existence and size
of long-term incentive plans may still have an effect on managerial decisions including

the decision to allocate capital. Thus, I control for such cash payments made to the CEO.

Model Speciﬁcation And Estimation

Introduced to the strategic management literature by McNamara, Deephouse and
Luce (2003), multilevel modeling (Bruk & Raudenbush, 1992: Raudenbush, Bruk,
Cheong & Congdon, 2000) is appropriate for the analysis of my model for three reasons.
First, the design of my study involves variables at three different levels —— at the business
segment level (business segment historical and social performance comparisons, over-

and underinvestment, business segment return homogeneity), at the ﬁrm level

101

(governance variables, level of diversiﬁcation, corporate effects and diversiﬁed ﬁrm
value) as well as at the industry level (environmental uncertainty). As such, my study
naturally lends itself to a multilevel modeling technique since each lower level is nested
within each higher level as well as within the hierarchy itself. That is, business segments
are nested within ﬁrms and ﬁrms are nested within industries, which implies a lack of
independence between observations since both business segments from the same ﬁrm as
well as fnms from the same industry may be more similar than segments from different
ﬁrms and industries, respectively. This lack of independence, or alternatively, the likely
existence of higher level constructs affecting lower level constructs would have resulted
in biased standard errors of regression coefﬁcients (Snijders & Bosker, 1999) had a
single-level approach such as regular OLS regression been used.

Second, modeling the decomposition of ﬁrm performance and the corporate effect
also requires a multilevel approach. Not only does a multilevel modeling technique avoid
a number of documented problems with alternative methods such as nested ANOVA,
variance components analysis (VCA), and the continuous variable model used by
Brush, Bromiley and Hendrickx (1999)”, but it also takes into account a number of
dependencies between industries, corporations, and business segments. Speciﬁcally, ,
business segments are not independent of industries as industry environments and
conditions affect ﬁrm conduct (Bain, 1956: Porter, 1980). At a higher level, corporations

and industries may also enjoy reciprocal inﬂuences as suggested by McGahan and Porter

 

2" A main problem with nested AN OVA is that the ﬁxed dummy coding consumes a large amount of
degrees of freedom while VCA has been shown to be insensitive to small effect sizes (Brush & Bromiley,
1997). In addition, the more advanced continuous variable method also has drawbacks because it divides
ﬁrms into a limited number of groups based on the number of segments and then models each with separate
equations. This requires that ﬁrms with either more or less segments be discarded which restricts
generalization to a larger and more inclusive population of ﬁrms.

102

.9- '1"
, .1p 1 ul
A

Avg“ ‘1'
.& J‘Sﬁ

(2002: 838) “the covariance between industry and corporate-parent effects is potentially
important because, for example, a diversiﬁed ﬁrm may be more likely to expand into
particular types of industries”. Taken together, this means that business segment
performance is cross-nested or dependent on both the corporation as well as on the
industry, implying that industry, corporate, and business segment effects are not
independent of each other. Thus, to properly model the described dependencies, a
multilevel speciﬁcation (such as STATA Multilevel mixed-effects linear regression,
STATA XT Mixed hereafter) is necessary to adjust for dependencies so as to avoid
biasing regression coefﬁcients.

Finally, antecedents to the corporate effect cannot be investigated using either of
the more popular single-level methods, because neither VCA nor ANOVA estimates of
industry, corporate, and business segment effects enable inferences about the importance
of strategy (Misangyi et al., 2006). Instead, by using dummy variables and often
including shared variance due to order of entry effects, they tend to represent “upper
bounds of each class of effects” (Misangyi et al., 2006), offering little information about
speciﬁc industry, corporate, and business segment factors associated with superior ﬁrm
performance. Thus, translated to the context of this study, single-level techniques would
not have allowed me to investigate the efﬁciency of the capital allocation process as an
antecedent to the corporate effect.

In sum, STATA XT Mixed allows me to pursue an alternative approach to
modeling the inherently multilevel capital allocation process in multidivisional ﬁrms. As
such, it addresses and avoids speciﬁc problems with prior methodologies as well as

correctly accounting for the lack of independence of observations. In contrast with

103

methods used by previous scholars not allowing the investigation of antecedents, STATA
XT Mixed allows me to investigate both the efﬁciency of the capital allocation process as
an important antecedent to the size of the corporate effect, as well as to investigate
multilevel antecedents to the capital allocation process itself.

In addition, STATA XT Mixed allows me to treat my longitudinal or panel data
set as multilevel in structure by assigning time to the ﬁrst level. That is, the time-series
dimension of my data set is represented at the ﬁrst level, while stable business segment,
corporate, and industry inﬂuences get represented at higher levels. This enables the
separation of cross-sectional information from information varying over time and the use
of different predictors at each level.

For those reasons, I used STATA XT Mixed27 to analyze the effects of multilevel
antecedents at business segment, corporate, and industry levels on the capital allocation
process (Figure 1), as well as to analyze performance implications of degrees of capital
allocation efﬁciency on the corporate effect (Figure 2). Additional effects of capital
allocation efﬁciency on diversiﬁed ﬁrm value and ﬁrm ROA were modeled using a
single-level panel data technique (STATA XT reg). Temporal aspects (time lags) of the
different variables are speciﬁed and discussed for both the antecedent model as well as

for models analyzing performance implications of capital allocation efﬁciency.

 

27 Note that I-ILM (Raudenbush et al, 2000) is another popular statistical program for ﬁtting multilevel
models. I-ILM and STAT A XT Mixed are, however, very similar when it comes to what models can be
speciﬁed as well as underlying estimation theory and computational algorithm to ﬁt such models. That is,
they both offer the choice of modeling effects on the dependent variable separated into a number of levels,
the choice between full maximum likelihood as well as restricted maximum likelihood (estimation
theories), as well the popular “expected-maximization” (EM) algorithm (Dempster, Laird, & Rubin, 1977)
as the computational algorithm of choice.

104

Antecedents to Capital Allocation Efficiency

Theory suggests that transient effects (i.e., varying over time) should be
distinguished from stable effects (Rumelt, 1991: Misangyi et al., 2006) and most studies
have n'ied to capture “variance over time effects” (McGahan & Porter, 2002: McGahan &
Porter, 1997: Roquebert et al., 1996: Rumelt, 1991) in explaining business segment
performance. However, due to the lack of theory guiding the characterization of strategic
factors as either transient or stable, the characterization is based on empirical assessment
of type of variance. Therefore, the initial step before examining associations between
multilevel antecedents and capital allocation efﬁciency is to determine whether each
antecedent should be treated as a transient or as a stable strategic factor. The difference is
that a transient strategic factor explains variance over time and should be entered
individually for each year, while a stable factor only explains cross-sectional variance
between corporations or business segments and should therefore be entered as the
average of observations over time. To determine if each strategic factor should enter as a

transient or as a stable factor, I used intra-class correlation (ICCl) analysis to examine

how much variance of each strategic factor occurs across time as opposed to in a cross-
sectional manner. Strategic factors where most variance occurs across time should be
entered as transient factors as opposed to those where most variance is cross-sectional in
nature allowing aggregation and stable effects (James, 1982: Bliese, 2000). Although no
absolute standard value for aggregation based on ICC. has been established (Avolio, Zhu,
Koh & Bhatia, 2004), I used a conservative value of 0.60 to allow for aggregation. This
cut-off value is above conventional recommended values for aggregation and multilevel

analysis, thus fulﬁlls required statistical support for aggregation to a higher level (Chan,

105

1998: Bliese, 2000). In addition, following Misangyi et al., (2006), I performed an ICC;

analysis to determine the reliability of each aggregated measure. Based on those analyses
(Table 2), I aggregated INSTITOWNER (institutional shareholder ownership),
DIVERSIFICATION (level of diversiﬁcation) and SOCPERF (social performance
comparisons) because a majority of its variance (81%, 86%, 65%, respectively) occurred
between organizations, while HISTPERF (historical performance comparisons),
COMPENSATION (contingent compensation), DYNAMISM (environmental
dynamism), and COMPLEX (environmental complexity) all had the majority of their
variance within unit and, thus, were left as transient effects. High levels of reliability
(ICCz) support the use of the aggregated measures. For the SRV (segment return
variance) and BLOCKOWNER (block holder ownership) measures, the variance is
relatively equally shared between the aggregate unit (business segment and corporation)
and across time, thus their ICC. values are very close to my designated cut-off. However,
because a sizable part of their variance occurs across time, and because I was hesitant to
set that variance to zero thereby reducing their power to explain variance over time to
zero as well, I elected to follow my rule and treat them as transient effects. Among
control variables, SIZE (ﬁrm size), TENURE (CEO tenure), and R&DINT (R&D
intensity) required aggregation. In addition, since a multilevel speciﬁcation requires a
dependent variable at the lowest level, I examined measures of capital allocation
efﬁciency for levels of transient versus cross-sectional variance. I found that all measures
of capital allocation efﬁciency vary substantially over time, none requiring aggregation,

which supports the use of a multilevel speciﬁcation.

106

Table 3. Intra-class Correlations (ICC) of Relevant Variables

 

 

 

Variable ICC( 1) ICC(2) Enter as
1. External Overinvestment (assets) 0,14 transient
2. External Overinvestment (sales) 0.08 transient
3. Internal Overinvestment (assets) 0.06 transient
4. Internal Overinvestment (sales) 0.02 transient
5. External Underinvestment (assets) 0.22 transient
6. External Underinvestment (sales) 0.04 transient
7. Internal Underinvestment (assets) 0.01 transient
8. Internal Underinvestment (sales) 0.03 transient
9. Organization size 0.92 0.99 stable
10. CEO tenure 0.89 0.99 stable
11. Cash ﬂow 0.58 transient
12. R & D intensity 0.70 0.95 stable
13. Cost of capital 0.22 transient
14. Long-term incentive plans 0.31 transient
15. Social Performance Comparisons 0.65 0.98 stable
16. Historical/social Performance Comp. 0.30 transient
17. Block holder ownership 0.57 transient
18. Institutional shareholder ownership 0.81 0.98 stable
19. Contingent compensation 0.27 transient
20. Environmental dynamism 0.08 transient
21. Environmental complexity 0.28 transient
22. Level of diversiﬁcation 0.86 0.98 stable
23. Business Segment Return Variance 0.58 transient

 

measures of capital allocation efﬁciency on level 1 (Time level) strategic factor

After performing the initial ICC analysis, I used STATA XT Mixed to regress

antecedents including HISTPERF, COMPLEXPROXY, BLOCKOWNER,

COMPENSATION, DYNAMISM, and COMPLEX. Subsequently, level 2 (Business-
segment level) stable strategic factors (SOCPERF) were used to explain between-business
segment variance in level 1 parameters (intercepts), while level 3 (Corporate level) stable
strategic factors (INSTITOWNER, DIVERSIFICATION) in turn were used to explain

between-corporation variance in level 2 parameters. In addition, a number of interaction

107

 

variables were used to analyze interactions between environmental uncertainty
(Dynarnism and Complex) and some of the antecedent relationships. Below is the full
model including control variables where equation 7 models stable organizational effects
for each ﬁrm j. The predicted values from this equation are included as the intercept in
the next equation (equation 6) which models stable business segment effects on each
segment i. Finally, the predicted values from that model are included as the intercept in

equation 5 which models transient effects on allocation efﬁciency across time t.

 

(5) Allocation Ineﬂicienciesa'j = 7:0,] + at] ij(Cost)tij + nzl'J(LTIP)tij + n3iJ(Cashﬂow)tij +
”413(mStPel‘Dtij + n5g(Complexproxy)tij +
F6iJIBIOCk0W11€rhij +n7iJ(Compensation)ti j +
”SleynamiSmhij + ﬂongomplexhij +
7’1 Orleynarnismi‘Blockownermj +
7‘ 11 ij (Dynamism*lnstitowner)tij +
”IZiKDYDaInism’“C0mpensation)tij +
”1313‘ (Dynamism*Diversiﬁcation)tij +
n14iJ(Complex*Blockowner)tij +
7r} 5,3(Complex*lnstitowner)tij +
7t]6i,(Complex*Compensation)tij +
n170(Complex*Diversiﬁcation)tij + en-J-

(6) no.)- = ﬂooj+l901j(SOCPef0ij+'ij

108

(7) Btm‘ = 7000 +7001(Size)j + 7002(Tenure)j + 7003(R&DINT)j +
7004(Institowner)j Hm5(Diversiﬁcation)j + u]-

Properly specifying temporal aspects of a model are important because it
integrates theory, design, and analysis, i.e., it increases the power and precision of
theories as well as strengthens and lends credibility to subsequent empirical results (e.g.,
Mitchell & James, 2001). With respect to the antecedent model, I can specify when each
strategic factor antecedent should enter the model in relation to the dependent variable
(capital allocation efﬁciency). That is, each strategic factor can enter the model either at
the same time or before the dependent variable, depending on if the effect is likely to be
instantaneous or working with a lag. However, because most ﬁrms plan ahead and decide
on budgets in the year prior to the actual allocation of funds, a one-year lag structure is
appropriate to reﬂect how antecedents affect the actual decision-making process
underlying a subsequent allocation. That is, since I measure the outcome of a capital
allocation process varying in efﬁciency, variables affecting the efﬁciency of that process
need to be entered as that process is taking place, not after it is completed. This indicates
that a one-year lag structure is appropriate.

Another consideration is whether segments belonging to single-business ﬁrms
should be included in the antecedent analysis together with segments belonging to
diversiﬁed ﬁrms. As explained, segments belonging to single-business ﬁrms allocate
capital between potential uses within the ﬁrm, and since they do, the efﬁciency of the
allocation process is likely to be affected by the same multilevel antecedents that affect
the allocation efﬁciency in diversiﬁed ﬁrms. For example, agency theoretical control

variables such as compensation and share holder concentration, are likely to be just as

109

effective in situations where ﬁrms only have one segment. Segments in single-business
ﬁrms are also just as likely to view both historical and social aspirations as appropriate
reference levels for performance. Thus, because single-business ﬁrms are not excluded
from making allocation errors and because the underlying mechanism with which my
multilevel antecedents work are unchanged when applied to (segments belonging to)
single-business ﬁrms, I have elected to include segments belonging to single-business
ﬁrms in the analysis of antecedents to capital allocation efﬁciency. Note, however, that
segments belonging to single-business ﬁrms only are included when external measures of
capital allocation errors are used. Because of the construction of internal measures of

allocation errors, they remain excluded.

Performance Implications: Corporate Effect

The ﬁrst step in investigating how capital allocation efﬁciency is related to the
corporate effect is to run an unconditional (null) STATA XT Mixed model without
predictors. In other words, measures of capital allocation efﬁciency are purposely left out
of the model, which allows for the partitioning of the dispersion in business segment
ROA across time, between business segments, and between corporations. Therefore,
estimates of how much variance is due to variation across time, between business
segments, as well as between corporations can be obtained from the unconditional
model”. In addition, by introducing industry effects at the business-segment level and
thereby accounting for their cross-nesting, individual effects (business-segment,

corporate, industry, and year effects) can also be calculated.

 

2‘ Note that these estimates also are incorporated in the full model as predictors at each level of analysis
explain both between-business segment and between-corporation variance.

110

Therefore, at Time level, business segment ROA at time t for business segment i

within ﬁrm j is a function of average business segment ROA over time (7:017) as well as a
random error (enj):

SegmentROAn'j = 71'01'1' + 6"]: (8)
Usual assumptions of normality for the error term applies such that e tij is assumed to

. . 2 . . . .
have a mean of zero and unrforrn variance of 0 wrthrn each of the tbusrness segments.
. . . . 2

Thus, total busrness segment ROA variance across trme rs o .

At the Business-segment level, the level 1 intercept (troy) representing mean ROA

performance of business segment i over time is regressed on average ROA of all business
segments in corporation j (Bogj) with a between-business residual (rij) accounting for
unexplained variance:

7Foij=l3ooj + "ij (9)
rij is assumed to be normally distributed with a mean of zero and with uniform business

segment variance within each ﬁrm of 17,.

At the Corporate level of analysis, the intercept of the business segment level

model (ﬂmj), representing mean ROA of business segments in corporation j, is now

111

modeled as an outcome varying randomly around a grand mean representing the mean of

all business segments in the sample (7000):

Booj = 7000 + llj (10)
Thus, this level investigates between-corporation variance with its own residual that is

also assumed to have a mean of zero and a between-corporation variance of 1'5,

As mentioned earlier, the unconditional model (Equations 8 through 10) can now

be used to calculate the amount of variance in business segment ROA overtime,
between-business segments, and between-corporations as a proportion of total variance.
That is, by using the variance of the residuals at each level, I can determine the

proportion of total variance that can be attributed to each level. Thus, the amount of

O O I 2 2 I 0
variance across trme rs o / (o + 1,, + {3 ), the amount of vanance between segments rs
2 g 0 O O 2
1",, / ( 0' + 1,, + 1'3 ), while the amount of variance between corporations is 1'3 / (0' + 1,,

+ 1'5).

In addition, by adding year effects to the ﬁrst level (reduced model=R) and
industry effects to the second level (complete model=C), I can in conjunction with the
unconditional models also estimate individual business segment, corporate, industry as

well as year effects”.

 

 

29 When estimating individual effects, the cross nesting of industry effects on business segment
Perfonnance needs to be accounted for by introducing industry effects at the second level (business-
segment level) of analysis.

112

Time level: SegmentROAn'j = 7t01j+ £1,3(Year)tij + en'j (R) ( 11 )

Business-segment level: 71'0ij = [300]: + rij (12)
Corporate level: [3001- = 7000 + ,uj (13)
Time level: SegmentROAnj = nojj+ 7t];j(Year)tij + en'j (C) (14)
Business-segment level: 71'0ij = [3001' + ,8011(Industry)ij + rij ( 15)
Corporate level: ﬂooj = 7000 + pl: (16)

That is, the year effect is calculated by using the proportional difference in residual

2
variance across time between unconditional and reduced models (R): (02 U - 0’2 R Y( (0'

+ ta. + rﬂ) U- Following the logic of entering industry effects at the business segment

level and allowing for cross-nesting of industry effects at both business-segment and
corporate levels, industry effects are calculated by measuring the reduction of variance as
the industry dummies are entered into the model. Thus, industry effects are calculated as

the proportional reduction of variance of the business segment residual as we go from the

reduced to the complete model: (I, R - In C )/( (02 + 1,, + 1:3) R.

Following similar logic of cross—nesting between levels, business-segment and
corporate effects are calculated by accounting for the possibility that some between
business-segment and between corporations variance may be due to industry and year
effects. Therefore, each needs to be adjusted for industry and year effects using the

unconditional, reduced, and complete equations above. Hence, business-segment effects

2
are calculated as (1,, u + 1'7, (3 - 1,, R )/( (0' + I” + rﬂ) R» while the corporate effect is

113

2
calculated as (1)3 u + 1'3 C - 113 R )/( (0' + 1“,, + 13) R. Finally, to adjust variance across
. . . . 2 2
trme for year effects, the followrng formula rs appropriate: (0 u + In C - 1,, R )/( (a + In-

+Tﬁ)R.

At this point, the unconditional model can be extended to include measures of
capital allocation efﬁciency as predictors. Thus, I am now able to investigate the impact
of capital allocation efﬁciency on the corporate effect by introducing measures of capital
allocation efﬁciency at the appropriate level of the unconditional model. Before doing so,
however, I determine if each measure of capital allocation efﬁciency as well as control
variables should enter the model as transient or stable effects, i.e., how much variance of
each measure occurs across time versus in a cross-sectional manner. This was done
previously using intra-class correlation (ICCl) analysis to estimate transient and stable
variance for each individual measure (see Table 2). All measures of capital allocation
efﬁciency and all but three control variables (ﬁrm size, CEO tenure, and R&D intensity)
varied signiﬁcantly over time and will therefore be introduced at the ﬁrst level (Time
level) since they are more likely to explain business-segment ROA over time.
Conversely, ﬁrm size, CEO tenure, and R&D intensity, showing relatively more variation
across ﬁrms will be aggregated and introduced as stable effects at the Corporate level.
Thus, control variables (with the exception of ﬁrm size, CEO tenure and R&D intensity),

measures of capital allocation efﬁciency, and the business segment ROA dispersion
Variable (SRV) enter at time level. Firm size, CEO tenure, and R&D intensity, ﬁnally,

enter at the corporate level forming the following model:

114

SegmentROAn'j =7t0ij + 7:151(Cost)tij+ 7t2ij(LTIP)tij + 7t3iJ(Cashﬂow)tij + (17)
mil-(Measures of capital allocation efﬁciency)tij + 7t5iJ(SRV)tij +

705,3(Measures of capital allocation efﬁciency*SRV)tij + etij
”Oil. zﬂOOj+rlj (18)

[300]“ = 7000 + Y]00(SlZC)j +yzoo(Tenure)j + Y300(R&DINT)j (19)

Using this model, I can determine whether each strategic factor (measures of
capital allocation efﬁciency and business segment return homogeneity) has a statistically
signiﬁcant effect on business segment ROA. In addition, because STATA XT Mixed
incorporates the calculation of variance components, the amount of variance explained by
strategic factors at each level as well as the portion of total variance captured by each
effect can be determined. That is, similar to the variance partitioning of the unconditional
model previously described, unconditional, complete, and reduced models are compared
to estimate the amount of variance each factor explains. Thus, the association between
inefﬁciencies and the corporate effect (H11) is analyzed in two steps. First, the
signiﬁcance of the parameter estimate for each measure of capital allocation efﬁciency is
established. Second, the amount of variance explained is calculated based on applicable
variance components for each statistically signiﬁcant (p<0.05 or better) measure of
capital allocation efﬁciency as it is introduced in the model. For example, models

including only control variables are compared to extended models also including
measures of capital allocation i.e., a comparison of the level 1 residuals between the

models. Associations between business segment return variance and corporate effects

115

(H12) and moderation effects of segment return variance (H13) are established in a
similar manner of classical hypothesis testing.

Furthermore, time lags are speciﬁed using the same underlying logic as in the
antecedent model (Antecedents to Capital Allocation Eﬁiciency): causal inferences
require predictors to be lagged relative to dependent variables unless the effect is
instantaneous. It is likely that misallocations of capital have lagged effects on
performance outcomes. That is, effects of “too much” capital invested in low prospect
segments and “not enough” capital invested in high prospect segments will gradually feed
through to corporate effects, diversiﬁed ﬁrm value, and ﬁrm ROA from the time the
allocations are made and until new allocations are implemented in the next time period.
Thus, because a new budget is decided on each year to be implemented the following
year, a one-year lag is a reasonable length of time for the temporal interplay between
these variables. Therefore, measures of capital allocation efﬁciency, business segment
return homogeneity, as well as control variables enter the model in the year prior to the
outcome variables.

Finally, the fact that segments belonging to single-business ﬁrms are excluded
from models using measures of allocation errors with internal reference points, but not
excluded in models with external reference points, provides a natural way to distinguish
between effects of diversiﬁed and single-business ﬁrms. In addition, it also provides a
compromise between those that believe that the corporate contribution is more

appr'Opriately represented when segments belonging to both diversiﬁed and single-
segment ﬁrms are included (e.g., McGahan & Porter, 1997), and those that believe that

only Segments belonging to diversiﬁed ﬁrms should be represented in corporate effects

116

models (e.g., Bowman & Helfat, 2001: Adner & Helfat, 2003: Brush, Bromiley &

Hendrickx, 1999).

Performance implications: Diversiﬁed Firm Value and Firm ROA

The other two ﬁrm performance outcome dependent variables, diversiﬁed ﬁrm
value and ﬁrm ROA, are not measured at the lowest level of analysis, making a
multilevel speciﬁcation inappropriate for analysis. Therefore, I chose to aggregate
measures of capital allocation efﬁciency to the corporate level to match their higher
levels of analysis, which would enable me to use a single-level technique (such as OLS)
to analyze the data.

Regular OLS, however, is seldom appropriate to analyze panel data sets because
of the potential for correlation between the error structures across panels. That is, since
panel data sets include cross-sections of data each containing a time series, the error term
is two-dimensional, inviting opporttmities for bias both across time as well as within
cross-sectional unit.

In my data set, I have both a temporal (year) and a cross-sectional unit of
observation (ticker), which subsequently requires a more robust approach than regular
OLS can provide. Therefore, I am using STATA’s panel data approach (XT reg), which
uses a more ﬂexible generalized least squares (GLS) technique that more appropriately
deals with common panel data problems, including cross-panel heteroscedasticity and
serial correlation of errors within panels (STATA Longitudinal /Panel-Data Reference
Manual). Further, as an extra precaution against potentially biased (lower) standard
errors, I am using robust errors (option vce robust in STATA XT reg) based on the

White/Huber/sandwich formula to further correct for heteroscedasticity and serial

117

correlation of errors. In addition, because serial correlation of errors within panels are
accounted for and STATA does not offer any post-estimation commands to test for ﬁrst-
order autocorrelation (e. g. Durbin-Watson) within this particular panel data approach, I
ran the affected models with a panel data approach that speciﬁcally corrects for ﬁrst-
order autocorrelation. In other words, this model corrects for autocorrelation by adjusting
coefﬁcients for the potential bias. Comparing results from the two models, I found no
substantial differences in coefﬁcients and thus concluded that autocorrelation is not a
problem in the affected models.

To determine whether tests of performance implications on diversiﬁed ﬁrm values
and ﬁrm ROA should be run with random or ﬁxed effects estimators, a Hausman test
(Hausman, 1971) was performed. For performance implications on Firm ROA, this test
was conclusive and non-signiﬁcant meaning that the more efﬁcient random effects
estimator is the best choice as it is not likely to bias estimates. Thus, the random
estimator was used in regressions with Firm ROA as the dependent variable. In
regressions with diversiﬁed fum value as the dependent variable, however, test were
inconclusive as the Hausman test was signiﬁcant in some regressions, but not in others.
Thus, I decided to run all regressions with diversiﬁed ﬁrm values as the dependent
variable with ﬁxed effects to guard against possible biases from using the more efﬁcient,
but more bias prone random estimator.

Although measures of capital allocation errors with external reference points
allow for the inclusion of segments belonging to single-business ﬁrms, regressions with
diversiﬁed ﬁrm values were constrained to only include segments belonging to

diversiﬁed ﬁrms because the measure is constructed to compare diversiﬁed ﬁrm values

118

 

with potential freestanding segment values. Thus, for single-business fnms, the
comparison would be compromised as it would entail a comparison with itself resulting

in potential bias. Therefore, segments tied to single-business ﬁrms are excluded.

 

119

CHAPTER 4: RESULTS

In this chapter, I present results from both the antecedent part of my full model as
well as from the part describing performance implications of capital allocation efﬁciency,
thus results are delineated into four sections: Antecedents to Capital Allocation Efﬁciency
describes how multilevel antecedents affect capital allocation efﬁciency, Performance
Implications: Corporate Ejfect describes performance implications of capital allocation
efficiency on the corporate effect, Performance Implications: Diversiﬁed ﬁrm value
describes similar effects on diversiﬁed ﬁrm values, and Performance Implications: F inn
ROA, ﬁnally, describes effects of capital allocation efﬁciency on ﬁrm ROA. As described
earlier, both single—business and diversiﬁed ﬁrms (and their segments) are included in
models using measures of capital allocation errors with external reference points, but
excluded from similar measures using internal reference points. Single-business ﬁrms
are, however, excluded from models describing performance implications on diversiﬁed
ﬁrm values. Table 4 presents means and variance as well as correlations among

variables.

Antecedents to Capital Allocation Efficiency

In this model, the dependent variable (capital allocation efﬁciency) is measured
using both internal and external reference points, but only regressions with external
reference points are shown because sample sizes were insufﬁciently small with internal
reference points. Low sample sizes can be explained by the exclusion of segments
belonging to single-business ﬁrms in internal, but not in external measures. Therefore,

since low sample size effectively made it impossible to use measures based on internal

120

 

5.0- 5.0- 5.0 N00- No.0- 00.0 5.0 00.0 N00- 5.0- 5.0 00.0 00.0 2.0 5.0 2.0 3.8.8> E3.3. .coEwom 385mg. .5m
m0.0 .m.0 00.0 00.0 N00 w0.0 5.0- 5.0- No.0 50.0- 00.0 00.0 .00- No.0- 5.5.0 5m.0 :0333.m8>.3..o.o>o.. .0N
M00- m0.0 00.0 50.0 m0.0 5.0 No.0- 5.0- 2.0 00.0 50.0- 00.0- 50.0- 50.0- mmd 2.10 ....8...Eoo.88...=o..>:m .05.
N00- m0.0- 5.0- 5.0 5.0.0 5.0 00.0 00.0 00.0- 000. No.0 5.0 00.0 5.0- 00.0 00.0 Em.Ea=5.u?.:oE:o..>=m. .5
5.0 2.0 5.0 N00 2.0 0.0 5.0 .0.0- m0.0- 5.0 00.0 No.0 00.0 5.0- 002. 003” 803.3.8053 ..8wc3cou .mm
00.0- mmd 2.0 2.0 2.0 2.0 .00. 5.0- 5.0- 5.0- .0.0- 000- N00- 50- 0m.0 .00 0:38:30 8......88..m?.8..33m:. .NN
N00- .0.0 .00. 5.0- 00.0- 50.0- 5.0- 5.0- 00.0- 00.0 00.0 5.0- 5.0- 5.0- 2.0 0N0 0.58:2... 83.0.. V.uo.m. ..N
No.0- 2.0 2.0 .50 .00. 5.0 50.0- 5.0- 2.0 5.0- 00.0- 00.0- 00.0- 00.0- 2.0 00.0 88.89.80 388808.228m .05.
5.0 0.0 .m.0 0N0 5.0 3.0 N00- 50- No.0 2.0- .50. N00 50- 00.0 mmd 50.0 88.89.80 388808.. 33.8.32 .0.
N00 mad N00 N00 5.0 00.0 00.0 00.0 00.0 5.0.0- 5.0- 5.0 00.0 5.0- 02... wvm m8... M9.3.80... 88.0.84 .2
No.0- No.0- 5.0- N00- 5.0.0- .00. 5.0- 5.0- 5.0 No.0 5.0- 00.0 00.0 00.0 v~.5 0N0 33.03.0580 .5.
5.0- mmd- 2.0- 25.0. 0.0 5.0 .v.0 No.0 wad- 50.0 5N.0 2.0 5.0 N00- .cmd 0.0 5.885930% .0.
00.0 0N0 23.0 00.0 00.0 2.0 00.0- 00.0 5.0 2.0- 00.0- 00.0 5.0- 5.0 0.0 0.0 3o... ..maU .2
00.. 00.0 m0.0 00.0 00.0 5.0 N00 m0.0 5.0- No.0 5.0 No.0- No.0 5.0 005 00m 8.5.8. OmU .2
00.. 2 .0 and 00.0 . ..0 5.0.0- 5.0- 5.0 0. .0- 5.0- 00.0 N00- 00.0- m0.~ 0.10 Gowwe: 83m :o.?~_.8w.0 .m .

00.. 24.0 5.0 2.0 00.0 5.0- 00.0 5.0- 00.0- 00.0- 00.0- No.0 2.0 00.0 <0”. =88wom. .N.

00.. 00.0 2.0 00.0- 00.0 5.0 2.0- 00.0- 00.0 5.0- N00 2.0 0.0 <ON. E...» .2

00.. .5.0 2.0 00.0 NWO- 000- 00.0 5.0 00.0 5.0 N00 5.0- $8.30 o:?> Eh". 3836833 .0.

00.. 00.0 5.0- 50.0- 2.0- 5.0- w0.0 00.0 00.0 N50 2.0 383.3 o:?> 8...”. 83.335 .0

00.. .800 5.0 5.0 S: a}. 5.0- 5.0- 2.0 No.0 Ame—30.88.88.823 .38.... .w

00.. 2.0 2.10 S: a). 5.0- 00.0- m0.0 5.0 3330.88.38.8th .88.... .5

00.. 00.0 mad- 5.3.0- 3.. E: 00.0 2.0 $8.30 88.582.38.83 .3833 .0

00.. 2.0- 210- ..E a... .00 00.0 .3833 388.88.8ch .88...m. .m

00.. .500 00.0 N00 3.0 00.0 Go?$..8.=.3>=.8>0 .88.... .v

00.. 00.0 5.0 50.0 No.0 3.8an ..8....3>...8>O .88.... .m

00.. wmd 5.10 50.0 38.3.. 82588390 3.58%. .N

00.. 2.0 No.0 5833.88.38330388...... ..

v. m. N. . . 0. 0 w 5 0 n v m N . .08 .88.). o..?.8>

 

32.2888 Ea seesaw 2.3.888 .4 as...

121

 

 

-F

P

N. N .m- CH. 3— I.
A Iv-UI-nuuonunv-vv Inlnv-=-.d--Av-

x. c. . v. t r .:.....<
V aha-Id I-dnulqunauml- ..vbu-AQu-nUIQA-

.w- 9.2:...-

 

 

8.. 8...- 8...- 8... 5...- 8... 8... 8...- 8... 8...- 8... m... 8...- ...... m... pant; cagegmumagim .R
8.. 8... 8... 8... 8... 8...- 8... 8... 8... 5.... 2...- 8... 2.... R... S..8.-..eo>....o_o>3 .8
8.. mm... 8...- 5...- 5.... w... 8... 8... 5.... N...- 8... 8... no... 85.858.335.38 .8

8.. 8...- 8...- 8...- 8...- 8... 8...- 8.... 8...- 8... 8... 8... ew.e_.=..._s=oe=e_>=m .8

8.. 8... 8...- 8...- 5... 8... 8... 8... 8... 82. 88 8.3285828588 .8

8.. on... 8... o... 8... 8... 8... 2... 8... .m... 8.283.:2..2.o§_m388...... .8

8.. a... 8...- 8...- 8... s... 5...- m... 8... 9.283802... .8... ..N

8.. 2... 8... 5...- .2...- 8... 2... 8... 28.5.88 85.585288 .8

8.. 8... 8...- z...- 8... mm... 8... 28.5.58 855.86.. .352... .2

8.. 5...- 8...- 8... a8. .8 as... 2.885.283 .w.

8.. 2.... 8...- 8.. 8... 858.380 .:

8.. 8...- ..m... 2... baggeaqm .2

8.. S... o... 3856.2

E 8 mm 8 MN 8 a 8 a. w. t o. m. .3 :82 28...;

 

28:32.8. 23.38.80 0.... 8.6..«5 9......88G .v 8..—ah.

122

reference points, I elected to focus on external measures of over and underinvestment for
the purpose of testing this model.

Using extemal measures of over and underinvestment, low sample sizes also
turned out to be a problem, although much less severe than for internal measures. Low
sample sizes were mainly due to the inclusion of CEO tenure because this variable had
substantially fewer usable observations than other variables. Thus, after running similar
regressions both with and without CEO tenure to make sure that this control variable did
not have any material impact on my results, I eliminated CEO tenure in all regressions. In
addition, I also eliminated another control variable, ﬁrm cash ﬂow (Cash ﬂow), because
it exhibited strong collinearity with the existing group of variables. It is possible that the
collinearity may be due to a common denominator (assets) of some of my measures as
described in Wiseman (working paper). However, the underlying mechanism causing this
problem may be more complex, considering that cash ﬂow causes no apparent problem in
other models with similar common denominators. For example, the cash ﬂow variable is
not collinear in models analyzing relationships between inefﬁciencies in the capital
allocation process and diversiﬁed ﬁrm value, even though those are also scaled by assets
(see Table 153). That is, if the common denominator was the only culprit causing the
collinearity, this would most likely be reﬂected in all regressions having assets as the
common denominator.

With the exclusion of CEO tenure and cash flow, sample sizes range from 9,055
to 10,234 in models with control variables only. From those levels, further reductions
occur in two steps as historical and social performance aspirations are added followed by

governance variables. In the ﬁrst step, observations available for analysis are reduced

123

because historical and social aspiration levels are available for only part of the sample. In
addition, only segments with historical and social performance below (above) current
performance are eligible for inclusion when testing for overinvestment
(underinvestment). For example, out of a total sample of 65,713 segments, 25,059 total
historical reference levels (38.1%) are available with only 12,283 (18.7%) of those
representing instances where historical performance is below current performance. Thus,
this decrease in the full sample corresponds to a decrease to roughly 2,500 observations
as historical performance aspirations are added to models with control variables.

In the second step, governance variables30 are added, which further reduces the
sample because block holder ownership, institutional shareholder ownership, and
contingent compensation are missing anywhere from 66.24% to 76.47% of the time
(Table 1). Thus, with the inclusion of governance variables, sample sizes are further

reduced to between 451 and 914 in ﬁnal models.

Antecedents to overinvestment

Tables 5A and 5B provide results of tests for overinvestment. Control variables
exhibited few relationships with overinvestment in models where overinvestment is
scaled by assets, but both organizational size and R & D intensity exhibited expected
positive relationships in models where overinvestment is scaled by sales (Model 1B).
This indicates that the added complexity of both larger and more R & D focused
organizations may make it more difﬁcult for such organizations to allocate capital, thus

increasing the likelihood of overinvestment errors.

 

3° Note that variables measuring diversiﬁcation and environmental uncertainty enter as well, but that they
only have a very marginal effect on sample sizes (Table l)

124

.82.... .....v....._.. .8. x. .. .o..v..<

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

:3 :3. :3 owe coo coo mmoo 2
5d 8.0 cod cod cod cod 00.0 NE0A 00.x.
anew cmdm thm E .3. 3.3. .mdh 3.. N E0 20>»
0.... mod nwd 5.. mud vm. mus 8.00.0.5
NV... and =0....0.-:E0>_Q .0 BS; * 50.0.9000 Sam.
0N... mod 5.8300500 .00m:..:00 * 50.0—9:00 .25
Ed- 0.5.0030 3:032:00. * DUE—0:50 $5.

5.. 0.5.0030 .020; .305 ... 53.07800 .25
cm... 2.. 00.80....895 .0 .03: * E30555 .25

$5 and 000850500 E09550 * E2EN§Q .>:m.

Gd 0.5.0030 3:003:00. ... 0.28.8.0 .25

who 0.580030 .020: .305 * E08050 .>:m

Ned Nod mm. 9.8.0800 .3:0E=0..>=m
:6- mm. _ - om. _ - €0.82... .0.:0E=0...>:m

0m. .- em..- 3... 1.2. <3..- <2...- 00..8..._E0>5 .0 .004
o _ .o- ..Nd 8.0 m _ .0. mod. 3 .o- 00.300an0 503.800
5.0. 00.0- ..md- QEmBEsc 52050.20 8:002:05

... _ «em- ..bod- *~ _ .N- 0280030 .020; ..005
13w _ .v azolaé a: 34. m:..w..00E00 00:08.0...0n. Eoom
3.6.1..» .**N..w 13.8... 0000......E00 00:05.0...9. 30:05.:

Exo- hmd- two- 3.0- 9m...- mvd- o. .o- .208 .0 .000
$6. Nvd- mvd- ood- Sd- omd- cod- 920.... D a. y.
m. ._ mad 5.. mod- Nod- 26. and 0% :0....~.§w.0
<5 <0 <m <9 <m <N <. .0002 _

 

 

 

 

A3080 .3 00.30 .:0§8>=...0>O .0585— " >Gv hon—0.0.bm— 00..-30:4.- _3.na0 00 800—00035. .<m 030,—.

125

AZ.v-:t s»..- .vb—SUI u...::-I.u>:m-..v>av -.:-L.:i..- I. \la- 11-1... II..-

 

.831. .52.: .8. v0 ., .073

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

..wm va va 000 000 000 0:0 2
N00 N00 00.0 N. .0 v. .0 3.0 00.0 N.00A 00.0
CNN 0. .NN mm. . N 0nd. N00. 0.0m. Nm.0N N .00 0.03
00. . - 0N0- 0. .0 N00- mm..- ..m .0- wm. 0008.0.
00.0- 00.00.08.030 .0 .0>0-. * 5.8.0500 .>0m

.N.0- 0000300000 .00w0..000 ... 5.8.0800 .>0m.

2.0 0.00.0030 0.00.5003. * 5.8.0.000 .>0m.

0.00.0030 80.00 ..00.m. ... 5.8.0.000 .>0m.

Quo- 0m.0- 00.0- 00..00_-..w.0>.0 .0 .03.. ... 80.0.0000 .>0m.
0m. N00 N00 0000309000 .00w0..000 ... E30093. .>0m.
0. .0- 3.0- 0.00.0030 00000.03. .._ 0.0.0058. .>0m.
5.0 0.00.0030 80.00 ..00.m. .. 0.0.0.0ch .>0m.

.0.0 0nd 5.8.0.000 .0.00E00..>0m.

00.0 8.0 0. . .- 00.0 0.0.0.0000 .0.00E00..>0m.
00.0 00.0 00.0 med 0 . .0 ..N.0 000.000.0830 .0 .0>0..
0.0 .m.0 00.0- 00.0- no.0 no.0- 00.00.000.000 0.0000000
00.0 mm .0 00.0 0nd 0.030030 80.000800 00000.03.
00.0. 050- 0.00.0030 80.00 002m.

0*...NN... , 18.0.0 33mm... 300800.00 000030.80 0000.0;
unwhﬁ Ihﬁu 3.2..N 30080—000 0000.05.80 0.0.8.0.:

N. .0 v . .0 0. .0 m...0- E0- 050- 00.0- .00000 .0 .30
0N... .md- 3.0- *0m.N *mvN *0vN ***NN.v 5.300. 0 0% y.
3.0- vmd- 00.0- 00.0- 0.0 .00- **Nv.m 0N... 00..0~.00w.0
0.: ..0 ..m a. an ..N m. .0005. J

 

00.00 .3 00.000 .000500>0...0>0 .0080... n >00 5.00.050 00:30.3. 03.000 3 0.0000035. .mm 0.00...

126

Hypothesis la predicts that business segment performance below aspirations for
such performance leads to overinvestment. Consistent with prediction, measures based on
both historical and social aspirations were signiﬁcant in models scaled by assets (Model
2A-7A, p<0.001), as well as by sales“ (Models 23, p<0.01: Model 5373, p<0.001).
Thus, Hypothesis 1a is supported.

Hypotheses 3a and 3b offer the prediction that higher levels of shareholder
concentration reduce instances and magnitudes of overinvestment. Consistent with
prediction, block holder ownership was found to decrease overinvestment in models
where overinvestment is scaled by assets (Model 5A-7A, p<0.05), while institutional
ownership had no effect on overinvestment. The association between block holders and
overinvestment, however, is not consistent across scalars and only signiﬁcant when
measures of social aspirations are used. Thus, Hypothesis 3a, suggesting a negative
association between block holder ownership and overinvestment is partially supported,
While Hypothesis 3b is not supported.

Hypothesis 43 suggests that higher levels of compensation contingent on ﬁrm
Performance may be able to reduce overinvestment. This hypothesis is not supported as I

ﬁnd no signiﬁcance in associations between contingent compensation and
°Verinvestment in my sample.

Hypothesis 5a offers the prediction that higher levels of ﬁrm diversiﬁcation
incrEtases overinvestment. This hypothesis is also not supported as parameter estimates

fail to achieve signiﬁcance in all models except Model 2A-4A where a non-hypothesized

negative association is weakly supported (p<0.10).

\

31
Note that coefﬁcients in Models BB and 4B could not be interpreted due to model non-signiﬁcance.

127

Hypothesis 6a and Hypothesis 7a predicting positive association between

environmental uncertainty and overinvestment are also not supported in any of my

models with the current data.

Finally, the next three sets of hypotheses (H8a-b, H9a-b and HlOa-b) offer
predictions about effects where environmental uncertainty moderates previously
described relationships between level of diversiﬁcation, shareholder concentration, and
contingent compensation, on the one hand, and inefﬁciencies in the capital allocation
process on the other. Following previous logic, I have separated inefﬁciencies into over
and underinvestment to make possible separate predictions for each. Neither Hypothesis
Sa-b, 9a-b, nor lOa-b is, however, supported indicating no evidence that environmental

uncertainty moderates any of the described associations. Thus, overall, my data does not

support any moderating effects of environmental uncertainty.

Antecedents to underinvestment

Tables 6A and 6B provide results of tests for underinvestment. Control variables
exhibited different relationships with underinvestment depending on scaling. In the
model scaled by assets, R & D intensity exhibits an expected positive association,
indicating that the added complexity of more research focused organizations may
contribute to additional underinvestment errors. In models scaled by sales, however, the
aSSoCiation between R & D intensity and underinvestment is negative, indicating a
dc":l‘ﬁasing tendency of more research focused ﬁrms to commit underinvestment errors.
Ful'thermore, organization size exhibits an unexpected negative association with
un‘lﬁtl‘investment with both scalars, indicating that larger ﬁrms may commit fewer

underinvestment errors. Finally, cost of capital exhibits positive associations indicating

128

that ﬁrms with higher funding costs may be more likely to commit underinvestment
errors.

The caveat that a common denominator may inﬂuence associations between R &
D intensity and underinvestment (when both are scaled by sales) applies, but is unlikely

because the association is negative. In addition, associations between overinvestment
scaled by sales and R & D intensity are positive, which conﬂicts with the previous
ﬁnding. Thus, associations involving R & D intensity may more likely be driven by
statistical chance than by a systematic bias due to a common scalar.

Hypothesis 1b predicts that business segment performance above aspirations leads
to lmderinvestment. Parameter estimates are non-signiﬁcant in models with historical
aSpirations, but actually decreases underinvestment in models with social aspirations and
underinvestment scaled by assets (Model 5A-7A). Thus, Hypothesis 1B is not supported.

Hypothesis 2a and Hypothesis 2b predict associations between shareholder

concentration and underinvestment. Although parameter estimates are signiﬁcant for
bIOCk holder ownership (Model ZB-4B), they are in the opposite of the hypothesized
direCtion. Note that parameter estimates in Models 2A-4A cannot be interpreted due to
mode] non-signiﬁcance. Thus, Hypothesis 2a and 2b are not supported.

Hypothesis 4b suggests a negative association between contingent compensation

and underinvestment. In contrast with prediction, this hypothesis is not supported.

Hypothesis 5b predicts a positive relationship between level of diversiﬁcation and

underinvestment. This hypothesis is partially supported as the relationship is signiﬁcant

in r“Odels where underinvestment is scaled by sales (Model ZB-3B, p<0.052 Model 4B-

129

.oo.V0*** ..o.V0** .no. V0 * .o. .V0<

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

000 000 000 00.. wow wow 09.0 2
000 N00 N00 m0.0 00:0 .00 00.0 N.00A 00.0
mNd. 0N.NN 00.... 00.0 0N0. 000 0000 N .00 0.03
..m.m .00 N00 m . .w N00. SmN. 8.0.. 000.00.
N00 mm. 0000000830 .0 .0>0-. * 30.0.0000 .00
00.0- 0N... 0000300000 00030000 * 008.0000 .00
00.0 0.008030 .0000203. ... 0.8.0000 .00
00.0 0.00.0030 80.00 0.00.0 * 000.0000 500

10. .- wm0- 000.000.0830 .0 .0>0-. * 00.00000 .00

00.0- 00.0- 0000300000 00w00000 * 00.00000 .00

000. 0.008030 0000.003. ... 00.00000 .00

0... 0.008030 .0200 0.00.0 ... 00.00000 .00

00.0 N00 0.00.0000 0000000000
00.0 .m. . . 0N.0 00.0- 00.00000 000000.300

00.0 000 00.0 00.0- 00.0 .00- 00000_-..0.0>.0 .0 .0>0-.
v. . . - mm. .- 00.0- wN.0 M00 0. .0 00000000000 00w00000
wN. m0. _ m . .. 0.00.0030 80.000.000 .0000303.
<0... *3.N <05. 0.00.0030 80.00 0.00.0

**Nv.n- .2..de- 1006. 0000000000 00000.0..00 0.00m
00.0- 000- mud- 300000000 00000.0..00 0.0.5.0.:

00.0 00.0 00.0 000 N00 000 *0..N .00000 .0 .000
.00 N00 000. m. ._ .2 Nm. .2... .0.m 3.300. 0 a. 0
w . .0 3.0 00.0 3.0 3.0 00.. 03.2.3.0- 0~.0 000000030
<0 <0 <m <0 <m <N <. .0003. _

 

.000 0. 0.0. 00.00.00... 000.. u >0. 0.000. 008.0. .005 s 053330 .0. 2......

130

.00.v0*** ..0.V0** .00. V0 * .0..V0<

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.0 0.0 0.0 .00 .00 .00 0mg. Z
00.0 00.0 00.0 00.0 00.0 00.0 00.0 N.00A 00.0
00.00 0N. .0 3.00 00.mN. 00.Nm. 00.0N. 040.00. N .00 0.0.3
Nm.0 0. .N. .m.N. mwh. v0.0N Nvém w0.0m. 0000.00.
“_A .0.N- 0.0 00..—000.0030 .0 .0>0.. * 00.0.0000 .>00
0.0.0 0. .0 0000000000 00000000 * 300.0000 .00
30.0 0.00.0030 00003000. ... 5.00.0000 .>00
0. .0- 0.00.0030 00.00 0.00.0 ... 3.00.0000 .>00

*va- w.0- 00.02.00.030 .0 .050 ... 00.00000 .>00

..0- 00.0- 0000000000 00000000 * 0.0.0.0000 .>00

.50- 0.00.0030 00003000. ... 00.00000 .>00

.00.. 0.00.0030 .00.00 0.00.0 ... 00.0.0000 .>00

*N0.N- 050. 50.0.0000 000000.300
0N. N00 0N0- <00. . - 00.00000 0000000000

*00.N imﬁN 18.0 <00. *.N.N *00.N 000350.030 .0 .030
00.0- <00. .- <0... . - NN.0 2.0 0 . .0 0000000000 00000000
00.. 0... Wm... 0.00.0030 00.000000 00003000.
<2... “*me <0... 0.00.0030 00.00 0.00.0

00.. m... .0.— 0000000000 00000.0..00 0.00m
0N0 mu; NN.. 0000000000 00000.0..00 00.0.0.2

N00 00.0 00.0 00.0 N00 00.0 *mmN 00.000 .0 .000
“3.2.00.0- timed- 03.3850- ***0N.0- ***ho.0- ***0v.0- ***0m.. .- 5.000.... n— 00. m
......0- 00.0- .00- 00.. m0. . ... is." .00. 00.0 00.00.0030
0.. 00 mm 00 mm 0N 0— .0002 L

 

0.23. 3 8.8. ...oa.§__.s_...0 .2000 u >0. 050...... 0008.... 0.0.0 2 05.005. .0. 0.00

131

p<0.10: Model SB-6B, p<0.01: Model 7B, p<0.05). Thus, Hypothesis 5b is also only

partially supported as associations are not consistent across my two scalars.

Hypothesis 6b suggesting a positive association between environmental

dynamism and underinvestment is not supported in either models scaled by assets or

sales.

Hypothesis 7b predicts a positive association between environmental complexity

and underinvestment. Although environmental complexity explains variance in

underinvestment scaled by sales (Model 7B), the direction of the association is opposite

of the hypothesized direction. Thus, Hypothesis 7b is not supported.

Hypotheses 8a-b, 9a-b, and lOa-b predicting moderating effects of environmental
uncertainty on associations between antecedents and underinvestment are not supported
in any of my models. Although interactions explain variance in a few models,
associations are in the opposite of the hypothesized directions. That is, any evidence of an
interaction effect where higher levels of environmental uncertainty exacerbates the effect
of diversiﬁcation as well as reduces the effectiveness of both monitoring and contingent

compensation is not present in the current analysis. Thus, Hypotheses 8a-b, 9a-b, and

103-b are not supported.

Performance Implications: Corporate Effect

I will present results from this model in two steps. First, I will show results of the
valﬁance partitioning of segment ROA into components using the unconditional model
and reduced models with time and industry effects added. Second, I will show results as
the unconditional model is extended to include predictors (measures of capital allocation

efﬁ(tiency and business segment return variance), which allows me to obtain estimates of

132

variance components as well as parameter estimates. All extended models are run with
the same limited number of control variables described in the previous section (CEO
tenure and ﬁrm cash ﬂow are eliminated) after I determined that their exclusion had no
discernible impact on signiﬁcance and size of parameter estimates. To further increase
usable sample sizes in regressions, over and underinvestment were combined into an
overall measure of capital allocation errors (Table 9 and Table 10), although separate
effects of over and underinvestment were also investigated in Tables 11 through 14.
While acknowledging that prior variance decomposition studies (e. g., McGahan & Porter,
1 997: Brush, Bromiley & Hendrickx, 1999: Misangyi et al., 2006: Hough, 2006), have
not used control variables in modeling corporate effects to, for example, partial out
effects from ﬁrm characteristics that may be beyond the control of managers, I chose to
include control variables to be able to differentiate between effects of capital allocation
efficiency on business segment ROA and potential alternative explanations. Doing so,
however, suggests that my estimates of explained variance may not be exactly
Comparable to prior studies - my estimates are likely to be lower due to the included
Controls accounting for some of the variance in business segment performance. In
addition, segments belonging to diversiﬁed and single-business ﬁrms are included in the
analysis when measures of capital allocation errors with external reference points are

used, while only diversiﬁed ﬁrms are included as measures with internal reference points

are used.

Unconditional variance partitioning
Table 7 presents results of my estimation of variance components and individual

eff(Bets (year, industry, corporate, business-segment, and time) for models without

133

predictor variables. Calculated variance components, although not directly used to test
hypotheses, are used to measure the variance explained by signiﬁcant predictors.
Using this unconditional model, the largest part of total business segment variance
is across time (61.2%) followed by variance between business segments (25.0%) and
variance between corporations (13.8%). Those three components are further narrowed
down by comparing unconditional, reduced, and complete models as described in the
section on model speciﬁcation and estimation. Results can be found at the bottom of
Table 7: year effects account for 0.1 %, industry effects for 1.3 %, corporate effects for
l 2 -5 %, business-segment effects for 25.0%, while the proportion of variance explained
by time makes up 61.1% of total variance. All variances are highly signiﬁcant (p<0.001).
Effect sizes differ from previous studies with similar multilevel methods (e. g,
Misangyi et al. 2006: Hough, 2006) and from earlier studies using VCA or ANOVA
Variance decomposition techniques (e. g., McGahan & Porter, 1997: Rumelt, 1991). In
Particular, I ﬁnd lower industry effects (1.3%) compared to industry effects in past
research, ranging from 4 percent (Rumelt, 1991) to 9.4 percent (McGahan & Porter,
1997). In addition, my corporate effect is generally higher (12.5 %), while my business
Segnilent effect (25 %) is generally lower than comparable effects in past research (Table
8)- This discrepancy can be explained by my sample covering a later time period (1998
thrcHigh 2006) as opposed to previous sample periods ending in 1999 or earlier. Thus, my
Saulple includes effects of FAS 131, requiring ﬁrms to further split up broader “industry”
or “line of business” segments into more narrow operating segments. Thus, ﬁrms on
aVerage are likely to report more segments in this study as opposed to previous studies

ending in 1999 or earlier. The higher number of business segments tend to increase

134

Table 7. Performance Implications: Corporate Effects Variance Partitioning

Variance P‘
Estimate value

Unconditional model

Irvel l variance (across time). Crij 0.37688 35409 0.000

0.15412 7882 0.000
0.08471 5395 0.000

I..evel 2 variance (between business-segments), rij

Ice vel 3 variance (between corporations). llj

Variance across time as a % of total variance

61.2%
Variance between business segments as a % of total variance 25.0%
Variance between corporations as a % of total variance 13.8%

Model inclusive of year eﬁ‘ects at Level I

Lavel 1 variance (across time). etij 0.37625 35409 0.000

0.15338 7882 0.000
0.08321 5395 0.000

Level 2 variance (between business-segmentS). fij

Level 3 variance (between corporations), llj

Model inclusive of year (Level I ) and industry effects (Level 2)
Level 1 variance (across time). erij 0.37592 35409 0.000

0.15328 7882 0.000
0.07526 5395 0.000

Level 2 variance (between business-segments). rij

Level 3 variance (between corporations), llj

Total variance explained by year effects

0. l %

TOtal variance explained by industry effects 1.3%
TOtal variance explained by corporate effects 12.5%
TOtal variance explained by business-segment effects 25 .0%
Mariana explained by time 61 , 1%

 

colT><>rate effects and decrease business segment effects by directly transferring variance
betWeen the two effects (Bowman & Helfat, 2001). This is easy to see when considering
a Sirrgle-segment ﬁrm (under PAS 14) that is now required to split up a larger “industry-
like” segment into two or more smaller ‘operating-like’ segments to conform to FAS 131.

Thus, because single-segment ﬁrms by deﬁnition have no corporate effects, the change in

135

 

 

reporting regulations will result in larger corporate effects and smaller business segment
effects.

To test the sensitivity of effects to the number of business segments ﬁrms are

reporting, I also decomposed business segment performance of diversiﬁed ﬁrms only.
That is, I excluded all single-business ﬁrms from the analysis, which should increase
corporate effects and decrease business segment effects based on the above argument. In
mat analysis, corporate effects indeed increased to 13.8 %, while business segment
effects decreased to 18.9 % (Table 8). Thus, the increase in business segments had the
forecasted effect in my sample.

The described transfer of variance from business segment to corporate effects as
tile number of reported segments increase also affects the variance allocated to industry
effects. This is because industry effects in this study and other studies (e.g., Misangyi et

al- , 2006) are represented by industry dummies entered at the business segment level of
the multilevel speciﬁcation. Thus, since my reported business segment effects (25.0 %)
are substantially lower than previous studies (Table 8), it follows that my industry effects
logi<:ally also could have a tendency to be lower. In sum, my more current sample period
inlpacts variance estimates of corporate and business segment effects which, in turn,
affects estimates of industry effects by lowering such effects compared to previous
Studies.
The relative size of reported corporate and business segment effects also
c()l‘responds with those reported by Hough (2007). That is, the ratio of corporate to
buSiness segment effects using proportional Rs (Brush et al., 1999) is exactly the same as

that reported by her indicating that corporate effects are smaller than business segment

136

scram
rform

throug’

a erro
bCIwei
(30.2 4
higher
transie
mum}:

captur.

dECOm

effects by 0.71 to 132. This is signiﬁcant because she uses information on “operating
segments while datasets covering earlier time periods contain line of business or industry
information” (page 52). Thus, although her sample covers a shorter time period (1995
through 1999) than mine, her sample is similar to mine in that it reports less aggregated
operating segments which could result in relatively larger corporate and smaller business
segment effects compared to earlier studies using more aggregated “industry” or “line of
bus iness” reporting. Furthermore, absolute differences in effects between our two studies
can be attributed to the substantially smaller proportion of transient variance (described
as error variance) in her study. Thus, because most of her business segment variance is
between segments (65.5 %), it is not surprising that her absolute estimates of corporate
(20- 2 %), business segment (40.2 %), and also industry effects (5.3 %) are substantially
higller than my corresponding estimates (Table 8). Reasons for the smaller proportion of
tI'arlsient variance in her study may include a shorter sample period as well as a different
multilevel model where she characterizes transient variance as error variance and thus
captures that portion as whatever variance is left over after all effects have been entered.
In addition, effects are often not directly comparable between variance
decomposition studies because samples differ also in a number of other ways. With
respect to my study, my sample includes all ﬁrms (manufacturing and service ﬁrms),
While some studies (e.g., Rumelt, 1991: Roquebert et al., 1996) limit theirs to
matlufacttu'ing ﬁrms. My sample also includes both diversiﬁed and single-business furns
With up to 10 segments, while Brush, Bromiley, and Hendrickx (1999), for example,

liIIlited theirs to only diversiﬁed firms with either 3 or 4 segments. Finally, some studies

\

3
2 Ratio = square root (variance of corporate effects/variance of segment effects). My reported ratio = sqrt
(12.5/25) = 0.71. Hough (2007) reported ratio = sqrt (202/401) = 0.71

137

 

Mam
“Source
Years I
ern

No. of

Qifn

Year

their

COHU

Hyp

andr

Estin

TABLE 8. Comparison of Effects With Selected Previous Studies“

 

 

    
  
  
   
 
  
   
  

Arrfelt Arrfelt Misangyi et Hough Mcggxzrn & Rumelt
(diversrﬁed al. (2006) (2006) (1997) (1991)
ms only)
Method Multilevel Multilevel HLM Multilevel ANOVA VCA
So urce of data Compustat Compustat Compustat Compustat Compustat FTC
Years covered 1998-06 1998-06 1984-99 1995-99 1981-94 1974-77
Sectoral coverage All All All All All Manuf.
N o - of observations 35 .409 15,147 10,633 19,405 58,132 10,866
96 of total variance
ear 0.1 0.1 0.8 <1.0 0.3 0.1
Industry 1.3 0.2 7.6 5.3 9.4 4.0
Olporation 12.5 13.8 7.2 20.2 9.1 1.6
Business segment 25.0 18.9 36.6 40.1 35.1 44.2
ime 61 . 1 67.0 47.8 N/A N/A N/A
1' N/A N/A N/A 34.5 46 1 44.8

 

 

 

 

 

 

 

 

*Note that the Brush, Bromiley, and Hendrickx (1999) study is excluded because their method (2SI.S)
limits their sample to ﬁrms with either 3 or 4 segments. Thus, their sample is very different from other
Studies (including those selected for comparison) and also substantially smaller (1,605 segment years in the
three- segment set, 692 segment-years in the four-segment set).

have used stratiﬁed sampling techniques (e.g., Misangyi et al. 2006) in effect limiting
their samples to only a portion of available ﬁrms and segments. Thus, not only are
saInple periods different, but a myriad of other differences between samples may also

cOutribute to the difﬁculty of comparing effect sizes across studies.

HYDOtheses testing
Hypothesis 11 predicts a negative association between capital allocation errors
and the corporate effect. This prediction is tested, as described earlier, in two steps by
loOking at both parameter estimates as well as at variance components. Parameter
estimates are given for all models, while estimates of variance components are

eInphasized when parameter estimates are signiﬁcant (p<0.05 and better). Effects of

138

maul
rtferen
undem

presen‘

corpor

_

lAll

h§

Tune 16
Cost of
Allocar
xiocar
Segre]

I. SR‘
Allocal

'SR
illocal

‘SR
Carpo,
@ganl

R&D

Illlerce

 

capital allocation efﬁciency are tested with measures based on both internal and external

reference points. In addition, to differentiate between effects of overinvestment and

underinvestment, individual effects of overinvestment and underinvestment are also

presented.

Table 9 provides results from testing the effect of allocation efﬁciency on the

corporate effect using internal reference points. Hypothesis ll, predicting a negative

Table 9. Performance Implications: Corporate Effects

Allocation Errors:

 

 

Internal 1

 

Time level

 

Cost of capital

-l.90"

-0.25

 

Al location Errors (assets)

0.

18

-O.24

-0.23

-0.26

-0.24

 

Allocation Errors (sales)

-0.60

-0.68

-0.26

-0.23

 

Segment return variance

-2.03*

 

-2.00*

 

( SRV)
Allocation Errors (assets)

0.80

-l.80"

- l .76"

-1.80"

-0.89

-2.20*

 

* S RV
Allocation Errors (sales)

-2.l6*

 

*SRV

 

Corporate level **

-2.23*

 

0‘ ganization size

 

R & D intensity

 

 

Intercept 17.78

 

20.44

16.10

19.78

19.74

21.32

20.36

20.43

 

Wald Chi2 3.63

 

Ptob >Chi2

0.05
N

0.42
0.8 l

0.67
0.71

3.83

8.32

4.20

7.30

12.23

 

\

30584

 

 

 

0.28

 

0.08

0.12

0.05

 

Va

11729

16321

11399

11399

 

 

 

11820

11399

0.02

 

11399

 

 

%ce components

Level 1 variance. etij

 

Level 2 variance, rij
w 3 variance. p'L

\

 

0.4733

0.4733

 

0.0703

0.0699

 

LeVel 1 variance

0.0461

0.0458

 

Explained

 

10:3] variance explained

 

 

 

 

 

 

 

 

0.00%

 

 

*Not calculated due to model partial or non-signiﬁcance

'"Models with corporate level controls are not shown due to lack of convergence

139

0.10%

influence of capital allocation errors on the size of the corporate effect, is supported when
capital allocation errors are scaled by sales in Model 8 (p<0.05), but not in Model 7
because model non-signiﬁcance suggests that the signiﬁcant coefﬁcient cannot be
interpreted. In addition, when capital allocation errors are scaled by assets, the
association is not signiﬁcant and neither is the model (Model 2). Thus, with respect to
measures of allocation errors using internal reference points, Hypothesis 11 is only
partially supported.

Table 10 provides results from testing the effect of allocation efﬁciency on the
corporate effect using external reference points. Direct effects of allocation errors are
supported in regressions with assets as scalar (Model 2, p<0.001) and explains
approximately 11.80 % of time level variance and 7.42 % of total variance. In addition,
direct effects are supported also with sales as a scalar (Model 7, p<0.05), explaining

1 1 -96 % of time level variance and 7.52 % of total variance. Thus, overall H11 and

as sociations between capital allocation errors and segment performance are supported in
rIZIOdels with both types of reference points and scalars with the exception of when
intel‘nal measures are scaled by assets.

With respect to the two scalars, I acknowledge that the scalar itself may affect
how the measure of business segment performance is interpreted. That is, when scaled by
aSSets, business segment performance may be interpreted as a measure of the efﬁciency
with which assets are used, while interpreted as a measure of margins when scaled by

Sales. Although potential differences in interpretation are not directly supported by Table

9 and 10 since external measures are signiﬁcant regardless of scalar, further investigation

140

8:25; .59 .3 So 852:; m 33......

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

seams ...penmd genes 3:598 8:22? .88.

$3.: $86 $8.: 3:53... 8:25» _ .93

2 mod moved 2 mod 336 386 4.: decaf? m _o>3

amo— .o 2.86 oped v8. .6 3.3 .o .5 .351? N .93

mmed owned 836 amend emomd :6 .0825; _ .33

2:26 So.» eoeetexs

much N51 ENE mCRm Feb :9: :9: 53. $28 exam 2
e _ .c 2...: N _ .o cod cod cod N. .o 3.0 cod med NED A poi

x _ .m. cow. new. 5.9 wcdm NOON Eh mod 8. _m $.m NEU 255

e _ .m. 3.8 wcam was 3 .m. m. _ .NN emdm o. e. 5.8 we: gauche:—
cwd $6 320:: D Q m

bed wed vim coceﬁcaweo

3:: 23:5th

Rd 3.3- >zm... $233 225 5:32?

3.3%.»: ***wx.m- 35* $.83 225 5:32?

3%..” £3. _- 3x..- ﬁned- hm..- <mw._- owe >Mm 85:? ESE EoEwum
who mm..- to. _ - *mm.m- 22:3 form cos—522

ovd- Nod- we. _ - 1.3:..- 3633 E85 5:822

cad w _ .o w _ .o *mbN- Sac w _ .o w _ .o w _ .o ......Owd- <8. _ - .553 .8 .80

323 25.5

o. o w n e n v m. N N _a=._2xm "ﬂatm— =e=aoe=<

 

 

 

 

 

 

 

 

 

 

 

 

38% 3235 “25.335 asset: .2 use

141

of such 1

dissena:

separate
.hsocia
11mm
1.13016:
is sca‘e
inrema
Signiﬁt‘
Model

and im

 

of such differences could be an interesting avenue for future extensions of this
dissertation.
To allow for a more precise test of H11, capital allocation errors are, once more,

separated into over and underinvestment to allow for separate predictions and tests.
As sociations between overinvestment and business segment performance support
Hypothesis 11. External overinvestment is signiﬁcant in models using both scalars
(T able 11), although only weakly signiﬁcant (p<0.10) in Model 4 where overinvestment
is scaled by sales and controls at the corporate level are used. Associations between
ititemal overinvestment and segment performance (Table 12) are, however, only
Sigrlifrcant in models without control variables at the corporate level (Model 1, p<0.05:

Model 3, p<0.01). Thus, with respect to overinvestment and measures using both external

311d internal reference points, H11 is only partially supported.

Table 11. Performance Implications: Corporate Effects

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

L Overinvestment: External 1 2 3 4

Time level

Cost of capital 0.44 -0.27 0.46 003

External Overinvestment (assets) -6.83*** -6.37***

External Overinvestment (sales) -6.14*** -1.65"
Corporate level

Organization size 1.73" 1.76"

R & D intensity -8.28*** -8.10***
Intercept 29.66 23.01 29.58 22.49
Wald Chi2 46.91 I 10.4 37.94 72.39
Prob > Chi2 0.00 0.00 0.00 0.00
N 15431 6824 15514 6889

 

 

 

 

 

"p<.10, * p< .05, **p<.01, ***p<.001

142

 

Table 12. Performance Implications: Corporate Effects

 

1 Overinvestment: Internal

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 2 3 4

Time level

Cost of capital -0.20 -0.66 -0.22 0.69

Internal Overinvestment (assets) -2. 12* -0.11

Internal Overinvestment (sales) -3.42** - 1.02
Corporate level

Organization size -0.65 -0.65

R & D intensity 545*” 6.44***
Intercept 19.59 8.30 19.21 7.64
Wald Chi2 6.35 30.68 11.73 43.8
Prob > Chi2 0.02 0.00 0.00 0.00
N 4169 23 19 4203 2346

 

 

 

"p<.10, * p< .05, **p<.01, ***p<.001
Associations between underinvestment and business segment performance offer
more ambiguous support for H11. As presented in Table 13, internal measures33
I‘epresenting effects of segments belonging to diversiﬁed ﬁrms only are not related to
business segment performance, while measures of underinvestment calculated with
GXternal reference point scaled by sales (Table 14: Model 3 and 4) exhibits a non-
hyPothesized positive association with the dependent variable. Thus, with respect to
underinvestment, Hypothesis 11, suggesting a negative relationship between
underinvestment and the corporate effect is not supported.
Hypothesis 12 predicts a negative relationship between the variance of business
segment ROA within ﬁrms and the corporate effect. This hypothesis is partially
supported as parameter estimates for segment return variance are signiﬁcant in some

models where allocation errors are included as well (Table 9: Model 8 and 9, Table 10:

 

33 Note that Model 2 and Model 3 did not converge due to a number of potential reasons including (but not
limited to) a lack of variance to allow the calculation of variance components.

143

Model 6). Thus, overall, a negative association between segment return variance and the

corporate effect is partially supported.

Table 13. Performance Implications: Corporate Effects

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I Underinvestment: Internal 1 2* 3* 4

Time level

Cost of capital -0.32 009

Internal Underinvestment (assets) -0.63

Internal Underinvestment (sales) 0.58
Corporate level

Organization size 0.2

R & D intensity 163"
Intercept 5.00 3.21
Wald Chi2 0.49 3.32
Prob > Chi2 0.78 0.5
N 7462 5363
*Models did not converge

Table 14. Performance Implications: Corporate Effects

 

 

 

 

 

 

[ Underinvestment: External 1 2 3 4
Time level
Cost of capital -2.16* -1.41 -2.22 — l .40
External Underinvestment (assets) -0.55 054
External Underinvestment (sales) 936*“ 7 .43***

 

 

Corporate level

 

 

 

 

 

 

 

 

 

 

 

 

 

Organization size 3.90*** 4.56***

R & D intensity -9.99*** -7.75***
Intercept 6.07 4.64 -3.80 -3.93
Wald Chi2 5.01 124.59 92.26 180.6
Prob > Chi2 0.08 0.00 0.00 0.00
N 11899 7663 11959 7712

 

"p<.10, * p<.05,**p<.01,***p<.001

144

Hypothesis 13, ﬁnally, suggesting a moderating effect of segment return variance
on the association between allocation errors and business segment performance is
supported in models with allocation errors constructed with both intemal (Table 9) and
external (Table 10) reference points. The associations, however, are not consistently
supported across models with both scalars and corporate level control variables. That is,
the association is supported in Model 9 (p<0.05) where internal reference points are used,
while only supported with assets as a scalar when external reference points are used

(Model 5, p<0.001: Model 6, p<0.001). Thus, Hypothesis 13 is partially supported.

Performance Implications: Diversified firm value

Table 15A—B presents results from my estimation of relationships between capital
allocation efﬁciency and diversiﬁed ﬁrm values. In particular, Table 15A presents results
Where both independent and dependent variables are scaled by sales, while Table 153

Presents results from regressions scaled by assets. In contrast with previous models
eStimating effects of antecedents on capital allocation efﬁciency and of capital allocation
efficiency on the corporate effect, sample size and collinearity considerations did not
interfere with my analysis thus the full range of control variables (including CEO tenure
and cash ﬂow) were included in regressions. Also in contrast with previous models,

regressions involving external measures were constrained to only include diversiﬁed

ﬁrms.

Diversiﬁed firm value scaled by sales (Table 15A)

Control variables exhibited relationships with diversiﬁed fn'm value in expected

as well as unexpected directions. Organization size exhibited a negative association with

145

my measure of diversiﬁed ﬁrm value”, indicating higher discounts for larger ﬁrms, while
CEO tenure exhibited a generally expected (positive) relationship, indicating lower
discounts (or higher ﬁrm values) when ﬁrms are run with more experienced CEOs. Cash
ﬂow and cost of capital, on the other hand, exhibited somewhat surprising associations
with the dependent variable given Jensen’s work on how cash rich organizations invest
(e- g., Jensen, 1886: 1991: 1993). That is, according to Jensen, ﬁrms with access to ample
and cheap investment capital tend to invest in ways that could lower ﬁrm values, not
increase them as I ﬁnd evidence of. Finally, R&D intensity exhibited a positive
association with diversiﬁed ﬁrm values, while long-term incentive plans was unrelated to
Observed discounts in diversiﬁed ﬁrms.
Models 2A-6A assesses associations between overinvestment and diversiﬁed ﬁrm
Value (Hypothesis 14a) as well as between underinvestment and diversiﬁed ﬁrm value
(Hypothesis 14b) using sales to scale relevant measures. Thus, the hypotheses assess the
I“Elationship between capital allocation errors and the degree of discount (or premium)
exhibited by diversiﬁed ﬁrms as compared to their potential value if their business
Segment were stand-alone entities. Hypothesis 14a predicts a negative relationship
between overinvestment and diversiﬁed ﬁrm value. This hypothesis was assessed in
Models 2A and 3A. Contrary to prediction, however, both internal and external measures
of overinvestment exhibited positive relationships with my measures of diversiﬁed ﬁrm
value (p<0.001). indicating that, although overinvestment explains variance in diversiﬁed
ﬁrm value, it is in the opposite of the hypothesized direction as overinvestment tend to

increase ﬁrm values. In addition, Model 6A uses the Berger and Ofek measure of

 

34 The diversiﬁed ﬁrm value construct is measured as actual ﬁrm value over imputed ﬁrm value if all
business segments instead were freestanding.

146

overin
signiﬁ
over'm
1651111:
thresi
captu.
belov
perfc
alloc
pore}

OVer

1 F11
Ore
Ca:
Co:
L0:

Ex:
[nu

Inn
0v

eff/7.27:7

overinvestment to assess effects on diversiﬁed ﬁrm values with only very marginal
signiﬁcance in the expected direction (p<0.10). The difference between my measures of
overinvestment and the measure used by Berger and Ofek may account for the divergent
results in associations between overinvestment and diversiﬁed ﬁrm value. That is, the
threshold level set for overinvestment is much more severe in their measure only
capturing ﬁrms with very poor future prospects (Tobin’s q below 0.25 as opposed to
below 0.5), thus they likely capture segments with little or no chance to reverse poor
performance. With my less strict threshold, however, it is likely that some segments
allocated extra investment may be able to turn their performance around, thereby

potentially accounting for the positive association between my measures of

OVerinvestment and diversiﬁed ﬁrm value.

Table 15A. Performance Implications: Diversiﬁed Firm Value

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Exed effects model: Scaled by

sales 1A 2A 3A 4A 5A 6A
Organization size -9.49*** -9.45*** -9.26*** -9.54*** -9.41*** .2.08*
CEO tenure 2.09'“ 2.08alt 1.86" 1.86" 2.09* 1.43
Cash flow 10.68“” 10.73““M 10.32*** 10.81*** 10.73*** 2.01*
R & D intensity 465*" 4.66*** 4.51*** 4.70m 4.66*** 0.42
Cost of capital -2.62** -2.64** -2.61** -2.65** -2.63""'I -2.28*
Long-term incentive plans 0.02 0.01 0.03 -0.05 -0.06 -1.15
External Overinvestment (sales) 382*"
Internal Overinvestment (sales) 2.92"
External Underinvestment (sales) 233*
Internal Underinvestment (sales) 0.32
Over_berger & Ofek -1.67"
Intercept 7.63 7.66 7.53 7.67 7.63 0.57
F (number of variables, DF) 33.23 29.16 28.36 30.59 28.69 4.13
Prob > F 011) 0.00 0.00 0.00 0.00 0.00
N 3179 3179 3179 3179 3179 277
"p<.10, * p< .05, **p<.01, ***p<.001

 

147

 

Hypothesis 14b predicts a negative association between underinvestment and

diversiﬁed ﬁrm value. This hypothesis was assessed in Models 4A and 5A. Contrary to

prediction, the external measure exhibited a positive association with diversiﬁed ﬁrm

values, while the internal measure exhibited a non-signiﬁcant relationship.

In sum, therefore, although both overinvestment and underinvestment explain

variance in diversiﬁed ﬁrm value, it is contrary to the predicted direction. Thus, neither

Hypothesis 14a nor 14b are supported.

Diversiﬁed firm value scaled by assets (Table 15B)

Control variables exhibit slightly different relationships than in the prior model.

Although organization size, CEO tenure, cost of capital, and cash flow have similar

effects on the dependent variable as in the previous model (15A), R&D intensity, and

‘

Table 15B. Performance Implications: Diversiﬁed Firm Value

 

\:xed effects model: Scaled by
assets

13

23

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3B 4B 5B GB

Organization size -9.29*** -9.3o*** ~9.26*** -9.34*** -9.22*** -270"
CEO tenure 2.32* 2.23* 2.28* 2.30* 2.32* 1.64"
Cash ﬂow 859*" 8.58*** 8.48*** 8.55*** 8.58*** 3.77***
R & D intensity -155 -154 -159 -154 -154 -2.22*
Cost of capital -4.38*** -4.36*** -4.40*** -4.37*** —4.37*** -2.02*
Long-term incentive plans 2.59" 2.59* 2.69“ 2.59* 2.60" -0.61
External Overinvestment (assets) 187"

Internal Overinvestment (assets) [80"

External Underinvestment (assets) 0.76

Internal Underinvestment (assets) -0.44

Over_berger & Ofek - l .96*

Intercept 9.40 9.41 9.36 9.42 9.34 2.80
F (number of variables, DF) 32.23 28.20 27.81 27.77 27.66 6.30
Prob > F 0.00 0.00 0.00 0.00 0.00 0.00
N 2825 2825 2825 2825 2825 268

Ap<.10, * p< .05, **p<.01, ***p<.001

 

 

 

 

148

 

long-term i1
1DC€1111V€ p'

Res
6A in two
significant
prediction
models. T
that exhib

H3p0thes

Perform:
T.
ROA. In
excluded
generallf
the rate (
indicari:
FURhErr
but POSi
may be
a530cm

may hat

FirmR

 

long-term incentive plans do not. R&D intensity is now non-signiﬁcant, while long-term
incentive plans instead is positively related to the dependent variable.

Results from Models 2B-6B differ from those reported for previous Models 2A-
6A in two respects. First, direct effects of overinvestment are now only weakly
signiﬁcant (Model 2B, p<0.10: Model 3B, p<0.10), but still in a direction contrary to
prediction. Second, underinvestment is not related to diversiﬁed fum value in any of my
models. Therefore, with the exception of the Berger & Ofek measure of overinvestment

that exhibited an expected negative association with diversiﬁed ﬁrm value, neither

Hypothesis 14a nor 14b is supported.

Performance Implications: Firm ROA

Tables 16A and 16B present result of the test of allocation efﬁciency on Firm
ROA. In these models, control variables (with the exception of cash flow that was
exeluded because it is measured in essentially the same way as the dependent variable)
generally exhibited strong relationships with ﬁrm ROA. Only ﬁrm size was unrelated to
the rate of return on assets, while cost of capital had a negative effect on Firm ROA,
indicating a plausible negative relationship between cost of funding and ﬁrm returns.
Furthermore, CEO tenure and long-term incentive plans exhibited an equally plausible,
but positive relationship implying that more experienced and better compensated CEOs
may be able to raise asset returns. Surprisingly, observed R&D intensity had a negative
association with Firm ROA, indicating that more research and development focused ﬁrms
may have lower returns on assets.
In order to assess performance implications of capital allocation efﬁciency on

Firm ROA, Hypothesis 15a predicts a negative association between underinvestment and

149

Firm ROA, while Hypothesis 15b predicts a negative association between overinvestment
and Firm ROA. Consistent with prediction, Model 4A and Model 5A show that internal
measures of underinvestment were negatively related to Firm ROA for both scalars
(p<0.01). In addition, one of the external measures of underinvestment was signiﬁcant
(Model 2A, p<0.01) offering additional support for hypothesis 15a, predicting a negative
as sociation between underinvestment and Firm ROA.
Hypothesis 15b predicts a negative relationship between overinvestment and Firm
ROA. This prediction is partially supported as Model 2B is signiﬁcant in the predicted
direction (p<0.001), while remaining models returned non-signiﬁcant parameter
estimates. Overall, therefore, Hypothesis 15a, predicting a negative relationship between
underinvestment and Firm ROA is relatively well—supported as three out of four possible
measures are signiﬁcant in the predicted direction. Hypothesis 15b, however, is only
Partially supported as only 1 out of 5 models are signiﬁcant in the predicted direction.

Speciﬁcally, only external overinvestment scaled by assets was signiﬁcant.

Table 16A. Performance Implications: Firm ROA

y

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

L Underinvestment 1A 2A 3A 4A 5A
Organization size -1.55 -0.53 -0.40 -2.21* - l .43
CEO tenure 2.46** 2.92" 3.09"”I 0.63 1.72"
R & D intensity -4.12*** -4.53*** -2.95** -l.62" -2. 19*
Cost of capital -4.07*** -6.16*** -7.23*** -3.29** -2.22*
Long-term incentive plans 482*" 4.20*** 3.75*** 5. l7*** 4.14***
External Underinvestment (assets) -2.97**

External Underinvestment (sales) 0.04

Internal Underinvestment (assets) .28?"

Internal Underinvestment (sales) -320"
Intercept 16.14*** 13.24*** 16.16*** 18.27*** 19.09***
Wald Chi 2 67.25 84.42 81.23 64.41 31.54
Prob > Chi2 0.00 0.00 0.00 0.00 0.00

N 4634 1912 2207 1051 1358

 

 

 

 

 

 

"p<.10, * p< .05, **p<.01, ***p<.001

150

 

TABLE 16B. Performance Implications: Firm ROA

 

 

l Overinvestment 18 23 3B 4B 53 on
Organization size -1.55 -1.82" -2.07* - l .93" -1.70" -0.71
CEO tenure 2.46* 1.45 1.75" 2.96** 2.91 ** 0.48

R & D intensity -4.12*** -1.28 -3.69*** -5.03*** -3.39** -l.73"
Cost of capital -4.07*** 0.43 -1.75" -3.26** -3.66*** -3.54***
Long-term incentive plans 4.82*** 0.67 2.77** 3.42** 0.86 0.97
External Overinvestment (assets) -3.50***

External Overinvestment (sales) 0.15

Internal Overinvestment (assets) 059

Internal Overinvestment (sales) 0.41
Over_berger & Ofek 0.86
Intercept 16. 14*** 12.60*** 9.56*** 19.91*** 20.00*** 13.72***
Wald Chi 2 67.25 18.05 25.02 54.08 31.88 17.34
Prob > Chi2 0.00 0.00 0.00 0.00 0.00 0.00

Ii 4634 649 7 16 1068 871 354

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

"p<.10, * p< .05, **p<.01, ***p<.001

Summarizing the described ﬁndings of my two models investigating antecedents

and performance implications, antecedents primarily affects allocation efﬁciency through

the effect of performance aspirations. Performance shortfalls in relation to aspirations for

Such performance generally increase overinvestment as predicted, while instances when

current performance is above aspirations do not, as predicted, lead to increased

underinvestment. Instead, my results indicate that current performance below aspirations

tends to decrease underinvestment. Further, governance variables affect allocation

efﬁciency through the effect of block holders generally acting to reduce allocation to

business segments. That is, my results indicate that block holders tend to both decrease

overinvestment and increase underinvestment for ﬁrms in my sample. Environmental

uncertainty, ﬁnally, did not materially affect allocation efﬁciency in my study.

Effects of allocation efﬁciency on performance outcomes differ across the three

measures of ﬁrm performance. Both over and underinvestment reduce ﬁrm ROA as

151

predicted, while only overinvestment reduces corporate effects. Diversiﬁed ﬁrm value,
however, is affected in the Opposite direction by overinvestment — excessive allocation of
capital to segments with relatively poor prospects tends to increase ﬁrm values. Finally,

business segment variance affects corporate effects as both direct and moderated effects

are supported.

152

 

CHAPTER 5: DISCUSSION, LIMITATIONS, AND EXTENSIONS
Discussion
Introduction

In this dissertation, I have argued that the capital allocation process is an
important, albeit by management researchers, a fairly overlooked corporate level process
that may have strong implications for ﬁrm performance. The central argument is that a
number of multilevel antecedents affect the efﬁciency of this process which, by
extension, also affects the level of performance of diversiﬁed as well as the single-
business ﬁrm. This is in contrast with much of the prior literature on capital allocation
that has been more concerned with building elegant models of capital allocation based on
ad—hoc assumptions (e. g., Scharfstein & Stein, 2000: Stein, 1997) than with empirical
investigations of performance implications of capital allocation efﬁciency in ﬁrms. Prior
research has also shown virtually no interest in antecedents to the efﬁciency of the capital
allocation process.

This dissertation, therefore, may be viewed as an initial step in building
knowledge about the capital allocation process, including knowledge about drivers tied to
the efﬁciency of the process, as well as performance implications of allocation efﬁciency
within the boundaries of the fum. Speciﬁcally, my results show that the level of
efﬁciency of the capital allocation process is an important factor in the performance of
diversiﬁed as well as single—business ﬁrms. In addition, my results, although somewhat
mixed, also show how a number of antecedents determine, as well as offer opportunities
for stakeholders to exercise control over the capital allocation process thereby affecting

its efﬁciency. Below, I will discuss implications of my ﬁndings in four sections,

153

including performance implications of capital allocation efﬁciency on diversiﬁed ﬁrm
value, ﬁrm ROA, and on the corporate effect, in that order. Following that, I will describe

implications of my results examining antecedents to capital allocation efﬁciency.

Performance Implications on Diversiﬁed firm value

More the grained ﬁndings of this study also support the above interpretation that
the capital allocation process has important implications for ﬁrm performance. Namely,
inefﬁciencies in the capital allocation process explain signiﬁcant variance in diversiﬁed
ﬁrm value. More speciﬁcally, I ﬁnd evidence of a positive association between
overinvestment and diversiﬁed ﬁrm value, indicating that overinvestment instead of
decreasing ﬁrm value, actually increases such value in relation to potential stand-alone
ﬁrm value if all business segments were freestanding. A number of explanations subject
to future veriﬁcation emerge for this surprising, counterintuitive, and interesting ﬁnding.
First, overinvestment, although predicted to be a main driver of inefﬁcient capital
allocation resulting in worse ﬁrm performance for the diversiﬁed ﬁrm may, after all, not
reduce ﬁrm performance. Instead, there may be processes at work whereby ﬁrms use
private knowledge not shared by capital markets about segments likely to show
signiﬁcant and quick improvement if properly fertilized with extra capital allocations.
That is, there may be opportunities for overinvestment in select business segments
potentially resulting in quick improvements of segment performance. Thus, ﬁrms may
use this knowledge to their advantage by improving performance in ways initially
overlooked by capital markets, but eventually acknowledged in the form of higher stock
prices for ﬁrms successful in picking segments worthy of extra investment capital. By

extension, since the reference level used to delineate business segments with relatively

154

ll»

 

better prospects from those with relatively worse prospects is the median (or average)
such prospect, the implication is that the average ﬁrm may be skilled at picking
improvement targets that, although technically on the low side of the reference point,
may be very close to the reference point and therefore offer potentially quick and easy
ﬁxes resulting in higher fn'm values.

With respect to improving segments close to the reference point, it is interesting
to note how the Berger & Ofek measure of overinvestment suggests a limit to the value-
increasing effect of overinvestment. That is, selective overinvestment leading to higher
ﬁrm values may not apply to ﬁrms with prospects that are much below the median of
such prospects. Thus, segments with very low prospects (such as those below the 25th
percentile) may be poor candidates for overinvestment since their segment performance
will not likely increase as a result of additional allocations.

A second explanation relies on the reasonable assumption that investors use
heuristics or short-cuts (e.g., Kahneman, Slovic & Tversky, 1982), to deal with the
complexities inherent in judging the performance of business segments and ﬁrms. Prior
research has suggested a publicity effect with legitimacy beneﬁts (e.g., Zuckerman, 1999:
Pollack & Rindova, 2003) if the ﬁrms managers and their actions are visible enough to be
reported in the business press. That is, since investors may be unable to perfectly judge
the efﬁciency and effectiveness of each decision (including decisions to allocate capital)
made by the ﬁrm, the visibility of each action or decision may be interpreted as a signal
containing information on how well the ﬁrm is doing. Thus, since ﬁrms betting heavily
on certain business segments may be more visible (and more likely subjects to press

coverage) than those with more conservative investment strategies, investors may

155

anticipate higher performance from the former group of ﬁrms. Because distinctive actions
such as heavy overinvestment are often attributed to the CEO by business press
journalists (Hayward, Rindova & Pollock, 2004), CEO celebrity effects may also add to
the visibility of the ﬁrm and its managers. In some sense, therefore, overinvestment may
signal to shareholders that managers are taking prompt and strong action in the hope of
raising overall ﬁrm performance. Hence, if investors believe that signal, they may bid up
stock prices of overinvesting ﬁrms potentially creating and reinforcing this positive
association between overinvestment and ﬁrm value. However, against the background
that overinvestment tends to reduce ﬁrm performance as measured by both the corporate
effect and ﬁrm ROA, this signal may be less than rational as overinvestment may be
more likely to reduce ﬁrm performance. Thus, investors relying on this heuristic to make
decisions may inadvertently bid up share prices of ﬁrms that are making decisions more
likely to hurt, than to increase their ﬁrm performance.

Finally, a third post hoc explanation for the divergent effects of over and
underinvestment on ﬁrm value is that both internal and external reference levels of
investment may be too low, potentially biasing measures of overinvestment. That is, the
hurdle for overinvestment (average level of investment) may be set too low, resulting in
many segments being deemed to overinvest, although they may in reality receive just the
Optimal amount of investment. This implies that average levels of capital investment in
diversiﬁed ﬁrms may be lower than optimal. Investors seem to strongly support this
contention by rewarding fnms overinvesting with signiﬁcantly higher ﬁrm values

compared to those that do not. That is, the positive association between overinvestment

156

and ﬁrm value could be interpreted as investors viewing the average level of ﬁrm
investment as too low, and signaling for it to be raised.

Implications of the ﬁnding that overinvestment increases ﬁrm values depend on
which explanation(s) are deemed most feasible. Still, clear overall implications can be
delineated for managers of diversiﬁed ﬁrms and for academic scholars, respectively. For
ﬁrms, the positive association between overinvestment and ﬁrm values implies that "
managers may want to look over their current allocations — those allocations may be
insufﬁciently low to allow for maximization of existing proﬁt and growth opportunities.
Firms in cash intensive industries, ﬁrms in early stages (e.g., start-up ﬁrms), and those
actively involved in M&A activities, may be particularly vulnerable with respect to
insufﬁcient rates of investment. Further, suggested visibility effects imply that investors,
often hindered from objective insights into the internal workings of ﬁrms, use heuristics
to substitute for such insights. Thus, ﬁrm managers should acknowledge that investor
may use heuristics to form judgments of complex processes (such as the capital allocation
process) that may drive ﬁrms share prices.

For scholars, the main implication may be to avoid assuming that overinvestment
(as deﬁned by existing measures) has negative implication for diversiﬁed ﬁrm
performance. Instead, selective overinvestment based on private information not shared
by markets may, at least temporarily, be supportive of ﬁrm performance and lead to
higher share prices. Ultimately, however, incongruence between theories suggesting
negative performance implications of overinvestment and opposite empirical results
needs to be reconciled with a better measure of capital allocation errors. That is, a

measure based not on the distance between actual allocations and average levels of

 

157

investment, but between actual allocations and optimal levels is needed. Thus, scholars
may want to focus on constructing models for calculating ﬁrm speciﬁc optimal levels of
capital allocation that could subsequently become a starting point for estimating capital
allocation errors. Further, scholars may also want to acknowledge and incorporate into
models predicting performance that heuristics, often used to interpret, predict, and judge
overall ﬁrm performance, may also be used by investors to judge the performance and
rationality of more fme- grained corporate level process in the diversiﬁed ﬁrm. That is,
investors may use heuristics to judge the efﬁciency of the capital allocation process.
Thus, to obtain a more thorough understanding of how the capital allocation process
affects ﬁrm performance, scholars may be well advised to study how investors not only
use heuristics to judge the effectiveness of the capital allocation process, but also how
those heuristics affect share prices, the latter to get a more complete picture of how
capital allocation errors (or the absence of such errors) affect ﬁrm value.

With regards to underinvestment, I ﬁnd some evidence that underinvestment in
relatively better performing business segments is positively associated with diversiﬁed
ﬁrm value, meaning that underinvestment tends to increase actual ﬁrm value in relation
to a ﬁrm’s hypothetical value if no underinvestment existed. Although I am hesitant to
draw conclusions based on this ﬁnding because it was not consistently supported in my
analysis, the value-increasing effect of underinvestment still supports my theoretical
argument that over and underinvestment should be treated as two separate capital

allocation errors with independent effects on ﬁrm performance. That is, since

158

overinvestment clearly is a different construct from underinvestment”, researchers
should no longer continue to assume that they are dependent, or that underinvestment
automatically follows overinvestment as has been common practice in the literature on
capital allocation (e. g., Berger & Ofek, 1995). Instead, each can exist on its own with

independent effects on diversiﬁed ﬁrm value.

Performance Implications on Firm ROA

Inefﬁciencies in the capital allocation process also explain signiﬁcant variance in
ﬁrm ROA. In fact, compared to associations between capital allocation efﬁciency and
diversiﬁed ﬁrm value, associations with ﬁrm ROA are more in line with predicted
results. That is, underinvestment has consistently negative performance implications by
reducing ﬁrm return on assets, while overinvestment also reduces ﬁrm ROA, just not as
consistently across all measures and scalars. This inconsistency in results between
overinvestment and ﬁrm ROA may be due to, as previously suggested, the hurdle rate for
overinvestment set too low, allowing instances of optimal allocation to be characterized
as overinvestment which, if true, could potentially account for the inconsistency of

overinvestment ﬁndings.

Performance Implications on Corporate Effects
Finally, capital allocation efﬁciency also explains signiﬁcant variance in business

segment ROA and the corporate effect. Inefﬁciencies (over and underinvestment

 

35 The difference between the two constructs becomes evident when considering that well-performing
segments cannot overinvest while under-performing segments cannot underinvest. Therefore, a ﬁrm may
have only well-performing (underperforming) segments in its portfolio which makes overinvestment
(underinvestment) impossible. In addition, there are additional uses of capital (e.g., buy-backs of company
shares, mergers & acquisition activity) that cash-rich ﬁrms can pursue instead of automatically allocating
excess capital to business segments.

159

 

 

combined into one measure) based on external reference points explain between 11.80%
and 11.96% of level 1 (time-level) variance and between 7.42% and 7.52% of total
variance depending on scalar. Thus, capital allocation errors explain signiﬁcant variance
in business segment performance.

This variance, however, is not stable across time and may therefore not be
recognized as variance related to the corporate effect. That is, corporate effects have
traditionally been viewed as stable effects (e.g., McGahan & Porter, 1997), or
performance differences between ﬁrms that are invariant over longer periods of time. A
more contemporary way of viewing corporate effects, however, may be to recognize that
corporate level inﬂuences may also be dynamic, something ﬁrst suggested by Adner and
Helfat (2003). Thus, according to that more contemporary view and logic, inefﬁciencies
in the capital allocation process, accounting for substantial variance in year-to-year
variations of business segment ROA should be viewed as a “dynamic” type of corporate
effect.

Considering the evidence of associations between capital allocation errors and
business segment performance, individual effects of over and underinvestment can be
used to differentiate between over and underinvestment, and to determine exactly what
drives the overall effect of allocation errors on business segment performance.
Interestingly, only overinvestment corresponds with the previous ﬁnding, i.e., exhibiting
a negative association with business segment performance. Underinvestment, on the other
hand, seems to have a positive effect on business segment performance as evidenced by a
positive association in a few models. The discrepancy in effects between over and

underinvestment poses an interesting question for future extension of this dissertation —

160

 

 

why does overinvestment, but not underinvestment affect business segment performance
and the corporate effect in the hypothesized direction? One possible reason subject to
future veriﬁcation may be that the level of business segment variance is different between
ﬁrms in the two categories — overinvesting ﬁrms may have segments with higher
variance in their segment retums as compared to ﬁrms that underinvest. This will be
elaborated on in the next section.

Associations involving segment return variance also warrant further attention. The
question of whether the level of segment return variance biases the variance
decomposition process has been much debated in the literature (e.g., Bowman & Helfat,
2001: Adner & Helfat, 2003), with as of yet, no agreement or even a single empirical test
of this potential bias. I, however, ﬁnd evidence that the level of segment return variance
has the potential to bias estimates of corporate effects. That is, signiﬁcant direct effects of
allocation errors suddenly become nonsigniﬁcant as segment retum variance is added to
the model. In addition, I ﬁnd evidence of an interaction effect between capital allocation
errors and segment return variance as the interaction explains variance up and above the
explanatory power of each component. Thus, this supports researchers questioning the
variance decomposition methodology (e.g., Bowman & Helfat, 2001: Adner & Helfat,
2003: Brush, Bromiley & Hendrickx, 1999), and suggests that future variance
decomposition research may need to control for the variance of segment returns within
ﬁrms as to not bias regression coefﬁcients and results.

As mentioned earlier, it is possible that this bias could help explain the
discrepancy between over and underinvestment and their individual effects on business

segment performance. Suppose segments with poor prospects that overinvest tend to

161

belong to ﬁrms with relatively high variance of their segment returns, while segments
with relatively better prospects that underinvest tend to belong to ﬁrms with relatively
lower variance of their segment returns. In that case, the discrepancy in effects of over
and underinvestment may be partly driven by the variance of underlying segment returns
- they strengthen the association between overinvestment and business segment
performance, while weakening the association between underinvestment and business
segment performance. This explanation will be investigated in future extensions of this

dissertation.

Antecedents to Capital Allocation Efficiency

The analysis of antecedents to capital allocation efﬁciency offers several ﬁndings
that contribute to our understanding of what drives and determines the efﬁciency of the
capital allocation process. From those ﬁndings, several implications emerge.

First, managers in charge of allocating capital are not only affected by
considerations tied directly to the decision at hand, but also by their aspirations for
business segment performance in combination with current performance. Consistent with
predictions, aspirations for such performance have powerful effects on the capital
allocation decision. Shortfalls in relation to aspirations for segment performance tend to
increase investment levels leading to overinvestment in poorly performing business
segments. This result holds true for performance shortfalls in relation to both relative and
to historical performance, which suggests that managers actively monitor and reacts to,
not only segment performance relative to industry competitors, but also to prior

performance of the particular focal segment.

162

Instances when current segment performance exceeds aspirations for such
performance do not as hypothesized, however, lead to increased levels of
underinvestment. Instead, my results suggest that managers exceeding their aspirations
for segment performance may act in a more rational manner by reducing instances and
magnitudes of underinvestment. At ﬁrst glance, this may seem to contradict the referent-
dependent logic underlying aspiration driven predictions of choice behavior. However,
there may be other forces at work to partially, or fully, offset these aspiration driven
inﬂuences. As previously discussed, organizations may have established routines (Nelson
& Winter, 1982) in the form of strong norms or rules governing the allocation of capital
(e.g., Bower, 1986: Shin & Stulz, 1998). That is, organizations may, for example,
automatically raise allocations each year based on increases in cash flows or proﬁts of the
ﬁrm. Alternatively, as also suggested by previous research (e.g., Scharfstein & Stein,
2000), allocations may be tied to the cash ﬂow of each individual business segment thus
may increase when business segments generate more cash, and decrease when the
generate less. Based on described norms and mles of allocation, allocations to well-
performing business segments may increase each year accounting for this empirical
result.

Second, although effects of agency-theoretical variables on the efﬁciency of the
capital allocation process are somewhat disappointing, my results nonetheless
demonstrate that large shareholders (block holders) may have the power and interest to
affect allocations, and thereby the efﬁciency of the allocation process. The effect of block
holders, however, is only partly in line with predictions and is therefore not consistently

beneﬁcial to owners by aligning interests, including reducing the risk differential (Coffee,

163

 

1988) between owners and protection-minded managers in charge of allocating capital.
That is, I ﬁnd evidence that block holders reduce levels and magnitudes of
overinvestment, but that they seem to increase underinvestment at the same time. Thus,
they are effective in loss contexts by reversing underlying managerial tendencies to
allocate excess capital to relatively less well-performing segments, but not in gain
contexts where they exacerbate protection-minded managerial tendencies to allocate
insufﬁcient amounts of capital to better performing segments by limiting investment in
promising business segments. In effect, therefore, block holder’s combined inﬂuence on
the capital allocation process may be a general tendency to lower allocations - both when
ﬁrms are overinvesting as well as when they are underinvesting in their respective
segments.

The response of block holders to overinvestment, although consistent with
predictions, is interesting because it can be viewed as contradicting the previously
discussed positive association between overinvestment and diversiﬁed ﬁrm values. That
is, block holders as an important share holder in most ﬁrms have a tendency to bid up
share prices of diversiﬁed ﬁrms that are overinvesting, while at the same time working to
lower allocations of the same ﬁrms. Notwithstanding that effects on diversiﬁed ﬁrm
values are stronger and more consistent as compared to direct effects of block holders in
antecedent models, this is an interesting contradiction in results that could be due to a
number of reasons subject to future veriﬁcation. First, the difference could be because of
important differences between ﬁrms in the two models — diversiﬁed as well as non-
diversiﬁed (single-segment and single-business) ﬁrms are allowed in antecedent models,

but only diversiﬁed ﬁrms are allowed in diversiﬁed ﬁrm value models. Thus, it could be

164

 

that block holders favor overinvestment only in diversiﬁed ﬁrms, but not in less
diversiﬁed ﬁrms. Results from the antecedent model supports this explanation as block
holders only have an allocation-reducing effect in models where both diversiﬁed as well
as non-diversiﬁed ﬁrms are included (e.g., models with external reference points). In fact,
when only diversiﬁed ﬁrms are included (models with internal reference points), block
holders have no effect on overinvestment. Second, the difference in block holder effect
may also stem from the presumably two different functions being performed by block
holders in the two models - investing in models investigating diversiﬁed ﬁrm value, but
monitoring in antecedent models. In fact, it corresponds to the anomaly pointed out by a
number of authors (e.g., Coffee, 1988) - agency theory presumes that share holders as
investors are risk neutral, but that that they have less of a risk appetite (risk averse) as
managers. Although monitoring is not necessarily the same as managing, there are
important similarities in that large share holders are presumed to monitor managerial
activity and intervene when managers act without their best interest in mind. Thus, my
ﬁnding corresponds to agency-theory and predictions of how managers act in their dual

roles as both investors and managers.

Limitations and Extensions

Limitations of this study are mainly related to my measures. In particular,
measures of inefﬁciencies in the capital allocation process have limitations. Although the
complexity of measuring the efﬁciency of a process (particularly one as complex as the
capital allocation process) in an objective manner, the relative lack of prior work and
guidance within capital allocation research, as well as the lack of studies using closer

performance measures (as evidenced in many calls to study the performance of closer

165

 

processes, e.g., Adner & Helfat, 2003: Misangyi et al. 2006) may have contributed, there
are speciﬁc limitations including the methodology underlying measures used that warrant
ftu'ther attention. Below, I will outline some of those limitations.

The common structure of capital allocation efﬁciency measures relying on
reference points creates a number of problems related to empirical analyses. First, the
reference point structure limits the number of instances of over and underinvestment that
may be identiﬁed in any given sample. That is, ﬁrrrrs are deemed to overinvest
(underinvest) in a particular business segment only if two conditions are fulﬁlled: (1),
that the business segment has below (above) median prospect as measured by a reference
point, and (2), that the particular business segment receives more (less) capital as
compared to the average investment of relatively higher (above median) and (lower)
prospect segments”. In addition, measures are not sensitive to excessive investment in
segments deemed to be above median prospects, nor are measures sensitive to excessive
underinvestment in segments deemed to be below median prospect, which leaves out
possible instances of capital allocation errors. Thus, measures of capital allocation
efﬁciency may be too stringent in deﬁning which business segments are committing
capital allocation errors and which ones are not. Many ﬁrms, therefore, may be
overinvesting as well as underinvesting in their different business segments, but may not
be categorized that way potentially leading to overlooked instances of capital allocation
errors. In fact, this probably contributed to insufﬁcient sample sizes when modeling

antecedents with internal measures of capital allocation errors. Although this limiting

 

3° The logic underlying these measures dictates that some cut-off is used to differentiate between segments
with high and low prospects. Because of the symmetry of over and underinvestment, this cut-off has to be
at the 50'“ percentile or less. Thus, the current cut-off is the least restrictive that it could be while still
retaining the logic of the measures. In fact, Berger and Ofek (1995) use a much more severe 25Ill percentile
cut-off to differentiate between segments with high and low prospects.

166

inﬂuence on sample sizes may not bias results, it contributes to the difﬁculty of ﬁnding
consistent evidence of associations between both antecedents and capital allocation errors
as well between capital allocation errors and performance outcomes.

Another problem with the reference point construction of allocation measures is
that internal measures are more restrictive than external measures. That is, internal
measures only allow the inclusion of diversiﬁed ﬁrms, while external measures allow for
the inclusion of both diversiﬁed and single-business ﬁrms. Although this difference
between measures allows a comparison between effects of diversiﬁed versus single—
business ﬁrms, it also strongly limits the number of useable observations in models with
internal measures of allocation efﬁciency. In fact, this is the underlying reason why the
number of observations was insufﬁcient in antecedent models with internal allocation
errors.

A third problem with the reference point construction is the inclusion of a high
level of noise into the resulting measures. In particular, external measures of allocation
errors assume that all segments in a particular industry have similar prospects. Although
this is motivated by the difﬁculty of applying a measure of future prospects (e.g., Tobin’s
q) to business segments, this likely adds noise to my measures as within-industry
differences between segment prospects are not reﬂected. This symmetrical assignment of
prospects within industries may also contribute to lower variance of the measure itself.
Thus, the additional noise and the lower variance, together, reduces useful variance and
thereby further adds to the difﬁculty of achieving signiﬁcant results when using measures

of capital allocation efﬁciency in empirical analysis.

167

Finally, reference centric capital allocation measures are at risk of bias because
they use the average level of capital allocation to determine whether a ﬁrm is allocating
capital in an optimal manner, allocating too much, or not enough capital to its respective
business segments. Thus, as previously described, this introduces the potential for bias
since the rational or optimal level of capital allocation may not correlate with the average
level used as a reference level in my measures of capital allocation efﬁciency. Therefore,
if the reference level is not optimal, neither predictions involving antecedents and their
effect on capital allocation efﬁciency, nor performance implications of such efﬁciency
can be completely trusted. Thus, although my current measures of capital allocation
errors are a worthwhile evolution from measures used by previous scholars (e.g., Berger
& Ofek, 1995), they may still overlook many instances of ﬁrms committing capital
allocation errors due to stringent threshold levels. In addition, they may also include
much noise, be biased by the use of non-optimal reference points, and thereby also offer
lower levels of useful variance to model.

My measures and their impact on the number of observations may also affect the
ability to generalize my ﬁndings over a larger population of ﬁrms and segments. In
particular, antecedent models may be susceptible to this issue because they include
allocation (allocation errors) as well as aspiration measures -— both of which create sharp
losses of observations as the original sample is divided into appropriate groups for the
purpose of hypotheses testing. That is, my sample is ﬁrst divided into two groups of
segments based on segment prospects — below median prospect segments are in one
group to enable a test of overinvestment, while above median prospect segments are

divided into a different group to enable a test of underinvestment. Within those groups,

168

 

segments are further divided into groups based on if current operating performance is
above or below aspirations for such performance. Thus, two distinct groups of
observations are formed depending on whether the test is for overinvestment (below
median prospect and below performance aspirations), or for underinvestment (above
median prospect and above performance aspirations). Observations not ﬁtting either
group are discarded. Therefore, allocation and aspiration measures necessitate a drastic
reduction in observations as segments are divided into appropriate dissimilar groups to
enable hypotheses testing. These groups are further reduced by the inclusion of other
independent as well as control variables.

To assess a potential sample bias due to differences between fnms and segments
in my antecedent models versus the overall population of ﬁrms and segments, I
performed three different t-tests comparing means of four ﬁrm and segment
characteristics. In the ﬁrst test, I compared means of variables in individual models of
over and underinvestment (N=451-914) with means of the same variables in the
population sample (N=30,039-53,465). In the second test, I compared means of variables
in individual models of over and underinvestment (N=451-914) with observations
speciﬁcally tied to only over or underinvestment (N=9,842-26,843). This was done to
avoid comparing segments operating below (or above) median prospects with a
population of segments operating both below and above median prospects. In the third
test, I compared means of segments making allocation errors with the population means.
This last test can be viewed as an alternative way of avoiding to compare segments
operating below (or above) median prospects with a population of segments operating

both below and above median prospects.

169

 

Results of the three tests (Table 17-19) suggest that means are not similar between
my models of antecedents to capital efﬁciency and population derived comparison
samples. That is, the alternative hypothesis (Ha) stating that means compared are
signiﬁcantly different from each other is supported in most cases. Implications of these
tests with regards to the existence of a possible sample bias may, however, still be
unclear for at least two reasons. First, in all three tests I compare distinct groups of
segments with a largely ungrouped population of segments. When looking at
underinvestment in the ﬁrst test, for example, I compare segments with both better
prospects and better performance (relative to the median segment and to competitor
and/or historical performance, respectively), to all segments regardless of prospects and
performance. When looking at underinvestment in the second test, I still compare a
distinct group of segments with current performance above aspirations to a sample of
segments performing both below as well as above performance aspirations. The third test
looks at allocation errors (over and underinvestment together), and compares means to
the full population of segments. This may be the most attractive test because it contains
observations from both over and underinvestment together, but it still compares those two
distinct groups to an ungrouped population. Thus, segments not being either below or
above both reference levels are excluded from the comparison. In sum, all three tests
have limitations because distinct groupings of segments are compared to a largely
ungrouped population of segments. Therefore, since the division of segments is based on
the dissimilarity of segments between groups, and because those groups are distinct from
each other, it may be unrealistic to expect segments to be similar in other respects

including those tested for.

170

Second, it is also unclear what level this test should be conducted at. On the one
hand, the antecedent investigation is at the business segment level as the dependent
variable is measured at the lowest level. On the other hand, antecedents are at all three
levels (industry, ﬁrm, and segment) — thus, should the burden of the test rest on the
similarity of segments, or the ﬁrms that those segments belong to? Put differently,
similarity between ﬁrms may not matter when looking at segment-level inﬂuences, while
similarity between segments may not matter when looking at ﬁrm-level inﬂuences.
Closely related to this is the added caveat that segment prospects are assigned based on
industry prospects, which could increase segment similarity within, but not between
industries.

Furthermore, the ﬁrst problem of comparing distinct groups to a largely
ungrouped population may indeed be difﬁcult to solve as the population of segments
cannot be made to replicate model groupings without actually replicating the models
themselves. Thus, even after performing the described tests, the possibility remains that a
sample bias exists and that I therefore may not be able to generalize my ﬁndings to a
larger population of ﬁrms and segments”.

What I can do, however, is to generalize my theoretical conclusions (Mook,
1983), which may further our understanding of the capital allocation process. That is, I
have argued theoretically and shown empirically that behavioral antecedents affect

allocation efﬁciency, and I can generalize those conclusions to the capital allocation

 

37 The ability to generalize my ﬁndings over a larger population of segments and ﬁrms may also
be affected by the limited number of observations available for R & D intensity. That is, ﬁrms, and by
extension segments, report corporate level R & D spending only around 55 % of the time (Table 1).
Although this caveat would affect all studies using R & D spending or similar variables, it suggests that my
results may not generalize to select industries where R & D may be an unlikely expense such as for
example retailing.

171

 

 

NvdVQ .on. 9:55:

weave ..I :54“:

Sex. .3 :55

 

oncva dun :55

 

 

 

save dun :55

Save .3 :55

Save .3 :55

 

£68 .ou. :55.—

 

 

 

8.ovo .3 :55:

Save dun e55

Sovo dun .c5a:

:évm .ou. t5a:

 

 

 

 

86X. .3 :55.—

Sevn dun Ema:

8.95 5"” Ema:

Sex. dun :55:

 

 

nﬂdwuzcaoz ..oeoscaoE
.353 ..o .80

omodmuzﬁwo—z $35522
9225 5 a. m

n$.mnuz=32 noose—.32
“Em Semi—390

canonuzcaog nova—=32

 

 

 

 

 

 

<O¢ EoEwom
3on2 nmwnz Saul movnz
Exam av <b.<m <v.<~ 352389
.252 “me ﬂash -mN .232 ﬁe 933—. .852 #3 03:. .252 36 93:.

 

Ease—anon .EuEﬁoEEo—ED 8 3=oeouou=<v Sacra. 29:8 «56:32:: 5: ~35.

 

8..ch .cum :5 a:

dem .3”. :5 a:

ccdvm 51 :5 5:

cocva .cu. :5 a:

Sova .31 :5 5:

 

 

 

8.ch 51:5 a:

5.on en :5 a:

dea duh :5 H...:

dem .cnm :5 H...:

8.on 5”” :5 5:

 

 

 

Save 51:5 5:

Sevm .on, :5 5:

oodvm .cnm :5 a:

Save sum :5 a:

Save .ou, :5 “$—

 

 

 

8.325 en :5 a:

Save .ouz :5 a:

Save .on_ :5 5:

Sevn 51:5 a:

codvm .cu. :5 5:

 

 

humuz

vauz

eeenz

EvHZ

eeeuz

Eoﬁzgoz 425.582
2253 .5 .50

omodmnzcaez .335562
55:25 O as m

metmmuzcae—z 4.335502
35 :o_EN:_sw.O

emmdnuzgeg 4255:32
<Om EuEmom

 

 

an .352 "mm «35,—.

 

me
-m—m .252 ”mm 035—.

 

me
.MN .252 "mm 935,—.

 

<h-<m
.282 "<m 035,—.

 

<v-<~

 

.252 Em osﬁ

Etc—meeﬁh.

 

Ease—anon 455392130 3 $=~e88=<v Sworn. «35am «aoenonoei JR— 035.

172

 

 

....ov: .ou. :5M:

adv: en, :55:

Rev: .oum :55:

iv: .ou: :5M:

 

 

 

 

 

Sexes". :5a:

8.9: .on_ :53:

86v: .3 :5a:

86v: en: 25::

 

 

 

Sex: .ou: :5a:

gov: d": :55:

86v: .on: :55:

gov: d": :55

 

 

 

 

86v: due :55:

86v: .oum :5a:

Sev: 5": :5a:

85v: 5": :5a:

 

 

SWSHZCNOE ..ovoﬁcmuz
.355 .5 .80

vﬁ.~_nz=82 65:52
9:55. Q a. x

NR.-uz=52 535502
on; 555530

8:35:52 ..oeoegoE
<OM 52:wa

 

 

 

 

 

 

Saul $an meanz meeuz
awn—m me <h-<m <vi<N 55555.9
.252 "mule—nah. in" .052 "We 935—. 55—2 "<0 «3:. 55—2 35 03:8

 

535.5%:— 535: J5Eumo>=€3=D 3 $5555<V :53? 2553565515 .mwu oiah

 

8.0x: .cum :5 a:

8.3!: dun :5 5:

ccdvm .cum :5 ”a:

8?: en: :5 5:

23v: .on, :5 5:

 

 

 

85v: 5”, :5 5:

93v: .cl :5 a:

8.9: .0”: :5 a:

adv: .on. :5 5:

codvm .0”: t5 “:1

 

 

 

Sex: .cnm :5 a:

8.3!: .cu. :5 a:

gov: .0”. :5 a:

ocdvm 6”” t5 5:

8.3x: .onm :5 5:

 

 

 

8.3: .3”, :5 5:

_c.:vm 5“: ...:Q a:

codv: .cum :5 a:

adv: .cn. :5 H~..:

8?: en :5 ”...:

 

 

bwmnz

vaHZ

eeenZ

5an

seer-Z

32.552552 4.355552
:3an .5 .50

viénzcauz 55.5522
bias 5 a m

o. m.c~u2::oz .....BECQo—Z
0N7 5:55.550

omw.mmuz:=uE 4.65552
<0: .55me

 

 

m: .252 "mm wish.

 

me
-mm 5.52 “mm wink.

av
-mN .352 "mm 53:.

 

 

<b-<m
.252 "<m 03:.

 

<v.<N
.252 “<m ~35.

 

38232.9

 

535.352. 525: 455535530 3 355553 3:89 255: 5065515 .3: 03:9

173

 

adv: .ou. :5 a:

cdem .OH. :5 ”:3

iv: .cu. :5 5:

onov: .on. :5 H5.:

Gov: .cu. :5 a:

 

 

 

85v: .on. :5 5:

86v: .ou. :5 5:

oodvm .ou. :5 5:

86X. .3 :5 a:

8.9: .ou. :5 5:

 

 

 

86v: .cu. :5 5:

gov: .on. :5 a:

8.0!: .on. :5 a:

85v: .3 :5 a:

oodvm .on. :5 ”a:

 

 

 

85v: .3 :5 a:

86v: .on. :5 a:

gov: .on. :5 5:

oodv: d". :5 5:

85v: .ou. :5 5:

 

 

 

 

 

 

 

mm ..ovnzcwo—z Luge—.35.
.553 :0 .30

amodmuzcuuv’. +4.55:on
€2.25 5 a :

notmmuzcuos. 6.5.5.33.

3.: 5:35.55

om~.owuz:a2>. ..ouoEEBS.

 

 

 

<Oy. 52:me
SmWZ wan—NZ 2.2.1.2 cam—HZ mu..uz
m: .252 meéﬁ .222). 51mm .252 <h-<m .253. <v.<~ .25.). .2535m2é
“no .25 mm 025.. "no 2...: mm 2...; "no .25 mm 22:... 26 2.: <m wish "<9 .2:— <m uEah

 

35.3.2.2. 6.559 5.38.? 3 2.2.3254. 383—. 2958 5252.22: .3 225—.

174

 

process. In particular, current performance in relation to performance aspirations may
have distinct effects on the level of efﬁciency of the capital allocation process. Thus,
although I cannot say exactly under what circumstances and contexts this may happen, it
could happen, which may make it necessary for future research to include behavioral
antecedents when studying capital allocation efﬁciency in ﬁrms. In addition, my
conclusions may also add to our understanding of how the allocation process may impact
a number of well-deﬁned performance outcomes for ﬁrms. For example, based on my
theoretical arguments and empirical ﬁndings tying capital allocation efﬁciency with both
corporate effects and diversiﬁcation discounts, researchers working in those areas may
need to consider the impact of capital allocation efﬁciency not to “underspecify” their
theoretical as well as empirical models.

To overcome many of the measure-related limitations and to assure continued
theoretical and empirical progress within capital allocation research, a new approach to
measuring capital allocation efﬁciency may be necessary. Such approach should be
centered on a theory-driven optimal level of investment instead of being centered, as is
currently the case, on empirically observed average levels of actual allocated capital.
Within such a framework, a baseline optimal level of capital investment could be
calculated for each ﬁrm that would be congruent with each ﬁrms’ idiosyncratic
environment. That optimal level of capital investment could subsequently be broken
down into individual allocations to business segments. Thus, deviations from that optimal
allocation could be measured and translated into a more continuous measure of capital
allocation errors. This approach would likely generate additional instances of capital

allocation errors since previously unaccounted for instances will now be accounted for.

175

For example, overinvestment in relatively better performing segments, as well as
underinvestment in relatively worse performing segments could now be accounted for. In
addition, since these new theory-driven measures of capital allocation efﬁciency are
based on optimal levels of capital allocation, they may be both more precise and less
likely to be afﬂicted with biases compared to existing measures.

This new framework for measuring allocation errors would enable a number of
important extensions of the capital allocation literature. First, it allows an investigation
into optimal levels of capital allocation. That is, I have presented some evidence that
average levels of allocated capital in my sample may be too low as additional allocations
often lead to increased ﬁrm values. Thus, this opens up an interesting avenue for future
investigation assuming optimal levels can be established. Second, with more precise
measures of capital allocation efﬁciency, offering additional power in empirical analysis,
additional antecedents and more complicated structures of antecedents could be
investigated with increased precision. In particular, future scholars could investigate
interesting interactions not only between corporate and environmental factors following
this dissertation, but also between the corporate level factors themselves. For example, in
line with Devers er al., (2008), scholars could investigate how individual parts of CEO
pay packages (e. g., cash pay, stock options, and restricted stock) may interact in complex
ways to affect managerial risk preferences, thereby also potentially affecting the
efﬁciency of the capital allocation process.

Third, more precise and efﬁcient measures of capital allocation efﬁciency would
also increase the power and reach of investigations into performance outcomes of capital

allocation efﬁciency. In particular, closer performance outcomes (other than diversiﬁed

176

ﬁrm value and the corporate effect) that are derivatives of more traditional accounting
and market based measures may become possible, or at least easier, to model with better
measures of allocation efﬁciency. In addition, it may also become possible to estimate
relationships between the capital allocation process and other processes, allowing for a
more fme- grained insight into the internal workings of the ﬁrm.

A ﬁnal beneﬁt of this new framework would be a stronger link between
applicable theory and empirical results. That is, less noisy, more bias-free and efﬁcient
measures closely linked with an optimal capital allocation process may be more likely to
produce results corresponding with applicable theory. Thus, empirical studies of capital
allocation errors will be more likely to show consistent and unambiguous results across
different samples, types of error (over and underinvestment), and scalars used. In
addition, the increased empirical power stemming from more efﬁcient measures could be
used to test more advanced theoretical models looking at possible relationships between
antecedents and performance outcomes of capital allocation efﬁciency not only across

industries, but also within individual industries.

177

REFERENCES

Aboody, D., & Lev, B. (2000). Information asymmetry, R&D, and insider gains. Journal
of Finance 55: 2747-2766.

 

Adner, R., & Helfat, CE. (2003). Corporate effects and dynamic managerial capabilities.
Strategic Management Journal, 24: 1011-1025.

Aldrich, H. (1979). Organizations and Environments. Prentice-Hall, Englewood Cliffs,
NJ.

Amihud, Y., & Lev, B. (1981). Risk reduction as a managerial motive for conglomerate
mergers. The Bell Journal of Economics 12: 605-617.

 

Andrews, KR. (1987). The Concept of Corporate Strategy, 3rd edn. Richard D. Irwin:
Homewood, IL.

Argyres, N. (1996). Capabilities, technological diversiﬁcation, and divisionalization.
Strategic Management Journal, 17: 395-410.

Avolio, B.J, Zhu, W., Koh, W., & Bhatia, P. (2004). Transformational leadership and
organizational commitment: mediating role of psychological empowerment and

moderating role of structural distance. Journal of Organizational Behavior, 25: 951-
968.

Bain, J .S. (1956). Barriers to New Compptition. Cambridge: Harvard University Press.

Barclay, M., & Holdemess, CG. (1990). Negotiated block trades and comratg control.
Working paper, University of Rochester, Rochester, NY.

Barker, V.L., & Duhaime, LM. (1997). Strategic change in the turnaround process: theory
and empirical evidence. Strategic Management Journal, 18: 13-3 8,

Baysinger, B.D., Kosnik, R.D., Turk, T.A. (1991). Effects of board and ownership

structure on corporate R&D strategy. Academy of Management Journal, 1: 205-
2 14.

Berger, P & Ofek, E. (1995). Diversiﬁcation effects on ﬁrm value. Journal of Financial
Economics, 44: 39-65.

Bergh, D. (1997). Predicting divestiture of unrelated acquisitions: an integrative model of
ex ante conditions. Strategic Management Journal, 18: 715-731.

Berle, A.A. & Means, G.C. (1932). The modern comration and private property.
Harcourt, Brace & World, New York.

178

Bernardo, A.E., Luo, J ., & Wang, J .J .D. (2006). A theory of socialistic internal capital
markets. Journal of Financial Economics, 80: 485-509.

Bethel, J .E. & Liebeskind, J. (1993). The effects of ownership structure on corporate
restructuring. Strategic Management Journal, 14: 15-31.

Birkinshaw, J ., Braunerhjelm, P., Holm, U., & Terjesen, S. (2006). Why do some
multinational corporations relocate their headquarters overseas? Strategic
Management Journal, 27: 681-700.

Bliese, PD. (2000). Within group agreement, non-independence, and reliability:
implications of data aggregation and analysis. In Multilevel Theog, Research, and

Methods in Organizations: Foundations, Extensions, and New Directions, Klein
K.J., Kozlowski S.J (eds). Josey-Bass: San Francisco, CA: 349-381.

Bollerslev, T. (1986). Generalized autoregressive conditional heteroscedasticity. Journal
of Econometrics, 31: 307-327.

Bolton, MK. (1993). Organizational innovation and substandard performance: When is
necessity the mother of irmovation? Organization Science, 4: 57-75.

Bower, J .L. (1986). Managing the resource allocation process, Harvard Business School
Press, Boston, Massachusetts.

Bowen, H.P. & Wiersema, M.F. (2005). Foreign-based competition and corporate
diversiﬁcation strategy. Strategic Management Journal, 26: 1153-1171.

Bowman, E.H. & Helfat, CE. (2001). Does corporate strategy matter? Strategic
Management Journal, 22: 1-23.

Boyd, B.K. (1990). Corporate linkages and organizational environment: A test of the
resource dependence model. Strategic Management Journal, 11: 419-430.

Boyd, B.K. (1995). CEO duality and ﬁrm performance: A contingency model. Strategic
Management Journal, 16: 301-312.

Boyd, B.K., Gove, S., & Hitt, MA. (2005). Consequences of measurement problems in
strategic management research: the case of Amihud and Lev. Strategic
Management Journal, 26: 367-375.

Boyd, B.K. (1995). CEO duality and ﬁrm performance: A contingency model. Strategic
Management Journal, 16: 301-312.

Brickley, J ., Lease, R.C., & Smith, W. (1988). Ownership structure and voting on
antitakeover amendments. J ourni of Financial Economics, 20: 267-291.

 

179

 

Bromiley, P. (1986). Cogpprate capital investment: A behavioral approach. Cambridge
University Press, Cambridge.

Bromiley, P. (1991). Testing a causal model of corporate risk taking and performance.
Academy of Management Journal, 34: 37-59

Bromiley, P. (2005). The Behavioral Foundations of Strategic Management. Blackwell
Publishing: Malden, MA.

Bruk, A.S., & Raudenbush, SW. (1992). Hierarchical Linear Models. Sage: Newbury
Park, CA.

Brush, T.H. & Bromiley, P. (1997). What does a small corporate effect mean? A variance

components simulation of corporate and business effects. Strategic Management
Journal, 18: 825-835.

Brush, T.H., Bromiley, P., & Hendrickx, M. (1999). The relative inﬂuence of industry
and corporation on business segment performance: An alternative estimate.
Strategic Management Journal, 88: 109-133.

Campa, J .M. (1993). Entry by foreign ﬁrms in the United States under exchange-rate
uncertainty. Review of Economics and Statistics, 4: 614-622.

Campa, J .M., & Kedia, S. (2002). Explaining the diversiﬁcation discount. The Journal pf
Finance, 57: 1731-1762.

Carpenter, M.A., & Sanders, W.G. (2002). Top management team compensation: The

missing link between CEO pay and ﬁrm performance? Strategic Management
Joumal, 23: 367-382.

 

Carter, R., & Manaster, S. (1990). Initial public offerings and underwriter reputation.
Journal of Finance, 45: 1045-1067.

Chandler, A. (1962). Strategy and structure: Chapters in the Histog of American
Industrial Enterprise. MIT Press: Cambridge, MA.

Chan, D. Functional relations among constructs in the same content domain at different

levels of analysis: a typology of composition models. Journal of Applied
Psychology, 83: 234-246.

Child, J. 1972. Organization structure, environment and performance: The role of
strategic choice. Sociology, 6: 1-22.

Coase, R. 1937. The nature of the ﬁrm. Economica, 4: 386-405.

180

Coffee, J.C, Jr. (1988). Shareholders versus managers: The strain in the corporate web. In
J.C. Coffee, Jr., L. Lowenstein & S. Rose-Ackermman (Eds.), Knights, raiders, and

targets: the impact of hostile takeover, 77-134. New York: Oxford University
Press.

Comment, R. & Jarrell, GA. (1993). Comrate focus and stock returns. Rochester, N .Y.:
Univ. Rochester, Bradley Policy Res. Center

Cyert, R.M. & March, J .G, (1963). A behavioral theory of the ﬁrm, Prentice-Hall
(Englewood Cliffs, NJ)

Dempster, A.P., Laird, N.M., & Rubin, DB. (1977). Maximum likelihood from

incomplete data via the EM algorithm. Journal of Royal Statistical Society, Series
B, 39: 1-8.

David, P, Yoshikawa, T. & Chari, M.D.R. (2006). Strategic investments in Japanese
corporations: Do foreign portfolio owners foster underinvestment or appropriate
investment? Strategic Management Journal, 27: 591-600.

Davis, G. (1991). Agents without principles? The spread of the poison pill through the
intercorporate network. Administrative Science Quarterly, 36: 583-613.

Dess, G.G. & Beard, D.W. (1984). Dimensions of organizational task environments.
Administrative Science Quarterly, 19: 52-73.

Devers, C.E, McNamara, G., Wiseman, R.M., & Arrfelt, M. (2008). Moving closer to
the action: Examining compensation design effects on ﬁrm risk. Organization
Science, 19: 548-566.

Dow, J. & Gorton, G. (1997). Stock market efﬁciency and economic efﬁciency: Is there a
connection? The Journal of Finance, 3: 1087-1129.

Duncan, RB. (1972). Characteristics of organizational environments and perceived
environmental uncertainty. Administrative Science Quarterly, 17: 313-327

Eisenhardt, K.M., & Zbaracki, MJ. (1992). Strategic decision making. Strategic
Management Journal, 13: 17-37.

Eliasson, G. (1976). Business economic planning. London: John Wiley & Sons.

Fama, ER (1980). Agency problems and the theory of the ﬁrm. Journal of Political
Economy, 88: 288-321.

Fama, E.F., & Jensen, MC. (1983). Separation of control and ownership. Journal of Law
& Economics, 16: 301-325.

181

FASB (Financial Accounting Standards Board). (1997). Disclosures about Segments of

an Enterprise and Related Information. Statement of Financial Accounting
Standards No. 131. FASB: Norwalk, CT.

Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7:
1 17-140.

Finkelstein, S., & Hambrick, D. (1996). Strategic leadership: Top executives and their
effects on organizations, West Publishing Company, NY.

Folta, T.B., & O’Brien, J .P. (2004). Entry in the presence of dueling options. Strategic
Management Journal, 25: 121-138.

Gibbs, P. (1993). Determinants of corporate restructuring: The relative importance of

corporate governance, takeover threat, and free cash ﬂow. Strategic Management
qu_m_a_l_, 14: 51-68.

Granovetter, M. (1985). Economic action and social structure: the problem of
embeddedness. American Journal of Sociology, 91: 481-510.

Greve, HR. (1998). Performance, aspirations, and risky organizational change.
Administrative Science Quarterly, 43: 58-86.

Goll, I. & Rasheed, MA. (1997). Rational decision-making and ﬁrm performance: The
moderating role of environment. Strategic Management Journal, 18: 583-591.

Gomez-Mejia, L., & Wiseman, R.M. (1997). Reframing executive compensation: An
assessment and outlook. Journal of Management, 23: 291-374.

Hambrick, D.C. & Mason, RA. (1984). Upper echelons: the organization as a reﬂection
of its top managers. Academy of Management Review, 9: 193-206.

Hansen, G. & Hill. C. (1991). Are institutional investors myopic? A time-series study of
four technology-driven industries. Strategic Management Journal, 12: 1-16.

Hansen, G., & Wernerfelt, B. (1989). Determinants of corporation performance: the

relative importance of economic and organizational factors. Strategic Management
Journal, 10: 399-411.

Haveman, HA. (1993). Organizational size and change: Diversiﬁcation in the savings
and loan industry after deregulation. Administrative Science Quarterly, 38: 20-50.

Hay, D.A., & Morris, DJ. (1979). Industrial Economics. Oxford University Press:
Oxford.

Hayek, F. (1945). The use of knowledge in society. American Economic Review, 35:
519-530.

182

Hayward, M.L.L., & Hambrick, DC. (1997). Explaining the premiums paid for large

acquisitions: Evidence of CEO hubris. Administrative Science Quarterly, 39: 103-
127.

Hayward, M.L.A., Rindova, V.P., & Pollock, T.G. (2004). Believing one’s own press:

The causes and consequences of CEO celebrity. Strategic Management Journal, 25:
637-653.

Hayward, M.L.A., & Shimizu, K. (2006). De-commitment to losing strategic action:
Evidence from the divestiture of poorly performing acquisitions. Strategic
Management Journal, 27: 541-557.

Henderson, A.D., Miller, D. & Hambrick, DC. (2006). How quickly to CEOs become

obsolete? Industry dynamism, CEO tenure, and company performance. Strategic
Management Journal, 27: 447-460.

Herfmdahl, CC. (1950). Concentration in the steel indusg, unpublished Ph.D.
dissertation, Columbia University, New York.

Hirshlefer, D., & Thakor, A.V. (1992). Managerial conservatism, project choice, and
debt. The Review of Financial Studies, 5: 437-471.

Hill, C., & Snell, S. (1988). External control, corporate strategy, and ﬁrm performance in
research intensive industries. Strategic Management Journal, 9: 577-590.

Holmstrom, B. (1979). Moral hazard and observability. Bell Journal of Economics, 10:
74-91.

Hoskisson, R.E., Hitt, M.A., Johnson, R.A., & Moesel, DD. (1993). Construct validity of
an objective (entropy) categorical measure of diversiﬁcation strategy. Strategic
Management Journal, 14: 215-235.

Hough, J .R. (2006). Business segment performance redux: A multilevel approach.
Strategic Management Journal, 27: 45-61.

Hough, J .R., & White, MA. (2003). Environmental dynamism and strategic decision-

making rationality: An examination at the decision level. Strategic Management
Journal 24: 481-489.

 

James, LR. (1982). Aggregation bias in estimates of perceptual agreement. Journal of
Applied Psychology, 67: 219-229.

J acquemin, A.P., & Berry, CH. (1979). Entropy measure of diversiﬁcation and corporate
growth. Journal of Industrial Economics, 4: 359-369.

183

Jensen, M.C. & Meckling, W.H. (1976). Theory of the ﬁrm: Managerial behavior, agency
costs and ownership structure. Journal of Financial Economics, 3: 305-360.

Jensen, MC. (1986). Agency costs of free cash ﬂow, corporate ﬁnance and takeovers.
American Economic Review, 67: 61-74.

Jensen, MC. (1989). Eclipse of the public corporation. Harvard Business Review, 68: 61-
74.

Jensen, M.C. (1991). Corporate control and the politics of ﬁnance. Journal of Applied
Corpprate Finance, 4: 13-33.

Jensen, MC. (1993). The modern industrial revolution, exit, and the failure of internal
control systems. Journal of Finance, 48: 831-880.

Kabir, R., Cantrijn, D., & J eunink, A. (1997). Takeover defenses, ownership structure

and stock returns in the Netherlands: An empirical analysis. Strategic Management
Journal, 2: 97-109.

Kahneman, D., & Lovallo, D. (1993). Timid choices and bold forecasts: A cognitive
perspective on risk taking. Management Science, 39: 17-31.

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision making
under risk. Econometrica, 42: 262-291.

Kahneman, D., Slovic, P, & Tversky, A. (1982). Judgment Under Uncertainty: Heuristics
and Biases. Cambridge University Press.

Keats, B.W., & Hitt, MA. (1988). A causal model of linkages among environmental
dimensions, macro organizational characteristics, and performance. Academy of
Management Journal, 31: 570-598.

Kelly, W.A. (1951). A generalized interpretation of the Herﬁndahl index. Southern
Economic Journal, 48: 50-57.

Kroll, M., Wright, P., Toombs, L., & Leavell, H. (1996). Forms of control — A critical

determinant of acquisition performance and CEO rewards. Strategic Management
19u_m_a_l, 2: 85-96. ‘

Kuhberger, A. (1998). The inﬂuence of framing on risky decisions: A meta-analysis.
Organizational Behavior and Human Decision Processes, 1: 23-55.

Lamont, O. (1997). Cash ﬂow and investment: Evidence from internal capital markets.
Journal of Finance, 52: 83-109

184

Lang, L & Stulz, R. (1994). Tobin’s q, corporate diversiﬁcation, and ﬁrm performance.
Journal of Political Economy, 102: 1248-1280.

Lant, T.K. (1992). Aspiration level adaptation: An empirical exploration. Management
Science, 38: 623-644.

Lant, T.K., Milliken, F.J. & Batra, B. (1992). The role of managerial learning and
interpretation in strategic persistence and reorientation: An empirical exploration.
Strategic Management Journal, 13: 585-608.

Lant, T.K., & Montgomery, DB. (1987). Learning from the strategic success and failure.
Journal of Business Research, 15: 503-517.

IeBaron, D., & Speidell, LS. (1987). Why are the parts worth more than the sum? Chop

Shop, a corporate valuation model. In The Merger Boom, edited by Lynn E. Brown
and Eric S. Rosengren. Conference Series, no. 31. Boston, Fed. Reserve Bank.

Leland, H.E., & Pyle, DH. (1977). Informational asymmetries, ﬁnancial structure, and
ﬁnancial interrnediation. Journal of Finance, 32: 371-387.

Levitt, B. & March, J .G. (1988). Organizational learning. Annual Review of Sociology,
14: 319-340.

Li, M & Simerly, R.L. (1998). The moderating effect of environmental dynamism on the

ownership and performance relationship. Strategic Management Journal, 19: 169-
179.

March, J .G. & Shapira, Z. (1992). Managerial perspectives on risk and risk taking.
Management Science, 33: 1404-1418.

March, J .G. & Simon, HA. (1958). Organizations. Wiley, New York.
Maksimovic, V. & Phillips, G. (2002). Do conglomerate ﬁrms allocate resources

inefﬁciently across industries? Theory and evidence. The Journal of Finance, 2:
721-767.

McConnell, J.J., & Servaes, H. (1990). Additional evidence on equity ownership and
corporate value. Journal of Financial Economics, 27: 595-612.

McGahan, A. & Porter, M. (1997). How much does industry matter, really? Strategic
Management Journal, Summer Special Issue, 18: 15-30.

McGahan, A. & Porter, M. (2002). What do we know about variance in accounting
proﬁtability? Management Science, 48: 834-851.

185

McGahan, A.M., & Porter, ME. (2005). Comment on ‘industry, corporate and business-
segment effects and business performance: A non-parametric approach’ by Rueﬂi
and Wiggins. Strategic Management Journal, 26: 873-880.

McNamara, G., Deephouse, D.L., & Luce, R.A. (2003). Competitive positioning within
and across a strategic group structure: the performance of core, secondary and
solitary ﬁrms. Strategic Management Journal, 2: 161-181.

Meyer, M., Milgrom, P, & Roberts, J. (1992). Organizational prospects, inﬂuence costs,
and ownership changes. Journal of Economics and Management Strategy, 1: 9-35.

Miller, J .S., Wiseman, R.M., & Gomez-Mejia, LR. (2002). The ﬁt between
compensation design and ﬁrm risk. Academy of Management Journal, 45: 745-756.

Milliken, F.J. (1987). Three types of uncertainty about the environment: State, effect, and
response uncertainty. Academy of Management Review, 12: 133-143.

Milliken, F.J . (1990). Perceiving and interpreting environmental change: An examination
of college administrators’ interpretation of changing demographics. Academy of
Management Journal, 33: 42-63.

Mikkelson, W. & Ruback, R. (1991). Targeted repurchases and common stock returns.
Rand Journal of Economics, 22: 544-556.

Mikkelson, W. & Ruback, R. (1985). An empirical analysis of the interﬁrm equity
investment process. Journal of Financial Economics, 14: 523-553.

Mintzberg, H. (1990). The design school: Reconsidering the basic premises of strategic
management. Strategic Management Journal, 11: 171-195.

Misangyi, V.F., Elms, H., Greckhamer, T., & LePine, J .A. (2006). A new perspective on
a fundamental debate: A multilevel approach to industry, corporate, and business
unit effects. Strategic Management Journal, 27: 571-590.

Mitchell, T.R., & James, LR. (2001). Building better theory: Time and the speciﬁcation
of when things happen. Academy of Management Review, 26: 530-547.

Montgomery, C.A., & Wemerfelt, B. (1988). Diversiﬁcation, Ricardian rents, and
Tobin’s q. Journal of Economics, 19: 623-632.

Mook, D.G. (1983). In defense of external invalidity. American Psychologist, 21: 379-
387.

Nelson, R.R., &. Winter, S.G. (1982). An Evolutiongy Theog of Economic Change.
Belknap Press/Harvard University Press: Cambridge.

186

Palich, L.E., Cardinal, L.B., & Miller, CC. (2000). Curvilinearity in the diversiﬁcation-
perforrnance linkage: An examination of over three decades of research. Strategic
Management Journal, 21: 155-174.

Palmer, T.B., & Wiseman, R.M. (1999). Decoupling risk taking from income stream

uncertainty: A holistic model of risk. Strategic Management Journal, 11; 1037-
1062.

Pollock, T.G., & Rindova, VP. (2003). Media legitirnation effects in the market for
initial public offerings. Academy of Management Journal, 46: 631-642.

Porter, ME. (1980). Compgtitive Strategy. Free Press, New York.

Pound, J. (1988). Proxy contests and the efﬁciency of shareholder oversight. Journal of
Financial Economics, 10: 53-83.

Rajan, R. Servaes, H. & Zingales, L. (2000). The cost of diversity: The diversiﬁcation
discount and inefﬁcent investment. The Journal of Finance, 1: 35-80.

Rasheed, A.M.A. & Prescott, J .E. (1992). Towards an objective classiﬁcation scheme for
organizational task environments. British Journal of Management, 3: 197-206.

Raudenbush, S.W., Bruk, A.S., Cheong, A.P., & Congdon, R. (2000). HLM5:
Hierarchical Linear and Nonlinear Modeling. Scientiﬁc Software International:
Lincolnwood, IL.

Ray, G., Barney, J.B., & Muhanna, A.W. (2004). Capabilities, business processes, and
competitive advantage: Choosing the dependent variable in empirical tests of the
resource-based view. Strategic Management J oumal, 25: 23-57.

Robins, J .A., & Wiersema, M.F. (2003). The measurement of corporate portfolio
strategy: Analysis of the content validity of related diversiﬁcation indexes.

Strategic Management Journal, 24: 39—59.

Roquebert, J ., Phillips, R. & Westfall, P. (1996). Market versus management: what drives
proﬁtability? Strategic Management Journal, 17: 653-664.

Ross, J ., & Staw, BM. (1993). Organizational escalation and exit: Lessons from the
Shoreham nuclear power plant. Academy of Management Journal, 36: 701-732.

Rumelt, R. (1974.) Strategy, Structure, and Economic Performance. Harvard Business
School Press: Boston, MA.

Rumelt, R. (1991). How much does industry matter? Strategic Management Journal, 12:
167-1 85.

187

Sanders, W.G., & Carpenter, MA. (2003). Strategic satisﬁcing? A behavioral-agency
theory perspective on stock repurchase program announcements. Academy of
Management Journal, 46: 160-178.

Scharfstein, D.S. (1997). The dark side of internal capital markets 11, Working paper,
MIT.

Scharfstein, D.S., & Stein, J.C. (2000). The dark side of internal capital markets:
Divisional rent-seeking and inefﬁcient investment. The Journal of Finance, 6:
2537-2564.

Scharfstein, D.S., & Stein, J .C. (1990). Herd behavior and investment. The American
Economic Review. 80: 465-480.

Scherer, F .M. ( 1980). Industrial MaLket Structure and Economic Performance. Rand
McNally, Chicago.

 

Schmalensee, R. (1985). Do markets differ much? American Economic Review, 75: 341-
35 1.

Schleifer, A., & Vishny, R. (1986). Large shareholders and corporate control. Journal of
Political Economy, 94: 461-488.

Schlingemann, F., Stulz, R. & Walking, R. (1999). Comrate focusing and internal
capital markets. Working paper, Fisher School of Business, Ohio State University,
Columbus.

Servaes, H. (1996). The value of diversiﬁcation during the conglomerate merger wave.
Journal of Finance, 51: 1201-1225.

Shin, H. & Stulz, R. (1998). Are internal capital markets efﬁcient? Quarterly Journal of
Economics, 113: 531-553.

Siggelkow, N ., & Levinthal, DA. (2003). Temporarily divide to conquer: Centralized,
decentralized, and reintegrated organizational approaches to exploration and
adaptation. Organization Science, 6: 650-669.

Siggelkow, N., & Rivkin, J .W. (2005). Speed and search: Designing organizations for
turbulence and complexity. Organization Science, 2: 101-122.

Simerly, R.L. & Li, M. (2000). Environmental dynamism, capital structure and
performance: A theoretical integration and an empirical test. Strategic Management
Journal, 21: 31-49.

Snijders, T.A.B & Bosker, R.J. (1999). Multilevel Analysis: an introduction to basic and
advanced multilevel modeling. Sage: London.

188

Spence, M. (1976). Informational aspects of market structure: an introduction. Quarterly
JournJal of Economics, 90: 591-597.

Staw, BM. (1976.) Knee-deep in the big muddy: A study of escalating commitment to a
chosen course of action. annizational Behavior and Human Performance, 16: 27-
44.

Staw, BM. (1981). The escalation of commitment to a course of action. Academy of
Management Review, 6: 577-587.

Staw, BM. (1997). The escalation of commitment. In Organizational Decision Mak_i_pg,
Shapira Z (ed). Cambridge University Press: New York: 191-215.

Stein, J .C. (1997). Internal capital markets and the competition for corporate resources.
The Journal of Finance, 1: 111-133.

Stulz, R. (1990). Managerial discretion and optimal ﬁnancing policies. Journal of
Financial Economics, 26: 3-27.

Thomas 111, LG. & Waring, Geoffrey. (1999). Competing capitalisms: Capital
investment in American, German, and Japanese ﬁrms. Strategic Management
Journal, 20: 729-748.

Villalonga, B. (1999). Does diversiﬁcation cause the ‘diversiﬁcation discount’?
Unpublished working paper, University of California, Los Angeles.

Von Krogh, G., Roos, J. & Slocum, K. (1994). An essay on corporate epistemology.
Strategic Management Journal, 15: 53-71.

Weick, K.E. (1979). The social psychology of organizing (2nd ed.). Addison-Wesley,
Reading, MA.

Westphal, J .D. (1998). Board games: How CEOs adapt to increases in structural board
independence from management. Administrative Science Ouarterly, 43: 511-537.

Wiersema, M.F. & Bantel, K.A. (1993). Top management team turnover as an adaptation
mechanism: The role of the environment. Strategic Management Journal, 14: 485-
504.

Williamson, O.E. (1975). Markets and hierarchies: Analysis and antitrust implications.
New York: Free Press.

Williamson, CE. (1985). The Economic Institutions of Capitalism. Free Press: New
York.

189

Williamson, QB. (1986). Economic Organization: Firms, Markets and PolicyLControl.
New York University Press, New York.

 

Wiseman, R.M. Making researchers out of monkeys: On the misuse of ratios in
management research. Working paper.

Wiseman, R.M., & Gomez-Mejia, L. (1998). A behavioral agency model of managerial
risk taking. Academy of Mﬂagement Journﬂ, 23: 133-153.

Wright, P., Ferris, S.P., Atulya, S., & Awasthi, V. (1996). Impact of corporate insider,
blockholder, and institutional equity ownership on firm risk taking. Academy of
Management Journal, 2: 441-463.

Zajac, E. & Westphal, J .D. (1994). The Costs and Beneﬁts of Managerial Incentives and
Monitoring in Large US. Corporations: When is More not Better? Strategic
Management Journal, 15: 121-142.

Zuckerman, E.W. (1999). The categorical imperative: securities analysts and the
illegitimacy discount. American Journal of Sociology, 104: 1398-1438.

190

 
        

     

  

STATE UN IV

llllllllllllll ll lMilllllHIIHHIIHllllllll

12