“tn-.53.. .3.

51... cm“
L _ i.
as ..
3.} (ll. .
{5:1}... .55.. a”.
7 25-:- 9, a.
1:125: .12:
v:{:!.:?”.
z. n. 23:"... A.
a : A... 9.4.
$3.63.... .
51.533...

I. 3 )1. n.

a... A y! .

tiffrgi.
II)»

\2 $1..
1 ) Q.- ’-

 

1 at... {i\d...1:.r:o|lixo.
)3}. 914‘ I.
3.11:... A
x. 3“!

3).. is!!!
a, 912:} .1 (.1). 7'31:-
(.3!

 

 

 

 

é

 

 

This is to certify that the

dissertation entitled

Corporate Governance in Regulated and Unregulated
Industries

presented by

Michael Samuel Cichello

has been accepted towards fulﬁllment
of the requirements for

Ph . D . degree in Finance

6% /. Aéd/«k

Maﬁr professor

Date ‘75”: ’6. 2000

MS U i: an Afﬁrmative Action/Equal Opportunity Institution 0-12771

 

 

LIBRARY

Michigan State

University

 

 

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

 

DATE DUE

DATE DUE

DATE DUE

 

5%‘3116 002902

 

 

 

 

 

 

 

 

 

 

 

 

 

 

11m common-peso.“

 

 

CORPORATE GOVERNANCE IN REGULATED AND UNREGULATED
INDUSTRIES

By

Michael Samuel Cichello

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Finance

2000

ABSTRACT

CORPORATE GOVERNANCE IN REGULATED AND UNREGULATED
INDUSTRIES

By

Michael Samuel Cichello

This dissertation contains three chapters that address issues in the area of
corporate governance. The ﬁrst chapter examines the use of ﬁnancial contracts by Silver
King Communications to control two downstream ﬁrms—Urban Communications and
Jovon Communications—who were afﬁliated with Silver King. These cases illustrate the
weakness of the prevailing focus on the ownership of equity with voting rights to
determine the locus of corporate control. In doing so, Chapter 1 provides insights into
the use of ﬁnancial contracting to avoid regulation, to assign control rights, and to
address costs associated with vertical relationships.

The second chapter examines the impact of the passage of the Energy Act in 1992
on the structure of pay for CEOs in the electric utility industry. Previous work by
Joskow, Rose and Wolfram (1996) and Joskow, Rose and Shepard (1993) has
documented both lower pay levels and lower sensitivities of pay to performance for
CEOs of regulated companies versus CEOs of unregulated companies. Using a sample of
228 CEOs from 1988 to 1998, this chapter conﬁrms ﬁndings by Kole and Lehn (1999)
that deregulation alters several facets of the corporate governance structure. Speciﬁcally,
the percentage of compensation from relatively ﬁxed components—salary and bonus—
decreases after the onset of deregulation, while the percentage of compensation from “at

risk” components—stocks and options—increases. Additionally, total yearly

compensation becomes more sensitive to the performance of the ﬁrm. It is also shown
that very sizable changes in the value of option and stock holdings of CEOs occur after
passage of the Energy Act. This effect cannot be solely attributed to the overall bull
market, as share ownership percentages of CEOs double to quadruple using various
measures of ownership.

Finally, the third chapter examines the methodology used in estimating pay-
performance sensitivities (PPSs). Previous work by Aggarwal and Samwick (1999) has
highlighted the importance of controlling for the variance of ﬁrms stock returns when
estimating PPSs. These authors estimate PPSs that are an order of magnitude greater for
ﬁrms of smaller stock return variances than for ﬁrms of larger variances. Using a
comparable sample of CEOs and non-CEO executives, I ﬁnd that when properly
controlling for ﬁrm size, the negative effect of the variance in stock returns on estimated
PPSs is greatly diminished for both CEOs and non-CEOs. In particular, when using
dollar returns as the measure of ﬁrm performance, it is crucial to properly control for ﬁrm
size. Regressions that use percentage returns as the measure of ﬁrm performance are not
as severely affected by this phenomenon. However, evidence shows that controls should

still be included for these regressions as well.

Dedicated, with love and gratitude, to my parents

iv

ACKNOWLEDGMENTS

I would like to express my appreciation to the members of my dissertation
committee—Dr. Charles Hadlock, Dr. Ted Fee, Dr. Michael Mazzeo, and Dr. Wallace
Mullin for their insightful guidance and support. I especially thank Charlie for serving as
my committee chairman and guiding me through the completion of my doctoral program.
His continuous support and encouragement have been invaluable. I also thank Dr. Robert
Kieschnick who provided insights for my research and encouragement for completion of
my degree.

I would also like to thank Dr. Richard Simonds, Dr. Kirt Butler, and Dr. Geoffrey
Booth for providing me with the opportunity to obtain my doctorate. Additionally, I
thank the Economics Department for their assistance. I thank fellow graduate students in
the ﬁnance and economic departments, especially Melinda Newman, Wei-ling Song,

Dan Hansen, Kathleen Beegle, Scott Baier, Jen Tracey, Paul Loetscher, Cece Howell, and
Jason Nanton for all of their help and support. I particularly thank Ramana Sonti, whose
assistance and friendship were crucial to the completion of my degree. I also thank all of
my friends, especially Elizabeth Ransom, Matt Kleiman, and the “swimming group” who
provided constant encouragement throughout my time in East Lansing.

I acknowledge ﬁnancial assistance from the Institute for Public Utilities at
Michigan State University. In particular, I thank Michelle Wilsey for her assistance in
securing a position at the Federal Communications Commission for ﬁeld work. I also
thank the staff members of the Federal Communications Commission for their insights

and support.

Finally, I would like to thank all of my family for their love and support. In
particular, I thank my parents for their continuous encouragement, and my brother Paul,
who discussed my research with me for countless hours and provided much needed

support in the later phases of my doctoral program.

vi

TABLE OF CONTENTS

LIST OF TABLES .................................................................................. ix
INTRODUCTION ................................................................................... 1

CHAPTER 1 THE USE OF FINANCIAL CONTRACTING TO CONTROL
DOWNSTREAM FIRMS: AN ANALYSIS OF SILVER KING

COMMUNICATIONS ......................................................... 6
Introduction ....................................................................... 6
A Brief Description of the Broadcast TV Industry and Its
Regulation .......................................................................... 9
1.3 Silver King Communications’ Use of Financing ........................... 13
1.3.1 Silver King Communications .......................................... 14
1.3.2 Urban Broadcasting Corporation ..................................... 16
1.3.3 Jovon Broadcasting Corporation ..................................... 20
1.3.4 Post-script ................................................................ 23
1.4 Analysis of Silver King Communications’ Use of Financing ............. 24
1.5 Summary ......................................................................... 27

CHAPTER 2 THE IMPACT OF DEREGULATION ON THE STRUCTURE AND
PAY-PERFORMANCE SENSITIVITY OF EXECUTIVE
COMPENSATION: AN ANALYSIS OF THE ELECTRIC UTILITY

INDUSTRY ..................................................................... 31

2.1 Introduction ..................................................................... 3 l

2.2 Relevant Literature and Hypothesis Development .......................... 34

2.3 Data and Methodology ......................................................... 36

2.4 Results ............................................................................ 43

2.4.1 Electric Utilities ........................................................ 43

2.4.2 Unregulated Companies ............................................... 52

2.5 Conclusion ........................................................................ 56
CHAPTER 3 A REEXAMINATION OF CROSS-SECTIONAL VARIATION IN

PAY-PERFORMANCE SENSITIVITIES .................................. 59

3. 1 Introduction ................................................... ' ................... 59

3.2 Relevant Literature and Hypothesis Development ......................... 61

3.3 Data and Methodology ......................................................... 64

3.4 Results ............................................................................ 72

3.5 Conclusion ....................................................................... 84

SUMMARY ......................................................................................... 85

vii

APPENDICES ...................................................................................... 87
Appendix A Tables for Chapter 2: The Impact of Deregulation on the
Structure and Pay-Performance Sensitivity of Executive
Compensation: An Analysis of the Electric Utility Industry. .. 88
Appendix B Tables for Chapter 3: A Reexamination of the Cross-
Sectional Variation in Pay-Performance Sensitivities ............ 97

REFERENCES .................................................................................... 120

viii

Table A1

Table A2

Table A3

Table A4

Table A5

Table A6

Table A7

Table Bl

Table BZ

Table B3

Table B4

Table B5

Table B6

Table B7

Table B8

LIST OF TABLES

Firm and CEO Characteristics ....................................................... 89
CEO Compensation Summary Statistics .......................................... 9O

OLS Estimates of CEO Pay-Performance Elasticities Using Salary

and Bonus for Electric Utilities (1988-1998) ...................................... 92
OLS Estimates of CEO Pay-Performance Elasticities Using Flow

for Electric Utilities (1988-1998) ................................................... 93
CEO Age and Turnover Variables ................................................. 94
OLS Estimates of CEO Pay-Performance Elasticities Using Salary

and Bonus for Unregulated Firms (1992-1997) .................................. 95
OLS Estimates of CEO Pay-Performance Elasticities Using Flow

for Unregulated Firms (1992-1997) ................................................ 96
Executive Compensation and Ownership Statistics, 1995 ....................... 98
Distributions of Firm Characteristics, 1993-1998 ................................ 99

Median Regression Estimates of Pay-Performance Sensitivities
for Measures of Firm-Speciﬁc Wealth (Dollar Returns),
1993-1998 ............................................................................ 100

Median Regression Estimates of Pay-Performance Sensitivities
for Measures of Firm-Speciﬁc Wealth (Percentage Returns),
1993-1998 ............................................................................ 103

Median Regression Estimates of Pay-Performance Sensitivities
for Measures of Flow Compensation (Dollar Returns), 1993-1998 .......... 105

Median Regression Estimates of Pay-Performance Sensitivities
for Measures of Flow Compensation (Percentage Returns),
1993-1998 ............................................................................ 108

OLS Estimates of Pay-Performance Sensitivities for Measures of
F inn-Speciﬁc Wealth (Dollar Returns), 1993-1998 ............................ 110

OLS Estimates of Pay-Performance Sensitivities for Measures of
F inn-Speciﬁc Wealth (Percentage Returns), 1993-1998 ...................... 113

ix

Table B9 OLS Estimates of Pay-Performance Sensitivities for Measures of
Flow Compensation (Dollar Returns), 1993-1998 .............................. 115

Table BIO OLS Estimates of Pay-Performance Sensitivities for Measures of
Flow Compensation (Percentage Returns), 1993-1998 ......................... 118

INTRODUCTION

This dissertation contains three chapters that address issues in the area of
corporate governance. The ﬁrst chapter examines the use of ﬁnancial contracts by Silver
King Communications to control two downstream ﬁrms—Urban Communications and
Jovon Communications—who were afﬁliated with Silver King. The notion that control
resides in the hands of voting shareholders is central to the way that many people think
about corporate control issues. In corporate law, arguments are given for why corporate
control rights are typically assigned to voting common shareholders, the residual
claimants on the ﬁrm.l This characterization has served to focus much of the empirical
research on corporate control.2

Chapter 1 illustrates that corporate control does not always reside in the hands of
voting common shareholders. Speciﬁcally, my study focuses on two cases to illustrate
the use of options in ﬁnancial contracts, particularly in conjunction with debt contracts, to
shift control of a corporation from shareholders to others. I show that Silver King
Communications was able, despite holding no voting securities in these entities, to use
options in ﬁnancial contracts to obtain control over these corporations, and to achieve
some of the beneﬁts of vertical integration that were blocked by regulation.
Consequently, I also provide evidence that is consistent with Cheung’s (1983) argument

that regulation will inﬂuence contract design.

 

' See for example, Chapter 3 of Easterbrook and Fischel (1991).
2 See Weston, Chung, and Hoag (1990) for a survey of empirical research on corporate
control.

The second chapter examines the impact of the passage of the Energy Act in 1992
on the structure of pay for CEOs in the electric utility industry. Historically, CEOs of
regulated companies have seen signiﬁcantly lower pay than their counterparts at
unregulated companies. Academic researchers have also found that CEOs of regulated
ﬁrms experience sensitivities of pay to performance that are signiﬁcantly lower than
unregulated company CEOs.3 However, sincesubstantial deregulation has occurrediand
continues to occur in the regulated sector of the economy, it is unclear whether these
differences are still present. Additionally, this deregulation provides an interesting
setting to study the relationship between regulatory environment and corporate
governance structure. Furthermore, assessing the effect of changes in regulation on the
contractual arrangements between managers and ﬁrms may allow us to better understand
the effects of deregulation on corporate performance.

The electric utility industry is a useful industry to study these changes for several
reasons. First, the sheer size of the industry warrants further investigation. U.S. electric
utilities had total revenues of almost $300 billion in 1998. Second, the industry has
relatively homogenous ﬁrms in the sense that most ﬁrms use similar technologies to
produce similar services. As Parrino (1997) notes, these types of industries yield
competitive labor markets because it is relatively easy to evaluate performance of CEOs,
and human capital is readily transferable across ﬁrms. Finally, an important deﬁning
event in this industry—the passage of the Energy Act in 1992—issued a strong signal to .

both the market and to the ﬁrms in the industry that deregulation was becoming a reality.

 

3 See Joskow, Rose and Wolfram (1996) and Joskow, Rose and Shepard (1993) for
details.

Similar to Kole and Lehn’s (1999) work on the airline industry, 1 document a
strong shift in certain corporate governance characteristics. The median percent of total
yearly compensation composed of salary and bonus decreases dramatically, from 89% in
1992 to 62% in 1998. This result is in marked contrast to results for a sample of
unregulated company CEOs, where there is very little change for the median CEO over
this period. CEOs of electric utilities also now receive options at a much higher rate,
with over half of CEOs in 1998 receiving options, compared to only 26% in 1992.
Additionally, the change in the value of CEOs’ option holdings and stock holdings
increased dramatically. For those receiving options, median size of holdings went from
$27,000 in 1988 to $348,000 in 1998. Likewise, median stock holdings of CEOs has
increased from $188,152 in 1988 to $1,643,234 in 1998. Results for the mean CEO are
even more dramatic. These results are not solely driven by the bull market of the 19905,
as median (mean) ownership by CEOs increased from 0.02% (0.07%) in 1992 to 0.05%
(0.13%) in 1998. When exercisable options are included in the analysis, the numbers go
from 0.025% (0.106%) in 1988 to 0.083% (0.225%) in 1998.

Meanwhile, tenure and turnover of CEOs have changed as well. Tenure as CEO
decreased from 5 to 4 years for the median CEO from 1988 to 1998. At the same time,
the median age for CEOs decreased from 57 in 1992 to 55 in 1998. More CEOs were
hired from outside the ﬁrm, with 17% of electric utility CEOs being outsiders in 1992
and 29% in 1998.

Finally, Chapter 3 examines the methodology used in estimating pay-performance
sensitivities. The study of the relationship between compensation of high-ranking

employees and the performance of their respective companies is of central importance to

the theory of the ﬁrm. Although academics have long studied this relationship, there has
been a relative explosion of such studies in the late 19805 and particularly in the 19905.4
This is due in large part to a few factors. First, public scrutiny of escalating executive
compensation appears to have reached a high point in the early 19905. Second, academic
researchers have been able to obtain much more detailed data on executive compensation
due to both increased effort and to enhanced reporting requirements by the Securities and
Exchange Commissions With these richer data sets now available, better proxies for
compensation variables of interest could be used to better test existing theory.

At the same time, researchers have been trying to improve the methodology of
studying the pay-performance relationship in order to better understand managerial
incentives. Recently, Aggarwal and Samwick (1999) have made such an improvement,
emphasizing the importance of properly controlling for cross-sectional differences when
studying the pay-performance relationship. Using a standard principal-agent model as
their motivation, they highlight the importance of controlling for the variability of ﬁrm
stock returns in pay-performance speciﬁcations. Their results appear to be quite
dramatic, providing evidence that ﬁrms with more variability in their returns have much
lower pay-performance sensitivities (PPSs) than comparable low variance ﬁrms. In some
regression results, they ﬁnd that a ﬁrm with median sample variance has a PPS over 10

times that of a maximum sample variance ﬁrm.

 

4 See Murphy (1999) for details on the rapid increase in the number of academic articles
published in this area in the 19905.

Beginning with ﬁscal year 1993, companies were required to annually report individual
salary, bonus, other annual compensation, restricted stock grants, long term incentive
payouts, option grants, and all other compensation for the top ﬁve paid executives.
Because companies were required to report data for the previous three years, detailed and

Chapter 3 provides evidence that this effect is in large part due to ﬁrm size.
Using a comparable sample, I show that it is important to control for not only the
variability in returns, but also for the size of the ﬁrm when estimating pay-performance
sensitivities. This is particularly important for regressions that use dollar returns as the
measure of ﬁrm performance. My results show that when examining comparably-sized
ﬁrms, the PPS for high variance ﬁrms is still less than that of low variance ﬁrms.
However, I ﬁnd that there is oftentimes an equal or greater “size effect” than “variance
effect.” In fact, when properly controlling for ﬁrm size, the difference between high and

low variance ﬁrms is negligible under some speciﬁcations.

 

reliable data is available for much of the 19905. Additional details on option holdings for
these same executives was also required, which had not previously been required.

CHAPTER 1

THE USE OF FINANCIAL CONTRACTING TO CONTROL DOWNSTREAM
FIRMS: AN ANALYSIS OF SILVER KING COMMUNICATIONS

1.1 Introduction

The notion that control resides in the hands of voting shareholders is central to the
way that many people think about corporate control issues. In corporate law, arguments
are given for why corporate control rights are typically assigned to voting common
shareholders, the residual claimants on the ﬁrm.6 These voting rights typically allow the
holder to vote on major corporate decisions, such as who shall belong to the board of
directors. Under corporate law, the board of directors is the corporate body that has'the
power to hire, monitor, ﬁre or reward management of the ﬁrm. In turn, management
exercises operational control over the ﬁrm, directing day-to-day business operations and
ﬁnancing. This characterization has served to focus much of the empirical research on
corporate control.7

Nevertheless, E. Fama once argued that:

"...ownership of capital should not be confused with ownership of the
ﬁrm. Each factor in a ﬁrm is owned by somebody. The ﬁrm is just the set of
contracts covering the way inputs are joined to create outputs and the way receipts
from outputs are shared among inputs. In this “nexus of contracts” perspective,

ownership of the ﬁrm is an irrelevant concept. Dispelling the tenacious notion
that a ﬁrm is owned by its security holders is important because it is a ﬁrst step

 

6 See for example, Chapter 3 of Easterbrook and. Fischel (1991).
7 See Weston, Chung, and Hoag (1990) for a survey of empirical research on corporate
control.

toward understanding that control over a ﬁrm’s decisions is not necessarily the
province of security holders."8

While not going as far as F ama, my study does illustrate that corporate control
does not always reside in the hands of voting common shareholders. Speciﬁcally, my
study focuses on two cases to illustrate the use of options in ﬁnancial contracts,
particularly in conjunction with debt contracts, to shift control of a corporation from
shareholders to others.

The usage of options embedded in debt contracts to transfer control rights is not a
new theme in the corporate governance literature. Since Smith and Warner (1979),
ﬁnancial economists have studied the use of covenants to confer control rights on
bondholders so as to protect bondholders’ interests when they conﬂict with stockholders’
interests. Through the use of restrictive bond covenants, bondholders must typically
approve any major ﬁnancing and/or investment decision. Further, as a result of the
innovative structuring of leveraged buyouts (LBOs), some have even questioned the

distinctions between debt and equity in corporate law.9

For example, Jensen (1989) has
noted that in the case of LBOs, new bondholders of the highly levered ﬁrm are granted
powers giving them substantial control over the ﬁrrn’s operations.

Nevertheless, this literature has primarily focused on the use of covenants or

embedded options in debt contracts to resolve conﬂicts between security holders in a

ﬁrm. One missing element in this research involves the issues that arise when the

 

8 Fama (1980) at page 289.

9 See papers presented at the Federal Reserve Bank of Boston’s conference that were
published in Kopcke and Rosengren (1991), which make the point that taxes and
regulation have motivated a number of innovations in the design of securities. Our case
study focuses upon how regulation motivates such innovation.

investor or security holder is also a competitor or a supplier. Clearly the incentives of
such investors are different from investors that are focused only on the direct ﬁnancial
returns from investing in the ﬁrm. Such investors will likely have an interest in
inﬂuencing the competitive behavior of the entities in whom they invest. A number of
researchers have begun to examine these possibilities.10 However, consistent with the
corporate control literature, this new research has focused on effects of a competitor or a
supplier taking an equity position in the target ﬁrm.

In contrast, I focus on the effects of a supplier using other types of ﬁnancial
contracts to inﬂuence the behavior of an outlet. Speciﬁcally, I examine Silver King
Communications’ provision of ﬁnancing to two of its outlets, Urban Broadcasting
Corporation and Jovon Broadcasting Corporation. Regulation of the broadcast television
industry provides signiﬁcant incentives for private parties to use ﬁnancing, particularly
non-equity ﬁnancing, to reallocate control rights. I show that Silver King
Communications was able, despite holding no voting securities in these entities, to use
options in ﬁnancial contracts to obtain control over these corporations, and to achieve
some of the beneﬁts of vertical integration that were blocked by regulation.
Consequently, I also provide evidence that is consistent with Cheung’s (1983) argument
that regulation will inﬂuence contract design.

To present my evidence and arguments, I organize the chapter as follows. Section
1.2 provides sufﬁcient background information about the broadcast television industry
and its regulation to illuminate the motivation of some ﬁrms in this industry to use

ﬁnancing to transfer control in non-traditional ways. Section 1.3 describes the cases

 

'0 See Allen and Phillips (1998) and Gilo (1996) for examples of such research.

under study: the ﬁrms, their ﬁnancial arrangements, and the evidence on the extent to
which control rights were given to Silver King Communications. Section 1.4 discusses

the implications of the cases, and section 1.5 concludes the chapter.

1.2 A Brief Description of the Broadcast TV Industry and Its Regulation

The broadcast television industry can be described as comprising three levels of
activities - the production of video programming, it’s distribution to media outlets, and
its delivery to viewers. Some ﬁrms engage in all three activities, others in just one
activity. Since the infancy of television, broadcast television networks (“networks”) have
been key players in this industry. Networks aggregate video programs into program
packages that they distribute to broadcast television stations (“stations”), that in turn
deliver the programming to viewers who can receive their signal.

Because of public concerns over media concentration, the broadcast television
industry has been and continues to be heavily regulated.'1 Many of the regulations were
primarily intended to limit the markets that networks could compete in and their extent of
vertical integration. For example, until recently, broadcast television networks were
limited in the ﬁnancial interests they could obtain in the programming they distributed.12
Elimination of these particular regulations made it feasible for the Walt Disney Company
(Disney) to buy Capital Cities ABC Inc. (ABC), and so allowed ﬁrms to begin to

integrate the production, distribution, and delivery of video programming.

 

“ See Chapters 2 and 13 of Head, Sterling, and Schoﬁeld (1994) for further discussion of
this point and the history of broadcast regulation.
'2 See FCC 95-382, Report & Order, Federal Communications Commission (1995).

Three sets of remaining rules are particularly relevant to the focus of my study—
rules limiting the number of broadcast television stations that a single entity can “own”,
rules deﬁning “ownership,” and rules regulating the contractual relationships between
networks and their “afﬁliates.” Afﬁliates are independently owned broadcast television
stations that contract with broadcast television networks to broadcast the network’s
programming.

The ownership rules prevent a network from owning as many broadcast television
stations as it might like, either within a local market or across the nation. Until August
1999, one rule limited a ﬁrm to owning one station per local market.13 A second rule
limits the number of stations that a single entity can own across the nation, or across

14

different “local” markets. This second rule is more important to a network as it forces

a network to afﬁliate with separately owned stations in different markets in order to

achieve the economies of scale possible in program distribution.15
While the local and national ownership rules limit the number of television

stations that an entity can “own”, they leave open the question of what constitutes

“ownership”, or in legal terms, what constitutes an “attributable interest.” Attribution

rules seek to identify those relationships with a licensee that would confer sufﬁcient

 

'3 On August 6, 1999 (FCC 99-209), the FCC modiﬁed this local ownership rule to allow
a single entity to own up to two broadcast television station in the same local market,
subject to various conditions. The FCC also changed the deﬁnition of a ‘local’ market
from one based on signal contours, to one based upon Designated Market Areas (DMAs).
DMAs are A.C. Nielsen’s delineation of local television markets. A.C. Nielsen serves
advertisers by measuring the viewership of television programming. Nevertheless,
during the period of our study this regulation limited entities to owning a single station in
a local market. . p

'4 On August 5, 1999 (FCC 99-208), the FCC modiﬁed the way it calculated audience
reach under this rule, but retained the rule.

10

inﬂuence on an entity such that it might have “a realistic potential to affect” the
programming and other decisions of the licensee.‘6 Following corporate law, the Federal
Communications Commission (“FCC”) focuses on voting common stock and considers
an entity to have an “attributable interest” in a television station if the entity owns ﬁve
percent or more of the voting stock of the licensee of the station. ’7 An important
exception to this benchmark is when there is one entity that owns more than 50% of the
voting common stock. In this case, other entities can hold more than 5% of the voting
common stock and not have an attributable interest in the licensee.

Finally, the national ownership rule raises questions about whether or not the
contractual relationships between a network and its independently owned afﬁliates should
be regulated since the motivating concern for the rule was to limit the media control that
a single entity could exercise. As a result of these concerns, there are a number of rules
regulating the contractual relationship between a network and its afﬁliate. Three of these
“network/afﬁliate” rules warrant particular attention. One rule prevents a network from
having an option on an afﬁliate’s broadcast time. Another rule prevents a network from
contracting with a station to be its exclusive outlet. The third and most important rule

gives an afﬁliate an unconditional right to reject its network’s programming.l8 Together,

 

'5 See Owen and Wildman (1992) for further discussion of the economics of the
broadcast television industry.

'6 See Attribution of Ownership Interests, 97 FCC 2'”, Federal Communications
Commission (1984). .

'7 See 47 Code of Federal Regulations (C.F.R.) §73.3615. On August 5, 1999 (FCC 99-
207), the FCC modiﬁed this rule, partly in response to cases as reported in this chapter.
’8 See 47 CPR. §73.658 (e). The right to reject rule, described therein gives an afﬁliate
the right to reject network programming in order to better serve the public interest. But
since the afﬁliate determines what serves the “public interest”, the licensee has
effectively an unconditional right to reject network programming.

11

these rules shift the risks associated with introducing new programming onto the
network, while increasing the free-riding and hold-up costs borne by networks.19

Independently owned afﬁliates are analogous to independently owned franchise
outlets in some attributes, and analogous to independently owned retail outlets in other
attributes. Like franchise outlets, broadcast television stations afﬁliated with a network
typically agree to certain operating procedures and to use the network’s brand in
identifying the station.20 However, unlike a franchise outlet, but more like a retail outlet,
affiliates will broadcast the programming of other video program distributors and their
own self-produced programming (e.g., local news).

Regardless of whether viewed as franchise outlets or as retail outlets, the
broadcast network/television station relationship has characteristics that motivate a
network to own rather than afﬁliate with its outlets. For instance, an independently
owned afﬁliate sometimes ﬁnds it more proﬁtable to not broadcast network programming
in order to show other programming. When it does so, it imposes costs on the network
through reductions in the advertising revenue (due to fewer viewer exposures), while
free-riding on the beneﬁts of its network’s investments in brand identity and
programming. For this and other reasons, a broadcast network will prefer to own rather
than afﬁliate with many of its broadcast television station outlets, whether those stations

are viewed as franchise outlets or as retail outlets. An important indication of this

 

'9 See Besen and Krattenmaker (1984) for further analysis of these rules and their costs.
20 See Lafontaine and Masten (1995) for further discussion of the attributes of a
franchise relationship.

preference is the fact that broadcast networks have repeatedly attempted over the years to
relax or eliminate the FCC ’5 national ownership rule.2|

In summary, networks have tremendous incentives to own rather than afﬁliate
with their distribution outlets. Further, FCC regulations not only limit a network’s ability
to own its outlets, but also exacerbate the costs of its relations by limiting the contracts
that a network can write with an afﬁliate. 22 For these reasons, networks have incentives
to ﬁnd other contractual methods of either reducing these costs or obtaining control over

their afﬁliates.
1.3 Silver King Communications’ Use of Financing

The focus of this chapter is on Silver King Communications’ relationships with
Urban Broadcasting Corporation (“Urban”) and Jovon (“Jovon”) Broadcasting
Corporation. These relationships came to the FCC’s attention when Roy Speer sought to
transfer control of Silver King Communications to Silver King Management, a company
controlled by Barry Diller. In response, Urban ﬁled a petition to deny the transfer,
arguing that Silver King Communications had unduly exercised control over it. At this
time, Urban was involved in bankruptcy proceedings, having been forced into bankruptcy

by Silver King in July 1995. Simultaneously, Urban was trying to force a restructuring of

 

2' See McClellan and Albinisk (1999) for a discussion of how the conﬂicts between
networks and their afﬁliates over the national ownership rule has recently grown more
virulent.

22 It is obvious that simply owning some of their afﬁliates’ stock reduces some of the
costs identiﬁed above by allowing the network to beneﬁt from their afﬁliates’
opportunistic behavior. However, existing attribution rules limit the extent to which
networks can use an equity investment to reduce these costs.

13

Silver King’s claims on it, invoking in its petition a ﬁling by Jovon. Jovon was currently
ﬁghting with Silver King Communications over the use of it’s broadcasting time and
claimed similar contractual relationships with Silver King Communications. As a result
of these ﬁlings, the FCC investigated Silver King Communications’ relationship with
both Urban and Jovon.

I use the public records that resulted from these investigations and Security &
Exchange Commission (SEC) ﬁlings to develop my discussion. 23 I try to reduce a large
volume of complex legal and ﬁnancial material to its essentials with respect to the issues
of concern in this chapter. Before discussing these relationships, I provide some

background information about Silver King Communications.

1.3.] Silver King Communications

Silver King Broadcasting Co., Inc., later to become Silver King Communications,
Inc. (“Silver King”) was incorporated in July 1986 as a wholly-owned subsidiary of
Home Shopping Network, Inc. (“HSN”). As part of a strategy to increase viewership of
programming produced by Home Shopping Club, Inc. (“HSC”), another wholly-owned
subsidiary of HSN, Silver King began acquiring UHF television stations. HSC sold
various consumer goods and services through its live, customer-interactive retail sales
programming. This programming was delivered to broadcast television stations, cable

television systems, and satellite dish receivers on a full and part time basis.

 

23 Details of Silver King’s contractual relationships with Urban and Jovon were obtained
from ﬁles in FCC Docket numbers 96-258 and 96-89. General information on the
companies was obtained from SEC ﬁlings.

By 1990, Silver King and its wholly-owned subsidiaries owned and operated 12
independent full-power UHF television stations, including one satellite television station.
These stations afﬁliated with and broadcast HSC retail sales programming. Further, they
served 10 of the largest metropolitan television markets in the United States and reached
an audience of 29 million television households. Thus they had, at that time, one of the
largest potential audiences of any owned and operated independent television
broadcasting group in the United States.

Despite the large audience achieved through ownership of television stations,
Silver King was interested in establishing a more comprehensive national distribution
chain for the Home Shopping Club programming. However, it was limited in expanding
its outlets and its relationships with its outlets by the aforementioned FCC rules. At the
time, the national ownership limit was 12 television stations. In other words, an entity
could have an “attributable interest” in no more than 12 television stations. Given that
the US. comprises 211 Designated Marketing Areas (“DMAs”), Silver King needed to
afﬁliate with many more broadcast television licensees in order to achieve a complete
national distribution chain.

In December 1992, Silver King was spun off from HSN. The contracts that are
described below were entered into when Silver King was a wholly-owned subsidiary of
HSN. Any contracts entered into with Silver King prior to its spin-off from HSN were to
be honored by Silver King after the spin-off. Despite being separated from its sister
company HSC, which produced the Home Shopping programming, Silver King remained

tied to HSN, and, thus, to HSC, through large equity blockholders of both companies.

15

With the above background on Silver King and the FCC rules in play, I now turn to
consider Silver King’s relationships with two companies that Silver King attempted to
include in its distribution chain, and that ultimately fought Silver King for control by

appealing to the FCC to intervene.

1.3.2 Urban Broadcasting Corporation

Urban Broadcasting Corporation was formed on April 18, 1989, with a mission of
constructing and operating a new television station, WTMW(TV), in Arlington, Virginia.
The principal player in this endeavor was Theodore White, who had been granted the
television station construction permit by the FCC. White had two major needs in
bringing his station on the air—ﬁnancial capital and expertise in the construction of a
television station. These were provided by Silver King Virginia, Inc. (“Silver King
Virginia”), a subsidiary of Silver King.24 As previously noted, Silver King owned and
operated 12 television stations by this time, thus developing considerable expertise in the
construction and operation of broadcast television stations. Silver King also had ample
capital to fund this construction.

Silver King beneﬁted from this arrangement in a number of ways. In addition to
the ﬁnancial return on its invested capital, Silver King also achieved another outlet for

Home Shopping Club programming. Given that Silver King had reached the FCC-

 

24 Technically, at the time of Urban’s incorporation, Silver King of Virginia was named
HSN Broadcasting of Virginia, and Silver King was named Home Shopping Network.
To avoid confusion, this chapter will consistently refer to these entities as Silver King of
Virginia and Silver King.

imposed limit of controlling 12 television stations, WTMW(TV) provided an outlet for
Home Shopping Club programming in the Washington, DC. area — a major DMA.

On March 22, 1990, Urban and HSC ﬁled an afﬁliation agreement with the FCC. This
agreement called for Urban to provide 24 hours per day of Home Shopping Club network
programming, Monday through Saturday, and 20 hours per day of such network
programming on Sunday. In addition to the four hours of non-network programming on
Sunday, the agreement allowed for ﬁve minutes per hour of local programming and
commercials -- for an additional two hours per day each day.

Under Urban’s incorporation, the company was composed of two shareholders--
Theodore White and Silver King of Virginia. White subscribed to 5,500 Class A Voting
Shares, while Silver King of Virginia subscribed to 4,500 Class B Series 1 Non-voting
Shares. White’s contribution for the shares was the construction permit. Silver King of
Virginia’s contribution was $45,000. Silver King’s Class B Series I shares were
convertible to Class B Series 2 Voting Shares, but only at least 18 months after
WTMW(TV) began airing programming. The bylaws also included a right of ﬁrst refusal
clause for the sale of Class A stock, allowing Urban’s non-selling Class A shareholders
and Class B shareholders, respectively, the right to buy the Class A shares up for sale
before any outsider could.

Two especially interesting portions of the contractual arrangements between
Silver King and Urban were a put and a call option. The call option allowed the Class A
shareholders (White) to buy all of the Class B shares (Silver King’s) for $10 should HSC
default on compensation payments under any afﬁliation agreement. The put option gave

Silver King the right to require Urban to redeem its Class B shares at 45 percent of the

17

fair market value25 of Urban should certain “speciﬁed events” occur. Alternatively,
should a “speciﬁed event” occur, Urban could, at its discretion, sell all of its assets or
cause the sale of all of its stock. Although many “speciﬁed events” are listed,26 the
crucial ones regarding Silver King’s exertion of control were the failure of Urban to
afﬁliate with HSC or to perform its obligations under the afﬁliation agreement, and
default by Urban under any loan agreement.

The ﬁnal ﬁnancial tie between Urban and Silver King was a loan agreement
between Urban and HSN, dated March 22, 1990, to fund construction of the television
station. Urban was initially advanced $1.15 million, with the principal amount increasing
(to a maximum total of $5.45 million) as construction expenses were incurred. Urban
was required to repay the loan beginning 90 days after commencement of operations.

The terms of the loan called for 84 monthly amortized payments based on an annual
interest rate of 12.8%. The loan included a number of important covenants. First, it
restricted Urban’s outside borrowing to $50,000. Second, it prohibited Urban from
issuing shares, options, warrants or convertible securities. Third, Urban was required to
submit a budget, subject to Silver King’s approval, at least 20 days prior to each ﬁscal
quarter. Finally, should an event of default occur, Silver King could demand immediate
repayment of the loan principal and accrued interest and could also unilaterally cancel the
afﬁliation agreement between HSC and Urban. Events of default included failure to
perform under the afﬁliation agreement with HSC as well as failure to pay an installment

of the loan. Over the next several years, the principal loan amount was increased,

 

25 Should the event occur within ﬁve years of the agreement, Silver King would be
entitled to 49.5 percent of Urban’s fair market value.

18

reﬂecting additional expenses for station construction. By June 1993, the loan had been
increased to $10.5 million.

Taken individually, these contracts appeared to simply protect Silver King’s and
Urban’s business interests. For example, Silver King argued that banks issuing loans to
network afﬁliates often insist on terms giving the lender the right to call the loan should
the afﬁliate cancel the network afﬁliation agreement. The idea is that the afﬁliation
agreement is an important source of revenue to the television station that helps to secure
the loan. Likewise, Silver King’s 45 percent nonvoting equity stake in Urban, even if
converted to voting stock is not enough to control the ﬁrm, yet it allows Silver King to
share in the capital gains of the company. Silver King argues that its put option on this
equity stake merely gives it security against negative ﬁnancial implications of actions by
Urban. For instance, should Urban choose to disafﬁliate with Home Shopping Club
whenever it wanted to, then Silver King, as a nonvoting minority shareholder, would be
left powerless under a different arrangement than they had expected, yielding a different
ﬁnancial and risk position.

However, given Urban’s ﬁnancial constraints, the combination of these contracts
served to lock Urban into the afﬁliation agreement and to give Silver King several
avenues of control. If Urban or Silver King cancelled their afﬁliation agreement, Urban
would have to either obtain ﬁnancing to buy back Silver King’s 45 percent equity stake
or put thecompany up for sale. The ﬁrst option, due to the covenants of the loan
agreement, would only be feasible if Silver King allowed third party ﬁnancing. The

second option was not likely to be palatable to White, as he would lose “control” of the

 

26 Other events include a change in control of Urban and any breach of Urban’s

19

company. Given that Silver King could force the aforementioned situation, this nexus of
contracts provided Silver King with the power to control Urban’s business operations.

In fact, there is ample evidence that Silver King did just this. First, Silver King
chose the engineering and law ﬁrms used in the construction of the television station.
Second, Silver King paid vendors used in the construction of Urban’s station directly for
their services. Urban provided both invoices from 20 vendors that were sent directly to
Silver King and the cancelled checks from Silver King to said vendors. Third, Silver
King management was much more highly involved in strategy sessions and business
meetings with the law and engineering ﬁrms involved in constructing the station.
Attendance lists of these meetings show that Silver King’s chief engineer, Al Evans,
attended 14 times the number of meetings that Theodore White did. Combining this
inﬂuence in day-to-day operations with the control of Urban’s ﬁnances, it is clear that
Silver King exerted tremendous control over Urban. The FCC agreed with this position,
arguing that Silver King’s activities “far exceeded a level of mere inﬂuence over, or

attributable interest in, Urban.”27

1. 3. 3 Jovon Broadcasting Corporation

Like Urban, Jovon Broadcasting Corporation is in the broadcast television
industry. Its principal stockholders are Joseph and Yvonne Stroud. In 1986, Joseph
Stroud was granted a construction permit to build a full-powered television station,

WJYS(TV), in Hammond, Indiana, which is in the Chicago DMA.

 

obligations under its Certiﬁcate of Incorporation or the Shareholder Agreement.

20

During July 1990, Jovon entered into an afﬁliation agreement with HSC. Jovon
agreed to broadcast Home Shopping Club programming for all hours except Monday
through Friday from 6:00 am. to 8:00 am. and Sunday from 6:00 am. to 10:00 am. For
broadcasting this programming, Jovon would receive $152 per hour (equaling
approximately $100,000 per month). The original agreement was for seven years, with a
clause that automatically renewed the agreement perpetually. The agreement allowed
HSC to unilaterally cease providing programming, thus terminating the agreement.
Although Jovon was entitled to make programming decisions, Jovon claimed that when it
attempted to do 50, Silver King declared Jovon in breach of their afﬁliation agreement.

The details of the three ﬁnancial agreements between Jovon and Silver King—a
loan, an equity option, and a put option—were very similar to those of Silver King’s
contracts with Urban. On August 7, 1990, Silver King agreed to loan Jovon $3.6 million
for construction and operation of the Channel 62 television station. As collateral for the
loan, Joseph and Yvonne Stroud pledged all of their voting stock to Silver King. Silver
King was also granted a ﬁrst position security interest in all of Jovon’s tangible and
intangible assets, excluding its FCC authorizations.

As in the Urban case, the loan agreement contained several restrictive covenants.
First, Jovon was not allowed to borrow in excess of $50,000 without prior approval from
Silver King. Second, Jovon was required to obtain Silver King’s written consent in order
to enter into any “contract or commitment” relating to its stock or assets involving
aggregate payments of more than $5,000. Third, Jovon was required to submit annual

budgets subject to Silver King’s approval. Fourth, Jovon needed prior written consent

 

27 Federal Communications Commission (1996) at para 54.

21

from Silver King to “suffer any material increase in excess of the reasonable range in the
broadcast industry in the same or similar markets” with respect to compensation payable
to any employee, or any bonus payment made to any employee, or any material change in
personnel policies, insurance beneﬁts or other compensation arrangements. This right
ceded Jovon’s day-to-day authority over personnel matters to Silver King. Fifth, should
Jovon fail to fulﬁll its afﬁliation agreement with HSC, they would be deemed in default
of their loan. Finally, a breach of the Option Agreement, described below, would also be
deemed an event of default.

The Option Agreement (or equity option) gave Silver King the right to acquire a
45 percent nonvoting equity interest in Jovon at an exercise price of $45,000. The option
existed until the later of the payment of Jovon's loan or the expiration of the initial term
of the afﬁliation agreement. The stock was convertible to voting stock upon exercise of
the option. As with the loan agreement, many restrictive measures were incorporated
into this contract. Jovon was obligated to provide Silver King with ﬁnancial information
and advance annual budgets, and was also precluded from building or acquiring another
broadcast station in any of the 50 largest television markets in the United States without
prior written consent from Silver King.

Finally, a put option gave Silver King the right to require Jovon to buy the
aforementioned exercised or unexercised option (at fair market value) should certain
events occur. Among these events were Jovon no longer being afﬁliated with HSC,
Jovon being in breach of its afﬁliation agreement, Jovon breaching its loan agreement, or
J ovon breaching its shareholder agreement. “Fair market value” would take into

consideration not only the value of the station under HSC programming, but also under

22

alternative programming formats. Thus, Silver King could force an end to the afﬁliation
agreement should it see the value of the station being enhanced under alternative
programming.

Given that the Strouds secured the loan with all of their shares of Jovon, they
could be forced to relinquish control of the company should they be deemed in default of
the loan. One possible scenario would involve Silver King initiating default by ending
the afﬁliation agreement. Since Jovon needed Silver King’s approval to obtain third
party ﬁnancing to repay the loan, the Strouds would have no way of retaining control of
the company under the existing system. Thus, as in the Urban case, Silver King
controlled Jovon by controlling its ﬁnances. Through the aforementioned bond
covenants, Silver King was also able to exert control over Jovon’s day-to-day operations.
Once again, the FCC supported these contentions, stating, “Even absent an equity interest
in Jovon, I believe that Silver King has the potential to inﬂuence the licensee

commensurate with that of a cognizable stockholder.”28

I. 3. 4 Post-script

After its investigations, the FCC found that there was an illegal transfer of control
from Urban to Silver King and ordered Silver King to reform certain terms of its
agreements with Urban and Jovon. The FCC was primarily concerned with Silver King’s
ability to force these companies into bankruptcy, which it had done to Urban, based upon

their performance as affiliates of the Home Shopping Club network. Subsequently the

 

28 Federal Communications Commission (1996) at para 120.

23

FCC initiated a proceeding to revise its attribution rules to better identify controlling
interests in broadcast television stations and concluded that proceeding on August 5,

1999 with the issuance of new attribution rules.29

1.4 Analysis of Silver King Communications’ Use of Financing

I think that the above cases demonstrate a number of interesting points. First, and
perhaps most importantly, ﬁnancial contracts—other than voting equity—can be used to
reallocate control rights between competitors or between a supplier and its outlets. While
these cases were focused on the latter relationship, control rights could have been
reallocated in a similar fashion amongst competitors. In addressing this point, it is
critically important to note that it was Silver King’s option to either force the sale of the
company, by exercising its put option, or to force the company into bankruptcy, by
exercising the embedded option in its debt contracts that gave Silver King control over
Urban and Jovon. In other words, Silver King’s option to force a change in control gave
it control over Urban and Jovon.

Second, I suggest that this insight also indicates how other types of collusive
agreements might be enforced. Unless one argues that investing in a television station is
a negative NPV investment, then both Urban and Jovon could have raised funds in
alternative ways. Thus, Urban and Jovon both voluntarily entered into their contractual
relationships with Silver King. They were in effect colluding with Silver King to avoid

aspects of the FCC’s regulation of broadcasting. Such collusive agreements, as Tirole’s

 

29 See Report & Order (FCC 99-207) for further discussion of the changes in the

(1992) survey points out, raise questions about the enforceability of such agreements. As
Tirole further points out, the literature on collusive agreements often assumes that side
contracts are enforceable but does not spell out. the mechanism. For my cases, the side
contracts had terms giving Silver King the option to force a change in the control of
either Urban or Jovon upon their failure to perform. Thus, the option to force a change in
control served as an enforcement mechanism in these collusive agreements.30

The fact that these arrangements were initially voluntary arrangements also raises
the question as to why Urban and Jovon were willing to cede so many control rights to
Silver King. I think that Rajan and Zingales’s (1998) notion that access to key inputs
deﬁnes power relations within a ﬁrm is also important for understanding the bargaining
position of ﬁrms vis-a-vis one another in contracting. Rajan and Zingales argue that it is
control over access to a key input or inputs that determines who should exercise control
over a ﬁrm. I argue that these same considerations explain why Urban and Jovon
voluntarily ceded signiﬁcant powers to Silver King. Silver King not only provided
Urban and Jovon with ﬁnancing, but it also provided them with programming, which is
essential to the viability and value of any broadcast television station.31 In exchange for
these two key inputs, Urban and Jovon provided Silver King with an outlet for its

programming.

 

attribution rules.

30 There have been cases, such as the case of Turner Broadcasting, described in
Holdemess and Sheehan (1991), in which the change in control was effected by the
conversion of a security holders’ ﬁnancial interest into voting common stock that would
give the security holder control over the ﬁrm. Again, however, it is the ability to change
control that serves as the enforcement mechanism.

3 I See National Association of Broadcasters (1994) for evidence on the ﬁnancial
implications of afﬁliation.

25

Access to outlets also explains why Ted Turner was willing to voluntarily enter
into an agreement with several cable companies to cede some of the control rights of his
Turner Broadcasting company to them. Holderness and Sheehan (1991) examine the use
of ﬁnancing with preferred stock to monitor Ted Turner, a majority shareholder in Turner
Broadcasting. While Holderness and Sheehan explain why the cable companies had an
incentive to provide Turner with ﬁnancing, they do not adequately explain why Turner
was willing to accept these companies’ terms. After all, Turner should have been able to
turn to others to raise funds. The key to Tumer’s motivation was that while his
programming was important to these cable companies, their willingness to distribute his
programming was critical to its value. Consequently, both sides controlled access to key
inputs for one another, and the relative value of those inputs was reﬂected in the terms of
their ﬁnancial agreements and side contracts.

An interesting illustration of both this point and my earlier point about the use of
embedded options is provided by the recent AT&T/TCI merger.32 In their joint proxy to
their shareholders, AT&T and TCI discussed the separation of TCI cable operations and
its programming operations (Liberty Media ventures). To effect this separation, AT&T
proposed to create two classes of tracking stocks. What is particularly interesting about
this transaction is that while AT&T states that it holds all the equity in the New Liberty
Media Group (and thus is exposed to whatever regulations apply to such equity interests),

Dr. John Malone, the former chairman of TCI, and his associates will control the

 

32 There were no FCC rules that would have required the arrangements in question.
Thus, regulation does not appear to be the impetus for the contractual arrangements in
this case, in contrast to our prior cases.

26

company by selecting the board.33 This control by Dr. John Malone and his associates is
protected by the identiﬁcation of “Triggering Events” (much like Silver King used with
Urban and Jovon), such as AT&T trying to appoint a majority of Liberty’s board. If a
“Triggering Event” occurs, then “the assets and businesses of the New Liberty Media
Group would be transferred to a new entity managed by Liberty Management LLC,
which is a separate entity owned by the current ofﬁcers of Liberty Media Corporation,
unless the Triggering Event is waived by Liberty Management LLC.”34 The current
ofﬁcers of Liberty Media Corporation are Dr. John Malone and his associates. Thus,
once again it is the option to force a change in control that allows an entity that does not
own the company’s voting equity to exert control over a company.35 I submit that the
reason that AT&T agreed to these provisions is that they needed the expertise of Dr. John
Malone and his associates to develop their video programming assets — an area of

endeavor totally new to AT&T.

1.5 Summary

Broadcast television regulations motivate ﬁrms in this industry to use creative

ﬁnancing arrangements in order to address incentive problems created by these

 

33 In the Proxy Statement/Prospectus, AT&T/TCI state that “although AT&T will own all
of the equity interests in the New Liberty Media Group and, initially, all of the common
stock of Liberty Media Corporation, the incumbent directors of Liberty Media
Corporation at the Effective Time (and their successors) will be able to control most
aspects of the day-to-day business of Liberty Media Corporation and its subsidiaries
following the Merger.”

34 See page 14 of AT&T and TCI’s joint Proxy Statement/Prospectus
(http://www.att.com/ir/ep/tci_merger/pr__proposed_transactions.html).

27

regulations. I study Silver King Communications’ use of ﬁnancing to gain control over
two separate companies that were its afﬁliates. From this examination, I have provided
strong evidence that ﬁrms can reallocate control over a ﬁrm through options embedded in
non-equity ﬁnancing. Further, I have been able to demonstrate how these same ﬁnancial
contracts can be used to either address incentive conﬂicts or to enforce collusive
agreements between different economic agents. Thus, I provide some insight into how
investors in a ﬁrm, who are also either that ﬁrrn’s competitors or suppliers, can use
embedded options in ﬁnancial contracts to align the interests of the different economic
agents.

This chapter suggests a number of lines of further research, of which I mention
just a few. First, in my cases, Silver King was very speciﬁc in its agreements about the
performance it expected of Urban and Jovon. In other, less regulated industries,
companies may not need to be so speciﬁc. For example, Allen and Phillips (1998)
suggest that partial equity investments of one ﬁrm into another ﬁrm, for which there is a
vertical product relationship, are used to align the interests of these parties. In these
cases, however, the expectations about performance are not speciﬁed in detail.
Consequently, it would be interesting to explore when and why these expectations are
made speciﬁc in some cases and not in other cases.

Second, prior research (e.g., Jensen (1989) or Harris and Raviv (1988)) on the
post-buyout performance of levered buyouts (LBOs) has tended to focus on the level of
debt taken on by these ﬁrms as inducing a change in their behavior. Opler’s (1993)

research suggests that a number of the debt contracts in these transactions had contingent

 

35 It is interesting to note that Dr. John Malone also played a major role in the structuring

28

features that conferred signiﬁcant control rights to debtholders in the LED.
Consequently, this chapter raises the question of whether it is the level of debt, the terms
of the debt, or a combination of both that signiﬁcantly inﬂuences the post-buyout
performance of LBOs.

Finally, I want to point out that my cases have broader applicability than may
appear on ﬁrst blush. The key FCC regulation that forced Silver King to pursue the
contractual arrangements that I studied was its national ownership rule. Without this
rule, broadcast networks would be able to own as many of their outlets as they wished.
However, what are antitrust laws if not, in part, limits on the ownership of assets or
companies in the economy? What distinguishes FCC’s ownership limits from antitrust
limits is a matter of degree rather than kind. As one might expect, where the constraint is
binding, we are more likely to ﬁnd economic agents seeking to relax the constraint.
Since both the FCC ’s ownership limit and antitrust laws focus on voting equity, it is
reasonable to expect that economic agents will use non-equity ﬁnancing to achieve what
they might have achieved through equity ﬁnancing.36 Thus, for any student of Cheung
(1983), it should be no surprise that in a speech before the FCC, Dan Rubinfeld, Chief
Economist of the Antitrust Division of the Department of Justice noted that increasingly

the cases coming before the Antitrust Division involve unique ﬁnancial arrangements that

 

of the ﬁnancing of Turner Broadcasting described in Holderness and Sheehan (1991).
36 A ﬁrm taking a non-voting equity position in another ﬁrm does not have to ﬁle a pre-
merger notiﬁcation under the Hart-Scott-Rodino Act (see
http//:www.ftc.gov/bc/hsr/hsrhtm for further information on pre-merger notiﬁcation
rules). Thus, as I have shown, it is possible to gain control of another ﬁrm while
avoiding antitrust review and regulation without investing in voting securities of the
target ﬁrm.

29

raise antitrust questions.37 I suggest that this indicates that antitrust enforcement will

need to consider the points raised in this chapter.

 

37 See Gilo (1996) for further discussion of the antitrust issues raised by ﬁnancial
arrangements between competitors.

30

CHAPTER 2

THE IMPACT OF DEREGULATION ON THE STRUCTURE AND PAY-
PERFORMANCE SENSITIVITY OF EXECUTIVE COMPENSATION: AN
ANALYSIS OF THE ELECTRIC UTILITY INDUSTRY

2.1 Introduction

Historically, CEOs of regulated companies have seen signiﬁcantly lower pay than
their counterparts at unregulated companies. Academic researchers have also found that
CEOs of regulated ﬁrms experience sensitivities of pay to performance that are
signiﬁcantly lower than unregulated company CEOs.38 However, since substantial
deregulation has occurred and continues to occur in the regulated sector of the economy,
it is unclear whether these differences are still present. Additionally, this deregulation
provides an interesting setting to study the relationship between regulatory environment
and corporate governance structure. Furthermore, assessing the effect of changes in
regulation on the contractual arrangements between managers and ﬁrms may allow us to
better understand the effects of deregulation on corporate performance.

The electric utility industry is a useful industry to study these changes for several
reasons. First, the sheer size of the industry warrants further investigation. U.S. electric
utilities had total revenues of almost $300 billion in 1998. Second, the industry has
relatively homogenous ﬁrms in the sense that most ﬁrms use similar technologies to

produce similar services. As Parrino (1997) notes, these types of industries yield

 

38 See Joskow, Rose and Wolfram (1996) and Joskow, Rose and Shepard (1993) for
details.

31

competitive labor markets because it is relatively easy to evaluate performance of CEOs,
and human capital is readily transferable across ﬁrms. Finally, there was an important
deﬁning event in this industry that issued a strong signal to both the market and to the
ﬁrms in the industry that deregulation was becoming a reality.

The deﬁning event for this study was the passage of the Energy Act in 1992,
which deregulated wholesale aspects of the electric utility industry. Non-utility
producers were allowed to participate in electricity generation and utilities were required
to grant non-utility generators access to transmission facilities. This increased
competition in the generation and wholesale electricity markets should spur improved
operating efﬁciency by existing utilities. The Act also signals the onset of deregulation
in the retail sector, as it allowed the Federal Energy Regulatory Commission (FERC) the
power to construct guidelines to enable this process. In 1995, FERC issued a ruling
allowing individual states to allow competition in the retail electricity market. With the
right to choose amongst competing suppliers of electricity, consumers (both industrial
and residential) will likely demand better service and lower prices. Thus, the passage of
the Energy Act ushered in a new era whereby the decisions of management likely have a
larger impact on ﬁrm performance.

The aim of this chapter is to provide a sense of how the career earnings proﬁle of
electric utility CEOs has evolved with the onset of deregulation. More speCiﬁcaIly, I
hypothesize that the structure of pay after passage of the Energy Act in 1992 will be
much more heavily weighted towards “at risk” long term compensation than it was prior
to passage. I also hypothesize that the sensitivity of pay to performance of the company

will increase post-Act. Finally, I hypothesize that there will be more turnover of CEOs,

32

CEOs will have shorter tenures at the helm, and that more CEOs will be hired from
outside the company than prior to the Act’s passage. All of these hypotheses are
consistent with arguments put forth by Kole and Lehn (1997, 1999) that regulation serves
as a substitute for monitoring and disciplining of executives.39 Once this regulation is
removed, other sources, such as better incentive contracts and greater board control must
be implemented to ﬁll the void.

Similar to Kole and Lehn’s (1999) work on the airline industry, 1 document just
such a shift in certain corporate governance characteristics. The median percent of total
yearly compensation composed of salary and bonus decreases dramatically, from 89% in
1992 to 62% in 1998. This result is in marked contrast to results for a sample of
unregulated company CEOs, where there is very little change for the median CEO over
this period. CEOs of electric utilities also now receive options at a much higher rate,
with over half of CEOs in 1998 receiving options, compared to only 26% in 1992.
Additionally, the change in the value of CEOs’ option holdings and stock holdings
increased dramatically. For those receiving options, median size of holdings went from
$27,000 in 1988 to $348,000 in 1998. Likewise, median stock holdings of CEOs has
increased from $188,152 in 1988 to $1,643,234 in 1998. Results for the mean CEO are
even more dramatic. These results are not solely driven by the bull market of the 19905,
as median (mean) ownership by CEOs increased from 0.02% (0.07%) in 1992 to 0.05%
(0.13%) in 1998. When exercisable options are included in the analysis, the numbers go

from 0.025% (0.106%) in 1988 to 0.083% (0.225%) in 1998.

 

39 Gompers and Lerner (1999) provide an interesting example of how another factor--
reputation-- can serve to monitor executives in a study of venture capital partnerships.

33

Meanwhile, tenure and turnover of CEOs have changed as well. Tenure as CEO
decreased from 5 to 4 years for the median CEO from 1988 to 1998. At the same time,
the median age for CEOs decreased from 57 in .1992 to 55 in 1998. More CEOs were
hired from outside the ﬁrm, with 17% of electric utility CEOs being outsiders in 1992
and 29% in 1998.

The remainder of the chapter is organized as follows. Section 2.2 presents the
relevant literature and develops hypotheses for testing. Section 2.3 describes the data set
used and outlines the methodology utilized in the chapter. Section 2.4 presents the results

of the analysis. Finally, Section 2.5 concludes the chapter.

2.2 Relevant Literature and Hypothesis Development

The standard way to approach the relationship between executive compensation
and company performance in the executive compensation literature is from an agency
theory perspective. Mirrlees (1974, 1976), Holmstrom (1979), and Grossman and Hart
(1983) provide a framework for the use of compensation plans that are designed to align
the interests of risk-averse, self-interested executives with those of shareholders. This
chapter proceeds with this underlying presumption, whereby managers’ incentives are
more closely aligned with that of shareholders through the extensive use of stocks and
options in compensation contracts.

The existing empirical literature on pay for performance for executives of
unregulated companies is quite extensive. There is consistent agreement that a positive

relationship exists between company performance and pay for CEOs. However, there is

34

disagreement over the degree of correlation. On the one hand, Jensen & Murphy (1990)
document a seemingly small responsiveness of CEO pay to changes in company
performance, concluding that salary and bonus increases by $0.02 for every $1,000
increase in shareholder wealth. They also ﬁnd that total CEO pay increases by $3.25 for
every $1,000 increase in shareholder wealth. More recent work by Hadlock and Lumer
(1997) and Hall and Liebman (1998), amongst others, ﬁnd different results. Hadlock and
Lumer (1997) provide evidence that the sensitivity of pay to performance has increased
dramatically from the late 19305 to the 19805. Hall and Liebman (1998) provide support
that the sensitivity of pay to performance is much greater than that documented by Jensen
and Murphy. They do so by pointing out the dramatic effect of the change in value of a
CEO’s stock and options holdings on the changes in his wealth and how these variables
overwhelm any changes in wealth due to changes in “direct” or total yearly
compensation.

The literature on pay for performance for regulated companies is less extensive.
Joskow, Rose, and Shepard (1993) conﬁrm early ﬁndings by Carol] and Ciscel (1982)
that CEOs of regulated companies are paid substantially less than CEOs of unregulated
companies. They ﬁnd evidence consistent with the theory that political pressures
constrain pay for CEOs of regulated companies. They also show that 95% of total yearly
compensation for CEOs of regulated companies comes from salary and bonus. Joskow,
Rose and Wolfram (1996) also conclude that political pressures may constrain top
executive pay levels, this time speciﬁcally for electric utilities. In particular, they ﬁnd
evidence that ﬁrms in environments ranked as more friendly to consumers receive lower

compensation than do CEOs of ﬁrms in environments ranked as more friendly to

35

investors. Additionally, Agrawal, Makhija, and Mandelker (1991) conclude that
managers of regulated ﬁrms do have incentives to maximize shareholder wealth, ﬁnding
a positive relationship between returns and executive pay while ﬁnding no relationship
between sales growth and executive pay.

The overall goal of this chapter is to explore how the career earnings proﬁle of
electric utility CEOs has evolved with the onset of deregulation. More speciﬁcally, I
hypothesize that the structure of pay post-1992 will be much more heavily weighted
towards “at risk” long term compensation than it was pre-1993. I also hypothesize that
the sensitivity of pay to performance of the company will increase after passage of the
Energy Act in 1992. Finally, I hypothesize that after the passage of the Energy Act, there
will be more turnover of CEOs, CEOs will have shorter tenures at the helm, and that
more CEOs will be hired from outside the company. Once regulation is removed, other
sources, such as better incentive contracts should ﬁll the void left from the removal of

regulation.

2.3 Data and Methodology

I began with all US. ﬁrms with primary SIC codes equal to either 4911 or 4931
from Compustat’s active and research databases for the years 1988 to 1998. Data on total
assets, dividend yield, and year-end market value were extracted from the Compustat
database. All subsidiaries and American Depository Receipts (ADRs) were eliminated
from the sample. I then matched this data up with yearly and monthly returns data from

the CRSP database. For the resulting companies with CRSP and Compustat data, I

36

collected over 1,000 corporate proxy statements from the Lexis-Nexis database and from
microﬁche ﬁles. From these statements, I extracted various compensation ﬁgures, such
as salary, bonus, restricted and incentive plan stock grants, options grants, stock and
option holdings, and other miscellaneous compensation items.40 I also collected various
characteristics of executives, such as age, tenure, and whether the CEO was hired from
within the company or not. When this information was incomplete, I used Dun &
Bradstreet’s Reference Book of Corporate Management to ﬁll in missing information on
executive characteristics. The resulting data set is an unbalanced panel of 103 ﬁrms and
228 CEOs, representing 1,068 company years.“

Using titles and compensation ﬁgures from the proxy statements, the individual
whom I deemed to be the “CEO” was determined in the following manner. If an
executive held the title of Chairman of the Board, CEO and President, then I coded him
as the CEO. Similarly, those with the title Chairman and CEO were also coded as the
CEO. If two people in a given year were both listed as CEO and Chairman”, then I
coded the one who earned the most money in that year as the CEO. Likewise, if one
person is denoted as Chairman and another as CEO, then I coded the one who earned

more money as the CEO.”44 Finally, if there is no one with the speciﬁc titles of

 

4° Proxy statements were available for all but ﬁve company years for these remaining
companies.

41 Of the 103 ﬁrms in the sample, 74 exist throughout the sample. The average ﬁrm is in
the sample for 10.37 (out of a possible 1 1) years.

42 This occurred in transitionary years, where the incumbent CEO was being replaced
mid-year.

43 There are 40 company years (representing 3.7% of the observations) for which one of
the two highest paid executives is listed as "Chairman." For 20 of these cases, the
Chairman is the second highest paid executive, and, thus, is not in the sample. In the
other 20 cases, the Chairman is the highest paid executive. For these 20 company years,

37

Chairman or CEO, then I denoted the President as the CEO.45 Throughout this chapter,
the person who I have denoted as CEO is consistently referred to as “CEO,” regardless of
the person’s actual title.

Summary statistics for the ﬁrms and CEOs represented in the sample are
presented in Panels A and B of Table A1. Panel A reports summary statistics for the
years 1988-1992, while Panel B reports summary statistics for the years 1993-1998. All
dollar ﬁgures are expressed in constant 1998 dollars.46 As seen in Table A1, these ﬁrms
are relatively large, having average revenues of $1 .7 billion in the earlier period and $2.4
billion in the later period. However, some large ﬁrms raise the average substantially, as
seen by the much lower median sizes. It should also be noted that the ﬁrms in the sample
have become larger, with standard measures of ﬁrm size—market value, sales, and total
assets—growing in real terms between 40% and 78% for the median ﬁrm across the two
time periods.

Table A1 also shows that average returns for electric utilities have decreased
substantially from the period before the Act to afterwards. At the same time, the
volatility of company returns, as measured by the standard deviation, has increased.

Summary statistics for CEO characteristics are also given. It appears that there is

little difference between periods for tenure characteristics of CEOs. However, this is due

 

15 of the second highest paid executives are listed as "President and CEO," four are listed
as "President," and one is listed as "Vice Chairman."

44 There are four company years for which the Chairman is the only person listed in the
executive compensation table.

45 For 23 company years (or 2.15% of the observations), the persOn whom I deem in
control is listed solely as "President."

46 Adjustments are made using the Consumer Price Index (CPI).

38

to aggregation of the data into just two periods. In Table A5, I present evidence of more
signiﬁcant changes between the two periods, using less aggregated data.

The ﬁrst line of evidence of the changing structure of CEO pay is provided by
constructing pay variables for all CEOs in the sample. Salary and bonus, restricted and
long term incentive plan stock grants, and other compensation were taken directly from
compensation tables in company proxy statements.47 The value of option grants was
computed using the Black-Scholes option pricing model. The exercise price and time to
expiration are taken from option grant tables in the proxy statements. The stock price
used is the closing stock price at year end in the year of the grant, taken from Compustat.
The dividend yield is drawn from Compustat as well. Finally the expected standard
deviation of returns is calculated from the previous 60 monthly returns, which are taken
from the CRSP database. All option grants are valued at calendar year end to be
consistent with other compensation valuations. 48 '

Stock holdings of CEOs are valued as the number of shares owned, as reported in
the “Security Ownership of Certain Beneﬁcial Owners and Management” table in
company proxy statements, multiplied by the year end closing stock price, taken from

Compustat. Stock ownership by CEOs is also measured as a percentage of outstanding

shares, where once again, the CEO’s number of shares are taken from company proxy

 

47 During the sample period, there has been a dramatic change in reporting requirements

for compensation of top executives. Prior to 1993, companies were only required to

report the combination of salary and bonus for the most recently completed ﬁscal year.

Beginning with proxy statements issued in 1993, companies had to individually report

salary, bonus, other annual compensation, restricted stock awards, long term incentive

plan payouts, option grants, and all other compensations for the previous three years.
Fiscal and calendar year end are the same for the entire sample of ﬁrms.

39

statements and the total number of outstanding shares are taken from Compustat.49 I
have provided two measures of stock ownership—one includes options that are
exercisable within 60 days in the CEO’s number of shares owned, while the other
excludes these options.

Option holdings are valued using the methodology developed in Core and Guay
(1998).50 The stock price, dividend yield, and the expected standard deviation of returns
are all calculated in the same way as above. Then, where possible, options are broken
into two categories—exercisable and unexercisable.5 "52 Within each category, the
average exercise price for the category is proxied by the current stock price minus the
quotient of the realizable value of options divided by the number of options. For
example, if the realizable value for (in-the-money) exercisable options is $1 million, the
CEO currently owns 500,000 exercisable options, and the current stock price is $25, then

the average exercise price is calculated as

25 - (1,000,000 / 500,000) = $23

 

49 The share ownership of CEOs is as of the proxy statement issuance date, while the
number of shares outstanding is as of the calendar year end. Typically, the proxy
statements are issued in the Spring after shares outstanding are calculated, although some
proxy statements are issued at calendar year end. The most common month of issuance is
March.

50 Core and Guay (1998) show that these valuations exhibit greater power and less bias
than alternative proxies that are commonly used.

5 I If these can not be differentiated, then the options are assumed to be unexercisable.
This appears to be appropriate given the fact that the only period for which these can not
be categorized is pre-1992. Most of the options held pre-1992 were either newly granted
or very recently granted options.

52 During the sample period, there has been a dramatic change in reporting requirements
for option holdings of top executives. Beginning with proxy statements issued in 1993,
companies had to individually report the number of unexercised options that were
exercisable, the number of unexercised options that were unexercisable, the dollar value
of in-the-money unexercised exercisable options, and the dollar value of in-the-money

40

Calculations are done for each CEO for both the exercisable and unexercisable option
groups. The average time to expiration is assumed to be six years for exercisable options
and nine years for unexercisable options. Using these parameters, the value of both
exercisable and unexercisable option holdings are calculated. Reported ﬁgures for option
holdings represent the sum of these two ﬁgures.

Finally, total yearly compensation (“ﬂow compensation”) is calculated as the sum
of salary, bonus, other annual compensation, restricted and long term incentive plan stock
grants, option grants, and all other compensation for a given year, as reported in the
compensation table in company proxy statements.

To characterize the pay for performance relationship, I run OLS regressions of the
change in log of compensation on the change in log of shareholder wealth, which can be
proxied for by the company’s yearly return. According to the previous literature,53 this
speciﬁcation appears to be relatively robust to outliers in the data. Speciﬁcally, I
estimate

A In (Compensation)p = or + [3 A In (Shareholder value)" (I)
z or + B Rein (2)
equation (2) for CEO i in year t, where Compensation can be any annual measure of a
CEO’s compensation and Ret is the annual return for CEO i’s company in year t, taken
from CRSP. The beta is interpreted as the elasticity of executive compensation with

respect to shareholder wealth.

 

unexercised unexercisable options for the ﬁve most highly paid executives of the
company.
53 In particular, see Hadlock and Lumer (1997) for a lucid discussion of this point.

41

I chose to exclude the ﬁrst and last years of a CEO’s tenure in calculating
sensitivities. If a CEO begins or ends his tenure mid-year, then any change calculated
from this ﬁrst or last year will be biased. For example, suppose a CEO is retiring in June.
He will likely get paid salary for half a year and may not receive any bonus.
Alternatively, an outgoing CEO may receive extraordinary compensation, serving as a
reward for his service. As I have seen several examples of both types exhibited in the
proxy statements, it is unclear as to which way the estimates would be biased. Therefore,
I exclude both the ﬁrst and last year of a CEO’s tenure in the regression analysis. This
reduces the number of observations from 832 to 630.

I ﬁrst run equation (2) using salary and bonus as the measure of compensation for
a couple of reasons. First, most of the literature on pay for performance uses this
measure, allowing me to compare my results to other work. Second, I have reliable data
for the entire sample period to do this. I also run equation 1(2) using ﬂow compensation
as the dependent variable. Flow compensation is a more comprehensive measure of a
CEO’s annual pay, being comprised of salary, bonus, stock and option grants, and other
annual compensation. However, the data for ﬂow compensation is only reliable for the
years 1990-1998 due to the enhanced reporting requirements imposed by the Securities
and Exchange Commission, as described above in Footnotes 10 and 15. Thus, in the ﬂow

compensation regressions, the data used is solely for the years 1990-1998.

42

2.4 Results

2.4.] Electric Utilities

As shown in the left-hand columns of Table A2, there has been a dramatic shift in
the structure of CEO pay from the period before passage of the Energy Act to the period
after.”55 In 1998 dollars, median CEO salary and bonus totaled $407,213 in 1992, and
rose to $656,042 in 1997. This 61% increase, representing a 10% annual growth rate, is
substantially larger than estimates from Murphy (1999), who ﬁnds average annual
increases of about 6% for the median CEO of his sample of unregulated S&P 500
companies, for the years 1992-1996.

At the same time, total real yearly (“ﬂow”) compensation increased by 110%, or
16% annually, from 1992 to 1998 for the median CEO. This was largely driven by the
increased usage of stock and option grants. Not only were these grants increasing in size
for CEOs, but more CEOs were being given ﬁrst-time stock and option grants than had
prior to the Act’s passage. For instance, only 26% of CEOs had option grants in 1992,
while 43% had options in 1997. While the median CEO’s real salary and bonus
increased by 61% from 1992 to 1997, the other components of his real ﬂow
compensation collectively increased by 377%. Thus, there appears to be a broadening of

grants to more CEOs, as noted by the fact that in 1992, the median CEO had no stock

 

54 I also examined these statistics for two subsamples-- (1) the 74 ﬁrms that are in the
sample for all 11 years and (2) the 89 ﬁrms that are in the sample in 1992 and 1997.
Results are very similar to those reported here.

43

grants, while in 1997 the median CEO received a stock grant of $88,734. Reinforcing
this shift, the median CEO’s percentage of ﬂow compensation coming from salary and
bonus, decreased from 89.1% in 1992 to 69.9% in 1997.56

The previous results document a substantial shift in the structure of CEO pay
post-Act. One would expect this to translate into an increased sensitivity of pay to
performance for these ﬁrms. Table A3 presents results examining this proposition.
Speciﬁcation (1) reports results from the OLS regression of the change in log of salary
and bonus on the company’s return. As shown, there is a strongly signiﬁcant relationship
between changes in the ﬁxed component of a CEO’s pay and the company’s market
return. The magnitude of the coefﬁcient, 0.1700,57’58 is much smaller than the estimate of
0.3983 that Murphy (1999) ﬁnds for a sample of 256 observations on S&P 500 utilities59
for 1990-1996 in Murphy (1999). However, this ﬁgure is comparable to the estimate of
0.2625 that Murphy (1999) ﬁnds for 2263 observations of S&P 500 industrials. It is
considerably less than the estimate of 0.491 8 for 399 observations on S&P 500 ﬁnance
companies.

As an executive’s salary is set at the beginning of a year, while the bonus is

typically stated as dependent on the ﬁrrn’s performance, as evaluated at year end, many

 

55 All of the following increases are monotonic, and even more dramatic for the year
1998. Statistics in Table 2, Panel B are presented for the year 1997 for direct comparison
to the sample of unregulated ﬁrms discussed below in Section B.

56 This decrease continues in 1998, where only 62.2% of the median electric utility CEO's
ﬂow compensation came from salary and bonus.

57 Regressions were also run for the two different periods (allowing for different
constants between periods). The results are qualitatively similar.

58 Aggarwal and Samwick (1999) argue that this speciﬁcation biases coefﬁcients towards
zero. Thus, these effects may be understated.

59 These utilities, as deﬁned by Murphy (1999) as all ﬁrms with two-digit SIC codes of
49, represent a broader set of utilities than my sample represents.

have argued that the company’s lagged return should be included in the speciﬁcation. As
shown in speciﬁcation (2), the coefﬁcient on same period return increases from
speciﬁcation (1), with last period’s return not being statistically signiﬁcant at
conventional levels.

To see if there is a difference in the elasticity of pay to performance between the
period leading up to the Energy Act (1988-92) and the period after passage (1993-98), I
run regression (3), which allows for a different elasticity for the period post-Act.
Although both coefﬁcients are signiﬁcant and represent quite different economic values,
with the post-Act elasticity being 75% of the pre-Act elasticity, they are not statistically
different (t = 0.969).60 The economic signiﬁcance of these coefﬁcients is also arguable.
The sample medians for CEO salary and bonus and ﬁrm size (measured by the ﬁrm's
market value) for the pre-Act period are $403,480 and $878 million, respectively.
Therefore, a 10% return for a median size ﬁrm is $87.8 million, corresponding to an
increase in salary and bonus of $8,279. This represents an increase of about 9.4 cents per
$1,000 increase in shareholder wealth. For the post-Act period, using median salary and
bonus of $582,327 and median market value of $1 .56 billion, a 10% company return
translates into an increase in salary and bonus of $8,848, or 5.7 cents per $1,000 increase
in shareholder wealth.

Regression (4) examines whether the performance of electric utility CEOs is

evaluated relative to other companies. In this speciﬁcation, the S&P 500 return is

 

60 This t-statistic is computed from the following regression

A In (Salary + Bonus) = or + B. Retit + B2 ( Pre93 * Reta)

where Pre93 is a dummy variable equal to one if the year is between 1988 and 1992.
Thus, the t-stat on B2 tests whether there is a statistical difference between the pre- and

45

included as an explanatory variable. Now, the coefﬁcient estimates for elasticities
between the two periods are substantially different, with an elasticity 36% less in the
post-Act period. However, once again, they are not statistically different (t = 0.934).
Equation (4) does provide strong evidence for relative performance evaluation (RPE), as
the coefﬁcients on the S&P 500 return are negative, quite large, and signiﬁcant. Using
median salary and bonus and market value ﬁgures for the two periods, a 10% company
return with no return on the S&P index yields a $13,331 ($12,293) increase in salary and
bonus for the pre-Act (post-Act) period. However, if the company earns a 10% return
and the S&P index return is 10% as well, then the increases in salary and bonus drop to
$3,716 and $4,315, respectively for the pre- and post-Act periods. In a very poorly
performing overall market, the effects are more dramatic. With a 10% company return
and a 10% loss in the S&P index, the median CEO's salary and bonus increases by
$22,946 ($20,271) for the pre- (post-) Act period.

Having established evidence for RPE, one might wonder whether the change in
regulation has coincided with a change in the level of RPE. Despite a noticeable
difference between the coefﬁcients, there is not a statistical difference in relative
performance evaluation between the two periods (t = 0.832).

Regression (5), which includes lagged company returns as an explanatory
variable, yields similar results to regression (3). Again, despite signiﬁcant coefﬁcients
for the two periods, the difference between periods is not statistically signiﬁcant (t =
0.550). Finally, regression (6) includes both current and lagged company returns and

current and lagged S&P 500 returns as explanatory variables. Once again, there is strong

 

post-Energy Act periods. All subsequently reported t-statistics for differences in periods

46

evidence for relative performance evaluation. Additionally, the apparent difference
between pre- and post-Act sensitivity of pay to performance is not statistically signiﬁcant
(t = 0.779).

The previous analysis does not provide evidence for an increasing sensitivity of
pay to performance post-Act. This should not be surprising, given that the measure of
pay chosen, salary and bonus, has become a much smaller portion of a CEO’s overall
compensation. A more comprehensive measure of CEO compensation would be ﬂow
compensation. Table A4 reports results for the same regressions as in Table A3, except
that the dependent variable in Table A4 is ﬂow compensation. Flow compensation, or all
reported compensation in a given year, includes salary, bonus, restricted and long term
incentive plan stock grants, option grants, and other miscellaneous compensation. Due to
changes in reporting requirements by the Securities and Exchange Commission, ﬂow
compensation can not be consistently estimated for the years 1988-89. Thus, I exclude
changes in ﬂow compensation for these years, leaving me with just two years of changes
(1990-91 and 1991-92) to include for the pre-Energy Act period. Regression (1)
establishes a basic positive relationship between changes in pay and company return.
Interestingly, the magnitude of the elasticity of compensation to shareholder wealth for
ﬂow compensation is substantially greater than that for salary and bonus. Since ﬂow
compensation includes stock and option grants, this result provides evidence that better
performing companies are rewarding their CEOs more through increased issuance of
stock and option grants than through increases in cash compensation. Thus, these results

are consistent with the use of more efﬁcient contracting.

‘

are from similarly run regressions.

47

Regression (2) points to a strong effect of lagged company returns on the change
in log of ﬂow compensation that is not present in the salary and bonus regressions of
Table A3. The elasticity of ﬂow compensation to lagged company performance is
positive and statistically signiﬁcant, possibly implying that CEOs are given an additional
reward with a lag upon better company performance. This reward appears to be given
with additional stock or option grants, rather than through increased salary and bonus.

Regression (3) appears to show an increased sensitivity of total yearly pay to
performance in the post-Act period, with the coefﬁcients for both periods being
statistically signiﬁcant, rising from 0.2813 to 0.4455. Using median ﬂow compensation
and market value ﬁgures for each of the two periods, there is a dramatic increase in
economic signiﬁcance. For a 10% increase in company return, the median CEO could
expect a $12,493 increase in ﬂow compensation pre-Act. For the post-Act period, a 10%
company return yields a $35,094 increase in ﬂow compensation. Despite the difference
not being statistically signiﬁcant at conventional levels (I = 1.37), there does seem to be
limited evidence for an increased sensitivity of pay to performance. Regression (4)
reports results of adding a benchmark retum—the return on the Standard and Poors 500
(S&P 500). The elasticities are not statistically different between periods (t = 0.30).
However, there is strong evidence of relative performance evaluation (RPE), as measured
relative to the market as a whole, particularly in the pre-Act period. For the median size
ﬁrm and the median pay CEO pre-Act, a 10% company return and a 10% S&P index
return translate into a $1,053 1055 in ﬂow compensation (or 1.2 cents per $1,000 increase
in shareholder wealth). Meanwhile, for the same CEO pay and ﬁrm size pre-Act, a 10%

company return with no return on the S&P index translates into a $27,092 increase in

48

ﬂow compensation. For the post-Act period a 10% company return accompanied by a
10% return in the S&P index yield a $23,364 increase in ﬂow compensation, while a 10%
company return and no return on the S&P index yield a $41,726 increase in ﬂow
compensation. This makes sense, as more of a CEO’s ﬂow compensation was salary and
bonus in this period. This also suggests that grants of both stock and options are not
based on RPE, as shown by the decrease in RPE post-Act, a period where grants were
increasing dramatically.

Regression (5) includes lagged company returns in the speciﬁcation.
Interestingly, there is a statistically signiﬁcant difference between periods for the
estimated coefﬁcients on lagged returns (t = 1.68). Regression (6), which includes the
benchmark return, the lagged benchmark return, and lagged company returns interacted
with the time period dummies, provides little additional insight to the analysis.

Although the changing structure of ﬂow compensation and its elasticity with
respect to company performance provide interesting results, as Hall and Liebman (1998)
point out, when evaluating the pay for performance relationship, one must pay close
attention to the change in an executive’s wealth due to changes in the value of his stock
and option holdings. The authors ﬁnd that the changes in wealth of CEOs due to changes
in values of stock and option holdings dwarf wealth changes due to changes in ﬂow
compensation.

Along these lines, Table A2 shows that in my sample, the mean (median) CEO
saw an increase in real value of share holdings from $739,937 ($289,867) in 1992 to
$4,111,551 ($1,205,378) in 1997. This represents a 456% (315%) increase. At the same

time, the mean (median) CEO’s option holdings value increased from $310,958 ($0) in

49

1992 to $976,802 ($172,055) in 1997. This represents a 214% increase for the mean
CEO.

One might argue that the increased value of option and stock holdings is not
surprising, given that through the mid to late 19905, the US stock market has experienced
phenomenal growth, producing very large stock price increases. However, the mean
(median) CEO’s percentage ownership of the company increased by 48% (104%) from
0.072% (0.024%) in 1992 to 0.107% (0.049%) in 1997. When options that are
exercisable within 60 days are counted in the CEO’s stock ownership, the mean (median)
CEO’s ownership stake increased by 88% (216%) from 0.106% (0.025%) in 1992 to
0.200% (0.079%) in 1997. Admittedly, electric utility CEOs’ stake in the ﬁrm is smaller
than that of the average unregulated company CEO. For instance, Hall and Liebman
(1998) ﬁnd for a sample of 368 large ﬁrms that the median CEO owned 0.14% of his
company in 1994.61 Despite this fact, it appears that the increase in stock ownership of
electric utility CEOs is quite strong. Further evidence of this trend is presented in the
next section.

Table A2 also shows that the sensitivity of option holdings to a 1% increase in
stock price has increased dramatically. For those holding options, the median dollar
increase for a 1% increase in stock price has risen from $7,682 in 1992 to $32,073 in
1997.

Having established a dramatic shift in the structure of pay for CEOs post-Energy
Act, I now turn to an examination of the tenure characteristics of the CEOs of electric

utilities. As shown in Table A5, there is limited evidence for a decrease in age of electric

50

utility CEOs. Although not a monotonic decrease, the median age for CEOs has
decreased from 57 years in 1992 to 55 years in 1998. At the same time, the tenure of
CEOs has seen little change at the median level. However, the turnover of CEOs in this
industry exhibits an interesting pattern. There appears to be quite a bit of turnover in the
years pre-Act, relatively little turnover in the years right after passage of the Act, and a
resurgence in turnover in the most recent two years.62 I hesitate to make any conclusions
from this limited data, but one explanation for this could be that there is some lag
between the decision to ﬁnd new management and the commencement of a new CEO.

What is particularly interesting from Table A5 is the increased percentage of
CEOs post-Act that were hired from outside of the company. I deﬁne an outsider as any
CEO who has worked for the company for less than three years when he begins his tenure
as CEO. Only 17% of CEOs in the industry were hired from outside their respective
ﬁrms in 1992. By 1998, nearly 30% of CEOs were outsiders. Supporting this, the last
column in Table A5 reports the percentage of new hires in a given year that were hired
from outside the company. One should be careful interpreting these percentages, as in
any given year the number of new hires ranges from 5 to 18 CEOs. However, it appears
that a greater percentage of new hires come from outside the company post-Act. In every
year since 1992, at least 25% of that year's new hires were hired from outside of the

company.

 

6’ Holderness, Kroszner, and Sheehan (1999) provide additional evidence that regulated
companies have historically had lower insider ownership than unregulated companies.
62 This is in stark contrast to results found in the banking industry by Hubbard and Palia
(1995), who ﬁnd a large increase in turnover with a loosening of regulation. The
difference is likely due to their analysis involving a more striking change in regulation.

51

2. 4. 2 Unregulated companies

In order to examine whether the changing structure of compensation for electric
utility CEOs can be differentiated from that of unregulated companies, I construct a
sample of compensation characteristics of unregulated company CEOs. This sample was
constructed from the ExecuComp database, which contains information on CEOs for the
years 1992-1997. Beginning with the entire database of CEOs, encompassing 9408
CEO-years, I excluded all CEOs of companies with SIC codes between 4000 and 4999 to '
eliminate CEOs of regulated companies. This reduced the sample to 8160 CEO years. I
then eliminated any observations where the value of option grants, the value of
shareholdings or the percent of the company owned were missing. This reduced the
sample to 7252 observations, representing 5582 observations of year-to-year
compensation changes. Finally, I eliminated any observations for which it was the
CEO’s ﬁrst or last year of tenure as CEO.63 The end result is 5172 observations for the
change in salary and bonus regressions and 5176 observations for the change in ﬂow
regressions.

As shown in Table Al, the unregulated ﬁrms are larger on average than the
electric utilities using market value, total assets, or sales as measures of ﬁrm size.
However, the median electric utility is larger than the median unregulated ﬁrm by all
three measures. Table A2 also shows that salary and bonus for the median CEO of an
unregulated ﬁrm is 24% - 35% higher than the median electric utility CEO. However,

salary and bonus represent a much larger percentage of ﬂow compensation for electric

52

utilities. For the median electric utility CEO, salary and bonus drops by 22% to only
70% of ﬂow compensation."4
Similar to the case of salary and bonus, the median unregulated company CEO’s
ﬂow compensation is 20% higher than the median electric utility CEO in 1992.
However, by 1997, the difference drops to only 5% due to a 110% increase in ﬂow
compensation for the median electric utility CEO. This result is quite likely due to a
large increase in stock grant values and not due to changes in option grant values. By
1997, the median electric utility CEO’s stock grant was $88,734. At the same time, the
median unregulated company CEO received no stock grant. Meanwhile, the median
unregulated company CEO saw a 254% increase in the value of his option grant between
1992 and 1997. In 1997, the median electric utility CEO still received no option grant.
This compensation structure shift makes sense because non-dividend protected
options are not as valuable to executive of companies paying relatively large dividends.
A relatively large percentage of increases in ﬁrm value are siphoned off to shareholders
as dividends, reducing option values at every ex-dividend date. Thus, these options
provide weaker incentives than stock grants do. The increased incentive effect of stock
ownership for electric utility CEOs is supported by the increased ownership percentage of
the company. Although electric utility CEOs’ percentage ownership of the company lags

considerably behind that of unregulated company CEOs, the electric utility CEOs

ownership has been growing much more rapidly. In 1997, the median utility CEO owned

 

63 As mentioned previously, it is uncertain how including these years will bias the
estimates.
64 By 1998, this percentage decreases to 62% for the electric utility sample.

53

109% more than the median utility CEO in 1992, as compared to a 41% increase for the
median unregulated CEO.“66

Table A2 also shows a large disparity in value of shareholdings between electric
utility and unregulated company CEOs. The 1992 median unregulated company CEO’s
shareholdings were 478% higher than that of the median electric utility CEO. However,
electric utility CEOs are seeing a greater increase in their shareholdings’ value, as shown
by the 365% increase from 1992 to 1997. The corresponding increase for the median
unregulated company CEO is 236%.67

Similar to electric utility CEOs, I attempt to characterize the pay for performance
relationship by running regressions based on equation (2) for the unregulated company
CEO sample. Results from these regressions are reported in Tables A6 and A7. The pay-
perforrnance elasticity reported in Table A6, where salary and bonus are the measure of
CEO compensation, are quite similar to those of electric utilities. Regression (l) in Table
A6 shows an elasticity of 0.1865, as compared to 0.1700 for utilities. One interesting
difference between samples is the disparity in coefﬁcient estimates of lagged returns.

While the estimate for electric utilities is positive and statistically insigniﬁcant at

conventional levels, it is negative and statistically signiﬁcant for unregulated companies.

 

65 This effect is even more pronounced in 1998, when the median electric utility CEO
owned 0.054% of the company, nearly 10% greater than the amount for 1997.

66 This measure of ownership is deﬁned as the number of shares (excluding options)
owned by the CEO, divided by the number of company shares outstanding. Information
on the number of shares including exercisable options is not readily available for the
unregulated company sample.

67 See Campbell and Wasley (1999) for an interesting discussion of the limitations of
using too much equity in attempting to align the interests of management and
shareholders.

54

Similar to the electric utility CEOs in Table A3, a strong negative relationship
exists between CEO pay and market performance for unregulated companies. However,
this effect appears to be substantially larger for unregulated companies, with a coefﬁcient
estimate of —0.I37068 to —0.1684 for electric utilities and —0.2354 to —0.2564 for
unregulated companies. Thus, it appears that the relative performance component of
salary and bonus for electric utilities is lower than for unregulated companies. I defer
comment on this issue until the subsequent discussion of ﬂow regressions below.

Finally, the R-squareds for the unregulated company regressions, though slightly
higher than those in Table A3, are comparable to those for utilities. It should not be
surprising that the larger sample of unregulated company CEOs has a higher R-squared.

Table A7 reports results for regressions using ﬂow compensation as the measure
of compensation for the unregulated company sample. In comparing the results of the
ﬂow regressions of regulated and unregulated company samples, some interesting
differences arise. First, the pay-performance elasticity is much stronger for electric
utilities (0.4110) than for unregulated companies (0.2113). This result supports
previously mentioned evidence that electric utility CEOs’ compensation packages have
shifted towards greater use of incentivizing components, particularly stock grants. With a
greater increase in the use of stock grants relative to increases in option grants and cash
compensation, increased stock ownership appears to be the compensation component that
is increasingly being used by strongly performing electric utilities.

Second, the negative relationship between CEO pay and the performance of the

market is much stronger for electric utilities than it is for unregulated ﬁrms. The

 

’8 The appropriate comparison coefﬁcient estimates from Table 3 are those on

55

169 to —0.3952 for electric utilities are much larger than the

coefﬁcient estimates of —0.233
estimates of -0.1575 to —0.1846 for unregulated companies. This once again suggests
that electric utilities are more efﬁciently using stock and option grants to incentivize its
CEOs

Third, the R-squareds are considerably higher for the electric utility regressions
than for the unregulated company regressions. This is a bit surprising, given the much
larger sample size of the unregulated company CEOs.

As in the case of the salary and bonus regressions, there is a discrepancy between
coefﬁcient estimates for the lagged returns variable. The relationship is positive and
statistically signiﬁcant for electric utilities, but negative and sometimes statistically
signiﬁcant for unregulated companies. These results are further evidence that electric

utilities reward strong performance, with some of these rewards being delayed a year. As

mentioned previously, this reward appears to be in the form of additional stock grants.

2.5 Conclusion

With the passage of the Energy Act in 1992, the move towards competition took a
large step forward in the electric utility industry. The law substantially deregulated the
wholesale aspects of the industry and opened the door for competition on the retail level.

At the time of this writing, 15 states had laws on the books allowing for competition in

 

Post92*S&Pt in regressions (4) and (6).
69 Once again, the appropriate comparison estimates from Table 4 are the coefﬁcient
estimates on Post92*S&Pt in regressions (4) and (6).

56

retail electric services.70 With this move towards competition, the decisions of executives
likely have more impact on returns to shareholders. Thus, one would expect a change in
many of the corporate governance characteristics of these ﬁrms.

Similar to the results of Kole and Lehn (1999), I document just such a shift. The
median percent of total yearly compensation composed of salary and bonus decreases
dramatically, from 89% in 1992 to 70% in 1997 and to 62% in 1998. Their counterparts
at unregulated companies saw a decrease from 95% in 1992 to only 93% in 1997. CEOs
of electric utilities also now receive options at a much higher rate, with 43% of CEOs in
1997 receiving options, compared to only 26% in 1992. Unregulated CEOs saw an
increase from 67% to 74% over the same time period. Additionally, the change in value
of electric utility CEOs’ option holdings and stock holdings increased dramatically. For
those receiving options, median size of holdings went from $27,000 in 1988 to $348,000
in 1998. Likewise, median stock holdings of CEOs has increased from $188,152 in 1988
to $1,643,234 in 1998. Results for the mean CEO are even more dramatic. These results
are not solely driven by the bull market of the 19905, as median (mean) ownership by
electric utility CEOs increased from 0.02% (0.07%) in 1992 to 0.05% (0.13%) in 1998.
When exercisable options are included in the analysis, the numbers go from 0.025%
(0.106%) in 1988 to 0.083% (0.225%) in 1998. Thus, it appears that electric utility CEOs
are receiving stronger incentives through the use of stock grants. Additional evidence for
this is shown by the greater sensitivity of pay for performance of ﬂow compensation, or

all annual compensation.

 

70 See National Regulatory Research Institute (1999) for more details on the status of
deregulation across various states.

57

Meanwhile, hiring and turnover practices at electric utilities have changed as well
post-Energy Act. The median age for CEOs decreased from 57 years in 1992 to 55 years
in 1998. At the same time, more CEOs were hired from outside the ﬁrm. In 1992, 17%
of the electric utility CEOs were CEOs who were originally hired from outside the
company. By 1998, 28% were outside hires. Taken together with the changing structure
of pay for CEOs, it appears that the early years of deregulation in the electric utility
industry have brought about fairly substantial shifts in certain elements of the corporate
governance structure. These shifts will likely continue as deregulation begins to occur in
more states on the retail level.

Overall, the analysis of electric utility CEOs' compensation packages conﬁrms the
trend in industrials towards greater usage of incentive compensation. However, the
analysis here also provides compelling evidence for stronger incentive schemes in the
electric utility industry coinciding with deregulation. These results suggest that proﬁt
maximization is very important in the post-deregulation phase for industries such as the
electric utility industry. Thus, other regulated industries undergoing such changes are

likely to see similar changes in incentive compensation schemes for its executives.

58

CHAPTER 3

A REEXAMINATION OF THE CROSS-SECTIONAL VARIATION IN PAY-
PERFORMANCE SENSITIVITIES

3.1 Introduction

The study of the relationship between compensation of high-ranking employees

and the performance of their respective companies is of central importance to the theory

of the ﬁrm. Although academics have long studied this relationship, there has been a

 

relative explosion of such studies in the late 19805 and particularly in the 19905.7| This is
due in large part to a few factors. First, public scrutiny of escalating executive
compensation appears to have reached a high point in the early 19905. During this time
period, the popular press detailed both dramatic pay raises for executives and large
company layoffs, causing a backlash against executive pay raises. This issue received so
much attention as to become part of the debates in the 1992 United States presidential
election, eventually resulting in changes in corporate tax law. More speciﬁcally, n
t.

Congress passed legislation which attempted to curb excessive executive pay by . t

 

disallowing corporations to expense any non-performance based pay for an executive in
excess of $1 million. Second, academic researchers have been able to obtain much more

detailed data on executive compensation due to both increased effort and to enhanced

 

7' See Murphy (1999) for details on the rapid increase in the number of academic articles
published in this area in the 19905.

59

reporting requirements by the Securities and Exchange Commission.72 With these richer
data sets now available, better proxies for compensation variables of interest could be
used to better test existing theory.

At the same time, researchers have been trying to improve the methodology of
studying the pay-performance relationship in order to better understand managerial
incentives. Recently, Aggarwal and Samwick (1999) have made such an improvement,
emphasizing the importance of properly controlling for cross-sectional differences when
studying the pay-performance relationship. Using a standard principal-agent model as
their motivation, they highlight the importance of controlling for the variability of ﬁrm
stock returns in pay-performance speciﬁcations. Their results appear to be quite
dramatic, providing evidence that ﬁrms with more variability in their returns have much
lower pay-performance sensitivities (PPSs) than comparable low variance ﬁrms. In some
regression results, they ﬁnd that a ﬁrm with median sample variance has a PPS over 10
times that of a maximum sample variance ﬁrm.

This chapter provides evidence that this effect is in large part due to ﬁrm size.
Using a comparable sample, I show that it is important to control for not only the
variability in returns, but also for the size of the ﬁrm when estimating pay-performance
sensitivities. This is particularly important for regressions that use dollar returns as the
measure of ﬁrm performance. My results show that when examining comparably-sized

ﬁrms, the PPS for high variance ﬁrms is still less than that of low variance ﬁrms.

 

72 Beginning with ﬁscal year 1993, companies were required to annually report individual
salary, bonus, other annual compensation, restricted stock grants, long term incentive
payouts, option grants, and all other compensation for the top ﬁve paid executives.
Because companies were required to report data for the previous three years, detailed and

60

However, I ﬁnd that there is oftentimes an equal or greater “size effect” than “variance
effect.” In fact, when properly controlling for ﬁrm size, the difference between high and
low variance ﬁrms is negligible under some speciﬁcations.

This chapter is organized as follows. Section 3.2 summarizes the relevant
literature on pay-performance sensitivities and develops hypotheses to be tested later in
the chapter. Section 3.3 describes the data and methodology used in the analysis.
Section 3.4 presents results on pay-performance sensitivities estimates using proper size

controls. Finally, Section 3.5 summarizes the ﬁndings and offers some conclusions.

3.2 Relevant Literature and Hypothesis Development

The standard way to approach the relationship between executive compensation
and company performance in the executive compensation literature is from an agency
theory perspective. Mirrlees (1974, 1976), Holmstrom (1979), Grossman and Hart
(1983), and Holmstrom and Milgrom (1987) provide a framework for the use of
compensation plans that are designed to align the interests of risk-averse, self-interested
executives with those of shareholders. A typical speciﬁcation that is estimated for
executive i working at ﬁrm j in year t is as follows,

mw=70+ylrtﬂ+ki+w+ep (l)
where (1),], is the executive’s compensation, 11:}, is the return to shareholders, is; is an

executive ﬁxed effect, it, is a year effect and at, is the error term. Using this

 

reliable data is available for much of the 19905. Additional details on option holdings for
these same executives was also required, which had not previously been required.

61

speciﬁcation, 7; is interpreted as the sensitivity of an executive’s pay to the ﬁrrn’s
performance, or its PPS. Typically, researchers calculate average pay-performance
sensitivities across a sample of executives or ﬁrms, thus implicitly assuming that all ﬁrms
have the same PPS.

Some researchers, such as Garen (1994) and Aggarwal and Samwick (1999), have
noted that much of the literature ignores key components of the principal-agent model
when estimating PPSs. One prediction of standard principal-agent models is that ﬁrms
with higher variances of stock returns should have lower PPSs than similar ﬁrms with
lower variance of stock returns. Aggarwal and Samwick (1999) provide evidence that
ignoring such cross-sectional differences leads to erroneous refutation of principal-agent
models. These authors document a substantial difference between PPSs for ﬁrms of high
and low variances, using the following speciﬁcation for executive i working at ﬁrm j in
year t,

«or = Yo + Yl a, + Y2 Hair) to. + 73 Ftozn) + l.- + 11: + 8:: (2)
where the. term F (621,) is the cumulative distribution function (CDF) of the variance of
returns for ﬁrms in the sample. The use of the CDF allows for an easy transformation of
coefﬁcient estimates into PPSs for various percentiles of the variance distribution. The
PPS for an executive is y. + y; F (621,). For instance, a ﬁrm with median stock return
variance has a PPS of y. + 0.5y2. The CDF takes on the values of zero and one for the
minimum and maximum observed variances in the sample, corresponding to PPSs of y]
and yt + yz, respectively. In some regression results, Aggarwal and Samwick ﬁnd that a
ﬁrm of median sample variance has a PPS over 10 times that of the maximum sample

variance ﬁrm.

62

 

A standard principle-agent model, such as that in Holmstrom and Milgrom
(1987), predicts that higher returns to shareholders will yield higher compensation, i.e., y.
> O and that ﬁrms with more variable stock returns will see lower PPSs for its executives,
i.e., y; < 0. Although the model predicts a higher sensitivity of compensation to
performance, it makes no prediction regarding the level of compensation with respect to
ﬁrm variance. Thus, there is no prediction for the sign of y3.

One potential problem is that equation (2) ignores the cross-sectional effect of
ﬁrm size on ﬁrm PPSs. Existing theory provides a few hypotheses regarding the
relationship between ﬁrm size and ﬁrm PPSs. Though not modeled formally, Watts and
Zimmerman (1986) argue that there may be a political effect on executive compensation
for large ﬁrms. They argue that since larger ﬁrms are more visible and thus more closely
scrutinized, there may be a public backlash against extremely large raises, even when
ﬁrm performance is outstanding. Schaefer (1998) provides a theoretical model which
allows for a negative relationship between PPSs and ﬁrm size that is also consistent with
large ﬁrm managerial teams having comparable PPSs to small ﬁrm teams. However, his
empirical evidence does not support this theory. Finally, Baker and Hall (1998)
introduce a model that allows CEO productivity to vary with ﬁrm size. They argue that
CEO productivity is increasing in ﬁrm size and that large ﬁrm CEOs are compensated for
this increase in productivity, offsetting the negative relationship between ﬁrm size and
PPS.

Additionally, several researchers have empirically documented a strongly
negative relationship between ﬁrm size and PPS, including Jensen and Murphy (1990a),

Hadlock and Lumer (1997), Schaefer (1998), Garen (1994), and Murphy (1999). In

63

particular, Hadlock and Lumer (1997) provide compelling empirical evidence that not
properly controlling for ﬁrm size yields a misspeciﬁcation of the model, potentially
resulting in invalid inferences from regression results.

This chapter does not attempt to explain why the negative relationship between
ﬁrm size and PPSs exists. Rather, the goal of this chapter is to test whether the
documented cross-sectional effect of the variance in ﬁrm returns on the pay-performance
sensitivity is due primarily to variation in returns or due to ﬁrm size. Thus, I decompose
the variance in stock returns variable in order to better control for ﬁrm size. At the same
time, I control for ﬁrm size in the level. Thus, I estimate the following equation,

too = Yo + VI to + o F(szt) 711': + 73 F(th) 19H" Y4 F(Ozjt) + 75 F(th) + 8.- + it: + a. (3)
where F(zj,) is the CDF of ﬁrm size, measured as the beginning of year market value of
the ﬁrm. For the aforementioned reasons, I predict that y] > 0 and that 73 < 0. I also
predict that y; < 0 , but that the economic signiﬁcance of this variable will be greatly

reduced relative to results for regression (2).

3.3 Data and Methodology

The data for this chapter comes from two main sources. Executive compensation
ﬁgures are taken from the ExecuComp73 data set, a Standard and Poor’s supplement to its
Compustat data set. Stock return data, which is used to construct measures of the

variance of ﬁrm performance, is taken from the CRSP data set.

 

7" I use the October 1999 release of ExecuComp.

The ExecuComp data set contains various compensation measures for the top ﬁve
executives at each of the ﬁrms in the S&P 500, S&P Midcap 400, and S&P SmallCap
600. Due to enhanced reporting requirements by the Securities and Exchange
Commission (SEC), compensation ﬁgures for executives at ﬁrms with ﬁscal years ending
after December 15, 1992 are quite detailed. ExecuComp has compiled data on annual
salary, bonus, stock and option grants as well as stock and Option holdings from proxy
statements ﬁled at the SEC. Thus, data for the years 1993-1998 can be used to estimate
changes in CEO wealth.

Panel A of Table B] provides summary statistics for the year 1995 for various
executive compensation measures used in regression analysis below. Throughout the
analysis, I separate executives into CEOs and non-CEOs, who are referred to as “other
executives.” All compensation ﬁgures are reported in thousands of 1995 dollars. “Flow
compensation” refers to annual compensation paid to executives in both long- and short-
term forms. Short-terrn compensation includes salary, bonus, and other annual payments
such as perquisites, preferential discounts on stock purchases, and gross-ups for tax
liabilities. Long-terrn compensation includes the value of stock option grants, the value
of restricted stock grants, payouts for long-terrn incentive plans and all other
miscellaneous compensation, such as contributions to beneﬁt plans and severance
payments.

As shown in Panel A of Table Bl, sample CEOs average ﬂow compensation for
1995 was $2.4 million, while other executives averaged $955,000 in ﬂow compensation.
However, the distribution of ﬂow compensation is highly skewed, as seen by the

presence of large outliers and by the large difference between mean and median ﬂow

65

compensation. The maximum ﬂow compensation is $65.6 million for CEOs and $34.7
million for other executives. Meanwhile, the mean ﬂow compensation is 69 and 63
percent greater than the median compensation for CEOs and for other executives,
respectively. For these reasons, the use of median regression analysis is appropriate, and
is discussed below.

The second row of Panel A shows that CEOs received an average raise of $72,000
from 1994 to 1995. Meanwhile, other executives received an average raise of $35,000.
Once again, the presence of large outliers skews the distribution, yielding a mean that is
148 (59) percent greater than the median for CEOs (other executives). The third row of
Panel A reports the change in the logarithm of ﬂow compensation, which approximates
the percentage change in ﬂow compensation. The mean and median CEO received a 4.8
and 4.3 percent increase in compensation from 1994 to 1995, while the mean and median
non-CEO received a 3.9 and 3.7 percent raise.

A5 is commonly known, ﬂow compensation does not provide the majority of
incentives for executives.74 For this reason, additional regression analysis is undertaken
using an executive’s change in ﬁrm-speciﬁc wealth. The fourth row presents such
compensation, which includes ﬂow compensation and changes in the market value of an
executive’s stock and option holdings. The average CEO’s ﬁrm-speciﬁc wealth
increased by $19.3 million from 1994 to 1995, while the average non-CEO’s ﬁrrn-
speciﬁc wealth increased by $3.2 million. Once again, large outliers skew the

distributions to the right for both CEOs and for other executives. For instance, one CEO

 

74 See Jensen and Murphy (1990b) and Hall and Liebman (1998) for documentation of
the substantial incentive effects of stock and option holdings relative to those of ﬂow
compensation.

66

had a phenomenal $5.8 billion increase in ﬁrm-speciﬁc wealth, while one non-CEO’s
wealth jumped by over $1 billion. The median increases in ﬁrm-speciﬁc wealth are $3.0
million and $972,000 for CEOs and for non-CEOs, respectively. Similar to regressions
using the previous dollar measures of compensation, the use of median regression
analysis seems appropriate for these measures as well.

One problem with including the change in value of stock option holdings in
executive wealth changes is that a potentially imprecise proxy is being used for changes
in wealth due to option holdings. SEC reporting requirements stipulate that companies
disclose only the value of “in-the-money” option holdings. Thus, a change in the value
of an executive’s option holdings is typically calculated using the value of these in-the-
money options. One could envision a situation where an executive has large holdings of
options that are just out-of-the-money. If the ﬁrrn’s stock price subsequently rises above
the options’ exercise price, then the value of these options is suddenly reported in
company ﬁlings with the SEC. One could also envision the opposite occurring, whereby
a large value of option holdings that is barely in-the-money suddenly gets removed from
company ﬁlings with the SEC due to a small stock price decline. For these reasons,
including changes in the value of option holdings in an executive’s change in ﬁrrn-
speciﬁc wealth may overstate the sensitivity of pay to performance.75 Thus, the change
in an executive’s ﬁrm-speciﬁc wealth excluding the effects of existing option holdings is
presented in row 5, providing a lower bound estimate of executives’ PPS. Using this
measure of compensation, the average CEO’s and non-CEO’s ﬁrm-speciﬁc wealth

increased by $16.8 and $2.4 million, respectively, from 1994 to 1995. The medians are

67

again substantially lower than the means, with median changes of $2.3 million and
$763,000 for CEOs and other executives, respectively.

Additional evidence of the signiﬁcance of executives’ investments in their ﬁrms
in presented in Panel B of Table B1. Nearly 99 percent of CEOs owned stock in their
company in 1995, while almost 95 percent of other executives were shareholders in their
respective ﬁrms for that year. Conditional on owning stock, the mean and median CEO’s
shareholdings were worth $44.5 and $3.7 million, with one particular CEO’s
shareholdings totaling $16.6 billion. Meanwhile, other executives‘mean and median
holdings were valued at $6.3 million and $516,000, with the largest holdings valued at
$2.3 billion. At ﬁrst glance, it appears that executives own a substantial share of the
company, with CEOs averaging a 2.92 percent and other executives holding 0.41.
However these distributions are right-skewed, as seen by the median holdings of 0.26
percent for CEOs and 0.04 percent for other executives.

The third row shows that stock options are widely held by both CEOs and other
executives. Over 85 percent of CEOs held options in 1995, while nearly 88 percent of
other executives had existing option holdings. Conditional on holding options, the in-the-
money mean and median values76 of existing options were $5.8 and $1.0 million for
CEOs and $1.6 million and $406,000 for other executives. Finally, the fourth row
presents statistics on options granted during the 1995 year. Nearly 65 percent of CEOs

and nearly 69 percent of other executives were granted options in 1995. The values of

 

75 The effect of these biases may not be severe, as most options are issued at-the-money
and have been held for several years.

76 The value of existing in-the-money option holdings is simply calculated as the number
of options multiplied by the difference between the end of ﬁscal year stock price and the
option’s exercise price.

68

these options”, conditional on an executive receiving a grant, were $737,000 and
$209,000 for the mean and median CEOs and $297,000 and $83,000 for the mean and
median non-CEOs.

The ﬁrms represented in this sample are 1,500 of the largest ﬁrms in the United
States. Unlike much of the previous work in this area, there is a large disparity in ﬁrm
size, creating a large amount of skewness in the distribution of ﬁrm size. Column 1 of
Table B2 shows that the market value of sample ﬁrms ranges from $2 million to almost
$334 billion, with the mean and median ﬁrms having market values of $4.6 billion and
$1.0 billion. Additional evidence on the heterogeneity of sample ﬁrms is presented in
Columns 2 and 3. Column 2 shows that net income ranges from -$8.5 billion to $40.7
billion, with a mean and median of $203 million and $47 million. Column 3 shows that
sales range from $0 to $165 billion, with a mean and median of $3.5 billion and $992
million. It is precisely because of this wide disparity in ﬁrm characteristics, particularly
in ﬁrm size, that it is imperative to utilize proper controls when estimating PPSs.

Table 82 also presents the distribution of returns for ﬁrms in the sample. Annual
returns are drawn from the ExecuComp data set, which provides data on total percentage
returns to shareholders in a given ﬁscal year. I adjust these nominal percentage returns
using the consumer price index to calculate real percentage returns to shareholders, which
are presented in Column 4 of Table B2. Real annual shareholder returns range from a
loss of 97 percent to a gain of over 1,357 percent, with a mean of 16.7 percent and a

median of 11 percent.

 

77 These option grants are valued using the Black-Scholes option pricing formula, as
detailed in Standard and Poor’s (1995).

69

Column 5 of Table B2 presents the distribution of dollar returns to shareholders.
This variable is constructed using the value of percentage returns in Column 4 multiplied
by the previous year’s market value (in millions of dollars) of the company. Dollar
annual returns range from a loss of over $18 billion to a gain of nearly $118 billion, with
a mean of $907 million and a median of $88 million.

Column 6 of Table B2 presents the distribution of the standard deviation of
percentage returns for sample ﬁrms. For each company year in the sample I take the
previous 60 months of shareholder returns from the CRSP data set, adjust for inﬂation
using the CPI, and calculate a standard deviation.78 The standard deviation ranges from a
low of 2.98 percent to a high of 55.09 percent, with a mean and median of 9.93 and 8.98
percent.

Column 7 of Table BZ presents the distribution of the standard deviation of dollar
returns to shareholders for sample ﬁrms. This is calculated similarly to the standard
deviations in Column 6, with the exception that monthly dollar returns are used. The
monthly dollar returns are calculated by multiplying the monthly percentage returns by
the beginning of year market value of the ﬁrm.79 Thus, the standard deviations presented
are in millions of 1995 constant dollars. The standard deviation ranges from a low of $1
million to a high of $6.9 billion, with a mean and median of $175 million and $58
million. It is this variable that is crucial to the analysis. Because this standard deviation
is calculated based on returns that include the market value of a ﬁrm, ﬁrms with large

market values have large dollar standard deviations. In fact, the correlation coefﬁcient

 

78 Executives at ﬁrms for which there was not at least 48 months worth of returns were
dropped from the sample.

70

between the CDF of market value and the CDF of dollar standard deviations is 0.89. The
question of interest is whether it is ﬁrm size or variance of returns that yields the large
and signiﬁcant negative coefﬁcient on the interaction term in equation (2).

Using the speciﬁcation in equation (3), PPSs can be calculated in a way similar to
the methodology used in Aggarwal and Samwick (1999). For a given ﬁrm size and
variance, one can calculate a PPS. For example, for a median variance ﬁrm of median
size, the PPS is estimated as y; + 0.572 + 0.5y3. For a maximum variance ﬁrm of
minimum size, the PPS is estimated as yl + y;. The focus of this chapter will be on
whether the variance of company returns has a large negative effect on a ﬁrrn’s PPS,
when properly controlling for ﬁrm size. Thus, I will make frequent comparisons of PPSs
between similarly sized ﬁrms of different variances.

The pay-performance sensitivity has been estimated in various ways in p’revious
work. Researchers have used both dollar returns, measured as the dollar change in
market value of the ﬁrm over the year, and percentage returns. The popularity of
regressions using dollar returns may be due in large part to the ease of interpretation of
coefﬁcient estimates. Typically, variables are denominated in such a way that regression
coefficient estimates are interpreted as the dollar increase in executive compensation
corresponding to a $1,000 increase in shareholder value. One particular problem with
using dollar regressions is the sensitivity and precision of coefﬁcient estimates to large
outliers. Given the large outliers in this data set, I focus my attention primarily on
median regression analysis. which is less prone to the strong inﬂuence of outliers.

Median regression minimizes the sum of the absolute residuals rather than the sum of the

 

79 Once again, executives at ﬁrms for which there was not at least 48 months worth of

71

squares of the residuals, as in OLS. Given that the distributions of executive
compensation and ﬁrm performance are skewed to the right, estimated PPSs will be
smaller for median than for OLS regressions. Furthermore, the precision of PPS
estimates will increase using median regressions, as the median is a more robust measure
of the center of the data. Standard errors are calculated according to the bootstrap
procedure detailed in Gould (1992) with 50 replications.

I also estimate equation (2) using percentage returns to shareholders as the
measure of returns. Using this speciﬁcation, the estimated PPS is interpreted as

thousands of dollars of compensation for each percentage point increase in returns to

 

shareholders. Given that there is no ﬁrm size component directly imbedded in percentage
returns, there is no need for a second interaction term. Thus 73 equals zero in equation
(3). One can construct estimates of the PPS comparable to those estimates above by
using sample dollar ﬁgures for median ﬁrm size.

As a robustness check, I also estimate equation (2) using OLS. As previously

mentioned, executive ﬁxed effects are included in these regressions. These regressions

 

are extremely sensitive to the presence of large outliers. For this reason, the reported

in-
regression results are for a sample which excludes the top and bottom one percent of the 1
sample used for the median regressions. d.“

3.4 Results

 

returns were dropped from the sample.

72

As mentioned above, a popular way to examine the sensitivity of executive pay to
the performance of a company is to regress dollar changes in wealth on dollar returns.
Panel A of Table B3 presents results for such regressions for CEOs. Speciﬁcation (1)
reports results from the median regression of the dollar change in CEO wealth on dollar
company returns, the interaction of dollar company returns and the CDF of variance, and
the CDF of variance, following equation (1) above. Compensation is measured in 1995
thousands of dollars, while returns are measured in 1995 millions of dollars. Thus, the
coefﬁcient on ﬁrm performance, which is statistically signiﬁcant, translates into a $26.47
increase in CEO wealth for every $1 ,000 increase in ﬁrm value. The coefﬁcient on the
interaction term is negative and statistically and economically signiﬁcant. The CDF of
variance is also statistically signiﬁcant and positive.

Using this speciﬁcation, the PPS is estimated by adding to the ﬁrm performance
coefﬁcient the product of the interaction term coefﬁcient and the CDF of ﬁrm size. Thus,
the CEO of the smallest variance ﬁrm in the sample, i.e., F (02) = 0, should expect a
$26.47 increase in shareholder wealth for a $1,000 increase in shareholder wealth.
Meanwhile, the CEOs of the median, i.e., F(oz) = 0.5, and maximum, i.e., F(oz) = l,
variance ﬁrms, respectively, should expect a $26.47 — 0.5($24.68) = $14.13 and $26.47 -
$24.68 = $1.79 increase in wealth for every $1,000 increase in shareholder wealth. For
all three variance levels, these estimates are statistically signiﬁcant. Under this
speciﬁcation, the PPS for a maximum variance ﬁrm is estimated at close to eight times
the sensitivity of the median variance ﬁrm. These estimates are quite comparable to
those of Aggarwal and Samwick (1999), who ﬁnd estimates of $27.60, $14.52, and $1.45

for the CEOs of the minimum, median, and maximum variance ﬁrms, respectively.

73

Regression (3) in Panel A of Table B3 controls for the impact of ﬁrm size on the
PPS for CEOs, as detailed in equation (2) above. As the results show, the coefﬁcient on
performance is statistically signiﬁcant and slightly larger than that for regression (l ).
However, the coefﬁcient estimate on the interaction term of ﬁrm performance and the
CDF of variance is greatly reduced, falling from —24.676 to —14.571. This is due to the
effect of ﬁrm size on the PPS, as noted by the strongly signiﬁcant coefﬁcient estimate of
the interaction term of ﬁrm performance and the CDF of ﬁrm size. Additionally, ﬁrm
size entered in the level is also statistically signiﬁcant.

Similar to the methodology used in Aggarwal and Samwick (1999), one can
construct estimates of a CEO’s PPS for various ﬁrm sizes and variances. The estimated
PPS is y. + F (oz) 72 + F (2) y3. For example, for the smallest size ﬁrm with median
variance, a $1,000 increase in shareholder wealth would translate into a $21.39 increase
in CEO wealth. A CEO at a comparably-sized ﬁrm with the maximum variance should
expect only $14.11 for the same increase in shareholder wealth. For a median sized ﬁrm
with the median variance, a CEO should expect a $15.22 increase in wealth, while a CEO
at a similarly sized but high variance ﬁrm should expect only a $7.30 increase in wealth.
Finally, at a large ﬁrm, a median variance ﬁrm CEO should expect a $9.04 increase in
wealth, compared to a $1.75 increase in wealth for his counterpart at a maximum
variance ﬁrm. For all estimated ﬁrm variances and sizes, the PPS estimates are
statistically signiﬁcant at the one percent level or better. Thus, when equally sized ﬁrms
of different variances are compared, the median variance ﬁrm has a PPS from 1.5 to 5.2
times that of the maximum variance ﬁrm. This is dramatically less than the 7.9 times for

regression (1 ).

74

Regression (2) in Panel A of Table B3 reports results for the regression of the
change in CEO wealth excluding option holdings on ﬁrm performance, the interaction of
ﬁrm performance and the CDF of variance of returns, and the CDF of variance of returns.
Once again, ﬁrm performance is strongly signiﬁcant and positive, while the interaction
term is strongly signiﬁcant and negative. The change in CEO wealth for a $1,000
increase in shareholder wealth is $11.27, $5.93 and $0.58 for CEOs at the minimum,
median and maximum variance ﬁrms, respectively. These statistically signiﬁcant
estimates are very comparable to estimates of $12.55, $6.59, and $0.63 found in
Aggarwal and Samwick (1999). However, once ﬁrm size is accounted for, the effect of
the variance of ﬁrm returns on the PPS is greatly reduced. This is shown in regression
(4). Firm size in the level shows up as very signiﬁcant, while in the interaction term, it is
just below standard statistical signiﬁcance levels. Similar to the results from regression
(3), the difference in calculated PPSs between different variance ﬁrms is greatly
diminished. Using estimates from regression (2), the median variance ﬁrm CEO’s PPS is
over 10 times that of maximum variance ﬁrm CEO. Using estimates from regression (4),
this difference is 2 to 7.7 times, depending upon the chosen ﬁrm size.

Panel B of Table B3 presents results from the same regressions for non-CEO
executives. The interpretation of results in Panel B is similar to that of Panel A, although
the magnitude of the coefﬁcient estimates of interest for non-CEOs is very much less
than for CEOs. All variables are statistically signiﬁcant and of the expected sign. For
instance, regression (1) estimates PPSs of $5.77, $3.12, and $0.47 for the minimum,
median, and maximum variance ﬁrms, representing a factor of 6.7 times between median

and maximum variance ﬁrms. These are substantially larger than similar estimates for

75

 

 

CEOs in Panel A. However, once again, when ﬁrm size is accounted for, the difference
between different variance ﬁrms is substantially reduced, ranging from 2.2 to 5.4 times
the sensitivity when going from the median to maximum variance ﬁrm. When changes in
wealth due to option holdings are excluded from the measure of executive compensation,
the results are less dramatic. The Y3 estimate is statistically insigniﬁcant and the 7;
estimate is only slightly changed with the inclusion of the ﬁrm size variable. Thus, the
effect of not controlling for ﬁrm size when estimating sensitivities for other executives is
not as dramatic when existing options are not included in changes in ﬁrm wealth.

Table B4 presents results for the same speciﬁcations, with the exception that
returns to shareholders are measured as percentage returns. Once again, coefficient
estimators are all statistically signiﬁcant and of the expected sign. In comparing results
between speciﬁcations where ﬁrm size is explicitly controlled for to speciﬁcations for
which it is not, one can see that there is little difference between coefﬁcient estimates.
This should not be surprising, as percentage return regressions do not have ﬁrm size
imbedded in the interaction term. For instance, when the change in ﬁrm-speciﬁc wealth
is the dependent variable, as in regressions (1) and (3), then there is little difference
between the respective yr and y; estimates. This is further supported by the estimated
PPSs. For example, using estimates from regression (1) of Panel A of Table B4, the
estimated PPS for a ﬁrm of median variance is 160.487, which is interpreted as a
$160,487 increase in executive ﬁrm-speciﬁc wealth for a one percent increase in
shareholder value. For a ﬁrm of maximum variance, the estimated PPS is $117,299 for a

one percent increase in shareholder value. Thus, the PPS estimate at the median variance

76

is 1.4 times that of the estimate at the maximum variance. When ﬁrm size is explicitly
controlled for, as in regression (3) of Table B4, Panel A, this difference is also 1.4 times.

When using percentage returns in PPS regressions, it is common to translate these
percentage PPS estimates into dollar PPS estimates by computing them for the mean or
median size ﬁrm. In this case, given the skewness of the distribution, I will use the
median size ﬁrm, which has a market value of $1.03 billion. For a one percent increase
in the market value of this ﬁrm, or a $10.3 million increase in shareholder value, the
estimated PPSs for a ﬁrm of median and maximum variance are $15.58 and $113980,
respectively. These results are now directly comparable to the estimated PPS presented
for regression (1) of Table B3, Panel A. One should note that even without controlling
for ﬁrm size, percentage regressions are less sensitive to the well documented effect of
ﬁrm size on estimated PPSs. For a median size ﬁrm, the translated PPSs for regression
(2) of Table B4, Panel A are $7.19 and $5.09 for ﬁrms of median and maximum variance,
respectively.

Panel B of Table B4 reports estimates for other executives for the same
regressions. Once again, without even controlling for ﬁrm size, the difference between
PPSs for different stock return variances is not nearly as large as is implied when using
dollar returns in regressions. Using estimates from regression (1) in Panel B of Table B4,
for a ﬁrm of median size, the percentage PPSs translated into dollar PPSs are $3.41 and
$1.81 for median and maximum variance ﬁrms. The same estimates reported in Panel B
of Table B3 are $3.12 and $0.47. Results are similar when the dependent variable

excludes changes in wealth due to option holdings. One slight difference between non-

77

CEOs and CEOs is that the inclusion of ﬁrm size controls reduces the effect of returns
variance on estimated PPSs. Results from regressions (2) and (4) in Panel B of Table B4
show a reduction in the difference between PPS estimated at median and maximum
variances going from 3.7 times to 2.9 times.

The results from analysis reported to this point have been based on regressions
using changes in an executive’s ﬁrm-speciﬁc wealth as the dependent variable.
However, it has been fairly common for researchers to use yearly measures of
compensation that exclude stock and stock option holdings as the dependent variable,
even though it is well documented that the bulk of incentives for executives to increase
their wealth come from stock and option holdings. For this reason, I estimate PPSs using
various measures of ﬂow compensation, which includes both long- and short-terrn
compensation that an executive receives in a given year.

Despite the aforementioned weaker incentive effect of ﬂow compensation relative
to changes in an executive’s wealth from stock and stock option holdings, regressions
using measures of ﬂow compensation in Table B5 illustrate some interesting ﬁndings.
One can see that the magnitude of coefﬁcients is a fraction of those from regressions
where change in ﬁrm-speciﬁc wealth is the dependent variable. For CEOs, the effect of
including controls for ﬁrm size yields an increase in the economic signiﬁcance of
variance in ﬁrm stock returns. This is illustrated in regressions (1) and (4) of Table B5,
Panel A. However, when the change in ﬂow compensation is used as the measure of
compensation, the inclusion of size controls reduces the magnitude of the variance

interaction term, thus reducing the effect of stock return variance on the estimated PPS.

 

8° The median variance PPS is computed as l60.487/(0.01 * 1,030) = 15.58, while the

78

This is illustrated in regressions (2) and (S) of Table BS, Panel A. Interestingly, when the
change in log of ﬂow compensation, which approximates the percentage change in ﬂow
compensation, is used as the measure of executive compensation, the inclusion of ﬁrm
size controls causes the stock return variance interaction term to become positive and
statistically insigniﬁcant. The results for non-CEOs are much more dramatic. In all
regressions that exclude explicit controls for ﬁrm size, the coefﬁcient estimates of
interest are statistically signiﬁcant and of the expected sign and relative magnitude. Once
ﬁrm size is controlled for, the results are very different. The most dramatic results are for
ﬂow compensation and for the percentage change in ﬂow compensation. Regressions (4)
and (6) in Panel B of Table BS show that the inclusion of ﬁrm size controls results in the
stock return variance interaction term becoming statistically insigniﬁcant. This result is
further highlighted by examining estimated PPSs for various ﬁrm sizes and variances.
For instance, when ﬂow compensation is the dependent variable and without controlling
for ﬁrm size, a ﬁrm of median stock return variance has a PPS 3.3 times that of a ﬁrm
with maximum stock return variance. With controls for ﬁrm size, the difference ranges
from 1.1 to 1.4 times.

Table B6 presents results of regressions identical to those in Table BS with the
exception that percentage returns are used instead of dollar returns. Unlike the
percentage return regressions using change in ﬁrm-speciﬁc wealth as the dependent
variable, those using ﬂow compensation as the dependent variable appear to be sensitive
to the inclusion of ﬁrm size controls. Panel A of Table B6 shows that the coefﬁcient

estimates of interest are of the expected sign and relative magnitude for CEOs. However,

 

maximum variance PPS is computed as 117.299/(0.01 * 1,030) = 11.39.

79

for ﬂow compensation, the difference between a ﬁrm of median and maximum variance
is 5.7 times without ﬁrm size controls. The comparable difference with these controls is
2.9 times.

Panel B of Table B6 illustrates similar results for other executives. As before, the
variable estimates of interest have the expected sign and relative magnitude. Also, when
ﬂow compensation is used as the dependent variable, estimates are again sensitive to the
inclusion of ﬁrm size controls. The difference between PPSs at ﬁrms of median and
maximum variance is 9.2 times without ﬁrm size controls and 3.5 times with controls.
Decreases in the sensitivity of PPS to stock return variance are less dramatic when using
the change in ﬂow compensation and percentage change in ﬂow compensation dependent
variables, as shown in regressions (5) and (6) in Panel B of Table B6.

The results to this point illustrate that it is important to not only control for the
variance in a ﬁrrn’s stock returns but also for ﬁrm size, particularly when using dollar
returns as the measure of ﬁrm performance. The previous results used median regression
analysis due to the presence of very large outliers in both the dependent variable and in
measures of ﬁrm performance. As a robustness check, I also estimate the same
speciﬁcations as above using OLS regressions. However, I exclude the top and bottom
one percent of dependent variables and of measures of ﬁrm performance in running the
OLS regressions. Although this will likely bias the coefﬁcient estimates, the goal of
using OLS regressions is as a robustness check on the interpretation of results, not as a
precise estimator of coefﬁcients.

With these caveats in mind, I run the same speciﬁcations as above using OLS.

All OLS regressions include year effects and executive ﬁxed effects. Standard errors are

80

computed using the Huber-White correction and thus are robust to heteroskedasticity.
Panel A of Table B7 presents results using dollar returns when changes in executive ﬁrm-
speciﬁc wealth is the dependent variable. The results without controlling for ﬁrm size
are quite comparable to those in Table B3. All variable estimates of interest are
statistically signiﬁcant. Coefﬁcient estimates for regression (1) in Panel A of Table B7,
where there is no control for ﬁrm size, show estimates of $27.67 and -$25.74, compared
to $26.47 and -$24.68 for Table B3, Panel A. Thus, the magnitude of the estimates is
slightly larger than for median regressions, as is expected. When ﬁrm size is controlled
for, there is a dramatic reduction in the economic signiﬁcance of ﬁrm stock return
variance on a ﬁrm’s PPS. The coefﬁcient estimate on the stock return variance
interaction term increases from -25.74 to —4.99. Meanwhile, the coefﬁcient estimate on
the ﬁrm size interaction term is strongly statistically and economically signiﬁcant at —
29.63.

This difference is borne out in estimated PPSs. The estimated PPSs for a ﬁrm of
median and maximum variance in regression (1) in Panel A of Table B7 are $14.80 and
$1.92, respectively, representing a difference of 7.7 times. When ﬁrm size is controlled
for, the difference between PPSs for a ﬁrm of median and maximum variance ranges
from 1.1 to 2.6 times. These results point to a much larger signiﬁcance of ﬁrm size than
the results for median regression analysis.

When changes in wealth due to stock option holdings are excluded from the
dependent variable, the results are similar, but not quite as dramatic. Using results from

regression (2) in Panel A of Table B7, the estimated PPS for a ﬁrm of median variances

81

is 12.8 times that of a ﬁrm of maximum variance. However, once ﬁrm size is controlled
for, the difference ranges from 1.3 to 8.5 times, depending upon the ﬁrm size chosen.

Panel B of Table B7 reports results for other executives. Once again, OLS
regressions yield stronger sensitivity of results to inclusion of ﬁrm size than median
regressions do. All variables of interest are estimated with the expected sign and are
statistically signiﬁcant. The results for regression (1) in Panel B of Table B7 provide
PPS estimates of $4.79 and $0.95 for ﬁrms of median and maximum variance,
respectively, representing a difference of 5.1 times. When ﬁrm size is controlled for in
regression (3) in Panel B of Table B7, the differences range from 1.4 to 3.4 times,
depending on the chosen ﬁrm size. The results when existing stock option holdings are
excluded for the dependent variable are similar, though less dramatic.

Table B8 reports OLS results using percentage returns when the dependent
variable is the change in an executive’s ﬁrm-speciﬁc wealth. Regressions (1) and (2) in
Panel A of Table BS have no ﬁrm size controls. All coefﬁcient estimates are of the
expected sign, while the magnitude of estimates is greater than for similar speciﬁcations
using median regressions. For example, the estimated PPS for a ﬁrm of median variance
is interpreted as a $185,784 increase in CEO wealth for a one percent increase in ﬁrm
returns using OLS regressions and $160,487 using median regression analysis. This
translates into a $18.04 increase in executive compensation for a $1,000 increase in
shareholder value for a ﬁrm of median variance and $9.63 for a ﬁrm of maximum

variance.81 When ﬁrm size is controlled for, the difference between PPSs for different

 

8’ The median size ﬁrm has a market value of $1 .03 billion. For such a sized ﬁrm, this
translates into a PPS of 185.784 / (0.01 * 1,030) = 18.04 and 99.214 / (0.01 * 1.030) =
9.63 for the median and maximum variance, respectively.

82

variance ﬁrms is diminished, though not as dramatically as for regressions using dollar
returns. For instance, when change in ﬁrm-speciﬁc wealth is the dependent variable, as
in regression (1) in Panel A of Table BS, a ﬁrm of median variance is estimated to have a
PPS 1.9 times that of a ﬁrm of maximum variance. Examining results from regression
(3) in Panel A of Table B8, when ﬁrm size is controlled for, the difference is 1.7 times.

The results in Panel B of Table B8 point to a similar story. Once again, estimated
coefﬁcients are larger than for estimates from median regressions in Table B4, Panel B.
The difference between PPSs estimated at median versus at maximum variance widens
when OLS regressions are used, particularly for the speciﬁcation where existing stock
option holdings are excluded from the dependent variable. In these two regressions—(2)
and (4) in Panel B of Table B8—the difference is 4.0 and 2.5 times, respectively.

Table B9 presents results of OLS dollar return regressions using measures of ﬂow
compensation as the dependent variable. In general, these results are comparable to
results presented in Table B5, where the only difference is that median regression
analysis is undertaken. However, coefﬁcient estimates using OLS are less precise than
when median regression analysis is used. Nevertheless, when size is controlled for
properly, the effect of variance in ﬁrm returns on estimated PPSs is reduced.

Table BIO reports results for OLS percentage return regressions using measures
of ﬂow compensation as the dependent variable. In all speciﬁcations, for both CEOs and
for other executives, the. inclusion of a control for size decreases the effect of the variance
of ﬁrm returns on estimated PPSs. A rather dramatic example is seen by comparing PPS
estimates from regressions (l) and (4) in Panel A of Table B10. In regression (1), where

there is no ﬁrm size control, the PPSs for a median and maximum variance ﬁrm are 3.68

83

and 0.05, respectively, representing a difference of 69.4 times. When ﬁrm size is
controlled for in regression (4), the PPSs for median and maximum variance ﬁrms are
5.84 and 2.71, respectively, representing a difference of 2.2 times.

As one ﬁnal robustness check, the market value of the ﬁrm and the log of ﬁrm
market value were each separately used in place of the CDF of the market value of the
ﬁrm for all speciﬁcations. In all cases, the results were similar, and, thus are not reported

here.

3.5 Conclusion

The study of the relationship between executive pay and ﬁrm performance is of
substantial importance to academics, as it illuminates the incentive system at work in
public corporations. Fine tuning the techniques used in estimating this relationship is
crucial, as results, and thus, interpretations can be quite sensitive to the particular
speciﬁcation used. This chapter documents the importance of properly controlling for
ﬁrm size when estimating PPSs. In particular, when dollar returns are used as the
measure of ﬁrm performance, it is imperative that the speciﬁcation include size controls.
I estimate PPSs using several different speciﬁcations and ﬁnd that when ﬁrm size is
properly controlled for, the effect of the variance in stock returns on estimated PPSs is
greatly reduced. These results hold for samples of both CEOs and executives below the
CEO level. Still left unresolved is the question as to why ﬁrm size is so strongly

negatively related to estimated PPSs. This is left for future research.

84

SUMMARY

This dissertation contains three chapters that address issues in the area of
corporate governance. The ﬁrst chapter examines the use of ﬁnancial contracts by Silver
King Communications to control two downstream ﬁrms—Urban Communications and
Jovon Communications—who were afﬁliated with Silver King. These cases illustrate the
weakness of the prevailing focus on the ownership of equity with voting rights to
determine the locus of corporate control. In doing 50, Chapter 1 provides insights into
the use of ﬁnancial contracting to avoid regulation, to assign control rights, and to
address costs associated with vertical relationships.

The second chapter examines the impact of the passage of the Energy Act in 1992
on the structure of pay for CEOs in the electric utility industry. Previous work by
Joskow, Rose and Wolfram (1996) and Joskow, Rose and Shepard (1993) has
documented both lower pay levels and lower sensitivities of pay to performance for
CEOs of regulated companies versus CEOs of unregulated companies. Using a sample of
228 CEOs from 1988 to 1998, this chapter conﬁrms ﬁndings by Kole and Lehn (1999)
that deregulation alters several facets of the corporate governance structure. Speciﬁcally,
the percentage of compensation from relatively ﬁxed components—salary and bonus—-
decreases after the onset of deregulation, while the percentage of compensation from “at
risk” components—stocks and options—increases. Additionally, total yearly
compensation becomes more sensitive to the performance of the ﬁrm. It is also shown
that very sizable changes in the value of option and stock holdings of CEOs occur after

passage of the Energy Act. This effect cannot be solely attributed to the overall bull

85

market, as share ownership percentages of CEOs double to quadruple using various
measures of ownership.

Finally, the third chapter examines the methodology used in estimating pay-
performance sensitivities (PPSs). Previous work by Aggarwal and Samwick (1999) has
highlighted the importance of controlling for the variance of ﬁrms stock returns when
estimating PPSs. These authors estimate PPSs that are an order of magnitude greater for
ﬁrms of smaller stock return variances than for ﬁrms of larger variances. Using a
comparable sample of CEOs and non-CEO executives, I ﬁnd that when properly
controlling for ﬁrm size, the negative effect of the variance in stock returns on estimated
PPSs is greatly diminished for both CEOs and non-CEOs. In particular, when using
dollar returns as the measure of ﬁrm performance, it is crucial to properly control for ﬁrm
size. Regressions that use percentage returns as the measure of ﬁrm performance are not
as severely affected by this phenomenon. However, evidence shows that controls should

still be included for these regressions as well.

86

APPENDICES

87

APPENDIX A

TABLES FOR CHAPTER 2: THE IMPACT OF DEREGULATION ON THE
STRUCTURE AND PAY-PERFORMANCE SENSITIVITY OF EXECUTIVE
COMPENSATION: AN ANALYSIS OF THE ELECTRIC UTILITY INDUSTRY

88

Table A1
Firm and CEO Characteristics
This table presents summary statistics on ﬁrm size and characteristics of CEOs from the
two different data sets used in this chapter. Panel A reports statistics on electric utilities
for the years 1988-92, while Panel B reports statistics for electric utilities for the years
1993-1998. Panel C reports statistics on unregulated companies for the years 1988-92.
Market value, Sales, and Total Assets are reported in millions of constant 1998 dollars.

Panel A. Electric Utility Data Set Summary Statistics (1988-92)

 

 

 

 

Wariable N Mean Median s.d. Minimum Nfaﬁmum]
Market value 502 $1,630 $ 878 $1,910 $ 11 $12,200
Sales 502 $1,677 $ 911 $1,810 $ 24 $10,296
Total Assets 502 $4,753 $2,554 $5,213 $ 36 $24,188
Stock Ret (%) 495 16.9 16.5 18.9 -66.7 96.4
Age of CEO 490 56.6 58 6.6 34 70
Yrs as CEO 496 5.1 4 4.7 0 24
Yrs employed 492 23.7 27 12.7 0 48
by company

Panel B. Electric Utility Data Set Summary Statistics (1993-1998)

[Variable N Mean Median s.d. Minimum Maximum]
Market Value 566 $2,700 $1,560 $3,150 $ 11 $23,300
Sales 566 $2,351 $1,352 $2,702 $ 25 $19,942
Total Assets 566 $6,296 $3,572 $6,968 $ 48 $39,514
Stock Ret (%) 562 12.4 10.8 21.1 -47.2 107.1
Age of CEO 546 56.3 57 5.4 39 69
Yrs as CEO 560 5.0 4 4.1 0 21
Yrs employed 554 20.8 21 12.3 0 45

by company

Panel C. Unregulated Company Data Set Summary Statistics (1992-1997)

 

 

[Vaﬁable N Mean Median s.d. Minimum Maximunq
Market Value 7264 $3,408 $ 797 $ 9,623 $ 2 $235,751
Sales 7260 $3,066 $ 777 $ 8,656 $ 0 $165,530
Total Assets 7265 $6,838 $ 831 $24,607 $ 6 $385,479
Stock Ret (%) 7107 24.0 17.0 50.6 -99.0 896.0

89

Table A2

CEO Compensation Summary Statistics
This table presents summary statistics on compensation of CEOs for a data set of electric
utility industry CEOs and for a data set of unregulated company CEOs. Panel A reports
statistics for the year 1992, while Panel B reports statistics for the year 1997. All ﬁgures

are in 1998 constant dollars.

Panel A. CEO Compensation Summary Statistics (1992)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Electric Utilities

N Mean Median Stand dev
Salary & Bonus 100 $422,036 $407,213 $205,823
Value of Option Grants 100 115,294 0 815,865
Value of Stock Grants 100 118,366 0 606,820
Flow Compensation 100 715,222 481,807 1,765,178
Value of Shareholdings 100 739,937 289,867 1,623,860
Value of Option Holdings 100 310,598 0 1,234,708
Percent owned (incl. options) 100 0.106% 0.025% 0.306%
Percent owned (excl. options) 100 0.072% 0.024% 0.183%
SB as Percent of Flow Comp. 100 82.6% 89.1% 17.0%
Percent of CEOs with options 100 26.0% ---- ----
Option holdings sensitivity 33 $24,035 $7,682 $55,185

Unregulated Companies

N Mean Median Stand dev
Salary & Bonus 806 $652,920 $503,283 $536,015
Value of Option Grants 806 513,010 157,645 1,876,938
Value of Stock Grants 806 200,335 0 1,353,585
Flow Compensation 806 920,759 580,263 1,554,740
Value of Shareholdings 806 24,200,870 2,044,585 203,361,300
Value of Option Holdings ---- ---- ---- ----
% own (incl options) ---- ---- ---- ----
% own (excl options) 806 2.46% 0.27% 6.17%
SB as % of Flow Comp. 806 85.37% 94.53% 19.08%
% of CEOs w/ options 806 67.0% ---- ----
Option holdings sens ---- ---- ---- ----

 

 

 

 

 

90

 

 

Panel B. CEO Compensation Summary Statistics (1997)

Electric Utilities

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

N Mean Median Stand dev
Salary & Bonus 90 $708,691 $656,042 $380,726
Value of Option Grants 90 266,140 0 570,998
Value of Stock Grants 90 272,155 88,734 536,555
Flow Compensation 90 1,325,681 1,012,154 1,244,147
Value of Shareholdings 90 4,111,551 1,205,378 10,200,000
Value of Option Holdings 90 976,802 172,055 2,150,692
Percent owned (incl. options) 90 0.200% 0.079% 0.468%
Percent owned (excl. options) 90 0.107% 0.049% 0.272%
SB as Percent of Flow Comp. 90 67.8% 69.9% 21.4%
Percent of CEOs with options 90 43.0% ---- ----
Option holdings sensitivity 48 $50,369 $32,073 $58,040

Unregulated Companies

N Mean Median Stand dev
Salary & Bonus 1,397 $1,207,629 $885,767 $1,287,902
Value of Option Grants 1,397 1,980,557 558,458 6,067,272
Value of Stock Grants 1,397 418,552 0 1,279,041
Flow Compensation 1,397 1,872,073 1,060,655 3,481,457
Value of Shareholdings 1,397 88,728,510 6,866,920 789,575,500
Value of Option Holdings ---- ---- ---- ----
% own (incl options) ---- ---- ---- ----
% own (excl options) 1,397 2.98% 0.38% 6.77%
SB as % of Flow Comp. 1,397 81.74% 92.74% 22.8%
% of CEOs w/ options 1,397 74.0% ---- ----
Option holdings sens ---- ---- ---- ----

 

 

 

 

 

91

 

 

Table A3
OLS Estimates of CEO Pay-Performance Elasticities Using Salary and Bonus for
Electric Utilities (1988-1998)

This table reports coefﬁcient estimates for the regression of change in log of salary and
bonus on various independent variables for a sample of electric utility CEOs . Ret
represents annual company stock return, as computed by CRSP; Pre93 is an indicator
variable equaling one for years before 1993; Post92 is an indicator variable equaling one
for years after 1992; and S&P represents the annual return on the S&P 500, as computed
by CRSP. t-statistics are reported in parentheses. All regressions use White robust

standard errors. Signiﬁcance at the 1%, 5%, and 10% levels are denoted by m, u, and ‘,
respectively.

 

Dependent Variables
Indep Variables

Alog(SB) Alog(SB) Alog(SB) Alog(SB) Alog(SB) Alog(SB)

(1) (2) (3) (4) (5) (6)

Constant .0967 .0862 .0956 .1169 .0865 .1113
Rot, .1700'“ .1846’”

(4.80) (5.11)
Ret..l .0594
(1.45)

Pre93*Ret, .2052’" .3304‘” .2030‘" 3309“"
(4.18) (3.00) (3.88) (3.02
Post92*Rett .1516‘" .2111‘” .1715” .2322 "
(3.67) (3.70) (4.12) (3.80)
Pre93*S&Pr -2383“ -2251"
(-222) (-202)

Post92*S&P( -.1370‘ ----- -.1684‘
(-1.88) (-1.68)

Pre93*Rett-t .0597 .0212
(1.06) (0.26)

Post92*RetH .0506 .0236
(1.07) (0.31)

Pre93*S&Pt-t -.0130
(-0.14)

Post92*S&Pp1 .0460
(0.47)

R2 .0407 .0455 .0422 .0565 .0461 .0588
F-stat 23.01 13.09 12.30 7.15 6.66 3.98
Sample Size 630 629 630 630 629 629

92

Table A4

OLS Estimates of CEO Pay-Performance Elasticities Using Flow for Electric

Utilities (1990-1998)

This table reports coefﬁcient estimates for the regression of change in the log of ﬂow
compensation on various independent variables. Flow compensation includes salary,
bonus, stock & option grants & all other compensation in a given year. Ret represents
annual company stock return, as computed by CRSP; Pre93 is an indicator variable
equaling one for years before 1993; Post92 is an indicator variable equaling one for years
after 1992; and S&P represents the annual return on the S&P 500, as computed by CRSP.
t-statistics are reported in parentheses. All regressions use White robust standard errors.
Signiﬁcance at the 1%, 5%, and 10% levels are denoted by m, ", and ’, respectively.

 

Dependent Variables
Indep Variables
Alog(Flow) A log(Flow) Alog(Flow) Alog(Flow) Alog(Flow) Alog(Flow)
(1) (2) (3) (4) (5) (6)
Constant .1028 .0689 .1051 .1493 .0677 .1045
Rett .41 10“" .4723’”
(5.14) (5.86)
net.-. .2106“
(2.24)
Pre93*Rett .2813‘" .6100‘" .3787‘” .6343‘"
(2.71) (2.81) (3.33) (2.83
Post92*Ret( .4455‘” .5297‘" .5220‘" .6478 ”
(4.92) (3.81) (5.72) (4.34)
Pre93*S&P, -.6337'” -4727‘
(-2.50) (1.68)
Post92*S&Pr ---- ---------- -.233l ----- -.3952
(-133) (-173)
Pre93*Ret.-. .0504 -.0859
(0.45) (-O.63)
Post92*Ret(-t .2876‘" .1312
(2.52) (0.69)
Pre93*S&Pt-t .1055
(0.58)
Post92*S&Pt-t .2958
(1.29)
R2 .0645 .0783 .0673 .0803 .0874 .1052
F-stat 26.41 17.61 13.30 7.36 9.19 5.10
Sample Size 508 507 508 508 507 507

93

Table A5

' CEO Age and Tenure Variables
This table reports yearly statistics on age and tenure variables for CEOs in the electric
utility industry. Age is the median age of CEOs; % Outside is the percentage of CEOs in
a given year who were hired as CEO from outside of the ﬁrm; Ten (pre-CEO) is the mean
number of years that the average CEO has worked at the company before becoming
CEO; Ten (CEO) is the median number of years that the CEO has served as CEO;
Turnover is the percent of CEOs in a given year who are new to their position; and Out |
new is the percentage of new CEOs who are outsiders. A CEO is deemed an outsider if
he has worked at the company for less than 3 years before becoming CEO.

Year N Age % Outside Ten (Pro-CEO) Ten (CEO) Turnover Out | new

1988 101 59.0 13.3 19.5 5.0 ---- ----
1989 101 58.0 17.2 18.8 4.5 18.8 37
1990 100 57.0 19.4 18.4 4.0 13.0 23
1991 99 57.0 17.3 17.9 3.0 18.2 11
1992 100 57.0 17.2 18.4 3.0 14.0 0
1993 100 57.0 19.2 17.1 4.0 9.0 33
1994 99 57.0 22.4 16.7 4.0 13.1 31
1995 100 58.0 25.3 16.2 4.0 5.0 40
1996 98 57.0 23.7 16.3 5.0 7.1 29
1997 90 56.0 27.8 14.1 5.0 13.3 25
1998 79 55.0 29.6 13.6 4.0 21.5 27
All 1068 57.0 20.9 17.1 4.0 11.9 24

94

Table A6

OLS Estimates of CEO Pay-Performance Elasticities Using Salary and Bonus for

Unregulated Firms (1992-1997)

This table reports coefﬁcient estimates for the regression of change in log of salary and
bonus on various independent variables for a sample of unregulated company CEOs. Ret
represents annual company stock return, as computed by CRSP. S&P represents the
annual return on the S&P 500, as computed by CRSP. t-statistics are reported in
parentheses. All regressions use White robust standard errors. Signiﬁcance at the 1%,
5%, and 10% levels are denoted by m, n, and 1', respectively.

 

Dependent Variables
Indep Variables
A log(SB) A log(SB) A log(SB) A log(SB)
(1) (2) (3) (4)
Constant .1133 .1197 .1625 .1655
Rett .1865‘” .1902‘" .2002‘” .2060‘"
(11.83) (11.54) (11.80) (11.46)
lien.l -0003” -.0004"‘
(-2.65) (-3.47)
S&P, —.2354‘” -2564’”
(-6.12) (-6.51)
S&P,l .0615‘
(1.62)
R2 .0621 .0634 .0697 .0724
F-stat 139.97 68.72 72.09 35.24
Sample Size 5172 5053 5172 5053

95

 

Table A7

OLS Estimates of CEO Pay-Performance Elasticities Using Flow for Unregulated

Firms (1992-1997)

This table reports coefﬁcient estimates for the regression of change in the log of ﬂow
compensation on various independent variables for a sample of unregulated company
CEOs . Flow compensation includes salary, bonus, stock & option grants & all other
compensation in a given year. Ret represents annual company stock return, as computed
by CRSP. S&P represents the annual return on the S&P 500, as computed by CRSP. t-
statistics are reported in parentheses. All regressions use White robust standard errors.
Signiﬁcance at the 1%, 5%, and 10% levels are denoted by m, u, and ', respectively.

 

Dependent Variables
Indep Variables
A log(F low) A log(F low) A log(Flow) A log(Flow)
(1) (2) (3) (4)
Constant .1001 .1075 .1331 .1366
Rett .2113‘" .2148'” .2204‘” .2261
(11.93) (11.63) (11.79) (11.45)
Rm.-. -.0002 -0003"
(-142) (.2.00)
S&P. -.1575‘“ -.1846‘"
(-332) (-3.82)
S&P“ .0682
(1.47)
R2 .0472 .0476 .0492 .0505
F-stat 142.43 68.24 69.55 33.31
Sample Size 5176 5057 5176 5057

96

APPENDIX B

TABLES FOR CHAPTER 3: A REEXAMINATION OF THE CROSS-
SECTIONAL VARIATION IN PAY-PERFORMANCE SENSITIVITIES

97

 

at. _ m mm now $3.8 Nut; oom 5mm $3.3 950% :2qu 50:30

 

 

 

 

000.00 woe wee; $3.3 owwK—m coo; Nana $3.2 0.03m SE :0 305:0
$mmdm $36 $36 $v~.m0 $omd $N0.N 00:30 85 :0 Beam
omﬁmnﬁm Em mmmd $3.3 0858.3 owcﬁ 303» $0043 060% SE :_ :2003
5:002 :002 :30 0:3 53:02 :002 :30 0:3
x02 3:020:00 3:020:00 E080: x32 3:020:00 3:020:00 E080:
m0>zzo0xm 050 £00EO 0>::00xm 0035
m2: .:_._m..0:30 33:9 0:: 0.08m .m .055
305:0
33.8.— oomdi- m3. moim $0.m 35636 omNJmT KNN $5.3 mm 3 minus w:_::_0xm
5.003

03.08.— 09“.va m3 ~01». 000% 35.35% owmdi- 08.». :Nd— m2; £603-65 E0m:e:0

 

 

mm_.m 09mm- :8. 08. Name: C. _ .m Kmd- mvo. wvo. we: A305w2 :_ 0w:e:U
mmmdm wooﬁv- mm mm NOW: $34.34 5N0. _ m- om mm me: 30C :m 0m:e:U
35.3” cm 3% mm: Nagy 2360 mm com; mmmd we: 0:0:80U 30.:

x02 52 5:002 :00: 2 x02 52 5:02 :002 Z

m0>sao0xm 050 30050 0>cso0xm C030

 

wee 5033:2580 033:00xm be 00330: .< .055

 

.300: 9:00: 0:: 0:00:05 :_ 0358: we 538:8 wEota :0::0 m0_0r_0m-xom_m 05 0: @6500: 0023 0:0 350% :0_E0 E250

.03—.0: :0 0.05305 33 E 08 35%.: 3:00 =< .m0>::00x0 0.088 :0 $520: :030 0:0 0.00am :0.“ 32 :0: 85:06 9:30:30

300:0: m. .050: mama—0:: :0mmm0cw0: :_ 00m: 5:859:00 :0 3.59008 BEBE: 03.: 05 :0.“ 33 :0: 005w: 35:0: < 0:0:

.m0>::00x0 :050 0:0 mOmU £0: :0: 3:30:30 0:: 5:359:00 :0 3530:: 0:093 :0 35:03 300883 3:085 038 mi...
mag £033.05 :_._m..0:3O 1:: 5:353:50 032.00me

~m 03.;

98

 

 

$5 $3. 003. So.» 80.0 30.» so.” 2
m: 8.0 80 $2 ”mam SN 23. :82
3.0.0 3.2 3.3: 0203 230: 30.8 £32 2:
3m 8.: 0:3 ~30 No.00 80 0.9.0 8
mom 8.: E on? 030 RN mm}. on
02 on: 5 2.8 $3 02 300 E
2 8.2 E 2.2 ms: 00 $0: 8
mm 03 mm 8.2 90 S 08.: 00
S 3.0 mm 80 m8 om 30 8
0m 2: E- 3.0.. NS 2 E. 2
:- 20 8- N3;- 0; o m; cm
2 3.0 SN- 82- a: .2- 3: 2
a 00.0 300;- SS- 3 RD;- S o
E 60 5 0.0 5 Q :0
wagom .8:oQ mar—30M— Eoo-Sm ﬂow—osohmsm gov—osemnm 23:09.00—
90 0.200300 no 5060300 3 waged ow meaoM o:_0>
0:00:05 0.000200% hug—OD _0:::< Eco-BL _0:::< moﬁm 0:50:— 32 “00—802

 

80:00 32 0:000:00 ¢0 80:08 8 008000: mm .0830: 00:00 m0 5003 20:08 o0 800008 05 86: 0000—00—00 m_ 0033 .0550z
.820Q\0 000.80.30Q 0.800000% ammo 80¢ 0000 86: :00» 03800 :0>_w 0 80¢ 2808 cc w:_0000a 05 00>0 0000—00—00 3 0.550%
0200.030 20.08.30Q 0.808% 80:00 0:000:00 mag ¢0 805:: 8 000000: 0:0 .0003 00010: 000% ¢0 w:_::_w0n 0080 803000053
00 0.530% 300.00% 805% 00 0000—00—00 000 0023 80300000080 00 0.550% .8on 3:25. .000 0000 0800:00xm 0.0000 0:0
0000:00m 80¢ :83 00 8030008000 0: 0.550% 800.00% BEEV .000 0000 0800—0000:”; 0.000: 0:0 0:00:00m 80¢ :30: 000 80:00
0:000:00 mag 00 8258 8 0080008 .0030 0:0 .0885 :02 .0503 00008: 03800 05 8 08¢ 00¢ 8003880 08000:: 0300 00:.

03782 500.3355 .5: 0.. 82.230:
8 2......-

99

Table B3

Median Regression Estimates of Pay-Performance Sensitivities for Measures of

F irm-Speciﬁc Wealth (Dollar Returns), 1993-1998

This table reports coefﬁcient estimates for the regression of changes in ﬁrm-speciﬁc
wealth on various independent variables for a sample of executives at 1,500 of the largest
US. ﬁrms. Panel A presents results for CEOs, while Panel B presents results for non-
CEO executives. Change in Wealth is the change in ﬁrm-speciﬁc wealth for an
executive, while Excluding Existing Options is Change in Wealth not including executive
stock option holdings. Performance is the dollar value of annual returns to shareholders;
CDF (Var) is the cumulative distribution function of sample ﬁrm variances; CDF (M10 is
the cumulative distribution function of beginning of year market values of sample ﬁrms.
All regressions include year effects. Bootstrapped standard errors based on 50
replications are presented in parentheses below each estimate. Compensation is
measured in thousands and dollar returns are measured in millions of 1995 constant
dollars. The pay-performance sensitivity is a. = y. + F(oz) y; + F(MV) Y3. Signiﬁcance

at the 1%, 5%, and 10% levels are denoted by ***, **, and *, respectively.

 

 

 

Panel A. CEOs
Change in Excluding Change in Excluding
Wealth Existing Wealth Existing
Options Options
(1) (2) (3) (4)
Coefﬁcients from Median Regressions
7, Performance 26.471‘" 11.273“ 28. 675‘” 11 .452‘”
(951)" (.629)... (1. 320)" (. 664)”
y; Perf *CDF(Var) -24.676 -lO.696 -14. 571 -7. 513
(1.051) (.667) (4.123) (2.348)
73 Perf *CDF(MV) -------------- ‘12. 351 -3.378
(4.709) (2.317)
74 CDF(Var) 1,084’" 2,558’” 219 555”
(214) (144) (380.) (243)..
75 CDF(MV) -------------- 793 2, 263
(423) (230)

 

100

Estimated Pay-Performance Sensitivities

 

an estimated at:

Median variance 14.133‘” 5.925‘" --------------
(.441) (.301)
Maximum variance 1.795." .577". --------------
(.194) (.091)
an. estimated at:
Median variance -------------- 21 .390‘” 7.695‘"
(Minimum size) (2.696) (1.223)
Maximum variance -------------- 14.105‘” 3939‘
(Minimum size) (4.614) (2.304)
Median variance -------------- 15.215m 6.006."
(Median size) (596).. (305).
Maximum variance -------------- 7.929 2.249
(Median size) (2.265) (1.148)
Median variance -------------- 9.039". 4.317".
(Maximum size) (2.129) (1.172)
Maximum variance -------------- 1.754 .560
(Maximum size) (.239) (.108)
N 6,618 6,618 6,618 6,618

 

lOl

Panel B. Other Executives

 

 

 

 

 

Change in Excluding Change in Excluding
Wealth Existing Wealth Existing Options
Options
(1) (2) (3) (4)
Coefﬁcients from Median Regressions
v. Performance 5 771‘” 1. 842‘" 5. 860‘" 1. 732‘"
(.129) ("060) (.134) (.063)
72 Perf * CDF(Var) -5. 302 -1 695 -4. 123 -l. 516
(.140) (.066) (419) (.254)
73 Perf * CDF(MV) -------------- -1.273 -.071
(.403) (.249)
74 CDF(Var) 666’” 1,044‘" 312‘” 253‘"
(25) (19) (50) (39)
y5 CDF(MV) -------------- 418’" 924‘”
(49) (40)
Estimated Pay-Performance Sensitivities
a. estimated at:
Median variance 3.120." .994". --------------
(.062) (.028).
Maximum variance 4.69 .147 --------------
, (.029) (.012)
a] estimated at:
Median variance -------------- 3.798‘" .974'”
(Minimum size) (.213) (.127)
Maximum variance -------------- 1.736." .216
(Minimum size) (.400) (.245)
Median variance -------------- 3.162.” .939."
(Median size) (.061) (.027)
Maximum variance -------------- 1.100". .181
(Median size) (.200) (.121)
Median variance -------------- 2.526." 904‘”
(Maximum size) (.209) (.128)
Maximum variance -------------- .464." .146”.
(Maximum size) (.031) (.015)
N 21.754 21,754 21,754 21,754

 

102

Table B4
Median Regression Estimates of Pay-Performance Sensitivities for Measures of
Firm-Speciﬁc Wealth (Percentage Returns), 1993-1998

This table reports coefﬁcient estimates for the regression of changes in ﬁrm—speciﬁc
wealth on various independent variables for a sample of executives at 1,500 of the largest
US. ﬁrms. Panel A presents results for CEOs, while Panel B presents results for non—
CEO executives. Change in Wealth is the change in ﬁrm-speciﬁc wealth for an
executive, while Excluding Existing Options is Change in Wealth not including executive
stock option holdings. Performance is the annual percentage returns to shareholders,
measured in percentage points; CDF (Var) is the cumulative distribution function of
sample ﬁrm variances; CDF (M V) is the cumulative distribution function of beginning of
year market values of sample ﬁrms. All regressions include year effects. Bootstrapped
standard errors based on 50 replications are presented in parentheses below each
estimate. Compensation is measured in thousands of 1995 constant dollars. The pay-
performance sensitivity is on. = 71 + F(02) 72. Signiﬁcance at the 1%, 5%, and 10% levels
are denoted by ***, **, and *, respectively.

 

 

 

Panel A. CEOs
Change in Excluding Change in Excluding
Wealth Existing Wealth Existing
Options Options
(1) (2) (3) (4)
Coefﬁcients from Median Regressions
7. Performance 203.675‘” 95 749’" 204.036'“ 88. 388‘”
(14.054)‘ (6. 463) (13.61 7.). (4. 978)”
72 Perf * CDF(Var) -86. 376 -43. 366 -87. 675 -35. 072
(19. 661) (9.982) (20. 23 4) (8.052).
73 CDF (Var) -527 -93 2, 774 2, 501
(205) (186) (297) (151)”
74 CDF(MV) -------------- 6, 601 5,816
(358) (201)

 

Estimated Pay-Performance Sensitivities

 

a. estimated at:

Median variance 160.487‘“ 74.066’" 160.199’” 70.852’“
(6.018) (2.863) (6.352) (2.478)

Maximum variance 117.299‘” 52.382'" 116.361‘” 53.316‘“
(8.258) (4.943) (10.000) (4.463)

N 6.618 6,618 6,618 6,618

 

103

Panel B. Other Executives

 

- Excluding

 

 

 

 

Change in Change in Excluding
Wealth Existing Wealth Existing
Options Options
(1) (2) (3) (4)
Coefﬁcients from Median Regressions
71 Performance 51.697’" 23.510‘" 50.004‘" 19.339‘”
(1.468 (.593) (1.712 (.759)
‘72 Perf* CDF(Var) -33.048 ” -19886‘” 31,129 " -15340‘”
(2.061) (.797 (2.236) (.932)
73 CDF(Var) -193'" -206‘ ‘ 711‘" 588‘”
(31) (23) (33) (21)
74 CDF(MV) -------------- 1,905‘" 1,763‘”
(50) (24)
Estimated Pay-Performance Sensitivities
an estimated at:
Median variance 35.173‘" 13.567‘" 34.439‘" 11.669‘"
(.634) (.261) (.769) (.359)
Maximum variance 18.649‘" 3.624'" 18.875’" 4.000‘”
(.879) (.318) (.868) (.340)
N 21,754 21 ,754 21 ,754 21 ,754

 

104

Table B5
Median Regression Estimates of Pay-Performance Sensitivities for Measures of
Flow Compensation (Dollar Returns), 1993-1998

This table reports coefﬁcient estimates for the regression of ﬂow compensation measures
on various independent variables for a sample of executives at 1,500 of the largest US.
ﬁrms. Panel A presents results for CEOs, while Panel B presents results for non-CEO
executives. Flow C ompens is the annual compensation paid to executives in both long-
and short-term forms. Performance is the dollar value of annual returns to shareholders;
CDF (Var) is the cumulative distribution ﬁmction of sample ﬁrm variances; CDF(MV) is
the cumulative distribution function of beginning of year market values of sample ﬁrms.
All regressions include year effects. Bootstrapped standard errors based on 50
replications are presented in parentheses below each estimate. Compensation is
measured in thousands and dollar returns are measured in millions of 1995 constant
dollars. The pay-performance sensitivity is a] = 7) + F(oz) 72 + F(MV) 73. Signiﬁcance
at the 1%, 5%, and 10% levels are denoted by ***, **, and *, respectively.

 

Panel A. CEOs
Flow Change Percent Flow Change Percent
Compens in Flow Change Compens in Flow Change
in Flow in Flow
(1) (2) (3) (4) (5) (6)

 

Coefﬁcients from Median Regressions

 

71 Performance 1.171” .838‘" .00037"‘ 1025"” .855‘” .00056‘"
080) (.099) (.00003) (.175) (.107) (.00004)

72 Perf‘CDF(Va1-) -.964‘" -.785"‘ -.00037"‘ -1200 ” -.552’ .00006
(.209) (.104) (.00003) (.526) (.312) (00005)
73 Perf‘CDF(MV) --------------------- .377 -.249 -.00061

.. m n (531)” (.3440) (.00008)‘

74 CDF(Var) 2,572 161 .03218 1,154 74 -.07058
(81) (28) (.0149) (1 10 n (44). (.03425)

‘75 CDF(MV) ““"7 --------------- 1,670 1 14 .09367
(1 17) (49) (.03933)

t

105

Estimated Pay-Performance Sensitivities

 

a. estimated at:
Median variance

Max variance

0:; estimated at:
Median variance
(Minimum size)
Max variance
(Minimum size)
Median variance
(Median size)
Max variance
(Median size)
Median variance
(Maximum size)
Max variance
(Maximum size)
N

.689”‘

(080)

t.‘

.207

(049)

6,667

.446."

(048)

.0.

.053

(016)

6,667

.00019“‘
(00002)
.000001

(000001)

.579“”

(196)
.303
(333)

t

.454

(048)
.178
(165)
.329"'
(160

‘O

.053

(.020)
6.667

.00058‘”
(00006)
.00061""
(00008)
.00028‘"
(00002)
.00031‘"
(00004)
400003
(00003)
.0000007
(000002)

6,667

 

106

Panel B. Other Executives

 

 

 

 

 

Flow Change Percent Flow Change Percent
Compens in Flow Change Compens in Flow Change
in Flow in Flow
(1) (2) (3) (4) (5) (6)
Coefﬁcients from Median Regressions
71 Performance .358‘” .237‘" .00028‘” .341’" .250‘" .00037’”
(.029) (.017) (00001) (.029) (.015) (00002)
72 Peri‘CDF(Var) -.294"” -.227”‘ -00028‘” -.056 -.135"‘ -.00001
(.033) (.019) (00001) (.099) (.049) (00003)
73 Perf‘CDF(MV) --------------------- -.222‘" -.105"‘ -.00036‘"
(.1 10) (.052) (00004)
74 CDF(Var) 985‘” 53"” .00997 453‘” 28" -.02745
(12) (5) (00770) (22) (12) (02050)
75 CDF(MV) --------------------- 613‘" 30‘" .03210
(21) (11) (02086)
Estimated Pay-Performance Sensitivities
an estimated at:
Median variance .211‘” .124'” .00014‘" ---------------------
(.013) (.008) (000007)
Max variance 064"” .010‘” -.000001 ---------------------
(.007) (.003) (000001)
011 estimated at:
Median variance --------------------- 313”" .183‘" .00037‘"
(Minimum size) (.063 (.030) (.00002
Max variance --------------------- .286’ ’ .115“ .00036‘ ‘
(Minimum size) (10.9.). (053). (0000;)
Median variance --------------------- .202 .130 .00019
(Median size) (.014) (.007) (.000008
Max variance --------------------- .174‘” .063" .00018"
(Median size) (.055) (.027) (.00002)
Median variance --------------------- .091 ‘ .077‘“ ‘ .000006
(Maximum size) (.050) (.024) (.00002
Max variance --------------------- .063‘” .010‘" -.000001 "
(Maximum size) (.055) (.003) (0000005)
N 24,872 24,872 24,872 24,872 24,872 24,872

 

107

Table B6
Median Regression Estimates of Pay-Performance Sensitivities for Measures of
Flow Compensation (Percentage Returns), 1993-1998

This table reports coefﬁcient estimates for the regression of ﬂow compensation measures
on various independent variables for a sample of executives at 1,500 of the largest US.
ﬁrms. Panel A presents results for CEOs, while Panel B presents results for non-CEO
executives. Flow Compens is the annual compensation paid to executives in both long-
and short-term forms. Performance is the annual percentage returns to shareholders,
measured in percentage points; CDF (Var) is the cumulative distribution function of
sample ﬁrm variances; CDF (M V) is the cumulative distribution function of beginning of
year market values of sample ﬁrms. All regressions include year effects. Bootstrapped
standard errors based on 50 replications are presented in parentheses below each
estimate. Compensation is measured in thousands and dollar returns are measured in
millions of 1995 constant dollars. The pay-performance sensitivity is 017 = 77 + F(oz) 72.
Signiﬁcance at the 1%, 5%, and 10% levels are denoted by ***, **, and *, respectively.

 

Panel A. CEOs
Flow Change Percent Flow Change Percent
Compens in Flow Change Compens in Flow Change
in Flow in Flow
(1) (2) (3) (4) (5) (6)

 

Coefﬁcients from Median Regressions

 

71 Performance 19.694'” 6.355‘” .00371‘" 13771” 6.280‘” .00331‘”
(1950 (663 (.00039) (1.624) (595) (00042)

72 Perf* -17.781 ” -3947 “ —.00122” -10.888 “ -3924 ” -.00065
CDF(Var) (2.393) (.824) (00058) (1.866) (.939) (00064)
73 CDF(Var) -376‘” -31 -.03194" 1,299‘" 112‘” .01056
(74) (24) - (01572) (75) (23) (01629)

7.. CDF(MV) --------------------- 3,734‘" 391‘" .09570‘”
(84) (30) (01577)

 

Estimated Pay-Performance Sensitivities

 

a] eStimated at: 0“ it. ##$ .O. ﬁt. it.
Median variance 10.804 4.382 .00310 8.327 4.317 .00298
(.927) (.373) (00021) (.804) (.275) (00017)

Max variance 1.913 ‘ 2.409‘” .00249’" 2.884’” 2.355’” .00266’"
(.883) (.422) (00033) (.628) (.487) (00030)
N 6.667 6,667 6,667 6,667 6,667 6,667

 

108

Panel B. Other Executives

 

Flow Change Percent Flow Change Percent
Compens in Flow Change Compens in Flow Change
in Flow in Flow

(1) (2) (3) (4) (5) (6)

 

Coefficients from Median Regressions

 

71 Performance 6.175‘" 1903*” .00288‘“ 4.688‘" 1.769‘” .00277‘”
(.410) (105)" (00013) (.258)" (.104) (00019)

it. it! #t# t.‘

72 Perf“ -5. 818 -1. 330 -.00139 -3. 911 -1.153 - .00123

CDF(Var) ( 553) (.141) (00023) (.332) (144) ( 00030)
72 CDF(Var) -1 17 -3 -.00472 469 37 .0151 1

(17) (4) (00914) (12) (4) (00853)

74 CDF(MV) --------------------- 1,342‘” 110‘” .05204‘”

(14) (7) (00842)

 

Estimated Pay-Performance Sensitivities

 

017 estimated at: m . . m m
Median variance 3.266 1.239" .00218“ 2.732 1.192 .00215‘"
(.166) ( 052) (00008) (.128) (050) (00009)

Max variance .357‘ 574 .00148‘” .776‘” .615 .00154‘“
(.200) (.066) (00016) (.145) (.067) (.00017)
N 24.872 24,872 24,872 24,872 24,872 24,872

 

109

Table B7
OLS Estimates of Pay-Performance Sensitivities for Measures of Firm-Speciﬁc
Wealth (Dollar Returns), 1993-1998

This table reports coefﬁcient estimates for the regression of changes in ﬁrm-speciﬁc
wealth on various independent variables for a sample of executives at 1,500 of the largest
US. ﬁrms. Panel A presents results for CEOs, while Panel B presents results for non-
CEO executives. Change in Wealth is the change in ﬁrm-speciﬁc wealth for an
executive, while Excluding Existing Options is Change in Wealth not including executive
stock option holdings. Performance is the dollar value of annual returns to shareholders;
CDF (Var) is the cumulative distribution function of sample ﬁrm variances; CDF (M10 is
the cumulative distribution function of beginning of year market values of sample ﬁrms.
All regressions include year effects and executive ﬁxed effects. Heteroskedasticity-
robust standard errors are presented in parentheses below each estimate. Compensation
is measured in thousands and dollar returns are measured in millions of 1995 constant
dollars. The pay-performance sensitivity is 017 = 77 + F(oz) 72 + F (MV) 73. Signiﬁcance
at the 1%, 5%, and 10% levels are denoted by ***, **, and *, respectively.

 

 

 

Panel A. CEOs
Change in Excluding Change in Excluding
Wealth Existing Wealth Existing
Options Options
(1) (2) (3) (4)
Coefﬁcients from OLS Regressions
7. performance 27.668‘" 13.173‘" 36.201” 16.380‘"
(1.637 (1.131 (1.673) (1.270)
72 Perf” CDF(Var) -25.743 ” -12.636 ” 4986’ 6377‘"
(1.752) (1.210) (2.877 (2.389)
73 Perf“ CDF(MV) -------------- -29.631 " 9580‘”
(3.433) (2.748)
74 CDF(Var) 2,945 3,409 -2,237 -264
(3,022) (2,649) (2,921) (2,651)
7, CDF(MV) -------------- 12,845‘" 11,406‘”
(2,697) (2,263)

 

110

Estimated Pay-Performance Sensitivities

 

(11 estimated at:
Median variance

Maximum variance

01. estimated at:
Median variance
(Minimum size)
Maximum variance
(Minimum size)
Median variance
(Median size)
Maximum variance
(Median size)
Median variance
(Maximum size)
Maximum variance
(Maximum size)

N

14796”
(.771)

it.

1.924
(.207)

6.855‘”

(.534)

ﬁt‘

.537
(.153)

111

33.708'"
(2.262)

it!

31.215
(3.402)

ﬁt.

18.893
(.790)

0“

16.400
(1.692)

4.077’”
(1.422)
1.584‘"
(.208)

6,132

13.191‘”
(1.746)
10.003‘”
(2.709)
8.401‘”
(.596)
5.213‘”
(1.340)
3.611‘“
(1.199
.423"
(.160)

6,123

Panel B. Other Executives

 

 

 

 

 

Change in Excluding Change in Excluding
Wealth Existing Wealth Existing
Options Options
(1) (2) (3) (4)
Coefﬁcients from OLS Regressions
77 Performance 8. 633‘” 3.350'” 10 078‘” 3 675‘”
(408) (.301)" (. 476) (. 360)"
72 Perf * CDF(Var) -7. 686 -2.997 -4. 228 -2. 798
(.445) (.325) (. 832).. (.686)
73 Perf* CDF(MV) -------------- -4. 961 -.538
(.947) (.793)
74 CDF(Var) 1,917" 1,783‘" 391 794
(893) (669) (863) (651)"
75 CDF(MV) -------------- 4,191 3, 349
(838) (638)
Estimated Pay-Performance Sensitivities
a7 estimated at:
Median variance 4.790‘" 1.851‘” --------------
(.192) (.143)
Maximum variance .947m .353”. --------------
(.076) (.056)
a. estimated at:
Median variance -------------- 7.964‘” 2276‘"
(Minimum size) (.613) (.500)
Maximum variance -------------- 5.850‘” .877
(Minimum size) (.934) (.778)
Median variance -------------- 5.484’" 2007‘"
(Median size) (.223) (.169)
Maximum variance -------------- 3.370." .608
(Median size) (.463) (.384)
Median variance -------------- 3.003‘" 1738’”
(Maximum size) (.415) (.349)
Maximum variance -------------- .889m .339”.
(Maximum size) (.078) (.058)
N 21,147 21,122 21,147 21,122

 

112

Table B8

OLS Estimates of Pay-Performance Sensitivities for Measures of Firm-Speciﬁc

Wealth (Percentage Returns), 1993-1998

This table reports coefﬁcient estimates for the regression of changes in ﬁrm-speciﬁc
wealth on various independent variables for a sample of executives at 1,500 of the largest
US. ﬁrms. Panel A presents results for CEOs, while Panel B presents results for non-
CEO executives. Change in Wealth is the change in ﬁrm-speciﬁc wealth for an

executive, while Excluding Existing Options is Change in Wealth not including executive

stock option holdings. Performance is the annual percentage returns to shareholders,
measured in percentage points; CDF (Var) is the cumulative distribution function of
sample ﬁrm variances; CDF (M 10 is the cumulative distribution ﬁmction of beginning of
year market values of sample ﬁrms. All regressions include year effects and executive
ﬁxed effects. Heteroskedasticity-robust standard errors are presented in parentheses
below each estimate. Compensation is measured in thousands of 1995 constant dollars.
The pay-performance sensitivity is 017 = 77 + F (02) 72. Signiﬁcance at the 1%, 5%, and

10% levels are denoted by ***, **, and *, respectively.

 

 

 

 

 

Panel A. CEOs
Change in Excluding Change in Excluding
Wealth Existing Wealth Existing
Options Options
(1) (2) (3) (4)
Coefﬁcients from OLS Regressions
77 performance 272.355‘” 130.677‘” 287.130‘" 143.704’”
(16.567 .. (11.282). (17.145)” (11.310).
72 Perf * CDF(Var) -173.142 -81.079 -l65.379 -73.280
(21.458) (15.099) (21.741). (15.128)
73 CDF(Var) -8,167 -5,129 -6,412 -3,619
(2,388) (1,823) (2,394)" (1303)..
74 CDF(MV) -------------- 26,545 22,194
(3,664) (2,449)
Estimated Pay-Performance Sensitivities
an estimated at: m m m m
Median variance 185.784 90.137 204.440 107.064
(8.681)” (5275.1. (10.777). (5701)..
Maximum variance 99.214 49.598 121.750 70.424
(10.319) (6.498) (13.217) (7.178)
N 6,132 6,123 6,132 6,123

 

113

Panel B. Other Executives

 

 

 

 

 

Change in Excluding Change in Excluding
Wealth Existing Wealth Existing
Options Options
(1) (2) (3) (4)
Coefficients from OLS Regressions
7. Performance 91.628‘” 43.391‘” 97.195‘” 47.054'"
(4.639) (3.436) (4.829) (3.526)
72 Perf * CDF(Var) -66.634"‘ .37.260”‘ -64.009”‘ -35.539”‘
(6.006) (4.245) (6.045) (4.253)
73 CDF(Var) -1,867‘" -980” -1,322‘ -622
(678) (496) (682 (497
7, CDF(MV) -------------- 9,805 ” 6,439 "
(1,099) (760)
Estimated Pay-Performance Sensitivities
017 estimated at:
Median variance 58.311‘“ 24.761 65.191 29.284‘”
(2.472)” (1.68)). (2.984)”. (1 £92.)"
Maximum variance 24.994 6.131 33.186 11.514
(2.955) (1.690) (3.571) (1.942)
N 21,147 21,122 21,147 21,122

 

114

Table B9
OLS Estimates of Pay-Performance Sensitivities for Measures of Flow
Compensation (Dollar Returns), 1993-1998

This table reports coefﬁcient estimates for the regression of ﬂow compensation measures
on various independent variables for a sample of executives at 1,500 of the largest US.
ﬁrms. Panel A presents results for CEOs, while Panel B presents results for non-CEO
executives. F low Compens is the annual compensation paid to executives in both long-
and short-term forms. Performance is the dollar value of annual returns to shareholders;
CDF (Var) is the cumulative distribution function of sample ﬁrm variances; CDF (M10 is
the cumulative distribution function of beginning of year market values of sample ﬁrms.
All regressions include year effects and executive ﬁxed effects. Heteroskedasticity-
robust standarderrors are presented in parentheses below each estimate. Compensation
is measured in thousands and dollar returns are measured in millions of 1995 constant
dollars. The pay-performance sensitivity is 017 = 77 + F(O'z) 72 + F(MV) 73. Signiﬁcance
at the 1%, 5%, and 10% levels are denoted by ***, **, and *, respectively.

Panel A. CEOs

 

Flow Change Percent Flow Change Percent
Compens in Flow Change Compens in Flow Change
in Flow in Flow

(1) (2) (3) (4) (5) (6)

 

Coefﬁcients from OLS Regressions

 

.t’

71 Performance .450‘" .838‘" .00035‘” .764 1.120‘" .00047‘”
(.158) (.167) (.00006) (.189) (.194) (00007)

72 Perf“ -.372" -.833‘" -.00035”’ -.102 -.183 .00002
CDF(Var) (.171) (.181) (00006) (.367) (.368). (00011.).
73 Perf* --------------------- -.598 -944 -.00049
CDF(MV) (.426) (.422) (00014)
74 CDF(Var) 542 350 .06914 -208 104 .06954
(422) (441) (16660) (416) (459) (17507)
75 CDF(MV) --------------------- 2,704‘" 774‘ -.08506

(380) (420) (16869)

 

115

Estimated Pay-Performance Sensitivities

 

017 estimated at:
Median variance

is

.00018”‘
(00003)
.0000006
(000008)

.422"
(078)
.005
(029)

.264"‘
(074)
.078“"
(026)

Max variance

017 estimated at:
Median variance
(Minimum size)
Max variance
(Minimum size)
Median variance
(Median size)
Max variance
(Median size)
Median variance
(Maximum size)
Max variance
(Maximum size)
N 6,341

6,188 6,316

116

.713”‘

(272)
.662
(424)

.414

(090)
.363’
(212)
.115
(181)
.064"
(027)
(2341

1028"‘

(270)
.937"
(419)

ii.

.556

(091)
.465”
(209)
.084
(182)
4007
(029)
((188

.00048”‘
(00009)
.00049‘”
(00014)
.00023‘“
(00003)
.00024‘”
(00007)
400001
(00006)
4000004
(000008)

6,316

Panel B. Other Executives

 

 

 

 

 

Flow Change in Percent Flow Change Percent
Compens Flow Change Compens in Flow Change
in Flow in Flow
(1) (2) (3) (4) (5) (6)
Coefﬁcients from OLS Regressions
7] Performance .143‘" .364‘” .00032‘" .118‘ .356‘” .00036‘"
(.052) (.059) (00003) (.063) (.066) (00004)
72 Perf" -.084 -.338‘” -.00032"’ -.404”‘ -.391‘” -.00017‘”
CDF(Var) (.050) (.065) (00003) (.122) (.138) (00006)
73 Perf* --------------------- .346" .062 -.00020”‘
CDF(MV) (.143) (.151) (00007)
74 CDF(Var) ---------------------- 241 8 .09302
(116) (124) (09053)
75 CDF(MV) 483‘“ 32 .05805 944’” 101 -.16749”
(120) (124) (08778) (114) (119) (08372)
Estimated Pay-Performance Sensitivities
017 estimated at:
Median variance .101‘” .195‘” .00016"" ---------------------
(024) (028) (00001)
Max variance .059‘" .026” -.000004 ---------------------
(.010) (.012) (000004)
017 estimated at:
Median variance --------------------- -.084 .160‘ .00027‘"
(Minimum size) (.091) (.093) (.00005)
Max variance --------------------- -.287 -.036 .00019‘”
(Minimum size) (.141) (.150 (.00007)
Median variance --------------------- .089 .191‘ ‘ .00018‘”
(Median size) (.029) (.031) (.00002)
Max variance --------------------- -.114 -005 .00009‘”
(Median size) (.070) (.074) (.00004
Median variance --------------------- .262‘" .222’" .00008‘ ‘
(Maximum size) (.061) (.069) (.00003)
Max variance --------------------- .060‘” .026” -.000005
(Maximum size) (.011) (.012) (000004)
N 24,1 10 24,161 23,973 24,1 10 24,161 23,973

 

117

Table BIO
OLS Estimates of Pay-Performance Sensitivities for Measures of Flow
Compensation (Percentage Returns), 1993-1998

This table reports coefﬁcient estimates for the regression of flow compensation measures
on various independent variables for a sample of executives at 1,500 of the largest US.
ﬁrms. Panel A presents results for CEOs, while Panel B presents results for non-CEO
executives. Flow Compens is the annual compensation paid to executives in both long-
and short-term forms. Performance is the annual percentage returns to shareholders,
measured in percentage points; CDF (Var) is the cumulative distribution function of
sample ﬁrm variances; CDF (M 10 is the cumulative distribution function of beginning of
year market values of sample ﬁrms. All regressions include year effects and executive
ﬁxed effects. Heteroskedasticity-robust standard errors are presented in parentheses
below each estimate. Compensation is measured in thousands and dollar returns are
measured in millions of 1995 constant dollars. The pay-performance sensitivity is 017 = 77
+ F(oz) 72. Signiﬁcance at the 1%, 5%, and 10% levels are denoted by ***, **, and *,
respectively.

 

Panel A. CEOs
Flow Change Percent Flow Change Percent
Compens in Flow Change Compens in Flow Change
in Flow in Flow
(1) (2) (3) (4) (5) (6)

 

Coefﬁcients from OLS Regressions

 

10 Performance 7305‘" 8.579‘" .00415"" 8.980‘" 9.180‘” .00424‘”
(1565) (1.762) 1.000691 (1.612) (1.793) (00070)

72 Perf* -7251‘” -7008‘” -.00277‘" 6275"” -6697’" -.00273”‘
CDF(Var) (2.009) (2.312) (00096) (2.007) (2.312) (.00096)
72 CDF(Var) -120 -2 .23718‘ 86 71 .24871”

(306) (320) (12536) (302 (318 (12540)
74 CDF(MV) --------------------- 3,158 ” 1,096 " .16364

(404) (430) (16793)

 

Estimated Pay-Performance Sensitivities

 

017 estimated at:

Median variance 3.679‘" 5.075’” .00277‘" 5.843‘” 5.831'” .00288‘"
(.738) (.81 1) (00033) (879)" (909) (00036)

Max variance .053 1.571 .00138 2.705 2.483 .00152
(.810) (.940) (00046) (.979) (1.053) (00049)

N 6,341 6,188 6,316 6,341 6,188 6,316

 

118

Panel B. Other Executives

 

 

 

 

 

Flow Change Percent F low Change Percent
Compens in Flow Change Compens in Flow Change
in Flow in Flow
(1) (2) (3) (4) (5) (6)
Coefﬁcients from OLS Regressions
7l Performance 2796‘" 3.803‘" .00331‘" 3.615‘” 4.005'" .00336‘"
(453)” (550)” (00033). (.463)" (.563 .. (00034).
72 Perf * -2.877 -3.l39 -.00218 -2.506 -3.049 -.00116
CDF(Var) (.590) (.703) (.00046) (.592) (.701) (.00046)
73 CDF(Var) -97 38 .06702 -12 58 .07190
(90) (104) (06660) (89) (104) (06683)
74 CDF(MV) --------------------- 1,388‘" 340‘” .08128
(123) (126) (08521)
Estimated Pay-Performance Sensitivities
017 estimated at:
Median variance 1.358‘” 2.233‘" .00222‘” 2.362’” 2.480‘” .00228‘“
(.207) (.250) (.00016) (.238) (.275) (.00017)
Max variance -.081 .664” .00113‘” 1.109‘" .956'“ .00120‘"
(.232) (.265) (.00021) (.273) (.282) (.00022)
N 24,1 10 24,161 23,973 24,110 24,161 23,973

 

119

REFERENCES

120

REFERENCES

Aggarwal, Rajesh, and Andrew Samwick, 1999, The Other Side of the Tradeoff: The
Impact of Risk on Executive Compensation, Journal of Political Economy 107, 65-105.

Agrawal, Anup, Anil Makhija, and Gershon Mandelker, 1991, Executive Compensation

and Corporate Performance in Electric and Gas Utilities, Financial Management, 113-
124.

Allen, J.W. and G. M. Phillips, 1998, “Corporate Equity Ownership and Product-Market
Relationships”, University of Maryland Working Paper.

Baker, George and Brian Hall, 1998, CEO Incentives and Firm Size, Working paper,
Harvard University.

Besen, S. and T. Krattenmaker, 1984, Misregulating Television: Network Dominance and
the FCC, Chicago: University of Chicago Press.

Campbell, Cynthia, and Charles Wasley, 1999, Stock-based Incentive Contracts and

Managerial Performance: The Case of Ralston Purina Company, Journal of Financial
Economics 51, 195-217.

Carol], Thomas, and David Ciscel, 1982, The Effects of Regulation on Executive
Compensation, The Review of Economics and Statistics 64, 505-509.

Cheung, S.N.S., 1983, “The Contractual Nature of the Firm, ” Journal of Law and
Economics 26, 1-21.

Core, John and Wayne Guay, 1998, Estimating the Incentive Effects of Executive Stock
Option Portfolios, University of Pennsylvania Working Paper.

Dun & Bradstreet, 1988-1999, Reference Book of Corporate Management.

Easterbrook, F. and D. Fischel, 1991, The Economic Structure of Corporate Law,
Cambridge: Harvard University Press.

Fama, Eugene F., 1980, “Agency Problems and the Theory of the Firm,” Journal of
Political Economy 88, 288-307.

Fama, Eugene F. and Michael C. Jensen, 1983, “Separation of Ownership and Control,”
The Journal of Law and Economics 26, 301 -325.

Federal Communications Commission, 1984, Attribution of Ownership Interests, 97 FCC
2d (1984) 997.

Federal Communications Commission, 1995, Report & Order, MM Docket 95-39; FCC
95-382 (Review of the Fin Syn Rules).

Federal Communications Commission, 1996, Memorandum Opinion and Order and
Notice of Apparent Liability, FCC 96-258, June 6, 1996.

Federal Communications Commission, 1999, Report & Order, FCC 99-207 .

Garen, John, 1994, Executive Compensation and Principal-Agent Theory, Journal of
Political Economy 102, 1 175-1 199.

Gilo, David, 1996, “Passive Investment as an Anticompetitive Commitment,” John M.
Olin Discussion Paper No. 189, Harvard Law School.

Gilson, Stuart, 1990, “Bankruptcy, Boards, Banks, and Blockholders: Evidence on
Changes in Corporate Ownership and Control When Firms Default,” Journal of
Financial Economics, 27, 355-387.

Gompers, Paul, and Josh Lerner, 1999, An Analysis of Compensation in the US. Venture
Capital Partnership, Journal of Financial Economics 51, 3-44.

Gould, W.W., 1992, Quantile Regression with Bootstrapped Standard Errors, Stata
Technical Bulletin 9, 19-21.

Grossman, Sanford and Oliver Hart, 1983, An Analysis of the Principal-Agent Problem,
Econometrica 51, 7-45.

Grossman, Sanford and Oliver Hart, 1986, “The Costs and Beneﬁts of Ownership: A
Theory of Vertical and Lateral Integration,” Journal of Political Economy 94, 691-719.

Hadlock, Charles and Gerald Lumer, 1997, Compensation, Turnover, and Top
Management Incentives: Historical Evidence,” Journal of Business 70, 153-187.

Hall, Brian and Jeffrey Liebman, 1998, Are CEOs Really Paid Like Bureaucrats?, The
Quarterly Journal of Economics 63, 653- 69].

Harris, M. and A. Raviv, 1988, “Corporate Control Contests and Capital Structure,”
Journal of Financial Economics 20, 55-86.

Head, 8., C. Sterling, and L. Schoﬁeld. 1994, Broadcasting in American: A Survey of
Electronic Media, Seventh edition, Boston: Houghton Mifﬂin Company.

122

Holderness, Clifford, Randall Kroszner, and Dennis Sheehan, 1999, Were the Good Old
Days that Good? Changes in Managerial Stock Ownership Since the Great Depression,
Journal of Finance 54, 435-469.

Holderness, C. and D. Sheehan, 1991, “Monitoring an owner: The case of Turner
Broadcasting”, Journal of Financial Economics 30, 325-346.

Holmstrom, Bengt, 1979, Moral Hazard and Observability, The Bell Journal of
Economics 10, 74-91.

Holmstrom, Bengt and Paul Milgrom, 1987, Aggregation and Linearity in the Provision
of Intertemporal Incentives, Econometrica 55, 303-328.

Hubbard, R. Glenn, and Darius Palia. 1995, Executive Pay and Performance: Evidence
from the US. Banking Industry, Journal of Financial Economics 39, 105-130.

Jensen, Michael C., 1989, “Eclipse of the Public Corporation,” Harvard Business Review
Vol. 5, 61-74.

Jensen, Michael and Kevin J. Murphy, 1990a, Performance Pay and Top-Management
Incentives, Journal of Political Economy 98, 225-264.

Jensen, Michael and Kevin J. Murphy, 1990b, CEO Incentives: It’s Not How Much, but
How, Harvard Business Review, May/June.

Joskow, Paul, Nancy Rose, and Andrea Shepard, 1993, Regulatory Constraints on CEO
Compensation, Brookings Papers: Microeconomics, 1-72.

Joskow, Paul, Nancy Rose, and Catherine Wolfram, 1996, Political Constraints on
Executive Compensation: Evidence from the Electric Utility Industry, RAND Journal of
Economics 27, 165-182.

Kole, Stacey and Kenneth Lehn, 1997, Deregulation, the Evolution of Corporate
Governance Structure, and Survival, American Economic Review 87, 421 -425.

Kole, Stacey and Kenneth Lehn, 1999, Deregulation and the Adaptation of Governance
Structure: The Case of the US. Airline Industry, Journal of Financial Economics 52, 79-
117.

Kopcke, R.W. and E. Rosengren (editors), 1991, Are the Distinctions between Debt and
Equity Disappearing?, Federal Reserve Bank of Boston, Conference Series #33.

Lafontaine, F. and SE. Masten, 1995, “Franchise contracting, organization, and
regulation: Introduction,” Journal of Corporate Finance 2, 9-3 7.

123

McClellan, S. and P. Albinisk, 1999, “It’s getting ugly”, Broadcasting & Cable (January
18), 11-24.

Mirrlees, J ., 1974, Notes on Welfare Economics, Information, and Uncertainty, in Essays
on Economic Behavior Under Uncertainty, ed. by M. Balch, D. McFadden, and S. Wu.

Mirrlees, J., 1976, The Optimal Structure of Incentives and Authority Within an
Organization, The Bell Journal of Economics 7, 105-131.

Murphy, Kevin J ., 1999, Executive Compensation, in Handbook of Labor Economics Vol
3, edited by Orley Ashenfelter and David Card.

National Association of Broadcasters, 1994, “A Financial Proﬁle of TV Station by
Network Afﬁliation: 1989-1993”.

National Regulatory Research Institute, 1999, Status of Electric Industry Restructuring.

Opler, T., 1993, “Controlling Financial Distress Costs in Leveraged Buyouts with
Financial Innovation,” Financial Management 22, 79-90.

Owen, B. and S. Wildman, 1992, Video Economics, Cambridge, MA: Harvard University
Press.

Parrino, Robert, 1997, CEO Turnover and Outside Succession: A Cross-Sectional
Analysis, Journal of Financial Economics 46,

Rajan, R. and L. Zingales, 1998, “Power in a Theory of the Firm,” Quarterly Journal of
Economics, 387-432.

Schaefer, Scott, 1998, The Dependence of Pay-Performance Sensitivity on the Size of the
Firm, Review of Economics and Statistics 80, 436-443.

Smith, C. and J. Warner, 1979, “On Financial Contracting: An Analysis of Bond
Covenants,” Journal of Financial Economics 7, 117-161.

Standard and Poor’s, 1995, S&P Compustat ExecuComp: The Executive Compensation
Database, Englewood, CO: McGraw-Hill.

Tirole, J ., 1992, “Collusion and the theory of organizations,” in Advances in Economic
Theory. Sixth World Congress, Volume 11, edited by Jean-Jacques Laffont, Cambridge:
Cambridge University Press.

Watts, Ross and Jerold Zimmerman, 1986, Positive Accounting Theory. Englewood
Cliffs, NJ: Prentice-Hall.

124

Weston, F., K. Chung, and S. Hoag, 1990, Mergers. Restructuring. and Corporate
Control, Englewood Cliffs, N.J.: Prentice-Hall.

125

'WWWWWWWP'
||1|1:111.111111111111I1111111111111
31293 021019090