Disentangling CEO compensation: A simultaneous examination of time, industry, and firm‐level effects

• Author: Sujit Sur,James Cordeiro,Michel Magnan
• DOI: http://doi.org/10.1002/cjas.1304
• Published date: 01 March 2015
• Date: 01 March 2015

Disentangling CEO compensation: A

simultaneous examination of time, industry,

and firm-level effects

Sujit Sur*

Carleton University

Michel Magnan

Concordia University

James Cordeiro

SUNY Brockport

Abstract

Despite decades of research, how CEO compensation is de-

termined remains an enigma. Drawing on agency, manage-

rial hegemony, and institutional theoretical perspectives, we

use hierarchical linear modelling—a multilevel analytic

technique—to examine how f‌irm-, industry-, and time-level

effects drive CEO compensation in US corporations. Results

show that while cash salary is mostly driven by f‌irm-specif‌ic

factors, equity-based compensation responds to time-level

effects with f‌irm- and industry-level effects playing a mar-

ginal role. We argue that such evidence is consistent with

the institutionalization of the CEO compensation determina-

tion process through the widespread adoption of benchmark

peer-group comparisons. Such practices underlie economy-

wide changes in CEO compensation that are increasingly

disconnected from other fundamental f‌irm- or industry-

specif‌ic factors. Copyright © 2015 ASAC. Published by

John Wiley & Sons, Ltd.

Keywords: CEO compensation, variance decomposition,

multilevel analysis, hierarchical linear modelling, level-wise

compensation determinants

Résumé

Malgré des décennies de recherches, le mode de compensa-

tion des PDG demeure énigmatique. La présente étude uti-

lise le modèle linéaire hiérarchique –technique d’analyse

multiniveaux –pour examiner comment des actions au

niveau de l’entreprise, de l’industrie et du temps inf‌luencent

la compensation des PDG au sein des compagnies

américaines. Elle s’appuie sur l’agence, l’hégémonie

managériale et les perspectives théoriques sur l’institution.

Les résultats montrent que si la rémunération est motivée

surtout par des facteurs propres à l’entreprise, la compen-

sation basée sur l’équité par contre est inf‌luencée par des

facteurs au niveau du temps, tandis que les facteurs au

niveau de l’entreprise et de l’industrie ne jouent qu’un

rôle marginal. Ces résultats cadrent bien avec

l’institutionnalisation du processus de détermination de la

compensation des PDG consécutive à l’adoption, sur une

grande échelle, de l’indice de référence des comparaisons

avec les groupes de pairs. Ce sont de telles pratiques qui

sous-tendent les changements qui touchent à la compensa-

tion des PDG à l’échelle de l’ensemble de l’économie et

qui sont de plus en plus en porte-à-faux avec les autres

facteurs fondamentaux caractéristiques de la f‌irme ou de

l’industrie. Copyright © 2015 ASAC. Published by John Wi-

ley & Sons, Ltd.

Mots-clés : compensation des PDG, décomposition des

variances, analyse multiniveaux, modèle linéaire

hiérarchique, déterminants de la compensation par niveau

The authors thank participants at the Association Académique Internationale de gouvernance and ASAC conferences for their comments as well as workshop

attendees at Institut universitaire de Monaco and Université de Nice-Sophia Antipolis. Michel Magnan acknowledges f‌inancial support from the Stephen A.

Jarislowsky Chair in Corporate Governance, the Institute for the Governance of Private and Public Organizations, and the Lawrence Bloomberg Chair in Accoun-

tancy and SSHRC.

*Please address correspondence to: Sujit Sur, Carleton University, 1124 Colonel By Drive, Ottawa, ON, K1S 5B6, Canada. Email: Sujit.Sur@carleton.ca

Canadian Journal of Administrative Sciences

Revue canadienne des sciences de l’administration

32:30–46 (2015)

Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/CJAS.1304

Can J Adm Sci

32(1), 30–46 (2015)Copyright © 2015 ASAC. Published by John Wiley & Sons, Ltd. 30

The scale and scope of CEO compensation now make it

an economic phenomenon in its own right, and thus a con-

cern for shareholders and stakeholders alike. For example,

Bebchuk and Grinstein (2005) found that in US f‌irms, the

magnitude of executive compensation relative to f‌irm’s earn-

ings increased from around 5% in 1993–1995 to around

10% in 2001–2003. Looking at the period from 1978 to

2011, Mishel and Sabadich (2012) observed that the in-

crease in CEO compensation, which they put at more than

725%, is substantially greater than stock market and corpo-

rate earnings growth. From a practical angle, compensation

may be viewed as the smoking gun of board governance pro-

cesses (Bebchuk & Fried, 2004). From an academic angle,

there has been extensive research on the determination of

the components and level of CEO compensation over the

years, mostly relying on cross-sectional designs, with no

def‌inite conclusion yet to be reached.

Here we aim to address what def‌ines the evolution over

time in CEO compensation within f‌irm-, industry-, and

temporal-specif‌ic trends. To this end, we conceptually and

methodologically revisit the ways in which executive com-

pensation determination have traditionally been examined.

Conceptual arguments explaining the underlying trends in

CEO compensation typically rest on two contrasting per-

spectives (Frydman & Jenter, 2010). On one hand, the man-

agerial hegemony perspective posits that powerful managers

are able to set their own pay, thus inf‌lating compensation

levels in ways that are not in shareholders’interests (Crystal,

1991). Ineff‌icient compensation contracting between the

f‌irm and CEO results from CEOs being in control of the

pay setting process. Bebchuk and Fried (2004) argued that,

in practice, there has been a decoupling of pay and perfor-

mance (Bertrand & Mullainathan, 2001). Hence, the increas-

ing trend in CEO compensation may be an illustration of the

executives’ability to garner a greater share of f‌irm value as a

result of weak governance and a broken process (Stiglitz,

2012, pp. 309, 316). On the other hand, high and rising ex-

ecutive compensation levels may ref‌lect optimal contracting

in a competitive market that seeks and rewards managerial

talent, while mitigating managerial opportunism and malfea-

sance. Contracts between shareholders—or the board on

their behalf—and a CEO aim to minimize agency costs

given a set of constraints while steering the CEO toward ac-

tions benef‌itting shareholders (Holmstrom, 2005).

However, being micro-economic oriented (i.e., assessing

only f‌irm and industry specif‌icinf‌luences), hegemony and

agency perspectives do not offer a comprehensive explanation

for the evolution of CEO compensation across f‌irms and

industries. Frydman and Jenter (2010) tentatively suggested

that a third, intermediate perspective is needed. In response

to their call, we consider that the institutionalization perspec-

tive may explain macro trends in CEO compensation. For in-

stance, institutionalization maps well with the peer grouping

process—that is, the normative practice that determines how

rewards are shared between CEOs and shareholders in an

uncertain and ambiguous bargaining environment. Hence, in-

stitutionalizedpressures exerted on setting compensation span

time and extend beyond f‌irm-and industry-specif‌icinf‌luences

(Balkin, 2008).

Supporting our three-pronged conceptual perspective,

we apply multilevel analysis as it allows a f‌iner analysis of

trends underlying CEO compensation. While OLS regres-

sions capture and ref‌lect cross-sectional and longitudinal

trends in the underlying data, a multilevel analysis allows as-

sessment of additional dimensions. For instance, in an anal-

ysis of CEO compensation data going back to 1936,

Frydman and Jenter (2010) showed that the relation between

f‌irm size and executive compensation, while weak until the

late 70s, is now much stronger. In addition, they found that

compensation mix is migrating toward stock options and

equity-based incentives and away from salary and other

components. However, they provided no clues as to the po-

tential drivers underlying such trends and were thus unable

to parse out the relative importance of the factors at different

levels of analyses, namely temporal, f‌irm, and industry.

We simultaneously consider the three conceptual per-

spectives and utilize multilevel modelling to analyze com-

pensation for all available US CEOs in the Standard &

Poor’s Execucomp database over the 1997–2006 period. Re-

sults show that while f‌irm-specif‌ic factors primarily drive

cash salary, equity-based compensation responds essentially

to time-level effects (independent of f‌irm performance).

Since the major part of CEO compensation is now equity-

based, our f‌inding suggests an institutionalization process

that is potentially driven by peer group comparisons.

We synthesize the paper’s contribution along the fol-

lowing lines. First, prior research typically adopts a single

conceptual perspective, typically either managerial power

(managerial hegemony) or optimal contracting within a mar-

ket for talent (agency). Frydman and Jenter (2010) con-

cluded that, while useful, both perspectives provide only a

partial explanation of the available evidence. We consider

that the integration of these perspectives, with the addition

of the institutionalization perspective, has the potential to

contribute to the ongoing debate about CEO compensation.

Second, Tosi, Werner, Katz, and Gomez-Mejia (2000)

concluded that f‌irm size overwhelms f‌irm performance as a

determinant of CEO compensation, while Frydman and Saks

(2010) observed a strong relation between f‌irm size and

compensation in recent years. Yet, their reliance on linear re-

gression models such as OLS precludes def‌inite conclusions

as cross-sectional and longitudinal trends are combined and

are diff‌icult to disentangle, with the procedure and sequence

of variables considered affecting their relative importance.

In contrast, multilevel analysis allows for assessment of the

relative importance of each level of analysis (i.e., f‌irm, indus-

try, and temporal) and specif‌ic variables within each level,

holding constant the other levels and variables (i.e., the esti-

mation is not conditional on the sequence of analysis, as it

would be in step-wise regressions). Furthermore, multilevel

DISENTANGLING CEO COMPENSATION SUR ET AL.

Can J Adm Sci

32(1), 30–46 (2015)Copyright © 2015 ASAC. Published by John Wiley & Sons, Ltd. 31