Are Incentive Contracts Rigged By Powerful CEOs?
∗
Adair Morse, Vikram Nanda, and Amit Seru
August, 2009
∗
Acknowledgements:We thank Lawrence Brown, Robert Bushman, Daniel Ferreira, Yaniv Grinstein, Camp-
bell Harvey (the editor), Charles Hadlock, Steve Kaplan, Joshua Rauh, Morten Sorensen, Anjan Thakor, the
associate editor, an anonymous referee, and the seminar participants at Arizona State, Alabama, Chicago GSB,
Georgia State, Georgia Tech, Michigan, Vienna, Washington St. Louis (Olin), American Finance Association
2006 meetings, European Finance Association 2006 meetings and FEA conference at UNC for their helpful com-
ments and suggestions. The first and third authors are from University of Chicago, Booth School of Business;
e-mail: adair.morse@chicagobooth.edu, amit.seru@chicagobooth.edu. The second author is from Georgia Tech;
email: vikram.nanda@mgt.gatech.edu. We are responsible for all the errors.
Are Incentive Contracts Rigged By Powerful CEOs?
Abstract
We argue that powerful CEOs induce their boards to shift the weight on performance measures to-
wards the better performing measures, thereby rigging the incentive part of their pay. The intuition
is developed in a simple model in which some powerful CEOs exploit superior information and lack
of transparency in compensation contracts to extract rents. The model delivers an explicit structural
form for the rigging of CEO incentive pay along with testable implications that rigging is expected to
(1) increase with CEO power; (2) increase with CEO human capital intensity and uncertainty about a
firm’s future prospects; and (3) negatively impact firm performance. Using measures of CEO power and
board independence on a large panel of firms in the U.S., we find support for these predictions. Rigging
accounts for at least 10% of the sensitivity of compensation to performance measures and is increasing
in CEO human capital and volatility of a firm’s future prospects. Moreover, the portion of incentive
pay that is predicted by power is associated with negative subsequent future stock performance of the
order of 0.8% and operating performance of 7.5% per year. Overall, the results provide evidence against
the agency substitution theory and support instead the entrenchment skimming theory. Our results
advo c ate for requiring ex ante disclosure of incentive contract terms.
I. Introduction
Incentive contracts are generally thought to be one of the governance cornerstones available
to a board for aligning managers’ interests with those of shareholders (e.g., Holmstrom, 1979).
Such contacting may be especially important for inducing CEOs to m aximize shareholder value
when agency problems are severe (Eaton and Rosen, 1983). In this paper, we cast a negative
light on the extent to which incentive contracting can counter agency problems and substitute
for good governance arrangements. We argue that when a board is relatively weak, incentive
contracting may itself become a source of s hareholder value leakage. In particular, we explore
the possibility of CEOs influencing their incentive pay by manipulating the manner in which
performance is measured.
The intuition is straightforward: Boards use a variety of measures to gauge CEO perfor-
mance. At the same time, a CEO will often know when her performance along one dimension
is going to be stronger than along another dimension. The CEO can use her influence over the
board, either directly or indirectly through other insiders, to slant her performance assessment
toward the better-performing measures. More specifically, insiders may induce the board to
make the CEO’s incentive pay more dep e ndent on measures of performance that they know
are doing well. The example of Home Depot CEO Robert Nardelli can help fix the idea. A
footnote in Home Depot’s 2004 proxy statement said that his long-term incentive pay would
be calculated by looking at the total return to shareholders over the three-year performance
period and comparing that to an established peer group of retailers. By that measure, Nardelli
had bombed. Home Depot’s stock fell since he took over in December 2000; meanwhile, rival
Lowe’s shares had soared. In 2005’s proxy, however, the footnote changed: he was to receive
incentive pay if the company achieved specified levels of average diluted earnings per share –
a measure by which Home Depot lo oked far more successful. We term this action of ex post
shifting the weight toward better performing measures rigging of the incentive contract.
Given the substantial literature on CEO compensation, it may be helpful to briefly distin-
guish our notion of contract manipulation from other analyses of CEO rent extraction. Adams,
Almeida and Ferreira (2005) find a direct relationship between CEO power and higher compen-
sation. We likewise find evidence that powerful CEOs get paid more. However, our focus is on
how CEOs use their power to manipulate incentive contracts and generate rents from what is
ostensibly performance based pay - over-and-above the level effect of power on compensation.
We build on this existing literature by presenting evidence that power weakens the effectiveness
of incentives. Our paper also differs from agency studies looking at the relationship b etween
power and the use of incentive pay versus c ash compensation (e.g., Mehran, 1995). Unlike
these papers, our theoretical argument is that incentive contracts are, at bes t, only partially
effective in compensating for weak board governance. Moreover, our empirical results imply
that to isolate the full effect of actual, nonrigged incentives, it is important to remove portions
of incentive pay that are rigged and have the effect of weakening CEO incentives.
Our rigging argument is also quite distinct from the idea that CEOs are compensated
1
for luck (Bertrand and Mullainathan, 2001; Garvey and Milbourn, 2006).
1
The notions are
similar to the extent that rigging and pay from luck do not reflect CEO skill or effort and both
happen in weak corporate governance environments. Otherwise, the notions are very distinct.
In the luck studies, boards pay CEOs for the firm doing well for reasons completely outside the
control of the firm: hence, it is the measure of performance (e.g., no adjustment for common
industry factors) that is wrong. In our rigging story, on the other hand, insiders undertake
active, intentional manipulation to shift weights toward firm-specific measures of performance
that are doing better – i.e., the measures of performance may be appropriate, but it is the
structure of the incentive contract that is manipulated. Our empirical design distinguishes
between these stories and finds strong support for rigging after removing any luck effect.
Finally, rigging is facilitated by the lack of complete disclosure about incentive contract
terms ex ante. Hence, we believe that our paper is the first academic study to provide empirical
evidence on the need for greater ex-ante disclosure of CEO incentive contracts. We note that
there is s ome recognition of this issue among compensation consultants and in the media. For
instance, a recent article in WSJ (15 March, 2006) reports that “Smart companies are increas-
ingly turning to pay for performance. But even that approach has its pitfalls. Performance
measures are seldom made public – for competitive reasons, companies say – and are open to
manipulation”. In the same vein, a recent study by a large compensation consulting firm, Wat-
son Wyatt, indicates the difficulty outside shareholders have in evaluating CEOs performance
reward by reporting that 46% percent of the top 100 US companies did not disclose the actual
goals on which they based rewards under their 2006 annual incentive plans.
The paper begins with a simple model that gives us an explicit structural form on the
incentive portion of the CEO wage function and guides our empirical tests. A key feature of
the model is the notion of a compromised board. We assume that compromised boards act in
the interest of CEOs and have few reservations about increasing their compensation — as long
as the payments do not reveal the board to be compromised and trigger shareholder outrage a
la Bebchuk, Fried and Walker (2002). The prior probability of a compromised board increases
with CEO power, though outsiders do not know whether a particular board is compromised
or not. In equilibrium, compromised boards conceal their type by using compensation policies
that mimic the contracts offe red by independent boards.
We make use of a familiar one-period contracting environment in which a risk-averse CEO
provides costly but unobservable effort. We extend the model by allowing for a state variable
that is observed by the board, though not by outsiders. Independent boards offer the CEO an
optimal contract conditional on the state. The set of possible contracts that could be offered
by the independent boards forms the complete set of contracts that shareholders exp ec t to
see in equilibrium and, thereby, limits the contract choices of a compromised board as well.
Despite the pooling in compensation arrangements , we show that asymmetric information and
the lack of transparency in CEO compensation arrangements can allow som e powerful CEOs
1
Oyer (2004) shows that, in contrast to the pay for luck argument, an optimal contract could allow the CEOs
to be compensated for luck type factors due to increase in their outside options.
2
to rig their compensation without triggering shareholder outrage. Specifically, the flexibility
provided to the compromised board in the form of multiple performance measures allows some
powerful CEOs to use ex post information to extract rents by manipulating the terms of their
incentive based pay. Rigging is thus a yet undiscussed mechanism falling under the concepts
of Bebchuk and Fried’s (2004) compensation camouflaging.
The model provides us with rich empirical predictions. First, rigging of incentive pay should
increase with CEO power. Second, the severity of rigging should increase with the human capi-
tal intensity of the CEO and uncertainty about the firm’s future prospects, reflecting the extent
to which performance weights can be shifted. Finally, firm performance should be decreasing
in the amount of rigging. The intuition is that rigging distorts the incentive mechanism and,
by weakening the CEO’s ex ante incentives, negatively impacts firm performance and value.
To test these predictions, we use a panel of 1,119 firms in the U.S. over the period 1993-2003.
Our empirical tests use industry adjusted performance measures to ensure that the effects we
document are after controlling for the type of luck factors in Bertrand and Mullainathan (2001)
and Garvey and Milbourn (2006). We collect measures for power directly from firm proxies
which is richer and more extensive than the readily available dataset choices. This enables
us to look at multiple facets of CEO power. For our measures of CEO influence over the
board, we c reate an index of CEO personal power and two measures of board weakness based
on the prop ortion of insiders and the percentage of board appointed by the CEO. Overall,
consistent with predictions, we find that CEOs with power rig incentive compensation: The
pay of powerful CEOs has greater sensitivity to the better performing of stock returns and
return on assets in a given period. Additionally, rigging is significantly higher when the human
capital intensity of the CEO and uncertainty about the firm’s future prospects is high.
Our rigging re sults hold under a number of robustness tests and alternative regression
specifications. In particular, rigging is apparent in the data using firm fixed effects with
alternative dynamic error structures. The magnitude of rigging is economically large and
accounts for about 10-30% of the incentive pay sensitivity to performance. Moreover, we show
that rigging is not limited to periods with high stock or accounting performance, and that
CEOs rig compensation contracts to take advantage of both stock and accounting returns. In
additional tests, we also find that factors associated with stronger governance – such as greater
institutional holdings – may moderate the amount of rigging.
Finally, in line with our expectations, we s how that rigging is negatively related to future
firm performance. To conduct this test, we project incentive compensation onto power variables
and economically-motivated variables. We find that the proportion of incentive pay predicted
by power variables has a significant negative association with subsequent firm value (Q-ratio),
operating return on as se ts and stock performance. A firm with one standard deviation above
mean rigged incentive pay faces a drop of 4.8% in firm value, 7.5% in operating return on
assets and about 0.82% drop in four factor risk-adjusted sto ck returns relative to the sam ple
average, over the subsequent two year period.
3
