Complexity of CEO Compensation Packages

Ana Albuquerque
Boston University Questrom School of Management
Catlica-Lisbon Business & Economics
albuquea@bu.edu

Mary Ellen Carter

Carroll School of Management
Boston College
maryellen.carter@bc.edu

Luann J. Lynch
Darden Graduate School of Business
University of Virginia
LynchL@darden.virginia.edu

November 2015

Abstract

We examine the extent to which the complexity of CEO compensation packages is consistent with the
efficient contracting or the rent extraction view. Using samples of firms from ExecuComp and Incentive
Lab from 2006 2012, we construct both an ex-post and an ex-ante measure of the complexity of CEO
compensation and examine their relation with both economic determinants of contract complexity and
excess CEO pay. We find that proxies for firm complexity are associated with more contract complexity,
after controlling for other economic characteristics, CEO characteristics, and the number of consultants
hired by the firm. In addition, we find that excess CEO pay is positively related to the complexity of CEO
compensation packages. As a whole, our results provide support for both the efficient contracting view
and the rent extraction view of the design of CEO compensation packages. To validate the rent extraction
view, in further analysis we show that the component of contract complexity related to excess pay is
associated with poor future performance. Finally, we document that contracts have become more complex
after 2009 but that they have also become clearer from a linguistic perspective.
_____________________________________________________________________________________
We gratefully acknowledge the financial support of Boston College, Boston University, Catlica-Lisbon, and the
University of Virginia Darden School Foundation. We appreciate helpful comments from Eli Bartov and seminar
participants at Bocconi University and Erasmus University Rotterdam. We thank Steve Lacey for his help
computing our readability scores, and Helen Erwin and Brendan Robertson for research assistance.

JEL classification: J33, M52

Keywords: CEO compensation, contracts, complexity, incentives

1. Introduction

In 2006, the Securities and Exchange Commission (SEC) began requiring expanded

disclosure of executive compensation contracts. These new disclosures provide investors with

more visibility into the many performance measures, benchmarks, and performance horizons

being used in determining executive pay. While the expanded disclosure requirements increased

transparency around executive compensation packages, potentially allowing for easier detection

of compensation arrangements that are not in the best interest of shareholders, investors have

expressed concerns about the increasing complexity of these packages (Johnson, 2011; KPMG

2011; Conference Board 2013). With more complex contracts (e.g., using several performance

metrics, measured over different time periods and benchmarked to different peer groups), it can

be difficult for outsiders to detect rent extraction and the extent to which compensation is not tied

to the managers own performance (Bebchuk and Fried, 2003). In fact, SEC commissioner Troy

Paredes warns about information overload by stating (I)ronically, if investors are overloaded,

more disclosure actually can result in less transparency and worse decisions.1

While the complexity of compensation packages may offer the opportunity for managers

to obscure undeserved pay from the firm (the rent extraction view), it is also possible that the

complexity of executive compensation packages reflects the scope and complexity of the role of

the executive in managing a complex business organization (the efficient contracting view).

Complex compensation contracts can be the result of providing executives with incentives that

match the complexity of the business operations they manage (Kole, 1997). For example, the

business environment can be such that it is important to promote intangible assets such as

innovation, employee knowledge, and process improvement for which the use of non-financial

1
The Twelfth Annual A.A. Sommer, Jr. Lecture on Corporate, Securities and Financial Law, on October 27, 2011

1

performance metrics can be more suitable. The extent to which compensation complexity reflects

either view is an empirical question.

Using a sample of firms from ExecuComp and a sub-sample of those firms also in

Incentive Lab from 2006 to 2012, we construct two measures of CEO compensation complexity:

(1) from ExecuComp, we compute a Herfindahl Index based on the proportion of total

compensation accounted for by each component of compensation, and (2) from Incentive Lab,

we calculate an index of complexity that increases with the number of components of the

compensation contract, types of performance-based features, number of performance measures,

and number of periods over which performance is measured. Then, we examine the relation

between these measures of complexity and both economic determinants (the efficient contracting

view) and excess CEO pay (the rent extraction view).

We find that the complexity of CEO compensation packages is related to several

economic characteristics that reflect the scope and complexity of the organizations. Larger firms,

firms with more business segments, and firms with foreign operations have more complex CEO

compensation packages. Firms with more growth opportunities and more volatile operations

have less complex plans. While the latter result seems counterintuitive, the compensation plans

for firms with more growth opportunities tend to include a large equity component, consistent

with equity being used to align incentives of CEOs with those of investors in firms where

monitoring of management is more difficult. And the plans for firms with more volatile

operations tend to include a large salary component, consistent with fixed compensation being

more important than variable pay. This is consistent with the predictions of agency theory, where

risk-averse CEOs discount equity-based pay in the presence of firm risk, making it costly for

shareholders to rely on incentive pay (Holmstrom and Milgrom, 1987). The presence of these

2

dominant components leads to less complex packages. In addition, controlling for the impact of

CEO characteristics on contract complexity, we find that CEO tenure, proximity to retirement,

and founder CEO status are associated with less complex contracts, possibly reflecting lower

agency problems with long-serving CEOs holding large ownership stakes in the firm (e.g.,

Hermalin and Weisbach, 1998; Fahlenbrach, 2009). Finally, we find that excess CEO pay is

positively related to the complexity of CEO compensation packages, offering evidence consistent

with the rent extraction view and suggesting that more complex compensation packages are

accompanied by more excessive pay. Overall, our results provide support for both the efficient

contracting view and the rent extraction view in the design of CEO compensation packages.

To validate our findings of rent extraction, we test whether the component of contract

complexity related to excess pay is associated with future poor performance. If the excess pay

component of contract complexity captures the CEOs attempt to obfuscate rent extraction, then

it would be negatively associated with future performance. Our results confirm this prediction.

Finally, in descriptive tests, we analyze the level of contract complexity over time and

show that while contracts have become more complex recently, they have also become less

associated with firm characteristics but more associated with linguistic simplicity.

In the midst of heightened scrutiny by regulators, increased shareholder activism, and

more interest in the media, we contribute to the executive compensation literature by providing

new evidence on the complexity of compensation arrangements. Prior research has examined

whether the level of executive compensation is related to firm complexity (Black, Dikolli and

Dyreng, 2014) or whether the form of executive pay is related to firm characteristics (Kole,

1997). We extend this literature by examining the complexity in the compensation package itself,

rather than its level or form. We introduce two new measures of compensation contract

3

complexity that explore the details of compensation contracts disclosed on proxy statements. We

take advantage of an increase in compensation disclosure mandated by the SEC since 2006 and

the change in compensation contract structure (e.g., performance share units have become a

popular way of providing incentives to executives) to provide new measures that werent

previously available. In addition, we rely on two large samples of firms (2,200 firms from

ExecuComp and 1,100 firms from IncentiveLab) allowing us to analyze the drivers of

compensation contract complexity using a broader and more recent sample. Finally, we also

contribute to the literature examining disclosure complexity. While studies have considered the

readability of annual reports and their relation to performance (e.g. Li, 2008), we are the first to

study the readability of the Compensation Disclosure and Analysis (CD&A) section of the proxy

statement as it relates to the complexity of compensation.

The remainder of the paper is organized as follows. Section 2 offers background and

develops hypotheses. Section 3 discusses the sample, data, and methodology. Section 4

provides results. We provide further analyses of the rent extraction view in Section 5 and

complexity through time in Section 6. Section 7 concludes.

2. Background and hypothesis development

Two contrasting hypotheses may explain the complexity inherent in executive

compensation packages. The efficient contracting hypothesis would predict that the complexity

of compensation contracts arises as a result of firms attempts to design efficient compensation

contracts that reflect the economics of the situation the executive is charged with managing;

different executives assume roles with differing scopes and levels of complexity, and as a result,

are compensated through compensation plans that differ in their complexity. Prior research has

found that the level of executive pay varies with characteristics associated with firm complexity,

4

including size (see Smith and Watts 1992, Gaver and Gaver 1993, among others), growth

opportunities (Smith and Watts 1992), the volatility of the firms operating environment

(Demsetz and Lehn 1985), and industry and geographic diversification (Rose and Shepard 1997;

Duru and Reeb 2002; Bushman, Chen, Engel, and Smith 2004; Black, et al. 2014). While these

studies consider how the level of executive pay varies with proxies for complexity, they do not

consider whether the complexity of the contract itself is related to firm complexity.

Kole (1997) analyzes the use of stock options, restricted stock, and long-term

performance plans (now labelled non-equity incentive plans) for 371 Fortune 500 firms in 1980.

She finds that there is greater use of equity pay and lower use of performance plans in high R&D

firms, consistent with stock-based pay dominating accounting-based pay (typical in performance

plans) for research intensive firms. She also finds longer vesting periods in high R&D firms

consistent with greater uncertainty surrounding project payoffs and importance of specialized

knowledge in certain firms. We differ from Kole (1997) in several important ways. First, in her

study, contract complexity is defined as the form of pay and the contract length. We take a more

detailed look at contract complexity by examining many dimensions along which contracts can

be complex. Second, we consider a significantly larger and more comprehensive sample, over a

longer time period, in more recent years in which contracts have become sufficiently complex to

become the subject of scrutiny by regulators, shareholders, and the media alike.

In contrast to efficient contracting, the rent extraction view of executive compensation

(see, e.g., Bebchuk and Fried 2003) suggests that compensation packages will be set in a way to

obscure or camouflage the rationale behind them; the complexity of compensation packages can

limit the ability of outsiders to detect the extraction of rents by executives. In such cases, CEOs

may take advantage of complex arrangements to increase their compensation relative to what

5

shareholders expectation of pay given the measures of performance they can observe.

Consistent with firms attempting to obfuscate rent extraction, Laksmana, Tietz, and Yang (2012)

document an association between positive excess pay and CD&A readability. Bebchuk and Fried

(2003) suggest that firms use pay practices that make less transparent both the total executive

compensation and the extent to which that compensation is tied to his performance. Some have

asserted that shareholders may not be able to understand the complex compensation plans found

in many firms (e.g., Walker 2011; Bebchuk and Fried 2004; Hoffman, 2015). Indeed, results in

an experimental study by Gillenkirch, Hendriks, and Welker (2013) suggest that shareholders are

better able to anticipate the incentive effects of compensation when compensation packages are

less complex. We extend this literature by examining contract complexity directly.

If the efficient contracting explains the complexity of executive compensation packages,

then we expect compensation complexity to be greater in firms with greater organizational

complexity, as reflected in firm size, growth opportunities, volatility of operating environment,

and diversification. If the rent extraction is associated with the complexity of executive

compensation packages, then we expect compensation complexity to be greater in firms with

higher excess pay, as it is those firms that have a greater ability to obscure how pay is set.

3. Research Design

3.1 Sample and data

We use two samples in our study. Our first sample includes all firms in the ExecuComp

database in 2006 - 2012 that have sufficient data to compute the variables in our initial regression

of the determinants of compensation contract complexity (2,003 firms; 10,785 firm-years). Our

second sample includes all firms on the Incentive Lab database in 2006 - 2012 that have

sufficient data to compute the variables in our initial regression of the determinants of

6

compensation contract complexity (1,027 firms; 5,711 firm-years).2 We begin our sample period

in 2006 because that is the first year the SEC required increased compensation disclosures and

because ExecuComp changed its definition of total compensation in that year, making

comparisons with prior periods harder to interpret.

We obtain financial statement data from Compustat, stock return data from CRSP, and

executive compensation data from ExecuComp and Incentive Lab.

3.2 Methodology

3.2.1 Measures of compensation contract complexity

We construct two alternative measures of the complexity of compensation contracts.

First, using data from ExecuComp, we calculate a Herfindahl Index based on the proportion of

total compensation (ExecuComp variable TDC1) accounted for by each component of

compensation, as follows:

TDC1_HHjt = (SALARYjt/TDC1jt) 2 + (BONUSjt/TDC1jt) 2 + (NONEQ_INCENTjt/TDC1jt) 2 +

(OPTION_AWARDS_FVjt/TDC1jt) 2 + (STOCK_AWARDS_FVjt/TDC1jt) 2 +
(OTHCOMPjt/TDC1jt) 2 + (DEFER_RPT_AS_COMP_TOTjt /TDC1jt) 2

TDC1_COMPLEXjt =1- TDC1_HHjt

Where:

TDC1jt = total compensation for CEO of firm j in year t

SALARYjt = salary for CEO of firm j in year t
BONUSjt = cash bonus for CEO of firm j in year t
NONEQ_INCENTjt = non-equity incentives for CEO of firm j in year t
OPTIONS_AWARDS_FVjt = fair value of option awards for CEO of firm j in year t
STOCK_AWARDS_FVjt = fair value of stock awards for CEO of firm j in year t
OTHCOMPjt = other compensation for CEO of firm j in year t
DEFER_RPT_AS_COMP_TOT jt = aggregate amount of deferred compensation earnings that were
reported as compensation for CEO of firm j in year t.

2
The Incentive Lab database, compiled by the data and analytics firm Incentive Lab from firms proxy statements,
contains data regarding performance measures, performance goals, vesting, and other data for incentive awards to
named executive officers.

7

Higher values of TDC1_HH indicate more concentration or less complexity in compensation

contract, as a higher value suggests that one or a few components dominate the pay package

(e.g., a contract where 100% of compensation is from a fixed salary would yield a value of 1),

while a lower value suggests that each component contributes relatively equally to total pay. In

order to have a variable whose value increases with compensation contract complexity, we create

the variable TDC1_COMPLEX that is equal to one minus TDC1_HH. As a result, higher

dispersion of total compensation among different components of pay is associated with a higher

value for contract complexity (TDC1_COMPLEX). Because this measure is computed using

compensation granted or earned in a year, this is an ex-post measure of complexity.

Second, using data from Incentive Lab, we calculate an index of complexity

(COMPLEX_IL) based on the number of components determining the potential payouts from

incentive-based compensation contracts, as follows. We include one point for each type of

compensation (short-term cash bonus, long-term cash bonus, restricted stock, stock options), as

Kole (1997) suggests that the form of compensation is one aspect of complexity. Then, for each

component, separately, we include an additional point for each of the following: the award has

absolute performance conditions, the absolute performance condition is based on more than one

performance measure, the absolute performance condition is based on performance over more

than one time period, the award has relative performance conditions, the relative performance

condition is based on more than one performance measure, and the relative performance

condition is based on performance over more than one time period. These features have not been

considered in the literature related to compensation contract complexity to date.

Agency theory predicts that compensation contracts will be tied to performance measures

that capture the CEOs effort (Holmstrom, 1979). Prior research has documented that executive

8

compensation contracts are tied to accounting- and stock-based measures (see for example

Lambert and Larcker, 1987; Sloan, 1993; Murphy, 2000). Research has also shown that

non-financial performance measures are used to determine executive incentive pay when

financial measures are noisier, when growth opportunities are larger, and when innovation is

more important (see, for example, Bushman, Indjejikian, and Smith, 1996; Ittner, Larcker and

Rajan, 1997). Thus, more complex firms are likely to use more performance measures to capture

different aspects of the CEOs actions. Further, complex firms are more likely to consider both

absolute and relative performance conditions. While relative performance insulates the CEO

from uncontrollable events, more complex firms may have a more difficult time identifying

similar peer firms. Finally, more complex firms may use multiple measurement periods. If

contract length is associated with the realization of project payoffs (Kole, 1997), complex firms

may consider multiple periods to cover the complexity inherent in their businesses.

A higher value of COMPLEX_IL is associated with higher compensation contract

complexity. For example, a more complex contract will use more forms of incentive pay, that

require meeting more than one performance measure, and that computes performance over more

than one time period; a less complex contract will use fewer forms of incentive pay that have

only time-vesting provisions. Because this measure is computed using incentive compensation

characteristics from which realizations will be computed, this is an ex-ante measure of

complexity. Appendix A summarizes this computation in tabular form. Appendix B provides an

example from DEF14-A to illustrate how the different characteristics of the compensation

contract are used to compute the COMPLEX_IL measure.

Figure 1 and Table 1 present descriptive statistics related to our measures of

compensation contract complexity. As shown in Table 1 Panel A, the mean (median) value of

9

TDC1_COMPLEX is 0.569 (0.616), with a standard deviation of 0.167. By construction, the

value of TDC1_COMPLEX can range from 0 to 0.8573; its value at the first and ninety-ninth

percentile is 0.1316 and 0.7733, respectively, showing a wide range of contract complexity

across firms. The mean (median) value of COMPLEX_IL is 4.106 (4.000), with a standard

deviation of 1.998. As shown in Appendix A, the value of COMPLEX_IL can be as high as 28.

However, a value that high would require that the firm award all four types of incentive

compensation, with both absolute and relative performance conditions within each type, based on

a more than one performance measure over more than one of time period. As this rarely occurs in

practice, it is not expected that the range in values of this variable reach its maximum value.

Indeed, its value at the first and ninety-ninth percentile is 0 and 9, respectively. Our contract

complexity measures are positively correlated (correlation of 0.35, p<0.05) yet seem to capture

different dimensions of contract complexity as the correlation is moderate.

Figure 1 graphically presents our measures by year over the sample period. The

complexity of compensation packages increased steadily over the sample period, consistent with

claims advanced in the popular press. COMPLEX_IL increased from 3.71 in 2006 to 4.64 in

2012; TDC1_COMPLEX increased from 0.566 in 2006 to 0.577 in 2012. To further investigate

whether the increase is statistically significant, we compare the complexity values over the

sample period of 2006 to 2008 to its value over the 2010 to 2012 period. We leave 2009 out to

have a balanced number of years in each sub-period. The results in Panel B of Table 1 confirm

that the increase in the mean and median values for both complexity measures increased in the

post-period and that the difference is statistically significant.

The mean values of our complexity measures by industry are presented in Panel C. The

3
At one extreme, a CEO who receives only salary would have a Herfindahl index of 1 and therefore
TDC1_COMPLEX would be 0. At the other, a CEO who receives compensation from all seven components equally
would have a Herfindahl index of 0.143 and thus TDC1_COMPLEX equal to 0.857.

10

complexity of compensation is highest in the Chemicals and Utilities industries using both our

measures of complexity; the Money industry has the lowest value of COMPLEX_IL and the

second lowest value of TDC1_COMPLEX, suggesting relative low complexity of compensation

in that industry as expected given its business focus.4

3.2.2 Economic determinants of compensation contract complexity

To examine whether the complexity of compensation is economically driven, we include

several proxies for economic determinants.

Size. Larger firms likely are more complex and more difficult to manage, as they have

more resources about which managers must make decisions and a larger scope of operations

(Smith and Watts 1992; Gaver and Gaver 1993; Himmelberg, Hubbard, and Palia 1999) and a

greater tendency to decentralize (Christie, Joye, and Watts 2003). In addition, it may be more

difficult to monitor managers of larger firms (Eaton and Rosen 1983). As a result of such

complexities, we expect that they have more complex compensation packages. We measure size

as the natural log of the market value of equity (lnMV).5

Growth opportunities. Firms with higher growth opportunities have businesses that are

more difficult to manage than firms with lower growth opportunities and are more difficult to

monitor because outside constituents cannot easily observe the firms investment opportunities

(Smith and Watts 1992). In addition, they tend to be more decentralized (Christie, Joye, and

Watts 2003). As a result, they may have more complex compensation packages, as compensation

contracts have performance measures that are tailored to the specific characteristics of each

business (Bushman et al. 1996). Alternatively, these firms may rely heavily on equity to

4
Studies of compensation typically exclude firms in the utilities and financial services industries. In untabulated
robustness tests, our findings are unchanged if we also exclude firms in these industries.
5
In robustness tests we proxy for firm size using the natural log of the book value of assets and find similar results.

11

compensate their CEOs in an attempt to align the incentives of CEOs with those of shareholders

(Smith and Watts 1992; Gaver and Gaver 1993; Himmelberg, Hubbard, and Palia 1999); a heavy

reliance on any one component of pay results in lower values of complexity as we measure it.

We measure growth opportunities as the ratio of the sum of the market value of equity and the

book value of total liabilities to the book value of assets (MTB) and the ratio of R&D expenses

to the book value of assets (R&D).6

Volatility of business operations. More volatile firms place more risk on the CEO. Thus,

it is possible that to compensate those CEOs for assuming that higher risk, more volatile firms

rely more heavily on a fixed salary; again, a heavy reliance on any one component of pay may

result in lower values of complexity as we measure it. On the other hand, it is more difficult to

monitor managers of more volatile firms (Demsetz and Lehn 1985). That higher difficulty in

monitoring may result in more complex compensation packages; alternatively, it may result in

compensation packages based primarily on equity, which may result in lower values of our

measures of complexity. We measure the volatility of business operations as the natural log of

the standard deviation of monthly stock returns (lnSTDRET).

Complexity of business operations. Firms with more complex business operations likely

have more complex compensation packages. We include three measures of business complexity.

For the first two measures, we use segment data obtained from Compustat. We calculate the

natural log of the number of business segments (lnSEGMTS). We also calculate a Herfindahl

Index based on the proportion of revenues accounted for by each segment (see also Jennings, Seo

and Tanlu 2014); we subtract this Index from 1 so that higher values of this variable reflect

greater complexity (SEGMTS_HH). As an example, the conglomerate General Electric has a

6
Missing observations for R&D expenses are assumed to be zero, because firms with negligible R&D expenses are
not required to report them separately.

12

high value of 0.82 for this measure of complexity. In their footnotes, firms must report financial

data on a segment basis, based on the segments used for the internal management of operations.7

Most firms in our sample provide segment data based on business lines. However, some firms

report that they manage internal operations geographically; for these firms, we use the reported

geographic segment information to construct our variables. Segment data is not reported for 210

firms (representing 992 firm-years). For these firms, we assume that they have only one segment.

Finally, we include an indicator variable to capture whether the firm has foreign operations,

based on whether the firm reports pretax income from foreign operations (PIFO). We expect that

firms that have more complex operations (greater lnSEGMTS, SEGMTS_HH, and PIFO = 1)

have more complex compensation plans.

3.2.3 The rent extraction view

To examine whether rent extraction plays a role in the complexity of compensation

packages, we include an estimate of excess pay (EXCESSPAY). We estimate excess pay

following Core, Guay, and Larcker (2008), where excess pay is the residual pay from an

expected CEO compensation model that controls for economic determinants such as CEO tenure,

firm size, book-to-market, stock return, accounting return, whether the firm belongs to the

S&P500, and industry and year controls. If rent extraction is a motive for complex contacts, we

expect a positive relation between excess pay and the complexity of compensation contracts.8

7
SFAS 131, implemented in 1997, requires that firms disclose information about each reportable segment in the
same manner that management views operating segments for internal decision-making purposes. All firm years in
our sample are subject to this standard.
8
For relatively simple compensation contracts, models of expected pay may be more accurate. Thus, our proxy for
excess pay may contain more measurement error for firms with complex compensation packages. For example, it
may be the case that in complex firms, performance measures include non-financial measures or individual
performance objectives that are missing in our model. Further, those measures may be negatively correlated with
observable measures of firm performance (ROA or stock returns), leading the appearance of excess pay. Though we
attempt to address this concern in later tests, our results may be affected by this measurement error.

13

3.2.4 Additional control variables

We include several additional variables to control for other factors that may explain

contract complexity. First, we include three characteristics about the CEO. We control for CEO

tenure as prior studies suggest that longer serving CEOs may be less subject to agency conflicts

due to less uncertainty about ability (Hermalin and Weisbach, 1998; Dikolli, Mayew, and Nanda,

2014). We also control for founder CEO as they are hypothesized to have lower principal-agent

conflicts due to greater equity holdings (Fahlenbrach, 2009). Thus, long-serving and founder

CEOs may not require complex contracts with multiple incentives. These proxies have also been

associated with CEO power, which may imply greater agency conflicts. Because of these

conflicting views, we control for their effects on compensation without directional predictions.

We also control for proximity to retirement, as Gibbons and Murphy (1992) find that older CEOs

that are close to retirement should receive higher incentive pay, which could be more complex,

to substitute for declining incentive alignment due to career concerns. However, Yermack (1995)

and Bryan et al. (2000) find no evidence that CEOs close to retirement hold more stock options.

As a result, we control for these CEO specific characteristics but we offer no predictions

regarding the complexity of their compensation contracts.

Second, we include the number of compensation consultants hired by the firm. While the

use of consultants is generally not thought to be related to rent extraction (Cadman, Carter, and

Hillegeist 2010), firms needing more complex contracts may be more likely to hire consultants.

Or, consultants hired by the firm may recommend more complex contracts to justify their

retention. Although the direction of causality is not clear, we expect that the use of consultants is

related to the complexity of the contract.

14

3.2.5 Multivariate regression

We estimate the following regression of compensation complexity on proxies for the

economic determinants of compensation complexity and our estimate of excess pay.

COMPLEXITYjt = + 1 lnMVjt + 2 MTBjt + 3 R&Djt + 4 lnSTDRETjt + 5 lnSEGMTSjt + 6

SEGMTS_HHjt + 7 PIFOjt + 8 EXCESSPAYjt + 9 lnTENUREj + 10 RETIREj
+ 11 FOUNDERj + 12 NumCONSULTj + j INDj + t YEARt +it (1)

Where:

COMPLEXITYjt is one of the following:

TDC1_COMPLEX = 1 (Herfindahl Index based on proportion of total compensation

accounted for by each compensation component for firm j in year t)
COMPLEX_IL = index of compensation complexity computed from Incentive Lab data

And:

lnMVjt = natural log of the market value of equity of firm j at the beginning of year t

MTBjt = (market value of equity plus book value of total liabilities)/ book value of
assets of firm j at the beginning of year t
R&Djt = research and development expenditures / total assets for firm j at the
beginning of year t
lnSTDRETjt = natural log of the standard deviation of monthly stock returns for firm j for
the two years leading to year t
lnSEGMTSjt = natural log of the number of business segments for firm j in year t
SEGMTS_HHjt = 1 (Herfindahl Index computed based on the proportion of total revenues
for firm j in year t accounted for by each business segment)
PIFOjt = indicator variable equal to 1 if pretax foreign income for firm j in year t is
greater than zero; 0 otherwise
EXCESSPAYjt = the log of excess CEO total compensation based on the methodology in
Core, Guay, and Larcker (2008) for firm j in year t
lnTENUREjt = natural log of the CEO tenure for firm j in year t
RETIREjt = indicator variable equal to 1 if the CEO is older than 65 in year t.
FOUNDERjt = indicator variable equal to 1 if the CEO of firm j in year t is a founder of the
firm.
NumCONSULTjt = number of compensation consultants hired by firm j in year t
INDj = indicator variable for industry, based on Fama-French 12 industry
classification for firm j
YEARt = indicator variable for year t

15

 We winsorize the variables MTB and EXCESSPAY at the 1 and 99 percent level to

mitigate the influence of outliers. We compute standard errors that are cluster-adjusted by firm.

4. Results

4.1 Descriptive statistics for sample firms

We present descriptive statistics in Table 2. Sample firms have a mean (median) market

value of equity of $7,550 million ($1,578 million) and mean (median) market-to-book ratio of

1.809 (1.459). Just over 18% of firms report a loss. The mean (median) standard deviation of

returns is 0.116 (0.100). The mean (median) number of business segments is 2.48 (2.00). At least

25% of firms report only one business segment, and at least 25% of firms report 4 or more

business segments. Over half of the firms (58.3%) report pretax income from foreign operations.

The mean (median) excess pay for sample firms is 0.020 (0.068). CEOs have an average tenure

of 7.8 years. And among our sample CEOs, approximately 15% are founders and only 7% are

close to retirement. Firms hire an average (median) of one consultant. Approximately 13% of the

firm years have no consultant while 13% have more than one (untabulated).

4.2 Pearson correlations

Table 3 presents Pearson correlations among our independent variables. The correlation

between our two measures of compensation complexity is 35% (significant at p < 0.05),

suggesting that the two measures capture similar characteristics, but are not perfectly correlated.

That these measures are not perfectly correlated is not surprising given that TDC1_COMPLEX

is an ex-post measure of realized complexity while COMPLEX_IL is an ex-ante measure of

contract complexity in determining the formula by which realizations will be computed. Size is

positively correlated with complexity of compensation (p < 0.05), but interestingly, the

market-to-book ratio and volatility of returns are negatively correlated with compensation

16

complexity (all significant at p < 0.05). Compensation complexity is positively correlated with

the number of segments (11.9% and 9.9% with TDC1_COMPLEX and COMPLEX_IL,

respectively, both significant at p < 0.05), yet the correlation with operational complexity as

proxied by the business segment concentration (Herfindahl Index) is low and not significant.

Compensation complexity is positively correlated with the presence of foreign operations (8.1%

and 7.0%, respectively, both significant at p < 0.05) and with excess pay (26.2% and 26.5%,

respectively, both significant at p < 0.05).

While several of the correlations among the independent variables are significant, there

are a few notable correlations. Firm size is negatively correlated with volatility in returns

(-49.9%) and positively correlated with the number of consultants (23.7%). CEO tenure is

positively correlated with being both a founder CEO (28.3%) and close to retirement (38.3%).

4.3 Multivariate regression

Table 4 Panels A and B present results from estimating our regression, using

TDC1_COMPLEX and COMPLEX_IL, respectively, as our dependent variable. Column (1) in

each panel presents results from estimating the regression with only our proxies for the economic

determinants of compensation complexity. Those results are relatively consistent across our two

alternative measures of compensation complexity. Specifically, as expected, larger firms, firms

with more complexity in terms of business segments (captured either by the number of segments

or its Herfindahl Index), and firms with foreign operations have more complex compensation

packages. Firms with higher market-to-book ratios and greater volatility in returns have less

complex compensation packages. Further examination of compensation packages in those firms

offers insight into their less complex compensation plans. Specifically, the compensation

packages for CEOs in firms with more growth opportunities tend to include a large equity

17

component, consistent with equity being used to align incentives of CEOs with those of investors

in firms where monitoring of management is more difficult (Smith and Watts 1992; Gaver and

Gaver 1993; Himmelberg, Hubbard, and Palia 1999). The concentration of pay in equity results

in a lower value for TDC1_COMPLEX for these firms. In addition, firms with high growth

opportunities use fewer performance measures and evaluate performance over fewer periods, in

particular in those compensation components that relate to long term pay (long term cash

bonuses, stock, and stock options); such features result in lower COMPLEX_IL scores for these

firms. The compensation packages for CEOs in firms with more volatile operations tend to

include a large salary component, which results in a lower value for TDC1_COMPLEX; the

larger proportion of compensation from salary in these firms may result from using fixed pay to

compensate CEOs for assuming the greater risk associated with those firms.

In Column (2) for each panel, we include CEO characteristics and the number of

consultants. CEOs with longer tenure, closer to retirement, and that are founders have less

complex compensation packages, while CEOs in firms with more consultants have more

complex packages.

Column (3) in each panel presents results from estimating the regression including

EXCESSPAY. Across both measures of compensation complexity, the coefficient on

EXCESSPAY is positive and significant, consistent with firms with excessive executive pay

having more complex compensation contracts, possibly to obfuscate the excessiveness of that

pay. In addition, the results on our proxies for economic determinants of compensation

complexity are mostly consistent with results presented in Column (1). However, while the

presence of foreign operations is positively related to compensation complexity in Column (1),

its relation with complexity is not significant in Column (2) after the inclusion of EXCESSPAY.

18

5. Further examination of the rent extraction view

5.1 Linguistic complexity

We acknowledge that both EXCESSPAY and compensation contract complexity are

measured with error. If the measurement error in these two constructs is correlated, the relation

that we observe between them could be a spurious one. Accordingly, we include another variable

in our analysis to facilitate our examination of the rent extraction view.9 Specifically, we include

an estimate of the complexity of compensation disclosures in the firms annual proxy statement

(the Compensation Discussion and Analysis section). Prior research finds that disclosure quality

is related to firm performance (e.g., Lang and Lundholm 1993). More specific to our study, prior

research suggests a correlation between the complexity of firms disclosures and firm

performance. In particular, Li (2008) finds that firms annual reports are more complicated and

difficult to read for firms that are performing poorly and firms that have lower earnings

persistence when they are profitable. 10 Li interprets this evidence as consistent with an

obfuscation hypothesis managers acting opportunistically when writing their annual reports to

hide negative information from investors. We extend this possibility to our setting. On the one

hand, it is possible that if firms are attempting to obfuscate or camouflage excessive CEO pay to

avoid or minimize the outrage that results from outsiders recognition of rent extraction, they do

so by providing more convoluted compensation disclosures (Bebchuk and Fried, 2003).11 On the

other hand, it is possible that complex contracts are an attempt to design better incentive

alignment, in which case the firm may seek to enhance investors understanding of those

9
In untabulated robustness tests, we control for measures of corporate governance (CEO is chair, percent of board
members appointed by the current CEO, percent of independent directors, and percent of institutional investors).
These variables are generally not significant but, importantly, the coefficient on EXCESSPAY remains positive and
significant.
10
Unlike Li (2008), we find no relation between the complexity of CD&A disclosures and future ROA.
11
Bushee, Gow and Taylor (2014) document that in the setting of conference calls, manager driven complexity is
related to greater information asymmetry, and they interpret this as evidence of attempts at obfuscation. Laksmana et
al. (2012) find a positive association between excess pay and low readability.

19

contracts by providing more readable compensation disclosures. Thus, we use a measure of the

readability of the Compensation Discussion and Analysis section of the annual proxy statement.

Following Li (2008), we use a measure of the complexity of the Compensation

Disclosure and Analysis section of the firms annual proxy statement the Fog Index.12 The

FOG index combines the number of syllables per word and the number of words per sentence

into a measure of readability. FOG rates the readability of text on a grade school level; the higher

the level, the more difficult the text is to read.

To calculate this measure, we use the following process. First, we download the

Compensation Discussion and Analysis section of each sample firms annual proxy statements

(DEF14A) for the years of 2006 to 2012, excluding the required compensation tables, and save

those disclosures as text files. Then, we read those text files through a text analysis program,

which computes, for each disclosure, the Fog Index of readability.13

Column (3) of Table 4 Panels A and B present the results of estimating equation (1)

including the measure of readability of the Compensation Disclosure and Analysis section of the

annual proxy statements. We note three primary observations from these tests. First, the

coefficient on FOG is negative. This is not consistent with the rent extraction hypothesis that

predicts a positive relation with contract complexity and lower readability. Instead, our results

are consistent with firms with more complex compensation packages trying to make disclosures

more readable consistent with efficient contracting. Second, the coefficient on EXCESSPAY

remains positive and significant in all regressions in both panels, offering consistent evidence

that some pay package complexity reflects a need to camouflage excessive pay. Finally, a few of

12
Li (2008) also uses the length of the annual report as an indication of readability. We do not use this measure
because there will be a mechanical relation between the number of words in the CD&A and our measures of
complexity. Firms that use more elements of pay or that have more conditions around performance-based pay will
be required to use more words to describe the pay packages.
13
Our mean (median) FOG scores of 22.52 (22.44) are similar to those of Li (2008) of 19.39 (19.24).

20

the coefficients on variables capturing the economic determinants of compensation complexity

lose some of their significance. In particular, in Panel A, the coefficients on lnMV and R&D are

no longer significant at conventional levels once the disclosure complexity variables are included

in the regression. This may be due to the correlation with the disclosure variable; the correlation

between lnMV and FOG is -0.188. However, the coefficient on SEGMTS_HH becomes

statistically significant, suggesting that complexity as measured by diversity of business

segments is positively associated with contract complexity, once we control for disclosure

complexity. In Panel B, the inclusion of the complexity disclosure variables does not change

significantly the coefficients of the economic determinant variables.

As robustness tests, we replace FOG with two other measures of readability: the Flesch

Reading Ease Index, and the Kincaid Index. The Flesch Reading Ease Index rates the readability

of text on a 100-point scale; the higher the level, the easier the text is to read. To facilitate

comparability, we convert the score such that higher values reflect less readability by subtracting

the actual Flesch score from the sample maximum. Like FOG, the Kincaid Index rate the

readability of text on a grade school level with the higher the level indicating greater difficulty.

Our conclusions are unchanged using these alternative measures. More complex contracts are

associated with more readable disclosures and EXCESSPAY remains positive and significant.

5.2 Association with future performance

Excess pay that is associated with contract complexity may reflect payment for

compensation risk (Black et al. 2014). The increased complexity in compensation packages,

particularly in the growth in the number of incentive plans and performance metrics, brings

greater uncertainty about expected pay, for which companies must pay a premium for executives

to accept (Gosling, 2014). In order to disentangle whether the component of contract complexity

21

that is associated with excess pay captures rent extraction instead of additional pay for added

risk, we follow Core et al. (1999) and relate its association to future performance. If the excess

pay component of contract complexity captures attempts to obfuscate excess pay (the rent

extraction view), then it should be negatively associated with future performance. If, instead, the

excess pay component of contract complexity captures additional pay to the CEO to compensate

him for the added compensation risk resulting from a complex contract, then it should not be

negatively associated with future performance. We measure future performance by future

accounting performance because, as mentioned in Core et al. (1999), this prediction is stronger

for accounting than for stock return performance since, under efficient markets, stock returns

would fully price the future cash flow implications of excess pay. To test our assertions, we

estimate the impact of the components of contract complexity on future performance using the

following model:

Future ROAit = 0 + 1 ROA+3 Complex-Payit + 4 Complex-Econit +5 Complex-Errit +

6 LogSalesit + 7 Log STD ROAit + 8 INDj + 9 YEARt +it . (2)

We estimate this regression using two separate measures of Future ROA: (1) the one-year-ahead

annual return on assets, and (2) the average of the one-year and two-year-ahead annual return on

assets. We decompose contract complexity into several components: contract complexity that is

associated with excessive pay (COMPLEX_Pay), contract complexity that is explained by

economic determinants in Eq. (1) (COMPLEX_Econ), and contract complexity that is not

explained by the above variables (COMPLEX_ERR). The set of control variables follows Core et

al. (1999). STD ROA is the standard deviation of return on assets (stock returns) over the prior

four years. LogSales is the log of total sales revenues. We also include industry and year

fixed-effects to account for any change in future ROA that is industry or year specific.

22

 Table 5, Panel A presents the result of estimation Eq. (2), using TDC1_COMPLEX as the

complexity measure. Columns (1) through (3) present the results for the one-year-ahead ROA

and Columns (4) through (6) present the results for the average two-year-ahead ROA. We first

estimate the impact of total contract complexity on future ROA as a benchmark model. Columns

(1) and (4) show that total contract complexity is not (statistically) negatively associated with

future performance (coefficient on COMPLEX is not significant), casting doubt that overall

contract complexity is related to a desire to obfuscate excessive pay packages. Next, we

decompose contract complexity into several components and test their associations with future

performance. Columns (2) and (5) show that the coefficient on (COMPLEX_Pay) is negative

(0.14, t-stat = 4.37 in Column (2)), consistent with contract complexity due to excessive pay

capturing rent extraction and compensation packages that are not optimal to the firm. The

coefficient on the economic determinants (COMPLEX_Econ) is negative ( 0.21, t-stat = 2.63

in Column (2)), suggesting that contract complexity that is due to the complexity of the business

model is also associated with poorer future performance, which is unexpected. One possible

explanation for this relation is that with overly complex pay packages incorporating multiple

measures over multiple periods, CEOs lose sight of the underlying business. We then investigate

whether this lower future performance is due to the characteristics of firms with more complex

business models (and thus contract complexity). To do so, we replace the variable

COMPLEX_Econ by its economic determinants in Columns (3) and (6). As expected, we find

that characteristics associated with more complex firms larger (lnMV), invest heavily in R&D

(R&D), and have higher volatility (lnSTDRET) and higher dispersion of business segments

(SEGMTS_HH) are negatively associated with future performance. In addition, using this

expanded model in Columns (3) and (6), we continue to find evidence that contract complexity

23

due to excess pay is negatively associated with future performance ( 0.07, t-stat = 2.06 in

Column (3)).

Panel B presents results using COMPLEX_IL as the complexity measure. We find a

negative overall relation between contract complexity and future performance. In decomposing

complexity, we continue to find that the portion due to excess pay (COMPLEX_Pay) remains

negatively related to future performance. Together, the results in Panels A and B reinforce the

rent extraction explanation for contract complexity: contract complexity explained by excess pay

is related to poor future performance.14

6. Analyses through time

The results above show that contract complexity has increased during the sample period.

To test whether the determinants of compensation contract complexity have also changed

through this period, we replicate the results in Table 4 for two subsamples, the pre-period of

2006 to 2008 and the post-period of 2010 to 2012, dropping 2009 to create balanced panels.

These time periods also coincide with the pre- and post- financial crisis period, that latter of

which included heighted interest by regulators in executive compensation, particularly in

financial services firms. Among the points raised were concerns about tying compensation to

appropriate performance hurdles and over appropriate time horizons, and increased transparency

of the pay setting process.15 To the extent that these concerns permeated across the economy, we

14
An alternative explanation is that our measure of expected pay does not consider non-financial performance
measures and the use of those measures is more common when financial measures of performance are noisy. The
resulting measurement error in excess pay could therefore be correlated with poor accounting performance. While
the observed high pay that is negatively related to financial performance appears to be rent extraction (and widely
criticized in the popular press), in our setting this high pay could be justified by these non-financial performance
measures that are negatively correlated with future ROA. We are currently exploring ways to address this concern.
15
See Statement by Treasury Secretary Timothy Geitner on Compensation, June 2009.

24

might expect greater complexity of compensation arrangements along with greater clarity of

CD&A disclosures.16

Table 6, Panel A (B) presents the results using TDC1_COMPLEX (COMPLEX_IL) as

the dependent variable. Columns (1) and (3) in each Panel presents results from estimating the

simple regression model with only the proxies for the economic determinants of compensation

complexity. The results show that the ability of our economic determinants to explain contract

complexity has decreased from the pre- to the post-period for both complexity measures (e.g.,

adjusted R2 decreased from 0.092 to 0.069 (0.100 to 0.083) in Column 1 of Panel A (B)). This is

reinforced by the drop in significance of our proxies for complexity; in Panel A, ln_SEGMTS is

not significant in the later period and in Panel B, both SEGMNTS_HH and PIFO lose

significance. Interestingly, the results in Columns (2) and (4) of both panels show that the

coefficient on the linguistic complexity measure is only statistical significant in the post-period.

We interpret this as evidence that the language used becomes simpler in the post-period possibly

reflecting an effort by the SEC through the Comment letters intervention that disclosures should

be easy for investors to follow.

In sum, it seems that in the pre-period, the economic determinants matter most and the

disclosure language is unrelated to the contract complexity. In the post-period, contracts have

become more complex, yet the complexity of the language in compensation disclosures has

decreased. One potential explanation is that institutional investors and shareholder activism are

calling for more pay-for-performance and, as a result, contracts have become more complex and

not necessarily related to firm characteristics. At the same time, there are increasing calls for

greater clarity in compensation disclosure. In fact, the 2014 SEC Comments and Trends report

16
Compensation consultants provide an indication that these concerns were likely to permeate. See for example, a
June 12, 2009 alert letter sent to compensation professionals by Fred W. Cook & Co., Inc.
http://www.fwcook.com/alert_letters/06-12-09_Obama_Administration_Releases_Executive_Compensation.PDF

25

by E&Y reports that executive compensation disclosures account for 14% of the total number of

registrants that received comment letters from the SEC, and that the SEC staff focuses its

reviews on registrants Compensation Discussion and Analysis in an effort to promote more

direct, specific and clear executive compensation disclosure.

7. Conclusion

Kole (1997) suggests that compensation contract complexity can result from providing

executives with incentives that match the complexity of the business operations they manage.

Alternatively, managers can enter into more complex contracts to obfuscate their rent extraction

(Bebchuk and Fried 2003). By designing more complex contracts, managers can make it more

difficult for outsiders to detect the extent to which compensation is not tied to the managers

own performance. The desire to camouflage might lead to the adoption of inefficient

compensation structures that can result in poor managerial incentives and firm performance.

Using samples of firms from ExecuComp and Incentive Lab from 2006 2012, we

examine the extent to which executive compensation contract complexity is consistent with

either the efficient contracting view or the rent extraction view. We construct two measures of

the complexity of CEO compensation: (1) an ex-post measure the captures the relative

importance of different forms of pay, constructed from ExecuComp, and (2) an ex-ante index of

complexity that increases with the number of factors determining performance-based pay,

constructed from Incentive Lab data. Then, we examine the relation between these measures of

complexity and both economic determinants of that complexity (the efficient contracting view)

and excess CEO pay (the rent extraction view).

We find that our measures of the complexity of CEO compensation packages are related

to several economic characteristics that reflect the scope and complexity of the organizations

26

CEOs are charged with managing. In particular, larger firms, firms with more business segments,

and firms with foreign operations have more complex CEO compensation packages. Firms with

more growth opportunities and more volatile operations have less complex plans, on the other

hand, reflecting compensation packages based heavily on equity as an incentive alignment

mechanism for firms with high growth opportunities and on salary for firms with more volatile

operations as a way to compensate CEOs for assuming higher risk. These results lend support to

the efficient contracting view of compensation contract design. However, we also find that

excess CEO pay is positively related to the complexity of CEO compensation packages,

consistent with the rent extraction view. In addition, we find that the component of contract

complexity that is related to excess pay is associated with lower future accounting performance,

further supporting our interpretation that complexity of contract is related to the desire to

camouflage excess pay. Overall, our results provide support for both the efficient contracting

view and the rent extraction view in the design of CEO compensation packages. Finally, we

document that while contract complexity has significantly increased overtime, those contracts are

more clearly explained from a linguistic perspective but less associated with firms

characteristics.

27

 Appendix A

Component of pay Characteristics Complexity Score

Short Term Cash Bonus Has time conditions 1
Has absolute performance conditions 1
Number of performance measures > 1 1
Number time periods > 1 1
Has relative performance conditions 1
Number of performance measures > 1 1
Number time periods > 1 1
Long Term Cash Bonus Has time conditions 1
Has absolute performance conditions 1
Number of performance measures > 1 1
Number time periods > 1 1
Has relative performance conditions 1
Number of performance measures > 1 1
Number time periods > 1 1
Restricted Stock Has time conditions 1
Has absolute performance conditions 1
Number of performance measures > 1 1
Number time periods > 1 1
Has relative performance conditions 1
Number of performance measures > 1 1
Number time periods > 1 1
Stock Options Stock Has time conditions 1
Has absolute performance conditions 1
Number of performance measures > 1 1
Number time periods > 1 1
Has relative performance conditions 1
Number of performance measures > 1 1
Number time periods > 1 1

Potential Total Score 28

28

 Appendix B

ConocoPhilips has a score of 9 for fiscal year of 2010 using the COMPLEX_IL measure. Below, we
provide details regarding how the information from the proxy statement (DEF 14A) about incentive
contracts is used to calculate the COMPLEX_IL measure score.

Summary:

Component Characteristics Score

ST Cash Bonus Has absolute perf conds 1
No of perf measures > 1 1
Has relative perf conds 1
No of perf measures > 1 1
LT Cash Bonus 0
Restr Stock Has absolute perf conds 1
No of perf measures > 1 1
Has relative perf conds 1
No of perf measures > 1 1
Stock Options Has time conditions 1

Total Score 9

Information from Proxy Statement regarding short-term cash bonus is as follows:

In 2010, our Variable Cash Incentive Program (VCIP) program used both quantitative and qualitative
performance measures relating to the Company as a whole, including:
Ranking 1st in relative annual total stockholder return compared with our
performance-measurement peer group (ExxonMobil, Royal Dutch Shell, BP, Total, and
Chevron);
Ranking 2nd in percentage change and 3rd in absolute change in improvement in relative
annual adjusted return on capital employed compared with the same peer group noted
above;
Ranking 3rd in percentage and absolute change in relative annual adjusted cash return on
capital employed compared with the same peer group noted above;
Ranking 2nd in relative adjusted cash contribution BOE compared with the same peer
group noted above;
Our health, safety and environmental performance; and
Advancement and support of our key strategic initiatives and plans.

Note that BOE stands for barrel of oil extracted.

ConocoPhilips uses both relative and absolute performance measures in their short-term cash program,
which gives them 2 points. In addition, they have four relative performance measures and two absolute
performance measures. Hence, we further assign 1 point if each absolute (relative) performance measure
has more than one metric.
Information from Proxy Statement regarding restricted stock and stock options plans is as follows:

29

Our program targets generally provide approximately 50 percent of the long-term incentive award in
the form of stock options and 50 percent in the form of restricted stock units awarded under the PSP.

Stock Option ProgramThe Stock Option Program is designed to maximize medium- and
long-term stockholder value. . Our stock options have three-year vesting provisions
and ten-year terms in order to incentivize our executives to increase the Companys share
price over the long term.

Performance Share ProgramThe PSP rewards executives based on their individual

performances and the performance of the Company over a three-year period. Each year the
Committee establishes a three-year performance period over which it compares the
performance of the Company with that of its performance-measurement peer group using
pre-established criteria.

In Dec 2007, the Committee established the sixth performance period under the PSP, for the three-year
period beginning Jan 1, 2008, and ending Dec 31, 2010 . In determining awards under the PSP for this
period, the Committee considered quantitative and qualitative performance measures relating to the
Company as a whole, including:
Ranking 3rd in relative total stockholder return compared with our perfor-measurement peer group
(ExxonMobil, Chevron, Royal Dutch Shell, BP, and Total);
Ranking 6th in percentage change and 3rd in absolute change in relative improvement in adjusted
return on capital employed compared with the same peer group noted above;
Ranking 2nd in relative adjusted cash contribution per BOE compared with the same peer group noted
above;
Ranking 6th in relative adjusted income per BOE compared with the same peer group noted above;
Our health, safety and environmental performance;
Advancement and implementation of the Companys strategic plans;
Leadership development and succession planning.

Stock options granted are just time vested and thus receive a score of one. The restricted stock units are
granted based on both relative and absolute performance metrics, which again gives them 2 points. In
addition, they have four relative performance measures and three absolute performance measures. Hence,
we further assign 1 point if each absolute (relative) performance measure has more than one metric.

30

 References

Bebchuk. L., Fried, J. 2003. Executive Compensation as an Agency Problem. Journal of Economic
Perspectives 17: 71 92.

Bebchuk. L., Fried, J., and Walker, D. 2002. Managerial power and rent extraction in the design of
executive compensation. The University of Chicago Law Review 69: 751 846.

Black, D., Dikoli, S., and Dyreng, S. 2014. CEO pay-for-complexity and the risk of managerial diversion
from multinational diversification. Contemporary Accounting Research 31: 103 135.

Bryan S, Hwang L, Lilien S., 2000. CEO stock-based compensation: an empirical analysis of
incentive-intensity, relative mix, and economic determinants. Journal of Business 73: 661 93.

Bushee, B., Gow, I., and Taylor, D. 2014. Linguistic complexity in firm disclosures: Obfuscation or
information? Working paper. University of Pennsylvania.

Bushman, R. Chen, Q, Engel, E. and Smith, A. 2004. Financial accounting information, organizational
complexity and corporate governance systems. Journal of Accounting and Economics 37: 167 201.

Bushman, R., Indjejikian, R., and Smith, A. 1996. CEO compensation: The role of individual
performance evaluation. Journal of Accounting and Economics 21: 161 193.

Cadman, B., Carter, M.E., and Hillegeist, S. 2010. The incentive of compensation consultants and CEO
pay. Journal of Accounting and Economics 49: 263 280.

Christie, A., Joye, M. and Watts, R. 2003. Decentralization of the Firm: Theory and Evidence. Journal of
Corporate Finance 9: 3 36.

Conference Board. 2013. Supplemental Pay Disclosure: Overview of Issues, Proposed Definitions, and a
Conceptual Framework by The Conference Board Working Group on Supplemental Pay Disclosure.
White Paper, The Conference Board.

Core, J., Guay, W., and Larcker, D. 2003. Executive equity compensation and incentives: A survey.
Economic Policy Review 9: 27 50.

Core, J., Guay, W., and Larcker, D. 2008. The power of the pen and executive compensation. Journal of
Financial Economics 88: 1 25.

Demsetz, H., Lehn, K. 1985. The structure of corporate ownership: causes and consequences. Journal of
Political Economy 93: 1155 1177.

Dikolli, S. S., Mayew, W. J., Nanda, D. 2014. CEO tenure and the performance-turnover relation. Review
of Accounting Studies 19: 281327.

Duru, A. and Reeb, D. 2002. Geographic and industrial corporate diversification: The level and structure
of executive compensation. Journal of Accounting, Auditing, and Finance 17: 1 24.

31

Eaton. J. and Rosen, H. 1983. Agency, delayed compensation and the structure of executive
remuneration. Journal of Finance 38: l189 1505.

Fahlenbrach, R. 2009. Founder-CEOs, investment decisions, and stock market performance. Journal of
Financial and Quantitative Analysis 44: 439 466.

Gaver, J., and Gaver, K. 1993. Additional evidence on the association between the investment opportunity
set and corporate financing, dividend, and compensation policies. Journal of Accounting and
Economics 16: 125 160.

Gibbons R, Murphy KJ. 1992. Optimal incentive contracts in the presence of career concerns: theory and
evidence. Journal of Political Economy 100: 468 505.

Gillenkirch, R., Hendriks, A. and Welker, S. 2013. Effects of executive compensation complexity on
investor behavior in an experimental stock market. European Accounting Review forthcoming.

Gosling, T. 2014. Why pay for corporate high flyers is coming back to earth. Financial Times. February
3, 2014.

Hermalin, B. and Weisbach, M. 1998. Endogenously chosen boards of directors and their monitoring of
the CEO. American Economic Review 88: 96 118.

Himmelberg, C., Hubbard, R. and Palia, D. 1999. Understanding the determinants of managerial
ownership and the link between ownership and performances. Journal of Financial Economics 53:
353 384.

Hoffman, L. 2015. Activists new target: Executive pay. Wall Street Journal. June 12, p. C1.

Holmstrom, B., Milgrom, P. 1987. Aggregation and linearity in the provision of intertemporal incentives.
Econometrica 55, 303328.

Ittner, C., Larcker, D. and Rajan, M. 1997. The choice of performance measures in annual bonus
contracts. The Accounting Review 77 (2): 231-255.

Jennings, J., Seo, H. and Tanlu, L. 2014. The Effect of Organizational Complexity on Earnings
Forecasting Behavior. Working Paper.

Johnson, M. 2011. Investors concerned over pay complexity. Financial Times. February 13, 2011.

KPMG and Financial Executives Research Foundation. 2011. Disclosure overload and complexity:
Hidden in plain sight.

Kole, S. 1997. The complexity of compensation contracts. Journal of Financial Economics 43: 79 104.

Laksmana, I., Tietz, W. and Yang, Y. 2012. Compensation disclosure and analysis: Readability and
management obfuscation. Journal of Accounting and Public Policy 31: 185 203.

32

Lambert, R. and Larcker, D. 1987. An analysis of the use of accounting and market measures of
performance in executive compensation contracts. Journal of Accounting Research, 25 (Supplement):
85-125.

Li, F. 2008. Annual report readability, current earnings, and earnings persistence. Journal of Accounting
and Economics 45: 221 247.

Murphy, K. 2000. Performance standards in incentive contracts. Journal of Accounting and Economics,
30(3): 245-278.

Rose, N. and Shepard, A. 1997. Firm diversification and CEO compensation: Managerial ability or
executive entrenchment? Rand Journal of Economics 28: 489 514.

Sloan, R. 1993. Accounting earnings and top executive compensation. Journal of Accounting and
Economics, 16: 55-100.

Smith, C. and Watts, R. 1992. The investment opportunity set and corporate financing, dividend, and
compensation policies. Journal of Financial Economics 32: 263 292.

Yermack D. 1995. Do corporations award CEO stock options effectively? Journal of Financial
Economics 39: 237 69.

33

 Figure 1
Measures of compensation complexity by year over the sample period 2006-2012

Measures of compensa3on complexity over sample

period
5.0 0.60

Herndahl Index from compensa3on

Complexity score copmuted from Incen3ve

0.58
4.5
0.56

components
4.0
Lab data

0.54
3.5
0.52

3.0 0.50
2006 2007 2008 2009 2010 2011 2012

COMPLEX_IL TDC1_COMPLEX

34

 Table 1
Descriptive statistics for measures of contract complexity

This table provides descriptive statistics for our two measures of contract complexity. These measures are defined in
Section 3.2.1. Panel A reports overall statistics. Panel B reports statistics over two time periods in our sample. Panel
C reports the mean values by industry where industry is defined as Fama French 12 industry groups.

Panel A: Statistics for contract complexity measures

Std.
Variable N Mean Deviation 1% 50% 99%
TDC1_COMPLEX 10,780 0.569 0.167 0.134 0.616 0.773
COMPLEX_IL 5,711 4.106 1.998 1 4 9

Panel B: Contract complexity measures over time

Pre-Period (2006-2008) Post-Period (2010-2012) Difference
Variable N Mean Median N Mean Median Mean Median
t-value p-value
TDC1_COMPLEX 4,651 0.560 0.606 4,512 0.580 0.625 5.829 0.000
COMPLEX_IL 2,390 3.88 4.00 2,354 4.445 5.00 9.879 0.000

Panel C: Mean values of complexity variables by Fama French industry code

Using TDC1_COMPLEX Using COMPLEX_IL

IndFF12 N % Mean N % Mean
1 Consum NoDur 658 6.10 0.580 386 6.76 4.054
2 Consum Durables 309 2.87 0.587 120 2.10 4.192
3 Manuf 1,304 12.10 0.606 706 12.36 4.681
4 Enrgy 486 4.51 0.587 328 5.74 4.095
5 Chems 321 2.98 0.655 207 3.62 4.763
6 BusEq 2,108 19.55 0.530 1,098 19.23 3.935
7 Telcm 243 2.25 0.573 179 3.13 3.838
8 Utils 510 4.73 0.615 368 6.44 4.970
9 Shops 1,272 11.80 0.556 602 10.54 3.859
10 Hlth 894 8.29 0.584 416 7.28 4.204
11 Money 1,360 12.62 0.548 628 11.00 3.678
12 Other 1,315 12.20 0.561 673 11.78 3.756
Total 10,780 100% 0.569 5,711 100% 4.106

35

 Table 2
Descriptive statistics

This table reports descriptive statistics for our explanatory and control variables. MV (lnMV) = market value (natural log of) of equity, MTB = market value of
equity plus book value of total liabilities divided by the book value of assets, R&D = research and development expenditures / total assets, STDRET
(lnSTDRET) = the standard deviation (natural log of) of monthly stock returns for the prior two years, SEGMTS (lnSEGMTS) = the number of business
segments (natural log of), SEGMTS_HH = 1 (Herfindahl Index computed based on the proportion of total revenues accounted for by each business segment).
PIFO = indicator variable equal to 1 if pretax foreign income for firm j in year t is greater than zero and 0 otherwise, EXCESSPAY = the log of excess CEO total
compensation based on the methodology in Core, Guay, and Larcker (2008), lnTENURE = natural log of the CEO tenure, RETIRE = indicator variable equal to 1
if the CEOs age is greater than 65, FOUNDER = indicator variable equal to 1 if the CEO is a founder of the firm, NumCONSULT =number of compensation
consultants hired by the firm.

Std.
variable N Mean Dev. 25% 50% 75%
MV 10,780 7,550 23,726 600 1,578 4,733
lnMV 10,780 7.460 1.666 6.398 7.364 8.462
MTB 10,780 1.809 1.064 1.118 1.459 2.083
R&D 10,780 0.031 0.159 0 0 0.316
STDRET 10,780 0.116 0.080 0.072 0.100 0.138
lnSTDRET 10,780 0.107 0.061 0.070 0.095 0.129
SEGMTS 10,780 2.479 1.998 1 2 4
lnSEGMTS 10,780 1.094 0.563 0.693 1.099 1.609
SEGMTS_HH 10,780 0.384 2.470 0.00 0.407 0.713
PIFO 10,780 0.583 0.493 0 1 1
CEOTenure 10392 7.816 7.170 2.833 5.667 10.333
LnTenure 10392 1.891 0.764 1.344 1.897 2.428
RETIRE 10682 0.075 0.264 0 0 0
FOUNDER 8692 0.149 0.356 0 0 0
NumCONSULT 10780 1.017 0.558 1 1 1
EXCESSPAY 10,204 0.020 0.657 -0.338 0.068 0.420
 Table 3
Pearson correlations

This table reports the Pearson correlations of our measures of compensation complexity, firm complexity, excess pay and other variables. Compensation
complexity variables are defined in Section 3.2.1. FOG is defined in Section 5.1. All other variables are defined in Table 2. * indicates significance at p < 0.05.

1 2 3 4 5 6 7 8 9 10 11 12 13 14
1. TDC1_COMPLEX 1
2. COMPLEX_IL 0.350* 1
3. lnMV 0.156* 0.169* 1
4. MTB -0.095* -0.119* 0.179* 1
5. R&D -0.059* -0.014 -0.107* 0.160* 1
6. lnSTDRET -0.144* -0.114* -0.499* -0.081* 0.132* 1
7. lnSEGMTS 0.119* 0.099* 0.146* -0.110* -0.039* -0.092* 1
8. SEGMTS_HH 0.013 0.020 0.023* -0.029* -0.012 -0.016 0.018* 1
9. PIFO 0.081* 0.070* 0.141* 0.095* 0.092* 0.015 0.125* -0.005 1
10. EXCESSPAY 0.262* 0.265* 0.047* -0.129* 0.018 0.054* 0.065* -0.011 0.096* 1
11. LnTenure -0.051* -0.110* -0.013 0.035* 0.001 -0.046* -0.003 0.000 -0.033* 0.000 1
12. FOUN DER -0.079* -0.127* -0.067* -0.023* -0.012 0.025* -0.005 0.010 -0.028* -0.023* 0.283* 1
13. RETIRE -0.117* -0.191* -0.059* 0.051* 0.013 0.026* -0.015 -0.022* -0.045* -0.049* 0.383* 0.166* 1
14. NumCONSULT 0.146* 0.148* 0.237* -0.057* -0.042* -0.085* 0.059* 0.006 0.044* 0.157* -0.089* -0.050* -0.099* 1
15. FOG -0.072* -0.108* -0.188* 0.010 0.090* 0.142* -0.071* -0.014 -0.028* 0.021* 0.013 0.008 0.050* -0.041*

37

 Table 4

Regression of compensation complexity on proxies for economic determinants and rent extracti

This table provides results of OLS regressions of compensation complexity on proxies for firm complexity, ex
pay and other variables. All regressions include industry and year fixed effects. Panel A reports the results using
ExecuComp measure of complexity while Panel B reports results using the Incentive Lab measure. In each pa
Column (1) includes only firm characteristics, Column (2) adds CEO characteristics, Column (3) adds excess
and Columns (4) (6) include measures of linguistic complexity. FOG is described in Section 5.1. Compensa
complexity is defined in Section 3.2.1. All other variables are defined in Table 2. Standard errors are clustered
firm. Robust t-statistics are reported in parentheses. ***, **, and * indicate significance at p < 0.01, p < 0.05 and
0.10, respectively.

Panel A: Using TDC1_COMPLEX as measure of complexity

(1) (2) (3) (4)

VARIABLES
lnMV 0.01*** 0.00 -0.00 -0.00
(4.05) (0.75) (-0.19) (-1.51)
MTB -0.02*** -0.01*** -0.01*** -0.01***
(-5.57) (-4.08) (-2.80) (-2.82)
R&D -0.02** -0.07 -0.08 -0.02
(-2.34) (-1.23) (-1.48) (-0.40)
lnSTDRET -0.29*** -0.30*** -0.38*** -0.36***
(-5.11) (-4.85) (-6.48) (-6.03)
lnSEGMTS 0.01*** 0.01*** 0.01** 0.01**
(2.79) (2.72) (2.30) (2.00)
SEGMTS_HH 0.00 -0.00 -0.00 0.01
(0.62) (-0.34) (-0.36) (1.48)
PIFO 0.02*** 0.02*** 0.01* 0.01*
(3.31) (2.72) (1.72) (1.71)
lnTenure -0.01*** -0.02*** -0.02***
(-3.70) (-4.71) (-4.90)
RETIRE -0.03** -0.03** -0.03**
(-2.34) (-2.43) (-2.33)
FOUNDER -0.03*** -0.02** -0.02**
(-2.73) (-2.40) (-2.52)
NumCONSULT 0.02*** 0.01*** 0.01***
(4.67) (2.99) (2.74)
EXCESSPAY 0.06*** 0.05***
(9.86) (9.49)
FOG -0.00***
(-2.97)

Observations 10,780 8,564 8,513 7,943

Adjusted R2 0.0800 0.0888 0.1360 0.1369

38

 Table 4 (continued)

Panel B: Using COMPLEX_IL as measure of complexity

(1) (2) (3) (4)

VARIABLES
lnMV 0.25*** 0.21*** 0.23*** 0.21***
(5.08) (4.05) (4.68) (4.12)
MTB -0.30*** -0.25*** -0.19*** -0.20***
(-6.34) (-5.05) (-4.07) (-4.17)
R&D 0.77 0.81 0.96 0.77
(0.78) (0.72) (0.85) (0.68)
lnSTDRET -2.22** -1.89* -2.98*** -2.44**
(-2.20) (-1.65) (-2.82) (-2.25)
lnSEGMTS 0.05 0.03 0.00 0.01
(0.73) (0.40) (0.03) (0.09)
SEGMTS_HH 0.01*** 0.09 0.02 0.02
(3.50) (0.69) (0.15) (0.18)
PIFO 0.22* 0.08 -0.01 -0.02
(1.86) (0.64) (-0.12) (-0.20)
lnTenure -0.23*** -0.30*** -0.31***
(-3.42) (-4.75) (-4.68)
RETIRE -0.79*** -0.90*** -0.88***
(-3.84) (-4.61) (-4.45)
FOUNDER -0.65*** -0.53*** -0.53***
(-4.13) (-3.38) (-3.36)
NumCONSULT 0.33*** 0.21*** 0.19**
(4.11) (2.61) (2.45)
EXCESSPAY 0.89*** 0.88***
(12.34) (12.00)
FOG -0.05**
(-2.44)

Observations 5,711 4,582 4,569 4,365

Adjusted R2 0.1184 0.1528 0.2254 0.2251

39

 Table 5

The impact of compensation complexity on future performance

This table reports results of OLS regressions of future return on assets (Columns 1-3) and average ROA (Columns 4-6) on
compensation complexity (Columns 1 and 4), and on compensation complexity partitioned into the component predicted by
excess pay (COMPLEX_Pay), economic determinants (COMPLEX_Econ) and the unexplained component
(COMPLEX_Econ) (Columns 2 and 5). All regressions include industry and year fixed effects. In Columns 3 and 6, we
replace COMPLEX_Econ with the economic determinants of complexity. Panel A reports the results using the ExecuComp
measure of complexity while Panel B reports results using the Incentive Lab measure. Compensation complexity is defined
in Section 3.2.1. All other variables are defined in Table 2. Standard errors are clustered by firm. Robust t-statistics are
reported in parentheses. ***, **, and * indicate significance at p < 0.01, p < 0.05 and p < 0.10, respectively.

Panel A: Using TDC1_COMPLEX as measure of complexity (COMPLEX)

(1) (2) (3) (4) (5) (6)
VARIABLES ROAt+1 ROAt+1 ROAt+1 AvgROAt+1,t+2 AvgROAt+1,t+2 AvgROAt+1,t+2
ROA -0.42 0.49*** 0.38*** -0.41 0.51*** 0.37***
(-0.65) (12.29) (8.69) (-0.75) (5.72) (4.79)
COMPLEX 0.05 0.02
(1.46) (1.14)
COMPLEX_Pay -0.16*** -0.10*** -0.18*** -0.13***
(-4.72) (-2.91) (-4.53) (-2.68)
COMPLEX_Err 0.01 0.02** -0.00 0.01
(1.38) (2.33) (-0.15) (0.94)
COMPLEX_Econ -0.09** -0.04
(-2.29) (-0.63)
Fog 0.00 -0.00
(0.04) (-0.26)
lnMV -0.00 -0.02
(-1.53) (-1.32)
MTB 0.03*** 0.03***
(8.68) (7.13)
R&D -0.24*** -0.35**
(-3.73) (-2.44)
lnSTDRET -0.15*** -0.28*
(-3.07) (-1.72)
lnSEGMTS 0.00 0.00
(1.01) (0.19)
SEGMTS_HH -0.01** -0.01**
(-2.32) (-2.32)
PIFO 0.00 0.00
(0.94) (0.47)
lnTenure 0.00 0.00
(1.03) (1.29)
RETIRE -0.00 0.00
(-0.13) (0.20)
FOUNDER -0.01* -0.01
(-1.65) (-1.51)
NumCONSULT 0.00 0.00
(0.26) (0.75)
LogSales 0.00 0.00*** 0.01*** 0.01 0.01** 0.02
(0.34) (4.70) (2.60) (1.16) (2.44) (1.59)
LagSTDROA -0.01* -0.00*** -0.00** -0.01 -0.00** -0.00
(-1.68) (-2.67) (-2.04) (-1.59) (-2.32) (-0.95)

Observations 10,119 7,631 7,631 8,348 6,226 6,226

Adjusted R-squared 0.1069 0.2713 0.3094 0.1411 0.1764 0.2082

40

 Table 5 (Continued)

Panel B: Using COMPLEX_IL as measure of complexity (COMPLEX)

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

VARIABLES ROAt+1 ROAt+1 ROAt+1 AvgROAt+1,t+2 AvgROAt+1,t+2 AvgROAt+1,t+2

ROA 0.43*** 0.50*** 0.37*** 0.38*** 0.46*** 0.30***

(9.23) (8.10) (5.23) (10.10) (12.29) (6.62)
COMPLEX -0.00* -0.00*
(-1.81) (-1.68)
COMPLEX_Pay -0.01*** -0.01*** -0.01*** -0.01**
(-4.25) (-2.72) (-3.52) (-2.05)
COMPLEX_Err -0.00 -0.00 -0.00 -0.00
(-0.20) (-0.22) (-0.89) (-0.99)
COMPLEX_Econ -0.01** -0.01*
(-2.45) (-1.84)
Fog -0.00 -0.00
(-0.48) (-0.88)
lnMV -0.00 -0.01**
(-1.35) (-2.18)
MTB 0.02*** 0.03***
(5.67) (6.89)
R&D -0.23*** -0.16***
(-3.61) (-2.99)
lnSTDRET -0.11** -0.12***
(-2.51) (-2.78)
lnSEGMTS 0.00 0.00
(0.96) (0.05)
SEGMTS_HH -0.01** -0.01**
(-2.56) (-2.56)
PIFO 0.00 -0.00
(0.23) (-0.39)
lnTenure 0.00 0.00
(1.15) (0.61)
RETIRE 0.00 0.01*
(0.73) (1.66)
FOUNDER -0.01 -0.01**
(-1.62) (-2.18)
NumCONSULT 0.00 -0.00
(0.29) (-0.12)
LogSales 0.00 0.00 0.00* 0.00 0.00 0.01**
(0.77) (0.99) (1.72) (0.12) (0.30) (2.22)
LagSTDROA -0.00 -0.00 -0.00 -0.00* -0.00 -0.00
(-1.56) (-0.80) (-0.20) (-1.68) (-0.89) (-0.85)

Observations 5,353 4,261 4,261 4,468 3,536 3,536

Adjusted R-squared 0.2190 0.2968 0.3378 0.2423 0.3020 0.3600

41

 Table 6

Regression of compensation complexity on proxies for economic determinants and rent extraction
for two sample periods of 2006-2008 and 2010-2012.

This table provides results of OLS regressions of compensation complexity on proxies for firm complexity, excess
pay and other variables. All regressions include industry and year fixed effects. Panel A reports the results using the
ExecuComp measure of complexity while Panel B reports results using the Incentive Lab measure. In each panel,
Columns (1)-(2) are for the early sample period 2006-2008 while Columns (3) (4) are for the later period
2010-2012. Column (1) and Column (3) includes only firms characteristics, Column (2) and Column (4) add CEO
characteristics, excess pay, and a measure of linguistic complexity. FOG is described in Section 5.1. Compensation
complexity is defined in Section 3.2.1. All other variables are defined in Table 2. Standard errors are clustered by
firm. Robust t-statistics are reported in parentheses. ***, **, and * indicate significance at p < 0.01, p < 0.05 and p <
0.10, respectively.

Panel A: Using TDC1_COMPLEX as measure of complexity

(1) (2) (3) (4)

VARIABLES 2006-2008 2010-2012

lnMV 0.01*** -0.01** 0.01*** -0.00

(2.62) (-2.12) (3.24) (-0.90)
MTB -0.01*** -0.00 -0.02*** -0.01***
(-3.53) (-0.40) (-4.15) (-3.08)
R&D -0.13** -0.08 0.00 0.02
(-2.44) (-1.19) (0.07) (0.26)
lnSTDRET -0.59*** -0.62*** -0.20*** -0.34***
(-6.25) (-5.54) (-2.98) (-4.37)
lnSEGMTS 0.02*** 0.01** 0.01 0.01
(3.55) (2.07) (1.43) (1.39)
SEGMTS_HH 0.00 0.02*** 0.01 0.01
(0.17) (2.60) (1.22) (0.66)
PIFO 0.03*** 0.01 0.02** 0.01
(3.29) (1.46) (2.17) (1.17)
lnTenure -0.01*** -0.02***
(-2.83) (-3.84)
RETIRE -0.03** -0.03**
(-2.11) (-2.07)
FOUNDER -0.03** -0.03**
(-2.48) (-2.06)
NumCONSULT 0.01* 0.01**
(1.85) (2.54)
EXCESSPAY 0.05*** 0.05***
(7.56) (5.99)
FOG -0.00 -0.01***
(-1.22) (-3.10)

Observations 4,651 3,157 4,512 3,607

Adjusted R-squared 0.0921 0.1417 0.0689 0.1219

42

 Table 6 (continued)

Panel B: Using COMPLEX_IL as measure of complexity

(1) (2) (3) (4)

VARIABLES 2006-2008 2010-2012

lnMV 0.24*** 0.20*** 0.29*** 0.23***

(4.51) (3.39) (4.75) (3.90)
MTB -0.30*** -0.21*** -0.28*** -0.19***
(-5.85) (-4.15) (-4.05) (-2.81)
R&D 0.75 1.70 0.58 0.12
(0.58) (1.23) (0.49) (0.09)
lnSTDRET -3.79** -2.59 -0.69 -2.10*
(-2.24) (-1.25) (-0.55) (-1.70)
lnSEGMTS 0.13 0.08 -0.02 -0.05
(1.48) (0.86) (-0.21) (-0.52)
SEGMTS_HH 0.01*** 0.31* -0.06 -0.20
(4.30) (1.93) (-0.41) (-1.29)
PIFO 0.29** 0.03 0.18 -0.06
(2.08) (0.21) (1.16) (-0.37)
lnTenure -0.27*** -0.32***
(-3.28) (-3.60)
RETIRE -1.04*** -0.83***
(-3.96) (-3.47)
FOUNDER -0.58*** -0.50**
(-3.16) (-2.49)
NumCONSULT 0.09 0.36***
(0.97) (3.49)
EXCESSPAY 0.79*** 0.93***
(8.62) (9.52)
FOG -0.03 -0.07**
(-1.02) (-2.38)

Observations 2,390 1,792 2,354 1,922

Adjusted R-squared 0.1002 0.2206 0.0834 0.1913

43