Many people today believe that U.S. executives are paid too much while others believe that the size of their compensation packages are justified. For this assignment, choose a peer-reviewed article to
JOURNAL OF MANAGEMENT ACCOUNTING RESEARCHAmerican Accounting AssociationVol. 32, No. 1DOI: 10.2308/jmar-52477
Spring 2020
pp. 177–201
Heaping of Executive Compensation
Bjorn N. Jorgensen
The London School of Economics and Political Science
Paige H. Patrick
University of Illinois at Chicago
Naomi S. Soderstrom
The University of Melbourne
ABSTRACT:We investigate whether compensation grants are subject to‘‘heaping,’’the tendency of less informed
individuals to provide round values when reporting estimates of discrete data. We document that an unexpectedly large
number of CEOs receive round compensation (i.e., evenly divisible by 100,000 and/or 10,000). We investigate
whether, consistent with heaping, the frequency of round compensation varies with proxies for boards of directors’
effort in setting compensation. We find that round compensation is more common when boards have characteristics
suggesting they provide weak oversight of compensation and thus face more uncertainty in estimating compensation.
We also find less frequent round compensation when boards face stronger pressure from external stakeholders,
encouraging boards to expend additional cognitive effort in setting compensation. Further, consistent with weak
oversight of compensation, round compensation tends to be higher than non-round compensation. However, we do not
find a consistent association between this higher, round compensation and future firm performance.
JEL Classifications:G30; G41; M40; M46.
Data Availability:Data are available from the public sources cited in the text.
Keywords:executive compensation; heaping; rounding; say on pay.
I. INTRODUCTION
E
mpirical research on executive compensation typically appeals to models based on agency theory to help explain how
boards of directors incentivize executives. These models assume numerous functional forms for compensation.
However, to our knowledge, none of these models predict that the distribution of compensation will include
discontinuities. Some models do predict that individual CEO compensation may re ect discontinuous earnings distributions
(e.g.,Healy 1985;Melumad 1989;Fedyk 2007); however, these models do not suggest that the distribution of CEO
compensation across many heterogeneous rms and years will include discontinuities.
While analytical models based on the agency theory framework do not predict discontinuities in the distribution of
executive compensation, research that re ects behavioral phenomena does predict that discontinuities will occur. We
investigate whether the heuristic response to uncertainty known as‘‘heaping’’partially explains the observed distribution of
We thank two anonymous referees, Shane S. Dikolli (editor), Brian Cadman, Frank Moers, Steve Rock, Hongtao Shen (discussant, 2014 JMAR/CJAS
conference in Wuhan, China), and seminar participants at the University of Aachen, the Universities of British Columbia, Oregon, and Washington
Conference, the University of Colorado Boulder, The University of Edinburgh, the University of Hohenheim, IE Business School, INSEAD, LaTrobe
University, The London School of Economics and Political Science, Maastricht University, Macquarie University, Monash University, Peking University,
and The University of Queensland for helpful comments.
Bjorn N. Jorgensen, The London School of Economics and Political Science, London, England, U.K.; Paige H. Patrick, University of Illinois at Chicago,
Department of Accounting, Chicago, IL, U.S.A.; Naomi S. Soderstrom, The University of Melbourne, Department of Accounting, Melbourne, Victoria,
Australia.
Editor’s note: Accepted by Shane S. Dikolli, under the Senior Editorship of Karen L. Sedatole.
Submitted: August 2017
Accepted: April 2019
Published Online: July 2019
177 executive compensation. 1Heaping is the tendency to provide estimates ending in digits that are the largest divisors of the base
number system. Heaping arises when people make point estimates of measures about which they lack precise information
(Turner 1958). That is, when individuals have imprecise information, they provide estimates that follow a predictable pattern:
the estimates tend to be multiples of the base of the number system, i.e., round numbers. The existence of heaping is widely
acknowledged in disciplines such as demography, where researchers frequently rely on self-reported information by
imperfectly informed individuals. Studies in diverse elds nd evidence of heaping when individuals are uncertain about the
actual values of measures, such as income (Turner 1958), birthweight (Barreca, Guldi, Lindo, and Waddell 2011), blood
pressure, and age (Roberts and Brewer 2001). For example, individuals are less likely to provide round estimates when they
provide their own ages but are more likely to provide round estimates when they provide estimates of others’ ages. In
accounting, empirical researchers document discontinuities in distributions of analyst and management forecasts, both of which
are reported estimates (e.g.,Herrmann and Thomas 2005;Bamber, Hui, and Yeung 2010;Zhou 2010;Dechow and You 2012;
Athanasakou and Simpson 2016). Consistent with heaping, these studies nd that both analysts and management tend to report
round values of forecasted earnings when they are less certain of the estimates they are providing.
Prior literature on executive compensation in accounting and nance largely explores whether compensation is structured
in a fashion consistent with agency theory models that represent competitive market forces, or with managerial power over
compensation (seeFrydman and Jenter 2010for a review). We build upon the existing understanding of how boards set
compensation by investigating the role of heaping in determining compensation. We explore whether the distribution of
compensation is consistent with heaping behavior. If executive compensation re ects heaping, the distribution of executive
compensation will exhibit abnormally high levels of round compensation (i.e., compensation that is a multiple of the base ten
numerical system) and these round values of compensation will vary predictably with proxies for the level of directors’
uncertainty about executive performance and optimal compensation design.
To address whether distributions of compensation exhibit abnormally high levels of round compensation, we present
evidence of the frequency of round CEO compensation values. We document that an unusually high number of executives earn
round compensation.
2Figure 1 illustrates this phenomenon for different components of CEO compensation. In our sample,
36.1 percent of CEO salaries, 40.9 percent of bonuses, and 23.8 percent of option grants are evenly divisible by either 100,000
or 10,000.
3This phenomenon does not appear to be based upon a mechanical relation between compensation and the
underlying performance metrics upon which it is based, since two commonly used performance measures, earnings per share
and stock price, end in zero only 9.3 percent and 11.3 percent of the time, respectively. Thus, we provide evidence that
executive compensation exhibits abnormally high frequencies of round compensation.
We perform three sets of analyses to assess whether round values of compensation vary predictably with proxies for the
level of directors’ uncertainty about executive performance and the design of optimal compensation, which would suggest that
the distribution of CEO compensation is consistent with heaping. First, we consider whether round compensation is more
common when boards have characteristics that indicate weak oversight of compensation. In these rms, we expect that
directors will be less likely to have acquired and processed the information necessary to make precise estimates about executive
performance. This may arise for a number of reasons: (1) weak boards are unlikely to challenge powerful executives who take
measures to obfuscate relevant information about performance; (2) directors on such boards are less likely to be responsive to
pressure from outside parties; or (3) such directors are not motivated by reputation concerns. Thus, if round compensation
results from heaping, it will be positively associated with proxies for weak board oversight. A substantial stream of literature
demonstrates that certain board characteristics are consistent with weaker oversight of compensation (e.g.,Core, Holthausen,
and Larcker 1999;Coles, Daniel, and Naveen 2014). Our evidence is consistent with this prediction. We show that round
compensation is more common when the board has a greater proportion of insiders and a greater proportion of directors
appointed by the CEO.
Second, we investigate whether stakeholder pressure is associated with round compensation. Prior research suggests that
boards increase attention and effort in setting compensation in the presence of more active stakeholders (e.g.,Johnson, Porter,
and Shackell 1997;Thomas and Martin, 1999;Core, Guay, and Larcker 2008;Ertimur, Ferri, and Muslu 2011;Abernethy,
Kuang, and Qin 2015;Ertimur, Ferri, and Oesch 2018). If boards increase the cognitive effort they expend on setting
compensation when they face greater stakeholder pressure, they will be more informed, reducing uncertainty about executive
performance and about how to set optimal compensation. As such, if round compensation is consistent with heaping, greater
1While heaping is a behavior that arises in the face of uncertainty, we do not argue that heaping is necessarily economically irrational. As we discuss
below, heaping may be consistent with boards expending ef cient levels of effort on costly information acquisition and processing.
2We limit our measure of rounding to compensation components that are rounded to 100,000s or 10,000s of dollars or shares granted to capture
meaningful levels of rounding.
3We exclude observations of $0 or 0 shares from our analyses. Thus, these values represent the frequencies with which non-zero compensation is
rounded.
178Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 stakeholder pressure will be associated with a lower incidence of round compensation. We examine the change in the frequency
of round compensation following passage of the Say on Pay provision of the Dodd-Frank Act, 4which requires an advisory
shareholder vote on executive compensation packages. This passage of compensation-related regulation is a reasonable proxy
for changes in stakeholder pressure because it is at least partially exogenous to individual rms’ governance environments.
Additionally, this identi cation strategy exploits within- rm variation in stakeholder scrutiny surrounding regulatory changes,
allowing us to make stronger inferences about the effect of stakeholder scrutiny on the frequency of rounding. We nd evidence
consistent with our expectations. Boards are less likely to grant round values of compensation components after passage of Say
on Pay, particularly for performance-based pay, which was the focus of that regulation. In sensitivity tests, we nd a similar
reduction in rounding of option grants following passage ofStatement of Financial Accounting Standards No. 123R(SFAS
123R)—Share-Based Payment(FASB 2004).
5
Third, we investigate whether, controlling for rm performance, round CEO compensation tends to be higher than non-
round CEO compensation.Bebchuk and Fried (2003)suggest that self-interested managers take advantage of opportunities to
extract rents from rms in the form of excess compensation. Round compensation as a result of heaping suggests that boards
granting round compensation face more uncertainty when setting compensation than do boards granting non-round
compensation. Self-interested CEOs may take advantage of this uncertainty to gain excess compensation. This suggests that
round compensation will be greater than non-round compensation for a given level of performance. We nd evidence
FIGURE 1
Distributions of Compensation
Panel A: CEO Salary Compensation
Salary is as reported by ExecuComp in $1,000s. The width of bins in the histogram is $20,000. The sample size for this histogram is 10,299 CEO-years
from 2007 to 2014. For legibility, we exclude 122 observations in which CEOs were granted more than $2,000,000 in salary compensation, resulting in a
sample size of 10,177 observations.
(continued on next page)
4Dodd-Frank Wall Street Reform and Consumer Protection Act (U.S. Congress 2010).5Statement of Financial Accounting Standards No. 123: Accounting for Stock-Based Compensation(FASB 1995) allowed rms to account for option
compensation using the intrinsic value method on the option grant date (usually zero) and required disclosure of the grant date fair value of option
compensation expense. SFAS 123R required rms to recognize option expense at grant date fair value rather than intrinsic value.
Heaping of Executive Compensation179
Journal of Management Accounting Research
Volume 32, Number 1, 2020 consistent with this prediction. After controlling for rm-level determinants of compensation, we show that total compensation
is higher when it is round.
These analyses provide consistent evidence that heaping frequently occurs when boards set executive compensation. Our
nal analysis investigates whether heaping is detrimental to rm performance. On the one hand, heaping may be associated
with poorer future rm performance if heaping occurs when boards do not provide strong oversight of compensation, allowing
self-interested managers to extract material rents via compensation. On the other hand, heaping of compensation may result
from an ef cient allocation of board attention. Boards have many tasks in addition to setting CEO compensation, but a limited
amount of attention to allocate across these competing tasks. If these competing tasks are economically important, a lack of
attention to the details of compensation can be desirable. We nd mixed evidence of an association between round
compensation and future rm performance. As such, we are unable to conclude whether, on average, heaping of executive
compensation is detrimental to rm performance, or represents ef cient allocation of limited board attention across multiple
tasks.
In sum, we provide evidence that, in addition to being driven by agency cost concerns and managerial power as shown in
prior literature, executive compensation also re ects behavioral tendencies. Speci cally, we demonstrate that executive
compensation exhibits characteristics consistent with heaping, which is the tendency of uninformed individuals to provide
round numbers when they estimate discrete, quantitative data. We show that round compensation is common and that the
attributes of this round compensation are consistent with less-informed boards. These ndings contribute to our understanding
of how boards determine compensation. We also contribute to the literature on discontinuities in executive compensation
distributions. Prior research documents that an unusually high number of executives earn salaries of $1 million (Rose and
Wolfram 2000,2002) or of $0/$1 (Hamm, Jung, and Wang 2015;Loureiro, Makhija, and Zhang 2015). In addition to these
focal points, we document an unusually high number of executives earning round compensation, that is, compensation
components that are rounded to 100,000s or 10,000s of dollars or shares granted.FIGURE 1 (continued)
Panel B: CEO Bonus Compensation
Bonus is as reported by ExecuComp in $1,000s. The width of bins in the histogram is $20,000. The initial sample size for this analysis is 1,998 non-zero
bonus grants. For legibility, we exclude 67 observations in which CEOs were granted more than $5,000,000 in bonus compensation, resulting in a nal
sample size of 1,931 observations.
(continued on next page)
180Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 We also extend the literature on heaping to a new setting—compensation. Heaping has been widely acknowledged in other
disciplines, and has received some attention in the nancial accounting literature, as well. Prior studies in accounting nd
evidence of heaping in analyst and management forecasts (e.g.,Herrmann and Thomas 2005;Bamber et al. 2010;Zhou 2010;
Dechow and You 2012;Athanasakou and Simpson 2016). We expand upon these prior studies by demonstrating that heaping
affects additional rm stakeholders, boards of directors and rm executives.
II. LITERATURE REVIEW
Psychology literature explores how individuals retrieve uncertain values from memory, and explains individuals’ reliance
on arithmetic prototypes (i.e., round numbers) when reporting these uncertain values (e.g.,Huttenlocher, Hedges, and Bradburn
1990). Social sciences research also observes this behavior when individuals make estimates of uncertain information, even
when individuals are not retrieving this information from memory.
6The social science literature coins the term‘‘heaping’’to
describe the tendency to provide estimates ending in digits that are the largest divisors of the base number system. In this
literature, heaping is considered a heuristic response to uncertainty that arises when people make point estimates of measures
about which they lack precise information.Simon (1990)andShah and Oppenheimer (2008)link heuristics to reduced effort
associated with decision processes.FIGURE 1 (continued)
Panel C: CEO Non-Equity Incentive Compensation
Non-equity incentive compensation is as reported by ExecuComp, in $1,000s. The width of bins in the histogram is $20,000. The initial sample size for
this analysis is 7,889 non-zero non-equity incentive grants. For legibility, we exclude 312 observations in which CEOs were granted more than $5,000,000
in non-equity incentive compensation, resulting in a nal sample size of 7,577 observations.
(continued on next page)
6For example, individuals tend to report round numbers for estimates of income (Turner 1958), age (e.g.,Shryock, Siegel, and Larmon 1980;Roberts
and Brewer 2001;A’Hearn, Baten, and Crayen 2009), weight and height (Rowland 1990), and blood pressure (Wen, Kramer, Hoey, Hanley, and Usher
1993).
Heaping of Executive Compensation181
Journal of Management Accounting Research
Volume 32, Number 1, 2020 In accounting, a handful of studies investigate whether heaping is present in distributions of earnings estimates.Herrmann
and Thomas (2005 ),Zhou (2010), andDechow and You (2012) nd that analysts’ forecasts of EPS exhibit more clustering
around multiples of ve cents than do actual EPS. Consistent with heaping, these studies nd that analysts who round EPS
forecasts appear to be less informed, exert less effort, and have fewer resources.Bamber et al. (2010)investigate whether
heaping is present in management’s earnings forecasts. In addition to re ecting uncertainty, they nd that round management
forecasts are opportunistically biased.
Research in nance and accounting documents in other settings rounding that does not necessarily result from heaping. In
the nance literature, prior research documents clustering of share prices or returns on round values, which may represent either
dealer collusion (Christie, Harris, and Schultz 1994;Christie and Schultz 1994) or naturally occurring discontinuities in
distributions (Osborne 1962;Grossman, Miller, Cone, Fischel, and Ross 1997;E. Johnson, N. Johnson, and Shanthikumar
2011;Bhattacharya, Holden, and Jacobsen 2012).Amiram, Kalay, and Ozel (2013)show that bond coupon rates are set in
increments of eighths. In the nancial accounting literature, researchers nd patterns consistent with rounding of unscaled
reported earnings and of reported earnings per share (EPS).
7These researchers ascribe the frequency of rounding to earnings
management to meet or beat earnings benchmarks.Zhou (2010)andCheong and Thomas (2011)document unusual patterns in
forecasts of EPS.Amiram, Bozanic, and Rouen (2015) nd that deviations of nancial statement information from predicted
distributions indicate nancial reporting irregularities.FIGURE 1 (continued)
Panel D: CEO Option Grants
Options are the number of options granted as reported by ExecuComp, in 1,000 shares. The width of bins in the histogram is 20,000. The initial sample
size for this analysis is 5,540 non-zero option grants. For legibility, we exclude 36 observations in which CEOs were granted more than 2,000,000 options,
resulting in a nal sample size of 5,504 observations.
(continued on next page)
7See, among others,Carslaw (1988),Thomas (1989),Hayn (1995),Burgstahler and Dichev (1997 ),Degeorge, Patel, and Zeckhauser (1999),Das and
Zhang (2003),Bamber et al. (2010),Jorgensen, Lee, and Rock (2014 ),Malenko and Grundfest (2014 ), andBurgstahler and Chuk (2015,2017).
182Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 Prior research also explores discontinuities in the distribution of executive compensation that are not related to heaping:
salaries of $0/$1 and of $1 million.Hamm et al. (2015)andLoureiro et al. (2015)investigate the prevalence of and reasons
for salary grants of $0 or $1. More closely related to our research,Rose and Wolfram (2000,2002)andPerry and Zenner
(2001)identify increased clustering in executive salaries at $1 million subsequent to introduction of Internal Revenue Code
Section 162(m), which limits the tax deductibility of top executives’ non-performance-based compensation to $1 million per
year.
III. HYPOTHESIS DEVELOPMENT
Is the Distribution of Executive Compensation Consistent with Heaping?
Determining executive compensation schemes and values is a substantial task for boards of directors and the information
acquisition and processing required for this task are costly activities for directors. However, pressure from other board
members, shareholders, other stakeholders, or reputation concerns can provide directors incentives to engage in information
acquisition and processing, i.e., to gather and understand more, or more precise, information (e.g.,Bertrand and Mullainathan
2001;Bebchuk and Fried 2003). Directors who undertake less information acquisition or spend less time processing
information related to executive performance and compensation practices will be, on average, less informed than those who
expend more effort, resulting in greater uncertainty about CEO performance and how to reward it. This leads to our rst
research question: Is the distribution of executive compensation consistent with heaping? We generate three hypotheses from
this research question.FIGURE 1 (continued)
Panel E: CEO Equity Incentive Grants
Equity incentive shares are the number of incentive shares granted as reported by ExecuComp in 1,000 shares. The width of bins in the histogram is
10,000. The initial sample size for this analysis is 4,991 non-zero equity incentive grants. For legibility, we exclude 209 observations in which CEOs were
granted more than 500,000 shares, resulting in a nal sample size of 4,782 observations.
(continued on next page)
Heaping of Executive Compensation183
Journal of Management Accounting Research
Volume 32, Number 1, 2020 If heaping occurs when boards set executive compensation, we expect to observe a disproportionate number of round
values of compensation grants. We also expect the frequency of round compensation to vary predictably, as we discuss in this
section.
Numerous studies demonstrate that certain characteristics of boards of directors are consistent with weak oversight of
managers and, speci cally, of executive compensation (e.g.,Core et al. 1999;Coles et al. 2014). In such rms, we expect that
directors will be less likely to have acquired and processed the information necessary to make precise, certain estimates about
executive performance. This may arise because boards do not challenge powerful executives who take measures to obfuscate
relevant information about performance, resulting in less precise information for decision-making. This may also arise if the
directors of such boards are less responsive to pressure from outside parties or are not as motivated by reputation concerns. We
thus expect boards with characteristics consistent with weak oversight of compensation to have less, or less precise, information
about executive performance. In turn, if heaping is re ected in compensation, these less-informed boards will be more likely to
provide round values of compensation. This leads to our rst hypothesis:
H1:Rounding of compensation is negatively associated with board strength.
Pressure from shareholders and other outsiders may in uence boards of directors to improve oversight of compensation.
Bertrand and Mullainathan (2001)andBebchuk and Fried (2003)refer to pressure from stakeholders as‘‘outrage.’’According
to this point of view, when faced with pressure from stakeholders, boards are less likely to adopt forms of compensation that
embarrass directors or executives, or that might reduce shareholders’ willingness to re-elect board members. Empirical
evidence suggests that boards adjust compensation practices due to stakeholder pressure.Johnson et al. (1997 ) nd that boards
respond to negative publicity by granting smaller increases in compensation and by increasing the sensitivity of cash pay to
performance.Core et al. (2008)obtain similar results but demonstrate that these effects may be short-lived.Thomas and Martin FIGURE 1 (continued)
Panel F: CEO Other Equity Grants
Other equity shares are the number of other equity shares granted as reported by ExecuComp in 1,000 shares. The width of bins in the histogram is 10,000.
The initial sample size for this analysis is 5,709 non-zero other equity grants. For legibility, we exclude 172 observations in which CEOs were granted
more than 300,000 shares, resulting in a nal sample size of 5,537 observations.
184Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 (1999)document that, following shareholder compensation proposals, boards reduce CEO compensation. Consistent with these
results,Ertimur et al. (2011) nd that shareholder activism targeting executive compensation results in reduced excess CEO
compensation.Ertimur et al. (2018)observe a $1.2 million reduction in CEO compensation following proxy advisors’
recommendations that shareholders withhold votes from directors due to concerns about pay-for-performance sensitivity.
Abernethy et al. (2015) nd that, while rms in the United Kingdom adopted performance-vested stock option plans to quell
public outrage, powerful executives in uence the choice of performance measures in those plans.
Evidence on the effect of stakeholder pressure suggests that boards will increase their oversight efforts and that this
increase in effort extends to compensation. Thus, when stakeholders increase their focus on compensation, we expect boards to
expend more effort gathering information about executive performance and in determining compensation design and payments.
As a result, boards will have more precise information about executive performance, and less uncertainty, resulting in less
round compensation. Thus, if round compensation results from heaping, we expect a negative association between the
frequency of rounding and stakeholder pressure. This leads to our second hypothesis:
H2:Rounding of compensation is negatively associated with stakeholder pressure.
We next investigate the association between the level of compensation and whether compensation is round.Bebchuk and
Fried (2003)suggest that self-interested managers take advantage of opportunities to extract rents from rms in the form of
excess compensation. Consistent withBebchuk and Fried (2003), many studies nd evidence of CEO opportunism when
boards exhibit characteristics of weak oversight (e.g.,Core et al. 1999;Coles et al. 2014;Abernethy et al. 2015). If round
compensation is consistent with heaping, then round compensation values can be indicative of boards that face more
uncertainty when setting compensation. In these cases, self-interested CEOs may take advantage of these boards’ uncertainty to
gain excess compensation. Thus, if round compensation results from heaping, then, for a given level of rm performance,
compensation will be higher when it is round. This leads to our third hypothesis:
H3:Ceteris paribus,round compensation is higher than non-round compensation.
Is Heaping Associated with Firm Performance?
In our rst three hypotheses we investigate whether round compensation arises from a heuristic response to uncertainty,
heaping. We posit that uncertainty about optimal executive compensation arises because information acquisition and processing
are costly, but that pressures from other board members and stakeholders will lead some boards to invest more in information
acquisition and processing about compensation. In this section, we investigate whether heaping is associated with rm
performance. On the one hand, if heaping results when boards do not provide strong oversight of compensation and allow self-
interested managers to extract material rents via compensation, heaping may be associated with poorer rm performance. On
the other hand, heaping of compensation may result from an ef cient allocation of board attention. That is, directors have
numerous responsibilities, all of which require costly information acquisition and processing. As such, some boards may
rationally choose to invest less in information acquisition and processing when setting compensation, as their limited attention
has greater impact when focused on other decisions or activities. Because of the con icting nature of these predictions, we state
our fourth hypothesis in the null form:
H4:Rounding of compensation is not associated with future performance.
IV. SAMPLE AND PRELIMINARY EVIDENCE ON ROUND COMPENSATION FREQUENCY
Sample Design
Our initial sample includes CEOs from rms covered by ExecuComp for the period 2007 through 2014. We begin our
sample in 2007 because the SEC increased compensation disclosure requirements for top executives’ equity holdings and
cash bonus payments beginning in December 2006. We focus our analyses on CEOs because their compensation is likely to
be the most visible. Our sample is reduced due to use of data from RiskMetrics, Compustat, and CRSP for our analyses. The
resulting sample includes 10,897 CEO-years. We further exclude 598 CEOs in their rst year of employment to alleviate
concerns that our results are driven by any tendency to round compensation in the rst year of employment. Our main sample
thus includes 10,299 CEO-years, although we place additional restrictions on our data for several of our tests. We discuss
these exclusions in the research design for each test and include the sample size associated with each test in the tables. Table
1 presents distributions of regression variables. Variable de nitions are presented in the table and are described in the text
below.
Heaping of Executive Compensation185
Journal of Management Accounting Research
Volume 32, Number 1, 2020 Preliminary Evidence on the Frequency of Round Compensation
The rst step in our analysis is to explore the prevalence of round values for all forms of compensation. We consider six
types of compensation grants, salary (ExecuComp variable SALARY ), discretionary bonus (ExecuComp variable BONUS),
non-equity incentive cash compensation (ExecuComp variable NONEQ_INCENT), grants of options (ExecuComp variable
OPTION_AWARDS_NUM), and equity compensation (ExecuComp variables EQ_TARG and SHARES_GRT).
8
Figure 1, Panel A, presents a histogram of salary compensation for our initial sample of 10,299 CEOs. We truncate this
histogram at $2 million for presentation purposes. Consistent withRose and Wolfram (2000,2002), Figure 1, Panel A, suggests
that a disproportionately large number of CEOs earn exactly $1 million. Of the observations in the bin containing $1,000,000
through $1,020,000, 495 represent observations in which CEOs earned exactly $1 million in salary. In addition to the high
frequency of salaries equal to $1 million, Figure 1, Panel A, shows unusually high frequencies of salary compensation at other
values. One large irregularity in the distribution occurs in the rst bin, where salary is $0–$20,000. The distribution of salaries
within the bin is not uniform; the bin is dominated by a signi cant number of CEOs (81) whose cash compensation is no more
than $1. The frequency of these observations is consistent withHamm et al. (2015)andLoureiro et al. (2015 ). However, we
TABLE 1
Distributions of Selected Regression Variables
Variable Mean Min. Median Max. Obs.
PercentRoundAll0.247 0.000 0.250 1.000 10,299
PercentRoundPerformance0.180 0.000 0.000 1.000 9,785
PercentRoundDiscretionary0.361 0.000 0.000 1.000 10,279
Dual0.540 0.000 1.000 1.000 10,299
Insiders0.208 0.000 0.182 0.857 10,299
AppointedbyCEO0.395 0.000 0.375 1.000 10,299
PostSOP0.508 0.000 1.000 1.000 10,299
TotalCompensation6,002.913 0.001 4,147.181 156,077.912 10,299
PerformanceCompensation4,664.123 0.000 3,010.320 151,141.922 10,299
DiscretionaryCompensation1,077.851 0.000 855.000 77,926.000 10,299
ROA
tþ1 0.091 1.223 0.081 1.126 10,299
Compensation components, taken from ExecuComp, are round if they are evenly divisible by 100,000 or 10,000 dollars or shares. The sample size for this
table is 10,299. Not all CEOs were granted either performance-based compensation or discretionary compensation. Following prior literature, we use the
natural logs of compensation variables in our empirical analyses.
Variable De nitions:
PercentRoundAll¼the number of round components in a compensation package divided by the number of compensation components granted;
PercentRoundPerformance¼the number of round performance-based components in a compensation package divided by the number of performance-
based components granted. Performance-based components include non-equity incentive compensation (the variable NONEQ_INCENT), the
number of plan-based stock awards (EQ_TARG), the number of other shares granted (SHARES_GRT), and the number of options awarded
(OPTION_AWARDS_NUM);
PercentRoundDiscretionary¼the number of round discretionary components in a compensation package divided by the number of discretionary
components granted. Discretionary components include salary (SALARY ) and bonuses awarded but not classi ed as part of a performance-based
plan (BONUS);
Dual¼an indicator variable set to 1 if the CEO is also the chairman of the board and 0 otherwise;
Insiders¼the percentage of board members who are insiders;
AppointedByCEO¼the percentage of outside board members appointed by the CEO;
PostSOP¼an indicator variable set to 1 if the observation is from 2011 or later;
TotalCompensation¼compensation as reported in ExecuComp, in thousands of dollars (TDC1);
PerformanceCompensation¼the sum of compensation from the four performance-based compensation components, as reported in ExecuComp
(NONEQ_INCENT, OPTION_AWARDS_FV, and STOCK_AWARDS_FV ), in thousands of dollars. Note that STOCK_AWARDS_FV
summarizes the dollar value of EQ_TARG and SHARES_GRT, which are reported in shares;
DiscretionaryCompensation¼the sum of salary and discretionary bonus compensation from ExecuComp, in thousands of dollars; and
ROA
tþ1 ¼return on assets measured as EBIT/average total assets measured in yeartþ1.
8We assume that boards grant equity compensation in numbers of shares rather than in estimated dollar values. We thus study the number of options or
stock shares granted. This assumption biases us against nding signi cant results if rms grant estimated dollar values of equity instead of numbersof
options or shares.
186Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 note unusually high frequencies of salary compensation throughout the distribution, primarily at values evenly divisible by
$100,000 and, in particular, around $500,000, $1.5 million, and $2 million.
We note similar irregularities around values divisible by $500,000 in the histograms of bonus compensation (Panel B) and
non-equity incentive compensation (Panel C ). The pattern is less obvious when we investigate the distributions of equity
granted in Panels D through F, although we observe irregularities at values evenly divisible by 100,000 shares. Overall, the
histograms of the distributions of cash and equity compensation provide preliminary evidence that compensation tends to be
disproportionately divisible by 100,000.
Since our histograms present compensation levels in $20,000 bins, we cannot conclude that the spikes we observe are due
to rounded compensation (i.e., compensation that is evenly divisibly by 100,000 or 10,000 units). In further examination of the
spikes in the histograms, we investigate the prevalence of observations that end in zero.
9We restrict this analysis to non-zero
values of each type of compensation grant. In Table 2, we present the frequency of the last digit being zero based on whether
the amount is evenly divisible by $1,000,000 (or 1,000,000 shares), $100,000 (or 100,000 shares), or $10,000 (or 10,000
shares). While rounding to values evenly divisibly by 1,000, 100, or 10 is common in our sample, we treat these observations as
non-round in subsequent analyses because they are unlikely to be economically signi cant to either rms or executives.
10 For
ease of exposition, we refer to values evenly divisible by 10,000 units as round.
Table 2 demonstrates that discretionary bonuses are the most likely form of compensation to be round (40.9 percent of
non-zero observations are evenly divisible by $10,000), followed by salary (36.1 percent of observations are evenly divisible
by $10,000). We note that forms of compensation that are more likely to be qualitative assessments of CEO performance (i.e.,
salary and discretionary bonuses) are those forms of compensation that are most likely to be round. This is consistent with
heaping, because these forms of compensation more naturally lend themselves to estimation, presenting opportunity for
heuristic responses to come into play. Performance-based compensation is a quantitative assessment of CEO performance and
will therefore re ect less uncertainty. If the frequency of round compensation does not result from heaping, we would not
expect the frequency of round compensation to vary systematically with whether the compensation tends to be quantitative or is
more likely to be based on qualitative considerations.
Table 2 also provides evidence that equity incentive shares and other equity are the least likely to be round (10.1 percent
and 11.1 percent of observations are evenly divisible by 10,000 shares, respectively). Non-equity incentive and option grants
are very likely to be round; 18.4 percent of non-equity incentive grants and 23.8 percent of option grants are evenly divisible by
10,000 units. In comparison, two commonly used performance measures, earnings per share and stock price, end in zero only
9.3 percent and 11.3 percent of the time, respectively (not tabulated).
TABLE 2
Frequency of Round Compensation Values for CEO-Year Observations with Compensation Greater Than or Equal to
$100,000
Discretionary Performance-Based
Salary BonusNon-Equity
Incentive OptionsEquity
Incentive
SharesOther
Equity
Shares
Divisible by 1,000,000 ($ or shares) 5.3% 4.9% 2.4% 0.8% 0.2% 0.1%
Divisible by 100,000 ($ or shares) 14.4% 17.6% 7.4% 7.8% 2.6% 1.8%
Divisible by 10,000 ($ or shares) 16.4% 18.4% 8.6% 15.2% 7.3% 9.2%
Percent of observations evenly divisible by 10,000 36.1% 40.9% 18.4% 23.8% 10.1% 11.1%
Number of non-zero observations 10,218 1,998 7,889 5,540 4,991 5,709
This table presents the frequency with which non-zero observations of compensation components are evenly divisible by 10,000. All data are from
ExecuComp. The base sample for this table comprises 10,299 CEO-years from 2007 to 2014.
9Benford’s Law shows that initial digits appear with differing frequencies in many datasets. As discussed inBeer (2009)and shown byDiaconis (1977 ),
this effect dissipates in digits farther from the left-most digit, so that, in large data sets with large values, the expected frequency of most digits
converges to 10 percent in places at least two digits from the left. Since we are interested in the right-most digit, we cannot draw inferences regarding
the expected frequencies of values ending in zero from either Benford’s Law orDiaconis (1977 ).
10 We con rm that our results are qualitatively similar when we de ne round as any value ending in zero, other than zero itself.
Heaping of Executive Compensation187
Journal of Management Accounting Research
Volume 32, Number 1, 2020 Overall, our descriptive analyses lead us to conclude that rounding is extremely common and that rounding occurs not just
around the last digit, but also at much larger magnitudes. In subsequent analyses, we investigate the determinants of rounding
and the associations of rounding with compensation levels and rm performance.
V. TESTS OF HYPOTHESES
Discretionary and Performance-Based Compensation
Our primary analyses focus on components of compensation that tend to be explicitly based on rm performance, rather
than on components that are discretionary in nature. By de nition, performance-based compensation is based on CEO (or rm)
performance; therefore, these grants are the result of a quantitative assessment. If directors expend more effort to acquire or
process information about a CEO’s performance, they should be able to provide more precise values of compensation.
Discretionary compensation, however, is based on a qualitative assessment. Thus, even if directors expend additional cognitive
effort in setting discretionary compensation, substantial uncertainty in the estimate will remain, because the inputs are not
quantitative. That is, additional cognitive effort may not be useful in resolving the uncertainty around discretionary
compensation grants due to their qualitative nature. If round compensation results from heaping, we expect discretionary
compensation to be round more frequently than performance-based compensation regardless of whether the board expends
additional effort to acquire and process information. The descriptive results reported above are consistent with this conjecture.
In addition, because we do not expect discretionary compensation components to vary with proxies for board effort, by
including these analyses, we can provide additional evidence that our results are not spurious.
Round Compensation, Board Oversight, and Stakeholder Pressure
Primary Analyses
Our rst two hypotheses relate to when rms are more likely to grant round compensation. H1 predicts that compensation
is more likely to be round in rms with weaker boards of directors, and H2 predicts that round compensation is negatively
associated with stakeholder pressure. To test these hypotheses, we estimate the following equation:
PercentRound
t¼aþb 1Dual tþb 2Insiders tþb 3AppointedByCEO tþb 4PostSOP tþb 5SDROA tþb 6SDReturns t
þb 7Digits tþb 8NumComponents tþYearFEþIndustryFEþe t ð1Þ
For total compensation, we create a composite measure that captures how much of the CEO’s compensation is round,
PercentRoundAll. PercentRoundAllis the number of round compensation components granted divided by the total number of
compensation components granted.
11 We next aggregate compensation components by whether the compensation granted is
discretionary or performance-based because, as discussed above, our hypotheses should hold only for performance-based
compensation grants. We label salary and discretionary bonus compensation as‘‘discretionary,’’and all other components (i.e.,
non-equity incentive plan compensation, option grants, equity incentive shares, and other equity grants) as‘‘performance-
based.’’PercentRoundPerformance(PercentRoundDiscretionary) is the number of round, non-zero, performance-based
(discretionary) compensation components divided by the total number of non-zero, performance-based (discretionary)
compensation components granted. We present distributions of these variables in Table 1. Overall, approximately 25 percent of
granted compensation components are round. Consistent with Table 2, discretionary compensation is more likely to be round
than is performance-based compensation. On average, 18 percent of performance-based compensation is round, whereas 36
percent of discretionary compensation is round. Sample sizes are smaller when we split compensation into performance-based
and discretionary components because some CEOs do not receive any performance-based compensation, and other CEOs do
not receive any discretionary compensation.
To test the association between board oversight and the frequency of round compensation, we examine measures of board
oversight used in prior research. If the CEO has greater power over the board of directors, the CEO will be more likely to be
able to withhold or obfuscate information about performance without repercussions from board members. In turn, the board
will have less precise information upon which to base estimates of CEO effort. Additionally, board members of rms with
powerful CEOs may be less responsive to outside pressure, or reputation concerns, than they are to pressure from CEOs. Our
11 For example, suppose a CEO were granted three compensation components: round salary, round non-equity incentive compensation, and a non-round
number of options.PercentRoundAllwould take a value of 0.667 (two round components divided by three components granted). Further,
PercentRoundDiscretionarywould take a value of 1, because the only discretionary component granted (salary) is round. Finally,
PercentRoundPerformancewould take a value of 0.5, because one of the two performance-based components is round; non-equity incentive
compensation is round while the option grant is not round.
188Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 rst measure of CEO power over the board isDual,an indicator variable set to one if the CEO is also the chairman of the board
and 0 otherwise. Prior literature posits that these CEOs exert more control over the board of directors (e.g.,Core et al. 1999).
Core et al. (1999)also suggest that boards with more insiders (Insiders) are less independent of management, allowing CEOs
more power over decision-making.Insidersis the percentage of board members who are current or former employees or family
members of current employees. We also include a measure of co-opted boards, measured by the number of outside directors
appointed by the CEO as a percentage of board size (AppointedByCEO). These directors may be more sympathetic to the CEO,
and more likely in uenced by the CEO (Core et al. 1999;Coles et al. 2014). Data on director characteristics are from
RiskMetrics, and data used to determine the CEO hire date are from ExecuComp.
To test the association between round compensation and stakeholder pressure, we include an indicator variable,PostSOP,which
equals 1 if the observation is from 2011 or later, and 0 otherwise. We use the passage of compensation-related regulation on executive
pay to proxy for stakeholder pressure (Sheehan 2012). The Say on Pay provision of the Dodd-Frank Act was passed in response to
increased public concern regarding the nature and reporting of executive compensation. While passage of compensation-related
regulations is the result of public concern, passage is not the result of an individual rm’s governance environment or pay practices.
As a result, prior literature considers these regulation changes as primarily exogenous shocks to governance. These regulatory
changes provide more robust evidence that any association between stakeholder pressure and changes in the frequency of rounding
does not result from endogeneity than do other measures of stakeholder pressure, such as shareholder proposals or media attention.
The Say on Pay provision requires rms to allow shareholders to vote on executive compensation packages. Although results
of the vote required by Say on Pay are only advisory,Ertimur, Ferri, and Oesch (2013)document that more than half of rms with
adverse Say on Pay vote outcomes change their compensation plans.Lo, Yang, and Zhang (2014 ) nd that boards respond to
shareholder disapproval by amending compensation policies. Taken together, these studies suggest that Say on Pay increased
boards’ accountability to shareholders for executive compensation, and therefore likely increased their effort in determining
executive compensation. In our main analyses we investigate changes in the frequency of round compensation around passage of
the Say on Pay provision of the Dodd-Frank Act.
12 Balsam, Boone, Liu, and Yin (2016 ) nd that in anticipation of Say on Pay,
rms made CEO compensation more performance-based.Abernethy et al. (2015)also nd evidence that outrage leads to a greater
emphasis on performance-based pay. This is consistent with our focus on performance-based compensation components.
If round compensation results from heaping and is consistent with weak board oversight (H1), we expect positive
coef cients ofDual(b
1),Insiders(b 2), andAppointedByCEO(b 3) in Equation (1). If rounding is negatively associated with
shareholder pressure (H2), we expect a negative coef cient ofPostSOP(b
4) in Equation (1).
In addition to our variables of interest, Equation (1) includes controls for the level of rounding in compensation that are
unrelated to board effort. The standard deviation of return on assets (SDROA) controls for the noise in performance measures,
because noisy performance measures provide less precise estimates of performance and may result in more heaping by the
board of directors.SDROAis measured as the standard deviation ofROA
tfor the three years ending in yeart,whereROA tis
earnings before interest and taxes scaled by average total assets. We also control for volatility in the operating environment,
which may impair the board’s ability to precisely measure CEO performance. We include the standard deviation of returns
(SDReturns), which is the standard deviation of monthly returns for the 36 months ending in yeart. Two control variables
capture potential mechanical reasons that round compensation might result. First, we control for the number of digits in
compensation granted (Digits), because larger values are more likely to be round. Second, we control for the number of
components granted (NumComponents), though we do not have a prediction for the direction of the expected association. We
include industry and year xed effects in all speci cations.
Because our dependent variables are bounded between zero and one, we followPapke and Wooldridge (1996 )and
estimate Equation (1) using a probit fractional response model.
13 We include year and industry xed effects to control for
unobservable economic factors that are correlated with both the frequency of rounding and the determinants of interest.
Table 3, Panel A presents the main results of estimating Equation (1).
14 We present the results of testing H1 in Columns 1
through 3. In this speci cation we do not include the variablePostSOP. Focusing on the rst column, in which the dependent variable
re ects the level of rounding across performance-based compensation components (PercentRoundPerformance), we nd evidence
consistent with our predictions. We nd a positive and signi cant association ( p,0.10, one-tailed) between two of our proxies for
CEO control over the board of directors,InsidersandDual,which is consistent with our rst hypothesis (H1) that boards round
compensation more when CEOs have more control over boards. However, we nd no evidence that appointment by the CEO is
associated with rounding of performance-based compensation;b
3is not signi cantly different from 0. In Columns 4 through 6 we
add the indicator variable for the passage of Say on Pay. Consistent with H2, we nd that the percentage of round compensation
12 In sensitivity tests, we investigate changes in the frequency of round option grants following implementation of SFAS 123R, which focused on
accounting for option grants.
13 Our inferences are unchanged if we use an ordered logit or ordered probit model instead of a fractional response model. 14 Our results are consistent when we include each variable of interest separately, then stack all variables together. For parsimony, we only report results
when we include all variables in the same regression.
Heaping of Executive Compensation189
Journal of Management Accounting Research
Volume 32, Number 1, 2020 TABLE 3
Predictions of Round Compensation Components
Panel A: Main Effects of Board Strength and Say on Pay
PercentRound
t¼aþb
1Dual
tþb
2Insiders
tþb
3AppointedByCEO
tþb
4PostSOP
tþb
5SDROA
tþb
6SDReturns
tþb
7Digits
tþb
8NumComponents
t
þYearFEþIndustryFEþe
t
ð1Þ
Pred.
SignPercentRoundRefers to:
PercentRoundPerformance
(1)PercentRoundDiscretionary
(2)PercentRoundAll
(3)PercentRoundPerformance
(4 )PercentRoundDiscretionary
(5)PercentRoundAll
(6 )
Dualþ0.052* 0.046 0.009 0.021 0.045 0.008
(1.43) ( 1.15) (0.33) (0.58) ( 1.10) ( 0.28)
Insidersþ0.336** 0.263* 0.434*** 0.284* 0.266* 0.401***
(1.94 ) (1.37 ) (3.28) (1.61) (1.38) (3.00)
AppointedByCEOþ0.063 0.011 0.061 0.084 0.012 0.072*
(0.89) ( 0.15) (1.14 ) (1.18) ( 0.15) (1.34 )
PostSOP 0.344*** 0.016 0.193***
( 8.85) (0.39) ( 6.86 )
SDROAþ1.351*** 0.691* 0.944** 1.350*** 0.691* 0.944**
(2.64 ) (1.34 ) (2.30) (2.61) (1.34 ) (2.29)
SDReturnsþ0.048 0.012 0.035 0.050 0.012 0.034
(0.92) ( 0.17 ) (0.84 ) (0.96 ) ( 0.17 ) (0.82)
Digitsþ0.073*** 0.451*** 0.238*** 0.081*** 0.450*** 0.245***
(2.60) (10.60) (8.96 ) (2.86 ) (10.56 ) (9.19)
NumComponents? 0.211*** 0.050 0.192*** 0.207*** 0.049 0.190***
( 9.94 ) ( 1.13) ( 13.33) (
9.68) ( 1.11) ( 13.15)
Constant 0.946*** 3.277*** 1.837*** 0.791*** 3.285*** 1.762***
( 4.71) ( 12.43) ( 9.52) ( 3.87 ) ( 12.44 ) ( 9.10)
Observations 9,785 10,277 10,299 9,785 10,277 10,299
R
2
0.056 0.053 0.079 0.063 0.053 0.082
(continued on next page)
190Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 TABLE 3 (continued)
Panel B: Predictions of Round Compensation Components Including Interaction Terms withPostSOP
Pred.
Sign (1) (2) (3)
Dualþ 0.005 0.034 0.004
( 0.10) ( 0.68) ( 0.11)
Post
Dual? 0.052 0.019 0.009
(0.93) ( 0.32) ( 0.23)
Insidersþ 0.037 0.408** 0.282**
( 0.18) (1.85) (1.81)
Post
Insiders? 0.732*** 0.294 0.263*
(2.92) ( 1.14 ) (1.46 )
AppointedbyCEOþ0.127* 0.002 0.103*
(1.43) ( 0.02) (1.46 )
Post
AppointedbyCEO? 0.092 0.019 0.059
( 0.85) ( 0.16 ) ( 0.73)
PostSOP 0.490*** 0.095 0.220***
( 5.76 ) (1.07 ) ( 3.61)
SDROAþ1.370*** 0.689* 0.949**
(2.62) (1.34 ) (2.29)
SDReturnsþ0.048 0.011 0.033
(0.91) ( 0.16 ) (0.79)
Digitsþ0.082*** 0.450*** 0.245***
(2.91) (10.56 ) (9.20)
NumComponents? 0.206*** 0.049
0.190***
( 9.64 ) ( 1.10) ( 13.15)
Constant 0.728*** 3.329*** 1.752***
( 3.52) ( 12.46 ) ( 8.92)
Observations 9,785 10,277 10,299
R
2
0.064 0.054 0.082
***, **, * Indicate statistical signi cance at the 1 percent, 5 percent, and 10 percent levels (one-tailed), respectively.
In Panel A the dependent variablePercentRoundrefers toPercentRoundPerformance(Columns 1 and 4 ),PercentRoundDiscretionary(Columns 2 and 5), orPercentRoundAll(Columns 3 and 6 ), as
de ned in the notes to Table 1. In Panel B the dependent variablePercentRoundrefers toPercentRoundPerformance(Column 1),PercentRoundDiscretionary(Column 2), orPercentRoundAll(Column 3),
as de ned in the notes to Table 1. The sample size for this table is 10,299 CEO-years from 2007 to 2014; we exclude observations in which the CEO was not granted discretionary ( performance-based)
compensation from Column 1 (Column 2). We estimate Equation (1) using a probit fractional response model. Standard errors are clustered by rm. This analysis includes industry and year xed effects.
Variable De nitions:
Dual¼an indicator variable set to 1 if the CEO is also the chairman of the board and 0 otherwise;
Insiders¼the percentage of board members who are insiders;
AppointedByCEO¼the percentage of outside board members appointed by the CEO;
PostSOP¼an indicator variable set to 1 if the observation is from 2011 or later;
SDROA¼the three-year standard deviation ofROA
tmeasured as EBIT/average total assets measured in yeart;
SDReturns¼the standard deviation of the prior 36 months of returns, ending in the year for which compensation is granted;
Digits¼the number of digits in values ofPerformanceCompensation, DiscretionaryCompensation,orTotalCompensation,respectively; and
NumComponents¼the number of performance-based compensation components (Column 1), the number of discretionary compensation components (Column 2), or number of compensation
components in a compensation package (Column 3).
Heaping of Executive Compensation191
Journal of Management Accounting Research
Volume 32, Number 1, 2020 components is negatively associated with the passage of Say on Pay ( p,0.01, one-tailed), suggesting that boards round
compensation less when stakeholders pay more attention. These associations are consistent with the explanation that round
compensation results from heaping: the frequency of round compensation in performance-based compensation is positively
associated with characteristics of boards that provide weak oversight of compensation, and negatively associated with shareholder
pressure.
The evidence in Columns 2 and 3, and 5 and 6 is also consistent with our expectations. For example, in Column 6, where
the outcome variable isPercentRoundAll,we detect positive associations between both the fraction of the board appointed by
the CEO, and the fraction of the board who are insiders. We also nd that the associations between round compensation and
proxies for weak board oversight and stakeholder pressure tend to be driven by performance-based compensation, rather than
by discretionary compensation. That is, with one exception, the associations between our variables of interest and the outcome
variable are signi cant in Columns 3 and 6, where the outcome variable isPercentRoundAll,but are not signi cant in Columns
2 and 5, where the outcome variable isPercentRoundDiscretionary. The exception is that round discretionary compensation is
more common when the board has a higher proportion of insiders.
We also nd that the association between the noise in performance measures (SDROA) and the percentage of round
compensation components is positive and signi cant (b
5.0) in all speci cations, again driven by performance-based
compensation, which is consistent with boards of directors heaping when performance measures are noisier, resulting in more
uncertainty about CEO performance. We detect no association between the standard deviation of returns (SDReturns) and the
frequency of rounding.
15
In additional analyses, presented in Table 3, Panel B, we consider whether the change in the level of rounding following
passage of Say on Pay was the result of changes in associations between board characteristics and rounding frequency. To do
so, we estimate a version of Equation (1) in which we interact all variables with the post-Say on Pay indicator,PostSOP.
Speci cally, the increase in stakeholder pressure may have affected rms with weaker oversight more strongly than rms with
stronger oversight. If this is the case, the interaction terms ofPostSOPand our measures of oversight (Dual, Insiders,and
AppointedbyCEO) will be negative, and the main effect of thePostSOPindicator will be zero. We nd some evidence that the
positive association betweenInsidersand the level of round performance-based compensation is driven by observations from
the post-Say on Pay period, and some evidence that board members appointed during the CEO’s tenure were more likely to
grant round performance-based compensation prior to Say on Pay. However, no interaction term betweenPostSOPand any of
our measures of oversight is negative. We continue to detect a signi cantly negative association betweenPostSOPand the
frequency of round compensation. These results suggest that, on average, boards were affected by the increased pressure from
stakeholders associated with Say on Pay regardless of the strength of their compensation oversight.
Additional Tests of Hypothesis 2
We perform two additional tests of the effect of changes in the frequency of rounding following the passage of
compensation-related regulation. In these additional tests, we examine changes in the frequency of rounding around Say on Pay
and around SFAS 123R, an earlier regulatory change that took effect in December 2005, affecting the treatment of option
compensation expense.
As an identi cation strategy, we use within- rm analysis to isolate the effect of shareholder pressure on rounding. In this
changes analysis, we hold the rm constant, which controls for rm-level effects and facilitates identi cation of the impact of
compensation-related regulations on boards’ effort in setting compensation. FollowingJagolinzer (2009), our comparisons
within rms do not require assumptions about the functional form of how rm, board, corporate governance, and executive
characteristics map into outcomes. Holding the rm constant between the pre- and post- periods also controls for the
compensation size. Compensation size may impact the likelihood of rounding, since higher magnitudes of compensation are
more likely to be round.
Our rst test of the effect of compensation-related regulation on rounding focuses on whether, on average, boards
reduced heaping following passage of Say on Pay. Our sample for this test includes two years before and two years after the
adoption of Say on Pay, resulting in an initial sample size of 6,986 CEO-years; we again exclude rst-year CEOs. We next
partition this sample into pre-Say on Pay (2009–2010) and post-Say on Pay (2011–2012). For tests of the change in the
frequency of round compensation for each type of compensation, we exclude grants of $0, $1, or zero shares from our
15 In unreported analyses, we estimate Equation (1), excludingDigits,by compensation decile. Results fail to provide evidence that our results are driven
by the size of compensation granted. We estimate Equation (1) by compensation decile rather than number of digits, because we nd evidence that
some rms manage compensation up or down around round values. When we estimate Equation (1) by compensation decile, our results are consistent
with the analyses we present in Table 3. We note that the associations between our variables of interest and rounding of discretionary compensation
components are stronger, however.
192Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 analyses. We further restrict the sample to include only rms that granted each component of compensation in every year
from 2009 through 2012.
Table 4, Panel A reports the results of tests of the association between passage of Say on Pay and round compensation
frequency. For each type of compensation (i.e., performance-based, discretionary, and total), we compare the mean of the
frequency of round compensation values in the pre- and post-Say on Pay periods. Consistent with our expectations, we nd
that boards were less likely to grant round compensation amounts after Say on Pay. Our measure of rounding for
performance-based compensation components is signi cantly lower following passage of Say on Pay; it decreased from 21.3
percent in the pre-Say on Pay period to 16.6 percent in the post-Say on Pay period ( p,0.01). While we nd no statistically
signi cant effect for discretionary compensation components, our composite measure of rounding is signi cantly lower
following passage of Say on Pay; it decreased from 26.9 percent in the pre-Say on Pay period to 24.2 percent in the post-Say
on Pay period ( p,0.01).
In Table 4, Panel B we investigate the change in the frequency of round option grants (a performance-based compensation
component) from before and after SFAS 123R. We limit this test to option grants because SFAS 123R speci cally addressed
option compensation, unlike Say on Pay, which had implications for all forms of compensation. For this test, we include the
years 2004 through 2007 and partition the sample into pre-SFAS 123R (2004–2005) and post-SFAS 123R (2006–2007 )
periods. We exclude observations of zero options from our analysis, resulting in a sample size of 4,169 CEO-years with non-
zero option grants. We again restrict our sample to include rms that granted options in each year of the sample period.
16 We
nd that boards are less likely to grant round option compensation following passage of SFAS 123R; the frequency of rounding
to amounts evenly divisible by 100,000 or 10,000 decreased from 43.3 percent to 37.6 percent. We nd that this change is
signi cantly different from zero ( p,0.01, one-tailed). Taken together, the evidence in Table 4 indicates that regulation that
highlights the treatment of compensation is associated with lower frequencies of round compensation. This, in turn, is
TABLE 4
Changes in Frequency of Round Compensation Values around Passage of Compensation-Related Regulation
Panel A: Change in the Frequency of Round Compensation Values Pre- and Post-Say on Pay
Percent of Round
Performance-Based
Components
PercentRound
PerformancePercent of Round
Discretionary
Components
PercentRound
DiscretionaryPercent of
Round
Components
PercentRound
All
Pre-Say on Pay % Rounded 21.3% 35.2% 26.9%
Observations 2,354 2,736 2,746
Post-Say on Pay % Rounded 16.6% 36.6% 24.2%
Observations 2,354 2,736 2,746
Pre- versus Post- t-statistics 5.376*** 1.131 3.518***
Panel B: Change in the Frequency of Round Number of Options Granted Pre- and Post-SFAS 123R
Pre-SFAS 123R % Rounded 43.3%
Observations 866
Post-SFAS 123R % Rounded 37.6%
Observations 866
Pre- versus Post- t-statistics 2.401***
*** Indicate statistical signi cance at the 1 percent level (one-tailed).
The frequencies reported in Panel A refer toPercentRoundPerformance(Column 1),PercentRoundDiscretionary(Column 2), orPercentRoundAll
(Column 3), as de ned in the notes to Table 1. The frequencies reported in Panel B are the percentage of non-zero option grants that are evenly divisibleby
100,000 or 10,000 shares. The pre-SOP period is 2009–2010, and the post-SOP period is 2011–2012. The pre-SFAS 123R period is 2004–2005, and the
post-SFAS 123R period is 2006–2007. The sample size for Panel A is 6,986 CEO-years; only non-zero values of compensation are included in this
analysis, and we exclude rst-year CEOs. We further restrict the sample for Panel A to include only rms that granted each component of compensation in
2009, 2010, 2011, and 2012. The sample size for Panel B is 6,676 CEO-years from 2004 to 2007, of which 4,169 are non-zero option grants. We further
restrict the sample for Panel B to include only rms that granted options in 2004, 2005, 2006, and 2007.
16 Our results are qualitatively similar if we do not restrict the sample such that each rm appears in both the pre- and post-SFAS 123R periods.
Heaping of Executive Compensation193
Journal of Management Accounting Research
Volume 32, Number 1, 2020 consistent with the round compensation we observe resulting from heaping: when boards increase the effort they expend in
setting compensation, they are less likely to provide round compensation grants.
Round Compensation and Compensation Levels
To test H3, that compensation is higher when boards grant round compensation, we test whether, controlling for economic
determinants of compensation and determinants of rounding, the level of compensation is higher when compensation is round.
To do so, we estimate the following equation:
Compensation
t¼aþb 1PercentRound tþb 2Dual tþb 3Insiders tþb 4AppointedByCEO tþb 5SDROA tþb 6SDReturns t
þb 7Digits tþb 8NumComponents tþd 1 8 CompensationControlsþYearFEþIndustryFEþe t
ð2Þ
In our primary analysis, we estimate Equation (2) with the dependent variable measured as performance-based
compensation (PerformanceCompensation), which is the sum of non-equity incentive plan compensation, the value of option
grants, and the value of stock grants. We also report results using the dependent variableDiscretionaryCompensation,de ned
as the sum of salary and discretionary bonus, and asTotalCompensation,de ned as the sum of all compensation components.
We again expect our results to be weaker for discretionary compensation than performance-based compensation. All
compensation measures are in logs.PercentRoundrefers toPercentRoundPerformance, PercentRoundDiscretionary,and
PercentRoundAllmeasured as discussed above. If round compensation is consistent with heaping, and the corresponding lack
of board effort in setting compensation allows self-interested managers to gain excess compensation, thenb
1will be greater
than 0 for performance-based compensation.
We include controls for the probability of rounding, as in Equation (1),Dual, Insiders, AppointedByCEO, Digits,
NumComponents, SDROA,andSDReturns,which are likely also correlated with the level of compensation. In addition, we
include common economic predictors of compensation levels (Core et al. 2008).ROA
t(ROA t 1 ) is return on assets measured as
earnings before interest and taxes scaled by average total assets measured in yeart(t 1).RET
t(RET t 1 ) is the 12-month buy-
and-hold market return for scal yeart(t 1).LogSales
t 1 is the log of sales in yeart 1.BTM t 1 is the book value of assets
divided by market value of assets in yeart 1.LogTenureis the log of CEO tenure.PercentOwnedis the percent of total shares
owned by the executive, as reported in ExecuComp. We again include year and industry xed effects.
Table 5 presents the results of estimating Equation (2). We nd that the coef cient ofPercentRoundPerformance,reported
in Column 1, is positive and signi cant ( p,0.01, one-tailed), which is consistent with our hypothesis that, controlling for rm
performance and other economic determinants of compensation, round compensation is higher than non-round compensation.
The coef cient of 0.084 implies that performance-based compensation is 8.4 percent higher when all its components are
rounded compared to when no components are rounded.
17 In Column 2, the association between discretionary compensation
rounding and the level of discretionary compensation is not statistically different from zero. However, re ecting the
performance-based compensation results, the coef cient ofPercentRoundAllis signi cantly positive in Column 3. These
associations are consistent with heaping leading to round compensation: when boards are less informed or expend less
cognitive effort in setting compensation, both round compensation and excess compensation result.
Round Compensation and Firm Performance
The above analyses are consistent with heaping occurring when boards determine executive compensation. We observe
abnormally high levels of round compensation values, and round compensation is consistent with a lack of information
acquisition and processing regarding compensation. In our nal analysis, we test whether round compensation is associated
with future rm performance, controlling for common predictors of performance and determinants of rounding. We estimate
the following equation using OLS regression:
ROA
tþ1 ¼aþb 1PercentRound tþb 2Dual tþb 3Insiders tþb 4AppointedByCEO tþb 5SDROA tþb 6Compensation t
þb 7ROA tþb 8LogSales tþ1 þYearFEþIndustryFEþe t ð3Þ
We proxy for future rm performance using ROA from the period after which compensation is granted (ROA
tþ1 ),
measured as earnings before interest and taxes scaled by average total assets measured in yeartþ1. We measure rm
performance using ROA following prior literature that demonstrates an association between current year compensation and
17 Since the independent variable in Equation (2) is natural logarithm of performance-based compensation, the economic magnitude of moving
PercentRoundPerformancefrom 0 to 1 is calculated as: 8.8 percent¼exp(0.084 ) 1.
194Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 TABLE 5
Associations of Round Compensation with Compensation Levels
Compensation
t¼aþb 1PercentRound tþb 2Dual tþb 3Insiders tþb 4AppointedByCEO tþb 5SDROA tþb 6SDReturns t
þb 7Digits tþb 8NumComponents tþd 1 8 CompensationControlsþYearFEþIndustryFEþe t
ð2Þ
Predicted
AssociationCompensationRefers to:
Performance
CompensationDiscretionary
CompensationTotal
Compensation
PercentRoundRefers to:
PercentRound
PerformancePercentRound
Discretionary PercentRoundAll
PercentRoundþ0.084*** 0.018 0.177***
(3.24 ) ( 1.04 ) (6.83)
Dualþ0.031** 0.021 0.006
(1.73) (1.06 ) (0.31)
Insidersþ 0.139* 0.129 0.178**
( 1.56 ) ( 1.19) ( 1.99)
AppointedByCEOþ0.014 0.064 0.040
(0.28) ( 1.14 ) (0.79)
SDROAþ 0.083 0.174 0.007
( 0.36 ) ( 0.67 ) (0.03)
SDReturnsþ0.040 0.040 0.016
(1.27 ) ( 1.22) (0.58)
Digitsþ1.836*** 1.408*** 1.415***
(53.56 ) (11.47 ) (11.93)
NumComponents? 0.107*** 0.088** 0.161***
(10.38) (1.89) (16.68)
ROA
t þ0.042 0.250** 0.099
(0.31) ( 1.77 ) ( 0.73)
ROA
t 1 þ 0.276** 0.467*** 0.067
( 2.16 ) (3.23) ( 0.50)
RET
t þ0.039*** 0.002 0.057***
(2.40) ( 0.12) (2.94 )
RET
t 1 þ0.007 0.039*** 0.016
(0.52) (3.23) (1.14 )
LogSales
t 1 þ0.107*** 0.067*** 0.131***
(11.00) ( 2.64 ) (5.35)
BTM
t 1 0.304*** 0.197*** 0.267***
( 6.21) (4.01) ( 4.33)
LogTenureþ0.002 0.009 0.025
(0.10) ( 0.34 ) (1.13)
PercentOwned? 0.002 0.014*** 0.008***
( 0.86 ) ( 4.09) ( 3.54 )
Constant 5.660*** 1.929*** 3.215***
( 28.45) ( 3.48) ( 4.71)
Observations 9,785 10,277 10,299
Adj. R
2 0.887 0.680 0.767
***, **, * Indicate statistical signi cance at the 1 percent, 5 percent, and 10 percent levels (one-tailed), respectively.
(continued on next page)
Heaping of Executive Compensation195
Journal of Management Accounting Research
Volume 32, Number 1, 2020 future year ROA (e.g.,Core, Holthausen and Larcker 1999;Hayes and Schaefer 2000). If round compensation re ects
signi cant rent extraction by CEOs, then we should nd a negative association between the frequency of rounding and future
rm performance (b
1,0). If, instead, heaping results from ef cient allocation of board attention, we will not detect an
association between round compensation and future rm performance (b
1not different from 0).
We include several controls for determinants of future ROA. We measureROA
tandSDROAas previously discussed. We
measureLogSales
tþ1 as the log of sales in yeartþ1. We also include the level of compensation as a control to address concerns
about mechanical associations about the frequency of rounding and levels of compensation. We include industry and year xed
effects in all speci cations.
We present the results of estimating Equation (3) in Table 6. We nd that round performance-based compensation is
signi cantly positively associated with future rm performance. This result is consistent with ef cient allocation of board
attention. This positive association could result if current-period performance-based compensation re ects the board’s private
information about performance in periodtþ1, as proposed inHayes and Schaefer (2000). Further, this positive association
might arise if current period performance-based compensation is viewed as a‘‘gift exchange’’between the board and the
executive proposed byAkerlof (1982).
We note the signi cantly negative association between round discretionary compensation and future rm performance. As we
report in Column 2 of Table 5, we do not nd that discretionary compensation is higher when CEOs receive round discretionary
compensation components. Thus, we interpret this negative association between round discretionary compensation components
and future rm performance as indicating the presence of agency con icts, rather the direct cost of excess compensation when
compensation is round. When we consider total compensation in the third column, we nd no evidence that the percentage of
round compensation components is associated with future performance. We therefore caution that this result does not resolve
whether round compensation re ects ef cient or inef cient allocation of effort in setting compensation.
VI. ADDITIONAL TESTS
Changes in Non-CEO Compensation
One concern with our cross-sectional analysis of the change in round compensation frequency following passage of
compensation-related regulations is that boards may respond to regulation by expending effort to change only the most visible
compensation—that of the CEO. If so, we may observe different patterns in the response to compensation regulation for CEOs
than for other executives. To address this concern, we perform our analyses of changes in rounding around Say on Pay again,
but we construct the samples to include the CEO and the Chief Financial Of cer (CFO). We test whether the effect of the
regulation change was stronger for the most visible executive, the CEO, than for other executives. Again,Jagolinzer (2009) TABLE 5 (continued)
Compensationrefers toPerformanceCompensation(Column 1),DiscretionaryCompensation(Column 2), orTotalCompensation(Column 3).
PerformanceCompensationis the sum of compensation from the four performance-based compensation components, as reported in ExecuComp, in
thousands of dollars.DiscretionaryCompensationis the sum of salary and discretionary bonus compensation from ExecuComp, in thousands of dollars.
TotalCompensationis compensation as reported in ExecuComp, in thousands of dollars (TDC1). Following prior literature, we use the natural logs of
compensation variables in our empirical analyses.PercentRoundrefers toPercentRoundPerformance(Column 1),PercentRoundDiscretionary(Column
2), orPercentRoundAll(Column 3), as de ned in the notes to Table 1. The sample size for this table is 10,299 CEO-years from 2007 to 2014. We estimate
Equations (2) and (3) using OLS regression. Standard errors are clustered by rm. This analysis includes industry and year xed effects.
Variable De nitions:
Dual¼an indicator variable set to 1 if the CEO is also the chairman of the board and 0 otherwise;
Insiders¼the percentage of board members who are insiders;
AppointedByCEO¼the percentage of outside board members appointed by the CEO;
ROA
t(ROA t 1)¼return on assets measured as EBIT/average total assets measured in yeart(t 1);
SDROA¼the three-year standard deviation ofROA
t;
SDReturns¼the standard deviation of the prior 36 months of returns, ending in the year for which compensation is granted;
Digits¼the number of digits in values ofPerformanceCompensation, DiscretionaryCompensation,orTotalCompensation,respectively;
NumComponents¼the number of performance-based compensation components (Column 1), the number of discretionary compensation components
(Column 2), or the number of compensation components in a compensation package (Column 3);
RET
t(RET t 1)¼the 12-month buy-and-hold market return for scal yeart(t 1);
LogSales
t 1 ¼the log of sales in yeart 1;
BTM
t 1 ¼the book value of assets divided by market value of assets in yeart 1;
LogTenure¼the log of CEO tenure; and
PercentOwned¼the percentage of total shares owned by the executive, as reported in ExecuComp.
196Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 notes that such analyses control for rm-year characteristics by design because we match CEOs and CFOs within the same rm
and for the same year.
We present the results of these analyses in Table 7. This analysis includes only rms where the CEO and CFO of a rm
were granted each type of compensation in both 2009 and 2011, which allows us to make within rm-year comparisons.
18 We
exclude rst-year CEOs and rst-year CFOs from our sample. Table 7 provides evidence that CFOs are statistically
signi cantly less likely to receive rounded compensation grants both before and after Say on Pay than are CEOs. Table 7 also
reveals that rounding of CFO compensation generally decreased around Say on Pay for performance-based compensation.
TABLE 6
Associations of Round Compensation with Future Firm Performance
ROA
tþ1 ¼aþb 1PercentRound tþb 2Dual tþb 3Insiders tþb 4AppointedByCEO tþb 5SDROA tþb 6Compensation t
þb 7ROA tþb 8LogSales tþ1 þYearFEþIndustryFEþe t ð3Þ
Predicted
AssociationPercentRoundRefers to:
PercentRound
PerformancePercentRound
Discretionary PercentRoundAll
PercentRound? 0.003** 0.002** 0.000
(1.65) ( 1.69) (0.12)
Dual 0.001 0.001 0.001
(1.00) (0.64 ) (0.68)
Insiders 0.014*** 0.013*** 0.012**
(2.52) (2.42) (2.20)
AppointedByCEO 0.002 0.001 0.002
( 1.01) ( 0.62) ( 0.66 )
SDROA? 0.018 0.014 0.014
(0.39) (0.35) (0.35)
Compensation? 0.000 0.000 0.001*
(0.19) ( 0.55) ( 1.36 )
ROA
t þ0.824*** 0.822*** 0.823***
(38.68) (41.08) (41.06 )
LogSales
tþ1 þ0.003*** 0.003*** 0.004***
(5.06 ) (6.77 ) (6.51)
Constant 0.020*** 0.016*** 0.013**
( 2.76 ) ( 2.47 ) ( 1.98)
Observations 9,785 10,279 10,299
Adj. R
2 0.634 0.638 0.638
***, **, * Indicate statistical signi cance at the 1 percent, 5 percent, and 10 percent levels (one-tailed), respectively.
PercentRoundrefers toPercentRoundPerformance(Column 1),PercentRoundDiscretionary(Column 2), orPercentRoundAll(Column 3), as de ned in
the notes to Table 1.Compensationrefers toPerformanceCompensation(Column 1),DiscretionaryCompensation(Column 2), orTotalCompensation
(Column 3), as previously de ned. Following prior literature, we use the natural logs of compensation variables in our empirical analyses. The sample size
for this table is 10,299 CEO-years from 2007 to 2014. We estimate Equation (4 ) using OLS regression. Standard errors are clustered by rm. This analysis
includes industry and year xed effects.
Variable De nitions:
ROA
tþ1 (ROA t)¼return on assets measured as EBIT/average total assets measured in yeartþ1(t);
Dual¼an indicator variable set to 1 if the CEO is also the chairman of the board and 0 otherwise;
Insiders¼the percentage of board members who are insiders;
AppointedByCEO¼the percentage of outside board members appointed by the CEO;
SDROA¼the three-year standard deviation ofROA
t; and
LogSales
tþ1 ¼the log of sales in yeartþ1.
18 We use only one year pre- and post-Say on Pay for this analysis because when we restrict our sample to include each type of compensation for each
executive in all four years from 2009 through 2012 our sample is extremely limited.
Heaping of Executive Compensation197
Journal of Management Accounting Research
Volume 32, Number 1, 2020 Taken together, the results in Tables 3, 4, and 7 offer strong support for H2, that regulations governing compensation
reduce the prevalence of round compensation.
Sensitivity Analyses
Our main results are robust to several sensitivity tests. In our analyses, we do not differentiate between exactly $1 million
salary observations and other round salary observations. Many of these salary observations are likely round not because of
heaping but because the Internal Revenue Code limits the deductibility of non-incentive-based compensation to $1 million
(Rose and Wolfram 2000,2002). We obtain qualitatively and quantitatively consistent results for all hypotheses when we treat
these observations as non-round.
Finally, we assume that boards choose numbers of options or shares to grant for compensation, rather than choosing a
value of grants and calculating the number of options or shares necessary to provide a desired level of compensation. When we
assume that boards grant dollar values of option or equity compensation, we nd that the dollar value of options is evenly
divisible by $10,000 in 3.5 percent of cases and that the dollar value of equity grants is evenly divisible by $10,000 in 4.5
percent of cases. This is consistent with some boards roundingdollar valuesof option and equity grants, instead ofnumbersof
option and equity grant shares. If some boards grant round dollar values of option or equity grants, the corresponding number
of options or shares would also be round only if the fair value of the option or share is also round. This implies that results of
TABLE 7
Test of the Association between Round Compensation Grants and Passage of Say on Pay for CEOs and CFOs
Panel A: Percent of Round Performance-Based Components
CEO CFOCompare CEO
to CFO t-statistics
Pre-Say on pay % rounded 22.7% 14.8% 6.606***
Observations 1,309 1,309
Post-Say on pay % rounded 17.2% 9.6% 7.534***
Observations 1,309 1,309
Pre- versus Post- t-statistics 4.510*** 5.230***
Panel B: Percent of Round Discretionary Components
CEO CFOCompare CEO
to CFO t-statistics
Pre-Say on pay % rounded 34.5% 24.6% 6.353***
Observations 1,513 1,513
Post-Say on pay % rounded 34.2% 22.3% 7.731***
Observations 1,513 1,513
Pre- versus Post- t-statistics 0.200 1.631*
Panel C: Percent of Round Components
CEO CFOCompare CEO
to CFO t-statistics
Pre-Say on pay % rounded 27.4% 19.2% 8.120***
Observations 1,518 1,518
Post-Say on pay % rounded 23.8% 15.0% 9.590***
Observations 1,518 1,518
Pre- versus Post- t-statistics 3.400*** 4.693***
*** Indicate statistical signi cance at the 1 percent level (one-tailed).
The frequencies reported refer toPercentRoundPerformance(Panel A),PercentRoundDiscretionary(Panel B), orPercentRoundAll(Panel C ), as de ned
in the notes to Table 1. The pre-SOP period is 2009, and the post-SOP period is 2011. The initial sample size for this table is 7,156 CEO- or CFO-years
from 2009 and 2011; we exclude rst-year CEOs and rst-year CFOs. We only include matched pairs of CEOs and CFOs, both of whom receive non-zero
compensation in the pre- and post-periods. Finally, we require rms to be present in both 2009 and 2011. The resulting sample sizes for each analysis are
included in panels above.
198Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 our tests of the frequency of rounding option and equity grants (Figure 1, Panels D, E, and F, and Table 2) are likely
understated. We nd that our results in Tables 3 through 7 are qualitatively similar when we use the dollar value of options or
shares granted instead of the number of options or shares granted.
VII. CONCLUDING REMARKS
We investigate whether round CEO compensation results from heaping, which is the tendency of imperfectly informed
individuals to report round estimates of discrete, quantitative data (Turner 1958). We document that round compensation is
quite common. Since boards may possess imperfect information due to a lack of investment in information acquisition and
processing, we investigate whether round compensation is associated with various proxies for board inattention to
compensation, including proxies for weak board monitoring, shareholder pressure, and excess compensation. Consistent with
our expectations, we nd evidence that rounding is positively associated with proxies for weak board monitoring. We also nd
that the incidence of round compensation decreases after passage of compensation-related regulation (i.e., Say on Pay and
SFAS 123R), which is consistent with these regulations providing incentives for boards to exert greater effort to reduce
uncertainty when determining compensation. Additional results provide evidence that round compensation tends to be higher
than non-round compensation. This association suggests that round compensation could signal to shareholders that boards who
grant round compensation are less attentive to compensation than other boards. As expected, our results are driven by
performance-based compensation rather than discretionary compensation. We also investigate the association of heaping of
compensation with future rm performance, but do not nd consistent evidence. While we interpret our ndings from the
perspective of heaping as de ned in the accounting literature, we acknowledge that other elds de ne heaping as a broader
phenomenon that might also explain our results.
Our study is the rst to investigate whether heaping explains the distribution of executive compensation, but several
questions remain for future research. For example, we measure stakeholder pressure using changes in regulation. Other sources
of stakeholder pressure may also affect how boards set compensation and the effects of behavior on compensation. Prior
literature documents that publicity (e.g.,Johnson et al. 1997;Core et al. 2008), shareholder proposals (e.g.,Thomas and Martin
1999,Ertimur et al. 2011), and proxy advisor recommendations (e.g.,Ertimur et al. 2018) affect how boards set compensation.
Future research could further investigate whether these sources of pressure have similar effects on compensation. Finally, we
measure single dimensions of rm performance and monitoring. Round compensation may affect board decision-making in
other ways, such as the probability of CEO turnover.
Overall, our results suggest that corporate governance choices can have subtle effects on decision-making by boards. It
remains an open question whether round compensation results from a more general practice in rms to provide round numbers,
including earnings and earnings per share. An investigation of whether round compensation is associated with rounding of
other reported values might provide additional evidence of heaping.
REFERENCES
A’Hearn, B., J. Baten, and D. Crayen. 2009. Quantifying quantitative literacy: Age heaping and the history of human capital.The Journal
of Economic History69 (3): 783–808.https://doi.org/10.1017/S0022050709001120
Abernethy, M., Y. Kuang, and B. Qin. 2015. The in uence of CEO power on compensation contract design.The Accounting Review90
(4 ): 1265–1306.https://doi.org/10.2308/accr-50971
Akerlof, G. A. 1982. Labor contracts as partial gift exchange.The Quarterly Journal of Economics97 (4 ): 543–569.https://doi.org/10.
2307/1885099
Amiram, D., Z. Bozanic, and E. Rouen. 2015. Financial statement errors: Evidence from the distributional properties of nancial
statement numbers.Review of Accounting Studies20 (4 ): 1540–1593.https://doi.org/10.1007/s11142-015-9333-z
Amiram, D., A. Kalay, and N. B. Ozel. 2013.The bond discount puzzle. Working paper, Columbia University.
Athanasakou, V., and A. Simpson. 2016. Investor attention to rounding as a salient forecast feature.International Journal of Forecasting
32 (4 ): 1212–1233.https://doi.org/10.1016/j.ijforecast.2016.02.011
Balsam, S., J. Boone, H. Liu, and J. Yin. 2016. The impact of say-on-pay on executive compensation.Journal of Accounting and Public
Policy35 (2): 162–191.https://doi.org/10.1016/j.jaccpubpol.2015.11.004
Bamber, L. S., K. W. Hui, and P. E. Yeung. 2010. Managers’ EPS forecasts: Nickeling and diming the market?The Accounting Review85
(1): 63–95.https://doi.org/10.2308/accr.2010.85.1.63
Barreca, A. J., M. Guldi, J. M. Lindo, and G. R. Waddell. 2011. Saving babies? Revisiting the effect of very low birth weight
classi cation.The Quarterly Journal of Economics126 (4 ): 2117–2123.https://doi.org/10.1093/qje/qjr042
Bebchuk, L. A., and J. M. Fried. 2003. Executive compensation as an agency problem.The Journal of Economic Perspectives17 (3): 71–
92.https://doi.org/10.1257/089533003769204362 Heaping of Executive Compensation199
Journal of Management Accounting Research
Volume 32, Number 1, 2020 Beer, T. W. 2009. Terminal digit preference: Beware of Benford’s law.Journal of Clinical Pathology62 (2): 192.https://doi.org/10.1136/
jcp.2008.061721
Bertrand, M., and S. Mullainathan. 2001. Are CEOs rewarded for luck? The ones without principals are.The Quarterly Journal of
Economics116 (3): 901–932.https://doi.org/10.1162/00335530152466269
Bhattacharya, U., C. W. Holden, and S. Jacobsen. 2012. Penny wise, dollar foolish: Buy-sell imbalances on and around round numbers.
Management Science58 (2): 413–431.https://doi.org/10.1287/mnsc.1110.1364
Burgstahler, D., and E. Chuk. 2015. Do scaling and selection explain earnings discontinuities?Journal of Accounting and Economics60
(1): 168–186.https://doi.org/10.1016/j.jacceco.2014.08.002
Burgstahler, D., and E. Chuk. 2017. What have we learned about earnings management? Integrating discontinuity evidence.
Contemporary Accounting Research34 (2): 726–749.https://doi.org/10.1111/1911-3846.12301
Burgstahler, D., and I. Dichev. 1997. Earnings management to avoid earnings decreases and losses.Journal of Accounting and Economics
24 (1): 99–126.https://doi.org/10.1016/S0165-4101(97 )00017-7
Carslaw, C. A. P. N. 1988. Anomalies in income numbers: Evidence of goal oriented behavior.The Accounting Review63 (2): 321–327.
Cheong, F. S., and J. Thomas. 2011. Why do EPS forecast error and dispersion not vary with scale? Implications for analyst and
managerial behavior.Journal of Accounting Research49 (2): 359–401.https://doi.org/10.1111/j.1475-679X.2010.00387.x
Christie, W. G., and P. M. Schultz. 1994. Why do NASDAQ market makers avoid odd-eighth quotes?The Journal of Finance49 (5):
1813–1840.https://doi.org/10.1111/j.1540-6261.1994.tb04782.x
Christie, W. G., J. H. Harris, and P. H. Schultz. 1994. Why did NASDAQ market makers stop avoiding odd-eighth quotes?The Journal of
Finance49 (5): 1841–1860.https://doi.org/10.1111/j.1540-6261.1994.tb04783.x
Coles, J. L., N. D. Daniel, and L. Naveen. 2014. Co-opted boards.Review of Financial Studies27 (6 ): 1751–1796.https://doi.org/10.
1093/rfs/hhu011
Core, J. E., W. Guay, and D. F. Larcker. 2008. The power of the pen and executive compensation.Journal of Financial Economics88 (1):
1–25.https://doi.org/10.1016/j.j neco.2007.05.001
Core, J. E., R. W. Holthausen, and D. F. Larcker. 1999. Corporate governance, chief executive of cer compensation, and rm
performance.Journal of Financial Economics51 (3): 371–406.https://doi.org/10.1016/S0304-405X(98)00058-0
Das, S., and H. Zhang. 2003. Rounding-up in reported EPS, behavioral thresholds, and earnings management.Journal of Accounting and
Economics35 (1): 31–50.https://doi.org/10.1016/S0165-4101(02)00096-4
Dechow, P. M., and H. You. 2012. Analysts’ motives for rounding EPS forecasts.The Accounting Review87 (6 ): 1939–1966.https://doi.
org/10.2308/accr-50226
Degeorge, F., J. Patel, and R. J. Zeckhauser. 1999. Earnings management to exceed thresholds.The Journal of Business72 (1): 1–33.
https://doi.org/10.1086/209601
Diaconis, P. 1977. The distribution of leading digits and uniform distribution mod 1.Annals of Probability5 (1): 72–81.https://doi.org/
10.1214/aop/1176995891
Ertimur, Y., F. Ferri, and V. Muslu. 2011. Shareholder activism and CEO pay.Review of Financial Studies24 (2): 535–592.https://doi.
org/10.1093/rfs/hhq113
Ertimur, Y., F. Ferri, and D. Oesch. 2013. Shareholder votes and proxy advisors—Evidence from Say on Pay.Journal of Accounting
Research51 (5): 951–996.https://doi.org/10.1111/1475-679X.12024
Ertimur, Y., F. Ferri, and D. Oesch. 2018. Understanding uncontested director elections.Management Science64 (7 ): 3400–3420.https://
doi.org/10.1287/mnsc.2017.2760
Fedyk, T. 2007.Discontinuity in earnings reports and managerial incentives. Working paper, University of California, Berkeley.
Financial Accounting Standards Board (FASB). 1995.Statement of Financial Accounting Standards No. 123: Accounting for Stock-Based
Compensation. Norwalk, CT: FASB.
Financial Accounting Standards Board (FASB). 2004.Statement of Financial Accounting Standards No. 123R: Share-Based Payment.
Norwalk, CT: FASB.
Frydman, C., and D. Jenter. 2010. CEO compensation.Annual Review of Financial Economics2 (1): 75–102.https://doi.org/10.1146/
annurev- nancial-120209-133958
Grossman, S. J., M. H. Miller, K. R. Cone, D. R. Fischel, and D. J. Ross. 1997. Clustering and competition in asset markets.The Journal
of Law & Economics40 (1): 23–60.https://doi.org/10.1086/467365
Hamm, S. J. W., M. J. Jung, and C. Wang. 2015. Making sense of one dollar CEO salaries.Contemporary Accounting Research32 (3):
941–972.https://doi.org/10.1111/1911-3846.12136
Hayes, R. M., and S. Schaefer. 2000. Implicit contracts and the explanatory power of top executive compensation for future performance.
The RAND Journal of Economics31 (2): 273–293.https://doi.org/10.2307/2601041
Hayn, C. 1995. The information content of losses.Journal of Accounting and Economics20 (2): 125–153.https://doi.org/10.1016/0165-
4101(95)00397-2
Healy, P. M. 1985. The effect of bonus schemes on accounting decisions.Journal of Accounting and Economics7 (1-3): 85–107.https://
doi.org/10.1016/0165-4101(85)90029-1
Herrmann, D., and W. B. Thomas. 2005. Rounding of analyst forecasts.The Accounting Review80 (3): 805–823.https://doi.org/10.2308/
accr.2005.80.3.805 200Jorgensen, Patrick, and Soderstrom
Journal of Management Accounting Research
Volume 32, Number 1, 2020 Huttenlocher, J., L. V. Hedges, and N. M. Bradburn. 1990. Reports of elapsed time: Bounding and rounding processes in estimation.
Journal of Experimental Psychology. Learning, Memory, and Cognition16 (2): 196–213.https://doi.org/10.1037/0278-7393.16.2.
196
Jagolinzer, A. D. 2009. SEC Rule 10b5-1 and insiders’ strategic trade.Management Science55 (2): 224–239.https://doi.org/10.1287/
mnsc.1080.0928
Johnson, E., N. B. Johnson, and D. M. Shanthikumar. 2011.Round numbers and security returns. Working paper, University of
California, Irvine.
Johnson, M. F., S. Porter, and M. B. Shackell. 1997.Stakeholder pressure and the structure of executive compensation. Working paper,
University of Notre Dame.
Jorgensen, B. N., Y. G. Lee, and S. K. Rock. 2014. The shapes of scaled earnings histograms are not due to scaling and sample selection:
Evidence from distributions of reported earnings per share.Contemporary Accounting Research31 (2): 498–521.https://doi.org/
10.1111/1911-3846.12020
Lo, K., S. Yang, and J. L. Zhang. 2014.Say on Pay votes and compensation practices. Working paper, The University of British
Columbia.
Loureiro, G. R., A. K. Makhija, and D. Zhang. 2015.The ruse of a one-dollar CEO salary?Working paper, The Ohio State University.
Malenko, N., and J. A. Grundfest. 2014.Quadrophobia: Strategic rounding of EPS data. Working paper, Stanford University.
Melumad, N. D. 1989. Asymmetric information and the termination of contracts in agencies.Contemporary Accounting Research5 (2):
733–753.https://doi.org/10.1111/j.1911-3846.1989.tb00736.x
Osborne, M. F. M. 1962. Periodic structure in the Brownian motion of stock prices.Operations Research10 (3): 345–379.https://doi.org/
10.1287/opre.10.3.345
Papke, L. E., and J. M. Wooldridge. 1996. Econometric methods for fractional response variables with an application to 401(K) plan
participation rates.Journal of Applied Econometrics11 (6 ): 619–632.https://doi.org/10.1002/(SICI)1099-
1255(199611)11:6,619:AID-JAE418.3.0.CO;2-1
Perry, T., and M. Zenner. 2001. Pay for performance? Government regulation and the structure of compensation contracts.Journal of
Financial Economics62 (3): 453–488.https://doi.org/10.1016/S0304-405X(01)00083-6
Roberts, J. M., and D. D. Brewer. 2001. Measures and tests of heaping in discrete quantitative distributions.Journal of Applied Statistics
28 (7 ): 887–896.https://doi.org/10.1080/02664760120074960
Rose, N. L., and C. D. Wolfram. 2000. Has the‘‘million-dollar cap’’affected CEO pay?The American Economic Review90 (2): 197–202.
https://doi.org/10.1257/aer.90.2.197
Rose, N. L., and C. D. Wolfram. 2002. Regulating executive pay: Using the tax code to in uence chief executive of cer compensation.
Journal of Labor Economics20 (S2): S138–S175.https://doi.org/10.1086/338677
Rowland, M. L. 1990. Self-reported weight and height.The American Journal of Clinical Nutrition52 (6 ): 1125–1133.https://doi.org/10.
1093/ajcn/52.6.1125
Shah, A. K., and D. M. Oppenheimer. 2008. Heuristics made easy: An effort-reduction framework.Psychological Bulletin134 (2): 207–
222.https://doi.org/10.1037/0033-2909.134.2.207
Sheehan, K. 2012. Say on Pay and the outrage constraint. InResearch Handbook on Executive Pay,edited by R. S. Thomas and J. G. Hill,
255–283. Cheltenham, U.K., and Northampton, MA: Elgar.
Shryock, H. S., J. S. Siegel, and E. A. Larmon. 1980.The Methods and Materials of Demography. Washington, DC: U.S. Department of
Commerce, Bureau of the Census.
Simon, H. A. 1990. Invariants of human behavior.Annual Review of Psychology41 (1): 1–20.https://doi.org/10.1146/annurev.ps.41.
020190.000245
Thomas, J. K. 1989. Unusual patterns in reported earnings.The Accounting Review64 (4 ): 773–787.
Thomas, R. S., and K. J. Martin. 1999. The effect of shareholder proposals on executive compensation.University of Cincinnati Law
Review67 (4 ): 1021–1081.https://doi.org/10.2139/ssrn.160188
Turner, S. H. 1958. Patterns of heaping in the reporting of numerical data. InProceedings of the Social Statistics Section,248–251.
Washington, DC: American Statistical Association.
U.S. Congress. 2010.Dodd-Frank Wall Street Reform and Consumer Protection Act. Pub. L. No. 111-203, § 521, 124 Stat. 1376, 1899
(2010). Washington, DC: Government Printing Of ce
Wen, S. W., M. S. Kramer, J. Hoey, J. A. Hanley, and R. H. Usher. 1993. Terminal digit preference, random error, and bias in routine
clinical measurement of blood pressure.Journal of Clinical Epidemiology46 (10): 1187–1193.https://doi.org/10.1016/0895-
4356(93)90118-K
Zhou, L. 2010. Nickels not pennies: Granularity in analysts’ EPS forecasts and forecast revisions.Journal of Accounting, Auditing &
Finance25 (2): 201–233.https://doi.org/10.1177/0148558X1002500203 Heaping of Executive Compensation201
Journal of Management Accounting Research
Volume 32, Number 1, 2020 Copyright
ofJournal ofManagement AccountingResearchisthe property ofAmerican
Accounting
Association anditscontent maynotbecopied oremailed tomultiple sitesor
posted
toalistserv without thecopyright holder'sexpresswrittenpermission. However,users
may
print, download, oremail articles forindividual use.