THE ACCOUNTING REVIEW
Vol. 79, No. 2
pp. 409–436
Determinants and Effects of
Subjectivity in Incentives
Michael Gibbs
University of Chicago
Kenneth A. Merchant
Wim A. Van der Stede
University of Southern California
Mark E. Vargus
The University of Texas at Dallas
ABSTRACT: This study examines two questions: When do firms make greater use of
subjectivity in awarding bonuses? What are the effects of subjectivity on employee pay
satisfaction and firm performance? We examine these questions using data from a
sample of 526 department managers in 250 car dealerships. First, the findings suggest
that subjective bonuses are used to complement perceived weaknesses in quantitative
performance measures and to provide employees insurance against downside risk in
their pay. Specifically, use of subjective bonuses is positively related to: (1) the extent
of long-term investments in intangibles; (2) the extent of organizational interdependencies; (3) the extent to which the achievability of the formula bonus target is both difficult
and leads to significant consequences if not met; and (4) the presence of an operating
loss. Second, we find that the effects of subjective bonuses on pay satisfaction, productivity, and profitability are larger the greater the manager’s tenure, consistent with
the idea that subjectivity improves incentive contracting when there is greater trust
between the subordinate and supervisor.
Keywords: incentive compensation; discretionary bonus; subjective performance evaluation; performance; pay satisfaction.
Data Availability: Confidentiality agreements prevent the authors from distributing the
data.
This paper has benefited from comments by two anonymous reviewers, Shannon Anderson, Rajiv Banker, Tony
Davila, Harry Evans, Joe Fisher, Nick Gonedes, Raffi Indjejikian, Bill Lanen, Eddie Lazear, Michal Matejka, Kevin
J. Murphy, D. J. Nanda, Robin Poston, Michael Raith, Dhinu Srinivasan, Mark Young, and workshop participants
at Copenhagen Business School, Erasmus University Rotterdam, Harvard Business School, Hong Kong Polytechnic
University, HKUST (2002 Accounting Summer Symposium), London School of Economics, National Chengchi
University, Tilburg University, University of Gothenburg, University Pompeu Fabra, University of Southern California, The University of Texas at Austin, University of Wisconsin–Madison, University of Zurich (conference on
Management, Economics and Corporate Governance), Washington University in St. Louis, and the American Accounting Association 2002 Annual Meeting and 2003 Management Accounting Meeting. The authors also thank
Liu Zheng for research assistance.
Editor’s note: This paper was accepted by Terry Shevlin, Senior Editor.
409
Submitted January 2002
Accepted August 2003
410
Gibbs, Merchant, Van der Stede, and Vargus
I. INTRODUCTION
n important characteristic of most incentive contracts is the use of subjectivity in
evaluating and rewarding employees (Murphy and Cleveland 1995; Prendergast
1999). In bonus assignments, subjectivity can arise in several ways, which are often
used in combination: (1) all or part of a bonus is based on subjective judgments about
performance; (2) the weights on some or all quantitative measures are determined subjectively,1 or (3) a subjective performance threshold or ‘‘override’’ is used, in which case a
subjective determination as to whether to pay a bonus is made based on measured performance and other factors. Subjectivity is also an important element of implicit incentives given,
for example, through promotions, job assignments, and threat of termination. While theoretical research has suggested various plausible reasons for the use of subjectivity in the
assignment of bonuses, empirical testing of these theories is rare (Bushman et al. 1996;
MacLeod and Parent 1999; Hayes and Schaefer 2000; Murphy and Oyer 2003). This study
helps to fill this gap, providing an examination of the use of subjectivity in incentive
contracts using a unique dataset of 526 department managers in 250 car dealerships that
includes variables related to subjectivity and perceptions of its use and effects.
We explore two questions: What are the determinants of the use of subjectivity in
awarding bonuses? What are the effects of subjectivity on pay satisfaction and performance?
The findings suggest that subjective bonuses (sometimes called discretionary bonuses) are
used to mitigate distortions or reduce risk of bonuses tied by formula to quantitative performance measures (formula bonuses). We find that the use of subjective bonuses is positively related to: (1) the extent of long-term investments in intangibles; (2) the extent of
organizational interdependencies; (3) the extent to which the achievability of the formula
bonus target is both difficult and leads to significant consequences if not met; and (4) the
presence of an operating loss.
We also find that the use of subjectivity increases the manager’s satisfaction with the
pay scheme and is positively related to productivity and net profit when the manager has
long tenure. We interpret these findings as suggesting that when trust (proxied by the
manager’s tenure at the dealership) is higher, subjective incentives can be effective.
Section II reviews existing theory and develops hypotheses. Section III details the
empirical design and measures. Section IV presents the results. Section V concludes and
offers directions for future research.
A
II. THEORY
Subjectivity can be useful in mitigating various problems faced in assigning rewards
through formulas based on quantitative performance measures. The use of subjectivity allows evaluators to exploit any additional relevant information that arises during the measurement period to the benefit of both the firm and the employee. The firm can benefit
through improved incentive alignment and the employee can benefit through reduced risk.
In effect, subjectivity allows for the recalibration of incentives during or after the period,
which can be especially important if there are costs to changing or renegotiating formal
bonus contracts (Baker et al. 1988). Our main focus is on how subjectivity can play these
roles. However, we should not expect the use of subjectivity to always be effective. Because
it involves discretion, subjectivity works well only if the supervisor makes fair, unbiased
judgments and if the subordinate accepts the judgments and does not try to influence the
1
Quantitative performance measures are often termed objective. However, the employee or evaluator can often
manipulate the measures, and the measures may not measure exactly what they are thought to measure, so this
terminology is misleading. We avoid this by using the terms quantitative measures and formula-based incentives.
The Accounting Review, April 2004
Determinants and Effects of Subjectivity in Incentives
411
supervisor inappropriately. We argue that the use of subjectivity will have positive effects
on pay satisfaction and performance only when there is adequate trust between the subordinate and the supervisor.
The hypotheses that we develop build on the recent literature on the use of subjectivity
in executive incentive contracts. For example, Bushman et al. (1996) consider the use of
‘‘individual performance evaluations’’ (instead of corporate-wide measures like stock performance) for CEO incentives. While individual performance evaluations need not be subjective, it is an important part of what they discuss. They argue that individual evaluation
is likely to be used to improve multitask incentives (reduce distortions), or to reduce noise
in the overall performance evaluation. Similarly, Hayes and Schaefer (2000) argue that
implicit contracts (subjectivity) will be used when some elements of managerial performance are not observable using quantitative performance measures, or when such measures
are noisier. Finally, Murphy and Oyer (2003) argue that executive incentive contracts may
make use of subjectivity to reduce the possibility of manipulation of quantitative performance measures (a form of distorted incentives), or to reduce noise in quantitative measures.
They also emphasize that subjectivity is more likely to be effective when the firm has less
incentive to renege on promised subjective bonuses, which is similar to our discussion of
the role of trust.
Determinants of Subjectivity in Incentives
Effective performance measures provide accurate, informative, and timely indications
of the individual’s contributions to firm value (or other organizational goals), at low risk to
the employee (Holmstrom 1979). When quantitative performance measures are effective,
formula incentives are likely to be used intensively. However, quantitative performance
measures often distort incentives (e.g., because they are incomplete or prone to manipulation) or impose undue risk on the employee (e.g., because they include uncontrollables).
Thus, mitigating distortions to improve incentive alignment, and filtering out uncontrollables
to reduce risk, allow for stronger incentives. Our hypotheses involve ways in which subjectivity can be used to further one or both of these goals.
Use of Subjectivity to Mitigate Formula Bonus Distortions
Formula bonus completeness. Performance measures, especially accounting numbers,
can distort incentives because they inadequately account for (or ignore entirely) some dimensions of the employee’s job. Most jobs involve multiple types of employee efforts and
decisions. Ideally, compensation contracts should use all possible information about employees’ effects on firm value, weighted properly, so that incentives are appropriately balanced across different dimensions of the job (Holmstrom and Milgrom 1991; Feltham and
Xie 1994; Baiman and Rajan 1995; Bommer et al. 1995; Kaplan and Norton 1996).
At managerial levels, where jobs are complex, compensation contracts are almost invariably incomplete. Commonly, then, managers direct their efforts only to measured tasks
and may ignore other important-but-unmeasured tasks (e.g., they focus on improving shortterm profits at the expense of long-term client relations). Subjectivity can be used to reward
managers for value-enhancing efforts that are not easily quantified, and thus are not explicitly measured in the formula contract. This can mitigate potential distortions from incentives based on quantitative measures. Bushman et al. (1996) and Murphy and Oyer
(2003) present evidence suggesting that executive incentive contracts make greater use of
subjectivity when accounting measures are less correlated with stock returns, and MacLeod
and Parent (1999) find that more complex jobs (that is, with more tasks) are more likely
to be evaluated subjectively. Thus, we predict that:
The Accounting Review, April 2004
412
Gibbs, Merchant, Van der Stede, and Vargus
H1: The use of subjectivity in the assignment of rewards will be negatively related to
formula bonus completeness.
Short-term focus. It is often argued that a drawback to accounting measures is that
they induce a short-term focus (e.g., Jacobs 1991). By their nature, accounting measures
are backward-looking, and thus do not accurately reflect the effects of employees’ efforts
or decisions on future firm performance. This problem is particularly acute in situations
where investments in intangible assets are important (Lev 2001). Managers whose performances are evaluated in terms of accounting income are discouraged from making investments in intangible assets due to their conservative accounting treatment. For example,
Bushman et al. (1996), Hayes and Schaefer (2000), and Murphy and Oyer (2003) predict
that executive pay will make greater use of subjectivity in firms with more growth opportunities or longer product cycles. Subjectivity can be used to mitigate an excessively shortterm focus in situations where accounting measure conservatism is more pronounced. Thus,
we propose that:
H2: The use of subjectivity in the assignment of rewards will be positively related to
the short-term focus of the quantitative measures.
H3: The use of subjectivity in the assignment of rewards will be positively related to
the extent of long-term investments in intangibles.
Susceptibility to manipulation. Quantitative performance measures are also often susceptible to manipulation by employees (Holmstrom and Milgrom 1991; Baker et al. 1994;
Ittner et al. 2003). Employees often are more knowledgeable (especially in real time) about
their environment than their supervisors (Baker 1992). This information can be used to
improve firm value through better effort allocation and decision making, but it can also be
used to manipulate performance measures to the employee’s advantage. Using subjectivity
to leave the basis for evaluations somewhat vague lessens employees’ incentives and abilities to manipulate the measures. Further, assuming that evaluators have sufficient skill to
detect some manipulations, subjectivity allows the firm to engage in ex post settling up,
thereby also diminishing the incentives of managers to engage in non-productive manipulations. This suggests that:
H4: The use of subjectivity in the assignment of rewards will be positively related to
the manipulability of the quantitative measures.
Use of Subjectivity to Reduce Formula Bonus Risk
In addition to mitigating formula bonus distortions, subjectivity can also be used to
reduce formula bonus risk imposed on employees by filtering out uncontrollables. Uncontrollables are events whose impact on firm value the employee cannot affect; they are often
referred to as noise. Controllables are events whose impact on firm value can be affected
by, or are sensitive to, the employee’s actions. They may even be random, as long as the
employee can react to them, mitigating negative effects and exploiting positive effects
(Merchant 1987). Contracting theory suggests that the relative importance (weight) of performance metrics should be a decreasing function of their noise and an increasing function
of their sensitivity to employee effort or decisions (Holmstrom 1979; Banker and Datar
1989). The subsequent hypotheses reflect ways in which subjectivity can be used to either
exclude uncontrollables to reduce employee risk or include controllables to improve incentives’ sensitivity to managerial effort.
The Accounting Review, April 2004
Determinants and Effects of Subjectivity in Incentives
413
Prior studies suggest that filtering out ‘‘windfalls’’ from incentive pay is rare (Merchant
1989). One explanation is that risk-averse employees are primarily concerned with being
provided insurance against downside risk. Another concern with using subjectivity to filter
out ‘‘good luck’’ (positive measurement error) to reduce pay when employees earn formula
bonuses that are ‘‘too high,’’ is that employees might perceive this as increasing the performance standard when performance is better than expected (Gibbons 1987), which may
undermine trust. Finally, management is more likely to face substantial influence costs from
employees when performance is poor and formula bonuses are low, or zero, than when
bonuses are high. Consequently, we expect that subjectivity is more likely to be used to
provide insurance for downside risk, filtering out effects of bad luck rather than good luck.
The hypotheses below reflect this asymmetry.
Uncontrollable factors. In all but the simplest environments, it is impossible to account
for all random events with quantitative performance measures. Instead, the firm may feel
that supervisors, using judgment, are better able to take such factors into consideration.
This is similar to Bushman et al.’s (1996) prediction that executive incentive contracts make
greater use of subjectivity if accounting measures or stock returns are noisy. Thus, we
predict that:
H5: The use of subjectivity in the assignment of rewards will be positively related to
the extent to which the quantitative measures reflect factors outside the managers’
control.
Organizational interdependencies. Interdependencies may lead to greater use of subjectivity to both reduce employee risk and to better align incentives. One response to joint
production or interdependency is to use broader performance measures, such as firm performance. However, such performance measures usually impose risk on the employee due
to the adverse effects from actions of other employees (Baiman and Rajan 1995; Bommer
et al. 1995). Subjectivity can be used to insure employees against these adverse effects. An
alternative is to use narrower performance measures that are less risky, but this is likely to
distort incentives to cooperate with colleagues. In this case, subjectivity can be used to
encourage cooperation. Thus, for both reasons we propose that:
H6: The use of subjectivity in the assignment of rewards will be positively related to
the extent of organizational interdependencies.
Environmental unpredictability. The distinction between controllables and uncontrollables is often ambiguous. In many situations, an external factor is only partly controllable by the employee, thereby exposing the employee to some risk. Consider an unexpected
price cut by a major competitor. This is likely to negatively impact most financial performance measures. If this was truly an unexpected event, some of the effects of the new price
competition are noise that should be removed from evaluations. However, removing all of
the effects would reduce employee incentives to respond to the new situation in ways that
would moderate its impact on firm profit. In such situations, the ideal evaluation should
filter out the uncontrollable effects, but include any effects on firm value that can be changed
by the employee’s effort or decisions. Doing the former reduces risk, while doing the latter
motivates employees to respond appropriately to events as they unfold, making use of any
information as it is learned. Assigning bonuses subjectively can help accomplish both objectives, and this will be useful in more unpredictable and competitive environments (Raith
2001). Hence, we predict that:
The Accounting Review, April 2004
414
Gibbs, Merchant, Van der Stede, and Vargus
H7: The use of subjectivity in the assignment of rewards will be positively related to
the level of environmental uncertainty.
H8: The use of subjectivity in the assignment of rewards will be positively related to
the level of competition.
Recalibration. Subjectivity can also provide flexibility to alter other aspects of the
incentive contract as the environment changes. Incentive contracts often have explicit or
implicit performance targets or standards, which are used to communicate expected levels
of performance to employees (Murphy 2000). However, as with all aspects of the incentive
contract, optimal standards or expected performance levels may change if the environment
changes (Merchant 1989; Merchant and Manzoni 1989).
If performance is lower than expected because of uncontrollable factors (e.g., because
of an unexpected recession), employees might believe that the target performance level will
not be achieved. In such cases, formula bonuses provide little incentive. Subjective bonuses
are an alternative for keeping employees motivated when the fixed performance targets have
become unreasonably difficult to achieve (Merchant and Manzoni 1989).
Therefore, recalibrating incentives is likely to be important when the expected level of
performance is set high (stretch targets) so that the likelihood of missing the standard is
greater. This is a problem, particularly, when there are significant nonfinancial consequences
of failing to achieve targets (e.g., smaller budgets, less autonomy, likelihood of promotion,
or threat of termination). In the extreme, the combination of difficult targets and significant
financial and nonfinancial consequences places the greatest risk on employees. In contrast,
if the target is difficult to achieve, yet the penalty is trivial, then the employee bears minimal
risk. Thus, we propose that the use of subjective bonuses will be related to the joint effects
of target difficulty and the consequences of not meeting the performance target, or that:
H9: The use of subjectivity in the assignment of rewards will be positively related to
the difficulty of meeting a performance target that has high consequences for
failure.
Building on the logic from H9, a special case when subjectivity is relevant is when an
entity is operating in a loss condition. When this is the case, the manager will typically
receive zero formula bonus (especially if the performance measure is based on profit, as it
often is). The loss might have been caused by poor managerial performance, or by ‘‘bad
luck’’ that was (at least partially) uncontrollable by the manager, or both. To the extent that
it was caused by uncontrollable events, incentives can be improved by filtering out these
effects from the manager’s incentive compensation. Once again, it is likely that the only
way to do so is to use subjectivity. Moreover, although firms could write formula plans in
terms of reducing losses, observations in our pilot study suggest that upper management is
reluctant to promise formula bonuses when the entity is generating losses, even when losses
are being reduced. Therefore, we propose that:
H10: The use of subjectivity in the assignment of rewards will be positively related to
the occurrence of a loss.
Effects of Subjectivity in Incentives
While subjectivity often provides benefits, it can also add another form of risk—that
of unfair performance evaluations (Fulk et al. 1985; Prendergast and Topel 1993). It can
The Accounting Review, April 2004
Determinants and Effects of Subjectivity in Incentives
415
also expose employees to hindsight bias if evaluators fail to adjust their evaluations to take
into account that employees may have had different information than is available to evaluators ex post (Hawkins and Hastie 1990). If subordinates do not trust their evaluators to
make informed and unbiased performance assessments, then the result could be employee
frustration, demotivation, and turnover. Moreover, when evaluations are subjective, employees may attempt to inappropriately influence supervisors for better evaluations (Milgrom 1988), or distort their efforts in order to win favor from their manager (Prendergast
1993). These problems are reduced if the employee and the evaluator develop a working
relationship with greater mutual trust (Pennings and Woiceshyn 1987; McAllister 1995;
Baker et al. 1994; Murphy and Oyer 2003).
We have argued that subjectivity can be used to improve formal incentive contracts by
reducing employee risk and improving incentives. We have also argued that if the evaluator
is not fair and unbiased, subjectivity can impose substantial risk on employees. Thus, if a
firm relies on subjectivity to a greater degree, then it is not clear a priori what the impact
on employee pay satisfaction and performance would be. For this reason we do not predict
whether greater use of subjectivity will increase or decrease pay satisfaction and
performance.
However, trust between the subordinate and supervisor can alleviate implementation
problems, leading to more effective use of subjectivity in incentive schemes. Folger and
Konovsky (1989) find a positive relationship between trust and the employee’s satisfaction
with their performance evaluation and Lawler (1971) argues that greater trust increases the
effectiveness of incentive plans. Therefore, higher trust combined with greater use of subjectivity should increase employee satisfaction with the compensation system, as well as
improve performance. Thus, we predict that:
H11: The use of subjectivity in assigning rewards will be more positively related to
pay satisfaction the greater the level of trust.
H12: The use of subjectivity in assigning rewards will be more positively related to
productivity the greater the level of trust.
H13: The use of subjectivity in assigning rewards will be more positively related to
net profit the greater the level of trust.
III. METHOD
Data Collection Procedure and Survey Development
Other recent studies of subjectivity in incentives generally had to rely on indirect measures of subjectivity, due to lack of data. For example, Bushman et al. (1996) relied on the
degree of ‘‘individual performance evaluation’’ in CEO incentive contracts, using proprietary data from Hewitt Associates. The data do not distinguish quantitative individual performance measures from subjective evaluations of individual performance. However, it appears likely that subjectivity is an important component of individual performance
evaluation in their sample, especially since it is difficult to imagine many quantifiable
measures of individual performance at the CEO level. Hayes and Schaefer (2000) infer the
use of subjectivity from the degree of variation in CEO compensation that is unexplained
by variation in current observable firm performance. The assumption is that this reflects
subjective performance assessments available to the compensation committee but not observable to researchers. Murphy and Oyer (2003) use a survey from Towers Perrin that
includes information on the use of discretion in incentive policies for executives. Thus,
their data describes discretionary incentive policies, rather than actual subjective bonuses.
The Accounting Review, April 2004
416
Gibbs, Merchant, Van der Stede, and Vargus
Our sample broadens the study of subjectivity from these other studies, in part, by using a
sample other than top executives, but most especially by collecting and analyzing a dataset
on the actual use of subjective bonuses.
For this study, we collect data from a sample of department managers in auto dealerships in the U.S. This setting is advantageous for several reasons. First, we observe variation
in compensation practices across a sample of managers in similar jobs, in firms with similar
organizational structures in the same industry. Such homogeneity across firms is rare in
empirical research, and means we control implicitly for many variables that would otherwise
be difficult to control for. Second, the compensation packages are not as complex as those
used with senior executives. Our managers do not receive stock options, restricted stock,
or deferred compensation. Moreover, their formula bonuses consist largely of relatively
simple contracts based on accounting measures, and the rewards are essentially all cash
(Gibbs et al. 2003). Finally, our sample follows the growing trend toward industry studies
(e.g., Ichniowski et al. 1997).
We have no reason to believe that our results are specific to the auto dealership industry.
Subjectivity is used with virtually all incentive schemes, particularly those used at managerial levels. The hypotheses are not specific to this industry, or even to department managers; rather they should have wide applicability to almost any employment setting. We
also have no reason to believe that auto dealership managers are compensated in unusual
ways. In fact, they appear to be rather typical profit-center managers in terms of how they
are held accountable for results. Although jobs in car dealerships are similar in complexity
to managerial jobs in other retail settings (e.g., department stores), they are arguably simpler
than those of divisional managers in large organizations. However, this makes it less likely
that subjectivity will play a role in evaluation, which biases against us finding support for
the hypotheses. But it makes our measurement task easier, which makes our research design
stronger. Overall, the setting seems well suited for studying this topic.
Data collection took place in collaboration with a management-consulting firm that
specializes in working with auto dealerships. We visited a dealership to acquaint ourselves
with the auto dealership business, organization structures and practices, common problems
faced, and specialized language (e.g., fixed operations, floor plan, selling gross, spiffs). The
scope of our intended research required us to develop a customized survey because data
previously collected by the consulting firm were not systematically organized and not sufficiently complex to capture details about the variables we considered relevant.
For each dealership, we developed five surveys: (A) owner, (B) general manager, (C)
service department manager, (D) new car sales department manager, and (E) used car sales
department manager. The owner survey was shorter than the others, serving primarily as a
means for the owners to provide their consent to release their financial data, which had
been collected by the consulting firm in the normal course of business. The general manager
survey included questions about their compensation package and their firm’s economic and
competitive environment, strategy, management practices, and performance. The service,
new car, and used car department surveys were basically identical. They asked about the
respective manager’s compensation package, competitive environment, cooperation between
departments, self-rated performance, and various attitudes and demographics. An important
part of the survey asked the manager questions about their perceptions of properties of
performance measures used for formula bonuses, and of any discretionary bonus they were
eligible for.
We pilot-tested the survey in 24 dealerships. We mailed 120 surveys (24 ! 5), and 59
(49 percent) were returned (A:14!B:14!C:14!D:9!E:8). Based on the pilot study, we made
changes to the format but little change in content. The final set of five surveys was mailed
The Accounting Review, April 2004
Determinants and Effects of Subjectivity in Incentives
417
in March 1999 to 1,203 dealerships, with one follow-up. Each survey was labeled with a
unique consulting firm dealer identification number, allowing us to match the compensation
contracts to the dealership’s financial data.
We received 1,057 surveys (A:277!B:250!C:205!D:186!E:139); i.e., 18 percent [1,057
" (1,203 ! 5)]. A few (A:3!D:1!E:3) were blank, thus the usable number of replies per
section was (A) 274 (23 percent), (B) 250 (21 percent), (C) 205 (17 percent), (D) 185 (15
percent), and (E) 136 (11 percent).2 We have at least one survey from 326 different dealerships, or 27 percent of the 1,203 dealerships. Given our limited follow-up efforts, this
response rate appears reasonable.3
Measures
The unit of analysis is the individual department manager’s compensation contract.
Use of Subjectivity: The use of subjectivity in the assignment of rewards is the dependent
variable for testing H1 through H10. We measure the use of subjectivity in two ways: (1)
an indicator variable is set to 1 if any subjective bonus is given, 0 otherwise, and (2) the
subjective bonus earned as a percent of total compensation (sum of base salary, formula
bonuses, subjective bonus, and miscellaneous rewards).
Formula Bonus Completeness: We use the number of distinct quantitative measures in
the formula bonus as a proxy for formula bonus completeness (Number of Measures, ranging from 0 to 3).
Short-Term Focus: We develop two measures of short-term focus. First, we ask the department managers to indicate the extent to which their primary formula-based performance
measure causes them to focus on short-term goals (Performance Measure Short-Term Focus,
scaled from 1 # not at all to 5 # very high). Second, as a proxy for the extent of longterm investments in intangibles, the opposite of short-term focus, we use the amount spent
on personnel training (Training).
Susceptibility to Manipulation: We ask managers to indicate the extent to which the
primary formula-based performance measure motivates them to manipulate the measure to
meet the target (Performance Measure Manipulation, scaled from 1 # not at all to 5
# very high).
Uncontrollable Factors: We ask department managers to indicate the extent to which their
primary formula-based performance measure reflects factors outside their control (Performance Measure Controllability, scaled from 1 # not at all to 5 # very high).
Interdependencies: We ask managers to estimate the percentage of their time that they
spend interacting with managers in other departments (Departmental Interdependence,
scaled from 0–100 percent).
Environmental Unpredictability: We use two indicators. First, we measure Environmental
Uncertainty at the dealership level. We ask the general manager the set of five questions
listed in Table 4, Panel A, scaled from 1 # very low to 5 # very high. Factor analysis
reduced the five items to a single, composite indicator of environmental uncertainty. Second,
we measure Degree of Competition faced at the department level by asking each department
manager about the number of competitors in their relevant trading area.
2
3
The response rate for survey E was lower because 38 dealerships combined the new and used car departments.
Our instruction to these dealerships was to complete survey D and discard survey E.
Following the consulting firm’s suggestion to not aggressively seek dealerships’ cooperation with our study, we
sent only a single reminder letter (without questionnaire replacement) to nonparticipants after four weeks. Six
weeks after that, we did a telephone call follow-up targeting dealerships from which we had received an incomplete set of surveys, hoping that their partial participation was an indication of their interest in the study.
The Accounting Review, April 2004
418
Gibbs, Merchant, Van der Stede, and Vargus
Recalibration: First, we assess risk from missing performance targets from the interaction
of two variables: performance target difficulty and the consequences of failing to achieve
the target. We calculate Perceived Target Difficulty as 100 minus the perceived likelihood
that the performance target would be achieved (0–100 percent). Although this measure is
ex ante, we have every reason to believe it is highly correlated with actual ex post target
difficulty because we measure it partway through the year. Moreover, Merchant and Manzoni (1989) found that actual target achievement is highly correlated with an ex ante measure similar to ours. We assess Consequences of Failing-to-Achieve the Target by asking
department managers their perceptions about the extent to which failing-to-achieve the
target adversely affects their (1) operating autonomy, (2) pay raise, (3) prospects for promotion, and (4) continued employment, ranging from 1# not at all to 5 # very high. We
average the scores on these four items to form a composite indicator. Second, whether the
department faces a loss in the current period is a dummy variable: Department Loss equals
1 if the department reports a loss, 0 otherwise.
Effects of Subjectivity: To test H11 through H13, we measure pay satisfaction, sales
productivity, and net profit per employee. We assess Pay Satisfaction by asking department
managers about the extent to which they are satisfied with (1) the level of their salary; (2)
the level of their bonuses; (3) the way in which their bonus plans were designed; (4) the
way in which their bonus plans were implemented; and (5) the way in which their performance was evaluated, in general. The scale is 1 # not at all to 5 # very high extent.
Factor analysis shows that these items can be reduced to a single, composite scale of pay
satisfaction. We measure Sales Productivity and Net Profit per employee as the 1999 departmental sales and net profits, respectively, divided by the number of employees in the
department.
We do not have a direct measure of the degree of trust between the manager and the
supervisor. To proxy for trust, we use the number of years the department manager has
been at the dealership (Department Manager Experience). Tenure is a defensible proxy for
trust for several reasons. The longer a manager has been at the dealership, the longer have
been the opportunities for the manager and dealership to develop a good working relationship with each other. There also has been more time for the dealership to establish a
reputation for conducting fair evaluations (Baker et al. 1994). Similarly, Fulk et al. (1985)
note that prior performance appraisal experience is a major factor contributing to the degree
of trust. And, by revealed preference, managers who have stayed at the dealership longer
are more likely to be satisfied with the way their performance has been evaluated in the
past.
Control Variables: We use the log of annual dealership 1998 sales to control for size. Size
is important because it might correlate both with the type and amount of bonuses awarded,
as well as with a variety of other firm practices, such as the degree of decentralization. It
might also be correlated with the manager’s human capital and ability (Rosen 1982), since
larger dealerships may hire more talented managers. We also use 1998–1999 sales growth
and change in employment as control variables in the sales productivity and net profit per
employee regressions.
Measurement Validity: Some measures are objective because they are derived directly
from the dealership financial data (training expenditures, operating loss, size [sales], sales
productivity, net profit per employee, sales growth, and change in employment). Other
measures are subjective because they are derived from the survey. Some survey questions
are factual (raise, formula bonus, discretionary bonus, number of formula bonuses, number
of competitors, and years of experience); others are perceptual, i.e., they measure respondent attitudes by means of either single-item questions (performance measure short-term
The Accounting Review, April 2004
Determinants and Effects of Subjectivity in Incentives
419
focus, manipulation, controllability, target difficulty, and departmental interdependencies)
or multi-item questions (environmental uncertainty and fail-to-achieve consequences).
Because the (single-item) perceptual questions pose the greatest threat to construct
validity, we compare answers across respondents within the same dealership. For the three
performance measure-related variables (short-term focus, manipulation, and controllability),
we obtained 29 within-dealership pairs of identical formula contracts and found that 68 out
of 87 (3 ! 29) assessments of performance measure short-term focus, manipulation, and
controllability are identical or within one Likert-scale point (78 percent). Moreover, none
of the paired sample t-tests for short-term focus, manipulability, and controllability is significant (all two-tailed p $ 0.36), indicating that the responses in the pairs are valid and
reliable (i.e., understood and rated by different respondents in a similar way).
Out of the 26 pairs available for target achievability, 17 (65 percent) put the degree of
target difficulty (measured on a percentage scale) within a 25-percentage range, and the
paired sample t-test is insignificant (two-tailed p # 0.18). The replies for the composite
fail-to-achieve consequences measure also are consistent across department manager pairs:
21 out of 24 valid cases (88 percent) fall within one Likert-scale point, and the remaining
three fall within two Likert-scale points. The paired sample t-test is insignificant (two-tailed
p # 0.65).
For departmental interdependence, we correlate the responses from the new and used
car managers about their estimated cooperation with service within the same dealership.
For the 87 pairs with complete data, the new and used car managers’ responses are significantly and positively correlated (r # 0.24, p # 0.03). Given that the cooperation of new
and used cars with service does not necessarily have to be of the same intensity within the
same dealership, the sign and magnitude of the correlation provide reasonable assurance of
the measure’s validity. For environmental uncertainty, which is a dealership-level variable,
we do not have departmental, or other, corroborating data.
While these tests suggest that our measures have acceptable psychometric properties,
we caution that they do not rule out that measurement error associated with single-item,
perceptual measures can potentially bias the multivariate model estimates.
Descriptive Statistics
Table 1 shows the importance of the subjectively assigned bonuses relative to the department managers’ base salaries and formula bonuses. Overall, approximately one quarter
of the department managers (23 percent) received a discretionary bonus in 1998. Used car
department managers were the most likely to receive a discretionary bonus; service department managers were the least likely.
Table 2 shows summary statistics on discretionary bonuses conditional upon receipt
(as opposed to Table 1 that includes instances of zero bonuses). For the managers receiving
a discretionary bonus, the bonus is approximately 18 percent of total compensation. The
percentage is similar across departments. These magnitudes are similar to those found in
other studies based on quite different samples. For example, Bushman et al. (1996) found
for a sample of CEOs that roughly 35 percent received bonuses based on ‘‘individual
performance’’ (which may or may not be subjectively determined), and that these bonuses
were about 18 percent of salary at the median. In a sample of middle managers, Gibbs
(1995) found that 25 percent received a bonus based on a subjective performance rating,
and on average the bonus was about 12 percent of salary. Thus, the use of discretionary
bonuses is consistent with prior studies.
The Accounting Review, April 2004
420
Gibbs, Merchant, Van der Stede, and Vargus
TABLE 1
Average Composition of 1998 Compensation Packages
Base Salary
Service Department Managers
Average ($)
Median ($)
Percent Receiving
Average Pct. of Total Compensation
New Car Department Managers
Average ($)
Median ($)
Percent Receiving
Average Pct. of Total Compensation
Used Car Department Managers
Average ($)
Median ($)
Percent Receiving
Average Pct. of Total Compensation
All Department Managers Combined
Average ($)
Median ($)
Percent Receiving
Average Pct. of Total Compensation
Formula Bonus
Total Compensation:
Average # $61,422 / Median
$30,253
$27,795
$27,000
$13,200
90.00%
68.00%
56.00%
34.26%
Total Compensation:
Average # $78,428 / Median
$29,001
$40,754
$26,000
$20,012
79.23%
64.48%
44.89%
36.77%
Total Compensation:
Average # $72,195 / Median
$28,793
$36,705
$25,500
$19,571
85.04%
66.14%
47.12%
38.32%
Total Compensation:
Average # $70,189 / Median
$29,435
$34,539
$26,000
$15,700
84.90%
66.27%
49.80%
36.17%
Discretionary Bonus
# $60,650 (n1 # 200)
$2,093
0
20.00%
3.53%
# $70,000 (n2 # 183)
$4,640
0
23.50%
4.26%
# $63,650 (n3 # 127)
$3,678
0
27.56%
5.03%
# $64,000 (n4 # 510)
$3,399
0
23.14%
4.17%
Total Compensation consists of any or all of the following components: Base Salary, up to three Formula
Bonuses, Discretionary Bonus, and Spiffs.
Definitions:
Formula Bonuses are based on quantitative performance measures (e.g., department profit). Some contracts have
up to three formula bonuses, although the majority of the managers (60 percent) receive one formula bonus
only. Across departments, the first formula bonus is on average 85 percent of the total formula bonus. Also, the
first formula bonus is on average more than seven times larger than the second formula bonus.
Discretionary Bonus is based on the supervisor’s subjective judgment of the manager’s performance.
Finally, there are miscellaneous rewards (not reported above), sometimes called ‘‘Spiffs,’’ that are difficult to
characterize in a standard way, such as the use of promotional vehicles and certain incentives provided by the
car manufacturers (e.g., vacation trips). Although receipt of spiffs is common (about 63 percent of the managers
receive them), their economic significance is relatively low (about $4,593 for those who receive spiffs, compared
at $15,000 to $20,000 for those who receive a discretionary bonus—see Table 2).
Table 3 examines the relationship between the manager’s discretionary and formula
bonuses. The proportion of total compensation stemming from the discretionary bonus is
significantly larger in the absence of a formula bonus: 13 vs. 3 percent.4 Moreover, the
primary formula and discretionary bonuses, as percentages of total compensation, are significantly and negatively correlated (r # %0.14, two-tailed p & 0.01). This is consistent
with our conjecture that discretionary bonuses are used to mitigate perceived weaknesses
in bonus awards based on quantitative performance measures.
4
When measured as a percentage of 1998 base salary, instead of total compensation, the proportion discretionary
bonus is about 16 percent of base salary when no formula bonus is offered compared to about 8 percent when
a formula bonus is offered.
The Accounting Review, April 2004
421
Determinants and Effects of Subjectivity in Incentives
TABLE 2
Average 1998 Discretionary Bonus (Dollars and Percentage of Total Compensation)
When Discretionary Bonus $ 0
Pct. Receiving
(ni! / ni)
Average Discretionary
Bonus
a
Service Department Managers (n1! # 40)
New Car Department Managers (n2! # 43)
Used Car Department Managers (n3! # 35)
All Department Managers (n4! # 118)
a
40 / 200
43 / 183
35 / 127
118 / 510
#
#
#
#
20.00%
23.50%
27.56%
23.14%
Dollars
Pct. of Tot. Comp.
$10,728
$19,962
$14,290
$15,149
17.65
18.14
18.27
18.01
ni is taken from Table 1, Panel A.
TABLE 3
Substitution among Formula Bonuses and Discretionary Bonuses
(All Department Managers Combined)
n
No Formula Bonus in 1998
At Least One Formula Bonus in 1998
61
448
Average 1998 Discretionary Bonus
(Pct. of Total 1998 Compensation)
12.59 (1)
2.99 (2)
t ( p)
(1)–(2): 5.32 (0.00)
Table 4 presents descriptive statistics of the measures used in our analyses. Panel A
shows the survey-based measures, and Panel B describes various dealership financial data
for 1998 and 1999. Panel B shows that the size and profitability of the service, new car,
and used car departments, as well of the dealership as a whole, are generally consistent
across both years. In terms of sales, the new car departments are by far the largest; service
departments are the smallest. While the sales departments report the largest overall profit,
their return on sales (0.6 to 1.5 percent) is much lower than the roughly 7 percent reported
by service departments. Low margins for sales departments are consistent with observations
that the U.S. auto retail market is competitive. It also suggests that competition may be a
determinant of pay.
Large dealerships tend not to specialize, as they tend to be large in all departments;
service and sales departments’ sales [employment] are positively correlated (average r
# 0.65 [r # 0.44] in both years, two-tailed p & 0.01). The profitability of sales departments
does not appear to influence the profitability of service departments, as the correlation
between service and sales department profits is insignificant (average r # %0.05 in both
years, two-tailed p $ 0.50). In contrast, profits of new and used car departments are positively correlated in both years (average r # 0.24, two-tailed p & 0.01). Thus, it appears
that performance, and hence the potential use of subjective bonuses, varies across departments within dealerships, especially service versus sales.
Approximately 5 percent of the dealerships operate at a loss. The proportion of new
car departments reporting net losses in 1998 was more than double that reported by service
departments (35 versus 17 percent). To the extent that departmental losses require more
complex compensation arrangements, since no formula bonus may be awarded, this could
The Accounting Review, April 2004
422
The Accounting Review, April 2004
TABLE 4
Descriptive Statistics
Panel A: Survey-Based Measures of the Independent Variablesa
Number of Formula Bonuses [number]
Performance Measure Short-Term Focus [1–5 scale]c
Performance Measure Manipulation [1–5 scale]d
Performance Measure Controllability [1–5 scale]e
Departmental Interdependence [percent]f
Environmental Uncertainty [1–5 scale]g
Degree of Competition [number]h
Performance Target Difficulty [percent]i
Fail-to-Achieve Consequences [1–5 scale]j
Department Manager Experience [years]k
b
Alpha
1st
Quartile
Median
3rd
Quartile
Mean
Std. Dev.
—
—
—
—
—
0.58
—
—
0.82
—
1.00
3.00
1.00
2.00
10.00
2.20
14.00
5.00
2.00
3.00
1.00
4.00
2.00
3.00
10.00
2.60
24.00
10.00
3.00
7.00
2.00
4.00
4.00
3.00
20.00
2.80
50.00
25.00
3.50
12.00
1.50
3.39
2.62
2.86
17.67
2.56
41.30
19.73
2.77
8.73
0.98
1.15
1.36
0.94
14.81
0.47
60.63
22.44
0.96
7.62
1st
Quartile
1. Service Department
1998 Training Expenditures
1999 Training Expenditures
1998 Sales
1999 Sales
1998 Net Profitm
1999 Net Profitm
1998 Return on Sales
1999 Return on Sales
9,691
11,029
1,310,579
1,417,881
31,787
49,809
0.029
0.028
Median
3rd
Quartile
Mean
Std. Dev.
16,166
16,378
1,903,260
2,051,477
135,125
164,380
0.071
0.081
27,721
28,623
2,897,450
3,267,474
280,123
314,819
0.125
0.130
20,523
21,849
2,376,870
2,586,923
188,763
224,159
0.068
0.072
17,407
18,861
1,647,092
1,751,267
313,700
294,659
0.092
0.093
(continued on next page)
Gibbs, Merchant, Van der Stede, and Vargus
Panel B: Department and Dealership 1998 and 1999 Sales and Profit1
7,064
7,689
15,431,000
15,651,374
%70,654
%45,181
%0.005
%0.002
14,383
14,892
22,225,944
24,667,255
117,432
179,264
0.006
0.008
25,043
26,465
36,413,603
38,608,903
443,278
642,436
0.015
0.021
22,176
24,305
29,867,285
33,162,719
329,110
481,800
0.006
0.009
28,586
33,078
28,263,148
31,440,467
890,248
1,087,004
0.020
0.020
1,401
2,269
8,607,572
9,090,647
20,315
%20,680
0.002
%0.003
5,337
6,501
12,904,154
13,211,912
174,584
181,717
0.015
0.014
10,734
12,767
17,336,635
19,105,538
424,610
410,387
0.028
0.026
8,631
10,368
14,543,011
15,672,696
229,181
235,133
0.015
0.013
11,661
14,497
9,851,913
10,546,750
323,700
427,632
0.023
0.025
28,624,816
31,205,391
370,511
394,793
0.012
0.015
40,627,558
45,393,280
795,651
910,832
0.019
0.020
63,522,405
66,819,878
1,426,660
1,610,950
0.029
0.031
50,779,829
55,668,877
1,073,744
1,312,634
0.020
0.022
38,816,419
42,987,438
1,193,330
1,584,321
0.016
0.015
(continued on next page)
423
The Accounting Review, April 2004
2. New Car Department
1998 Training Expenditures
1999 Training Expenditures
1998 Sales
1999 Sales
1998 Net Profitm
1999 Net Profitm
1998 Return on Sales
1999 Return on Sales
3. Used Car Department
1998 Training Expenditures
1999 Training Expenditures
1998 Sales
1999 Sales
1998 Net Profitm
1999 Net Profitm
1998 Return on Sales
1999 Return on Sales
4. Dealership
1998 Sales
1999 Sales
1998 Net Profitm
1999 Net Profitm
1998 Return on Sales
1999 Return on Sales
Determinants and Effects of Subjectivity in Incentives
TABLE 4 (continued)
424
The Accounting Review, April 2004
TABLE 4 (continued)
Department-level measures (n # 526), except Environmental Uncertainty, which is measured at the dealership (n # 250). Summary statistics include Cronbach Alpha
(for multi-item measures only), quartiles (first quartile, median, third quartile), mean, and standard deviation.
b
Number of Formula Bonuses: The survey asked for information on up to three formula bonuses and had an open-ended question about formula bonuses beyond the
third. No respondents, however, reported eligibility for more than three formula bonuses.
c
Performance Measure Short-Term Focus: Extent to which the measure used for the primary formula bonus causes the manager to focus on short-term goals (1 # Not
at All, 2 # Low Extent, 3 # Medium Extent, 4 # High Extent, 5 # Very High Extent).
d
Performance Measure Manipulation: Extent to which the manager is encouraged to manipulate the performance measure used for the primary formula bonus to meet
the target (1 # Not at All, 2 # Low Extent, 3 # Medium Extent, 4 # High Extent, 5 # Very High Extent).
e
Performance Measure Controllability: Extent to which the measure used for the primary formula bonus reflects factors outside the manager’s control (1 # Not at All,
2 # Low Extent, 3 # Medium Extent, 4 # High Extent, 5 # Very High Extent).
f
Departmental Interdependence: Percentage of time department managers spend interacting with managers in other departments.
g
Environmental Uncertainty: Five items: (i) how predictable are the market actions of car dealerships with which you compete; (ii) how accurately can you predict
new car sales for the next year; (iii) how stable are the customer preferences and tastes for new car purchases; (iv) how stable are the legal constraints facing your car
dealership; (v) how stable is the economic environment facing your car dealership? Each item was scored as 1 # Very Low, 2 # Low, 3 # Medium, 4 # High, and
5 # Very High, and then reverse-coded so that higher scores correspond with higher degrees of environmental uncertainty. The scale was validated as part of a factor
analysis including multiple items measured at the dealership level. The composite scale reported here is the average of the five component item-scores. To prevent
that items with relatively large variance influence the overall multi-item scale score, we also constructed a composite score as the average of the standardized itemscores. We use this standardized score (" # 0.00 and # # 0.62) in the multivariate analyses in Table 5.
h
Degree of Competition: Number of competitors, i.e., the number of franchised / non-franchised service providers, the number of same / other nameplate new car dealers,
and the number of used car dealers located in the relevant trading area, respectively.
i
Performance Target Difficulty: 100 minus the percentage likelihood that the performance target for the primary formula bonus will be met.
j
Fail-to-Achieve Consequences: Multipart question: If you fail to achieve target performance for the primary formula bonus, to what extent do you believe that the
following will be adversely affected: (i) your operating autonomy; (ii) your pay raise; (iii) your prospects for promotion; and (iv) your continued employment? The
items were scored as 1 # Not at All, 2 # Low Extent, 3 # Medium Extent, 4 # High Extent, and 5 # Very High Extent. The scale was validated as part of a factor
analysis including multiple items measured at the department level. The composite score reported here is the average of the component item-scores. To prevent that
items with relatively large variance influence the overall multi-item scale score, we also constructed a composite score as the average of the standardized item-scores.
We use this standardized score (" # 0.00 and # # 0.81) in the multivariate analyses in Table 5.
k
Department Manager Experience: Number of years that the department manager has been working at the dealership.
l
Data from the consulting firm archival database for 1998 and 1999 for the 250 dealerships in the sample. Note that the sum of the sales and profits of the service and
the new and used car departments do not add up to the dealership totals. This is because most dealerships also have sales and profits associated with body and parts,
which were not part of the study.
m
In 1998, 17.4 percent of the service departments, 35.1 percent of the new car departments, 22.8 percent of the used car departments, and 6.2 percent of the
dealerships incurred a loss.
In 1999, 16.9 percent of the service departments, 30.2 percent of the new car departments, 27.8 percent of the used car departments, and 5.1 percent of the
dealerships incurred a loss.
a
Gibbs, Merchant, Van der Stede, and Vargus
Determinants and Effects of Subjectivity in Incentives
425
be one potential explanation for our observation that new car managers receive the largest
subjective bonus awards.
While we did not have theory to predict how the departments would differ in their use
of subjectivity in incentives, the observed differences in quantitative characteristics of sales
and service departments and differences noted in our field interviews suggest that we
distinguish between the two types of departments in our tests. Therefore, in subsequent
analyses we report findings for the pooled sample, as well as for service and sales departments separately.
Finally, the sample includes multiple observations of different managers from the same
dealership, which might result in correlated errors across these observations. As a robustness
check, we ran the analyses using Huber-White cluster-corrected standard errors. The correction made no difference in our inferences. This is consistent with the observed heterogeneity in contracts across departments within dealerships.5
IV. RESULTS
Determinants of Subjectivity in Incentives
Our hypotheses make predictions about whether subjectivity is used, rather than about
the size of the subjective bonus. Although a subjective bonus is more likely to affect the
manager’s incentives the greater its size, it is an empirical question whether incidence or
size of the subjective bonus is a better measure of the use of subjectivity. For these reasons,
Table 5 reports the tests of the determinants of both the incidence (Panel A) and size (Panel
B) of subjective bonuses for the pooled and individual service and sales departments.
Incidence of Subjective Bonuses
For the pooled departments, the Logit analysis in Table 5, Panel A, is consistent with
our predictions that the use of subjective bonuses is positively related to (1) the department’s
investments in training (H3), (2) the extent of departmental interdependencies (H6), and
(3) the extent that the achievability of formula bonuses is difficult and leads to significant
consequences if not achieved (H9).
First, we find that subjective bonus usage is significantly and positively related to the
level of investments in training, which we believe represents one of the few areas in a
dealership where current investments have long-term consequences.6 This result is consistent with subjective bonuses being used to encourage such long-term investments (H3).
However, contrary to our expectation, we find that the use of subjective bonuses is negatively related to the degree to which the formula bonus exhibits a short-term focus (H2).
A possible explanation is that the formula bonus is used to generate short-term sales, and
hence, there is no ‘‘weakness’’ to be mitigated with subjectivity. However, both results
together (investments in training versus short-term focus of the formula bonus) suggest that
5
6
The correlations among our independent variables (not reported) are generally small. Of the 26 significant
correlations (two-tailed p & 0.10), only three coefficients exceed 0.20, and none are greater than 0.30. Thus,
multicollinearity is not a concern.
We do not examine advertising. Although advertising can be a long-term investment to promote or establish
brand awareness, car dealerships use it primarily to increase sales in the short term, such as when inventories
increase. Similarly, our field-based insights suggest that dealerships that promote customer satisfaction, which
is another potential long-term focus indicator, would include a customer satisfaction measure, commonly called
a Customer Satisfaction Index (CSI), in the formula bonus. Consistent with this, we find that emphasis on
customer satisfaction, measured by a factor-analyzed six-item survey scale, is significantly greater (t # 6.12, p
& 0.01) for dealerships that include CSI in the department managers’ formula bonus (n # 73) compared to
those that do not (n # 442).
The Accounting Review, April 2004
426
The Accounting Review, April 2004
TABLE 5
Determinants of Discretionary Bonuses
Panel A: Logit Analysisa
Exp.
Sign
Service
New / Used
%0.574 (0.47)
%13.953 (0.17)
%0.736 (0.09)
%0.067 (0.42)
3.630 (0.35)
0.188 (0.28)
%0.501 (0.02)
%42.248 (0.45)
%0.052 (0.41)
%0.292 (0.21)
0.016 (0.16)
%0.499 (0.21)
0.001 (0.46)
[%0.002 (0.44)]
[%0.235 (0.31)]
0.034 (0.02)
0.870 (0.07)
%0.170 (0.38)
93
27%
0.012 (0.43)
%0.221 (0.10)
65.036 (0.05)
0.035 (0.41)
0.019 (0.47)
0.021 (0.05)
%0.272 (0.26)
0.002 (0.35)
[0.010 (0.17)]
[%0.295 (0.18)]
0.019 (0.03)
0.530 (0.13)
%0.052 (0.45)
150
27%
150.860 (0.02)
0.397 (0.09)
0.492 (0.15)
0.035 (0.20)
0.351 (0.32)
0.003 (0.43)
[0.046 (0.09)]
[%1.015 (0.08)]
0.009 (0.40)
%0.612 (0.34)
0.540 (0.25)
57
28%
(continued on next page)
Gibbs, Merchant, Van der Stede, and Vargus
Intercept
Number of Formula Bonuses
H1 [%]
Performance Measure Short-Term Focus
H2 [']
Department Training Expensesb
H3 [']
Performance Measure Manipulation
H4 [']
Performance Measure Controllability
H5 [']
Departmental Interdependence
H6 [']
Environmental Uncertainty
H7 [']
Degree of Competition
H8 [']
[A] Performance Target Difficulty
[B] Fail-to-Achieve Consequences
[A] ! [B] Interaction
H9 [']
Departmental Loss (Dummy 1 # Yes)
H10 [']
Size (Log of Dealership Sales)
Number of Observations
Discretionary Bonus $ 0
Service / New / Used
Panel B: Tobit Analysisc
%0.435 (0.25)
0.011 (0.30)
%0.030 (0.03)
6.223 (0.04)
0.012 (0.20)
0.002 (0.46)
0.001 (0.27)
%0.021 (0.29)
0.000 (0.13)
[0.001 (0.12)]
[%0.014 (0.31)]
0.002 (0.03)
0.067 (0.05)
0.016 (0.33)
150
27%
%0.655 (0.25)
%0.033 (0.17)
%0.020 (0.15)
%0.184 (0.42)
8.721 (0.03)
0.031 (0.06)
0.019 (0.27)
0.003 (0.10)
0.037 (0.22)
0.002 (0.06)
[0.005 (0.01)]
[%0.063 (0.09)]
0.003 (0.08)
%0.084 (0.20)
0.020 (0.35)
57
28%
28.059 (0.18)
0.008 (0.35)
%0.017 (0.30)
0.000 (0.41)
%0.046 (0.21)
0.000 (0.33)
[%0.000 (0.42)]
[%0.001 (0.49)]
0.002 (0.05)
0.079 (0.08)
0.001 (0.43)
93
27%
0.028 (0.17)
%0.048 (0.02)
Logit regressions with dependent variable set to 1 if 1998 Discretionary Bonus $ 0; 0 otherwise. See Table 4, Panels A and B, for definitions and descriptive
statistics of the independent variables. (One-tailed p-values in parenthesis.)
b
Department Training Expenses scaled by department sales.
c
Tobit regressions (left-censored at zero) with 1998 Percentage Discretionary Bonus of Total Compensation as the dependent variable. Same independent variables as
in Table 5, Panel A. (One-tailed p-values in parenthesis.)
a
427
The Accounting Review, April 2004
Intercept
Number of Formula Bonuses
H1 [%]
Performance Measure Short-Term Focus
H2 [']
Department Training Expenses
H3 [']
Performance Measure Manipulation
H4 [']
Performance Measure Controllability
H5 [']
Departmental Interdependence
H6 [']
Environmental Uncertainty
H7 [']
Degree of Competition
H8 [']
[A] Performance Target Difficulty
[B] Fail-to-Achieve Consequences
[A] ! [B] Interaction
H9 [']
Departmental Loss (Dummy 1 # Yes)
H10 [']
Size (Log of Dealership Sales)
Number of Observations
Discretionary Bonus $ 0
Determinants and Effects of Subjectivity in Incentives
TABLE 5 (continued)
428
Gibbs, Merchant, Van der Stede, and Vargus
a short-term/long-term trade-off is important in car dealerships. Beyond generating immediate sales, dealerships seem to have long-term concerns, such as those related to employee training. Thus, although this industry may not have the sharpest short-term/longterm trade-offs, as in pharmaceuticals, for example, it still does have them.
Second, the finding that subjective bonus usage is positively related to the extent of
departmental interdependencies (H6) has two possible, and not mutually exclusive, interpretations: subjective bonuses are used (1) to encourage and reward cooperation between
departments, or (2) to offer employees insurance against performance measurement noise
due to the influence of other departments.7
Third, the positive relationship between the use of subjective bonuses and the interaction between target difficulty and the consequences of failing to meet the target (H9) is
also subject to alternative explanations. One is that firms that award subjective bonuses use
this flexibility to set aggressive targets. An alternative interpretation is that aggressive firms
view discretionary bonuses as insurance, which they sometimes invoke to award a bonus
even though the formula-based objectives were not met. We examine the interaction of
these effects because, conceptually, the impact of target difficulty alone is unimportant if
no significant consequences are attached to missing the target.
Finally, in the pooled analysis we fail to find that the use of subjective bonuses is
related to formula-based contract completeness (H1); manipulability (H4), or controllability
of the formula bonus measures (H5); or to the level of environmental uncertainty (H7), the
degree of competition (H8), and the occurrence of a departmental loss (H10).
We observe several notable differences in the determinants of the use of subjective
bonuses when we examine service and sales departments separately. Specifically, while
some new results emerge, the pooled results appear to be driven by either the service or
sales departments. The use of subjective bonuses is significantly and positively related to
the level of service departments’ investments in training. Training expenditures are subject
to the horizon problem, that is, the department manager incurs an immediate expense in
return for anticipated, yet uncertain, future benefits (sales or profits) stemming from better
service quality. Given that formula bonuses typically depend on current sales or profits,
this result is consistent with subjective bonuses being used to reward or encourage longterm investments (H3).
Two of the other pooled results, however, appear to be driven by the sales departments.
First, the negative relationship between the use of subjective bonuses and short-term focus
may be attributable to the existence of numerous short-term incentive awards (commonly
called spiffs in this industry) that are used to provide incentives for immediate sales, thus
making subjective bonuses less prevalent to generate sales (H2). Second, sales managers
are more likely to receive subjective bonuses when sales targets are difficult to achieve and
when they are associated with more severe consequences if not met (H9). This is consistent
with field evidence that sales managers face more aggressive targets to encourage new
business, while service managers’ performance is less predicated on new business as they
rely on warranty work and nameplate business (i.e., it is common for customers to service
their car where they bought it).
7
To encourage departmental cooperation, the dealership could also use dealership-level measures in the formula
bonus. Empirically, however, we observe little use of dealership measures: only six department managers report
dealership net or gross profit as the primary metric in their formula contract. There are two reasons that interdependency might lead to subjectivity. The first is to encourage departmental cooperation, and dealership measures do that to some extent. But the other is to exclude more uncontrollables due to effects of colleagues on
the individual manager’s performance, and dealership-level measures, in fact, make that worse.
The Accounting Review, April 2004
Determinants and Effects of Subjectivity in Incentives
429
In the pooled results, the extent of departmental interdependence was positively related
to the use of subjective bonuses. However, we fail to find a similar result for either type
of department. This result may be due to small sample sizes for the subsamples.
Finally, three new results emerge in the department-specific regressions. In the service
departments, the use of subjective bonuses is associated with formula bonus completeness
(H1) and manipulability (H4). Service departments are likely to involve many more tasks
on which performance is difficult to measure, e.g., quality of repairs, customer satisfaction,
timeliness of service, and appropriate billing of hours and use of replacement parts. For
these reasons, contract completeness and performance measure manipulability might manifest themselves more in the service departments than in the sales departments. In the sales
departments, the use of subjective bonuses is significantly greater when the department
reports a loss, and thus where there is no, or a reduced, formula bonus award (H10). This
is consistent with a subjective bonus being used to recreate incentives that are removed
when the department is not profitable to encourage continued productive effort.
To summarize, the findings suggest that subjective bonuses are used to mitigate formula
bonus distortions caused by incompleteness (H1), short-term focus (H3), and susceptibility
to manipulation (H4). The findings also indicate that subjective bonuses are used to reduce
formula bonus risk by filtering out uncontrollables due to departmental interdependencies
(H6) and by recalibrating incentives in situations where stretch performance targets are not
met (H9) or where the department operates in a loss situation (H10).
Size of Subjective Bonuses
Table 5, Panel B, shows the results of a Tobit analysis examining the size of subjective
bonuses as a percentage of total compensation. The sales department results are identical
to those obtained with the prior Logit analysis. However, we generally obtain stronger
results for the service managers. For the service departments, we now also find that the
size of subjective bonus is significantly greater (1) when there are departmental interdependencies (H6), (2) when there is intense competition (H8), and (3) when the quantitative
measures have difficult targets and more severe consequences (H9).
Importantly, the size of subjective bonuses appears to be strongly related to the degree
of competition whereas the use of subjective bonuses was not. This implies that, when
subjective bonuses are used in competitive environments, a relatively greater amount is
awarded. For contract completeness, we find the opposite. Although incomplete contracts
appear to give rise to the use of subjective bonuses, they fail to systematically vary with
the amount of subjective bonuses awarded.
We further examine the Tobit results to evaluate the economic significance of the various determinants of subjective bonuses.8 An increase of 1 standard deviation in the target
difficulty and consequences interaction is associated with an increase of 67 percent in
subjective bonus (from 8.8 to 14.7 percent of total pay), while a 1 standard deviation
8
For a given vector of explanatory variables xi, the expected level of subjective bonuses can be computed as
E[Subjective Bonus!xi] # $(%!xi / #) ! (%!xi ' #&i), where &i # N(%!xi / #) / $(%!xi / #), and N and $ refer to the
standard normal density function and cumulative normal density function, respectively (Greene 2000). To evaluate the economic effects of the significant determinants over various ranges we set the value of the remaining
variables to their sample means when computing the expected subjective bonus award. We evaluate the expected
value of the subjective bonus using the mean values of the independent variables that were used in the regressions
in Table 5, which yields fewer observations (n # 150) than in Table 4 because of missing values.
The Accounting Review, April 2004
430
Gibbs, Merchant, Van der Stede, and Vargus
decrease corresponds to a 43 percent reduction in subjective bonus. Examining the economic impact of training expenses, performance measure manipulation, departmental interdependencies, and competition, we note that these determinants are associated with service department managers. For service managers, training expenses have the greatest impact
on subjective bonus awards. An increase of 1 standard deviation in training expenses is
associated with an increase in subjective bonus of 106 percent (from 4.7 to 9.7 percent of
total pay), while a decrease of 1 standard deviation results in a 57 percent reduction in
service managers’ subjective bonus. Changes in the level of departmental interdependencies
have the smallest impact on subjective bonus: an increase (decrease) of one standard deviation leads to a 55 (38) percent increase (decrease) in subjective bonus. The economic
effects of competition and performance measure manipulation lie between the prior two
results. For sales managers, the economic impact of departmental profitability is also significant: subjective bonus is 96 percent higher when the department reports a loss relative
to when it is profitable (14.1 versus 7.2 percent of total pay). We interpret these findings
as indicating that our determinants of subjective bonus provide significant economic incentives to managers.
Effects of Subjectivity in Incentives
To test the hypothesized effects of the use of subjective bonuses, we performed ordinary
least squares regressions in Table 6 with pay satisfaction (Panel A), sales productivity (Panel
B), and net profit per employee (Panel C) as the dependent variables, all measured in 1999.
In early 1999, the department managers learn the outcome of three components of their
1998 compensation: (1) their 1998–1999 raise; (2) their 1998 formula bonus; and (3)
whether they received a subjective bonus for 1998. We expect that these three elements of
compensation have an effect on pay satisfaction (which we measure in early 1999 at the
time of the survey) as well as subsequent 1999 performance.9 We measure 1999 performance along two dimensions: department sales and net profit per employee. We include a
sales productivity measure because subjective bonuses are more intangible, by definition,
than formula bonuses, and are designed to affect other things beyond profit, which is
primarily captured by the formula bonus.
Subjectivity may increase pay satisfaction, sales productivity, and net profit to the extent
that it reduces the employee’s risk and better aligns interests of the firm and employee.
However, this will be the case only if subjectivity is effective because there is adequate
trust between the employee and the firm. Therefore, Table 6 shows the results of the models
with and without the interaction term of subjective bonus and manager tenure (our proxy
for trust). We again present the results for the pooled and individual service and sales
departments. As additional control variables, we include dealership size in all regressions,
plus sales growth and changes in employment for the sales productivity and net profit
regressions. We include dealership size for all the reasons elaborated above, although the
possibility that larger dealerships (attract and) retain more talented managers is of particular
importance here. Moreover, size also captures potential economies of scale in Panels B and
9
We use (an extreme form of) 1998 performance (departmental loss) as an independent variable in the Table 5
analyses. Thus, if we investigate the effect of 1998 subjective bonus on 1998 performance we find a negative
relationship. Also, if we include the effect of 1998 formula bonus on 1998 performance we find, tautologically,
that they are positively related (because 1998 bonus is a function of 1998 performance; e.g., 2 percent of net
profit). In other words, within the same time period, bonuses are both determining performance (by motivating
employees to achieve rewarded performance) and being determined by performance (either by formula or through
losses leading to the use of subjectivity), which raises issues of endogeneity and simultaneity (Ittner and Larcker
2001). This illustrates the importance of specifying an appropriate temporal model to investigate the effects of
1998 subjective bonuses on subsequent 1999 outcomes.
The Accounting Review, April 2004
Panel A: Pay Satisfactiona
Exp.
Sign
Intercept
1998–1999 Raiseb
1998 Formula Bonusb
[A] 1998 Discretionary Bonusb
[B] Department Manager Experience
[A] ! [B] Interaction
Size (Log of Dealership Sales)
'
'
H11 [']
Service / New / Used
Service
New / Used
%1.830 (0.08)
%1.648 (0.10)
%2.320 (0.14)
%2.581 (0.11)
%1.713 (0.16)
%1.518 (0.19)
0.2!10%4 (0.06)
0.5!10%5 (0.00)
0.1!10%5 (0.34)
0.040 (0.22)
—
0.092 (0.11)
F # 5.76
(0.00)
R2 # 0.08
n # 357
0.2!10%4 (0.06)
0.5!10%5 (0.00)
[%0.1!10%5 (0.35)]
[%0.019 (0.42)]
1.257 (0.02)
0.082 (0.14)
F # 5.47
(0.00)
R2 # 0.09
n # 351
0.6!10%4 (0.06)
0.5!10%5 (0.00)
0.5!10%5 (0.32)
0.059 (0.35)
—
0.119 (0.16)
F # 2.05
(0.08)
R2 # 0.07
n # 138
0.5!10%4 (0.02)
0.5!10%5 (0.03)
[%0.4!10%4 (0.07)]
[%0.030 (0.42)]
4.915 (0.04)
0.139 (0.13)
F # 1.99
(0.07)
R2 # 0.08
n # 136
0.9!10%5 (0.26)
0.5!10%5 (0.00)
0.9!10%6 (0.37)
0.6!10%4 (0.50)
—
0.085 (0.19)
F # 4.36
(0.00)
R2 # 0.09
n # 219
0.1!10%4 (0.23)
0.5!10%5 (0.00)
[%0.6!10%6 (0.41)]
[%0.060 (0.32)]
1.387 (0.02)
0.073 (0.23)
F # 4.48
(0.00)
R2 # 0.11
n # 215
%10.86 (0.36)
%7.521 (0.40)
%0.575 (0.14)
%5
%0.4!10 (0.31)
%70.13 (0.04)
0.001 (0.00)
0.5!10%4 (0.00)
[0.9!10%4 (0.11)]
[2.603 (0.11)]
23.351 (0.03)
0.2!10%5 (0.00)
%0.535 (0.02)
1.069 (0.27)
F # 8.89
(0.00)
R2 # 0.21
n # 285
%0.495 (0.17)
%5
%0.4!10 (0.39)
%70.97 (0.04)
0.001 (0.00)
0.5!10%4 (0.01)
0.1!10%3 (0.04)
2.480 (0.11)
—
0.2!10%5 (0.00)
%0.658 (0.00)
1.338 (0.22)
F # 8.70
(0.00)
R2 # 0.18
n # 296
0.1!10%6 (0.41)
0.3!10%5 (0.15)
0.022 (0.27)
—
0.2!10%6 (0.00)
%0.018 (0.00)
0.073 (0.01)
F # 5.27
(0.00)
R2 # 0.22
n # 139
0.2!10%7 (0.48)
[0.1!10%4 (0.09)]
[%0.021 (0.29)]
%0.689 (0.15)
0.2!10%6 (0.00)
%0.017 (0.00)
0.078 (0.01)
F # 5.17
(0.00)
R2 # 0.25
n # 134
0.001 (0.03)
0.1!10%5 (0.49)
0.4!10%4 (0.27)
%1.416 (0.29)
—
0.1!10%5 (0.02)
%1.107 (0.00)
5.677 (0.01)
F # 3.50
(0.00)
R2 # 0.14
n # 157
0.001 (0.02)
0.1!10%4 (0.33)
[0.3!10%4 (0.33)]
[%1.768 (0.25)]
18.706 (0.08)
0.1!10%5 (0.01)
%0.986 (0.00)
5.553 (0.01)
F # 3.31
(0.00)
R2 # 0.16
n # 151
Determinants and Effects of Subjectivity in Incentives
TABLE 6
Effects of Discretionary Bonuses
Panel B: 1999 Sales Productivityc
'
'
H12 [']
(continued on next page)
431
The Accounting Review, April 2004
Intercept
1998–1999 Raise
1998 Formula Bonus
[A] 1998 Discretionary Bonus
[B] Department Manager Experience
[A] ! [B] Interaction
1998–1999 Sales Growthd
1998–1999 Change in Employmentd
Size (Log of Dealership Sales)
432
The Accounting Review, April 2004
TABLE 6 (continued)
Exp.
Sign
Service / New / Used
Service
New / Used
Panel C: 1999 Net Profit per Employeee
Intercept
1998–1999 Raise
1998 Formula Bonus
[A] 1998 Discretionary Bonus
[B] Department Manager Experience
[A] ! [B] Interaction
1998–1999 Sales Growth
1998–1999 Change in Employment
Size (Log of Dealership Sales)
'
'
H13 [']
%15.41 (0.42)
%69.33 (0.02)
%72.74 (0.02)
0.004 (0.00)
0.3!10%5 (0.00)
%7
%0.5!10 (0.49)
%6.953 (0.14)
—
0.4!10%7 (0.00)
%0.889 (0.13)
0.107 (0.49)
F # 10.37
(0.00)
R2 # 0.20
n # 296
0.004 (0.00)
0.4!10%5 (0.00)
[%0.2!10%5 (0.21)]
[%8.835 (0.05)]
72.363 (0.01)
0.3!10%7 (0.00)
%0.752 (0.12)
1.501 (0.37)
F # 13.24
(0.00)
R2 # 0.28
n # 285
0.4!10%5 (0.22)
0.1!10%6 (0.35)
0.4!10%6 (0.40)
0.327 (0.08)
—
0.2!10%6 (0.00)
%0.865 (0.00)
4.660 (0.01)
F # 4.58
(0.00)
R2 # 0.20
n # 139
0.4!10%5 (0.22)
0.1!10%6 (0.39)
[0.6!10%5 (0.09)]
[3.698 (0.06)]
%46.464 (0.13)
0.2!10%6 (0.00)
%0.834 (0.00)
4.851 (0.01)
F # 4.46
(0.00)
R2 # 0.22
n # 134
55.55 (0.38)
0.004 (0.00)
0.4!10%5 (0.00)
%6
%0.5!10 (0.44)
%12.25 (0.15)
—
0.4!10%7 (0.02)
%1.147 (0.24)
%2.234 (0.42)
F # 5.16
(0.00)
R2 # 0.20
n # 157
35.97 (0.40)
0.004 (0.00)
0.4!10%5 (0.00)
[%0.2!10%5 (0.29)]
[%15.02 (0.07)]
106.04 (0.02)
0.3!10%7 (0.00)
%1.031 (0.22)
%1.408 (0.44)
F # 7.29
(0.00)
R2 # 0.29
n # 151
OLS regressions with Pay Satisfaction as the dependent variable. (One-tailed p-values in parenthesis.)
Pay Satisfaction measured as the extent to which the respondents are satisfied with: (i) the level of their salary; (ii) the level of their bonuses; (iii) how their bonus
plans are designed; (iv) how their bonus plans are implemented; and (v) how their performance is evaluated, in general. Each item was scored as 1 # Not at All, 2
# Low Extent, 3 # Medium Extent, 4 # High Extent, and 5 # Very High Extent. Factor analysis reduced the five items to a single, composite indicator of pay
satisfaction, which we constructed as the average of the standardized item-scores (" # 0.00; # # 0.86; Cronbach Alpha # 0.91).
b
Refer to Table 1, Panel A, for the definitions and descriptive statistics of the 1998 Formula and Discretionary Bonus (dollars). The Raise is the change in base salary
from 1998 to 1999 (dollars).
c
OLS regressions with 1999 Sales Productivity as the dependent variable. Sales Productivity equals 1999 departmental sales divided by the number of employees in
the department. (One-tailed p-values in parenthesis.)
d
Sales Growth is the change in departmental sales from 1998 to 1999. Change in Employment is the change in departmental employment from 1998 to 1999. All
other independent variables are the same as in Table 6, Panel A.
e
OLS regressions with 1999 Net Profit per Employee as the dependent variable. Net Profit per Employee equals 1999 departmental net profit divided by the number of
employees in the department. (One-tailed p-values in parenthesis.) Same independent variables as in Table 6, Panel B.
a
Gibbs, Merchant, Van der Stede, and Vargus
%11.45 (0.45)
Determinants and Effects of Subjectivity in Incentives
433
C. We include sales growth and changes in employment to control for trends in performance
that might affect productivity or profit per employee.
Pay Satisfaction
Panel A shows that raises and formula bonuses have positive effects on pay satisfaction.
We had no priors about the effects on pay satisfaction of either use of subjective bonus or
tenure. Our prediction is that greater use of subjectivity should increase pay satisfaction
only if there is better trust between the manager and the firm, because trust should increase
the benefits and reduce the costs of subjectivity (H11). We find that this interaction is
significant as predicted. This result also holds for the service and sales departments separately. Thus, there is evidence in Panel A consistent with the idea that trust is important
for successful use of subjectivity.
Performance
The results in Panels B and C of Table 6 for productivity and profitability are similar
to those for pay satisfaction for the pooled departments. However, raises, formula bonuses,
and subjective bonuses appear to have no effect on service department productivity and
profitability. Raises, and subjective bonuses when trust is high (interaction term), both
appear to increase sales department productivity (H12) and profitability (H13).
In sum, the effects of compensation on pay satisfaction do not vary by department, but
the effects on productivity and profitability do. Subjective bonuses have a more positive
effect when there is better trust (or at least, a longer-term relationship) between the manager
and the firm, indicating that trust increases the benefits and reduces the costs of subjectivity.
These results suggest the importance of trust or relational contracting in making subjective
incentives effective.
V. DISCUSSION AND CONCLUSIONS
Subjective performance evaluations are present and important in virtually all jobs, from
the lowest levels to CEOs. Subjectivity can play a particularly important role in incentives,
increasing the alignment of interests between the employee and the firm and reducing
employee risk. Subjectivity should be most important in complex work environments, where
job designs involve multiple tasks and decision making. Similarly, it should be most important in unpredictable environments. Despite the apparent importance of subjectivity, it
has been little studied.
In this paper, we studied the determinants and effects of subjectivity in incentive systems. We discussed many ways in which subjectivity should theoretically play a role. More
importantly, we developed and administered an extensive survey of compensation practices
in U.S. car dealerships to collect data on the use of subjectivity and related issues to test
the theoretical arguments.
The tests reveal a number of interesting findings. Discretionary bonuses, those based
on subjective assignments of rewards, provide approximately 20 percent of the total compensation for managers who receive them. Slightly less than 25 percent of the managers
receive such bonuses in a given year. These are similar to results found in studies of top
executives and middle managers, indicating the widespread use of subjectivity in compensation practices.
We find that, when they are used, subjective bonuses seem to substitute for formula
bonuses, as subjective bonuses are larger where formula bonuses are not used. Specifically,
we find evidence that subjective bonuses are used to mitigate perceived weaknesses in bonus
awards based on quantitative performance measures, such as in situations where formula
The Accounting Review, April 2004
434
Gibbs, Merchant, Van der Stede, and Vargus
bonuses fail to adequately encourage investments with long-term impacts, such as in training, and where formula bonuses fail to encourage cooperation or where performance is
noisy due to the influence of other departments. The findings also suggest that subjectivity
is used to reduce employee risk or to re-create incentives, such as when managers face
difficult targets that have severe consequences or when they operate in a loss condition. In
support of the idea that subjectivity aids contracting with the manager when trust is higher,
we find that subjective bonuses are positively associated with pay satisfaction and performance when the manager has longer tenure at the dealership.
Some of our theoretical predictions, however, are not supported. We fail to find evidence
of a positive relationship between the use of subjective bonuses and some weaknesses in
the quantitative performance measures, such as their controllability. Any of several reasons
might have contributed to a lack of support for some of our hypotheses. The sample size,
while relatively large for a survey study, might have been inadequate to capture the relationships that surround the uses and effects of subjective bonuses. Measures of some variables, such as weaknesses of quantitative performance measures, are crude. This is to be
expected since the concepts that we are studying are inherently subjective.
There are also limitations in the theory. The theory that explains the uses and effects
of subjective bonuses is in an early stage of development. Subjective bonuses are only one
part of the firms’ compensation and reward package. The relationships among the various
package elements may be interactive, both among themselves and with various aspects of
the management setting. In contrast, existing theory is quite simplistic. It generally specifies
only linear relationships between two or a few variables, while ignoring interactive relationships, nonlinear relationships, and relevant ranges. Existing theory also ignores numerous potentially relevant variables that are descriptive of the totality of the organization’s
reward package and the specific situation in which the reward packages are used.
The few other studies of subjectivity (Bushman et al. 1996; Hayes and Schaefer 2000;
Murphy and Oyer 2003) made substantial progress relating indicators of use of subjectivity
to financial variables and firm characteristics. However, many interesting issues arising with
subjectivity are not easily studied using traditional datasets. By its nature, subjectivity involves concepts that are qualitative and, of course, subjective. It raises behavioral issues,
such as trust, and conflicts in perceptions between the employee and supervisor. Such issues
require a different approach, involving either field research or survey methods. While the
data collection methods that we have employed are difficult, we expect that further progress
on understanding the complexities of incentive systems, including the use of subjectivity,
can be made by collecting new types of variables to examine theoretical concepts that have
to date been given limited study.
REFERENCES
Baiman, S., and M. V. Rajan. 1995. The informational advantages of discretionary bonus schemes.
The Accounting Review 70 (October): 557–579.
Baker, G. P., M. C. Jensen, and K. J. Murphy. 1988. Compensation and incentives: Practice vs. theory.
Journal of Finance 43 (July): 593–615.
———. 1992. Incentive contracts and performance measurement. Journal of Political Economy 100
(June): 598–614.
———, R. Gibbons, and K. J. Murphy. 1994. Subjective performance measures in optimal incentive
contracts. Quarterly Journal of Economics 109 (November): 1125–1156.
Banker, R. D., and S. M. Datar. 1989. Sensitivity, precision, and linear aggregation of signals for
performance evaluation. Journal of Accounting Research 27 (Spring): 21–39.
The Accounting Review, April 2004
Determinants and Effects of Subjectivity in Incentives
435
Bommer, W. H., J. L. Johnson, G. A. Rich, P. M. Podsakoff, and S. C. MacKenzie. 1995. On the
interchangeability of objective and subjective measures of employee performance: A metaanalysis. Personnel Psychology 48 (Autumn): 587–605.
Bushman, R. M., R. J. Indjejikian, and A. Smith. 1996. CEO compensation: The role of individual
performance evaluation. Journal of Accounting and Economics 21 (April): 161–193.
Feltham, G. A., and J. Xie. 1994. Performance measure congruity and diversity in multi-task principalagent relations. The Accounting Review 69 (July): 429–453.
Folger, R., and M. A. Konovsky. 1989. Effects of procedural and distributive justice on reactions to
pay raise decisions. Academy of Management Journal 32 (March): 115–130.
Fulk, J., A. P. Brief, and S. H. Barr. 1985. Trust-in-supervisor and perceived fairness and accuracy
of performance evaluations. Journal of Business Research 13 (August): 301–313.
Gibbons, R. 1987. Piece-rate incentive schemes. Journal of Labor Economics 5: 413–429.
Gibbs, M. 1995. Incentive compensation in a corporate hierarchy. Journal of Accounting and Economics 19 (March–May): 247–277.
———, K. A. Merchant, W. A. Van der Stede, and M. E. Vargus. 2003. Performance measure properties and incentives. Working paper, University of Chicago, University of Southern California,
and The University of Texas at Dallas.
Greene, W. H. 2000. Econometric Analysis. Upper Saddle River, NJ: Prentice Hall.
Hawkins, S., and R. Hastie. 1990. Hindsight-biased judgments of past events after the outcomes are
known. Psychological Bulletin 107 (May): 311–327.
Hayes, R. M., and S. Schaefer. 2000. Implicit contracts and the explanatory power of top executive
compensation. Rand Journal of Economics 31 (Summer): 273–293.
Holmstrom, B. 1979. Moral hazard and observability. Bell Journal of Economics 10 (Spring): 74–91.
———, and P. Milgrom. 1991. Multitask principal-agent analyses: Incentive contracts, asset ownership, and job design. Journal of Law, Economics and Organization 7: 24–52.
Ichniowski, C., K. Shaw, and G. Prennushi. 1997. The effects of human resource management practices on productivity: A study of steel finishing lines. American Economic Review 87 (June):
291–313.
Ittner, C. D., and D. F. Larcker. 2001. Assessing empirical research in managerial accounting: A
value-based management perspective. Journal of Accounting and Economics 32 (December):
349–410.
———, ———, and M. W. Meyer. 2003. Subjectivity and the weighting of performance measures:
Evidence from a balanced scorecard. The Accounting Review 78 (July): 725–758.
Jacobs, M. T. 1991. Short-Term America: The Causes and Cures of Our Business Myopia. Boston,
MA: Harvard Business School Press.
Kaplan, R. S., and D. P. Norton. 1996. The Balanced Scorecard: Translating Strategy into Action.
Boston, MA: Harvard Business School Press.
Lawler, E. E. 1971. Pay and Organizational Effectiveness. New York, NY: McGraw-Hill.
Lev, B. 2001. Intangibles: Management, Measurement, and Reporting. Washington, D.C.: Brookings
Institute Press.
MacLeod, W. B., and D. Parent. 1999. Job characteristics and the form of compensation. Research
in Labor Economics 18: 177–242.
McAllister, D. J. 1995. Affect- and cognition-based trust as foundation of interpersonal cooperation
in organizations. Academy of Management Journal 38 (February): 24–59.
Merchant, K. A. 1987. How and why firms disregard the controllability principle. In Accounting and
Management: Field Study Perspectives, edited by W. J. Bruns, Jr., and R. S. Kaplan, 316–338.
Boston, MA: Harvard Business School Press.
———. 1989. Rewarding Results: Motivating Profit Center Managers. Boston, MA: Harvard Business
School Press.
———, and J. F. Manzoni. 1989. The achievability of budget targets in profit centers: A field study.
The Accounting Review 64 (July): 539–558.
Milgrom, P. 1988. Employment contracts, influence activities, and efficient organization design. Journal of Political Economy 96 (February): 42–60.
The Accounting Review, April 2004
436
Gibbs, Merchant, Van der Stede, and Vargus
Murphy, K. J. 2000. Performance standards in incentive contracts. Journal of Accounting and Economics 30 (December): 245–278.
———, and P. Oyer. 2003. Discretion in executive incentive contracts: Theory and evidence. Working
paper, University of Southern California and Stanford University.
Murphy, K. R., and J. N. Cleveland. 1995. Understanding Performance Appraisal: Social, Organizational, and Goal-Based Perspectives. Thousand Oaks, CA: Sage Publications.
Pennings, J. M., and J. Woiceshyn. 1987. A typology of organizational control and its metaphors. In
Research in the Sociology of Organizations, edited by S. B. Bacharach, and S. M. Mitchell,
75–104. Greenwich, CT: JAI Press.
Prendergast, C. 1993. A theory of ‘‘yes men.’’ American Economic Review 83 (September): 757–770.
———, and R. Topel. 1993. Discretion and bias in performance evaluation. European Economic
Review 37 (April): 355–365.
———. 1999. The provision of incentives in firms. Journal of Economic Literature 37 (March): 7–
63.
Raith, M. 2001. Competition, risk, and managerial incentives. Working paper, University of Chicago.
Rosen, S. 1982. Authority, control, and the distribution of earnings. RAND Journal of Economics 13
(Autumn): 311–323.
The Accounting Review, April 2004